Atomic spectrometry update: review of advances in the analysis of clinical and biological materials, foods and beverages

Contents

• 1 Reviews

• 2 Metrology, interlaboratory studies, reference materials and reference ranges

• 3 Sample collection and preparation
  • 3.1 Collection, storage and preliminary preparation
  • 3.2 Digestion, extraction and pre-concentration

• 4 Progress with analytical techniques
  • 4.1 Mass spectrometry
  • 4.2 Atomic absorption and atomic emission spectrometry
  • 4.3 Laser induced breakdown spectroscopy
  • 4.4 Vapour generation procedures and atomic fluorescence spectrometry
  • 4.5 X-ray spectrometry

• 5 Nanomaterials

• 6 Applications: clinical and biological materials
  • 6.1 Metallomics
  • 6.2 Imaging with MS, X-rays and LIBS
  • 6.3 Trace elements in pregnancy
  • 6.4 Elements as tags for indirect determinations
  • 6.5 Multi-element applications
  • 6.6 Progress for individual elements

• 7 Applications: drugs and pharmaceuticals, traditional medicines and supplements

• 8 Applications: foods and beverages
  • 8.1 Progress for individual elements
  • 8.2 Single and multi-element applications in food and beverages

• 9 Abbreviations

• 10 Conflicts of interest

• References

---

Abstract

This update covers publications from the second half of 2020 to the middle of 2021. Advances in analytical techniques and their applications relevant to clinical and biological materials, foods and beverages are reviewed in the text, highlighting their key features. Three tables complement the text, aiming to summarise technical details of interest for sample preparation based on extraction procedures, applications to clinical and biological materials and analysis of food and beverage samples. In this year’s review, we report on efforts to establish reference materials, reference procedures, reference networks and reference intervals, in some areas of clinical and food analysis. In terms of analytical developments, whereas ICP-MS and ICP-MS/MS are superior in their ability to achieve low LODs, there are also been an increase in the number of applications reporting sample pre-treatment involving SPE or LLE, to improve the LODs of poorer performing, but more affordable, techniques. Imaging is another expanding area, with several techniques being exploited to provide elemental profiles. Wider applications of instrumental developments, such as ICP-QQQ-MS and ICP-MS/MS, also provide analytical advantages in the area of clinical, food and beverages analysis and advances were reported. Authenticity and certification of origin are increasingly important features for certain food. Elemental profiling, coupled with statistical techniques, seems to be a very valuable tool to support these claims.

---

1 Reviews

This latest update adds to that from last year¹ and complements other reviews of analytical techniques in the series of Atomic Spectrometry Updates from the previous year.^2–6

Reflecting the considerable interest in nanoparticles, which we have featured in previous ASUs, Galazzi et al.⁷ reported the application of spICP-MS, scICP-MS, LA-ICP-MS, LC-ICP-MS and FFF-ICP-MS to studies of NP biotransformation, aggregation or agglomeration, interactions with biomolecules, and the impact of these events on potential toxicity. Their review described applications of these techniques to a wide variety of clinical and plant investigations, rather than any discussion of the instrumentation involved. The authors showed that spICP-MS and LA-ICP-MS are the more popular techniques used for quantification of NPs. Liquid chromatography coupled with ICP-MS was indicated as the most appropriate approach to investigate interactions with target molecules, a statement supported by the many examples from published work which they discussed. They also referred to FFF-ICP-MS, together with NP size-characterisation, while analysis of solid samples, using LA-ICP-MS, afforded information on the spatial distribution of particles in tissues.

Galazzi et al.⁷ referred also to scICP-MS and, in a tutorial review, Theiner et al.⁸ considered the instrumentation and applications for this technique. The developments in flow cytometry, from the 1950s to the present day, were described as an introduction to this review of the technologies associated with equipment, sample introduction and preparation, and calibration strategies. This was followed by accounts of recent applications, referring to single-cell analysis in suspensions, LA-ICP-MS and imaging mass cytometry.

For several years, these ASU reviews have included work involving the use of assays in which a metal label has been incorporated into an immunoreaction with the end-point measurement being by ICP-MS. The number of such papers has gradually increased and these assays are now featuring in review articles. With a title including the intriguing phrase ‘So far, so good, so what?’, Grindlay and colleagues⁹ expanded on the techniques and applications, emphasising the advantages of high diagnostic sensitivity and selectivity, multiplexed capabilities, large dynamic range and robustness for the analysis of both liquid and solid samples. They lamented, however that immunoassays based on ICP-MS are poorly recognised among the life sciences. Another review referred to immunoreactions used to label biomolecules in cells and tissues with metal elements. Using LA-ICP-MS, bioimaging could be achieved.¹⁰ These authors¹⁰ further discussed how amplification strategies based on the use of labels containing several atoms of an element, such as inorganic NPs, increase the sensitivity of the process. The concept of target labelling and signal amplification using NPs was also discussed by Larraga-Urdaz et al.¹¹

Procedures for the speciation of As in clinical and food samples have been discussed in considerable detail in previous Updates. Most of these featured some form of chromatography to separate the species prior to the measurement of As, but a few workers have shown that a degree of separation can be achieved based on the sample preparation. Welna et al.¹² reviewed non-chromatographic speciation of As, as made possible by HG and sample pre-treatment strategies, such as specific extraction–complexation, retention and co-precipitation procedures. Most of the procedures referred to speciation for As^III, As^V, DMA and MMA, but there was some consideration given also to arsenosugars, AB and AC. Among the examples presented were analyses of clinical samples, foods and beverages. Despite the evident enthusiasm shown by the authors for this approach, it is likely that chromatographic speciation will continue to be the more popular.

In a “Special Issue Article” for X-Ray Spectrometry, Fernández-Ruiz¹³ described the basic principles of TXRF spectrometry and then discussed applications to measure elements in serum and other biological fluids, proteins, tissues and metal containing drugs, emphasising the versatility and ease of use of this technique. A second review,¹⁴ in the same journal, summarised results from analyses of RMs or well-characterized biological samples, such as human fluids and animal or plant tissues, by TXRF following preparation by dilution or suspension. No real samples were analysed, but the results allowed the authors to suggest how the technique may be applicable to clinical studies.

Three reviews focused on analysis of foods. A comprehensive review by Lorenc et al.¹⁵ described speciation analysis of elements in food samples. With many examples to illustrate the topic, the authors showed how, using LC-ICP-MS, elemental speciation in a wide range of food types has been achieved. In addition to technical details, the review included tables to show typical reported concentrations compared with maximum permissible levels according to FAO/WHO, EU, Chinese and Canadian authorities. The authors recommended that guidance should be more widely introduced and be based on toxicological data, with results given by reliable analytical procedures. This review did not mention analysis using LIBS, but Velasquez-Ferrin et al.¹⁶ attempted to fill that gap by discussing the potential for the technique to advance the characterisation of components in food, in the future. Zanardi and co-workers¹⁷ discussed work from the last 10 years on the authentication of fish and sea foods from data on inorganic elemental composition. The analytical methods were noted, with ICP-OES seen as the most used, although it was acknowledged that ICP-MS was more prominent in recent publications, together with ICP-TOF-MS and ICP-MS/MS. Sample preparation procedures were included (see also Section 3.2 of this Update) together with a discussion of statistical data analysis programs. An overview of relevant publications was given in a table that listed sample types, techniques used, elements determined and how data were analysed.

Two ‘Critical Reviews’ provided updates that address longstanding areas of interest. That of Parsons et al.¹⁸ returned to the potential of dried blood spot samples as suitable specimens to assess exposure to chemical elements, while Timerbaev¹⁹ was concerned with the current status of ICP-MS and the development of metal-based drugs and diagnostic agents. These reviews are discussed in more detail in Sections 3.1 and 7, respectively.

2 Metrology, interlaboratory studies, reference materials and reference ranges

Given that measurement results are crucial to decisions in several areas of human activities, efforts to ensure their harmonisation and comparability have been in place for a long time at an international level. Under the Meter Convention, an international treaty signed in 1875, a structure of organisations working toward these aims was established (http://www.bipm.org), including Committees for specific areas of measurements. The Consultative Committee for Amount of Substance – Metrology in Chemistry and Biology (CCQM) pursues the objective of advancing the global comparability of chemical and biological measurement standards and capabilities. Key-comparison is the term used to indicate the comparison of measurement results, traceable to the SI, obtained by National Measurement Institutes and other designated laboratories, to underpin the Calibration and Measurement Capabilities (CMCs) of the participating institutes and demonstrate the international equivalence of the measurement standards involved. The key-comparison CCQM-K145 was carried out to compare measurement results, traceable to the SI, for 12 essential and toxic elements (As, Co, Cr, Hg, Mn, Mo, Ni, P, Pb, S, Sr and Zn) in lyophilised bovine liver.²⁰ The 30 participants used a variety of measurement techniques, including AAS, ICP-OES, ICP-MS, IDMS, NAA and XRF, whereas MAD was the most common sample pre-treatment applied. The results reported by participants were screened for consistency and anomalous values. Results exceeding the interval defined by the robust mean ± twice the robust SD were identified as outliers. The participants providing these results were asked to investigate the possible causes and reported technical issues, such as incomplete dissolution, instrumental instability, interference effects and blank correction, as the reasons for these anomalous results, that were then excluded from the assignment of the reference values. In addition, results obtained by comparison with commercial standards, for which the traceability chain was not established, were also excluded. The reference values were then assigned as the arithmetic mean, for data sets ≤7, or as the median, for data sets ≥8. The associated uncertainties were calculated combining the SD of the results and the uncertainties reported by participants, in the first case, and, in the second case by multiplying the median absolute deviation by 1.25 and dividing the product by the square root of the number of results included. Reference values (±uncertainties) were established as: As (n = 5): 10.57 ± 0.45 μg kg⁻¹; Co (n = 6): 126.07 ± 2.04 μg kg⁻¹; Cr (n = 11): 4.380 ± 0.050 mg kg⁻¹; Hg (n = 10): 15.75 ± 0.47 μg kg⁻¹; Mo (n = 8): 1.548 ± 0.004 mg kg⁻¹; Mn (n = 10): 5.745 ± 0.023 mg kg⁻¹; Ni (n = 17): 2.022 ± 0.025 mg kg⁻¹; P (n = 6): 11.40 ± 0.10 mg g⁻¹; Pb (n = 14): 144.65 ± 0.59 μg kg⁻¹; S (n = 3): 6.873 ± 0.099 mg g⁻¹; Sr (n = 5): 321.04 ± 3.41 μg kg⁻¹; Zn (n = 19): 456.20 ± 1.19 mg kg⁻¹. The relative expanded uncertainties, calculated using a coverage factor k = 2, ranged from 0.6% (Zn) to 8.6% (As). The results of this comparison, while setting targets for the best performance achievable for these determinations, also confirmed the measurement capabilities of most participating institutes, based on the “degree of equivalence”. This index is calculated as the difference between the participant’s reported result and the reference value, divided by the combined values of the result’s and the reference value’s expanded uncertainties. The authors stated that the results of CCQM-K145 can be used to support claims of measurement capability for inorganic elements in similar biological tissue and food samples.

Interlaboratory comparisons are particularly important in new areas of development, such as the reliable determination of nanomaterials in food for the purpose of providing appropriate information to the consumers, according to current EU legislation (Reg. (EU) 1169/2011). Titanium dioxide is an authorised food additive (E171), that may comprise a fraction of nanoparticulates with sizes below 100 nm. Geiss and co-workers²¹ tested the performances of a newly developed screening method, based on spICP-MS, in an interlaboratory study involving 7 experienced participants. The aim of the study was to determine the size distribution and concentration of TiO₂ particles in both the additive itself and in two types of sugar-coated confectionery (sugar-covered button-shaped chocolate candies and white chewing gum dragées). Particular attention was given to the optimisation of the sample preparation, in order to minimise any effect on the size distribution of the particulate. A standard operating procedure was provided to all participants, who applied spICP-MS for the analysis. In addition a set of the same samples was analysed by TEM, as the reference technique. The participants were asked to report the particle mean diameter, the most frequent diameter, the percentage of particles with a diameter below 100 nm, the particle number concentration and other cumulative particle size distribution parameters. The results were evaluated according to ISO 5725, without exclusion of outliers. The reference value and its uncertainty were estimated using robust statistics. For all parameters studied, the repeatability RSD ranged from 2.0% to 13.1%. However, although the reproducibility RSD varied from 6.8% to 26.9% for most of them, it ranged from 31.4% to 52.5% for the smallest and largest particle diameter, the total number of particles and the particle number concentration. The comparison with the results obtained by TEM indicated discrepancies for both the mean and median particle diameter, with spICP-MS reporting significantly higher values. Both the higher particle size LOD for spICP-MS (estimated as between 30 and 35 nm) and the difficulty of overcoming agglomeration in the sample preparation were indicated as possible factors explaining these differences. The authors concluded that the study provided a good estimate of the performance of the spICP-MS method, which was considered a promising starting point for further evaluation.

The measurement of Cd in blood and urine is performed routinely in occupational and environmental medicine. Efforts have been in place for many years to harmonise the performances of laboratories across the world and define requirements for their evaluation in external quality assessment schemes (EQAS). Owing to measures to reduce sources of exposure, both at the work place and from environmental sources such as dietary intake and smoking habits, the levels of Cd in blood and urine of the general population in industrialised countries have decreased, requiring lower LOQs and improved performances for their accurate determination. Cadmium was included in a list of first priority substances defined within the framework of the EU project HBM4EU, involving 30 European countries, aimed to harmonise and promote human biomonitoring studies in Europe. One of the objectives of the project was to develop a network of expert laboratories capable to determine low levels of Cd in clinical samples, with LOQs <0.05 μg L⁻¹ in urine and <0.15 μg L⁻¹ in blood. To this aim, as reported by Nubler et al.,²² an ILC for candidate laboratories (n = 38), and three EQASs, for an extended list of expert participants (n = 58), were organised, each round including two samples each of blood and urine at different concentration. The test materials were prepared from human urine and bovine blood, spiked with known amounts of Cd to reach appropriate concentrations. Homogeneity and stability tests were carried out according to ISO 13528:2015, by measuring the Cd concentrations by ICP-QQQ-MS at m/z 114. In the ILC, the assigned values and their uncertainties were determined from the consensus of the participants’ results (n = 21, urine; n = 19, blood), using robust statistics. For the EQASs, six expert laboratories, selected according to pre-defined criteria of high accuracy and reliable performance, were asked to analyse three samples for each matrix and concentration. The mean of the means of the reported results was used to establish the reference value. Its uncertainty was determined as the RSD divided by the square root of the number of expert laboratories submitting results. The assigned values for the ILC and the EQASs ranged from 0.055 to 0.451 μg L⁻¹, for urine, and from 0.105 to 0.749 μg L⁻¹, for blood, with relative uncertainties, depending on concentration, between 3% and 13%, for both matrices. The participants’ results were evaluated against a target SD of 25%, based on an expert opinion, using z scores, classified according to ISO/IEC 17043. For Cd in urine, the percentage of laboratories obtaining satisfactory z scores was ≥90%, except for the sample at the lowest concentration. The same parameter ranged from 76% to 100%, for Cd in blood. The reproducibility RSD varied between 8% and 36% for Cd in urine and from 9% to 28% for Cd in blood. Among the issues potentially affecting the performance of laboratories, internal standardisation and correction for interference from molybdenum oxide were discussed.

Certified RMs are necessary to support the traceability of analytical measurements and to demonstrate the validity of analytical methods, however, the availability of suitable CRMs is often a problem. In this year’s review, de Vega and co-workers²³ reported the characterisation by MC-ICP-MS of the new RMs IRMM-524A (Fe, 0.1 mm foil) and ERM-AE143 (Mg in HNO₃, 2% w/w) for their isotopic composition and assessed their applicability for the Fe and Mg isotopic analysis of geological and biological samples. First, the isotopic composition of the two new RMs was determined by a direct measurement approach, by comparison with the existing RMs (IRMM-014 and DSM3), using the bracketing technique. For Fe, the results for IRMM-524A showed very close agreement (±2SD, n = 5) for both δ⁵⁶Fe_IRMM-014 (+0.002 ± 0.019‰) and δ⁵⁷Fe_IRMM-014 (+0.003 ± 0.019‰), which was expected as both RMs are from the same original batch of Fe metal. For Mg, the values (±2SD, n = 20) were δ²⁶Mg_DSM3 = −3.291 ± 0.053‰ and δ²⁵Mg_DSM3 = −1.688 ± 0.037‰, indicating a slightly lighter isotopic composition than for DSM3. The intercept method was then applied using solutions of Fe (n = 46) and Mg (n = 17), obtained from a variety of geological or biological RMs and high purity materials. Biological samples (n = 10) were digested, prior to separation of the Fe and Mg isotopes by AEC and CEC, respectively. Geological materials (n = 21) required MAD with HF and HNO₃, as a first step, followed by evaporation to dryness and dissolution in aqua regia, prior to the chromatographic separation. The δ values were then calculated for these test samples, using both the conventional and candidate RMs, and regression analysis was performed. The y intercept provided the difference between the two calibration materials, with results comparing well with the direct approach and agreeing within the measurement uncertainties. The researchers concluded that the intercept method was superior due to the lower uncertainties determined. Therefore they proposed to assign the following values (±2SD) to the new RMs: δ⁵⁶Fe_IRMM-014 = −0.004 (±0.014‰), δ⁵⁷Fe_IRMM-014 = +0.005 (±0.024‰) and δ²⁶Mg_DSM3 = −3.295 ± 0.040‰, δ²⁵Mg_DSM3 = −1.666 ± 0.043‰. Furthermore, the geological and biological RMs analysed in this study represented a wide variety of complex matrices, for some of which δ values for Fe and Mg were reported for the first time, providing a valuable resource for future studies.

Isotopic patterns also provide a powerful tool to identify provenance of materials. The determination of stable isotopes is increasingly used in support to claims of food authenticity and to fight food fraud. Salmon is a highly valued fish and differences associated with its origin (e.g. farmed vs. caught fish, sustainable vs. conventional farming) may determine both customers’ preferences and market price. Potential illegal activities, such as product substitution, illegal harvesting or inappropriate farming, may pose a risk to consumers’ health and damage salmon fisheries and the environment. To support verification of salmon provenance, two new RMs, including certified values for genetic, fatty acid and stable isotope characteristics, are being developed (NIST RM 8256 “Wild-Caught Coho Salmon” and NIST RM 8257 “Aquacultured Coho Salmon”), based on fresh frozen powder homogenates of marine/estuarine and riverine fish, respectively. Aiming to enhance their potential, Christopher et al.²⁴ explored the possibility to determine both S and Sr isotopes in these RMs and the related real samples, exploiting the capabilities of ICP-MS/MS for their simultaneous determination after acid MAD. With respect to other ICP-MS instrumentation, ICP-MS/MS offered the advantage of selecting the ions of interest in the first stage, as well as allowing further selectivity, using O₂ as the reaction gas, in the second stage. Both the S and the Sr isotopes (³²S, ³⁴S and ⁸⁶Sr, ⁸⁷Sr) were measured simultaneously as their molecular oxides (³²S¹⁶O, ³⁴S¹⁶O and ⁸⁶Sr¹⁶O, ⁸⁷Sr¹⁶O) in the salmon-based RMs, thus avoiding interferences at the original m/z values, and were quantified using the bracketing technique. Measurements were carried out on 8 jars for each RM, sampled across the entire production lot using a random stratified sampling scheme. Sample aliquots (1 g of frozen material) were digested in a microwave oven with 4 mL of conc., high purity, HNO₃, at 1600 W. The temperature steps were: 125 °C in 10 min, then holding for 5 min, 210 °C in 10 min, followed by holding for 15 min. The digests, made up to 25 mL with high purity deionised water, were subsequently diluted from 3- to 20-fold. Aliquots (0.2 g) of the isotopic standards NIST SRM 987 (strontium carbonate) and NIST RM 8555 IAEA-S-2 (sulfur isotopes in silver sulfide) were also digested with 3 mL of conc., high purity, HNO₃, using a focusing microwave system, and the digests diluted. The acid content in both the diluted RM and isotopic standard digests presented to the instrument was then adjusted to 3%. The measurement procedure was based on the “concentration gradient method”, involving a sample-standard bracketing approach for calibration, where a series of sample and standard dilutions at decreasing concentrations are analysed. From these sets of data, time resolved analysis generated the regression lines between the concentrations of the respective isotope pairs for each element. Their slopes indicated the isotope ratios, after adjustment for dead time and mass bias correction. The ⁸⁷Sr : ⁸⁶Sr ratio (±expanded uncertainty, 95%) was determined as 0.70736 ± 0.00318 in NIST RM 8257 and as 0.71094 ± 0.00220 in NIST RM 8256. The corresponding values for the ³⁴S : ³²S ratio were evaluated as 0.04401 ± 0.00026 and 0.04436 ± 0.00027, respectively. Plotting the absolute values (±expanded uncertainty) of the ³⁴S : ³²S ratio vs. those of the ⁸⁷Sr : ⁸⁶Sr ratio for the two RMs provided evidence of the almost complete differentiation between the farmed and the wild salmon materials, thus supporting the validity and usefulness of the presented approach. Notwithstanding, the authors pointed out potential problems, such as the natural presence of Rb levels (with two natural isotopes at m/z 85 and 87) in the matrix RMs and real salmon samples, but not in the isotope standards. They also suggested future improvements to reduce the sources of uncertainty and improve instrumental precision.

This year we comment on three publications investigating reference ranges in different biological tissue types. Zaichick and Wynchank²⁵ used five methods to determine the mass fractions of 67 elements in 99 prostate tissue samples (37 autopsy samples from apparently healthy European Caucasian males and 62 samples from subjects with prostate cancer). A systematic analysis combined results from this study with other published data to present mass fraction values for 71 elements in normal and cancerous prostate tissues. Marked differences in element levels were observed between the two tissue types and four categories were proposed: major (or bulk), minor, trace and ultra-trace, for use when referring to the elemental content of the reference man.

In a contrasting study, acid MAD and ETAAS were used to determine reference levels for Ni in brain, stomach, liver, kidney, lung and heart tissue.²⁶ Selected subjects (n = 60) were non-occupationally or environmentally exposed to Ni. Samples were collected from organs with no visually evident pathology or injury within 24 h from death. Ni content was highest in liver and kidney and lowest in brain and stomach tissue.

Acid MAD and GFAAS were applied to determine Al content in brain samples from 20 non-neurologically impaired donors (n = 191, 0.01–9.28 μg g⁻¹ dw).²⁷ Associations between raised brain Al content and neurodegenerative diseases are well established, but this study provides valuable reference data from control brain tissues.

Elemental reference intervals in biological samples demonstrate population-based variability, hence locally derived ranges are commonly applied in clinical practice. This year, multi-element studies evaluated elemental content of blood, serum and urine in population cohorts from Germany, Belgium and Sweden.^28–30 Heitland et al.²⁸ presented concentration ranges for 73 elements in blood, serum, erythrocytes and urine from 102 non-exposed inhabitants in northern Germany. Complementing ICP-MS/MS and ICP-OES methods overcame spectral interferences for some elements observed in previous multi-element studies, however the authors accept that this work is limited by the number of subjects. Reference values and analytical limits (LOD and LOQ) were also derived for 18 elements, measured by ICP-MS, in blood and/or plasma, from 380 adults in Belgium.²⁹ Notwithstanding the modest sample size, the application of an a priori selection of the “reference” individuals, carried out by occupational physicians, applying pre-established inclusion and exclusion criteria, was considered a major strength of this study. Finally, a Swedish study evaluated the variability of 24 h excretion rates for 22 elements in urine samples from 60 healthy non-smokers.³⁰ Elements were quantified using SF-ICP-MS and their levels normalised against creatinine concentrations. Despite the small study size, this work provides potentially valuable information with respect to intra and inter-day variability, as well as intra and inter-individual variation, for the 24 h excretion of trace elements, that adds to previously published work.

A large study by Ha et al.³¹ reported gender and age specific paediatric reference intervals for whole blood trace element levels in infants under 1 year. Samples were collected from 13 446 infants attending routine outpatient assessments between 2007 and 2019 (7206 boys, 6240 girls). Study participants were stratified by age (<6, 6–9 and 9–12 months) and gender. Whole blood concentrations of Ca, Cu, Fe, Mg and Zn were determined by AAS and reference ranges were established according to CLSI standards. This study gives the largest dataset for Ca, Cu, Fe, Mg and Zn in this age group and benefits from using the same sample type and analysis method throughout.

3 Sample collection and preparation

3.1 Collection, storage and preliminary preparation

Dried blood spot samples were introduced for monitoring exposure to Pb more than 50 years ago, but, although the advantages of sample stability and ease of transport are recognised, so too are the problems associated with uneven spreading, potential for contamination and differences between capillary and venous blood. The development of dried blood spot sampling, primarily for Pb but also for other elements, and the analytical techniques that have been used, were described by Parsons et al.¹⁸ in a review that covers the published literature from initial reports in 1970. There is much of interest for those who appreciate the history of developments in science and the review brings the topic right up to date, describing results for a range of elements in samples on non-filter paper collection media, pre-screened plastic micro-collection devices and with modern volumetric absorptive microsampling devices. The authors concluded that these recent developments offered exciting potential for small scale sample collection, but that further work in non-laboratory settings was necessary to validate their use.

3.2 Digestion, extraction and pre-concentration

There has been an increase in the number of papers published compared with last year’s review period, especially papers describing sample pre-treatment involving SPE or LLE. There has, perhaps, also been an increase in the numbers of papers for which the information in the abstract did not match that presented in the body of the paper. The most egregious of such errors was probably the omission of an “m:” LODs were reported as single-digit ng L⁻¹ in the abstract, but were, in fact, single-digit ng mL⁻¹.

There is still interest in optimising extraction conditions. For the ICP-MS determination of Pd-based drugs, which are emerging as alternatives to platinum anticancer chemotherapeutics, in biological matrices (adipose tissue, muscle, liver, kidney, spleen, testis, heart, lungs, brain, blood and serum), the critical variables were optimised³² with two-level factorial and central composite designs. Samples (50 mg) were digested in tubes closed with a screw cap, with 900 μL HNO₃ + 300 μL HCl for 60 min in a 90 °C water bath. The method (LOD 0.001 μg L⁻¹) was validated by the analysis of in-house RMs for which both Pd and Pt were determined in all samples. The researchers pointed out that the method allows the processing of hundreds of biological samples simultaneously, with low reagent and sample consumption. The optimum conditions for the MAD³³ of cocoa beans with HNO₃ + H₂O₂, for the determination of Ba, Ca, Cu, K, Mg, Mn, P, S, Sr and Zn by ICP-OES, were selected by a modified Doehlert design. The figures of merit included residual acidity, DOC and accuracy. The procedure (200 mg of sample, final volume 25 mL) was validated by the analysis of NIST SRMs 2384 (baking chocolate) 1515 (apple leaves), 1570a (spinach leaves) and 8435 (whole milk powder), for which some of the values obtained were statistically different from the certified ones. The LODs ranged from 0.05 (Sr) to 200 (P) mg kg⁻¹. A UAD procedure has been developed³⁴ for the determination of essential elements (Cu, Fe, Mn, Zn) and non-essential or potentially toxic elements (Al, As, Cd, Cr, Pb) in the edible parts of some common vegetables (carrot, radish, cauliflower, pumpkin, and spinach) by FAAS and ICP-OES, in which 200 mg of sample were digested in hot HNO₃ + H₂O₂. The final volume was 25 mL. The method, whose LODs ranged from 0.01 (Mn) to 0.4 (Pb) μg g⁻¹, was validated by the analysis of NIST SRMs 1515 (apple leaves) and 1570a (spinach leaves). All analytes, except Al (not found in pumpkin or spinach) and As (not found in carrot, cauliflower or pumpkin), were found in all samples. The researchers concluded that the method was a valid alternative to “conventional acid digestion methods”, with the advantage of short analysis time and lower reagent consumption. In what is probably the first use of a hydrophilic IL to extract metals from dried foodstuffs – shiitake (Lentinus edodes) and agaric (Auricularia auricula) mushrooms, as well as tea leaves – Ling et al.³⁵ determined Cd and Pb by ETAAS after dissolution of 0.3 g of sample with 3 g of 1-butyl-3-methylimidazolium chloride and 0.5 mL conc. HNO₃, by simply heating on a hot plate for about 10 min. After cooling, “bio-macromolecule substances”, precipitated (with 15 mL H₂O), were filtered (0.22 μm micro-porous membrane) and washed with a further 15 mL of H₂O. The combined filtrates were diluted to 50 mL. The procedure was validated by the analysis of three CRMs TMQC0015 (shiitake mushroom), GBW10089 (agaric mushroom) and GBW10052 (green tea) and applied to materials purchased in local supermarkets, in which both elements were found. The LODs were stated as 0.02 (Cd) and 0.04 (Pb) μg L⁻¹, presumably corresponding to 0.3 and 0.7 μg kg⁻¹ in the solid samples, respectively. Results were also in agreement with those obtained by a lengthy acid digestion procedure involving HNO₃ + HClO₄, evaporation to dryness and redissolution in HNO₃. Two extraction methods for water-extractable As from sea cucumbers were compared³⁶ as was the order of extraction of (a) lipid-soluble and (b) water-soluble As species. The researchers found that extraction with water produced very similar results to extraction with TFA and H₂O₂, but that more lipid-soluble species were obtained if the extraction into hexane preceded those involving more polar solvents. Regardless of which extraction was performed first, a second extraction with DCM and MeOH was performed, so that four fractions were created: hexane, MeOH, TFA and residue. The As species in each fraction were determined by HPLC-ICP-MS and the procedure was validated by the analysis of CRMs NRCC DORM-4 (fish protein) and SQID-1 (cuttlefish). The distribution of As species in various body parts (body wall, tentacles, internal organ, skin and muscle) were also determined. Lin et al.³⁷ devised a “universal” method for the speciation analysis of As in various seafoods (fresh fish, shellfish and shrimp) that involved a MAE/MAD method and a single HPLC separation. For sample preparation, 0.1 g of dried sample was mixed with 6.0 mL of 20 mmol L⁻¹ HNO₃ and left for 12 h, then heated to 120 °C (details of vessels and type of oven are not given) for 30 min. After cooling and centrifuging, the residue was extracted with the same procedure and the two extracts were combined. The final volume was not given, but appeared to be variable (between 10 and 50 mL). This procedure did not change the speciation of As^III, As^V, AB, DMA or MMA, all of which were separated by CEC on a Dionex IonPac TMCS12A (4 × 250 mm) column by isocratic elution at 30 °C with a mobile phase of 5.0 mmol L⁻¹ HNO₃ + 5.0 mmol L⁻¹ EDTA. The presence of a relatively high concentration of an arsenosugar in some samples was confirmed by ESI-MS. This may be the first report of the simultaneous ion-exchange separation of both “anionic” and “cationic” arsenic species. No reference was made to any prior chromatographic work.

There were several reports of extraction of analytes from oils or waxes. In the case of margarine, an ultrasound-assisted DLLME procedure, based on melting of the donor phase, was devised³⁸ for the determination of Cd, Cu, and Fe by HR-CS-ETAAS. The sample (5 g) was sonicated at 60 °C with a mixture of 150 μL of 0.1 mol L⁻¹ HNO₃, EtOH and 3% (v/v) H₂O₂ (2 + 1 + 1). The mixture was centrifuged, then cooled (ice) to solidify the remaining sample, and 20 μL of the extract were taken for analysis. The method, validated by spike recoveries, had LODs of 0.03, 0.52, and 224 μg kg⁻¹ for Cd, Cu, and Fe, respectively. The procedure was applied to the analysis of five real samples, and all three analytes were found in all samples. Visual inspection of the results obtained after wet digestion of the sample and determination by ICP-OES probably shows no significant differences, but no statistical evaluation was presented. An even simpler procedure³⁹ involving dissolution in, or dilution with, xylene (1 + 19) was applied to the determination of Al, Ba, Ca, Cd, Cr, Cu, Fe, Mg, Mn, Mo, Ni, Pb, Si, Ti and V in solid fats (involving heating to 80 °C) and oils by ICP-OES. To avoid extinguishing the plasma, a 5 μL sample was introduced via a conventional nebuliser and an in-house made “high temperature torch integrated sample introduction system (HTISIS)”. The method was validated by spike recoveries and by comparison of the results obtained by an ICP-MS method involving “conventional” HNO₃ MAD (0.5 g sample, 20 g final mass of digest). In only 9 out the 55 cases evaluated on the basis of a t-test, were the results statistically different. The LODs obtained with the HTSIS system were up to 7-times lower that those obtained with a cyclonic spray-chamber. The possible advantages of this “dilute and shoot” method for the determination of elements that are difficult to mineralise with HNO₃, such as Si, were discussed. Carneiro et al.⁴⁰ developed a method in which the emulsion formed by mixing a sample of oil with an aqueous extractant was broken by heating. The optimal procedure, established by a D-optimal mixture design, consisted of vortexing a mixture of 3.0 mL of each oil sample, 1.0 mL of HNO₃ (30%, v/v) and 1.0 mL of Triton X-100 (30%, w/v) followed by heating to 90 °C for 20 min. The aqueous phase was separated by micropipette and analysed directly by MIP-OES. The procedure was applied to the determination of Al, Ba, Cr, Cu, Mn, Ni, P, Ti, and Zn in olive, avocado, babassu, coconut, licuri, primrose, pequi and grape seed oils. The results were compared with those obtained by MIP-OES after wet MAD with an HNO₃ + H₂O₂ mixture. A paired t-test indicated no significant difference. The LODs ranged from 0.4 to 4 μg kg⁻¹. A similar procedure⁴¹ was described for the determination of Cu, Fe, Mn, and Ni in margarine by ETAAS in which 4 g of melted margarine was vortexed with 2 mL of freshly prepared 7% Triton X-114 in 10% HNO₃ and then heated to 80 °C until three phases formed. The acidic aqueous middle phase (approximately 1.5 mL) was collected for the analysis of the metals. The upper organic phase contained only margarine and the lower phase was a surfactant-rich phase. The LODs were 5, 13, 2, and 23 μg kg⁻¹ for Cu, Fe, Mn, and Ni, respectively. The results for 11 samples were compared with those obtained following wet MAD with an HNO₃ + H₂O₂ mixture. No differences were observed for Cu, Fe and Ni, but results for Mn were significantly different.

Several descriptions of digestion in alkaline media have appeared. Although this is not a novel procedure per se, several authors mentioned this aspect of the method in the title of their article. Hwang et al.⁴² determined I in foods by ICP-MS after extraction with TMAH, a procedure they found to be superior to MAD with HNO₃–H₂O₂, which gave inaccurate results for the analysis of the CRMs SRM 1849a (infant/adult nutritional formula) and SRM 1548a (typical diet). The sample (0.1–1.0 g), 4.5 mL of H₂O and 1 mL of 25% TMAH were heated at 90 °C for 3 h. After cooling, the mixture was diluted to 25 mL with water and centrifuged (3000 rpm for 15 min). The method, with an LOD of 0.013 μg kg⁻¹ and LOQ of 0.04 μg kg⁻¹, was applied to the analysis of 123 different food samples. The range of concentrations observed, for the most numerous food categories analysed, were: from <LOQ to 125 μg kg⁻¹ for cereals and grain products (n = 17); from <LOQ to 623 μg kg⁻¹ for vegetables (n = 58) and from 5 to 34 μg kg⁻¹ for beans (n = 13). For seaweeds, although a limited number of samples was analysed, the authors reported that the I content was between 16 and 110 μg kg⁻¹ in three samples each of laver and sea mustard but, as previously reported, was much higher in kelp (n = 4, 3023–4470 μg kg⁻¹). To determine B in foods (almond, grape and orange flour) after dissolution in TMAH, a laboratory-made ETV sample introduction device, consisting of a Fe–Cr–Al coil, was developed⁴³ to overcome the problem of memory effects in ICP-OES. The procedure, optimised with a central composite design, involved the digestion of 200 mg of sample in Teflon vessels with 2.00 mL of 16.2% (w/w) TMAH at 60 °C for 120 min. Following dilution to 5 mL and centrifugation (at 3000 rpm for 10 min), 3.00 mL of the supernatant were removed and diluted to 10.00 mL. The procedure, with an LOD of 2 μg g⁻¹, was applied to the analysis of several real samples; B was only found in the almond flour. Results obtained with a conventional introduction system after MAD with HNO₃ and H₂O₂ were presented, and a visual inspection suggests that they were not significantly different from those obtained by alkali digestion and ETV introduction. A similar protocol was followed⁴⁴ for the determination of Cu, Mo, and Zn in beef. To 0.25 g sample were added 5.0 mL of H₂O and 1.5 mL 25% (w/w) TMAH followed by heating at 85 °C for 60 min, cooling and centrifuging. The final volume was 10 mL. Copper and Zn were determined by FAAS and Mo by ETAAS. The LODs were 0.5, 0.2 and 0.6 mg kg⁻¹ for Cu, Mo and Zn, respectively. The method was validated by the accurate analysis of CRMs ERM-BB184 (bovine muscle) and Embrapa-RM-AgroE3001a (bovine liver) and by comparison of the results with those of an acid MAD procedure (when Mo was determined by ICP-MS after HNO₃ + H₂O₂ digestion). The method was applied to rib plate and top sirloin samples provided by the National Institute of Meat of Uruguay. Chen et al.⁴⁵ extracted iSe species from enriched rice, tea and garlic with KOH, after evaluating a number of dilute aqueous extractants (H₂O, HCl, citric acid and CH₃COONH₄). The sample (0.2 g) was mixed with 10 mL 0.1 mol L⁻¹ KOH and heated in an ultrasonic bath at 78 °C for 30 min, followed by cooling, centrifuging, and filtering. The species were determined by anion-exchange HPLC-ICP-MS. The method showed LODs of 2.5 and 5 μg kg⁻¹ for Se^IV and Se^VI, respectively, and was validated by spike recoveries. When applied to 11 real samples, Se^IV was detected in all of them, but Se^VI only in one. Inorganic Se ranged from 0.6% to 18% of total Se, that was determined after MAD with HNO₃–H₂O₂. For the determination of Cu and Zn in human hair, several UAD procedures were compared⁴⁶ and dissolution in alkali (NaOH) selected as superior compared with water or HNO₃ + H₂O₂. Sample (200 mg) and 4 mL of 2 mol L⁻¹ NaOH were sonicated and then heated at 80 °C for 30 min; after cooling, 10 μL of the digest was spotted on filter paper for determination by LIBS. The LODs were 0.01 μg g⁻¹ and 0.4 μg g⁻¹ for Cu and Zn, respectively. Results for the analysis of three samples were compared with those obtained by a procedure involving MAD (HNO₃ + H₂O₂) and ICP-OES; although no statistical analysis was performed, visual inspection indicates that the differences were not significant. For the dissolution of rice for the selective determination of As^III and total iAs, Costa et al.⁴⁷ dissolved the samples in 40% v/v benzyl trimethylammonium hydroxide in water (pH 14), that was commercially available as a new reagent called Universol®. The successful validation of the procedure by the analysis of CRM IRMM-804 (rice flour) showed that no interconversion of species (including possible degradation of the methylated species to iAs) had taken place. The sample (150 mg) was digested with 0.5 mL Universol® at 60 °C for 30 min. After cooling, the mixture was diluted with H₂O to give a final volume of 5.0 mL, centrifuged (3000 rpm, 15 min) and filtered. The filtrate was then subject to additional processing to selectively extract As^III (see Table 1).

Table 1 Pre-concentration by liquid- or solid-phase (micro) extraction

Element	Matrix	Technique	Extraction mode/reagents	Procedure/comments	LOD/LOQ,^a μg L^−1	Validation	Ref.
a Unless stated otherwise.
Al	Food (rice, potato and spinach) cooked in Al pans	FAAS with slotted quartz tube atomiser	SPME with graphene oxide modified (4-phenyl) methanethiol nanomagnetic composite (Fe3O4@4-PhMT-GO) and elution with HNO3	Acid/peroxide digestion of 1 g. Diluted to 50 mL adjusted to pH 6. Extracted with 20 mg of Fe3O4@4-PhMT-GO (sonication and magnetic separation), solution discarded and analyte dissolved in 0.5 mL of 2 mol L^−1 HNO3. Al found in all samples	2	Comparison with ETAAS and ICP-MS methods and CRM SRM 1570a (spinach leaves) SRM 2670a (rice) SRM 3281 (cranberry fruit)	241
As^V, As^III	Water, bottled	ETAAS	SPME. Magnetic (Fe3O4) graphene oxide functionalized with cysteamine hydrochloride. Elution with HNO3	200 mL extracted with 20 mg at pH 4. Eluted with 0.5 mL 1 mol L^−1 HNO3. Total As determined after oxidation with MnO4^−. Analytes found in all samples	0.001	Spike recovery and CRM GBW08666 (arsenous acid solution) and GBW08667 (arsenic acid solution)	242
As^V, As^III	Rice	HG-AAS	SPE of As^III complex with APDC, in the presence of 0.1 mol L^−1 EDTA (as a masking agent) in 0.5% (v/v) antifoam-A, onto fine particles of tetra-n-butylammonium perchlorate. Redissolution in HCl	150 mg extracted with benzyl-trimethylammonium hydroxide, APDC added and made up to 50 mL. To 10 mL, EDTA and Antifoam added, followed by 0.9 mL of 0.2 mol L^−1 tetra-n-butylammonium bromide and 0.6 mL ion-pairing agent KClO4 (0.9 mol L^−1). A finely dispersed suspension formed that was vortexed, centrifuged, separated and leached with 1 mL conc. HCl. Hydroxylamine hydrochloride was selected as best reagent to convert As^V to As^III. No real samples were analysed	0.01 μg g^−1	CRM IRMM-804 (rice flour)	47
As, Cd	Wine (sorghum, multigrain, millet, sea snake, seahorse, petrel)	FAAS	DMLLE into DES DL-lactic acid/trioctylmethylammonium chloride	5 mL sample adjusted to pH 7, 0.2 g of NaCl (4%) added, followed by a mixture of 400 μL DES (as extraction solvent) and 300 μL methanol (as dispersion solvent) vortexed, sonicated, centrifuged and the upper layer analysed. Analytes found in most samples	0.30 (As), 0.080 (Cd)	Spike recovery	243
As^III	Foods (rice, brown rice, carrot, pepper, flour, milk powder, tomato, cabbage, garlic, chicken liver and fish), water (tap, well, bottled)	HG-AAS	SPME poly(methyl methacrylate-co-2-aminoethyl methacrylate) (PMaema), elution with ethanol HNO3	Sample (300 mg) digested with H2O2 + HNO3 made up to 100 mL. To 30 mL was added 110 mg of PMaema, pH adjusted to 4.3, shaken, centrifuged, aqueous phase removed and analyte eluted with 1.5 mL of ethanol. Analyte detected in all samples except garlic and milk powder	0.003	Spike recovery and CRMs: NIST SRM 1573a (tomato leaves) and NIST SRM 1568a (rice flour)	244
As (inorganic)	Food (rice, edible mushroom and sunflower seeds)	ICP-OES	DSPME of both species on hydrothermally synthesized cadmium sulfide nanoparticles (CdS NPs). Dissolved in HNO3	Dried, ground, sieved, 1 g MAD with HNO3 diluted to 10 mL, pH adjusted to 5, 10 mg of CdS NPs added and purged with N2 before capping, sonicated, centrifuged, separated and CdS NPs mixed with 600 μL of 1.0 mol L^−1 HNO3, sonicated to dissolve and solution made up to 3 mL. Analyte found in all food, though only 2 μg kg^−1 was found in rice	0.5 (As^III), 0.8 (As^V). Values given in abstract are erroneous	Spike addition and CRM ERM-BC211 (rice) NIST SRM 2669 (arsenic species in frozen human urine)	212
Ba, Cd, Cu, Ni, Pb, Zn	Meat (poultry, pork and beef)	ICP-OES	DSPME on graphene oxide modified with sodium hydroxide (NaGO), elution with HNO3	Digestion (1 g) in HNO3. Adjusted to pH 5 and diluted to 50 mL; 25 mg NaGO added, sonicated, centrifuged, separated and eluted with 2.5 mL of 1.5 mol L^−1 HNO3. No Ba, Cd, or Pb found. Cu only found in chicken, Ni and Zn found in all samples. Results of recovery experiments not given	0.01–0.2 μg g^−1	Spike recovery	245
Cd, Zn	Oil (fish, butter and margarine)	FAAS	DMLLE. DES (glycolic acid + mandelic acid) as disperser. Extraction into 3%, v/v HNO3	Sample (2 g) diluted to 7 mL with ethyl acetate. 750 μL disperser and 400 μL extractant were mixed and injected at 45 °C. Analytes found in most samples	0.1 (Cd), 0.2 (Zn)	Spike recovery and CRM EnviroMAT HU-1 (used oil)	246
Cd	Tea (eucalyptus and rosemary)	FAAS with slotted quartz tube atomiser	LLME. DPC complex extracted into DES magnetic nanofluid (choline chloride + phenol mixed with Fe3O4 NPs)	Hot water extraction, filtered. 8.0 mL adjusted to pH 10, DPC (0.05% in ethanol) added, vortexed, 50 μL extractant added, vortexed, magnetic field applied, upper phase decanted, mix with 150 μL 5 mol L^−1 HNO3. Standard additions calibration. It is not clear that Cd was found in any sample	0.3	Spike recovery and comparison of results with an ICP-OES method	247
Cd	Hair, water (tap and mineral), vegetables (cabbage, spinach, tomato)	FAAS	DLLME. Complex with 2-(6-methylbenzothiazolylazo)-6-nitrophenol (MBTANP) extracted into IL (1-hexyl-3-methylimidazolium tris(pentafluoroethyl)trifluorophosphate [HMIM][FAP]) with ethanol as disperser	Acid/peroxide digestion. pH adjusted to 7.0, and to 25 mL of digest was added 1 mL of 0.1% MBTNAP followed by 200 μL of IL and 400 μL of disperser. Ultrasonicated, cooled (ice) and centrifuged. Aqueous layer removed and IL phase diluted to 250 μL with acidified ethanol. Analyte found in all samples except mineral water	0.1	Spike recovery and CRM NRCC TMDA-51.3 (fortified water), NIST SRM 1570A (spinach leaves)	248
Cd, Cr, Cu, Mn, Ni, Pb	Oils (fish and essential: apricot, black cumin, pine turpentine, thyme, mint)	ICP-OES	SPME with gamma-Al2O3 functionalised with fluorescein (Al-Flr) and elution with HCl	Sample (4 g diluted to 10 mL with n-hexane) plus 500 mg of Al-Flr, vortex agitation for sorption and elution with 5 mL 0.3 mol L^−1 HCl. Analytes except Ni and Pb, were found in all samples	0.7 (Cd), 2 (Cr), 4 (Cu), 0.7 (Mn), 1 (Ni), 1 (Pb)	Comparison with MAD method, spike recovery into oil standards	249
Cd, Cr^VI	Beans (lentils, pea, cotyledon, wax bean)	ETAAS	DSPME (magnetic) of the DDTC complexes by strontium hexaferrite modified by CTAB [(SrFe12O19) @CTAB] NPs	MAD of 500 mg in HNO3. Diluted to 150 mL and adjusted to pH 5.5, 1.12 mL 10^−4 mol L^−1. DDTC, and 60 mg SrFe12O19@CTAB NPs were added, stirred and magnetically separated. Dissolved in 3 mL of methanol containing 5% (v/v) HNO3. Analytes found in all samples	0.09 (Cd) 0.08 (Cr) μg kg^−1	Spike recovery and pseudo-RM	250
Cd, Mn	Beverages (green tea and soluble coffee)	FAAS	LLE into aqueous two-phase system of PEG 4000 + K2HPO4	MAD (200 mg) in H2O2 + HNO3. Diluted to 10 mL and adjusted to pH 6.5. Procedure difficult to follow but 11.1 g of polymer, 0.501 g salt stock solution and 0.369 g water were mixed, manually shaken and left in an ultra-thermostatic bath for 18 h at 298.15 K. No Cd found	0.4 (Cd), 6 (Mn) μg kg^−1	Spike recovery, CRM NIST SRM 1575a (pine needles) and NIST SRM 1515 (apple leaves)	251
Cd	Flaxseed flour	AAS with thermospray introduction	LLME of DDTP complex by dodecanoic acid/THF supramolecular solvent	Extraction (200 mg) in 1.5 mol L^−1 HNO3. 10.0 mL mixed with 400 μL of 41.6% (m/v) DDTP vortexed, 1.09 mL of THF containing 45.3 mg mL^−1 dodecanoic acid was added, vortexed, heated at 60.0 °C and frozen. The aqueous phase was discarded and the organic phase was dissolved in 250 μL of EtOH/water (1 + 1 v/v). Cd found in five of six samples	0.1	Comparison with MAD method	86
Cd	Bivalve molluscs, water	FAAS	DLLME of complex with 2-(2-bromo-5-pyridylazo)-5(diethylamino)phenol (Br-PADAP) into 1,2-dichloroethane and trichloroethylene with ethanol as disperser	Sample (200 mg) digested with H2O2 + HNO3 neutralized with NaOH made up to 25 mL. To 5 mL pH adjusted to 5, 70 μL of 0.015% (m/v) Br-PADAP solution and 70 μL of a mixture of 1,2-dichloroethane and trichloroethylene (3 + 2) added, sonicated, centrifuged. Analyte was not detected in any samples	0.4	CRM BCR-713 (wastewater effluent) and NIST SRM 1566b (oyster tissue)	252
Cd, Co, Hg, Ni, Pb	Fish (11 varieties)	ICP-OES	DSPME into pectin-coated magnetic graphene oxide (pectin/Fe3O4/GO). Elution with HNO3	Fillets (0.5 g) were homogenized and digested in 25 mL of HNO3 + H2O2 at 85 °C for 2 h. Liquid portion was separated, vacuum dried to 1 mL, diluted to 50 mL and 1.5 mg of sorbent added. pH adjusted to 6.2, sonicated, sorbent magnetically separated, washed eluted with 1 mL of 1 mol L^−1 HNO3. No analytes found in 6 of the samples	0.01–0.2 μg g^−1	Spike recovery and CRM NIST SRM 1946 (lake superior fish tissue)	253
Cd	Food (sheep milk, dry tea, eggplant, honey, dry beans, rice, parsley, tomato, and spinach)	FAAS	DLLME of complex with patent blue V into 1-tetradecanol (as extraction solvent) in THF (as disperser) and H2O to create a “nanostructured supramolecular” solvent	Sample (1 g or 1 mL) MAD with HNO3 + H2O2 made up to 25 mL. To 5 mL, pH adjusted to 8.2, were added 67.2 mmol L^−1 of Patent Blue V and 480 μL of supramolecular solvent. Sonicated, vortexed, cooled and solidified organic phase transferred and melted. Analyte detected in all samples except sheep milk	15 μg kg^−1	Spike recovery CRM INCT-TL-1 (tea leaves) and NIST SRM 1570a (spinach leaves)	254
Cd species	Blood and urine	ETAAS	SPME with magnetic Fe3O4-supported naphthalene-1-thiol functionalized graphene oxide (Fe3O4@NpSH-GO, NMNT-GO). Elution with HNO3	Organic forms converted to inorganic by MAD with HNO3. To 1 mL, at pH 6, 20 mg of Fe3O4@NpSH-GO was added, sonicated, magnetically separated, analyte redissolved in 100 μL 0.5 mol L^−1 HNO3, diluted to 200 μL with H2O. Analytes found in all samples	0.01	Comparison with ICP-MS method and CRM NIST SRM 955c (caprine blood, levels 1, 2 and 3)	255
Co, Cu, Ni	Soft drinks (grape, apple, turnip juice), spice (cumin, mint, rosemary, thyme, parsley) vegetables (black and red peppers, lettuce, spinach)	FAAS	Magnetic SPME with a polyaniline (PANI)/polythiophene (PTH) co-polymer coating on Fe3O4 NPs. Elution with acidic thiourea	Acid extraction. 50–60 mL of sample adjusted to pH 10 extracted (manual shaking) with 100 mg of Fe3O4@coPANI-PTH, separated by external magnet, solution discarded. Eluted (resuspension, sonication, magnetic separation) with 0.5–1.0 mL of 0.2% thiourea in a 2 mol L^−1 HCl and 1 mol L^−1 HNO3. Co found in all samples except mineral water. No Ni was found in the apple juice, Cu was found in all samples. Only vegetables were analysed for Co	2 (Co), 1 (Cu), 3 (Ni)	Spike recovery and CRM TMDW-500 (drinking water) SRM 1570a (spinach leaves) and SRM 1573a (tomato leaves)	256 and 257
Co	Food (biscuit, chocolate wafer, corn, wheat, herbal tea, spinach mint) beverages (tap water, chocolate milk, cow milk, red wine)	FAAS	LLME of dithizone complex into DES of tetraheptylammonium chloride and oleic acid (THACl + OleA)	Sample (1 g) digested with H2O2 + HNO3 neutralized but final volume not given. To 5 mL, 2.4 μmol L^−1 of dithizone solution (volume not given) was added, 200 μL of DES (THACl + OleA; 1 + 1) was injected, mixed by repeated withdrawal and reinjection, cooled (ice); solidified DES phase was dissolved in 1.0 mL of acidic ethanol. Analyte found in all samples except tap water	0.04	Spike recovery and CRM NIST SRM 1573a (tomato leaves)	258
Co	Urine	FAAS with slotted quartz tube atomiser	LLME of complex with a Schiff base ligand, 2-[(Z)-[(2-aminophenyl)imino]methyl]-4-bromophenol into 1-decanol with CO2 effervescence assistance from tablet made of NaH2PO4·2H2O and Na2CO3	Sample (3 mL) MAD with HNO3 + H2O2 pH adjusted to 5. To 8 mL (buffered to pH 11) was added 2.5 mL of ligand solution (0.87 g L^−1 in ethanol), shaken, 0.50 mL of 1.0 mol L^−1 NaOH added, heated (45 °C), effervescent tablet containing 100 μL 1-decanol added, after gas evolution ceased, vortexed and centrifuged. Upper liquid layer removed and diluted to 150 μL with dilute ethanol (1 + 1). It is not clear whether Co was found in the sample	4	Spike recovery	259
Co, Ni	Food (pea, lentil, rice powdered milk) and beverages (fruit juice, milk, tap water)	FAAS with slotted quartz tube atomiser	DLLME of complexes with pyridylazo naphthol into 1-dodecanol as extractant and a deep eutectic solvent (tetrabutylammonium bromide : acetic acid) as a dispersant with solidification of the floating organic droplet. ACN as demulsifier	Dried, ground, MAD (1 g) with HNO3 + H2O2 made up to 10 mL, pH adjusted to 5, and injected into 75 μL of the extraction solvent (1-dodecanol) and 250 μL of the dispersive solvent (DES). Sonicated, and 250 μL of ACN (demulsifier) added. Cooled (ice). Upper solid layer dissolved in 100 μL ethanol. Analytes found in all samples except tap water and Co in milk	0.2 (Co), 0.4 (Ni)	Spike recovery and CRM NIST SRM 1643b (water), SRM 1568a (rice flour) and FAPAS QC material T07370 QC (brown rice)	260
Cr^VI and total Cr	Waters (tap), beverages (apple juice, cherry juice, red wine, carbonated drink, and sparkling water) and vegetables (mushroom, spinach, beans, tomato, lettuce, carrot, cabbage, celery and zucchini)	FAAS with air-acetylene flame	DLLME of complex with 4-hydroxy-2-[(E)-(4-sulfonato-1-naphthyl) diazenyl] naphthalene-1-sulfonate (azorubine) into supramolecular solvent (THF + tetrabutylammonium hydroxide + 1-decanol)	Acid/peroxide digestion (1 g). 125 mL adjusted to pH 6. Reagents added together with NaCl to adjust ionic strength, sonicated, centrifuged, aqueous layer removed and remaining phase diluted to 1 mL with 1 mol L^−1 HNO3. Analyte not found in any of liquid samples except tap water, but was found in all solid samples	0.03	Spike recovery and CRM GBW10052 (green tea) and NIST SRM 1643e (trace elements in water)	261
Cr^VI	Urine	FAAS	DLLME with solidified floating drop. Complex with DPC extracted into with ethanol as disperser and SDS surfactant/ion-pair reagent	To 10 mL of sample acidified (pH 3–4) DPC, SDS and salt added. Extracted into 100 μL 1-undecanol, sonicated stirred, centrifuged, cooled and solid removed and dissolved in 100 μL of acidified ethanol (0.1 mol L^−1 HNO3). Difficult to decipher optimum conditions. Analyte found in all samples	0.02–0.05	Spike recovery	262
Cr^III, Hg, Zn	Food and beverage (black tea, mineral water, tap water, onion, cheese, and rice)	ICP-OES	SPME by Armillae mellea immobilized nanodiamond	MAD (details not given) of 300 mL sample (pH 5) pumped at 2.0 mL min^−1 though mixture of 200 mg powdered Armillae mellea and 200 mg nanodiamond. Washed and eluted with 5.0 mL of 1.0 mol L^−1 HCl. No Hg found in any samples, Cr found only in tea and onion, Zn found in all samples	0.03 (Cr), 0.1 (Hg), 0.04 (Zn)	Spike recovery and CRM NIST SRM 1643e (trace elements in water) and NCSZC 73014 (tea leaves)	263
Cr species	Milk (cow, powder)	ICP-MS	DSPME with fibrous g-C3N4@TiO2 nanocomposites (FGCTNCs) elution with HNO3 for Cr^III and NaOH for Cr^VI	Total Cr: MAD with H2O2 + HNO3 made up to 60 mL. Residual and digestible Cr: digested with artificial gastric juice, filtered. Filtrate and residues digested as for total Cr. Cr^III and Cr^VI filtrate taken. 20 mL adjusted to pH 8.0 for Cr^III or 3.0 for Cr^VI, 10 mg of FGCTNCs added, sonicated, vortexed, centrifuged, desorbed in 1.0 mL eluent (0.5 mol L^−1 HNO3 for Cr^III or 0.1 mol L^−1 NaOH for Cr^VI). Analytes found in both samples	Milk powder (0.1 g): 110 pg g^−1 (Cr^III) and 260 pg g^−1 (Cr^VI); cow milk (2 mL) 5 pg g^−1 (Cr^III) and 13 pg g^−1 (Cr^VI)	Spike recovery and CRM GBW 10017 (milk powder)	222
Cu	Mushrooms (8 varieties)	FAAS	CPE of complex with 1-(2-pyridylazo)-2-naphthol (PAN) into Tergitol NP-7 surfactant	Acid/peroxide digestion, evaporated to dryness, dissolved in 0.1 mol L^−1 NaOH. 50 mL adjusted to pH 7 containing 0.15% (w/v) Tergitol NP-7, 1.4 × 10^−5 M PAN, warmed (25 °C) centrifuged, decanted and diluted to 2 mL with HNO3 solution containing 20% (v/v) MeOH. Some details of the procedure are missing; but analyte found in all samples	3	Spike recovery and CRM GBW 07603 (trace elements in bush branches and leaves) and GBW 07605 (tea)	264
Cu	Hemodialysis water	FAAS	SPE with powdered Moringa oleifera seeds with elution into HCl	Extractant contained in micropipette tip. 4 mL sample aspirated through 10 mg of adsorbent (250 μm particle size), eluted with 0.2 mL of 0.5 mol L^−1 HCl. No analyte found in sample	0.6	Spike recovery at 30 and 80 μg L^−1	265
Cu	Food (tea, coffee, chocolate, spinach and infant milk substitute)	FAAS	SPE. Magnetic (Fe3O4) polymerized allyl chloride functionalized with imidazole-4,5-dicarboxylic acid	Acid digestion. Other details hard to find. Eluent was 5 mL of 1 M HNO3. Analyte found in all samples	1	Sike recovery and CRM NIES No. 7 (tea leaves)	266
Hg	Fish oil	CVAAS	DLLME. APDC complex extracted into IL 1-butyl-3-methylimidazolium hexafluorophosphate [Bmim][PF6]	Acid/peroxide digestion of 100 mg with final volume 25 mL; 5 mL taken adjusted to pH 2; 120 μL IL added, vortexed, heated, cooled (ice); 60 μL transferred to the HG reaction vessel and 50 μL of 25% (m/v) HNO3 added. HG with SnCl2 solution. No Hg was found in any sample	0.04	Spike recovery IRMM BCR-060 (Lagarosiphon major) and ERM EC278 (mussel tissue)	118
Hg species	Edible seaweed (wakame, sea spaghetti and hijiki)	ICP-MS with HPLC	SPE on an ionic imprinted polymer synthesized by the precipitation polymerization method and using a ternary pre-polymerization mixture containing the template (MeHg), a non-vinylated monomer (phenobarbital), and a vinylated monomer (methacrylic acid). Elution with mercaptoethanol + methanol	Seaweed dried and ground, UAE (200 mg) with 10 mL of 0.1% (v/v) HCl, 0.12% (w/v) L-cysteine, 0.1% (v/v) mercaptoethanol solution. Centrifuged, separated, adjusted to pH 9. Loaded at 2 mL min^−1, column washed and eluted with 2 mL of 0.8% (v/v) 2-mercaptoethanol and 20% (v/v) methanol (pH adjusted to 4.5) at a flow rate of 1.0 mL min^−1. Evaporated to dryness and redissolved in 200 μL of mobile phase 90.4% (v/v) + 2-mercaptoethanol + 10% (v/v) methanol (pH 2.0). Both species found in all three samples	0.007 (MeHg) and 0.02 (Hg) μg kg^−1	CRM BCR-463 (tuna fish)	267
Mo	Lamb	MIP-OES	DLLME of the complex with 8-hydroxyquinoline (8-HQ) into CCl4 with MeOH as disperser	Procedure was optimised by a Box-Behnken design. Preparation was by MAD (100 mg) with HNO3 and H2O2, with final volume 30 mL. To 10 mL (adjusted to pH 3.8) was added 0.1 mL of 1% m/v 8-HQ and a mixture of 90 μL of CCl4 and 480 μL MeOH. Phases were separated by centrifugation (3000 rpm for 3 min) and the 90 μL supernatant, removed by micropipette, diluted to 500 μL with methanol. Analyte was found in all six samples	0.2 μg g^−1	CRM NIST SRM 1577c (bovine liver) and comparison of results with those of an ICP-MS method	268
Ni	Tea (green)	FAAS with slotted quartz tube atomiser	DLLME. Schiff base (E)-2,4-dibromo-6-(((3-hydroxyphenyl)imino)methyl)phenol. 1,2-Dichloroethane (the extraction solvent) was dispersed into the aqueous solution by generating fine droplets by air-assisted spraying	Hot water extract (2 g in 250 mL). 8 mL extract. pH adjusted to 8, 1 mL extractant (0.025% m/v), vortexed, extracted (0.375 g with vortexing), centrifuged, organic phase removed and evaporated, residue dissolved in 150 μL 65% HNO3. No analyte was found in the sample	4	Spike recovery	269
Ni, Pb	Food (cow milk, dry baby milk, tuna, shrimp, sugar (white and brown), salt, tea (black and green)) and tap water	ICP-OES	SPE. Magnetic (iron oxide NP) functionalized with Pleurotus ostreatus (dried and powdered)	Acid digestion (details in a previous publication). 400 mL adjusted to pH 5 loaded onto a column at 2 mL min^−1 and eluted with 5 mL of 1.0 mol L^−1 HCl. No Ni was found in any sample; Pb was found in four of 11 samples	0.02 (Ni), 0.04 (Pb)	CRM NCSZC 73014 (tea leaves), NWTM-15 (fortified water) and NIST SRM 1643e (trace elements in water)	263
Pb	Water (tap) and food (salted peanuts, chickpeas, roasted yellow corn, pistachios, and almonds)	FAAS	LLME. DES (iron^III chloride dissolved in phenol) containing alpha-benzoin oxime as complexing agent	Acid digestion. Adjust to pH 2. 150 μL of DES added to 2 mL of digest (1 g in 10 mL). Vortexed, centrifuged, aqueous layer removed, DES dissolved in 500 μL HNO3. Analyte only found in peanuts and corn	0.008	Spike recovery and CRM TMDA 64.2 (fortified water), TMDA 53.3 (fortified water), and NCSDC-73349 (bush branches and leaves)	282
Pb	Hair	ETAAS	SPE on nanocomposite of ZnCr layered double hydroxide and graphene oxide (GO/ZnCr LDH)	Digestion (30 mg) in H2O2 + HNO3. Diluted to 15 mL and adjusted to pH 6. 10 mg of extractant was packed between two small pieces of glass wool in a pipette tip. 5.0 mL sample loaded, rinsed and eluted with 1.5 mL of 1.0 mol L^−1 NaNO3 at pH 4.0. Analyte found in only two of six samples	0.1 μg g^−1	Spike recovery	270
Pb	Bee products (honey, mead, honey vinegar and honey beer)	ETAAS	DLLME of 1,5-diphenylcarbazide (DPC) complex into a magnetic IL [P6,6,6,14]2MnCl4 with ACN as dispersant	Sample (1 g) digested in H2O2 + HNO3. Evaporated to dryness and dissolved in 50 mL H2O. To 8 mL, in 2 mol L^−1 HCl and 0.5% (w/v) NaCl, was added 35 μL of ACN, 50 μL DPC (50 mg L^−1) and 150 μL of 50% (w/v) [P6,6,6,14]2MnCl4 in CHCl3. Vortexed. Magnetically separated. Analyte found in all samples	0.003	Spike recovery and CRM NIST SRM 1643e (trace elements in water)	240
Pb	Beverages (water, fruit juices)	ETAAS	DLLME of complex with pyridylazo naphthol (PAN) into CHCl3 with SDS as disperser. Emulsion broken with Triton X-100	Sample filtered, 5 mL MAD with H2O2 + HNO3 made up to 30 mL, adjusted to pH 9, 1.0 mL of 0.065 mol L^−1 PAN added followed by 0.5 mL CHCl3 with air-assisted mixing, 0.5 mL of 9.18% w/v SDS, with air-assisted mixing, finally 0.3 mL of 4.0% v/v Triton X-100. Analyte found in most samples	0.025	CRM SLR-6 (river water)	271
Pb	Food (onion, celery, carrot, tomato), water (tap and mineral)	FAAS	SPME into polymerized DES of TBAB and acrylic acid (1 + 2 molar ratio), poly(TBAB-2AA), elution with HNO3	Dried, MAD (mass not given) with HNO3 + H2O2 made up to 50 mL, pH adjusted to 6, 10 mg of poly(TBAB-2AA DES) added, shaken, centrifuged, upper solution decanted, analyte dissolved in 1 mL 5 mol L^−1 HNO3. Analyte found in all samples, except tap water	2	Comparison with ICP-MS method, spike recovery	272
Sb isotope ratios	Biological materials (plants, blood, urine)	HG-MC-ICP-MS	SPE on thiol-functionalized mesoporous silica (TSP) elution with HCl	Sample (500 mg or 3 mL) digested with HNO3 (+HF for plants), evaporated at 75 °C (to prevent Sb volatilization), re-dissolved in 5 mL of 1.5 mol L^−1 HNO3, sonicated and heated (hermetically closed) at 80 °C overnight. Aliquot evaporated at 75 °C dissolved in 10 mL of 0.5 mol L^−1 HCl and 0.5% w/v KI + ascorbic acid. Pumped through 0.2 g of TSP, washed successively with 5 mL of 0.5 mol L^−1 HCl and 6 mL of 2.5 mol L^−1 HCl (to desorb weakly retained elements), Sb was eluted with 6 mL of 6 mol L^−1 HCl	0.008	CRM plants NCS-DC-73349 (bush branches and leaves), BCR-482 (lichen), Clinchek® (urine control level II), Seronorm™ (whole blood II), NIST SRM 1643f (trace elements in water) and NRCC SLRS-6 (river water)	70
Se	Fish	HG-AAS	SPME on magnetic graphene oxide (GO)–gamma Fe2O3 NPs	MAD (600 mg) with HNO3, pH adjusted to 2, diluted to 40.0 mL, 60 mg of MNPs added, shaken, magnetically separated, eluted into 500 μL of a 0.1 mol L^−1 EDTA solution (pH 12), shaken, magnetically separated. No prereduction. Analyte was found in both samples	30 ng g^−1	CRM NRCC TORT-2 (lobster hepatopancreas), IRMM ERM-BB422 (fish muscle) NIST SRM 1577a (bovine liver) and SRM 1577b (bovine liver)	273
Se^IV, Se^VI	Milk formula, water (tap and bottled) and cereals (10 varieties)	ETAAS	DMLLE. DES (choline chloride + phenol) extractant for hydrophobic Se^IV complexed with DDTC	MAE with HNO3 and H2O2. Adjusted pH. After shaking, 0.5 mL of THF (to aid phase separation) was added, and, after further shaking, 0.2 mL of acidic EtOH was added to dilute the enriched analyte phase. Total Se determined after reduction with thiosulfate. Analyte found in all samples	0.03	Spike recovery and CRM BCR 189 (wholemeal flour)	52
Te	Water (tap, mineral)	FAAS with slotted quartz tube atomiser	DSPME into magnetic chitosan hydrogel. Adsorbent dissolved in HNO3	Sample (8.0 mL) adjusted to pH 6, 25 mg hydrogel added, mechanically mixed, magnetically separated dried (12 h), dissolved in 150 μL of HNO3. No information about detection of analyte in sample	20	Spike recovery	274

---

Other reports of the comparison of sample preparation procedures have appeared. For the TXRF determination of 20 elements in tea leaves, Maltsev et al.⁴⁸ compared four sample preparation procedures: suspension in H₂O, open vessel digestion with HNO₃ + H₂O₂, closed vessel MAD with HNO₃ + H₂O₂, and infusion with hot (90 °C) H₂O. All procedures concluded with the transfer of 10 μL to a non-siliconized quartz carrier and drying. The researchers concluded that results for the open-vessel digestion and MAD were similar but that the rapid suspension procedure was unsuitable because of particle size and absorption effects. Results were compared with those obtained by WDXRF (0.5 g of powder pressed into a pellet with boric acid) and only those for Ca, K and P were significantly different. The TXRF methods were applied to the analysis of CRM GBW07605 (GSV-4 tea leaves) and although no detailed statistical analysis was performed, visual inspection of the results shows possible significant differences for just a few elements. Results were presented for the analysis of 19 real samples. For the determination of Cd and Pb in cocoa beans and their derived products (liquor, cocoa powder and cocoa butter) by ICP-OES, two sample preparation procedures, MAD and dry-ashing, were evaluated.⁴⁹ For MAD 0.5 g of sample and 8 mL of conc. HNO₃ were allowed to stand overnight. Then, 2 mL of 30% (m/v) H₂O₂ were added and the vessels transferred to the oven and heated to 170 °C for 25 min. The final volume was not unambiguously specified but might have been 20 mL. For dry-ashing, 2.5 g of the sample were heated in pre-carbonized porcelain capsules on a hot plate and incinerated in a muffle furnace at 450 °C for 15 h. The ash was dissolved in dilute HCl and made up to volume with H₂O in 20 mL volumetric tubes. The methods were validated by the analysis of CRMs INCT-TL-1 (tea leaves) and NIST SRM 1547 (peach leaves). The researchers concluded that the dry-ashing procedure was better, based on the costs involved. The LODs were 0.5 and 7.0 μg kg⁻¹ for Cd and Pb, respectively. Some 90 samples of cocoa beans were examined, and the concentration range found for Cd was <0.0015–1.6 mg kg⁻¹ and for Pb was <0.022–2.5 mg kg⁻¹. For the determination of sodium ethyl(2-mercaptobenzoato-(2-)-O,S) mercurate (thiomersal) in vaccines, four oxidative pre-treatments for conversion to iHg were evaluated.⁵⁰ Of these, procedures involving (1) KBr/KBrO₃ and (2) KMnO₄ in dilute acids, in which 100 μL of vaccine solution was taken and the final volume was 25 mL, were suitable. After conversion, the excess oxidant was removed by reaction with ascorbic acid, and the iHg was determined by CV-AFS. The method was validated by spike recoveries and comparison of the results with those of a method involving MAD of 500 μL of vaccine with HNO₃ + H₂O₂ and making up to 25 mL. The LOD was 0.02 μg L⁻¹. For spike additions to five real samples (containing between 0 and 100 mg L⁻¹ of thiomersal) of 5, 10 and 20 μg L⁻¹, the recoveries ranged from 80.1–106% (method 1) and from 92.5–101% (method 2). Visual inspection of the results suggests that those obtained with the MAD procedure are significantly different from those of one or other of the “open vessel, room temperature” procedures. It is not clear whether concentrations were expressed with reference to Hg or to thiomersal. For the determination of 40 trace elements in bees and beehive products (beeswax, honey, pollen, propolis and royal jelly) by ICP-OES and ICP-MS, four sample digestion procedures were evaluated,⁵¹ three of which involved microwave-assisted heating, and the fourth involved heating in a water bath. For all procedures, the sample mass was 200 mg and the final volume was 10 mL (for ICP-OES) or 20 mL (for ICP-MS). The MAD procedures were as follows: (a) 1 mL 67% HNO₃, 0.5 mL 30% H₂O₂, and 1.5 mL of deionised water added to quartz vessels; (b) 1 mL 30% H₂O₂ and 4 mL aqua regia added to PTFE vessels; and (c) 1 mL 30% H₂O₂ and 4 mL aqua regia also in PTFE vessels. All vessels were heated at to 180 °C for 40 min. For the fourth method, autosampler tubes containing 1 mL of 67% HNO₃ and 0.5 mL of 30% H₂O₂ were heated to 95 °C for 30 min. The methods were validated by the analysis of CRM NIST SRM 1515 (apple leaves) and by spike recoveries. The researchers also monitored residual carbon content and residual acidity and concluded that the digests obtained by all four treatments were suitable for ICP-OES and ICP-MS analyses, except for the hot-water digests of pollen for ICP-MS analysis. They concluded that this method was a fast alternative to MAD for all elements in biomonitoring studies. The LODs for all the elements by ICP-MS were given and the results of the determination of 16 elements in several commercial products were tabulated. Only in pollen was As detected, and only Ce, Cr, Cs and Fe were detected in beeswax. A comparison was made⁵² between ultrasound-assisted DLLME and vortex-assisted emulsification LLME in a method for the determination of trace quantities of Se species in water samples used for the preparation of baby foods and for the determination of total Se in formula milk and baby cereal. Further details are given in Table 1. The researchers concluded that better LOD, RSD, enhancement factor and recoveries were obtained with the UA method and that the “green and low-cost” DES involved did not interfere in the ETAAS analysis.

Two papers by Flores and co-workers describe the application of microwave-induced combustion for sample preparation.^53,54 The first report concerned the determination of Br, Cl and I in granola by ICP-MS.⁵³ The procedure could tolerate a relatively high sample mass (up to 1.0 g) that was pressed into pellets and placed on quartz holders containing 15 mm diameter low-ash filter papers wetted with 50 μL of 6 mol L⁻¹ NH₄NO₃ solution, which acted as combustion aid and igniter solution, respectively. The holders were inserted into the quartz digestion vessels containing 6 mL of absorbing solution (50 mmol L⁻¹ NH₄OH). The vessels were closed, fixed on the rotor and pressurized with 20 bar of O₂. At the end of the process, the solutions were transferred to calibrated flasks and diluted to 25 mL with H₂O. The residual carbon content, determined by ICP-OES, was <25 mg L⁻¹. The method was validated by the accurate analysis of CRMs NIST SRM 1572 (citrus leaves) and NIST SRM 1547 (peach leaves). The LOQs were 25, 0.025, and 0.002 μg g⁻¹ for Cl, Br, and I, respectively, sufficient to detect all three analytes in six real samples, except for I in one. The second report⁵⁴ concerned the determination of Br, Cl, and I in blood by ICP-MS. The MIC procedure was applied to the analysis of blood spots (100 mg) on filter paper and to a study of the stability of the analytes in the spotted blood. No changes in halogen content were observed for samples stored on paper (with and without anticoagulant) in a desiccator at 20 °C for 90 d. A MAD-UV procedure was also applied to 150 mg samples of whole blood. In this procedure, Cd low-pressure microwave-induced discharge lamps (maximum power 10 W) with the emission at 228 nm and 326 nm were inserted into each quartz vessel. Similar results were obtained with each procedure. The LOQs were 0.06 and 0.04 for Br, 14 and 30 for Cl and 0.01 and 0.008 μg g⁻¹ for I, for the MIC and MAD-UV procedures, respectively. These were well below the concentrations encountered in three real samples. The methods were validated by satisfactory spike recoveries of Br, Cl and I, added as inorganic species and also as an organic I standard (T4-levothyroxine). The residual carbon content was decreased to about 8 mg L⁻¹ in the MAD-UV procedure when the digesting solution contained H₂O₂. This was well below that found to cause an enhancement in the analyte signals.

A theme common to all sample pre-treatment procedures is the separation of analytes from interfering matrix components. The severe polyatomic interferences and tailing effects from uranium on the determination of ultra-trace ²³⁷Np, ²³⁹Pu, and ²⁴⁰Pu, in the urine of exposed individuals by SF-ICP-MS and ICP-MS/MS, were removed by SPE.⁵⁵ Two multi-step sample preparation procedures, each of which started with 20 mL of urine, were described, both involving a CaF₂/LaF₃ co-precipitation step prior to dissolution and retention of the analytes on a 2 mL column of anion-exchange resin AG MP-1M. After washing and eluting the uranium, the Np and Pu were eluted with conc. HBr that was evaporated to dryness and the residue dissolved in 1 mL of HNO₃. The researchers demonstrated that to ensure good recoveries, it was important to decompose organic matter during sample preparation. The overall chemical fractionation between ²³⁷Np and ²⁴²Pu for the whole analytical procedure was 0.974 ± 0.064 (k = 2), allowing ²³⁷Np and Pu isotopes to be measured using ²⁴²Pu as a yield tracer with yields of 76 ± 5%. The ICP-MS/MS LODs were 0.025, 0.025, 0.015, and 0.020 fg mL⁻¹ for ²³⁷Np, ²³⁹Pu, ²⁴⁰Pu and ²⁴¹Pu, respectively, for a 20 mL urine sample. The method was validated by the analysis of three Pu-spiked urine reference materials from the Association for the Promotion of Quality Control in Radiotoxicological Analysis (PROCORAD), France, and by comparison of results with those obtained by conventional alpha spectrometry. The procedure developed successfully measured the concentrations of ²³⁷Np, ²³⁹Pu, ²⁴⁰Pu, and ²⁴¹Pu in urine samples collected during decorporation therapy with DTPA, following a Pu inhalation exposure accident in Japan. The researchers concluded that the method developed (twelve 20 mL samples in 9 h) was superior to the alpha spectrometry method, that required 500 mL of sample and took one week for 2 samples. They also concluded that, as it is difficult to measure ²³⁸Pu by ICP-MS due to the isobaric interference from ²³⁸U, and alpha spectrometry cannot distinguish between ²³⁹Pu and ²⁴⁰Pu, for optimal emergency response it would be better to measure long-lived ²³⁹Pu and ²⁴⁰Pu by SF-ICP-MS and ICP-MS/MS, and short-lived ²³⁸Pu by alpha spectrometry. For the determination of Cd, Cu, Ni and Pb in water and food samples (cabbage, spinach, tomato and black tea) by FAAS, Gouda et al.⁵⁶ devised a method involving precipitation of the analytes with 3-benzyl-4-p-nitrobenzylidenamino-4,5-dihydro-1,2,4-triazole-5-thiol (BNBATT). The sample (250 mg) was dissolved in 10 mL of conc. HNO₃ and 3.0 mL of (30%, v/v) H₂O₂. After heating to near dryness at 150 °C for 2.0 h on a hot plate, 10 mL of H₂O were added and, after filtering, the solution made up to 50 mL. Then 2.0 mL of (0.2% w/v) BNBATT solution was added, the pH adjusted to 7 and the mixture allowed to stand for 10 min before centrifuging. The supernatant was removed and the precipitate dissolved in 1.0 mL of conc. HNO₃ followed by dilution to 5.0 mL with H₂O. The method was validated by the accurate analysis of CRMs: NIST SRM 1570a (spinach leaves) and TMDA 51.3 and TMDA 53.3 (fortified waters) from the National Water Research Institute, Environment Canada and by spike recoveries. The LODs ranged from 0.3 to 0.8 μg L⁻¹ (corresponding to 6 and 2 μg kg⁻¹ in the solid samples) and the method was applied to the analysis of some real samples, in none of which was Pb detected, nor could Cd be detected in the black tea. For the determination of Hg in fish sauce, a method, incorporating pre-concentration by DGT, has been developed⁵⁷ in which the Hg was determined by AAS in an instrument (the Advanced Mercury Analyser AMA 254) that trapped (as a gold amalgam) the elemental Hg evolved on combustion of the sample in O₂, followed by thermal desorption from the amalgam. The researchers showed that Purolite S924, a polystyrene-based resin with thiol functional groups, was suitable as a binding gel and that the method could tolerate NaCl concentrations up to 50 g L⁻¹. For 24 h collection, the LOD was 0.07 μg L⁻¹, 10-times lower than the instrumental LOD. The method was validated by spike recoveries and the analysis of CRM ERM-BB422 (fish muscle) and ERM-CC580 (sediment), then applied to the analysis of 10 real samples (diluted 1 + 3 with H₂O), in all of which Hg was detected. The samples were also analysed directly without pre-concentration and no significant difference in the results was observed, though for two of the samples the Hg concentrations were below the LOD. The researchers pointed out that the direct analysis of fish sauce may damage the catalyst of the instrument, as corrosion of some metal parts of the device was observed after only a few samples had been analysed. For the determination of MeHg in fish, Li et al.⁵⁸ developed a rapid screening method in which a similar instrument, the direct sampling Hg analyser (Model 5E-HGT2321, Changsha Kaiyuan Hongsheng Technology Co., Ltd, Changsha, P. R. China), was used. The method was based on the removal of iHg by volatilisation, following reduction with Sn^II, and the assumption that all remaining Hg species were MeHg. Samples (0.5 g) were mixed with 10 mL of 5 mol L⁻¹ HCl (for 1 h in an ultrasonic bath at ambient temperature). After centrifuging (8000 rpm for 15 min), 1 mL of the supernatant was transferred, 1 mL of 10% (m/v) SnCl₂ was added and the solution agitated by ultrasound in a sealed box connected, via an activated carbon filter, to a vacuum pump. Finally, an aliquot of the residual solution was transferred to the Hg analyser. The method was validated by the analysis of CRM QC457B-2 (fish tissue), certified for the MeHg content of 507 ± 28 μg kg⁻¹, spike recoveries and by comparison of the results with those obtained by an HPLC-AFS method with post-column photo-oxidation, that was applied to five real samples and the CRM. The differences were not significant. The LOD was 0.6 μg kg⁻¹.

A large number of descriptions of methods incorporating LLE or SPE were published in the recent review period, for many of which the development was motivated by the need to decrease the LOD of a procedure in which analytes were quantified by FAAS. These and similar studies are summarized in Table 1. Much of the work described in these articles relates to finding the optimum conditions, usually meaning those that produce the highest sensitivity, which were assumed to also produce the lowest LOD, and so the relevant parameters were varied according to some optimisation strategy. Many of the procedures were described as “pre-concentration” though this really only applied in the case of liquid samples, when the analyte concentrations in the solutions introduced to the instrument really were higher than those in the original samples. For solid samples, many procedures represented a dilution: the concentration of analyte in the solution introduced to the instrument was lower than the concentration in the original solid sample. However, it should not be overlooked that these procedures also separated the analytes from potentially interfering matrix components and that this is probably an important feature of the methods developed.

4 Progress with analytical techniques

4.1 Mass spectrometry

Developments in sample introduction strategies for ICP-MS featured in a number of publications over this review period. Two papers described slurry nebulisation ICP-MS methods, in which food samples were finely ground and suspended in a dispersant-containing solution before direct nebulisation into the plasma. This approach offers a high throughput alternative to traditional digestion methods. The first procedure,⁵⁹ applied to the determination of heavy metals in rice and wheat flour, required particle sizes smaller than 3.3 to 3.5 μm to achieve efficient transportation and ionisation and to allow calibration against aqueous standards. The results of the analysis of SRMs ranged from 94% to 107% of the certified values and RSDs were between 0.4% and 6.5%. The LODs were comparable with those obtained by MAD and ICP-MS at 1.1 ng g⁻¹ for Hg and 3.5 ng g⁻¹ for As. In the second report,⁶⁰ a mixed gas plasma (Ar–N₂–H₂) was utilised to reduce the matrix effects that occur with slurry nebulisation. Flow injection analysis minimised deposition of solids on internal ICP-MS parts. The elements, Cd, Cu, Fe, K, Mo and Zn in 0.05% Na solution, were accurately determined with LODs from 0.02 mg kg⁻¹ (¹¹²Cd) to 40 mg kg⁻¹ (³⁹K) bulk sample and the method was further validated using a wheat flour SRM. Another study compared ICP-MS response factors across three types of nebuliser for various S-containing biomolecules relative to an inorganic S standard and made use of interference-free S detection with ICP-QQQ-MS.⁶¹ While the two regular flow nebulisers evaluated (concentric and cross flow) resulted in significantly lower response factors for proteins and hydrophobic peptides, a total consumption nebuliser demonstrated similar response factors across all molecules. This highlights the importance of selecting an appropriate sample introduction system if specific standards are to be avoided for biomolecule quantification.

Several reports described advances in sample introduction systems for time resolved ICP-MS analysis to determine elemental content of single cells or to quantify numbers or sizes of NPs. Sun et al.⁶² reported a bespoke sample introduction system, consisting of an ultralow volume autosampler with a controllable flow rate (5 to 5000 μL min⁻¹) and a customised cyclonic spray chamber with no baffle. This arrangement achieved transfer efficiencies up to 90%, enabling application to in vivo monitoring of AuNP dynamics in the blood of a single mouse. Size and concentration LODs for AuNPs in ultrapure water were 19 nm and 8 × 10⁴ particles per L, respectively. Two further groups investigated sample introduction strategies to reliably generate single cells prior to spICP-MS detection. Yu et al.⁶³ described a 3D water-in-gas microfluidic device that used Ar as the continuous phase, allowing online coupling with ICP-MS. The flow rates of gas and water could be adjusted to reduce the probability of signals coming from multiple cells. The approach was applied to the measurement of Cd and Zn in Cd-treated HepG2 cells and there was good agreement with total elemental concentrations determined by acid digestion and conventional ICP-MS. Another novel approach involved a droplet splitting microchip coupled to ICP-MS for quantitative determination of released Fe/Pt in single cells treated with FePtNPs.⁶⁴ To accomplish this, the device contained a T-junction channel designed to symmetrically split droplets following cell lysis. Then a magnet was used to separate released Fe/Pt from parent FePtNPs for ICP-MS detection. The device also had a second mode of operation to allow quantification of total Fe/Pt in single cells. The results obtained for released Fe/Pt in cells compared well with those determined by conventional ICP-MS after ultrasonic cell lysis and magnetic separation.

Following the trend in recent years’ reviews, ICP-QQQ-MS continues to have an increasing scope of applications owing to its advantages of effective interference removal and increased sensitivity. The techniques, scICP-QQQ-MS and spICP-QQQ-MS, were employed to quantitatively evaluate the uptake and biotransformation of TeNPs in bacteria, S. aureus and E. coli.⁶⁵ For scICP-MS, the elements Te and P (a constitutive cell element to estimate total number of bacterial cells), were detected in sequential runs with interference on P being overcome using O₂ mass shift mode. A high performance concentric nebuliser with a total consumption spray chamber enabled introduction of intact bacteria into the plasma with a transport efficiency of up to 60%. The LOD achieved for Te was 0.068 ± 0.008 fg cell⁻¹. Following cell lysis, spICP-MS analysis was performed to determine the number of TeNPs per cell and, in conjunction with TEM, to estimate the dimensions of the rod shaped TeNPs. The size LOD calculated (assuming cylindrical morphology) was 28 nm.

Over this review period, three publications involved ultra-sensitive detection of Ti by ICP-QQQ-MS using mixed O₂ and H₂ reaction gases. With this approach, the ⁴⁸Ti to [⁴⁸TiO]⁺ mass shift is monitored on m/z 64 while oxide interferences, e.g., [CaO]⁺, [SO₂]⁺ and [PO₂]⁺, are removed by a secondary mass shift reaction with H₂. The study by Fu et al.,⁶⁶ demonstrated greater sensitivity and lower background with this gas mixture compared with the more commonly used NH₃/He gases. The LODs for Ti in human serum ranged from 0.78 ng L⁻¹ for ⁴⁸Ti to 7.20 ng L⁻¹ for less abundant ⁵⁰Ti and RSDs were between 2.0 and 4.2%. Results were verified by comparison with SF-ICP-MS measurements. A similar interference removal strategy was used for the determination of TiO₂NPs by spICP-QQQ-MS in radish plant tissues,⁶⁷ where size LODs were reduced to 15 nm in pure water and 21 nm in a 50 mg L⁻¹ Ca containing matrix; the latter represented a significant improvement vs. published HR-ICP-MS methods. Meanwhile in other work,⁶⁸ a smaller size LOD (40 nm) for TiO₂NPs in a milk matrix was achieved with an spICP-QQQ-MS method with respect to an spHR-ICP-MS method (55 nm). Both methods were applied to determination of TiO₂NPs in food products containing the additive E171: chewing gum; chocolate and cake decorations.

Another ICP-QQQ-MS application that featured this year was ultra-trace detection of radionuclides for biomonitoring and radiological emergency response. A thorough validation of an ICP-QQQ-MS method for determination of ²²⁶Ra in 0.5 mL urine, following only a dilute and shoot procedure, was reported.⁶⁹ The use of ICP-QQQ-MS in no gas mode effectively removed polyatomic interferences due to the increased abundance sensitivity. An impressive LOD of 0.007 ng L⁻¹ was achieved, which is well below recommended action levels for ²²⁶Ra detection and lower than that reached by ICP-QMS. Good agreement was obtained with target values for SRM-spiked urine (bias from −4.7% to 6.2%). In another report,⁵⁵ the capabilities of ICP-QQQ-MS and SF-ICP-MS were compared with regard to the rapid analysis of ²³⁷Np and Pu isotopes, using yield tracer, ²⁴²Pu, in small volume urine samples. Prior removal of isobaric ²³⁸U was performed by way of CaF₂/LaF₃ co-precipitation and anion exchange chromatographic separation. The ICP-QQQ-MS method achieved similar LODs to SF-ICP-MS, between 0.015 and 0.025 fg mL⁻¹ for 20 mL of urine. The relatively small urine volume and the speed of analysis offer significant advantages over the traditional method of alpha spectroscopy.

The feasibility of using ICP-QQQ-MS to determine Sr and S isotope ratios to differentiate NIST salmon RMs (8256 wild caught and 8257 aquacultured) for food fraud testing was investigated.²⁴ Sample preparation consisted of MAD and dilution with no further matrix reduction or isotope separation strategies. In O₂ shift mode, oxide ratios, [⁸⁷Sr¹⁶O]⁺/[⁸⁶Sr¹⁶O]⁺ and [³⁴S¹⁶O]⁺/[³²S¹⁶O]⁺, were monitored to avoid isobaric interference on Sr from Rb and Kr, and polyatomic interference on S. The concentration gradient method, a measurement and data evaluation procedure for high precision isotope ratio determination with ICP-QMS, was followed and a standard sample bracketing calibration, in a manner similar to MC-ICP-MS, was employed. Accuracy of the Sr isotope measurements was validated by analysis of SRMs, JCp-1 (coral) and NASS-6 (seawater). There was reasonable agreement between the absolute ⁸⁷Sr : ⁸⁶Sr isotope ratios measured compared with MC-ICP-MS, however, precision on the ⁸⁷Sr : ⁸⁶Sr isotope ratios was notably poorer with ICP-QQQ-MS (in the order of 0.1 to 0.2% relative standard uncertainty for salmon RMs).

Developments in isotope analysis by MC-ICP-MS included novel couplings for sample introduction and simplified sample purification methods. Colin et al.⁷⁰ used HG-MC-ICP-MS for precise determination of Sb isotope ratios in complex solid matrices, including biological CRMs. Following MAD, a novel single-step purification method involving thiol-functionalised mesoporous silica powder efficiently isolated Sb from other potentially interfering elements. Employment of HG, a highly efficient sample transport system, enabled small quantities of Sb (40 to 100 ng) to be detected with increased sensitivity compared with the use of wet plasma or desolvation strategies. The HG-ICP-MS method achieved an LOD of 0.008 ± 0.001 μg L⁻¹ for total Sb. External reproducibility from independently digested and purified CRMs was found to be 0.05‰ for δ¹²³Sb. Another interesting piece of work⁷¹ reported total Cu and Cu isotope determination in microvolumes of serum (2 to 15 μL) for the purpose of diagnosis and treatment follow-up in Wilson’s disease. The total Cu method involved direct injection of only 1 μL of pre-treated serum followed by time resolved ICP-MS analysis, achieving an LOD of 3 μg L⁻¹. Isotopic Cu determination was performed by fsLA-MC-ICP-MS, where 1 μL of treated serum was double deposited onto pure Si wafers before ablation using a high repetition rate fs laser. Optimised LA parameters were: repetition rate, 6 kHz; speed, 30 mm s⁻¹ (5 μm spot separation) and fluence, 0.7 J cm⁻². Accurate determination of δ⁶⁵Cu in NIST 3114 at 1 mg L⁻¹ was achieved with an internal precision of 517 ppm. Two further papers described validation of procedures to isolate Ca and Mg in biological fluids and tissues prior to isotope analysis by MC-ICP-MS. The method by Grigoryan et al.⁷² consisted of sequential chromatographic separation of Ca and Mg from a single sample aliquot. Calcium isotope measurements were performed in cold plasma conditions to mitigate the occurrence of Ar-based ions. Expanded uncertainties for δ²⁶Mg and δ^44/42Ca were ≤0.12‰ and ≤0.29‰ respectively and good agreement of the Ca isotope data vs. double-spike TIMS was observed. Meanwhile, Goff et al.⁷³ developed a new three-stage elution method to separate Mg from the sample matrix and applied it to biological CRMs. External precision was reported to be 0.03‰.

Matrix effects leading to signal suppression or enhancement are well known in ICP-MS analysis of clinical materials. A detailed study by Seregina et al.⁷⁴ systematically investigated the impact of HNO₃, inorganic salts and organic compounds on ICP-MS response as well as considering strategies to reduce non-spectral interferences. A combination of optimised “robust” instrumental parameters (most importantly increased radiofrequency power and decreased Ar carrier gas velocity) achieved sufficient reduction of matrix effects to allow use of a single IS for all elements regardless of IP, albeit with a reduction in sensitivity compared with standard instrumental parameters. In other work,⁷⁵ Huang and co-workers assessed the use of internal standardisation to correct for matrix effects in spICP-MS analysis of complex biological samples (enzyme-digested liver tissue, urine and plasma). Applying time resolved analysis, ¹⁴⁰Ce was monitored in single mass mode, while the IS, ¹⁰³Rh, and ¹⁴⁰Ce, were monitored in dual mode. Experiments comparing mass concentrations of CeO₂ NPs by spCP-MS and conventional ICP-MS following MAD, demonstrated that using an IS was an effective and practical way to correct for matrix effects and improve the accuracy of size characterisation of CeO₂ NPs.

Whilst most of the progress with mass spectrometry in this review period related to ICP-MS, two studies employed TIMS for elemental analysis of biological materials. The first developed a method using ¹¹⁶Cd–¹⁰⁶Cd double spike correction for isotopic characterisation of Cd in biogenic CRMs.⁷⁶ A Cd separation procedure prior to isotopic analysis removed interfering Sn. The double spike was added to all samples before digestion at a 1 : 1 mass ratio to avoid Cd isotopic fractionation during the ion exchange purification. Intermediate precision for δ^114/110Cd was −0.005 ± 0.029‰, which was favourable with respect to MC-ICP-MS. Excellent agreement with literature values for CRMs, including pig kidney (ERM-BB186) and spinach leaves (NIST SRM 1570a), was observed. In another work,⁷⁷ ID-TE-TIMS was applied to the direct quantification of ultra-trace concentrations of radioactive ⁹⁰Sr in μL droplet samples. The set up obviated the need for a chemical separation procedure to remove isobaric or polyatomic interferences and use of the total evaporation method enhanced the detection efficiency. The method achieved an LOD of 0.029 fg and was applied to low volume biological samples such as tears, eye lashes and saliva.

4.2 Atomic absorption and atomic emission spectrometry

Although AAS is now a very mature technique there continue to be some new developments. Most relate to CS-AAS for multi-element analysis. Zambrzycka-Szelewa and co-workers⁷⁸ determined Ca, Cu, Fe, K, Mg, Mn, and Na in Polish beer by CS-FAAS. With the advantages of multi-element measurements and extended range of calibration curves, compared to conventional FAAS, results were obtained rapidly and with acceptable LODs, from 0.45 (Mn) to 94 μg L⁻¹ (Na). The analysis of a Polish mixed herb RM gave results between 94.6% and 103.7% (for microelements) and between 97.4% and 100.6% (for macroelements) of the stated reference values, with SDs between 1.1% and 7.3%. Element concentrations were reported for beers from Poland and nine other countries. Concentration differences were up to 30-fold.

An unusual application, determination of S in hair by HR-CS-ETAAS, was reported by Marrocos et al.⁷⁹ The procedure featured molecular absorption by CS at 257.961 and 258.033 nm. Samples were washed according to the IAEA methodology and digested with HNO₃ and H₂O₂. To avoid losses of volatile CS₂, Na₂WO₄ and Pd–Mg were used as permanent and diluent modifiers, respectively. The concentrations found for the CRM NCS DC73347a were reported and said to be consistent with the certified value, with readings at either wavelength. The published paper has few other results with details such as the LOD included in the ESI. Two publications described solid-sampling CS-ETAAS, one for the analysis of milk powder and infant formula, the second for dried blood spots. In the first, Gómez-Nieto et al.⁸⁰ reported measuring the concentrations of Cd and Cu. The method was comprehensively validated using ERM-BD150 skimmed milk powder, with optimisation of the temperature furnace heating programmes and chemical modifiers. The analytes were determined sequentially with Cd atomised first, the furnace allowed to cool and then reheated to atomise the Cu. Results were compared with those obtained when samples were first digested and analysed as liquid aliquots. The values obtained for the CRM agreed with the certified concentrations. In real samples, the measured Cu content was consistent with statements on the labels, whereas all Cd levels were below the analytical LOD. Vieira et al.⁸¹ developed a procedure for the simultaneous determination of Fe and Zn in dried blood spot samples. The work undertaken included assessments of blood volume and spot size, reproducibility among discs cut from the same dried spot, potential for contamination as well as validation of the heating programme and modifiers, used for the analysis. As the concentration of Fe and Zn are high in whole blood, the analytical lines at 307.572 nm and 446.16 nm were used for Fe and 307.589 nm for Zn. Samples were also analysed following acid digestion and measurement by HR-CS-FAAS. Concentrations given by the two procedures agreed at a 95% confidence level.

One other report of note employed HR-CS-AAS, using an electrically heated quartz furnace, to determine Se in food and environmental samples.⁸² The procedure involved generation of SeH₂ in a reaction cell connected to the quartz atomiser by a length of Perma Pure MD-050-48 Nafion membrane tubing, that removed any water vapour. Air was displaced from the reaction cell and atomiser by purging with Ar for 20 s before adding NaBH₄ to initiate the reaction and measurement, at 950 °C and a wavelength of 196.026 nm. A number of food and drinking water CRMs were analysed and the results were found to be between 94% and 105% of the certified values. Reported LODs were given as applicable to the sample volumes or masses taken for analysis and were 0.125 μg L⁻¹ or 0.062 mg kg⁻¹. Much of the technical details and results were published in the ESI report and not in the main publication. Despite the performance of this elegant approach, it is unlikely to be widely adopted when there are other well-established procedures already in use.

Xing et al.⁸³ described a procedure to measure Cd concentrations in grain samples that has similarities to that of Chirita et al.,⁸² but also unlikely to be much used. Xing and colleagues also used a laboratory designed electrically heated quartz tube atomiser, but this was mounted into a conventional AA spectrometer. Their device comprised a quartz tube wrapped with Ni–Cr heating coils, connected to the AA spectrometer via a similarly heated narrow tube. The analysis began as a sample boat with solid sample entered the first section of the quartz tube, kept at 35 °C. With a flow of air at 300 mL min⁻¹, the sample was dried for 20 s and the temperature was increased to 725 °C during 55 s to destroy the organic content of the sample and produce the Cd vapour. The Cd vapour was trapped by Kaolin filler in the distal section of the quartz tube. With an increase in temperature and a mixed N₂–H₂ atmosphere, the Cd vapour was conveyed to a N₂–H₂ flame in the spectrometer for atomisation. The work to develop the device and optimise the analytical conditions was described in detail. The LOD was 0.15 ng in a 200 mg sample. The analysis of 13 CRMs, with concentrations from 7.9 to 1280 ng g⁻¹, gave results ranging from 95% to 103% of the certified values, with RSDs between 1.1% and 9.6% (n = 6).

More than 30 years ago, a device for two-stage electrothermal atomisation was developed and became commercially available. In this latest review period, two groups have published work using an atomiser based on similar principles. Using a system designed in 2015 by Zakharov, Irisov et al.⁸⁴ described the measurement of As in urine which avoided lengthy sample preparation. Sample and modifier were placed into a graphite furnace in a HR-CS-AA spectrometer. Arsenic atoms formed at 2300 °C condensed on a tungsten probe positioned above the furnace’s sample introduction hole, while matrix components were removed by a flow of Ar gas. The probe was lowered into the furnace and reheated for a second atomisation, and interference-free measurement. The LOD with a 15 μL sample was 1.5 μg L⁻¹. Volzhenin et al.⁸⁵ used the same device to show that mussels harvested from Peter the Great Lake did not contain Cd, Pb and Zn levels above 8.21, 31 and 475 μg g⁻¹, respectively, and were therefore safe for human consumption.

Lemes et al.⁸⁶ used a thermospray flame furnace device to measure concentrations of Cd in flaxseed flour by AAS. Most of the work described was for the sample preparation but with an LOD of 0.10 μg L⁻¹ and a short sample preparation, the procedure was considered to be simpler, faster, low-cost and environmentally friendly compared to alternatives using acid digestion and ETAAS or ICP-MS.

There is little new to report for MIP-OES, but the work of Akhdhar et al. is of interest.⁸⁷ To determine F, a solution of Ca(NO₃)₂ was mixed with the test solution in the spray chamber. The CaF molecular emissions were investigated to determine lines with best sensitivity and minimal interference, leading to the choice of 530.455 nm. The optimised conditions were applied to the determination of F in infused tea solutions. In the absence of a suitable CRM, samples were also analysed using HR-CS-GF molecular absorption spectrometry, giving results that were not significantly different. Reproducibility is, however, less impressive with a reported RSD of 28%. The assay is deemed appropriate for use where the more expensive equipment is not available although the authors indicate that further work should be carried out to improve precision.

The report of Gorska et al.⁸⁸ included a description of different instrumental designs for GD-OES that may be used as alternatives to larger, more expensive equipment, opining that while flowing liquid anode-atmospheric pressure glow discharge (FLA-APGD) affords the lowest LODs, flowing liquid cathode-atmospheric pressure glow discharge (FLC-APGD) is resistant to matrix effects and achieves LODs similar to those obtained with ICP-OES. The discharge in the FLC-APGD analyser constructed by these workers was sustained in an open-to-air chamber between a tungsten nozzle and a vertically oriented tungsten rod. The acidified sample solutions served as the cathode and were pumped through the tungsten nozzle. Emitted radiation was captured by an imaging spectrometer. The system was applied to measurements of Ca, K, Mg and Na in fruit juices. The respective LODs were 5.69, 0.14, 0.63 and 0.02 μg L⁻¹ with RSDs of 1.02, 1.1, 0.78 and 0.89%. Samples were also analysed using ICP-OES giving consistent results for K and Na but lower values for Ca and Mg with many of the sample types. When repeated using calibration by standard additions, acceptable results were obtained, and the authors suggested that these elements may have formed strong complexes with organic acids in the samples. For similar reasons of size and cost, Peng et al.⁸⁹ considered GD-OES as an alternative technique to ICP-OES and ICP-MS. These workers adopted solution anode glow discharge (SAGD) as the radiation source as the sensitivity is high for volatile-forming elements, to determine concentrations of Pb in blood. A hollow titanium tube served as the cathode and was connected to a commercial hydride generator from which the volatile analyte was formed. Emission from the plasma was detected by a portable charge coupled device. The LOD, at 0.061 ng mL⁻¹, was shown to be similar or superior to other techniques except for HG-AFS, and the RSD at 40 μg L⁻¹ was 2.2%. Analysis of three CRMs gave results within the certified ranges and results from blood samples were also in agreement with those measured using ICP-MS. A similar arrangement, in principle, was designed by the same research group except that the samples were placed onto a tungsten coil for ETV, and the analyte transported by carrier gas to the discharge space.⁹⁰ The analyser was used to determine Cd and Pb in blood. The LODs were 0.4 and 1.2 μg L⁻¹ for Cd and Pb, respectively and RSDs were below 5%. Analysis of a whole blood CRM gave results within the certified ranges. Three samples were also analysed by ICP-MS, Cd was detected only in the sample with the highest concentration. The results for this analysis and for the measurements of Pb were in agreement with the ICP-MS results.

Cezario et al.⁴³ referred to the tendency for B to adhere to equipment, with attendant memory problems for analytical measurements. They suggested that memory effects are reduced with electrothermal vaporisers but that tungsten coils have limited lifetimes as they are prone to cracking. These workers developed an iron–chromium–aluminium vaporiser, coupled to a conventional ICP-OES analyser. It was suggested that the vaporiser would not have B-memory effects as it does not contain C and that Fe has previously been used as a modifier for B measurements by ETAAS. Experiments to demonstrate the robustness and memory properties of this coil vaporiser confirmed its fitness for purpose and the operating conditions were optimised. Using a 10 μL sample the LOD for B was 0.3 mg L⁻¹. Concentrations of B in samples of almond flour, prepared for analysis by dissolution in 16.2% w/w TMAH at 60 °C for 120 min were 9–18 μg g⁻¹ and were equivalent to the results determined by ICP-MS after acid digestion. No B was detected in orange and grape seed flour samples by either procedure.

4.3 Laser induced breakdown spectroscopy

Within this review period, the number of papers covering LIBS has significantly increased, perhaps reflecting the development and maturity of the technique. The application of LIBS as a clinical diagnostic tool has received significant attention. Chu and co-workers⁹¹ utilised LIBS for blood cancer detection after identifying CN, Ca, H, K, Mg, N, Na and O spectral lines as key for differentiation. Serum was spotted on to a boric acid pellet and allowed to dry. Samples from controls (n = 6), acute myeloid leukaemia (n = 6), chronic myelogenous leukaemia (n = 6), multiple myeloma (n = 6) and lymphoma patients (n = 8) were measured and a number of statistical approaches were applied. It was shown that the random subspace method combined with LDA provided identification of blood cancer type with 91% accuracy, which was a powerful result for conditions which usually rely on bone marrow biopsies for identification. In a similar approach, Chen et al.⁹² simply spotted serum onto filter papers for diagnosis and staging of multiple myeloma using the emission lines of C, Ca, CN, H, K, Mg, N, Na and O. Samples from 55 controls and 75 patients were analysed. Using PCA first to rationalise the data, k-nearest neighbour, SVM and ANN classifiers were applied, with all three achieving accuracies of >90%. Melanoma was the focus of a research by Khan et al.⁹³ who utilised LIBS with formalin fixed paraffin embedded tissues from normal samples (n = 110) and stage III C patients (n = 110). It was noted that enhanced signals for Ca, K, Mg, Na and P were observed in the melanoma samples. A number of statistical classification methods were then employed using 27 emission lines, with ANN and PLS-DA demonstrating 100% accuracy. Yue and co-workers⁹⁴ investigated the use of blood plasma for ovarian cancer diagnosis. The plasma was spotted onto graphite plates and dried prior to LIBS analysis. Samples were taken from 79 healthy volunteers, 34 patients with ovarian cysts and 63 with ovarian cancer. It was found that K was a key marker for differentiating between healthy and patient samples, with Ca, Mg and Na proving useful for separating cyst and cancer samples. A two-stage model was optimised and implemented, that achieved diagnostic sensitivity and specificity of 71.4% and 86.5%, respectively. This outcome was reasonable, but not as powerful as in previous studies discussed here. The reasons for this were not discussed by the authors. Alsharnoubi et al.⁹⁵ employed LIBS in the screening of serum samples for Ca, Cu, Fe and Zn for beta-thalassemia detection. Controls and patient samples were analysed, showing that Cu and Fe were statistically significantly higher in patients whereas Ca and Zn were lower.

Porizka et al.⁹⁶ uniquely combined elemental nanoparticle tagging with LIBS for cancerous tissue identification. In this work, the HER2 biomarker for breast cancer was used to demonstrate the feasibility of LIBS in comparison to traditional immunohistochemistry and immunocytochemistry techniques. The streptavidin-conjugated upconversion nanoparticles containing Y were incubated with HER2 positive and negative cell lines. The data provided unambiguous results and excellent agreement with histological methods with the additional benefit of multiplexing in the future by using additional elemental tags.

Zhang and co-workers⁴⁶ reported the determination of Cu and Zn in hair using LIBS following ultrasound-assisted alkali dissolution. Initially, three extraction solvents were compared, namely water, 2 M NaOH and 3 : 1 HNO₃–H₂O₂, using the CRM GBW09101b (human hair). The slurries were sonicated for 30 min, followed by heating at 80 °C for 30 min, then 10 μL of the resultant solution was dropped on to filter paper. Calibration was achieved through standard addition and the method was validated with the CRM and complementary analysis by ICP-OES. The alkali conditions provided higher signal responses than with water or acidic conditions, leading to the lowest LOD and RMSECV. The authors noted this could be due to the improved solubility of hair in alkali media leading to more energy available to excite the analytes, as well as the increased viscosity of the NaOH solution, which provided a lower penetration radius on the filter paper. The LODs achieved were impressive for LIBS: 0.0146 μg g⁻¹ Cu and 0.3517 μg g⁻¹ Zn. The researchers used the alkali approach on three human hair samples and compared the results against ICP-OES with relative errors of 4.9% for Cu and 7.6% for Zn, demonstrating the suitability of the method with quick sample preparation times.

Within this review period, a significant number of papers were published covering the quantitative analysis of food products by LIBS. Su et al.⁹⁷ considered Sargassum fusiforme (hijiki algae) for analysis by LIBS to assess the levels of As, Cd, Cr, Cu, Hg, Pb and Zn as hijiki is a popular health food product in Asia. A PLS calibration algorithm was implemented for quantitative analysis using the values from ICP-MS analysis. The workers also evaluated the impact of a noise elimination approach (threshold variables) and three variable selection methods (successive projections algorithm, uninformative variable elimination and variable importance in projection) for the LIBS spectra to improve the PLS modelling. However, it was found that the variable selection models had an adverse effect on the accuracy and prediction capability. Therefore, it was recommended to only use the noise elimination with PLS for hijiki samples. The models were effective for Cd, Cr, Cu, Hg, Pb and Zn but As did not perform well, yet the authors did not offer an explanation for this finding. Gamela et al.⁹⁸ determined Cu, K, Sr and Zn in cocoa beans by direct analysis. The values obtained using a PLS calibration model compared well against those of the reference method based on acid MAD and ICP-OES measurement, and the analysis of a number of CRMs gave results of acceptable accuracy (85–120%). The data was then combined with complementary values from EDXRF spectrometry to improve the calibration model that enhanced the performance for K. This ‘data fusion’ approach functioned better than the individual data sets, but obviously required significantly more laboratory time when using a total of three analytical techniques. Due to its use as an additive/preservative, a number of countries have regulations on the maximum allowable level of P in seafood products, therefore Tian and co-workers,⁹⁹ investigated the potential of LIBS for P detection. The samples were homogenised and immersed in standard solutions of varying P concentrations. The mixture was then dried, ground, mixed with microcrystalline cellulose then pressed into pellets before LIBS measurements. Using the molybdenum blue spectrophotometric method as the reference, linear regression, PLS and SVM were assessed. It was found that internal standardisation was required and the C 247.86 nm line was selected. A matrix effect was observed for the three seafood types under investigation (codfish, scallop and shrimp), which was an interesting finding given the general similarity of the matrices. Additionally, the intercepts of scallop and shrimp were far from zero, due to the original samples contain high levels of P and potentially affected by self-absorption. Overall, it was found that the SVM approach was most suitable for the seafood samples, achieving the best figures of merit, which is in contrast to other works noted in this section. However this finding does emphasise the severe matrix and analyte dependence of LIBS. Liu et al.¹⁰⁰ applied a dried droplet approach for the analysis of food samples. The CRM NIST SRM 1515 (apple leaves) alongside shop-bought flour, milk powder and pancake powder were acid digested and the solutions spotted onto hydrophobic PTFE filters to minimise any inhomogeneity from the solution dispersion and ringing effect. The digestion solutions were also spiked with Ca, K and Na and spotted to obtain the quantitative data using standard addition calibration. The LIBS method achieved a R² of >0.99 and LOQs of 1.9–48 mg kg⁻¹ for the three elements and matrices. Additionally, the data agreed very well with ICP-MS as the reference method, with a maximum error of 6.3%. It should be noted that there were significant differences between the figures of merit for the different sample types, once again highlighting the sensitivity of LIBS to the matrix.

Another trend from this past year has been the increase in studies utilising LIBS for provenance and authentication investigations. Two papers focussed on honey^101,102 as a food product often subject to fraud and adulteration. Stefas et al.¹⁰¹ simply ablated the honey samples directly, acquired from ten different floral nectar sources (e.g. almond, flower, tree, lavender, etc.). Fifty points across the sample surface were measured, generating a significant quantity of data with similar spectral profiles, therefore a number of statistical methods were applied and compared, namely PCA, LDA, SVM and RFCs. Initial data inspection showed Ca, K, Mg and Na were the most valuable for differentiation but emission lines from C, CN, H, N and O were also included. Furthermore, all approaches achieved accuracies greater than 95%, with LDA and SVM scoring over 99%. Fechner and co-workers¹⁰² also directly analysed honey samples (n = 49) from four provinces of Argentina by spark discharge-assisted LIBS. Similarly, statistical tools were applied to the data, finding that spectral pre-processing using generalised least squares weighting and means centering followed by either k-nearest neighbour or SVM algorithms achieved 100% correct classification. Additionally, the use of PLS-DA highlighted Ca, Cu, Fe, K, Mn and N as key markers for discrimination. Both of these papers clearly demonstrated the potential power of LIBS as a real-time quality control technique for honey authentication. Park et al.¹⁰³ combined measurements from diffuse optical reflectance spectroscopy (DORS) and LIBS in a ‘data fusion’ approach for provenancing of edible salt. The samples (n = 21 from seven countries) were simply ground for DORS analysis and then pressed into pellets for LIBS. Pre-processing by means centering was applied to the DORS spectra whereas for LIBS, normalisation to Na was used for the emission lines of Ca at 393 nm, K at 766 nm, Mg at 383 nm and Sr at 408 nm. Subsequently, both data sets were processed using PCA and the k-nearest neighbour algorithm, finding that the LIBS model provided highest classification accuracy, but DORS was much stronger for certain salt samples. Therefore, the ‘data fusion’ approach was implemented and the optimised weighted model achieved 100% accuracy even for samples that were previously difficult to differentiate by the individual techniques. Perez-Rodriguez et al.¹⁰⁴ described the application of spark discharge LIBS for the origin identification of rice. Four rice genotypes, namely Guri, Irga 424, Puita, and Taim, were selected and 72 samples were collected from the Corrientes province in Argentina. The rice was cryomilled and pressed into tablets for analysis. From the spectra, C, Ca, Fe, Mg, N and Na were selected as suitable emission lines. The workers used central composite design combined with SVM for modelling, achieving prediction accuracy of 96.4%, noting the method was more efficient and faster than other statistical techniques. In a similar vein, Peng-Kai et al.¹⁰⁵ (in Chinese) analysed ginseng from five regions of Northeast China by LIBS and applied machine learning approaches for provenancing. It was found that C, Ca, Fe, H, Mg, N and O were selected as variables, with PCA used to reduce the data dimensions. A back-propagation neural network and SVM algorithm were then compared after optimisation. Both models achieved accuracies greater than 99% but both also misclassified samples in two regions, most likely due to their geographical proximity. This highlights that the technique still requires close supervision and perhaps larger datasets or more advanced statistical techniques. Yao and co-workers¹⁰⁶ investigated an advanced modelling approach for the classification of Chinese tea leaves. The LIBS spectra highlighted Al (396.15 nm), C₂ (516.45 nm), Ca (393.37 nm, 396.84 nm), CN (388.34 nm), Fe (517.46 nm), K (766.49 nm) Mg (279.55 nm, 280.27 nm) and Mn (279.83 nm) as key emission lines that were then referenced to C (247.86 nm) as an IS. The researchers then applied the genetic algorithm technique to identify the penalty factor and kernel parameters for subsequent SVM modelling. Once optimised, the method achieved an average accuracy of 98.4% which was significantly higher than other approaches such as cross validation SVM or particle swarm optimisation SVM.

4.4 Vapour generation procedures and atomic fluorescence spectrometry

As for previous review periods, there is continued interest in the application of VG procedures with not only AFS, but also with AAS, OES and MS. Work that is discussed in detail in this section involves some novelty of the VG method and/or the combination of VG with the mode of detection. If the novelty concerns sample pre-treatment, the paper is discussed in Section 3.2 and also, if pre-concentration by LLE or SPE was involved, in Table 1. In addition to CVG, with either SnCl₂ (specifically for Hg) or BH, a number of reports of PVG have appeared. There is also a report of VG in a liquid anode GD. Several methods of increasing the sensitivity by trapping of the vapour-phase species followed by rapid release have been described including procedures involving headspace SPME, a DBD device, the surface of a tube-in-flame device, and the well-known gold amalgam trapping of Hg.

The considerable interest in the non-chromatographic speciation of As by HG is shown in a review (86 references) by Welna et al.,¹² although not every paper cited involved the analysis of clinical materials or foods (a number were concerned with the environment). Many papers described selective HG as the basis for iAs speciation, but there was also a substantial section describing separation, by LLE or SPE, prior to the HG step. A number of reaction schemes by which various organic As compounds were also determined were covered. The reviewers concluded that compared with HPLC methods, these non-chromatographic procedures are simpler, faster and “economically friendly”. They also pointed out that the combination of HG with AAS or AFS has detection capabilities comparable to those of HPLC-ICP-MS. It is to be noted that one reason for this is that the HPLC sample introduction system dilutes the analytes. They were also confident that further developments will be reported, including alternatives to chemical HG techniques, such as electrochemical generation. It is interesting to note that only one paper cited described PVG. Costa et al.⁴⁷ described the non-chromatographic determination of As^III, but, as the procedure is based on selective SPE, details are given in Table 1. Total iAs was determined after reduction of As^V with hydroxylamine hydrochloride. Digestion with a novel alkaline reagent (see Section 3.2) did not cause species interconversion.

Several reports of speciation analyses by HPLC-VG-AFS have appeared. In the ongoing studies by Hai and co-workers on aspects of arsenic trioxide treatment of patients suffering from acute promyelocytic leukemia, relevant As species were determined by HPLC-HG-AFS. In the latest in a series of papers, the analyses of CSF,¹⁰⁷ plasma¹⁰⁸ and red blood cells¹⁰⁹ were described. Although each paper contains all the details of sample preparation necessary to replicate the work, there is little information about the HPLC-HG-AFS part of the analysis. In the first two papers, readers were referred to two earlier publications from which the procedure was “adapted”. The third paper contained no references to prior analytical work and described an isocratic AEC separation at 30 °C on a Hamilton PRP X-100 column with a mobile phase consisting of 13 mmol L⁻¹ CH₃COONa, 4 mmol L⁻¹ KNO₃, 3 mmol L⁻¹ NaH₂PO₄ and 0.2 mmol L⁻¹ Na₂EDTA at pH 6. However, as this is identical to the procedure described in the first paper, it is concluded that there is nothing new in the HPLC separation. No information about the post-column conversion to hydrides was provided; the analyses were performed with a commercial instrument (LC-AFS 6500, Haiguang Instrument Co., Ltd., Beijing, China). In an interlaboratory comparison, Nakayama and co-workers¹¹⁰ determined MeHg and iHg in cord blood by both HPLC-CV-AFS and HPLC-HG-ICP-MS. For the AFS method the species were separated on a Phenomenex Luna 5U C18(2) column by isocratic elution with ACN + H₂O + 1.5 mmol L⁻¹ APDC. The eluent was merged with an acid carrier (10% HCl + 10% Br₂) and a reductant (2% SnCl₂ in 10% HCl) and passed through a UV digester at 75 °C. In the ICP-MS method the species were separated on a ZORBAX SB-C18 column at 15 °C by a mobile phase 5% (v/v) methanol, 0.1% (v/v) 2-mercaptoethanol, and 0.018% (v/v) HCl. The eluent was merged with 0.08% (w/v) NaBH₄ in 0.06% (w/v) NaOH. The spray chamber was maintained at 2 °C and 20% O₂ in Ar was added. The isotopes monitored were ²⁰²Hg, ¹⁹⁶Hg, and ²⁰⁵Tl, as IS. The LODs were 0.04 and 0.02 ng mL⁻¹ for MeHg and iHg, respectively, by the ICP-MS method, and 0.12 and 0.14 ng mL⁻¹ for MeHg and iHg, respectively, by the AFS method. The methods were validated by the analysis of CRMs Seronorm whole blood (level 2) and Quebec blood (PC-B-M 1201, 1203 and 1601) from the Institute National de Santé Publique du Québec. The validated ICP-MS method was applied to 101 maternal blood and 366 cord blood samples in the pilot study of the Japan Environment and Children’s Study (JECS) from which 50 cord blood samples were randomly selected and analysed by the AFS method. The researchers concluded that there was “relatively good agreement” between the results. They also pointed out that although the throughput of the ICP-MS method (60 samples with calibration standards, blanks, and QC samples per day, or 1200 samples per month) was 50% greater than that of the AFS method, the AFS system was five times less expensive. For the determination of Se species in Se-enriched grain crops (rice, soybean and sweet potato), Wei et al.¹¹¹ compared two chromatographic separation modes followed by UV-HG-AFS. They found that with an AEC method (Hamilton PRP X-100 and phosphate buffer solution as mobile phase) only five Se species (SeCys₂, MeSeCys, Se^IV, SeMet, and Se^VI) could be determined (SeUr was not detected), the sensitivities for organoSe compounds (especially for SeMet) were poorer and the retention time of Se^VI was much longer. Interestingly, they were able to determine SeUr with other anion-exchange systems: Dionex AS 11 with KOH as mobile phase and a ZORBAX SB-Aq column with citric acid as mobile phase. The method selected for comparison was based on an RP C18 column at 35 °C with a mobile phase of 25 mmol L⁻¹ phosphate buffer and 5 mmol L⁻¹ TBAB and 5% (v/v) MeOH at pH 6.0. The method was validated by the analysis of CRM (SELM-1) and by recovery experiments; the determination of total Se was validated by the analysis of GBW10045 (rice flour) and GBW(E) 080684a (rice flour). For speciation analysis the optimised sample preparation involved ultrasonic agitation of 250 mg with 5 mL of H₂O followed by the addition of proteinase K (20 mg) and MAE at 55 °C for 30 min. The cooled solution was centrifuged (10 000 rpm for 10 min), the supernatant was filtered through a 3 kDa filter by further centrifuging (5000 rpm for 15 min). Finally, the solution was filtered (0.22 μm nylon) and refrigerated. Assuming the final volume was 5 mL, the sample preparation represents a 20-fold dilution and thus the solution LODs between 0.8 and 2 μg L⁻¹ represent LODs in the samples between 16 and 40 μg kg⁻¹. The method was applied to three real samples, in none of which were MeSeCys, SeUr, Se^IV or Se^VI found; SeMet was found in all three, and SeCys₂ was found in the rice. Total Se was determined by HG-AFS after digestion with HNO₃ and HClO₄, evaporation, redissolution in hot HCl and addition of ferricyanide, which was not explained but may have something to do with maintaining the Se in the +4 oxidation state (as the +6 state is not BH-active). Typically, hot HCl quantitatively reduces Se^VI to Se^IV.

A number of methods with quantification by AFS have been described. For the determination of total Hg in hair as a biomarker for MeHg exposure in a 23 year-long study of Swedish pregnant women, Kippler et al.¹¹² reported using CV-AFS from 1996 to 1999 and ICP-MS from 2000 onwards. The reason for the change was not given, but the LOD was lower for the AFS method (0.03 mg kg⁻¹) than for the ICP-MS method (0.08 mg kg⁻¹). Sample preparation for the AFS method involved an alkaline digestion in which 20 mg of hair was digested with L-cysteine, NaOH and NaCl at 95 °C for 20 min, whereas the procedure for the ICP-MS method was an HNO₃ MAD of 50 mg for 30 min. In neither case was it clear what the final volume was, so the dilution due to the sample preparations could not be calculated, nor was it obvious that one method was significantly shorter than the other. No significant differences were found between the results for 14 different hair samples analysed by both methods. A further modification to the ICP-MS method (stabilizing the Hg in solution with gold rather than HCl) was evaluated by the analysis of 71 hair samples and again no significant differences were observed. Analytical quality was monitored by the analysis of several CRM for Hg in hair (IAEA 086, NCSZC81002b, GBW 09101) in each run. The researchers concluded that the exposure to MeHg via fish appears to be slowly declining among Swedish pregnant women. For the determination of sodium ethyl(2-mercaptobenzoato-(2-)-O,S) mercurate (thiomersal) in vaccines, the Hg was converted to iHg and measured by CV-AFS.⁵⁰ The main focus of the study was the evaluation of several different oxidative pre-treatments, which is described in more detail in Section 3.2. For the rapid screening of Hg in fruits and vegetables (tomato, lemon and orange) Zhang et al.¹¹³ generated Hg vapour, with a capillary liquid electrode discharge microplasma, that was subsequently swept into a commercial AFS instrument. The only sample preparation was the insertion of a plastic needle into the fruit or vegetable such that a juice drop was formed into which a stainless-steel capillary, already filled with dilute HNO₃, was inserted for 10 s. The inherent capillary action and the reduced pressure induced by the microplasma ensured that the juice was transported along the tube and sprayed into the plasma, which was sustained by an ac power supply connected between the liquid at the end of the capillary and a tungsten electrode. The Hg vapour generated by reduction of Hg ions by electrons in the plasma was swept into the two-channel AFS instrument by the flow of argon in which the plasma was sustained. Calibration was by standard additions, though it was not clear how this was achieved. A large background signal due to scattering from particles was corrected by monitoring the signal in the second channel, which was set for Cd. The LOD was estimated as 0.3–0.5 μg L⁻¹ and the method was validated by spike recoveries and comparison of the results with those obtained after acid MAD and conventional AFS analysis. Mercury was found in most samples at single-digit μg kg⁻¹ concentrations, though it was not clear how concentrations determined directly in the juice were converted to values for the entire fruit or vegetable. The researchers presented preliminary result to show the possibility of coupling the capillary liquid electrode discharge microplasma system with OES. The gas-phase pre-concentration in a DBD trap developed by Mao and co-workers, described in last year’s ASU¹ for the determination of As in blood by AFS, has been applied for the determination of Pb in blood¹¹⁴ and Sb in bottled water.¹¹⁵ The basis of the method is that following sample preparation, the analyte hydride was generated in a batch reactor by the usual BH-based chemistry and the evolved hydride decomposed on the quartz surface of the DBD trap so that all the element in the original sample volume introduced was collected. Thus, other vapour-phase constituents that are not trapped and which might cause interferences in the AFS measurement are removed. This also meant that dispersion processes prior to trapping had no effect on the signal shape, which was governed by the kinetics of release from the quartz surface when the composition of the gas was changed from one containing O₂ (Ar + air) to one containing H₂ (Ar + H₂). In this way, a second GLS device, introduced to eliminate the effects of foaming encountered in the case of Pb in blood, had no effect on sensitivity or LOD, which was 8 pg for a 2 mL sample volume (original blood sample was 0.05 mL). The procedure was validated by the analysis of GBW09139 (freeze-dried bovine blood), spike addition and comparison of the results with those of an acid MAD ICP-MS method. The researchers indicated that the method could determine Pb down to 0.16 μg L⁻¹ in blood samples in 10 min, including sample preparation and dilution. For the determination of Sb in bottled water,¹¹⁵ the DBT trap removed the depressive effect of water vapour, and there was thus no need for any Nafion drying tubes or columns of desiccants. For a 2 mL sample volume, the LOD was 9 pg, well below that needed to detect Sb in real samples from PET bottles and in the CRMs used for validation: NRCCRM BWZ6657-2016 (80 ± 4 ng g⁻¹ Sb in 1% HCl) and GSB 07-1376-2001 (29.8 ± 1.5 mg L⁻¹), from the Second Institute of Oceanography, State Oceanic Administration, Hangzhou, China. Spikes (5 μg L⁻¹) were also quantitatively recovered. The procedure was applied to an Sb leaching study, which showed that for PET bottles Sb was detectable (at low single-digit μg L⁻¹ concentrations) even after only 5 h at 25 °C.

A significant fraction of the work published in the current review period described procedures concerned VG with other atomic spectrometry techniques. Perhaps the most extensively studied was the determination of Sb isotope ratios by HG-ICP-MC-MS,⁷⁰ in which samples were purified by SPE on thiol-functionalized mesoporous silica powder (see Table 1). This step removed a number of elements that otherwise would interfere in the subsequent HG reaction with BH and contributed to the average LOD of 0.008 μg L⁻¹. The method was validated by the analysis of no fewer than 14 CRMs, many of which were environmental, though plants and human blood and urine were included. Accurate results were obtained and there was no evidence of isotope fractionation. When some pure Sb standards were examined, not all were found to be isotopically similar and the researchers pointed out the importance of choosing a common isotopic standard solution against which to compare already published and future Sb isotope data. Thus the community was urged to make efforts to inter-calibrate different in-house isotopic standards and, if possible, use the same standard solution going forward. Peng et al.⁸⁹ developed a HG-GD-OES (monitoring at 368.3 nm) method for the determination of Pb with an LOD of 0.06 μg L⁻¹ in which the hydride was introduced via the tubular cathode into a solution anode GD microplasma. After a full optimisation, the LOD was 0.06 μg L⁻¹. The method was validated by the analysis of CRMs GBW07311 (stream sediment), GBW07312 (stream sediment), and GBW07601a (GSH-1) (human hair), and by a comparison of the results with those obtained by ICP-MS. The method was also applied to whole human blood for which the sample digestion procedure (with conc. HNO₃) involved dilution of 0.2 g to a final volume of 20 mL. The two samples analysed contained about 50 and 100 ng g⁻¹ Pb. A method has been developed¹¹⁶ for the determination of bioaccessible Hg in which mercury species were converted to iHg by PVG and pre-concentrated by headspace SPME prior to desorption and measurement by MIP-OES. Formic acid was added to all solutions to give a final concentration of 5% v/v and volatile Hg vapour was generated by reactions induced by an 11 W UV-C lamp on passage of 10 mL through a quartz tube (3 mm id and 68 cm long) connected to a GLS containing a SPME fibre (50/30 μm DVB/CAR/PDMS-coated) in the headspace. After collection, the Hg was thermally desorbed in a laboratory-made device and transported to the 100 W MIP sustained in argon in a quartz tube in a TM₀₁₀ Beenakker-type water-cooled cavity, that was also part of a laboratory-made spectrometer. No citations to previous use of this spectrometer were given, and only minimal description was provided. The method was validated by the analysis of CRM ERM-CC580 (estuarine sediment) for total Hg content and of ERM-CE464 (tuna fish) for the MeHg content. The LODs were between 0.03 and 0.04 μg L⁻¹ for both species and the method was applied for monitoring of the bioaccessible fraction of mercury during incubation in simulated body fluid (Ringer’s solution) in the presence of selenium NPs as a potential Hg detoxifying agent. The researchers commented that high total Hg content in ERM-CC580 (132 mg kg⁻¹) allowed for substantial dilution (up to 2500-times) of the sample extract leading to the minimization of matrix effects, and that in the case of the tuna fish CRM, fatty acids, which are known to suppress photo-reduction of Hg species, were removed by LLE. Frois et al.¹¹⁷ compared CVG and PVG for the determination of Hg in fish by AAS. For each generation method, the Hg vapour was measured by a commercial dedicated Hg analyser (FIMS 400). The researchers also compared the effectiveness of optimisation methods and obtained better performance figures of merit with multivariate strategies than with a single-cycle alternating variable search method. They selected BH for the CV procedure and did not investigate the performance with SnCl₂, but selected HCOOH for the PVG procedure after investigating the performance with propanol and CH₃COOH. They validated both methods by the accurate analysis of one CRM, NRCC DOLT-4 (dogfish liver) and two Brazilian fish RMs. The researchers indicated that the solution LODs were 25 and 11 ng L⁻¹ for the CVG and PVG procedures respectively, but then miscalculated the LOD in the sample for the PVG procedure for which 150 mg of sample was diluted to 100 mL, whereas for the CVG method the final sample volume was 15 mL. The researchers also favoured PVG on the basis of cost (HCOOH is cheaper and more stable than BH) and claimed that the throughput of both techniques were “similar”, though, in fact, the sample preparation time and the analytical frequency of the PVG method were 25% and 67% worse, respectively than those for the CVG procedure. To obtain a similar LOD when using a conventional AA spectrometer, with a quartz tube atom cell mounted above the burner and an EDL light source, Menezes¹¹⁸ devised a DLLME procedure in which the Hg in a 5 mL sample was transferred into 60 μL of the IL extractant (see Table 1). The viscosity was decreased by the addition of 50 μL 25% (m/v) HNO₃ and the mixture transferred to a miniature batch reaction vessel (a 4 mL vial with cap and septum). Tin(II) solution was pumped in and the solution mixed by magnetic stirring, the evolved Hg was swept into the atom cell by a stream of argon. After optimising all the relevant parameters, the LOD in the initial aqueous solution was 45 ng L⁻¹. The method was validated by the analysis of a CRM. The abstract indicated that this was ERM-CE278 (mussel tissue), but the text indicated IRMM BCR-060 (Lagarosiphon major) an aquatic plant, and the procedure applied to the analysis of fish oil samples, for which the sample preparation procedure transferred all the Hg in 100 mg of sample to 25 mL, following digestion with HNO₃ and H₂O₂ (at 110 °C for 6 h) and pH adjustment with NaOH. In all five samples, the Hg concentration was below the LOQ of 38 μg kg⁻¹. A PVG procedure was developed¹¹⁹ for the determination of Cd by FAAS. In this case, samples containing 8 mol L⁻¹ propionic acid were pumped through a quartz tubing reactor and irradiated by UV-C radiation. On emerging from the reactor, the fluid stream was delivered via a concentric nebuliser (driven by argon) into a GLS and the separated gases transported via a pressed-cotton trap to the interior of a slotted quartz tube mounted in a fuel–lean air–C₂H₂ flame. After collection of the Cd on the interior wall of the tube, the argon stream was switched to H₂, when the Cd was rapidly released into the light path of the spectrometer giving a sharp, transient signal. The LOD of the optimised procedure was 2 μg L⁻¹. The method was applied to the analysis of tap water and a daphne tea extract (hot water), but only the results of spike recoveries were given, so it is not clear whether Cd was detected in the real samples. The researchers reported that calibration with matrix-matched standards was needed for the analysis of the tea, but did not describe what was entailed. Two reports^57,58 of the determination of Hg with a commercial AAS instrument in which Hg, released from the sample by combustion in a stream of oxygen, was collected on a gold amalgam trap and then thermally desorbed and measured are discussed in more detail in Section 3.2.

4.5 X-ray spectrometry

A comprehensive review of recent advances in X-ray spectrometry³ complements the applications with clinical and biological materials, foods and beverages covered within this Update. Imaging applications of X-ray spectrometry are described separately in Section 6.2 below.

Significant efforts have been made to improve quantitative analysis by EDXRF spectrometry within this review period. Machado et al.¹²⁰ described a novel approach using external calibration based on CRMs and a matrix correction factor. In total, fifteen animal tissue and plant/leaves CRMs were used to build calibration curves for Ca, Cu, Fe, K, Mn, P, S and Zn. The Compton-to-Rayleigh ratio was used for normalisation of the matrix effects. The researchers compared the effect on the calibration from using only the animal CRMs, only the plant/leaves CRMs and the combination of the two. The use of both groups provided the lowest error and improved the agreement with the certified values of NIST SRM 1577a (bovine liver). Most elements were within 13% deviation from expected values, with the exception of Ca where the animal tissue model had a deviation of 81% and the combined model had a 48% deviation – whilst it certainly improved the result, the authors did not further discuss this large deviation for Ca. The researchers then went on to apply the combined calibration to the analysis of human cancer tissues (endometrium and breast carcinoma) from five patients, with matched pairs of normal tissue. The results for some elements showed significant differences between healthy and tumour tissue, as determined by Wilcoxon Sign Rank test, although some elements experienced significant sample heterogeneity leading to extremely large error bars for Ca and Fe in particular. A single sample was also analysed by ICP-OES for K, P and Zn only, however, no error bars or statistical analysis were presented for the ICP-OES data, and with the log-scale of the graphic, it was difficult to conclude that agreement was observed. The paper presented a new XRF quantitation strategy but the validation data and application to tissues perhaps fell short. Carvalho et al.¹²¹ discussed the challenges of quantifying biological materials by EDXRF spectrometry due to the high levels of low Z elements (i.e. C, H, N, O). The researchers investigated this by comparing different calibration approaches for the analysis of Cu, Fe and Zn. The first method utilised fundamental parameters and five others were based on building the calibration from matrix CRMs and incorporating the Rayleigh and Compton scattering peaks for correction of matrix effects. The optimal results were obtained using the CRMs and correction with the Rayleigh/Compton ratio, providing results for two independent CRMs within 10% of the certified values. This was then applied to four human tissue samples, comprising of matched pairs of normal and tumour tissue (colon, lung, ovary and prostate). Here, the samples were lyophilised then ground into powder and pressed into pellets. The results showed some trends for Cu, Fe and Zn between the matched pairs of tissues. Following a similar approach, Ghidotti et al.¹²² focussed on honey authentication, using EDXRF spectrometry as a fast screening approach to support the ‘Protected Denomination of Origin’ (PDO) status which is frequently subject to fraud. The analytical method was validated in a previous publication but used forty seven matrix based RMs and CRMs to build the calibration curves for Ca, Cl, Cu, Fe, K, Mn, Ni, P, Rb and Zn, then a further twenty one materials for accuracy determination. Honey samples were obtained from three PDO regions in Spain (n = 183), a Spanish region without PDO (n = 18) and a further group of samples from outside these regions (n = 131), covering different botanical varieties from multiple countries. The honey samples were analysed directly without any preparation. Fit for purpose figures of merit were obtained. The data was then subjected to statistical classification analysis using SIMCA and PLS-DA, finding that PLS-DA was optimal for this data set. Generally, classification was accurate at the larger scale (e.g. PDO and non-PDO regions) but was poorer at distinguishing some finer features such as the difference between light and dark honeys, which may also be due to lower sample numbers. However, several trends were observed leading to good differentiation overall. The work demonstrated the potential use of EDXRF for honey authentication. Almeida et al.¹²³ described the direct analysis of ground coffee by EDXRF spectrometry. Quantitative analysis for Cd, Fe, Mn and Pb was achieved through external calibration using the concentrations from the samples obtained from MAD and ICP-OES analysis compared to the intensities of pelletised ground coffee by XRF spectrometry. Twenty samples were used for the calibration, obtaining R² > 0.99 and achieving LODs between 55–86 μg kg⁻¹, LOQs 165–258 μg kg⁻¹ and % RSD was 6–14% (n = 10), indicating good performance by direct analysis. Subsequently, five additional samples were quantified, obtaining values in good agreement with those from ICP-OES. The work demonstrated the potential for the direct analysis of coffee, but the validation would have benefitted by the inclusion of RMs or spiked samples. Gamela and co-workers⁹⁸ also utilised EDXRF for cocoa beans in combination with LIBS and ICP-OES – the paper is further discussed in Section 4.3.

The analysis of elemental impurities in pharmaceutical samples was investigated by Chowdhury et al.,¹²⁴ employing EDXRF to meet the ICH Q3D requirements (International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use) under GMP. Section Q3D splits the elemental impurities into categories, with this work focusing on class 1 and class 2A impurities, namely As, Cd, Co, Hg, Ni, Pb and V. The researchers selected a drug with a high maximum daily dose of 540 mg per day, therefore the maximum allowable limits for the impurities were in the ppm range as they are related to the dose. The drug samples were simply crushed and quantified using liquid calibration standards. The validation study included assessment of the linearity, LOD/LOQ, precision and accuracy (by spiking at three concentration levels). The results demonstrated that EDXRF spectrometry fulfilled the requirements for the ICH rules, which is a remarkable achievement. However, only one drug product was tested with a high dosage therefore the approach may struggle to meet the detection requirements in other cases.

The use of TXRF spectrometry has received significant attention and expansion to a number of interesting applications during this review period. Margui and co-workers¹⁴ presented a simple approach for multi-element analysis of biological samples by TXRF spectrometry. Dilution was applied to liquid samples and solid samples were suspended in either water or 1% Triton X-100. A number of CRMs (including plant materials, food and fish tissue) were measured alongside biological quality control materials for serum, plasma and whole blood. Additionally, seminal plasma samples that had been previously analysed by ICP-OES were included. The samples were spotted (5–20 μL) onto the quartz reflector and dried. The LODs were variable but generally in the range of 1–10 mg kg⁻¹ for Br, Cu, Fe, Mn, Ni, Rb, Sr and Zn. However, the limits were significantly higher for Ca and K (81–84 mg kg⁻¹ and 130–170 mg kg⁻¹ respectively), though it was considered that these elements are typically present in biological and plant materials at significantly higher levels therefore the LOD is not necessarily restrictive. The results from analysis of the CRMs and additional controls were generally within 90–110% of the expected values, but there were larger deviations at values close to the LOQ. Moreover, the researchers compared the impact of internal standardisation with Y against external calibration using the matrix CRMs. This improved the accuracies for low Z elements such as K significantly (75% vs. 98%), while mid and high Z elements displayed little difference. Overall, the work demonstrated the potential capability of TXRF spectrometry in a variety of matrices with relatively simple sample preparation. Following on from this, Margui et al.¹²⁵ reported the use of TXRF spectrometry for elemental analysis of seminal plasma (Br, Ca, Cd, Co, Cr, Cu, Fe, K, Mn, Ni, Pb, Se, Rb and Zn). Based on the previous work, the samples were simply diluted in 1% Triton X-100, then an aliquot of 5 μL was spotted onto the reflector and dried. The instrumental parameters were optimised and validated against spiked samples and a serum QC material. However, a number of elements were below the LOD for the serum material, relying on the recovery of the spiked sample at 0.9 mg kg⁻¹, which is significantly higher than the levels expected in seminal plasma. It was perhaps unsurprising that when the approach was applied to human samples (n = 21), only Zn could be reliably detected above the LOQ. The measurement of trace elements in saliva and dental caries from preschool children was another unusual matrix utilising TXRF spectrometry.¹²⁶ Samples were collected from children aged 36–72 months and were split into two groups: caries (n = 60) and non-caries (n = 60). The levels of Al, Cu, Fe, Mn and Zn were obtained from only 10 μL of saliva. However, little detail regarding the calibration or quality control measures were provided in the paper, other than the use of Ga as an internal standard. Following statistical analysis, Fe and Mn were significantly higher in the caries group compared to the non-caries, indicating a potential relationship. Whilst the accuracy was not verified, the work does demonstrate the ease of sample preparation and speed of analysis.

Two papers of note covered the application of TXRF spectrometry for food and dietary supplements. Beltran and co-workers¹²⁷ described the method validation for As, Cr, Hg and Pb in dietary supplements. The samples were digested using a mixture of HNO₃, HCl and H₂O₂, heated in a water bath at 80 °C for 8 h. The instrumental parameters were optimised by a central composite design, showing that two methods were needed as Hg required different conditions than As, Cr and Pb. Good figures of merit were obtained: LOQs were 1.35, 1.90, 2.50 and 1.95 μg L⁻¹ for As, Cr, Hg and Pb respectively; the recoveries were in the range 91–108% and good comparability was achieved for two samples also analysed by ICP-OES. The concentrations found in commercial supplements were in the range of <LOQ to 13.62 μg g⁻¹, demonstrating the suitability of the approach. Maltsev et al.⁴⁸ examined the impact of sample preparation technique on the measurement capability of TXRF spectrometry. Tea leaves were prepared by four methods, namely suspension in water, infusion in hot water, open vessel digestion and closed vessel microwave digestion. The influence of particle size and absorption effects were investigated. The method was validated with CRMs and complementary analysis by WDXRF, concluding that Ba, Br, Ca, Cu, Fe, K, Mn, Ni, P, Pb, S, Sr, Rb and Zn were suitable for TXRF measurement. Nineteen tea samples were then successfully analysed.

As is now tradition, this section of the review would not be complete without covering the in vivo detection of Pb in bone. Zhang and co-workers¹²⁸ compared a portable XRF system with KXRF. The study analysed the mid-tibia bone in seventy-one participants from three communities in Indiana, USA, that included ten occupationally exposed individuals. It was shown that the soft tissue thickness played a significant role in the accuracy of the portable instrument especially when the tissue thickness was >6 mm. The correlation between the two systems was r = 0.48 for all samples, but increased to r = 0.73 if the tissue thickness was <6 mm. It was also shown that increasing the measurement time for the portable system from 3 min to 5 improved the uncertainty by a factor of 1.4. Overall, the study demonstrated the potential of the system for large scale population studies though further consideration for accuracy and precision is still required, especially at lower Pb concentrations. A paper by Gherase et al.¹²⁹ described developments for the calibration of L-shell KXRF spectrometry for the determination of in vivo Pb in bone. In this work, the effect of soft tissue thickness was addressed with bone phantoms doped with various levels of Pb and a fixed amount of Sr and coated in polyoxymethylene, resin and wax to mimic the soft tissue at four different thicknesses between 1–4 mm. By utilising the Sr K-β/K-α ratio, Pb concentrations and Pb L-shell response, it was possible to derive accurate calibration without knowledge of the soft tissue thickness. The approach achieved an LOD of 20 μg g⁻¹. Portable XRF spectrometry for the analysis of Zn in human toenail clippings was the focus of a study by Fleming and co-workers,¹³⁰ as a non-invasive biomarker. The efficacy of the instrument’s proprietary software calibration was also assessed against the PyMca, software, developed by the European Synchrotron Radiation Facility for the fitting of X-ray spectra. The Zn concentrations were initially determined using ICP-MS and ranged between 32 to 140 μg g⁻¹, with a population (n = 60) average of 85 μg g⁻¹. The correlation between these data and those from the portable XRF instrument, using the in-built calibration model, was r = 0.60, but increased to r = 0.68 using the PyMca software. The portable technique was sensitive enough to detect Zn in nails reliably, but the agreement in absolute concentration still required further development, which the authors acknowledge. It will be interesting to see how this develops in the coming years.

In a study by Nelson and co-workers,¹³¹ a number of techniques were applied to investigate the distribution and speciation of micro- and nanoparticles in soft tissue adjacent to titanium and ceramic dental implants. The combination of SR-μXRF, SR-nanoXRF and XANES enabled identification of Ti particles from several μm to 100 nm in size, exhibiting two predominate states, specifically metallic Ti and TiO₂. Additionally, ceramic particles were observed in five of the eight samples. Both particle types were found at around 40 million particles per mm³. The use of complementary techniques provided a unique insight into the potential causes of inflammation and potentially the pathogenesis of peri-implantitis.

The use of X-ray based techniques for As speciation in food and beverages was featured in two studies. Zhang et al.¹³² reported the novel application of a benchtop/portable EDXRF system for As speciation in beverages. The Gutzeit method was combined with XRF spectrometry for rapid analysis at relatively low cost compared to standard lab-based approaches such as HPLC-ICP-MS. Here, iAs is reduced to arsine gas that is trapped with test discs impregnated with HgBr. These discs were then measured with the benchtop XRF instrument. The method was linear up to 133.3 μg L⁻¹ and achieved an LOD of 1.9 μg L⁻¹ and LOQ of 5.7 μg L⁻¹ by external calibration. Additionally, spike recoveries were in the range of 87–95% for spike amounts between 13–133 μg L⁻¹ but this dropped to 69% at the 6.7 μg L⁻¹ level, although this was close to the LOQ. The presence of DMA was also tested, finding that the DMA was only physically adsorbed to the discs rather than chemically bound and could be easily removed with an acetone wash step without impacting the iAs. The approach was tested using drinking water, apple juice and red wine, with the results comparable to those obtained by ICP-MS. The study demonstrated the potential of the approach as a rapid screening technique. Cho et al.¹³³ described the direct speciation analysis of As in various CRMs using XANES, aiming to enable As species detection in solid samples with minimal sample preparation, therefore reducing the risk of any species transformation. The researchers analysed 13 As compounds (organic and inorganic) to build the spectral library. Then, the method was applied to a number of seafood, cosmetic and environmental CRMs and candidate RMs. A variety of As species were identified across the different matrices and these results were compared, achieving similar profiles, to those obtained with IC-ICP-MS after extraction with H₃PO₄–NH₂OH·HCl. The paper demonstrated the potential of XANES for species screening across several matrices.

5 Nanomaterials

Most of the publications during this review year featured spICP-MS or scICP-MS to either characterise (size, numbers) nanoparticles in various sample types, or apply these approaches to assess uptake and effects in cells or tissues. Some included quantification rather than simply demonstrating the presence of particles in the analysed samples. The use of NPs for sample preparation/pre-concentration was mentioned in Section 3.2.

Within the setting of a European project (physicochemical characterization of nanoparticles in food additives in the context of risk identification; see EFSA Journal 2021:18 (6):EN-6678 and EFSA Supporting publication 2021:EN-6678, for details), development of methods to analyse additives in their ‘pristine’ state and in foods was tasked to a ‘Nanofood@’ project. From this work De Vos et al.¹³⁴ reported the characterisation of particles in the additive E174 (silver) and E174-containing confectionary. Using TEM and spICP-MS, it was shown that samples contain both NPs and flakes. Although the total number of flakes were fewer, these represented the largest mass of Ag. The sizes of NPs and flakes tended to be slightly larger in food products compared to the actual additive. Grasso et al.¹³⁵ used spICP-MS to determine the number and size of AgNPs in canned tuna, mackerel, anchovy and clam in order to assess discharge of NPs into the aquatic environment and the implications for human health. Procedures for extraction of NPs and digestion for total Ag measurement were validated and values determined for NP size, diameter, number, total Ag and Ag in NPs. Results followed a pattern of greatest in tuna followed by mackerel, clam and anchovy. Using estimated meal intakes, exposure would be well below the daily reference dose for Ag (5 μg kg⁻¹ b.w.). These workers also determined TiO₂NPs in the same samples.¹³⁶ Particle sizes and Ti concentrations were significantly greater in anchovy and clam, compared to tuna and mackerel, with the largest number of NPs found in clam. These results produced evidence of exposure to TiO₂NPs also in foods where they had not been intentionally added. Using AF4-MALS-ICP-MS, Li et al.¹³⁷ were able to characterise E171 (TiO₂NPs) in coffee creamer and instant fruit drink powders. Details of the methodology and calibration were explained. Particle size ranges were from approximately 25 to 550 nm in all sample types, consistent with previous reports. The authors pointed out that while A4-MALS gave structural information for the NPs, the ICP-MS results were necessary to confirm that they contained Ti.

Beside to the application of investigating NPs levels in food and beverages, characterisation of NPs in biological fluids is seen to be important, as there may be risks to human health associated with exposure. Methodologies to validate the particle sizes of AgNPs and AuNPs using spICP-MS and FFF-MALS-UV-ICP-MS, were developed using NP reference materials spiked into samples of blood, serum and urine.¹³⁸ The results obtained for linearity, LOD, resolution, repeatability, recovery and stability demonstrated that these methods are suitable to monitor metallic NPs in exposed and non-exposed populations. Similar experiments, using spICP-MS were reported for the characterisation of AgNPs in blood,¹³⁹ CeO₂NPs in urine, plasma and enzyme digested liver⁷⁵ and PtNPs in serum and urine.¹⁴⁰

Further to establishing acceptable analytical methodologies, there are reports of cellular and tissue uptake of NPs and possible metabolic effects. In a novel piece of work, Chen et al.⁶⁴ monitored the uptake of FePt-CysNPs into single cells, and the subsequent release of the Fe and Pt over the following 18 h. The technique involved a droplet-splitting microchip coupled online to an ICP-MS detector. The results indicated that while the released Fe had little effect on the endogenous cell content, there may be sufficient Pt to be cytotoxic. There is evidence to suggest that HgSe particles are formed during Hg detoxification. These ‘natural’ NPs are found in the brain tissue of various species, including humans. To investigate the metabolism of these particles, Cid-Barrio et al.¹⁴¹ determined the uptake and accumulation of HgSeNPs in cultured embryonic and human-tumour cell lines. Particles were first synthesised and characterised for use in the project. Uptake was assessed by cell viability, fluorescence of the NPs and measurement of Hg by ICP-MS. The results confirmed uptake of the HgSeNPs did occur and that there was no effect on cell viability. Both scICP-MS and spICP-MS, together with TEM, were used in a project reported by Gomez-Gomez et al.⁶⁵ Internalisation of TeNPs into bacteria (E. coli and S. Aureus) was demonstrated by TEM and, with scICP-TQ-MS, the mass of Te per cell was measured. Following cell lysis, the number, size and shape of particles were determined. Since exposure to TiO₂NPs in pregnant rodents has been shown to alter placental functions and that there is maternal–foetal transfer of NPs, with toxic effects on the foetus, Guillard et al. proposed an investigation with human subjects.¹⁴² Results from ICP-MS analysis of normal term placenta showed measurable concentrations of Ti. Half of the samples of meconium (an infant’s first stool sample) also contained measurable Ti (>0.01 mg kg⁻¹) using STEM-EDX, as isolated and clusters of TiO₂ particles. In further studies where a human placental model was perfused with food-grade TiO₂NPs, Ti was not detected in the perfusate although STEM-EDX showed NPs to be present. The amounts and concentrations of NPs were very low, but the authors suggested there is a need for a risk analysis for chronic exposure to TiO₂NPs during pregnancy.

A further series of publications focus on quantification of NPs in biological samples and foods.

The anti-bacterial properties of silver-containing creams applied to burns and other skin wounds, have been recognised for many years. However, investigating the processes of absorption and mode of action has proved challenging. In a report from laboratories in Italy, France and Scotland, Roman and colleagues¹⁴³ analysed specimens of skin from patients with mid-to-deep thickness burns or equivalent skin wounds, treated with dressings containing AgNPs or silver sulfadiazine. The spatiotemporal distribution and speciation of Ag was investigated by LA-ICP-MS and SR-μXRF/μXANES. With these techniques it was shown that Ag is rapidly released onto the wound surface, followed by a significant structure-dependent penetration into the damaged tissues. This is accompanied by sequential processes of metallic Ag dissolution, Cl⁻ complexation, change to metal-thiol protein complexes, and final mobilisation into deeper skin layers towards the vascular networks. Complete local clearance of Ag was observed after 12 days of treatment in the case of full healing. The results provide an insight into the dynamics of Ag in real human wounds.

As AuNPs have excellent properties as diagnostic and therapeutic agents, Sun et al.⁶² considered it important to investigate, in vivo, the metabolism in blood samples on a single particle basis. With a purpose built autosampler, having a flow rate in the range of 5–5000 μL min⁻¹ and a customized cyclonic spray chamber, the LODs for the NP size and number concentration were found to be 19 nm and 8 × 10⁴ particle per L, respectively. The system was applied to track the changes in size and concentration for AuNPs in very small samples of blood collected from a mouse. Following exposure, the concentration quickly reached a peak and then gradually decreased. The results demonstrated the potential for analysing and monitoring the size and concentration of NPs in ultralow-volume blood samples with low concentrations, to show the fate of NPs in vivo. Grønbæk-Thorsen et al.¹⁴⁴ also investigated the behaviour of NPs in blood. These authors used CE-ICP-MS to study SeNPs stabilised by a coating of polyvinyl alcohol (PVA). A fused silica capillary column was used to separate species produced when the SeNPs were incubated in human plasma. The electrophoretic and ICP-MS conditions were optimised and spiked samples were analysed at 5 h, 48 h and 5 days after incubation at ambient temperature. Peaks obtained were compared to those produced by PVA-SeNP, TMSe, SeGalac and Se^III. The electopherograms showed a peak for PVA-SeNP at 5 h with an additional small peak identified as Se^III appearing at 48 h. The Se^III peak was larger after incubation for 5 days but corresponded to less than 5% of the total Se. The results demonstrate what may prove to be an effective approach to investigations of SeNPs in biological samples.

Noting their widespread use in items such as cosmetics, food additives and medicines, Salou et al.¹⁴⁵ developed and validated a procedure to measure concentrations of TiO₂NPs in human urine by scICP-MS. Using the parameters set out in ISO/IEC 17025:2017, general requirements for the competence of testing and calibration laboratories, LODs for concentration and particle size, linearity, repeatability, reproducibility and recovery were determined using the developed method. Sonicated urine samples were diluted in 0.1% HNO₃ and spiked with TiO₂NPs (SRM 1898). These samples were then used to optimise the scICP-MS instrumental operating conditions and for the validation experiments. The concentration and particle size LODs were 0.05 ng mL⁻¹ and 50 nm, respectively. Repeatability and reproducibility were 14% and 18% respectively for concentration while for particle size, the values for both parameters were 6%. The assay was linear from 0.16 (the LOQ) to 2.6 ng mL⁻¹. Accuracy, derived from analysis of the SRM 1898 spiked materials was 98%. The authors suggested that the method is suitable to be used for human risk assessment after exposure to NPs. In a similar investigation to that of Salou et al., He and colleagues¹⁴⁶ also used spICP-MS, to determine Ag, Au, CeO₂ and ZnO NPs in simulated gastric fluid. The developed method had a concentration LOD of 135 particles per mL, metal LODs from 0.02 to 0.1 μg mL⁻¹ and particle size LODs between 15 and 35 nm. It was noted that while ZnONPs rapidly dissolved, AuNPs and CeO₂NPs formed aggregates with little dissolution. Both aggregation and dissolution were observed with AgNPs. The original NP concentration and NP size, gastric fluid contact time and temperature all influenced the final results, indicating that evaluation of the biological consequences of exposure to these NPs is a complex challenge.

Two publications presented results that feature NP quantification in foods. Taboada-Lopez et al.¹⁴⁷ used A4F UV spectroscopy to detect, and ICP-MS to quantify, AgNPs in clams, oysters and scallops after enzymatic hydrolysis. It was found that the AgNPs were associated with proteins, with the fractograms showing binding to different peaks in the different sample types. Measurement of total Ag by ICP-MS in acid digests of the extracts indicated mean concentrations of 3700 μg kg⁻¹ in frozen scallop, 990 μg kg⁻¹ in oyster and much lower amounts in clam and fresh scallop. It was concluded that the approach adopted allows assessing low levels of NPs. Similar results were obtained when the concentrations of Ti in TiO₂NPs were measured using ICP-QQQ-MS and HR-ICP-MS in other food types, chewing gum, chocolate candy and cake decorations.⁶⁸ The emphasis in both of these reports was to demonstrate the potential of the techniques used to extract NPs, to confirm the structures of the extracted NPs and for their final quantification.

6 Applications: clinical and biological materials

In this section, developments in various areas of work involving clinical and biological materials are discussed. In addition technical details of selected procedures are summarised in Table 2.

Table 2 Clinical and biological materials

Element	Matrix	Technique	Sample treatment/comments	Reference
Ag, Au (NPs)	Urine, blood and serum	spICP-MS, AF4-FFF-ICP-MS	spICP-MS: ^107Ag and ^197Au monitored with dwell time, 5 ms, and analysis time, 60 s. Size LODs were 7 to 14 nm. AF4-FFF-ICP-MS: separation of AgNPs and AuNPs with 0.1 M NaOH (pH 9.3) and 0.02% NaN3 respectively in 25 μL sample using a short channel flow cell containing a 10 kDa regenerated cellulose membrane and a 350 μm height spacer. Dwell time for ICP-MS detection was 100 ms. Size LODs were 2 to 5 nm	138
As species (As^III, As^V, DMA^V, MMA^V)	RBCs, plasma	HPLC-HG-AFS	Cell lysis of RBCs (100 mg) with mobile phase (13 mM CH3COONa + 4 mM KNO3 + 3 mM NaH2PO4 + 0.2 mM EDTA-2Na (pH = 6)) and 5% NH3. Haemoglobin-bound As released by 4 h incubation at room temperature with 30% H2O2 followed by centrifugation. 20% HClO4 added to lysed RBC supernatant or plasma (360 μL) to precipitate proteins. Supernatant filtered then species separated by isocratic elution at 1.0 mL min^−1 on an anion-exchange column	109
As	Urine	HR-CS-GFAAS	Direct analysis of urine using two-stage probe atomisation with a U-shaped W-probe. Urine (15 μL) with Pd-modifier (20 μg) dried using smooth heating up to 120 °C. Atomisation temperatures were 2300 °C. The LOD was 1.5 μg L^−1 and quantification range, 50 to 1000 μg L^−1	84
Br, Cl, I	Blood	ICP-MS	Samples prepared by either MIC, where blood was spotted onto square pieces of filter paper with absorbing solution, 50 mM NH4OH, or by MAD-UV in diluted HNO3 + H2O2	54
Ca, Mg isotope ratios	Blood, plasma, serum, urine and mouse tissues	MC-ICP-MS	Samples double digested with HNO3/H2O2. Sequential chromatographic separation: Mg isolated using AG50W-X8 strong cation exchange resin; Ca fraction re-run on same column to remove Fe before passing through an Sr-resin. Organic compounds removed by heating and fractions diluted to 4 mg L^−1 Ca and 150 mg L^−1 Mg for isotope determination. Ca isotope ratio measurements performed in cold plasma conditions at medium mass resolution in static collection mode	72
Cd, Pb	Blood	ETV-APGD-AES	Blood (0.5 g) diluted with 0.2% Triton X-100 and 10 μL dripped on to W-coil. Optimised ETV parameters: plasma gas, He/3% H2; drying current, 3.8 A; pyrolysis current, 3.8 A (Cd) and 4.6 A (Pb); vaporisation current, 8.5 A. Optimised APGD parameters: carrier gas flow rate, 450 mL min^−1; discharge gap, 12 mm; discharge current, 22 mA (Cd) and 24 mA (Pb). Emission lines 228.8 nm (Cd) and 368.3 nm (Pb) were selected. The LODs for Cd and Pb were 0.4 and 1.2 μg L^−1 respectively	90
Cd and Cd-metallothioneins	HepG2 and MCF-7 cells	Short column CE-ICP-MS	Separation via a 30 cm long polyamide-coated fused silica capillary. Running buffer (Tris–HNO3) incorporated 18 mM SDS. Isotope ^111Cd monitored in TRA mode. The LODs were 0.013 mg L^−1 for Cd-metallothioneins and 0.020 mg L^−1 for Cd^2+	148
Cd, Zn	HepG2 cells	scICP-MS	A 3D Ar enclosed droplet based microfluidic device. Optimised flow rates for single cell generation were 30 mL min^−1 for Ar gas and 0.5 μL min^−1 for water phase. Dwell time optimised at 5 ms	63
Cu and Cu isotope ratios	Serum	ICP-MS, fsLA-MC-ICP-MS	Serum pre-treated by MAD, separation of Cu on a Cu-specific resin and pre-concentration to 0.3 to 4 mg Cu L^−1. Total Cu determination from 1 μL pre-treated serum using microflow nebuliser and sample flow rate, 0.1 mL min^−1. Cu isotope ratios measured on 1 μL pre-treated serum deposited on ultrapure Si wafers. Laser parameters were: repetition rate, 6 kHz; scan speed, 30 mm s^−1; 5 μm spacing between lines and fluence, 0.7 J cm^2	71
Cu (exchangeable)	Serum	ICP-MS	Serum (300 μL) incubated with EDTA (3 g L^−1) at pH 7.0 to 8.0 for 60 min prior to ultrafiltration (10 kDa filter). The LOQ was 0.19 μM, RSDs ≤4.7% and recoveries, 94 to 102%	275
Cu species	Serum	LC-ICP-QQQ-MS	Serum pre-diluted 10-fold in H2O followed by inline 2-fold dilution and injection onto column for 18 s at 1 mL min^−1 (300 μL injection). Bound and extractable Cu species separated on resin that binds Cu at pH 6 to 8 with gradient elution using NH4CH3CO2/NH3 solution and 10% v/v HNO3. ICP-MS analysis in QQQ mode with gases, O2 (0.23 mL min^−1) and He (0.80 mL min^−1). Ion ^63Cu^+ monitored with dwell time, 100 ms, in TRA mode. Analysis time was 6 min	203
Cu, Zn	Hair	LIBS	Hair (200 mg) dissolved in 2 M NaOH and sonicated for 30 min before being heated at 80 °C for 30 min. Solution dripped onto filter paper for LIBS sampling. The LODs were 0.0146 and 0.3517 μg g^−1 for Cu and Zn respectively	46
Cu, Zn	Mouse kidneys	LIBS	10 μL drop of 1.53 mM Zn solution applied to paraffin embedded tissue sections, evaporated to give a dried 4 to 5 mm diameter drop and used as an IS. Optimal laser settings were: ablation lens focus 150 μm under sample surface; gate delay, 500 ns and analysis in Ar atmosphere using Ar purge	172
Cr species (Cr^VI, Cr-transferrin and Cr-albumin)	Serum	LC-ICP-MS	Separation of Cr species via monolithic convective interaction media diethylaminoethyl column with linear gradient elution from 100% buffer A (50 mM Tris–HCl + 10 mM NaHCO3 at pH 7.4) to 60% buffer B (buffer A + 2 M NH4Cl) in 10 min at flow rate, 1 mL min^−1. Recovery of separated Cr species was close to 100% and RSDs were <8%	149
Cr species	Hepatopancreatic tissue (CRM)	IE-HPLC-ICP-MS	Extraction from homogenised tissues (80 g) by incubation for 60 min with 10.8 mM EDTA solution (pH 10.5) and 1000 μg L^−1 53Cr^III IS solution. To prevent changes in Cr oxidation state, 1 M NH4NO3 added. Following centrifugation, supernatant incubated at 70 °C for 90 min to form [Cr^III–EDTA]^− complex. Separation of total Cr^VI and soluble Cr^III species achieved with 20 μL injection onto strong anion exchange column and isocratic elution with 30 mM NH4NO3 (pH 6). The ICP-MS operated in He mode with dwell time, 100 ms. The LODs and LOQs were ≤0.03 and ≤0.08 μg g^−1 respectively	150
F	Rat liver tissue	HR-CS-GFAAS	MAD digestion of ∼500 mg tissue with final HNO3 concentration of <2%. Thermal pre-treatment with Pd/Zr modifier (0.1% m/v Pd in 20 mg L^−1 Zr) with Ga (2 g L^−1) in HNO3 at 1100 °C. Modifiers, CH3COONa (1 g L^−1) and Ru^III nitrosyl nitrate (0.6 g L^−1), added before injection of sample (20 μL) with Ga (2 g L^−1) in HNO3 as the molecule-forming reagent. Maximum pyrolysis temperature was 550 °C and optimum molecule forming temperature was 1550 °C. Molecular absorption of GaF was measured at 211.248 nm. The LOD was 6 mg L^−1	211
Fe, Pt, FePtNPs	HepG2 cells	spICP-MS	Bespoke microchip for droplet generation, cell lysis and droplet splitting using a magnet for selective quantification of FePt released from intracellular FePt NPs. Microchip also operated in total-mode for quantification of total Fe/Pt in single cells. Optimal dwell time was 5 ms	64
Fe, Zn	DBS	HR-CS-GFAAS	Filter paper disks (3.4 mm diameter, equivalent to 3.86 μL blood) directly inserted onto W-coated “boat type” graphite platform and transported into GF using automatic solid sampler. 15% (v/v) H2O2 solution added as chemical modifier. Measurements performed at 307.572 and 446.165 nm for Fe and 307.589 nm for Zn. The LOQs were 10.2 and 1.2 mg L^−1 for Fe and Zn respectively	81
I	Urine	AMS	Samples pre-treated in autoclave with H2O2 then acidified with HNO3. I precipitated as AgI for AMS determination. Isotope ratios, ^129/127I, and total ^129I concentrations measured	276
Mg isotope ratios	CSF	MC-ICP-MS	Sample (5 μL) digested in 14 M HNO3 (1 mL) + 9.8 M H2O2 (0.25 mL) in closed beakers at 110 °C for 18 h. High-gain (10^13) Omega Faraday cup amplifiers used to monitor δ^24/26Mg by MC-ICP-MS. Expanded uncertainty for QC material was ± 0.16‰	277
Mo isotope ratios	Urine	MC-ICP-MS	Isolation of Mo on N-benzoyl-N-phenylhydroxylamine chromatography resin without acid digestion. Sample loading/wash solvent consisted of 0.1 M HF/1 M HCl mixture and elution solvent was 6 M HF. ^97Mo–^100Mo double-spike mass bias correction applied. External precision on δ^98/95Mo was >0.08‰	208
Np, Pu isotopes	Urine	SF-ICP-MS, ICP-QQQ-MS	Pre-treatment of 20 mL urine, spiked with 11.4 pg mL^−1 242Pu involved: organic matter decomposition performed either by HNO3/H2O2 digestion at 220 °C or NH4HF2 fusion at 250 °C; CaF2/LaF3 co-precipitation; centrifugation; addition of H3BO3 and 7.2 M HNO3; filtration; addition of NaNO2 and heating at 40 °C for 30 min. Pre-treated samples subjected to AG MP-1M anion exchange resin column separation, where U eluted with 7.2 M HNO3 and ^237Np/Pu isotopes with concentrated HBr. Final eluent evaporated to dryness and residue re-dissolved in 4% HNO3, containing 2.2 pg mL^−1 233U IS. The LODs ranged from 0.015 to 0.025 fg mL^−1	55
Pb	Blood	HG-DBD-AFS	Blood (2 mL), diluted 40-fold in HNO3, directly introduced into double gas–liquid separator. Plumbane generated by mixing with 3.5% (v/v) HNO3 and 1.4 mL reductant (30 g L^−1 NaOH + 8 g L^−1 KBH4 + 10 g L^−1 H3BO3 + 5 g L^−1 K3[Fe(CN)6]). Triple-layer coaxial quartz tube employed as DBD trap. Pb trapped with 9 kV discharging under 50 mL min^−1 air and released by 12 kV discharging under 110 mL min^−1 H2. The LOD was 0.16 μg L^−1 and total analysis time was <10 min	114
Pb	Blood, hair and CRMs (stream sediments)	HG-SAGD-OES	Acid digestion. Optimised HG parameters were: 0.8% K3[Fe(CN)6] additive; 2% NaBH4; carrier acid, 2% HCl; carrier gas, 700 mL min^−1 He. Optimised SAGD parameters were: discharge current, 40 mA; supporting electrolyte, 0.8 M HNO3; HNO3 flow rate, 3.3 mL min^−1, discharge current, 40 mA and discharge gap size, 2 mm. Pb emission line at 368.3 nm selected. The LOD was 0.061 ng mL^−1	89
Pd, Pt	Rat blood, serum and tissues	ICP-MS	Samples (50 mg) digested in closed tubes using 3 : 1 (v/v) HNO3 :  HCl at 90 °C for 60 min. The LODs for both elements were 0.001 μg L^−1 and intermediate precision RSDs were 0.52 to 1.53%	32
Pu	Urine	SF-ICP-MS	Incubation of urine (2 mL), spiked with 100 ng L^−1 242Pu tracer, and concentrated HNO3 for 5 min followed by addition of aqueous 4 M NaNO2 and reaction for 5 min. Automated low pressure chromatography system with CF-ThU-1000 column used to isolate ^239Pu. The LOD was 0.63 pg L^−1	278
Ra	Urine	ICP-QQQ-MS	Urine diluted with 2% v/v HNO3 and 100 ng L^−1 of IS, ^193Ir. ^226Ra determined in no gas QQQ mode. The LOD was 0.007 ng L^−1	69
S	Hair	HR-CS-GFAAS	Hair washed with H2O and acetone then digested with HNO3/H2O2 before a further 1000-fold dilution. Chemical modifiers, 15 μg Pd/10 μg Mg and 800 μg W used. Pyrolysis and vaporisation temperatures were 900 and 2400 °C respectively. Wavelengths, 257.961 and 258.033 nm, were monitored simultaneously	79
Se (NPs)	Plasma	CE-ICP-MS	Separation of SeNPs from dissolved Se species with modified CE-ICP-MS method involving coated fused silica capillary. Optimised sheath liquid flow from syringe pump was 14 μL min^−1. Optimal ICP-MS parameters were: carrier gas flow, 0.98 L min^−1; sampling depth, 5 mm and Ar make up gas flow rate, 0.1 L min^−1. Reaction gas, O2, at 0.4 mL min^−1, employed in QQQ mode and [^80Se^16O]^+ monitored	144
Sr	Tears, eyelashes, saliva, environmental standards and water	ID-TE-TIMS	No sample pre-treatment to remove Zr or other matrix elements. ID using ^86Sr spike. Optimised TE profiles involved: temperature ramp rate, 100 mA min^−1; Sr target intensity, 4 V and maximum filament current, 4500 mA. Maximum of 20 ng Sr (natural Sr and ^86Sr spike) loaded onto Re filament. The LOD was 0.029 fg without pre-concentration	77
Te, (P)	Bacterial cells	scICP-QQQ-MS, spICP-QQQ-MS	For single cell analysis, cell suspension, containing 10^5 cells per mL^−1, introduced into plasma at 10 μL min^−1via high performance concentric nebuliser and total consumption spray chamber. In TRA mode, ^126Te monitored “on mass” and, in a separate run, P monitored as [^31P^16O]^+ in QQQ mass-shift mode. Dwell time was 5 ms with run time, 2 min. The LOD was 0.068 ± 0.008 fg per Te cell. spICP-MS with same parameters was applied to lysed cells	65
Ti	Serum	ICP-QQQ-MS	Serum diluted in 1% (v/v) HNO3. ICP-QQQ-MS analysis with O2 : H2 reaction mixture (flow rates 0.4 and 2.0 mL min^−1 respectively) and [TiO]^+ monitored with O2 mass shift method. Optimised bias voltage was −5 V. The LODs for Ti isotopes ranged from 0.78 to 7.20 ng L^−1	66
U	Urine	HR-ICP-MS	Sample digestion in concentrated HNO3 with 0.5 ng g^−1 233U tracer. Isolation of U on Tru spec resin. Analysis in low resolution mode, taking 3 min per sample. Total U concentrations and ^235U/^238U isotope ratios determined. Procedural blanks were <2.5 pg	279
Zn	Toenails	XRF	Toenails washed and sonicated in acetone and H2O. Measurement of Zn with mono-energetic portable XRF instrument using three 300 s trials for each clipping. Excitation beam was Mo Kα X-ray at energy of ∼17.5 keV. Results assessed using factory-calibrated Zn concentrations and analysis of energy spectra for Zn characteristic X-rays	130
Various (7)	Central nervous system germ cell tumour tissue	LA-ICP-TOF-MS	Distribution of elements, Ca, Cu, Fe, Mg, P, S and Zn, studied in paraffin embedded tumour sections (5 μm thickness). Low dispersion laser connected to ICP-TOF-MS via aerosol rapid introduction system (Ar make up gas flow, 1.10 L min^−1, and He carrier gas flow, 0.6 L min^−1). LA sampling performed in fixed dosage mode, with repetition rate, 25 to 30 Hz, using a square spot size of 20 μm. Pixel size was 10 μm and fluence, 1 to 1.5 J cm^−2. ICP-TOF-MS with mass resolution 6000 used in standard operation mode allowed detection of ions from m/z 14 to 254	162
Various (20)	Plasma and HCT-116 cells	LA-ICP-TOF-MS	Automated micro-spotting of gelatine-based micro droplet standards (400 ± 5 pL) producing droplet sizes of ∼200 μm diameter. Low dispersion LA parameters were: fixed dosage mode; repetition rate, 200 Hz; 5 × 5 μm square spot size and fluence, 0.60 to 0.90 J cm^−2. The ICP-TOF-MS with mass resolution 6000 was operated in standard operation mode and detected ions from m/z 14 to 256. Analysis time was <5 min per standard	152
Various (10)	Mesothelioma cells containing asbestos fibres	LA-ICP-TOF-MS	LA parameters: 3 μm spot diameter; 50 Hz repetition frequency; 150 μm s^−1 scan speed and 4 J cm^−2 fluence. Gases, H2 and He, used in ICP-MS collision reaction cell and MS method scanned from Na to U every 25.5 ms. Elements, Al, Ca, Fe, K, Mg, Mn, Na, P, Si and Ti, used for imaging of asbestos fibres	160
Various (21)	Foetal urine	ICP-MS	Micro-uptake nebuliser with 0.1 mL min^−1 sample flow and low volume (20 mL) spray chamber enabled determination of 21 elements in 200 μL urine following dilution in HNO3. Measurements performed in no gas, O2 reaction (As and Se) and NH3 reaction (Al, Cr, Cu, Fe, Mn, V, and Zn) modes with Ge, Rh and Tb as IS’s. Analysis time was ∼5 min	199
Various (13)	Seminal plasma	TXRF	Seminal plasma (0.2 g) diluted 1 : 1 in 1% Triton X-100 aqueous solution and IS, Y, added at 5 mg kg^−1. Mixture homogenised, 5 μL deposited onto siliconized quartz glass reflector and dried using IR lamp. Measurement time, 2000 s, gave RSDs of approximately 10%. LODs were 1.9 and 2.6 mg kg^−1 for Ca and K respectively and 0.04 to 0.33 mg kg^−1 for other elements	125

---

6.1 Metallomics

A comprehensive review of recent advances in inorganic speciation complements the work with clinical and biological materials, foods and beverages covered in this update.⁴ Discussion of work relating to NPs, some of which is also relevant to this part of the Update, is included in Section 5.

Intracellular metabolism of the toxic metal Cd was investigated by Men et al.¹⁴⁸ It is well known that Cd combines with the sulfur-rich protein, metallothionein, but in this work other Cd species were also determined. Cell lysates were prepared from two cultured cell lines, HepG2 and MCF-7, in which Cd had been added to the growth medium. Cell viability was monitored using fluorescent probes for visualisation by microscopy and by staining to show apoptosis. The Cd species in lysate supernatant were separated by CE coupled to ICP-MS. Conditions for the electrophoresis were investigated and the broad Cd²⁺ peak was sharpened, with increase in sensitivity, by addition of SDS to the running buffer. Electrophoretograms of cell lysates presented peaks for Cd-metallothionein and a second species that was demonstrated to be Cd-GSH. The work provides an additional approach for investigations of metal metabolism.

Milačič and colleagues¹⁴⁹ used weak anion-exchange convective interaction media, diethylamino (CIM DEAE) short monolithic column disks (12.0 mm i.d, length 3.0 mm, bed volume 0.34 mL) seated into a nonporous self-sealing fitting matrix support ring, made of highly porous polyglycidyl methacrylate-co-ethylene dimethacrylate, to separate Cr species in serum. The eluate from the chromatographic column passed through a UV-Vis detector to monitor absorbance due to proteins at 278 nm, and then led to an ICP mass spectrometer. The system was used to investigate the kinetics of interactions between Cr^III and Cr^VI with serum constituents. Human serum samples were doubly spiked with enriched ⁵³Cr^III and ⁵⁰Cr^VI and incubated at 37 °C. Samples were taken for chromatographic separation and detection at intervals from 5 min to 48 h. It was seen that ⁵³Cr^III rapidly interacted with transferrin and that by 48 h more than 90% was bound, with the remainder associated with albumin. The ⁵⁰Cr^VI was reduced to ⁵⁰Cr^III and was bound to transferrin while 20% remained as ⁵⁰Cr^VI. These results showed that the main Cr species in human serum at physiological concentration levels was the Cr-transferrin complex.

Chromium speciation was also investigated by Pechancova et al.¹⁵⁰ who developed and validated a procedure to determine total Cr and soluble Cr species in extracts from chicken muscle samples using IE-HPLC-ICP-MS. The work concentrated on the extensive validation of the sample preparation and analytical methodology. The assay parameters were reported as LOD: 0.03 μg g⁻¹, LOQ: 0.08 μg g⁻¹, linearity: r = 0.9998, accuracy: 86–110%, precision: <10%, extraction recovery: 89–110%. Chicken tissue samples enriched with ⁵³Cr^III and ⁵⁰Cr^VI isotopes were analysed to investigate possible species interconversion during the procedure. The Cr^III was stable during all steps of developed analytical approach, while about 30% of added Cr^VI was reduced by the sample matrix components.

Following from previous work with Goji beers, a type of endosperm liquid, Alchoubassi et al.¹⁵¹analysed coconut water, also an endosperm liquid, to determine whether essential trace elements occur as low molecular weight complexes, which could prove to be a valuable food supplement. Speciation was investigated by SEC-ICP-MS and HILIC-electrospray-Orbitrap MS. The metal species identified included Cu complexes with phenylamine and nicotianamine, Fe complexes with citrate and malate, Mn complex with asparagine and Zn complexes with citrate and nicotianamine. The total element concentrations were in the ranges of 0.2–2.7, 0.3–1, 3–14 and 0.5–2 mg L⁻¹ for Cu, Fe, Mn and Zn, respectively, and varied as a function of the origin of the nut and its maturity.

6.2 Imaging with MS, X-rays and LIBS

Several papers during this review period have focussed on quantitative strategies for LA-ICP-MS. Schweikert et al.¹⁵² described a novel approach for the production of gelatine calibrants using a micro-spotter device. When combined with a prototype low-dispersion LA chamber coupled to ICP-TOF-MS for quasi-simultaneous multi-element analysis, high repetition rates (pixel acquisition rates > 200 Hz) and fast analysis times were achieved. The micro-spotter produced droplets of approximately 200 μm diameter from 400 ± 5 pL of gelatine spiked with the elements of interest, that produced 10 calibration standards within 1 h. The concentrations of the standard spots were confirmed by conventional solution-based ICP-MS following acid dissolution. The calibration curves achieved correlation coefficients of at least 0.995 and impressive LODs for imaging, i.e. <0.01 μg g⁻¹ for several elements. Proof of concept was demonstrated with spiked serum samples quantified for Cu, Fe, P, Pt and Zn at several concentration levels, spotted in the same way as the calibrants. Recoveries were in the range of 84–104% and RSDs were between 1.1 and 7.8% (n = 4). Additionally, the approach was applied to multicellular tumour spheroids exposed to cisplatin and oxaliplatin at clinically relevant levels. This demonstrated the quantitative imaging capability for multiple elements in single cells in less than 2 h, which is impressive. Billimoria et al.¹⁵³ published some interesting findings from calibrants prepared from different elemental species leading to different sensitivity responses using LA-ICP-MS. As no suitable CRMs or calibrants are available, typically researchers in this field rely on in-house materials produced from inorganic salts. However, little is known about whether this affects quantitative results for imaging applications in inhomogenous samples such as brain tissue slices. In this work, Fe and Se were selected as key elements in neurodegenerative disease and inorganic salts were compared with protein-based species, spiked into pig brain homogenate to produce a range of calibrants at physiological concentrations. Following a number of systematic experiments, the results for Fe were within experimental error for both species (inorganic Fe vs. ferritin). However, interestingly, this was not the case for Se (inorganic Se vs. extracted Se yeast protein). At lower laser fluences, ≤2 J cm⁻², that is relevant for biological tissue samples, there was a discrepancy of up to 36% between the regression slopes of the two species despite signal normalisation with ³²S¹⁶O as an IS. It was suggested that a higher laser energy was required to effectively release the Se from the Se-proteins and transport to the ICP. Additionally, the study included the use of a total consumption nebuliser and a source of carbon through the addition of MeOH, that significantly enhanced the Se signal intensity and improved the LOD. It was found that 25% v/v MeOH was an optimal balance between signal stability and enhancement, achieving an LOD of 0.032 mg kg⁻¹ for Se. The paper represents an important development in the journey towards quantitative tissue imaging. Arnaudguilhem and co-workers¹⁵⁴ described a multi-element approach implementing an aqueous soluble polymer for calibration (Cd, Co, Cu, Fe, Hg, Mo, Ni, P, Pt, Se and Zn) and internal standardisation (In, Ir and Rh). Dextran was used as the polymer, however elevated background levels of Fe and P in the material led to difficulties with the calibration for these elements and they were not analysed further. Additionally, Hg was lost during the spin coating and drying process despite additional attempts to stabilise it with excess chloride ions and cysteine. Both of these issues highlight the challenges of producing matrix-matched calibration standards for solid state analysis. The ICP-MS parameters were optimised to obtain maximum signal and stability. Perhaps unsurprisingly, it was found that the behaviour of the elements was related to their first ionisation potential. The final methodology was validated using in-house spiked kidney tissue homogenate and the results were compared to total acid mineralisation, achieving agreement within 5–29% for Co, Cu, Mo, Ni and Pt. However, these figures dramatically increased for Cd, Se and Zn, leading to concern about the authors conclusion that the method was suitable despite the large biases. The approach was then applied to two human tissue samples, but the authors noted further potential issues as the fresh tissue deposited on the coated slides could experience a drying effect in the LA chamber due to the He flow, therefore recommending the use of a cryocell in future work. This, combined with the validation results, does not inspire confidence for this quantification approach in comparison to those described in the other work covered in this review.

The use of LA-ICP-MS as a quantitative tool for disease identification and diagnosis has also received attention. Bishop et al.,¹⁵⁵ used LA-ICP-MS to improve the quantification of dystrophin in muscle tissue via Gd-labelled anti-dystrophin antibodies for the detection of Duchenne’s muscular dystrophy. Patients exhibit very low levels of dystrophin and testing in this field currently suffers from poor sensitivity and reproducibility with classical techniques. Control and diseased muscle tissue samples from both mice and human subjects (with different genetic mutations) were analysed by LA-ICP-MS alongside traditional immunohistochemistry and immunofluorescence methods. Gelatine standards doped with Gd were used with LA-ICP-MS to produce quantitative data as well as spatial tissue images. The study also included testing a number of statistical models for image processing of the muscular dystrophy tissues because the low levels of dystrophin against the background could cause noise and false identification, concluding that k-means clustering provided the optimal approach to overcome these issues. The analysis of mouse and human samples by LA-ICP-MS compared extremely well to the classical techniques with the added benefit of quantitative data. The human subjects had differing mutations and level of disease progression, that was clearly reflected in the images and Gd concentrations. The method had a number of benefits over standard histological procedures and could be multiplexed with other tagged-antibodies to provide further insight. Kim et al.¹⁵⁶ described the application of LA-ICP-MS imaging for the detection of Wilson’s disease as a significant improvement over current testing methodologies, in particular for early diagnosis where classical techniques can lack the required sensitivity. The researchers utilised liver biopsies from mouse models and compared histological staining and SEM-EDX with LA-ICP-MS. The results showed that rhodanine staining did provide clear results for a human tissue sample from a patient with Wilson’s disease but no difference between control and diseased mouse groups. The SEM-EDX measurement did not detect Cu in the liver samples or any specific aggregate sites. Whilst it was able to identify C, Ca, Cl, K, Na, O and Si, these were only relative levels as quantitative analysis was not possible. Conversely, the LA-ICP-MS method was able to clearly differentiate between controls and diseased models and provide Cu concentrations in addition to Ca, Fe, Mn, Na, and Zn using spiked homogenised tissues as calibrants. With the use of a custom Excel imaging template, visual maps of the elemental distribution were produced with clear differences between the controls and diseased tissues observed and trends found for other elements (e.g. Zn increased for disease models, Fe and Mn had irregular distributions across the liver tissue). The approach has potential to improve Wilson’s disease diagnosis however, LA-ICP-MS is far from routine in clinical settings and required 4–6 h analysis time for one sample. Gao and co-workers¹⁵⁷ utilised AuNPs for the investigation of amyloid beta peptide distribution in brain tissue with Alzheimer’s disease. The AuNPs were functionalised with anti-amyloid beta antibodies and incubated with brain tissue from mouse models followed by LA-ICP-MS analysis. The implementation of calibrants prepared from homogenised brain tissue enabled quantitative analysis. The high sensitivity of Au provided clear images of the localisation of amyloid beta, that was consistent with traditional immunohistochemical staining techniques, with the added benefit of the concentration via the Au level. The work could have benefit for understanding the impact of Alzheimer’s disease treatments and progression.

An unusual sample type was reported by Olszewska and Hanc¹⁵⁸ who had the rare opportunity to analyse a natal tooth from a 14 day old baby (extracted due to tooth mobility, difficulty feeding and potential health risk). Quantitative imaging was achieved using spiked tooth homogenate for Ca, Cd, Cu, Mg, Mn, Ni, Pb, Sr and Zn with P used as an IS. The approach was validated using the CRM NIST SRM 1400 (bone ash) and the total elemental levels in the tooth were determined by solution-based ICP-MS analysis after MAD. The correlation coefficients were >0.99, with RSDs in the range 2.5–7.5%, and the results obtained for the CRM were in the range 92–110% of the certified values, indicating method suitability. The LODs ranged from 0.08 μg g⁻¹ for Pb to 10 μg g⁻¹ for Ca. The results provided useful data for this uncommon sample type, giving insights into the prenatal and perinatal development of teeth. The data presented interesting findings, in particular the presence of Cd, Ni and Pb in the developing root region of the tooth was a concern, however the scarcity of the sample type and analysis of only a single tooth still leaves a significant knowledge gap.

Two papers from the same researchers,^159,160 also featured an unusual application, LA-ICP-MS was used as a tool for asbestos identification in malignant mesothelioma cell models. In the first work,¹⁵⁹ a two volume low dispersion LA cell was used in combination with SF-ICP-MS, achieving a resolution of 2 μm, with fast analysis times. Human mesothelioma cells were exposed to four mineral fibres (white, blue and brown asbestos and the non-asbestiform wollastonite) that contain different elemental ratios of Fe, Mg and Si. The resultant images enabled the localisation of the fibres and clear identification of the four types. In the second study,¹⁶⁰ the same fast LA system was utilised with ICP-TOF-MS to obtain pseudo-simultaneous analysis of the mass range 23 to 238 amu, significantly expanding the range of elements detected. The same cell exposure experiment as described in the first paper was performed. Principle component analysis was applied to the data using Al, Ca, Fe, K, Mg, Mn, Na, P, Si and Ti, identifying the four different fibre types, with the amosite fibres forming two distinct groups. There are two forms of amosite known (grunerite and cummingtonite) but it was not clear which were used in this work, therefore requiring further investigation. However, both papers demonstrated that the approach delivered asbestos classification in lung cell culture, offering a number of advantages over classical optical techniques, such as distinguishing between asbestos and non-asbestiform fibres, in relatively quick analysis times.

Bucker et al.¹⁶¹ studied the impact of La-based therapies on bioaccumulation within the body in comparison to Gd-based imaging drugs. In this work, tissues from two patients with long term exposure to La₂(CO₃)₃, a phosphate binder for chronic kidney conditions, were analysed by quantitative LA-ICP-MS to determine the ratio of Gd to La. It was found that La deposition was relatively low compared to the dosage and with respect to Gd, indicating La is more readily metabolised than Gd. Whilst Gd and La were co-localised in the tissues analysed, the Gd/La ratios varied significantly with the largest differences found in the thyroid (90 ± 17) and the cerebellum (270 ± 18). The data were from a very limited cohort so it was difficult to draw definitive conclusions, but the information presented suggests potentially separate metabolism mechanisms for the different REEs. Theiner and co-workers¹⁶² described the use of a low dispersion two volume LA cell with ICP-TOF-MS for the qualitative imaging of aggressive brain tumour germ cells. Two patients undergoing Pt-based chemotherapy and one control were analysed for Pt with endogenous elements such as Ca, Cu, Fe, Mg, P, S and Zn. A resolution of 10 μm was achieved leading to images with clear localities. This was confirmed with classical histological hematoxylin and eosin staining from adjacent tissue slices. It was noted that Pt was found at the highest levels in the necrotic regions which correlated with Ca, Mg and S, whereas the lowest elemental levels were observed in fibrotic scars. The study showed the active accumulation and action of the chemotherapy drug within the tumour.

The subject of sample preparation for elemental imaging was covered by Pushie et al.¹⁶³ who recommended that sucrose cryoprotection, commonly used for classical histological staining, should be avoided. By using left and right brain hemispheres from mice and rats, direct comparisons by either sucrose cryoprotection or rapid freezing in liquid nitrogen cooled isopentane could be performed. The tissue slices were analysed by XRF microscopy with the results showing that significant redistribution of Zn occurred in the hippocampus with the sucrose cryoprotection method compared with the rapid cooling approach. They concluded by urging others to avoid the use of this method especially when investigating the elemental distribution of labile metal ions. Nakanishi et al.¹⁶⁴ described the use of LA-ICP-MS for the analysis of 38 elements in human skin following simple sectioning as sample preparation. Although the authors concluded that the levels observed for several elements was in line with previously reported data, in this work they did detect Ba, I, Mo and Sr for the first time.

Turyanskaya and co-workers^165,166 focussed on the analysis of bone by LA-ICP-MS and X-ray techniques for different purposes in two papers. In the first,¹⁶⁵ a human bone-cartilage sample was used to directly compare LA-ICP-MS and μXRF spectrometry. Quantitative results were obtained for both techniques using the CRM IAEA H-5 (animal bone) and an in-house prepared cartilage standard. As μXRF spectrometry is a non-destructive technique, the exact same tissue sections were analysed by both methods. A human femoral head biopsy sample was used and two sections were prepared at a thickness of 15 μm and 41 μm, that captured both the bone and articular cartilage tissue, enabling a comparison of the two tissue types. Overall, there was good agreement between the techniques, although this was stronger with the 41 μm section. The LA-ICP-MS produced maps with higher resolution, permitting more feature identification, but both instruments could identify a “tide mark” at the bone-cartilage interface that was found to contain Ca at levels similar to those in bone, but significantly enhanced Zn concentrations. The study demonstrated the proof of concept and added some knowledge to the elemental analysis of cartilage. In the second work,¹⁶⁶ utilised SR-μXRF and SR-nanoXRF spectrometry to investigate the distribution of Gd in human cortical bone following exposure to MRI contrast agents 8 months prior. Three different synchrotron facilities (ANKA, Germany, ESRF, France and Diamond Light Source, UK) were employed on the sample, which is a unique opportunity given the limited access to these resources. The ANKA beamline provided the SR-μXRF data, with a resolution of ∼20 μm, and included quantitative backscattered electron imaging to estimate the level of mineralisation, enabling the derivation of % (m/m) Ca. The sensitivity for Gd was very low however the distribution correlated with bone features (cement lines and interface between mineralised matrix and vascular canals). Due to the low resolution, the sample was then analysed at ESRF and Diamond using SR-nanoXRF spectrometry. In both cases, the excitation energy was higher with the nano beam, that enabled the detection of other elements, such as Zn. The increased resolution allowed identification of finer structures in the elemental maps. It was found that Gd was present in the calcified bone regions and both Gd and Zn were co-localised at the cement lines, in a region approximately 10–15 μm wide. The study represented the first imaging analysis of Gd in exposed human bone tissues, adding valuable data to support understanding Gd deposition and potential release mechanisms.

A number of developments with SR-XRF for the imaging analysis of brain tissues were reported in this review period. Genoud et al.¹⁶⁷ utilised SR-nanoXRF imaging of substantia nigra from seven subjects with confirmed Parkinson’s disease. The researchers also utilised X-ray ptychography, an imaging technique where the object is moved laterally in the X-ray field to obtain many interference patterns, that are subjected to computational processes to generate an image, achieving a resolution of 13.5 nm. The SR-nanoXRF analysis focused on six key elements, namely Cu, Fe, K, P, S and Zn, with quantitation achieved through comparison to a thin film reference sample (AXO RF8-200-S2455). These techniques, alongside conventional staining methods, provided a powerful combination leading to highly detailed tissue images. This enabled examination of Lewy bodies, superoxide dismutase 1 deposits and neuromelanin, in combination with relevant elements, such as Cu and Zn, possibly implicated in Parkinson’s disease progression. The work highlighted the need for a combination of complementary techniques to gain maximum data from precious samples for complex diseases. Shi and co-workers¹⁶⁸ focussed on trace element distributions in brain tissue following ischemic stroke by SR-μXRF spectrometry. Using rat models, the spatial imaging of Ca, Cu, Fe, K and Zn was obtained across several time points following brain ischemia (3, 4.5, 6, 12 and 24 h, then at 3, 5, 7, 10, 14 and 28 days). A number of trends were revealed across the measurement period, providing a comprehensive picture of the elemental changes.

In a curious advance, Kondo et al.¹⁶⁹ reported the 3D analysis of a single hair strand by SR-μXRF spectrometry. The workers built a rotational stepping motor mechanism to add to the SR-μXRF setup and utilised the scattered X-ray intensity to determine the rotational adjustment for the XRF signal. Quantitation was achieved through production of an in-house reference sample using an electron gun and vacuum evaporation to layer Cr, Fe, Ni and Zr on Kapton film. A spatial resolution of 2 μm enabled detailed patterns in the cross-section and longitudinal direction of the hair strand to be visualised for Br, Cu, Fe and Zn. The authors noted the potential of the application for dietary, archaeological and forensic investigations.

It is not often that food products feature in the imaging section of this review, however, Gu et al.¹⁷⁰ applied multiple X-ray-based techniques to the analysis rice grains. This enabled detection of the chemical species as well as distribution information within the grain. It was found that the majority of Cd was present in the grain as Cd-thiolate, most likely due to thiol-rich proteins, with Cd-carboxyl and Cd-histidine covering the remaining forms. Furthermore, imaging analysis determined two patterns of Cd distribution, which were due to changes in the supply of soil porewater – the difference was quite marked as one form yielded an even distribution across the grain, whereas, in the second, it was concentrated in the outer layers, that would typically be removed during the milling process. The combination of XANES, SR-μXRF and μPIXE provided a comprehensive view of Cd in rice, alongside the total content of other elements such as As, Ca, Cu, Fe, K, Mg, Mn, P and Zn.

Imaging applications utilising LIBS are not a regular feature within this section of the review, however, two papers have covered the application of LIBS for tissue imaging. Kiss et al.¹⁷¹ performed 3D imaging of skin cancer tumours for Ca, K, Mg and Na in formalin fixed paraffin embedded tissues following standard histopathological diagnosis. Four tumour types were selected, namely cutaneous malignant melanoma, squamous cell carcinoma, basal cell carcinoma and a benign tumour for comparative purposes. Correlations were observed at the tumour margins and in regions of healthy and tumour tissues enabling clear differentiation in the images, matching well to the histological staining. The 3D images were generated by multiple slices of the tissue, with a depth resolution of approximately 150 μm. The tumour regions were visible and the intensities varied showing the edge of the tumour. With the use of a k-means clustering algorithm, the tumour features could be identified but the boundaries/demarcation required additional refinement. The authors concluded further work to the model was needed, as well as differentiation of cancer type, which was beyond the scope of the study, yet the data showed good promise for LIBS. Sindelarova and co-workers¹⁷² considered the optimisation of the LIBS parameters to improve sensitivity for trace elements in soft tissues by enhancing the level of Zn present by spotting known amounts of a Zn standard onto tissue sections. The mouse kidney was formalin fixed and paraffin embedded, then sliced into 10 μm sections and 10 μL droplet of 100 ppm Zn was added and dried (spot size was approximately 4–5 mm). Although the level of Zn within the drop varied in homogeneity and the size of the section varied, the total amount of analyte in the drop was the same. This enabled comparison between drops and provided enhanced signals for optimisation of the LIBS parameters. Following this, the Zn intensity was significantly higher in the tissue section compared to the original conditions. These settings were then applied to other matrices using the Zn dropping procedure, namely CRM BAM-310 (aluminium alloy), epoxy embedded radish, polystyrene, aluminium foil and a glass slide. The SNR was assessed showing large differences between the substrates, from around 15 for the aluminium foil to approximately 100 for kidney tissues, demonstrating the need for matrix optimisation. Tissue maps were also generated in around 15 min, significantly quicker than with LA-ICP-MS approaches.

6.3 Trace elements in pregnancy

Once again, various articles have described links between maternal element levels and adverse outcomes in pregnancy, although many of the findings confirm previously documented associations and have not been commented on in this review. Baser et al.¹⁷³ conducted a multi-element study looking at placental trace and toxic elements concentrations and second trimester spontaneous abortion. Digested placental samples from 30 women with second trimester abortion and 60 healthy term singleton pregnancies were analysed by ICP-MS for 9 toxic and 4 trace elements. Concentrations of As, Cd, Co, Hg, Mn, Pb, Sb, Se and Sn were elevated in the abortion group compared with controls. Element concentrations for all placental samples were below biologic exposure indices for blood indicating the need to derive acceptable levels in placental tissue.

Two studies of note investigated Cd exposure in pregnancy. In a birth cohort study, 11 heavy metals were quantified, using ICP-MS, in blood (n = 483) and urine (n = 512) samples collected from 20 to 28 weeks gestation and correlated with placental characteristics and birth weight documented in maternity records.¹⁷⁴ Elevated blood and urine Cd levels were associated with reduced placental and birth weight. Associations between prenatal Cd exposure and low birth weight are well documented, however this work adds to previous studies presenting a potential mediation role for the placenta between prenatal exposure and birth outcomes. Osorio-Yanez et al.¹⁷⁵ investigated metal (As, Cd and Pb) exposure and bone remodelling in 1054 pregnant women enrolled in the PROGRESS cohort study (Programming Research in Obesity, Growth, and Environment, and Social Stress). Blood metal levels were quantified by ICP-MS/MS and bone remodelling by quantitative ultrasound at the radius in the second and third trimester. Blood Cd concentrations were negatively associated with the radius bone density z-score, although reverse causality is not excluded and further studies are needed to confirm the findings.

A few publications in this review period described prenatal Pb levels and pregnancy outcomes. In the bone remodelling study by Osorio-Yanez et al.,¹⁷⁵ bone Pb was determined by K-shell XRF spectrometry and was also found to be inversely associated with radius z-score indicating reduced mineral density. Another large cohort study (n = 2174) explored the established association between maternal blood Pb and preeclampsia, revealing a non-linear dose-effect relationship with a cut-off point at 4.2 μg dL⁻¹.¹⁷⁶ Blood Pb levels were measured at 12–27 (+6) weeks gestation by AAS. Aung et al.¹⁷⁷ reported on a novel epigenome study that identified correlations between maternal metal levels (Mn and Pb) and DNA hypermethylation at sites involved in immune responses and nervous system development. Blood Cd, Hg, Mn, Pb and Se were measured by ICP-MS and DNA methylation analysed on the Illumina 450 K array for 97 women in the Early Autism Risk Longitudinal Investigation. The study is limited by sample size and further work is required to confirm any contribution the observed metal-induced DNA methylation may have towards adverse pregnancy outcomes.

While many studies evaluate maternal metal concentrations during pregnancy, few directly interrogate foetal exposure. Materno–foetal transfer of TiO₂particles from the food additive E171 was shown to occur during pregnancy through ICP-MS analysis of placental and meconium (an infant’s first stool) samples.¹⁴² Detectable levels of Ti were observed in all placentae (n = 22, range 0.01–0.48 mg kg⁻¹) and in 50% of the meconium samples (n = 8, range 0.02–1.5 mg kg⁻¹). Elemental mapping using STEM-EDX showed Ti and O present at particle deposits in both placentae and meconium. Ex vivo perfusion studies with an E171 suspension demonstrated materno–foetal transfer via detection of nanosized TiO₂ particles in foetal exudates.

6.4 Elements as tags for indirect determinations

Element tagging for immunoassay readout with ICP-MS has gained popularity over recent years and methodological developments are summarised in two review articles.^9,11 Numerous immunoassay formats have been evaluated, primarily using REE and transition metals with various signal amplifications strategies. While elemental MS detection can offer technological advantages for immunoassays in terms of LOD and multiplexing capability, instrumentation remains relatively complex compared to automated readout methods such as fluorescence, chemiluminescence and electrochemiluminescence. Despite the breadth of assays proposed, few have been validated against standard clinical guidelines and none have reached the routine clinical laboratory.

Element-tagged immunoassays to date are tabulated in the review by Torregrosa et al.⁹ and many have targeted tumour marker analysis. This review period was no different with proposed assays for CA19-9, CEA, AFP and hCG.^178–181 An AuNP-CA19-9 antibody probe ICP-MS based immunoassay demonstrated good correlation with electrochemiluminescence (R² 0.999) for CA19-9 analysis in 21 serum samples.¹⁷⁸ Following on from previous work on CEA, Jiang et al.¹⁷⁹ reported a duplex assay for determination of CEA and AFP using Eu and Sm tags with ICP-MS detection. This group applied the same methodology for hCG quantitation and both hCG and the duplex CEA/AFP immunoassays were validated to CLSI standards.^179,180 Zhang et al.¹⁸¹ described an alternative duplex assay for CEA and AFP based on DNA-conjugated AuNPs and rolling circular amplification with ICP-MS. Consistent with ELISA, analysis of human serum gave higher concentrations of CEA and AFP in liver cancer patients (n = 10) compared to healthy controls (n = 10). Using rolling circular amplification, protein quantitation was achieved with higher sensitivity but smaller dynamic range than with Eu/Sm tagging.^11,179,181

Yin et al.¹⁸² used Mucin 1 (MUC1) from the human breast cancer cell line MCF-7 to model a multifunctional platform based on capture and release for non-destructive enumeration of circulating tumour cells (CTCs), thus allowing for subsequent molecular and functional readouts. Capture of MUC1 by capture probes (MUC1 specific aptamer hybridized to partially complementary strand (initiator) and immobilised on magnetic beads) caused release of initiator strands that were then hybridised to Tb labelled complementary strands (substrates) immobilised on detection probes. Substrate was then cleaved from the detection probe by a specific endonuclease and Tb-labelled fragments subjected to ICP-MS detection. Benzonase nuclease was used to digest nucleic acids on the capture probe releasing MUC1 for re-culture. Recoveries between 82 and 105% were obtained for whole blood samples spiked with different concentrations of MCF-7 cells.

Other analytes targeted by elemental immunoassay formats include mycotoxins, thrombin and Plasmodium falciparum lactate dehydrogenase (PfLDH).^183–185 Multiplexed analysis of two mycotoxins (ochratoxin A and fumonicin B1) in wheat flour was achieved using capture probes (specific aptamers immobilised on magnetic beads) and different NP labels (CdSe-QD and AgNP), followed by ICP-MS determination of Cd and Ag.¹⁸³ Detection of PfLDH, with LOD (1 parasite μL⁻¹) comparable to PCR, was reported using a sandwich immunoassay format with AuNP-labelled mAb and ICP-MS readout.¹⁸⁵ This group speculated on future point of care applications for malaria control, should technology allow. Finally, thrombin was used as a model biomarker to demonstrate the applicability of graphene oxide–AuNP composites and spICP-MS in clinical diagnostics.¹⁸⁴ Thrombin aptamer modified AuNPs formed composites with graphene oxide via π–π interactions between ssDNA and the graphene oxide structure. Incubation with thrombin resulted in thrombin–aptamer binding and AuNP release for analysis by spICP-MS. Using the graphene oxide–AuNP platform, an LOD of 4.5 fM was attained for thrombin with a linear range of 10 fM to 100 pM. This method has the potential to be used for a wide variety of biomarkers owing to the versatility of the ssDNA–graphene oxide interaction.

Disease-related microRNAs (miRNAs) are an emerging class of biomarkers for which various ICP-MS based detection methodologies have been described using different amplification strategies and elemental tags.^186–189 Sandwich hybridization on streptavidin coated microtitre plates was used to detect three target miRNAs with three NP probes (Ag, Au and Pt), quantified by ICP-MS.¹⁸⁶ Improved selectivity, sensitivity and dynamic range was achieved in a multiple-metal-NP-tagging method with hybridization chain reaction amplification.¹⁸⁷ Here, three metal NP probes (Ag, Au and Pd) interacted with one target miRNA and ICP-MS response patterns were interrogated to enhance selectivity compared with other miRNA detection methods. Kang et al.¹⁸⁸ presented a multiplexed assay using three MNAzymes (multicomponent nucleic acid enzyme) paired with three target miRNAs and three lanthanide tags with MB-DNA probes. In a different MNAzyme-based approach, cleavage of linker DNA by MNAzyme in the presence of target miRNA resulted in smaller AuNP aggregate size detected by spICP-MS.¹⁸⁹

In addition to solution-based immunoassay platforms, elemental tagging with LA-ICP-MS and LIBS has demonstrated potential for quantitative biomarker imaging in different sample types.^{96,155,157,190} Quantification and localization of dystrophin in muscle tissue and amyloid beta in mouse brain tissue were achieved using Gd-labelled Abs and AuNP conjugates, respectively, with LA-ICP-MS detection.^155,157 Looking to optimise LA-ICP-MS readouts, Vlcnovska et al.¹⁹⁰ compared 4 different Ab-NP conjugates (AgNP, AuNP, CdTe-QD and EuNP) for LA-ICP-MS detection of p53 in dot blot analysis. Optimal results were obtained with AuNP, while AgNP gave large background signals and CdTe-QD and EuNP analysis displayed relatively high non-specific sorption. Most recently a tag-LIBS approach was presented⁹⁶ in which lanthanide (Y)-based photon-up-conversion NP were used to determine HER2 in cell pellets. The elemental signature of Y allowed distinction between HER2 positive and HER2 negative cells, however further work to reduce background signals would be beneficial.

6.5 Multi-element applications

6.5.1 Specimens analysed to investigate metallic implants and biomaterials. Following the considerable interest in work associated with metallic hip and other joint implants, the number of reports referring to these devices has greatly diminished. In our recent Updates, we have described the concerns relating to metallic dental implants. In this review period there is one relevant publication. Nelson et al.¹³¹ identified micro and nanosized particles in inflamed tissue around all samples from patients with titanium implants and in five out of eight ceramic implants. The particles were characterised for size, distribution and chemical speciation using SR-μXRF spectrometry, SR-nanoXRF spectrometry and XANES. The density of particles was estimated to be up to several 10 million per mm³. Titanium was detected as the metal and as TiO₂ and with Nb and a variety of other metal in the titanium implants while the particles from ceramic implants contained Zr with Cr, Fe, Hf, Sr, Y and Zn. This report identifies particle release from ceramic implants for the first time.

Essure is a small spring like device that, fitted onto the fallopian tubes, effectively blocks the eggs from moving into the womb and has been used as a permanent contraceptive. However, adverse effects have been reported by many women, leading to restrictions of the use of this device in several countries. The symptoms include pain, menorrhagia, and broad extrapelvic symptoms (persistent asthenia and fatigue, heart palpitations, tinnitus, pruritus, joint and/or muscularis pain, skin rashes, digestive disorders) which often resolve following removal of the device. It has been suggested that the pathophysiology may involve release of metals. The device is composed of Cr, Fe, Ir, nitinol, Pt, stainless steel and Sn–Ag. Parant et al.¹⁹¹ measured Cr and Ni by ICP-MS in fallopian tissue and peritoneal fluid obtained during removal of the Essure device from symptomatic and asymptomatic subjects. The observed concentration of the metals failed to show any relationship with the length of time between insertion and removal of the device nor with any of the reported symptoms.

6.5.2 Biological fluids and tissues. Elemental analysis in biological samples for diagnosis and prognosis of malignancies continues to be an area of interest. Martynko et al.¹⁹² proposed chemometric processing of macro and trace element concentration profiles in urine as a non-invasive screening protocol for prostate cancer (PCa). The concentrations of 19 elements were measured by ICP-OES and AAS in urine from 34 patients with biopsy confirmed PCa and 32 controls. With accuracies up to 89% for PCa status prediction, authors speculated that this method is superior to routine PSA measurements. While reported diagnostic sensitivity (100%) may be higher than that achieved with PSA, diagnostic specificity was not truly tested in this cohort as controls had no previous history of any prostate disease. Using a similar approach, a heavy metal score was presented to predict outcomes in acute myeloid leukaemia.¹⁹³ Serum samples were analysed by ICP-MS for 10 elements and a risk score derived based on how many element concentrations breached specific cut-off limits. Higher metal scores were associated with reduced survival at 6 months in 2 independent patient populations. Chen et al.⁹² reported on the use of serum-based LIBS, assisted by machine learning, for diagnosis and staging of multiple myeloma in a cohort of 130 subjects (75 myeloma patients and 55 healthy controls). Machine learning achieved over 90% accuracy for sample discrimination through analysis of spectral variations. Chemometrics was also applied to LIBS spectra for discrimination between melanoma and healthy skin tissues in formalin-fixed paraffin-embedded samples.⁹³

Two studies gave conflicting conclusions regarding the role of metal concentrations in the etiology of amyotrophic lateral sclerosis (ALS). In a prospective cohort study, Peters et al.,¹⁹⁴ analysed blood samples from 107 cases and 319 control subjects recruited in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort. Samples were collected before disease onset and the concentrations of As, Cd, Cu, Hg, Mn, Pb, Se and Zn were quantified by ICP-MS revealing potential roles for Cd, Pb and Zn in ALS etiology. Associations were observed between blood Cd (OR 2.04, 95% CI 1.08–3.87) and Pb (OR 1.89, 95% CI 0.97–3.67) with increased risk and Zn (OR 0.50, 95% CI 0.27–0.94) with decreased risk of ALS. In an observational study, the levels of Al, As, Cd, Co, Cu, Fe, Hg, Mg, Mn, Ni, Pb, Pd, Se, V, and Zn were measured in CSF samples from 37 ALS patients and compared to reference values.¹⁹⁵ Results confirmed previously documented associations between Cu and Se and ALS. Higher Cu and Mn concentrations were reported in patients with bulbar onset ALS compared to those with spinal onset, however its noteworthy that only 6 of the 37 patients had spinal onset ALS.

Elemental profiling in biological matrices is gaining popularity as a potential tool to discriminate between different clinical diagnoses and patient groups. Galusha et al.¹⁹⁶ added to previous work describing the elemental composition of bones in long-term PN patients. Bone samples collected at autopsy from 7 PN patients were analysed by ICP-MS/MS for 12 elements and results were compared to those obtained on 18 control bones from hip/knee replacement surgery. Enrichment of Cd, Mn and U, alongside depletion of Ca, Mg and P, were observed in PN bones compared to controls. However, the authors acknowledge that results are only indicative, given the small number of patients with different underlying diseases. Another preliminary study investigated trace element concentrations in post-mortem lung tissue to diagnose seawater drowning (SWD).¹⁹⁷ Measurement of 11 trace elements by ICP-MS showed increased levels of Br and Sr and decreased concentrations of Cd, Cu, Mn, Pb, Se and Zn in the SWD group (n = 29) compared to other causes of death (n = 45).¹⁹⁷ Although other forms of asphyxia were included, the study is limited by an absence of freshwater drowning cases in the control cohort.¹⁹⁷ Nakanishi et al.¹⁶⁴ developed a proof of concept method to simultaneously quantify 38 elements in human skin using LA-ICP-MS. The method was applied to 3 skin samples from healthy subjects, identifying Ba, I, Mo and Sr for the first time in human skin. Future studies are required to determine the applicability of LA-ICP-MS for discrimination of different skin disorders.

This year, novel biological sample types for elemental analysis included teardrop and foetal urine samples. Levels of As, Cd, Cr, Cu, Fe, Mn, Sr, V and Zn were measured by ICP-MS in 173 teardrop fluid samples.¹⁹⁸ Teardrop As, Cu and V concentrations were higher in healthy subjects (n = 129) compared with type 2 diabetics (n = 44), however these observations contradict previously reported trends in other sample types. Teardrop fluid collection presents an attractive alternative to invasive sample techniques, however more detailed information regarding elemental composition of this fluid is required before it can have any true clinical application. Sajnog et al.¹⁹⁹ utilised ICP-DRC-MS with a micronebuliser to quantify 21 elements in 58 low volume (200 μL) foetal urine samples collected during vesico-amniotic shunting. Whilst this is an interesting sample type, comparison of element concentrations with literature values were largely inconclusive and the clinical application is unclear, especially given the invasive sample collection procedure.

Previous reviews have commented on the uncertain clinical utility of toenails as samples to assess metal exposure.¹ This year, a study on 602 preschool children in São Paulo evaluated fingernail Cd and Pb concentrations for biomonitoring of subchronic exposure.²⁰⁰ Nail and blood Cd levels showed no correlation and only weak correlations were observed for nail and blood Pb concentrations (p < 0.05, r values for different variables 0.09–0.26). Data from this study confirm that fingernail Cd and Pb are not reliable biomarkers for exposure.

6.6 Progress for individual elements

6.6.1 Arsenic. This year has seen three publications by Guo et al. using HPLC-HG-AFS to investigate As speciation in clinical samples from patients with acute promyelocytic leukaemia (APL) treated with As trioxide.^107–109 Analysis of plasma samples (n = 305) revealed correlations between trough concentrations of As species and both the As₂O₃ dose and the treatment efficacy,¹⁰⁸ hence suggesting a need for As concentration monitoring in As₂O₃ therapy. A much smaller study (n = 6) focussed on As speciation in CSF samples from APL patients treated with either As₂O₃ or As₂O₃ combined with mannitol infusion.¹⁰⁷ This work elucidated a distribution trend of As species in CSF (As^III, DMA > MMA > As^V) and demonstrated increased As species penetration into the CSF in patients receiving mannitol. Finally, analysis of As species in plasma and RBC from 97 APL patients confirmed highest concentrations of DMA in plasma while high affinity binding of MMA with haemoglobin rendered this the predominant methylated As species in RBC.¹⁰⁹

6.6.2 Boron. An anthropological study by Arriaza et al.²⁰¹ demonstrated boron overexposure in ancient populations of the Atacama desert. Concentrations of B in ancient mummy hair (n = 144) ranged from 1.5 to 4 times average concentrations in contemporary hair and varied with geographic area. The authors speculate that chronic B overexposure in these populations may be attributed to ancient geogenic water contamination from the rivers of Northern Chile.

6.6.3 Cobalt. Ionic Co is a known erythropoietic stimulant, however testing is not currently routine owing to the number of available non-prohibited Co containing compounds and the absence of an agreed threshold value for Co in urine. Knoop et al. reported a proof-of-concept sports drug testing methodology for detection of illegal Co²⁺ supplementation.²⁰² Chromatographic separation in the LC-ICP-MS method enabled simultaneous determination of legitimate organic and atypical iCo in urine, plasma and concentrated erythrocytes from 20 volunteers enrolled in a Co excretion study. Administration of CoCl₂ yielded significantly higher urine Co levels than administration of cyanocobalamin, however the authors acknowledge that more validation work is required before a method like this could be adopted for accurate drug control quantitative analyses.

6.6.4 Copper. Recent work has focussed on developing different approaches for Cu determination in Wilson disease diagnosis. Quarles et al. reported a high throughput LC-ICP-MS method for determination of extractable Cu in serum.²⁰³ The method boasted a sample to sample time of 370 s (250 s run time and 120 s column conditioning). Extractable Cu/bound Cu ratios were able to clearly distinguish between Atp7b^+/− control (6.4 ± 3.6%) and Atp7b^−/− Wilson genotype (38 ± 29% in asymptomatic and 34 ± 22% in symptomatic) rats. Hepatic Cu content is considered the best diagnostic test for Wilson disease owing to pathological Cu accumulation in tissues such as the liver and brain. Hepatic ⁶³Cu quantitation was achieved using LA-ICP-MS in tissues from 11 Atp7b wildtype and 5 Atp7b^−/− mice with ¹³C standardisation.¹⁵⁶ All samples showed a homogeneous distribution of ⁶³Cu, that was elevated in Atp7b^−/− mice compared with Atp7b wildtype mice. The LA-ICP-MS approach overcomes sensitivity and contamination limitations associated with alternative techniques whilst offering higher throughput multi-element analysis. Simultaneous quantitation of Cu, Fe and Zn could be particularly relevant in Wilson patients treated with D-penicillamine and Zn formulations.

In a less typical application, Cu isotope ratios (δ⁶⁵Cu) were investigated to complement CA-125 antigen screening measurements for ovarian cancer diagnosis.²⁰⁴ Serum δ⁶⁵Cu was elevated in 44 ovarian cancer patients compared with 48 healthy donors (p < 0.001) while analysis of ovarian biopsies revealed lower δ⁶⁵Cu in tumour tissue (n = 11) than in healthy ovarian tissue (n = 10) (p < 0.05). The authors propose increased lactate and Cu transporter activity in ovarian tumours to be a causative mechanism for the observed differences in δ⁶⁵Cu and suggest further clinical studies are warranted to define δ⁶⁵Cu thresholds.

6.6.5 Gadolinium. As with previous years, evaluation of tissue retention of Gd following administration with Gd-based contrast agents (GBCAs) continues. Turyanskaya et al. used SR-XRF spectrometry to map Gd in a bone biopsy 8 months post MRI.¹⁶⁶ A specific Gd accumulation pattern was identified and Gd structures were aligned to histological bone structures, however it’s noteworthy that the MRI contrast agent administered to this patient was unknown. Analysis of autopsy hair samples by LA-ICP-MS showed correlations between Gd measurements, Gd dose and time since exposure.²⁰⁵ This gives potential for evaluating Gd exposure in subjects in vivo, however clinical utility may be hindered by inter and intra-individual variation of the Gd peak position along the hair shaft. Despite concerns regarding Gd toxicity via retention following GBCA administration, no clinical features of toxicity were present in 17 Gd exposed patients despite detectable Gd levels in samples of whole blood (median: 0.013 ng mL⁻¹, IQR: LOD to 0.884 ng mL⁻¹), plasma (median: 0.012 ng mL⁻¹, IQR: LOD to 0.046 ng mL⁻¹) and urine (median: 0.304 μg g⁻¹ creatinine, IQR: 0.070–3.702 μg g⁻¹ creatinine) analysed by ICP-MS.²⁰⁶ More work is required to evaluate the clinical significance of the reported Gd concentrations.

6.6.6 Iodine. Pinto et al. investigated I distribution across 14 different brain regions from reference subjects.²⁰⁷ Samples were collected from 52 cadavers without neurological or psychiatric diseases and extracted I was quantified using ICP-MS. Low levels of I were detected with an average concentration across all regions of 0.14 ± 0.13 μg g⁻¹ dw. Consistent with known links between thyroid hormones and brain development, I levels were heterogeneous across different brain regions and higher in areas associated with cognitive function (frontal cortex, caudate nucleus and putamen).

6.6.7 Molybdenum. Isotopic variability of Mo was described for the first time in human urine samples by Zhang et al.²⁰⁸ A MC-ICP-MS method was developed for determination of Mo isotopic variations based on N-benzoyl-N-phenylhydroxylamine purification and separation with double spike mass bias correction. Accuracy and precision, assessed using an IAPSO seawater standard and in-house urine standard respectively, were acceptable, analysis of 31 urine samples from healthy individuals (14 men and 17 women) gave a wide range of Mo concentrations (8.9–122.8 ng g⁻¹). The authors did not clarify the basis (by specific gravity or by creatinine concentration) for the expression of Mo levels in mass units. All samples were enriched with ⁹⁸Mo compared to NIST SRM 3134 (molybdenum standard solution). Potential gender specific differences in δ^98/95Mo were observed, however larger datasets are required to confirm this.

6.6.8 Selenium. Interrogated data from the National Health and Nutrition Examination Survey (NHANES) 1999–2000 identified an association between Se levels and learning disability (LD) in children aged 4–11 years (n = 10 760).²⁰⁹ Serum Se concentrations, measured by AAS, were lower in children diagnosed with LD (107.7 ± 2.7 ng mL⁻¹) than in those without known LD (112.8 ± 1.0 ng mL⁻¹). Although this was a large study, the conclusion is associational and further work is required to determine causation. A further limitation is potential subjectivity in the primary outcome as diagnosis of LD was reported by the children’s parents.

6.6.9 Silicon. Measurement of Si, as a novel in vivo exposure marker for silicosis diagnosis and pathogenesis, was explored using animal and human studies.²¹⁰ Serum Si levels were determined by ICP-MS, however, although instrument parameters were given, it is noteworthy that the publication does not report details of the method validation. Levels of serum Si in rats increased following exposure to Si dust with a peak at 16 h. Human population studies also demonstrated that serum and urine Si increased in subjects working in an iron mine for more than 1 year (n = 918), compared to the control group (n = 120). Furthermore, this increase occurred before elevations of pathogenic cytokines (TGFβ₁ and TNFα) were detected, peaking from 1 to 5 years post exposure.

6.6.10 Strontium. Rapid quantification of radioactive ⁹⁰ Sr in low volume samples was achieved using ID-TE-TIMS.⁷⁷ An LOD of 0.029 fg (0.15 mBq) was achieved from 1 μL samples without any pre-concentration steps in less than 1 h. The proposed ID-TE-TIMS method demonstrated favourable characteristics compared to radiometric analysis (LOD, sample volume, analysis time and precision) and ICP-MS (LOD and sample volume) and was successfully applied to low volume biological samples including tears, eyelashes and saliva.

6.6.11 Titanium. Fu et al.⁶⁶ documented a novel ICP-MS method, using O₂–H₂as the reaction gas to eliminate interferences and quantify ultra-trace levels of Ti in human serum. The method, with LODs ranging between 0.78 and 7.20 ng L⁻¹, depending on the Ti isotope, was used to measure Ti in serum samples from 26 patients with Ti implants and 28 controls. Serum Ti concentrations were significantly higher in the case group compared with controls and ICP-MS results were comparable to those obtained by SF-ICP-MS.

7 Applications: drugs and pharmaceuticals, traditional medicines and supplements

In a critical review, Timerbaev¹⁹ discussed developments with metal-based drugs and diagnostic agents since his similar review in 2014. He referred to monitoring drug–biomolecule interactions, cellular and tissue distribution, intracellular transformation and profiling active metabolites, all of which benefit from the sensitive and precise determination of metal-based drugs particularly ICP-MS-based methodologies, that offer new insights into drug activation and targeting chemistries. The review also included investigations of cellular responses to a drug, and high-resolution and quantitative tissue imaging. The author noted that, despite the considerable development with new drugs, there are few clinical applications and clinical assays. He suggested that in part this reflected that, usually, researchers are not analytical chemists and there are few suitable CRMs. However, he recognised the considerable development with imaging technologies and separation techniques – all of which are included in these annual ASUs.

To some extent the reports of Gawor et al.,²¹¹ de Oliveira et al.⁵⁰ and Vojtek et al.³² support Timerbaev’s observations. Gawor et al.²¹¹ assert that 20–25% of all commercial pharmaceuticals are fluorinated compounds, some F-containing compounds are harmless to health, but some are toxic and there are few available procedures for measuring concentrations of F in biological samples. Gawor et al., therefore developed a method using a HR-continuum source AA spectrometer, with a transversely heated graphite atomiser, to measure the molecular absorption of GaF generated in the gas phase of the atomiser. The procedure was applied to measure F concentrations in the liver of rats treated with the drug cinacalcet (route of administration is not specified). The study included measurement of cinacalcet in liver by HPLC-ESI-MS/MS and fluorinated peptides demonstrated by nano-UHPLC-ESI-MS/MS, with the SwissProt database used to identify the target peptides. While the analytical work is carefully reported, there is little evidence for assay performance apart from the LOD for GaF (6 μg L⁻¹) and there is no discussion of the clinical relevance or application of the procedure.

de Oliveira⁵⁰ developed a method to determine concentrations of thiomersal in vaccines by measurement of iHg by CV-AFS. The concentrations and conditions for oxidative decomposition of thiomersal were investigated and optimised. Of the four oxidising systems tested two, KBr/KBrO₃ and KMnO₄ successfully yielded iHg. The sample and oxidising agent were mixed, left at ambient temperature for 5 min and ascorbic acid (12% m/v) added to remove excess Br₂. A reference method, involving MAD with HNO₃ and H₂O₂, was used to compare the effectiveness of the described procedures. The analytical range, LOD and RSD were determined as 0.5–2 μg L⁻¹, 0.02 μg L⁻¹ and 3.16% at 2.5 μg L⁻¹, respectively. Recoveries of spiked thiomersal were in the range 80.1–1.06% with KBr/KBrO₃ and 92.5–101% with KMnO₄. Results were not significantly different from those given using the microwave digestion reference method.

It was suggested by Vojtek et al.³² that Pd-based drugs are emerging as alternatives to Pt-anticancer drugs and, therefore, they developed and validated an ICP-MS procedure for simple and fast analysis of Pd/Pt-based drugs in 11 biological matrices (adipose tissue, muscle, liver, kidney, spleen, testis, heart, lungs, brain, blood and serum). Samples were digested in closed tubes, using a mixture of HNO₃ and HCl, for 60 min in a 90 °C water bath. Validation results showed an LOD of 0.001 μg L⁻¹, and a linear dynamic range from 0.025 to 10 μg L⁻¹. Repeatability and intermediate precision (RSD) ranges were 0.02–1.9% and 0.52–1.53%, respectively, for both metals. The accuracy, assessed by comparison with a MAD reference method, varied from 83.5 to 105.1%. The authors proposed that, with low reagent and sample consumption, the method is suitable for analysis of novel Pd/Pt-based drugs in pharmaco-toxicokinetic and biodistribution animal studies that involve a large number of multi-organ samples.

8 Applications: foods and beverages

In addition to the discussion of individual papers and common features reported below, Table 3 summarise technical details of interest for selected procedures for the analysis of food and beverages, complementing those reported in Table 1.

Table 3 Food and beverages

Element	Matrix	Technique	Sample preparation/comments	LOD/LOQ^a, μg L^−1	Validation	Reference
a Unless stated otherwise.
As species	Sea cucumbers	ICP-MS	The distribution between lipids (polar and nonpolar) and water-extractable fractions was investigated. Two extraction methods for water-extractable As from sea cucumbers were compared (see Section 3.2) as was the order of extraction of (a) lipid-soluble and (b) water-soluble As species. Speciation was carried out by both AEC (Hamilton PRP-X100 with isocratic elution with 25 mmol L^−1 carbonate buffer) and CEC (Metrosept C6 with three-stage gradient: A H2O; B 50 mmol L^−1 pyridine at pH 2). Detailed results for four extracts of various body parts were given	LODs not given. LOQ 1 μg kg^−1 (DMA) 20 μg kg^−1 (AsSug-382 and MMA)	Total As in CRM NRCC SQID-1 (cuttlefish) and DORM-4 (fish protein)	36
As, Hg (As^III, As^V, MMA, DMA, AsB, Hg^II, MeHg)	Fish (freshwater tilapia, seawater tilapia, swordfish, freshwater bass and flatfish)	HPLC-ICP-MS	Samples (500 mg) were transferred to 15 mL PE centrifuge tubes and 5 mL of 1% (v/v) HCl and 0.1% (m/v) protease XIV were added. The mixture was heated (water bath) at 70 °C for 60 min, cooled, centrifuged (3743 g, 10 min), diluted to 10 mL and filtered (0.22 μm PVDF). The HPLC separation was complete in <4.5 min on a ZORBAX SB-Aq C18 column with gradient elution: (A) 5 mmol L^−1 1-octanesulfonate, 5 mmol L^−1 acetate buffer and 1% (v/v) IPA at pH 4.0; (B) 2 mmol L^−1L-cysteine in 1% (v/v) IPA (pH 4). A DRC with O2 (measurement of As at ^75As^16O^+ at m/z 91) improved the sensitivity for both elements. All species were found in all samples, except for As^III and Hg^II in DORM-3. LODs were significantly lower than any previously reported	0.005–0.007 ng mL^−1 (As) and 0.013–0.015 ng mL^−1 (Hg)	Spike recoveries and CRM NRCC DORM-3 (fish protein)	213
As (total and iAs), I and various (20)	Seaweed	ICP-MS	To 3–5 g of sample, 1–2 mL of H2O and 10 mL of conc HNO3 were added, followed by heating at 75 °C, cooling, dilution with H2O to 20 mL, of which 1 mL was diluted to 10 mL with 2% HNO3 + 0.5% HCl. Minimal information about the analysis was provided, and there are no details of how iAs was determined. 72 samples of European and Asian seaweed of 8 genera were analysed and the results correlated with genus, geographical origin and type of sample. Most elements (Ag, Al, Cd, Co, Cr, Cu, Fe, Hg, Mn Mo, Ni, Pb, Sb, Se, Tl, U, V and Zn) were found in all samples. The elements at the highest concentrations were Al, As, Fe, I, Mn and Zn	0.002 mg kg^−1	No details	225
As (total and H2O-soluble)	Total diet	ICP-DRC-MS with He, niobium IS or HPLC-ICP-MS/MS	In this first German total diet study, 870 pooled samples were prepared between December 2016 and May 2019. The As speciation was performed for rice, rice-based meals and products from the food groups “grains and grain-based products” and “composite dishes” as well as for all samples of the main food group “fish, fish products and seafood”. For total As, MAD of 400 mg sample with HNO3 and H2O2 (210 °C for 25 min), diluted to 25 mL. For H2O-soluble, to 1 g were added 20 mL of 0.02 mol L^−1 TFA acid followed by 6% (v/v) H2O2 to oxidise As^III to As^V, sonicated (15 min at 35 °C) and heated (water bath for 60 min at 95 °C). Sonication was repeated and the suspensions centrifuged. The species in the supernatant were separated by AEC-HPLC (reference to prior work given) and, if peak in void volume, by CEC. Quantifiable concentrations of total As were found in 63% of the pooled samples. Highest values were found in fish, fish products and seafood (mean 1.43 mg kg^−1 (AB was the predominant species)); iAs was the prevalent arsenic species in terrestrial foods (mean 0.02 mg kg^−1)	0.001 mg kg^−1 for moist food and 0.002 mg kg^−1 for dry food	Total As: CRM IRMM ERM-BB422 (fish muscle) and CRM BCR-274 (single cell protein). iAs and DMA: National Reference Laboratory for Metals, Federal Office of Consumer Protection and Food Safety (BVL), Berlin internal RM (rice and soybean)	280
Ca, Cu, Fe, K, Mn, N	Honey	LIBS (spark assisted)	Optimised conditions were Nd:YAG laser, 8 ns Q-switched, operated at 1064 nm and 50 mJ. The DC spark discharge (4300 V) was between two cylindrical tungsten electrodes 4 mm apart and 2 mm above the sample surface. The best classification (100%) based on geographical origin was obtained with spectral preprocessing by smoothing (Savitzky–Golay), generalized least squares weighting (GLSW) and mean centering, together with the k-nearest neighbor and SVM classification algorithms. It was concluded that N plays a fundamental role in discriminating the geographical origin. The results (only 76% prediction accuracy) for same 49 samples, obtained by ICP-MS, are available in an earlier publication from this group	Not given	Not applicable	102
Ca, Mg, Na, K	Honey	LIBS	Optimised conditions were Nd:YAG laser, 5 ns Q-switched, operated at 1064 nm, 1 and 10 Hz, and 60 mJ. Classification based on floral origin by different machine learning algorithms. (PCA, LDA, SVMs and RFCs) was >95% successful. The introduction to the paper is a helpful tutorial	Not given	Not applicable	101
Cd, Cu	Milk powder and infant formula	HR-CS-AAS	Direct solid sampling (0.3–0.9 mg). Palladium modifier added to samples and standards; after drying and ashing (900 °C), Cd was atomised at 1400 °C, then after changing the wavelength, Cu was determined at 2200 °C. Possible homogeneity problems were discussed and evaluated. Cu was found in all four samples, but Cd was only quantified in one	0.6 (Cd), 6 (Cu) μg kg^−1	CRM ERM BD150 (milk powder) and comparison of results with those of a HNO3 MAD procedure (500 mg in 10 mL)	80
Cd	Grain (rice, wheat, corn and soybeans)	ETV-FAAS	A laboratory constructed ETV device applied several heating stages: drying, ashing (in O2-rich atmosphere), and trapping of Cd on kaolin, vaporization under H2 and N2 and transport to spectrometer. Gases were delivered from a gas generation device and H2 was recycled by the formation of water and condensation. Sample was ground to pass a 60-mesh sieve. Analyte was found in all samples, each of which could be analysed in about 10 min, including sample preparation	0.15 ng g^−1 for a 200 mg sample	CRMs GBW(E)100349b (rice) GBW(E)100354b (rice) GBW(E)100350b (rice) GBW(E)100498b (corn). Comparison of results with those obtained with a MAD ETAAS procedure	83
Cl and various (14)	Spinach	WDXRF	About 8 g of powdered sample were pressed into a pellet 37 mm diameter and 3 mm thick. The concentrations of 15 elements (Al, Br, Ca, Cl, Cu, Fe, K, Mg, Mn, Na, P, Rb, S, Sr, Zn) present in three NIST SRMs were determined with the Spectra^plus software program (Quant Express) package (based on a fundamental parameter approach for standard-less analysis), for which full details were provided. The results evaluated by Duncan’s multiple range tests (p < 0.05) showed no significant difference between the measured and the certified values. The procedure was then applied to NIST SRM 1570a (spinach leaves) for which a value of 8000 ± 37 mg kg^−1 for Cl (not certified in 1570a) was obtained, together with results from the other 14 elements that were not significantly different from the certificate values. The researchers recommended that NIST “may include the present measured Cl concentration in the SRM 1570a to the list of certified values”		CRM NIST SRM 1515 (apple leaves), SRM 1573a (tomato leaves), SRM 1547 (peach leaves)	217
Cu, Mo, Zn	Beef	FAAS (Cu, Zn), ETAAS (Mo)	Sample (250 mg) was extracted with 5.0 mL of H2O and 1.5 mL 25% (w/w) TMAH by heating at 85 °C for 60 min; after cooling and centrifuging, the volume was made up to 10 mL (see Section 3.2). The method was applied to 20 rib plate and top sirloin samples provided by the National Institute of Meat of Uruguay and all analytes were found in all samples	0.5 (Cu), 0.6 (Zn) and 0.2 (Mo) mg kg^−1	CRMs ERM-BB184 (bovine muscle) and Embrapa-RM-AgroE3001a (bovine liver) and by comparison of results with those obtained after MAD	44
Cu, Fe, Mn, Zn	Coconut water (Cocos nucifera)	ICP-MS, HPLC-ICP-MS	The total element concentrations (MAD with HNO3 and H2O2) were in the range of 0.3–1, 0.2–2.7, 3–14 and 0.5–2 mg L^−1 for Cu, Fe, Mn, and Zn, respectively. Metal species, identified by SEC-ICP-MS and HILIC-ESI-MS, included Cu complexes with phenylamine and nicotianamine, Fe complexes with citrate and malate, a Mn complex with asparagine, and Zn complexes with citrate and nicotianamine. Sample stability was investigated, and it was concluded that ultra-centrifugation and storage at ≤4 °C preserved speciation for at least 4 months	Not given	Total concentrations by CRM HPS (North Charleston, SC) MFD (mixed food diet)	151
F	Tea (infusions)	MIP-OES	The CaF molecular emission from a commercial instrument (Agilent MIP-AES 4200) operating a N2 plasma was monitored at 530.455 nm. The sample solution was merged (1 + 1) with a 20 g L^−1 Ca solution. The LOD was 5-times higher than that of the MAS method and the OES method was unable to quantify F in three of the 10 tea infusions examined whose concentrations ranged from 3 to 8 mg L^−1	1.1 mg L^−1	Comparison of results with those of HR-CS-MAS (CaF at 606.429 nm). A paired t-test showed no significant differences	87
I	Foods (various)	ICP-MS	Sample (0.1–1.0 g), 4.5 mL of H2O and 1 mL of 25% TMAH was heated at 90 °C for 3 h (see Section 3.2). After cooling, the mixture was diluted to 25 mL of water and centrifuged (3000 rpm for 15 min). Method applied to 123 different samples of milk, infant formula, meat, cooked, food, vegetable, oil, seasoning, cereals and grain, products, bean, vegetable, fruit, sweetener, laver, sea mustard, kelp. The highest concentrations were found in laver, sea mustard and kelp. No I was found in the sweeteners. An ISE method had inadequate LODs	0.013 μg kg^−1	CRMs NIST SRM 1849a (infant/adult nutritional formula) and SRM 1548a (typical diet)	42
Sb	Bottled water	HG-AFS	Stibine, generated from reaction of 2 mL of acidified sample with BH, was trapped in the presence of O2 by a DBD at 75 kV. After 50 s, the Sb was released into a carrier containing H2 at 6.8 kV and transported to the AFS instrument (see Section 4.4). Analyte was found in three real samples, but in a leaching study from bottles made of different materials (polypropylene, PET, polycarbonate, and HDPE), Sb was found only in the water stored in PET	9 pg (in a 2 mL sample)	Spike recoveries and CRM NRCCRM BWZ6657-2016 (Sb in 1% HCl) and GSB 07-1376-2001	115
Se (inorganic species)	Rice, tea and garlic (selenium enriched)	ICP-MS	Alkaline sample extraction (200 mg and final volume 10 mL) solubilised only inorganic Se (see Section 3.2). Total Se was determined after MAD (500 mg) with HNO3 and H2O2; final volume 25 mL. Inorganic species were separated by AEC on a Hamilton PRP X-100 column with 5-step gradient elution, using 40 mmol L^−1 (NH4)2HPO4, pH 5.0, and 60 mmol L^−1 (NH4)2HPO4, pH 6.0. The chromatography was extensively optimised for both inorganic and organic Se species. In 11 real samples, Se^IV was detected in all, but Se^VI was detected in only one. Inorganic Se ranged from 0.6 to 18% of total Se	2.5 (Se^IV) 5 μg kg^−1 (Se^VI)	Spike recoveries	45
Various (15)	Fats and oils	ICP-OES	Samples were dissolved in (by heating to 80 °C), or mixed with, xylene (1 + 19) (see Section 3.2). To avoid extinguishing the plasma, a 5 μL sample was introduced via a conventional nebuliser and an in house-made HTISIS. The only information given was that it contained a 9 mL inner volume glass single-pass spray chamber with a copper heating coil wound around whose temperature was controlled at 400 °C. The method was applied to the determination of Al, Ba, Ca, Cd, Cr, Cu, Fe, Mg, Mn, Mo, Ni, Pb, Si, Ti and V in 11 solid fat and oil samples. Most elements were detected in all samples, though Ni was only detected in one sample	Ranged from 0.2 (Mg, Mn, Ti) to 18 (Pb) μg kg^−1	Spike recoveries and by comparison of the results obtained by MAD (0.5 g sample, HNO3 20 g final mass) and ICP-MS	39
Various (16)	Swordfish	Thermal desorption AAS, ICP-OES, ICP-MS	Lateral slices (8 per fish) were freeze dried and homogenised. Hg was determined by AAS, directly on the solid. For the other elements, MAD of 200 mg with conc. HNO3 (180 °C 10 min), cooled, 0.5 mL of H2O2 added and reheated. Digested samples were evaporated to 0.5 mL and diluted to 25 mL, prior to determination by ICP-OES (Ca, Fe, K, Mg and Na) or ICP-MS (As, Cd, Co, Cr, Cu, Mn, Ni, Pb, Se and Zn). Dark muscle was notably enriched in Cu and Fe and had enhanced concentrations of Cd, Mn, Se and Zn. The rest (all detected) were equally distributed between dark and white muscle	LOQs: 0.01 ng (Hg), 10 mg kg^−1 (Ca, Fe, K, Mg and Na), 0.004–1 mg kg^−1 (As, Cd, Co, Cr, Cu, Mn, Ni, Pb, Se and Zn)	CRM NRCC TORT-2, (lobster hepatopancreas), IAEA-436 (tuna homogenate)	281
Various (40)	Bees and beehive products (beeswax, honey, pollen, propolis and royal jelly)	ICP-OES and ICP-MS	Several sample preparation methods were evaluated (see Section 3.2). A water-bath extraction (95 °C for 30 min with 200 mg sample, 1 mL of 67% HNO3 and 0.5 mL of 30% H2O2, leading to a final volume of 10 or 20 mL) was found satisfactory. Results for the determination of 16 elements (Al, As, B, Ba, Be, Bi, Ca, Cd, Ce, Co, Cr, Cs, Cu, Fe, Ga and K) in several commercial products were given. Only in pollen was As detected, and only Ce, Cr, Cs and Fe were detected in beeswax	ICP-MS only. Ranged from single-digit ng kg^−1 (Nb) to single-digit mg kg^−1 (Ca, K, P, Si)	Spike recoveries from each sample material and CRM NIST SRM 1515 (apple leaves)	51

---

8.1 Progress for individual elements

8.1.1 Arsenic. There continues to be a significant number of papers relating to As determination and As speciation. Ahmad et al.²¹² synthesised CdS nanoflowers to extract As from rice, waters, mushroom and seeds. Due to the small particle size of the CdS nanoflowers direct slurry analysis was possible and LODs of 0.5 ± 0.2 and 0.8 ± 0.2 ng L⁻¹ for As^III and As^V were reported. Good agreement was obtained with the expected values of RMs and spiked samples for both As^III and As^V species.

Different parts of various sea cucumber species from different locations underwent As speciation testing using a variety of extractions including extraction with increasing/decreasing polarity sequence and addition of TFA to water. In this multivariable test, the key conclusions obtained were that TFA extracted more As species than water alone and that the composition of As in different sea cucumber species varied greatly, with Apostichopus species showing a prevalence for AB and the Cucumaria species having higher iAs concentrations.³⁶

A combined As and Hg speciation method using HPLC-ICP-DRC-MS was shown to be effective for analysis of fish samples.²¹³ Optimised sample conditions were established for a MAE using a 1% HCl–0.1% protease mixture and extracts underwent HPLC separation using a ZORBAX SB Aq C18 column, followed by ICP-MS detection. Key As species (As^III, As^V, AB, DMA, MMA) and Hg species (iHg and MeHg) were shown to have good separation. The LODs ranged from 0.005 to 0.007 ng L⁻¹ for As species and from 0.013 to 0.015 ng L⁻¹ for Hg species. Analysis of RMs and spiked samples yielded results in the range 93–105% of the expected values. Good extraction recoveries between 95 and 97% indicate that this could be an effective simplified combined method for As and Hg speciation.

The effect of cooking foods (rice, garlic, asparagus) in As^V contaminated water was studied by Clemente et al. ²¹⁴ XANES was utilised to identify As species in raw and cooked food as well as samples that had undergone bioavailability extraction. The study showed that the presence of As^III compounds increased when foods were boiled and the type of food had an impact on the extent of reduction of As^V to As^III, potentially linked to the antioxidant properties of those foods: rice (12%) < garlic (13%) < asparagus (55%).

Lin et al.³⁷ developed a MAE method followed by IC-ICP-MS for As speciation including AB, As^III, As^V, DMA, MMA and arsenosugars. Freeze dried samples were digested in 20 mmol L⁻¹ HNO₃. A spiked blank and spiked samples underwent the same digestion treatment to assess the efficacy of the method that gave recoveries >92%. The method was shown to be universally suitable for a variety of seafood (seaweed, fish, shellfish and shrimp) samples, with potential to provide a universal methodology for As speciation.

8.1.2 Halogens. As in previous years iodine had featured as the subject of several studies, but interestingly there has also been a number of papers on the other halogen elements, Br, Cl and F. Hwang et al.⁴² compared ion selective electrode (ISE) with acid or TMAH digestion followed by ICP-MS to measure I. Initially CRMs underwent all three procedures. The first (ISE) was found to be too insensitive for general foods but useful for seaweeds, where levels are typically higher. Acid digestion was rejected due to poor reproducibility. Digestion with TMAH was selected to perform a broad study covering many different food groups. Most food types were found to have levels of I in the μg kg⁻¹ range (reported by the authors as μg 100 g⁻¹), and, as previously reported, I concentrations in seaweeds were typically in the mg kg⁻¹ range (given as mg 100 g⁻¹) and g kg⁻¹ for kelp. Different seaweed species have significantly different I concentrations, with lavers having a typical concentration of 23 mg kg⁻¹, sea mustards containing around 100 mg kg⁻¹ and kelps having concentrations of around 4000 mg kg⁻¹. It was concluded that ISE may be useful for quantification of I in seaweeds but for general I analysis in foods TMAH/ICP-MS is an acceptable approach.

Interestingly, in a study conducted by Milinovic et al.²¹⁵acidic digestion in aqua-regia of lyophilised samples, followed by ICP-OES, was shown to give acceptable results for I determination. The author proposed that the generation of the non-volatile IO₃⁻ and iodo-chlorine species allowed for successful determination of I. Again the study demonstrated that different seaweed species had varying concentrations of I, ranging from 33 to 391 μg g⁻¹ and the author also noted the importance of the differences when looking at I dietary intakes.

Levels of I with respect to dietary intake were investigated in various wholefood macroalgae, and in food products and dietary supplements, containing macroalgae,²¹⁶ employing TMAH/protease digestion followed by ICP-MS analysis. The levels of I in seaweeds varied considerably, as previously discussed, and so did the I concentration in the food products analysed, likely due to the use of different seaweed species in the ingredients. The authors also showed that there were considerable concerns about the reliability of labelling of I content in food supplements, with label claims of I being both exaggerated and under reported across the samples studied. The authors concluded that control of I in the diet would be tricky due to the risk of over consumption from certain foodstuffs or insufficient intake due to poor labelling of supplements.

In an unusual application, WDXRF spectrometry was used to measure multiple elements in a range of CRMs with a particular focus on Cl. Typically XRF spectrometry is not sufficiently sensitive to perform Cl determination, but the WDXRF results showed good correlation with the reference values. The authors²¹⁷ proposed that this may be a suitable technique for determining reference values for Cl in CRMs.

Microwave induced combustion combined with ICP-MS was investigated to measure Br, Cl and I content in granolas.⁵³ The MIC methodology was optimised for sample size, combustion conditions and absorbing solution. Once optimised conditions were found, the method was evaluated using CRMs of dried plant materials, as no granola CRMs were available. An agreement of >94% with reference values was obtained for all elements. Six different granola samples and their corresponding spiked aliquots were also analysed, obtaining spike recoveries >95%. The LOQs were 0.025, 25 and 0.002 μg g⁻¹, for Br, Cl and I, respectively. Chlorine was found to be present at between 322 and 896 μg g⁻¹, Br was found consistent, being between 0.618 and 0.980 μg g⁻¹, whereas a high variation in I concentrations was observed, from below the LOQ up to 0.181 μg g⁻¹. The authors suggested that the low carbon levels allow for accurate determination of halogen elements in granola.

Fluorine in teas was measured by MIP-OES by Akhdhar et al. ⁸⁷ Direct measurement of F is not possible by MIP-OES due the energy of the plasma being insufficient to produce measurable levels of excited F atoms. To overcome this problem, infusions of tea were mixed with CaCl₂ at a ratio of 1 + 1 to generate CaF and the emission of this strongly bound molecule at a wavelength of 530.455 nm was used for measurement. This approach was shown to be suitable for measurement of F in tea infusions with a LOQ of 1.1 mg L⁻¹ as results comparable to the reference method (HR-GF molecular absorption spectrometry) were obtained.

8.2 Single and multi-element applications in food and beverages

8.2.1 Dietary intake studies. Sarvan and co-workers²¹⁸ performed intake studies for Hg in the German diet. They measured total Hg and MeHg in fish and mushrooms as these are recognised common sources of Hg. Methylmercury was extracted using nitrogen distillation and measured using ICP-MS. Total Hg was measured using a direct mercury analyser. By evaluating dietary intake with respect to MeHg rather than total Hg, a more accurate assessment can be carried out as the approach focuses on the highly toxic MeHg form rather than on the total Hg content. The study showed that Hg was commonly present in fish as MeHg, representing from 65 to 118% of the total, whereas in mushrooms, the levels of MeHg were mostly below the LOQ (0.01 mg kg⁻¹), and where the level was measured it accounted for only 2% of the total Hg content. The study showed that in most cases, population exposure to excessive MeHg was unlikely, with the exception of the groups aged from 14 to 24 years old with a high fish consumption, due to higher consumption of tuna in these age groups.

Quantitative correlation between iAs intake and urinary excretion of iAs and its metabolite MMA was established in a duplicate diet study with 150 participants.²¹⁹ The geometric mean of iAs intake was determined as 0.349 μg per kg per day and that of excreted As (iAs and MMA) was 5.20 ng mL⁻¹ (corrected for specific gravity) or 4.05 ng mg⁻¹ creatinine. Correlation coefficients between iAs intake and excreted As were 0.544 for creatinine corrected values and 0.458 for urinary concentrations corrected for specific gravity. The authors suggest that the observed relation log₁₀(daily intake) = 0.451 × log₁₀(excretion) + 0.814, using creatinine corrected values, could be used to estimate iAs dietary intake from measurements of urinary As in epidemiological studies.

8.2.2 Human milk and infant formula. Differences in essential and toxic element concentration were established in human milk concentrate by Oliveira et al. ²²⁰ The team measured elemental levels in human milk and human milk concentrate samples to ascertain the potential risks associated with providing concentrate human milk to pre-term babies. Concentrate samples were prepared by lyophilising 50 mL of milk and reconstituting them with a 75 mL milk aliquot. Samples underwent acid digestion and analysis using ICP-MS to measure the levels of Al, As, Cd, Cr, Fe, Hg, Mn, Ni, Pb, Se, Sn, and Tl. The concentration process only showed an increase in concentration of Mn (0.80 μg L⁻¹) and Se (6.47 μg L⁻¹). The study concluded that human milk concentrate could be safely given to pre-term infants, due to general low levels of toxic elements and increase in the nutritional elements Mn and Se.

A review of analytical approaches to determine Ca in milk and formula products was carried out by Masotti et al. ²²¹ In this in-depth review multi and single element methods were assessed. It was concluded that there is a shift towards ICP-OES from the well-established FAAS due to reduction in costs and multi-element capability. The authors also noted the usefulness of XRF as an in situ technique for use by farmers and manufacturers.

8.2.3 Dairy products. Chromium speciation was carried out in cow’s milk using fibrous g-C ₃ N ₄ nanocomposite solid-phase microextraction (FGCTNCs).²²² Liquid and dried milk samples underwent MAD with H₂O₂ + HNO₃ for total Cr measurements by ICP-MS. Fresh milk samples were also assessed for digestible and residual fractions by incubating with simulated gastric juice and filtering to give the digestible and residual aliquots. These aliquots were further split and acid digested for measurement of total Cr or treated with the FGTNCs for speciation. Chromium species were extracted at different pH values, Cr^III at pH 8.0 and Cr^VI at pH 3.0. The method was verified by analysis of a milk powder CRM, with the sum of the species yielding 104% of the certified value for the total Cr content, and recoveries from spiked samples, giving values between 91 and 102%. This study demonstrated it was a suitable technique for the estimation of the bioavailability of Cr species in milk.

Gómez-Nieto et al.⁸⁰ utilised the difference in atomisation temperature between Cd (900 °C) and Cu (2100 °C) to develop a direct solid sample method for their sequential determination in milk powders and infant formula by HR-CS-GFAAS, achieving LOQs of 2.1 μg kg⁻¹, for Cd, and 1.7 mg kg⁻¹, for Cu. Optimisation was carried out using a skimmed milk powder CRM (ERM-BD150), that was also prepared by MAD for comparison. The procedure was shown to be a robust method for the direct analysis of these elements, achieving good agreement with the expected values when applied to a range of different RMs.

In a large study,²²³ 11 commercially available infant formulas available in Italy were screened for 40 elements, for the purpose of assessing infants’ intake and possible related health risks. Among these, Al, As, B, Be, Cr, Nb, Sb, Te, V, W and Zr, were found to have levels below the LOQ in over 30% of the samples and were therefore excluded from further statistical analysis. In particular, Al was not detected in 73% of the samples. Descriptive statistics (arithmetic mean, SD, median and IQR) were calculated for the remaining elements. The observed concentrations varied from more than 2000 μg g⁻¹ for Ca and K to less than 1 ng g⁻¹ for Tl and U. In particular, the levels of Cd, Ni, Pb, and Zn were within the ranges: 0.001–0.006 μg g⁻¹, 0.040–0.100 μg g⁻¹, 0.0006–0.0026 μg g⁻¹ and 10.2–20.2 μg g⁻¹, respectively, in agreement with other studies carried out within the EU. The authors assessed the dietary intake of Cd, Mn, Ni, Pb and Zn and the related potential risks in comparison with provisional tolerable daily or weekly intake indications from EFSA, FAO/WHO and other advisory committees. They found the intake of these elements to be always lower than the stated tolerable limits, but advised that the contribution of the water used for the reconstitution of the powdered formulas should also be considered.

8.2.4 Fats and oils. Corn oil, vegetable oil and animal fats were analysed with a multi-element dilute and shoot ICP-OES method.³⁹ Samples were diluted 1 + 19 in xylene before being introduced into the ICP-OES. A standard cyclonic sample introduction system and a high temperature torch integrated sample introduction system (HTISIS) were compared, along with a MAD procedure. With the dilution in xylene, both sample introduction methods showed a marked improvement in LOQs for most elements, as compared to MAD, with HTISIS exhibiting lower LOQs for all elements compared to cyclonic introduction. The HTISIS also gave significant improvement to the spiked recoveries, that ranged from 20 to 80% with the cyclonic sample introduction system and from 90 to 110% with the HTISIS for all elements in all oil types. The authors proposed this approach as a suitable, easy method for oil analysis.

8.2.5 Cereals. Shi et al.⁵⁹ investigated direct nebulisation of solid particles for the determination of the As, Cd, Hg and Pb content in wheat and rice by ICP-MS. These two grains form the basis of the majority of diets across the globe, so ensuring that the levels of these toxic elements are very low is important. In the method optimisation, CRMs were milled for increasing lengths of time and particle size distributions were measured by laser diffraction and emission SEM. The samples were then analysed and the results compared with reference values. Unsurprisingly, it was found that smaller particle size obtained by longer milling obtained the best accuracies. Samples were milled to ∼1 μm, then were dispersed in 0.5% polyethylene imine and the slurry analysed by ICP-MS, calibrated using aqueous standards. A comparison was carried out on the same samples analysed after MAD. With the optimised method, the analysis of two NIST SRMs (1568b rice flour and 1567b wheat flour) yielded results in good agreement with the reference values for all elements (94–107%). Precision ranged from 0.4 to 6.5% and LOQs varied from 1.1 ng g⁻¹ (Hg) to 3.5 ng g⁻¹ (As). A further four wheat and five rice RMs were assessed by the method and good correlation was also seen with their reference values or the results obtained with MAD.

8.2.6 Vegetables, fruit, mushrooms and nuts. The concentrations of Ba, Cd, Co, Cr, Cu, Fe, K, Li, Mg, Mn, Mo, Pb and Zn were measured in raw and cooked creole beans.²²⁴ Optimised acid digestion conditions were used to prepare samples for MIP-OES analysis. In addition, simulated gastric, saliva and intestinal juices were applied to extract Ba, Cu, Fe, Mn, Pb and Zn from raw and cooked beans in a bioaccessibility study. The MIP-OES method was validated using spiked bean samples at three different concentrations, that gave recoveries ranging from 81% to 113%, with RSDs <10%. In addition, two CRMs (CRM-Agro FT_012016 tomato leaves and NIST SRM 1846 infant formula) and digestions of mixtures of the CRMs with the beans were also successfully analysed, with accuracies within the same range. For the studied elements, a high variability was seen between different batches of beans. The cooked beans had lower concentrations of the target elements due to leaching of elements from maceration during cooking. The bioaccessible fraction of Ba, Cu, Fe, Mn, Pb and Zn in three different types of creole beans was significantly lower than their respective total concentrations, measured in both the raw and cooked beans. The authors pointed out that the cooking process may significantly influence the mineral content of beans and that investigating this process may provide important nutritional information, since beans are consumed mostly after cooking.

Seaweeds, commonly consumed in Asian cuisines, are finding popularity in western diets. In a study covering 72 different seaweeds from 8 genera, from Europe and Asia, samples were assessed for 20 elements, including As, iAs and I.²²⁵ As there are currently no guidelines for seaweed consumption, studies such as these offer useful information to form risk assessments. The majority of elements were found at low levels but the authors highlighted some interesting trends. The genera of the seaweed has a huge impact on the concentration of I with significant levels of I measured in brown seaweeds (Phaeophyta) and nearly 7000 mg kg⁻¹ determined in one kelp sample. Red seaweeds were found to typically have the highest toxic element contents. The author also noted that one Korean seaweed contained a total As content of 44 mg kg⁻¹ of which 7.1 mg kg⁻¹ was iAs.

8.2.7 Fish and seafood. Grasso et al.^135,136 published 2 papers investigating TiO₂and Ag nanoparticles in seafood. Multiple samples of canned tuna, mackerel, anchovies and clams were obtained from the local market and digested under alkaline conditions using TMAH, to preserve the integrity of the particles. Analysis was carried out by spICP-MS. The total concentrations of Ti and Ag were also assessed in the same samples, using acid digestion and ICP-MS analysis. Across all samples, AgNPs were found to be 26–28 nm in size, with a concentration between 0.44 × 10⁷ and 2.28 × 10⁷ particles g⁻¹. A greater variability was observed for TiO₂, with typical sizes ranging from 73 to 113 nm and concentrations within the range 1.47 × 10⁷ to 1.05 × 10⁸ particles g⁻¹. Since AgNPs and TiO₂ are not typically used in the manufacture of canned fish products and are not expected in the packaging, the authors concluded that the levels seen are likely to be from the environment and bio-accumulating in all fish types. AgNPs were also the subject of a study carried out by Taboada-Lopez.¹⁴⁷ Bivalve molluscs underwent enzymatic treatment without ultrasound and AF4-UV-ICP-MS analysis for the quantitation of AgNPs. The samples were also digested and analysed by ICP-MS for the total Ag content. The ratio of NP to ionic Ag was found to be higher than previously reported: 4% (oyster), 6% (frozen scallop), 39% (fresh scallop) and 56% (clam), which the authors attributed to the softer sample digestion methods employed.

8.2.8 Drinking water and non-alcoholic beverages. Field flow fractionation hyphenated to Multi Angle Light Scattering and spICP-MS (AF4-MALS-ICP-MS) was applied by Li et al.¹³⁷ to study TiO₂ nanoparticles. Standards for TiO₂ NPs are problematic due to particle aggregation, but, using utilising multiple detection techniques, polystyrene NPs can be used to give an accurate calibration and will also increase the detection range. Comparison of the two detection methods showed good agreements in the patterns following separation by AF4. Multi angle light scattering offers a powerful technique for particle sizing, whereas the ICP-MS offers elemental profiling to ensure that the particles being measured are specifically TiO₂. The method was optimised and used to assess TiO₂ NPs in coffee creamers and powdered fruit drinks. It could be used as a cost effective alternative to methods such as TEM.

A range of herbal and fruit teas were used to assess the presence and bioavailability of a large number of elements (Ag, Al, As, Ba, Be, Ca, Cd, Co, Cr, Cs, Cu, Fe, Ga, Hg, K, Li, Mg, Mn, Mo, Ni, P, Pb, Pt, Rb, Sb, Se, Si, Sn, Sr, Ti, Tl, V and Zn). The sample preparation involved full digestion, infusion as teas and preparation of the bioavailable fraction in simulated gastric and intestinal conditions. Infused teas showed variable leaching efficiencies in excess of 50%, but also some elements were not detected in some extractions. These results were consistent with those reported in previous studies. The bioavailability study was carried out on the infused tea samples and showed high extraction of Ca, K, Mg, Mn, P, Rb and Si whereas the toxic elements As, Cd, Hg and Pb were not detected. This allowed the authors²²⁶ to conclude that drinking teas provide a useful source of nutritional elements with a low risk of exposure to toxic elements.

Antimony in bottled water was analysed using AFS with dielectric barrier discharge as a pre-concentration technique by Liu and co-workers.¹¹⁵ An LOD of 5 pg mL⁻¹ equating to a seven-fold concentration factor was obtained, together with good agreement with CRMs’ values (95–104%) and RSDs <8%. The dielectric barrier was charged with 7.5 kV under a gas flow of 600 mL min⁻¹ Ar mixed with 80 mL min⁻¹ air for 50 s. The sample was then discharged at 6.8 kV under a flow of 120 mL min⁻¹ of H₂. The developed method was applied to assess the Sb levels in water kept in PET bottles over a range of temperatures and time intervals. Between 25 and 60 °C, for up to 72 h, the level of Sb did not rise above 2.5 μg L⁻¹, but at 80 °C the level of Sb increased rapidly, reaching levels of 10 μg L⁻¹ by 72 h.

Coconut water is a useful source of minerals such as Cu, Fe, Mn and Zn and when consumed as organic complexes, these elements are more bioaccessible. To study the properties of this beverage, Alchoubassi et al.¹⁵¹ applied SEC-ICP-MS and HILIC alongside total elemental analysis by ICP-MS. The results showed that each element had a different predominant organic salt, often as a significant percentage of the total elemental content: Cu phenylalanine (92%), Fe malate/citrate (87%), Mn aspartate (47%) and Zn nicotianamine (62%). The elements present in these forms demonstrate the suitability of coconut water as a source of these nutritional elements.

8.2.9 Authenticity. Over the last year numerous journal articles have been published covering chemometrics combined with spectroscopic techniques to establish food origin, authenticity and vintage. A variety of analytical methodologies and statistical techniques have been employed. Cerutti²²⁷ used HTSIS and ICP-MS to perform direct analysis of predominant REEs in wines. The fingerprint of Dy, Er, Eu, Gd, Ho, Nd, Pr, Sm, Tb, Tm and Yb allowed for discrimination of wines from twelve regions. Barium, Co, Cr, Mn, Ni Pb and V allowed for a second fingerprint to further discriminate wines from another five regions.

Pistachio nuts were analysed using ICP-OES. Applying PCA to the data allowed for discrimination of different cultivars with a sensitivity of 98.0% and selectivity of 99.6%. The authors²²⁸ noted the key indicator elements were Ca, Cu, Fe, Mg, Sr and Zn. Pistachios were also the subject of a study carried out by Kalogiouri et al.,²²⁹ in which REEs were used to identify the geographical origin of samples from Greece and Turkey. The quantification results were further analysed using permutational analysis of variance, nonmetric multidimensional scaling, PCA and hierarchical clustering to investigate the similarities between the pistachios. It was found that a simplified comparison of La and Sm concentrations following a decision tree model could predict the origin of pistachios.

Elements such as Br and Cl, as well as Ca, Cr, Cu, Fe, K, Mn, Ni, P, Rb, S, Sr and Zn were determined by EDXRF spectrometry and the information was used by Fiamegos et al.²³⁰to define a profile for Pimenton de la Vera, a paprika with PDO status. Two statistical methods, namely PCA and PLS-DA, were used for modelling. The best results were obtained using PLS-DA, with a sensitivity of 100%, a specificity of 91% and an accuracy of 96%. It was shown that Fe, Mn and Sr were key indicators for Pimenton de la Vera samples.

Foschi et al.²³¹ applied PLS-DA and LDA to investigate the geographical origin of lentils. Eighty-nine samples from three regions were analysed by ICP-OES for 15 elements. In prediction, 100.0% (LDA) and 92.6% (PLS-DA) of 27 external samples were correctly assigned.

Mangoes from Europe, South America and Africa were profiled using ICP-MS and stable isotope analysis. It was shown that ¹⁵N, ¹⁸O, ³⁴S along with As, Ba, Cs, Cd, Pb and Se, were the key indicators to distinguish Spanish mangoes from those from other countries. PLS-DA was found to give correct classification of 97.6% for the elemental analysis and 100% for the isotope analysis. Muñoz-Redondo et al.²³² combined both methods to improve their performance for assessing the origin of mangoes, particularly those from Spain with a PGI status.

Blueberries were analyses for their B, Ca, Cu, Fe, Mg, Mn, Na, P and Zn content by ICP-OES by Kuang and team.²³³ The data was then processed using LDA, multilayer perception neural network and SVM. The last two techniques were shown to be the most successful chemometric methods with a 95% success rate in determining regional origin, based on a dataset of 148 blueberry samples from three regions.

Sparkling wines from Brazil, Argentina, France and Spain were analysed by ICP-OES. Twelve elements (Al, B, Ba, Ca, Cu, Fe, K, Li, Mg, Mn, Na and Sr) underwent chemometric analysis by LDA, PCA and logical regression. Boron, K and Na provided 94% accurate classification of country of origin. Calcium, K and Na allowed for discrimination between South American and European wines, Mn allowed to distinguish between Brazilian and Argentinian wines and the ratio B/K clearly separated Spanish and French wines. The authors²³⁴ surmised these simplified comparisons could prove useful to prevent fraudulent activities in the sparkling wine industry.

Walnut samples from 10 countries and across three harvests were analysed by ICP-MS.²³⁵ Forty-seven elements were selected for the chemometric study, but the authors noted that the levels of REEs were too low to be considered for this study. Aluminium, Ba, Co, Cu, Fe, Mo, Ni and Sr were found to be the most significant indicators. With a predictive accuracy of 73% when assessing on a global level, at a regional level this approach achieved improved accuracies of 91%, 77%, and 94% for France, Germany, and Italy, respectively.

Basmati rice is a highly prized commodity. Demonstrating rice authenticity is important to prevent fraud both for the consumer and for the import duties, as basmati rice is exempt when imported into the EU. To this aim, Arif et al.²³⁶ applied ICP-MS together with chemometrics, using data-driven SIMCA. Thirty-five elements were included in the model for the geographical discrimination of rice grown in the Basmati region of Pakistan, that achieved sensitivity and specificity of 100% and 98%, respectively. Rice was also the subject of two papers published by Pérez-Rodríguez and co-workers. In the first,²³⁷ ICP-OES and PCA/LDA were used to investigate adulteration of rice flours. Pure and adulterated flour samples were differentiated with overall accuracy and specificity in the ranges 72–88% and 80–100%, respectively, using models based on mineral features. Classification performance was improved to >91% using PCA-based data fusion and allowed to identify adulterated rice flour samples with a 100% success rate. In the second work, alternative analytical methods were used to assess the botanic origins of rice.¹⁰⁴ Spark discharge-LIBS was applied to measure the content of C, Ca, Fe, Mg, N and Na in 72 samples from four rice varieties. This information was then analysed using SVM, with the algorithm parameters optimised using a central composite design to find the best classification performance. These models achieved 96.4% correct predictions in test samples and showed sensitivities and specificities between 92 to 100%.

Pu’ur teas are highly valued, and aged teas are highly sought after. Liu et al.²³⁸ employed HCA, PCA, PLS-DA, BP-ANN and LDA to build authentication models for predicting the Pu’er tea with different production years, from data for 86 elements measured by ICP-MS and ICP-OES. The performance of the different analytical models gave varying results, with the supervised models (PLS-DA, BP-ANN, and LDA) giving preferential performance over the unsupervised models of HCA and LDA. The key-markers at identifying the vintage of tea were ⁵⁵Mn, ⁶⁸Zn and ²⁰³Tl, and the BP-ANN algorithm gave the best performance, with 100% accuracy in the prediction of the year of vintage. Liu and co-workers²³⁹ used similar techniques to look at different blends of Chongqing tuo teas. The predictions from PLS-DA and LDA were better than those of HCA and PCA, with LDA achieving precision of 100%. Cadmium, Co, Cs, Mo, P and Rb were shown to be the most important markers.

8.2.10 Bee products. This year has also seen many papers relating to honey and bee products, no doubt fuelled by the increased interest in bee welfare and the health benefits of honeys. Two papers were published describing the determination of the geographical region of origin for honeys. In the first study, Fechner et al.¹⁰² reported that applying SVM algorithms to the results of SD-LIBS analysis, the PDO status of honeys from Argentina could be determined with 100% effectiveness. In addition, PLS-DA showed that the key indicator elements were Ca, Cu, Fe, K and Mn. Ghidotti et al.¹²² investigated the potential of EDXRF spectrometry, coupled with chemometric techniques, for identification of Spanish honeys with PDO against others from Spain and different countries. The concentrations of Ca, Cl, Cu, Fe, K, Mn, Ni, P, Rb and Zn were measured and underwent PLS-DA data manipulation. The method proved very good at discriminating dark honeys even with a low number of learning samples, but lighter honeys were poorly defined and required a greater number of learning samples.

Honeys of a particular botanical origin are particularly desirable. Stefas and co-workers¹⁰¹ applied LIBS to discriminate honeys from different plant origins, including, almond, carob and orange. Unsurprisingly, analysis by LIBS gave strong spectral lines for C, H and O for all honeys, so the lines for inorganic components were used. Calcium, K, Mg and Na were shown to be measurable in all samples. The data obtained was then processed using PCA, LDA SVM and RFCs. All four techniques achieved accuracies better than 95%, showing that LIBS may be a simple and useful method for determining the botanic origin of honey.

Dispersive liquid–liquid microextraction followed by GFAAS was used for the analysis of Pb in honey products, including honey, mead, honey vinegar and honey beer. As Fe can cause interference in the analysis of Pb by GFAAS, Fiorentini et al.²⁴⁰ developed an extraction method using alternative magnetic ionic liquids. Samples were prepared with thermal digestion in HNO₃ and H₂O₂ and extracted into the dispersant. The optimal dispersant was found to be 50% [P_6,6,6,14]₂MnCl₄ in CHCl₃. This method was shown to have an extraction efficiency of 97% with an LOD of 3 ng L⁻¹. Spiked samples of all matrices gave recoveries from 95 to 100% demonstrating the suitability of the method.

Approaches for digestion of samples of bees and beehive products were assessed for determining 40 elements by ICP-MS and ICP-OES. Both MAD and open vessel digestion methods were investigated with various acid mixtures. The authors⁵¹ noted that CRMs were not widely available for the materials under study, so CRMs based on various other matrices as well as spiked samples were used to assess method performance. For the two preparation techniques, similar spiked recoveries were obtained (from 80% to 130%). It was noted that open vessel digestion gave acceptable results and offered a simpler approach than MAD for sample preparation. In most cases HNO₃ and H₂O₂ proved sufficient for digestion, but for some types, aqua-regia may be required to give a better determination.

9 Abbreviations

AA	atomic absorption
AAS	atomic absorption spectrometry
AB	arsenobetaine
ac	alternating current
ACN	acetonitrile
AEC	anion-exchange chromatography
AES	atomic emission spectrometry
AF4	asymmetric flow
AF4-FFF-ICP-MS	AF4 field flow fractionation-ICP-MS
AF4-MALS-ICP-MS	AF4 multi-angle light scattering-ICP-MS
AFS	atomic fluorescence spectrometry
AFP	alpha fetoprotein
AMS	accelerator mass spectrometry
ANN	artificial neural networks
APDC	ammonium pyrrolidinedithiocarbamate
APDG	atmospheric pressure glow discharge
APL	acute promyelocytic leukaemia
ASU	Atomic Spectrometry Update
BH	borohydride
BPANN	back-propagation artificial neural networks
CAR	carboxen
CCQM	Consultative Committee for Amount of Substance – Metrology in Chemistry and Biology
CE	capillary electrophoresis
CEA	carcinoembryonic antigen
CEC	cation-exchange chromatography
CI	confidence interval
CLSI	Clinical and Laboratory Standards Institute
CRM	certified reference material
CS	continuum source
CS-AAS	continuum source-AAS
CSF	cerebrospinal fluid
CV	cold vapour
CV-AFS	cold vapour-AFS
CVG	chemical vapour generation
DA	discriminant analysis
DBD	dielectric barrier discharge
DBS	dried blood spot
DCM	dichloromethane
DDTP	o,o′-diethyldithiophosphate
DES	deep eutectic solvent
DGT	diffusive gradient in thin film
DLLME	dispersive liquid–liquid microextraction
DMA	dimethylarsenic
DOC	dissolved organic carbon
DORS	diffuse optical reflectance spectroscopy
DRC	dynamic reaction cell
DTPA	diethylenetriaminepentaacetic acid
dw	dry weight
EDL	electrodeless discharge lamp
EDX	energy-dispersive X-ray
EDXRF	energy dispersive X-ray fluorescence
EFSA	European Food Safety Agency
ELISA	enzyme-linked immunosorbent assay
EQAS	external quality assessment scheme
ERM	european reference material
ESI	electrospray ionisation
ETAAS	electrothermal atomic absorption spectrometry
EtOH	ethanol
ETV	electrothermal vaporisation
EU	European Union
FAAS	flame atomic absorption spectrometry
FAO	UN Food and Agricultural Organization
FFF	field flow fractionation
fsLA-MC-ICP-MS	femtosecond laser ablation multicollector-ICP-MS
GBCA	Gd-based contrast agent
GD	glow discharge
GLS	gas liquid separator
GMP	good manufacturing practice
GSH	glutathione
GFAAS	graphite furnace atomic absorption spectrometry
HCA	hierarchical clustering analysis
hCG	human chorionic gonadotropin
HDPE	high density polyethylene
HG	hydride generation
HG-DBD-AFS	hydride generation in situ dielectric barrier discharge-AFS
HILIC	hydrophilic interaction liquid chromatography
HPLC	high performance liquid chromatography
HR	high resolution
HTISIS	high temperature torch integrated sample introduction system
IAEA	International Atomic Energy Agency
IAPSO	International Association of the Physical Sciences of the Ocean
iAs	inorganic arsenic
IC	ion chromatography
ICH	International Council for Harmonisation
ICP	inductively coupled plasma
ID	isotope dilution
id	internal diameter
IDMS	isotope dilution mass spectrometry
ID-TE-TIMS	isotope dilution total evaporation thermal ionisation mass spectrometry
IE	ion-exchange
iHg	inorganic mercury
IL	ionic liquid
ILC	inter-laboratory comparison
IP	ionisation potential
IR	infrared
IS	internal standard
iSe	inorganic Se
IQR	interquartile range
KXRF	K shell XRF
LA	laser ablation
LC	liquid chromatography
LDA	linear discriminant analysis
LIBS	laser induced breakdown spectroscopy
LLE	liquid–liquid extraction
LLME	liquid–liquid microextraction
LOD	limit of detection
LOQ	Limit of quantification
μXANES	micro XANES
μXRF	micro XRF
μPIXE	micro particle induced X-ray emission
m/z	mass-to-charge ratio
MAD	microwave-assisted digestion
MAD-UV	microwave-assisted ultraviolet digestion
MAE	microwave-assisted extraction
MAS	molecular absorption spectrometry
MC	multicollector
MeHg	methylmercury
MeOH	methanol
MeSeCys	methylselenocysteine
MIC	microwave-induced combustion
MIP	microwave induced plasma
MMA	monomethylarsenic
MRI	magnetic resonance imaging
MS	mass spectrometry
MS/MS	tandem MS
NAA	neutron activation analysis
NIST	National Institute of Standards and Technology
NP	nanoparticle
NRCC	National Research Council of Canada
OR	odds ratio
PCA	principal component analysis
PCa	prostate cancer
PCR	polymerase chain reaction
PDO	protected designation of origin
PDMS	polydimethylsiloxane
PET	polyethyleneterephthalate
PfLDH	Plasmodium falciparum lactate dehydrogenase
PGI	protected geographical indication
PIXE	particle-induced X-ray emission
PLS	partial least squares
PLS-DA	partial least squares discriminant analysis
PN	parenteral nutrition
PSA	Prostate specific antigen
PTFE	poly(tetrafluoroethylene)
PVA	polyvinyl alcohol
PVDF	polyvinylidene fluoride
PVG	photochemical vapour generation
QC	quality control
QD	quantum dot
QMS	quadropole MS
QQQ	triple quadrupole
RBC	red blood cells
REE	rare earth element
RFC	random forest classifier
RM	reference material
RMSECV	root mean square error of cross validation
RSD	relative standard deviation
SAGD	solution anode GD
sc	single cell
scICP-MS	single cell ICP-MS
SD	standard deviation
SD-LIBS	spark discharge-assisted LIBS
SDS	sodium dodecylsulfate
SEC	size exclusion chromatography
SeCys2	selenocystine
SEM	scanning electron microscopy
SeGalac	Se methylseleno-N-acetylgalactosamine
SeMet	selenomethionine
SeUr	selenourea
SF	sector field
SI	Système International d’unités – International System of Units
SIMCA	soft independent modelling of class analogy
sp	single particle
SPE	solid phase extraction
spICP-MS	single particle ICP-MS
SPME	solid phase microextraction
SR-μXRF	synchrotron radiation-μXRF
SRM	standard reference material
SR-nanoXRF	synchrotron radiation nanoXRF
SR-XRF	synchrotron radiation XRF
STEM	scanning transmission electron microscopy
SVM	support vector machine
TBAB	tetrabutylammonium bromide
TE	total evaporation
TEM	transmission electron microscopy
TFA	trifluoroacetic acid
THF	tetrahydrofuran
TIMS	thermal ionisation mass spectrometry
TMAH	tetramethylammonium hydroxide
TMSe	trimethylselenonium ion
TOF	time-of-flight
TRA	time resolved analysis
Tris	tris(hydroxymethyl)aminomethane
TXRF	total reflection XRF
UAD	ultrasound-assisted digestion
UAE	ultrasound-assisted extraction
UHPLC	ultra high performance liquid chromatography
UV	ultraviolet
UV-C	ultraviolet C radiation
UV-Vis	ultraviolet-visible
VG	vapour generation
vs.	versus
WDXRF	wavelength dispersive XRF
WHO	World Health Organisation
XANES	X-ray absorption spectroscopy near-edge structure
XRD	X-ray diffraction
XRF	X-ray fluorescence

10 Conflicts of interest

There are no conflicts to declare.

11 References

1. M. Patriarca, N. Barlow, A. Cross, S. Hill, A. Robson, A. Taylor and J. Tyson, J. Anal. At. Spectrom., 2021, 36(3), 452–511.
2. E. H. Evans, J. Pisonero, C. M. M. Smith and R. N. Taylor, J. Anal. At. Spectrom., 2021, 36(5), 868–891.
3. C. Vanhoof, J. R. Bacon, U. E. A. Fittschen and L. Vincze, J. Anal. At. Spectrom., 2020, 35(9), 1704–1719.
4. R. Clough, C. F. Harrington, S. H. Hill, Y. Madrid and J. Tyson, J. Anal. At. Spectrom., 2021, 36(7), 1326–1373.
5. S. Carter, R. Clough, A. Fisher, B. Gibson, B. Russell and I. Whiteside, J. Anal. At. Spectrom., 2020, 35(11), 2410–2474.
6. J. R. Bacon, O. T. Butler, W. R. L. Cairns, O. Cavoura, J. M. Cook, C. M. Davidson and R. Mertz-Kraus, J. Anal. At. Spectrom., 2021, 36(1), 10–55.
7. R. M. Galazzi, K. Chacon-Madrid, D. C. Freitas, L. F. da Costa and M. A. Z. Arruda, Rapid Commun. Mass Spectrom., 2020, 34(Suppl 3), e8726.
8. S. Theiner, K. Loehr, G. Koellensperger, L. Mueller and N. Jakubowski, J. Anal. At. Spectrom., 2020, 35(9), 1784–1813.
9. D. Torregrosa, G. Grindlay, L. Gras and J. Mora, Microchem. J., 2021, 166, 106200.
10. A. Lores-Padin, P. Menero-Valdes, B. Fernandez and R. Pereiro, Anal. Chim. Acta, 2020, 1128, 251–268.
11. A. L. Larraga-Urdaz, M. L. F. Sanchez, J. R. Encinar and J. M. Costa-Fernandez, Anal. Bioanal. Chem., 2022, 414(1), 53–62.
12. M. Welna, A. Szymczycha-Madeja and P. Pohl, Molecules, 2020, 25(21), 4944.
13. R. Fernández-Ruiz, X-Ray Spectrometry, 2021, pp. 1–15.
14. E. Margui, J. Jablan, I. Queralt, F. Bilo and L. Borgese, X-Ray Spectrometry, 2021.
15. W. Lorenc, A. Hanc, A. Sajnog and D. Baralkiewicz, Mass Spectrom. Rev., 2022, 41(1), 32–50.
16. A. Velasquez-Ferrin, D. V. Babos, C. Marina-Montes and J. Anzano, Appl. Spectrosc. Rev., 2021, 56(6), 492–512.
17. M. O. Varra, S. Ghidini, L. Husakova, A. Ianieri and E. Zanardi, Foods, 2021, 10(2), 270.
18. P. J. Parsons, A. L. Galusha, Y. X. Cui, E. M. Faustman, J. C. Falman, J. D. Meeker and K. Kannan, J. Anal. At. Spectrom., 2020, 35(10), 2092–2112.
19. A. R. Timerbaev, J. Anal. At. Spectrom., 2021, 36(2), 254–266.
20. J. Wang, J. Chao, C. Wei, H. Li, Q. Wang, P. Song, H. Lu, Y. Zhou, Y. Tang, S. Wang, L. Yang, K. Nadeau, I. G. Pihillagawa, M. E. Johnson, L. L. Yu, T. Naykki, T. Sara-Aho, R. P. Zambra, R. Napoli, O. Rienitz, J. Noordmann, C. Pape, J. Towara, C. Tsz-Chun, C. Hei-Shing, A. Stakheev, V. Dobrovolskiy, T. Stolboushkina, A. Glinkova, S. Taebunpakul, U. Thiengmanee, N. Kaewkhomdee, C. Uribe, E. Carrasco, A. Botha, P. Fisicaro, C. Oster, D. A. Ahumada, J. P. Abella, C. S. Segura, R. Shin, S. L. P. Deborah, F. Dewi, B. T. M. Kiat, W. Y. Zongrong, L. H. Wah, C. Haraldsson, J. Merrick, L. Antin, I. White, H. Goenaga-Infante, S. Hill, J. Entwisle, R. Jacimovic, T. Zuliani, V. Fajon, Y. H. Yim, S. W. Heo, K. S. Lee, J. W. Lee, Y. Lim, T. O. Okumu, M. Ndege, L. Wangui, S. Z. Can, F. G. Coskun, M. Tunc, P. Giannikopoulou, E. Kakoulides, K. Inagaki, S. Miyashita, H. Klich, R. Jebali, N. Chaaban, L. Bergamaschi, E. Sobina, T. Tabatchikova and P. Migal, Metrologia, 2020, 57(1A), 08013.
21. O. Geiss, I. Bianchi, C. Senaldi, G. Bucher, E. Verleysen, N. Waegeneers, F. Brassinne, J. Mast, K. Loeschner, J. Vidmar, F. Aureli, F. Cubadda, A. Raggi, F. Iacoponi, R. Peters, A. Undas, A. Muller, A. K. Meinhardt, E. Walz, V. Graf and J. Barrero-Moreno, Food Control, 2021, 120, 107550.
22. S. Nubler, M. E. Lopez, A. Castano, H. Mol, M. Schafer, K. Haji-Abbas-Zarrabi, D. Bury, H. M. Koch, V. Vaccher, J. P. Antignac, D. Dvorakova, J. Hajslova, C. Thomsen, K. Vorkamp and T. Goen, Int. J. Hyg. Environ. Health, 2021, 234, 113711.
23. C. G. de Vega, S. M. Chernonozhkin, R. Grigoryan, M. Costas-Rodriguez and F. Vanhaecke, J. Anal. At. Spectrom., 2020, 35(11), 2517–2529.
24. S. J. Christopher, D. L. Ellisor and W. C. Davis, Talanta, 2021, 231, 122363.
25. V. Zaichick and S. Wynchank, Radiat. Environ. Biophys., 2021, 60(1), 165–178.
26. D. Dudek-Adamska, T. Lech, T. Konopka and P. Koscielniak, Biol. Trace Elem. Res., 2021, 199(6), 2138–2144.
27. C. Exley and E. Clarkson, Sci. Rep., 2020, 10(1), 7770.
28. P. Heitland and H. D. Koster, J. Trace Elem. Med. Biol., 2021, 64, 126706.
29. P. Hoet, C. Jacquerye, G. Deumer, D. Lison and V. Haufroid, Clin. Chem. Lab. Med., 2021, 59(4), 729–742.
30. L. Barregard, D. G. Ellingsen, B. Berlinger, S. Weinbruch, F. Harari and G. Sallsten, Int. J. Hyg. Environ. Health, 2021, 233, 113693.
31. F. Ha, Y. Wu, H. Wang and T. Wang, Biol. Trace Elem. Res., 2022, 200(1), 1–12.
32. M. Vojtek, E. Pinto, S. Goncalves-Monteiro, A. Almeida, M. P. M. Marques, H. Mota-Filipe, I. Ferreira and C. Diniz, Anal. Methods, 2020, 12(39), 4806–4812.
33. R. E. Gamela, V. C. Costa, M. L. de Mello, J. C. Romero, F. A. C. Amorim and E. R. Pereira, J. Braz. Chem. Soc., 2020, 31(10), 2082–2090.
34. I. Ul-Haq, E. Ahmed, A. Sharif, M. Ahmed and W. Ahmad, Food Anal. Methods, 2021, 14(11), 2262–2275.
35. Y. Ling, F. Luo and S. Zhu, LWT–Food Sci. Technol., 2021, 142, 111077.
36. Z. Gajdosechova, C. H. Palmer, D. Dave, G. Jiao, Y. F. Zhao, Z. J. Tan, J. Chisholm, J. Z. Zhang, R. Stefanova, A. Hossain and Z. Mester, Environ. Pollut., 2020, 266(Pt 2), 115190.
37. Y. Lin, Y. Sun, X. Wang, S. Chen, Y. Wu and F. F. Fu, Microchem. J., 2020, 159, 105592.
38. J. S. Almeida, L. A. Meira, A. D. S. Nascimento, G. L. Santos, V. A. Lemos and L. S. G. Teixeira, Food Anal. Methods, 2021, 14(3), 596–605.
39. S. Martinez, R. Sanchez, J. Lefevre and J. L. Todoli, J. Anal. At. Spectrom., 2020, 35(9), 1897–1909.
40. A. F. Carneiro, C. N. Carneiro, L. D. Pires, L. S. G. Teixeira, S. M. Azcarate and F. D. Dias, Talanta, 2020, 219, 121218.
41. Y. B. Kurtulus, D. Bakircioglu, A. C. Babac, S. Yurtsever and N. Topraksever, J. AOAC Int., 2020, 103(5), 1256–1263.
42. J. Hwang, J. S. Park, S. Y. Kim, N. E. Song, M. Yoo, Y. Choi, J. S. Nam and D. Seo, J. Anal. Chem., 2020, 75(11), 1399–1403.
43. P. S. D. Cezario, M. C. L. do Nascimento, A. S. Luna and J. S. de Gois, Microchem. J., 2020, 157, 104875.
44. F. Iaquinta, F. Tissot, L. L. Fialho, J. A. Nobrega, M. Piston and I. Machado, Food Anal. Methods, 2021, 14(1), 156–164.
45. S. Z. Chen, L. P. Liu and D. J. Tang, Anal. Lett., 2021, 54(11), 1809–1825.
46. S. Zhang, Y. Chu, S. Ma, F. Chen, D. Zhang, Z. Hu, Z. Zhang, H. Jin and L. Guo, Microchem. J., 2020, 157, 105018.
47. B. E. D. Costa and N. M. M. Coelho, J. Food Compos. Anal., 2021, 95, 103686.
48. A. S. Maltsev, E. V. Chuparina, G. V. Pashkova, J. V. Sokol’nikova, O. V. Zarubina and A. N. Shuliumova, Food Chem., 2021, 343, 128502.
49. A. P. F. de Oliveira, R. F. Milani, P. Efraim, M. A. Morgano and S. A. V. Tfouni, Food Control, 2021, 119, 107455.
50. M. J. de Oliveira, F. A. S. Cunha and J. C. C. Santos, J. Anal. At. Spectrom., 2021, 36(4), 740–746.
51. M. L. Astolfi, M. E. Conti, E. Marconi, L. Massimi and S. Canepari, Molecules, 2020, 25(18), 4263.
52. R. Shahid, T. G. Kazi, H. I. Afridi, F. N. Talpur, A. Akhtar and J. A. Baig, Anal. Methods, 2020, 12(43), 5186–5194.
53. G. T. Druzian, M. S. Nascimento, U. Cerqueira, C. G. Novaes, M. A. Bezerra, F. A. Duarte and E. M. M. Flores, Food Chem., 2021, 344, 128677.
54. S. R. Waechter, P. Dalla Vecchia, J. S. Barin, E. M. M. Flores and F. A. Duarte, Talanta, 2021, 226, 122157.
55. G. S. Yang, J. Zheng, E. Kim, S. Zhang, H. Seno, M. Kowatari, T. Aono and O. Kurihara, Anal. Chim. Acta, 2021, 1158, 338431.
56. A. A. Gouda, R. El Sheikh, W. S. Hassan, N. Gouda and H. A. Khadrajy, Int. J. Environ. Anal. Chem., 2020, 1–12.
57. P. Divis, M. Reichstadter, Y. Gao, M. Leermakers and J. Krikala, Foods, 2020, 9(12), 1858.
58. S. Q. Li, C. C. Hao, P. Z. Xing, X. L. Xia, T. P. Liu and X. F. Mao, At. Spectrosc., 2020, 41(5), 211–217.
59. C. Z. Shi, W. Guo, L. L. Jin and S. H. Hu, RSC Adv., 2020, 10(70), 42993–42997.
60. R. Teuma-Castelletti and D. Beauchemin, J. Anal. At. Spectrom., 2020, 35(12), 2820–2825.
61. L. Cid-Barrio, F. Calderon-Celis, J. M. Costa-Fernandez and J. R. Encinar, Anal. Chem., 2020, 92(19), 13500–13508.
62. Y. Sun, N. Liu, Y. Wang, Y. Yin, G. Qu, J. Shi, M. Song, L. Hu, B. He, G. Liu, Y. Cai, Y. Liang and G. Jiang, Anal. Chem., 2020, 92(22), 14872–14877.
63. X. X. Yu, B. B. Chen, M. He and B. Hu, Anal. Chem., 2020, 92(19), 13550–13557.
64. Z. N. Chen, B. B. Chen, M. He and B. Hu, Anal. Chem., 2020, 92(18), 12208–12215.
65. B. Gomez-Gomez, M. Corte-Rodriguez, M. T. Perez-Corona, J. Bettmer, M. Montes-Bayon and Y. Madrid, Anal. Chim. Acta, 2020, 1128, 116–128.
66. L. Fu, H. Xie, J. Huang, X. Chen and L. Chen, Anal. Chim. Acta, 2021, 1165, 338564.
67. J. Wojcieszek, J. Jiménez-Lamana, L. Ruzik, M. Asztemborska, M. Jarosz and J. Szpunar, Front. Environ. Sci., 2020, 8, 100.
68. J. Noireaux, S. López-Sanz, J. Vidmar, M. Correia, L. Devoille, P. Fisicaro and K. Loeschner, J. Nanopart. Res., 2021, 23(4), 102.
69. G. Xiao, Y. Z. Liu and R. L. Jones, Radiat. Prot. Dosim., 2020, 191(4), 391–399.
70. F. Colin, R. Eleonore, F. Remi and C. Corinne, J. Anal. At. Spectrom., 2021, 36(4), 776–785.
71. M. C. García-Poyo, S. Bérail, A. L. Ronzani, L. Rello, E. García-González, B. Lelièvre, P. Cales, F. V. Nakadi, M. Aramendía, M. Resano and C. Pécheyran, J. Anal. At. Spectrom., 2021, 36(5), 968–980.
72. R. Grigoryan, M. Costas-Rodriguez, R. E. Vandenbroucke and F. Vanhaecke, Anal. Chim. Acta, 2020, 1130, 137–145.
73. S. Le Goff, E. Albalat, A. Dosseto, J. P. Godin and V. Balter, Rapid Commun. Mass Spectrom., 2021, 35(10), e9074.
74. I. F. Seregina, M. A. Bolshov and K. Ossipov, Spectrochim. Acta, Part B, 2021, 177, 106034.
75. Y. Y. Huang, J. T. S. Lum and K. S. Y. Leung, J. Anal. At. Spectrom., 2020, 35(10), 2148–2155.
76. J. Borovicka, L. Ackerman and J. Rejsek, Talanta, 2021, 221, 121389.
77. C. Ito, R. Shimode, T. Miyazaki, S. Wakaki, K. Suzuki and Y. Takagai, Anal. Chem., 2020, 92(24), 16058–16065.
78. E. Zambrzycka-Szelewa, E. Nalewajko-Sieliwoniuk, M. Zaremba, A. Bajguz and B. Godlewska-Zylkiewicz, Molecules, 2020, 25(15), 3402.
79. V. C. P. Marrocos, R. A. Gonçalves, F. G. Lepri and T. D. Saint’Pierre, Spectrochim. Acta, Part B, 2020, 174, 106008.
80. B. Gómez-Nieto, V. Motyzhov, M. J. Gismera, J. R. Procopio and M. T. Sevilla, Microchem. J., 2020, 159, 105335.
81. A. L. Vieira, E. C. Ferreira, S. R. Oliveira, F. Barbosa and J. A. G. Neto, Microchem. J., 2021, 160, 105637.
82. L. Chirita, E. Covaci, A. Mot, M. Ponta, A. Gandea and T. Frentiu, J. Anal. At. Spectrom., 2021, 36(2), 267–272.
83. P. Z. Xing, X. Li, L. Feng and X. F. Mao, J. Anal. At. Spectrom., 2021, 36(2), 285–293.
84. D. S. Irisov, R. K. Musin and Y. A. Zakharov, Spectrochim. Acta, Part B, 2021, 178, 106146.
85. A. V. Volzhenin, N. I. Petrova, T. E. Romanova and A. I. Saprykin, Anal. Lett., 2021, 54(14), 2293–2303.
86. L. F. R. Lemes and C. R. T. Tarley, Food Chem., 2021, 357, 129695.
87. A. Akhdhar, M. Schneider, S. Hellmann, A. Orme, E. Carasek, E. M. Krupp and J. Feldmann, Talanta, 2021, 227, 122190.
88. M. Gorska and P. Pohl, J. Anal. At. Spectrom., 2021, 36(7), 1455–1465.
89. X. X. Peng, M. Y. Zhao, M. L. Yuan and Z. Wang, Talanta, 2021, 225, 121995.
90. L. Qian, Z. Lei, X. Peng, G. Yang and Z. Wang, Anal. Chim. Acta, 2021, 1162, 338495.
91. Y. W. Chu, F. Chen, Z. Q. Sheng, D. Zhang, S. Y. Zhang, W. L. Wang, H. L. Jin, J. W. Qi and L. B. Guo, Biomed. Opt. Express, 2020, 11(8), 4191–4202.
92. X. Chen, Y. Zhang, X. H. Li, Z. H. Yang, A. C. Liu and X. Yu, Biomed. Opt. Express, 2021, 12(6), 3584–3596.
93. M. N. Khan, Q. Wang, B. S. Idrees, G. Teng, X. Cui, K. Wei and J. Chiou, J. Spectrosc., 2020, 2020, 1–11.
94. Z. Q. Yue, C. Sun, F. Y. Chen, Y. Q. Zhang, W. J. Xu, S. Shabbir, L. Zou, W. G. Lu, W. Wang, Z. W. Xie, L. Y. Zhou, Y. Lu and J. Yu, Biomed. Opt. Express, 2021, 12(5), 2559–2574.
95. J. Alsharnoubi, Y. Nassef, R. F. Fahmy and M. Gamal, Lasers Med. Sci., 2021, 36(5), 957–963.
96. P. Porizka, K. Vytiskova, R. Oborilova, M. Pastucha, I. Gabris, J. C. Brandmeier, P. Modlitbova, H. H. Gorris, K. Novotny, P. Skladal, J. Kaiser and Z. Farka, Microchim. Acta, 2021, 188(5), 147.
97. L. Su, W. Shi, X. Chen, L. Meng, L. Yuan, X. Chen and G. Huang, Food Chem., 2021, 338, 127797.
98. R. R. Gamela, E. R. Pereira and F. M. V. Pereira, Food Anal. Methods, 2021, 14(3), 545–551.
99. Y. Tian, Q. Chen, Y. Q. Lin, Y. Lu, Y. Li and H. Lin, Spectrochim. Acta, Part B, 2021, 175, 106027.
100. Y. C. Liu, Y. K. Hsieh, Y. W. Chu, I. Ahmed, Z. L. Hu, M. S. Khan, S. Y. Zhang, F. A. M. Manno, S. H. C. Manno, M. Lian, C. F. Wang, L. B. Guo and C. D. Lau, J. Anal. At. Spectrom., 2020, 35(10), 2224–2230.
101. D. Stefas, N. Gyftokostas and S. Couris, Spectrochim. Acta, Part B, 2020, 172, 105969.
102. D. C. Fechner, T. V. Silva, M. G. Nespeca, A. L. Vieira, J. A. Gomes, D. Santos, R. G. Pellerano and E. C. Ferreira, At. Spectrosc., 2021, 42(3), 148–153.
103. J. Park, S. Kumar, S. H. Han, S. H. Nam and Y. Lee, Spectrochim. Acta, Part B, 2021, 179, 106088.
104. M. Pérez-Rodríguez, P. M. Dirchwolf, T. V. Silva, A. L. Vieira, J. A. G. Neto, R. G. Pellerano and E. C. Ferreira, Food Chem., 2020, 331, 127051.
105. D. Peng-Kai, Z. Shang-Yong, Z. Ke-Xin, W. Ji, G. Xun, H. Zuo-Qiang and L. Jing-Quan, Acta Phys. Sin., 2021, 70(4), 040201.
106. M. Y. Yao, G. R. Fu, T. B. Chen, M. H. Liu, J. Xu, H. M. Zhou, X. W. He and L. Huang, J. Anal. At. Spectrom., 2021, 36(2), 361–367.
107. M. Guo, Q. Zhao, S. Fan, Z. Wu, L. Lin, H. Chen, Y. Gao and X. Hai, Br. J. Clin. Pharmacol., 2021, 87(10), 4020–4026.
108. M. Guo, J. Zhou, S. Fan, L. Li, H. Chen, L. Lin, Q. Zhao, X. Wang, W. Liu, Z. Wu and X. Hai, Expert Rev. Clin. Pharmacol., 2021, 14(4), 503–512.
109. S. X. Guo, X. Y. Wang, C. L. Gao, Z. Q. Wu, H. Z. Chen, L. W. Lin, M. H. Guo, Y. H. Gao and X. Hai, Toxicol. Lett., 2021, 347, 78–85.
110. M. Iwai-Shimada, Y. Kobayashi, T. Isobe, S. E. Nakayama, M. Sekiyama, Y. Taniguchi, S. Yamazaki, T. Michikawa, M. Oda, H. Mitsubuchi, M. Sanefuji, S. Ohga, N. Mise, A. Ikegami, R. Suga and M. Shimono, Toxics, 2021, 9(4), 82.
111. Y. Wei, J. Zhang, S. Qiu, Q. Huang, L. Yuan, L. Chen, T. Dai, T. Tu, B. Zhang, H. Yan and W. Li, Food Anal. Methods, 2021, 14(7), 1345–1358.
112. M. Kippler, I. Gyllenhammar, A. Glynn, M. Levi, S. Lignell and M. Berglund, Environ. Pollut., 2021, 268(Pt A), 115712.
113. A. I. Zhang, Y. Lin, J. H. Yang, L. B. He, Y. R. Deng, X. D. Hou and C. B. Zheng, Analyst, 2021, 146(12), 3852–3857.
114. M. T. Liu, L. Ding, J. X. Liu, X. F. Mao, X. Na and Y. B. Shao, Anal. Sci., 2021, 37(2), 321–327.
115. M. Liu, L. Ding, J. Liu, X. Na and X. Mao, Anal. Lett., 2021, 54(6), 990–1004.
116. M. Borowska and K. Jankowski, Talanta, 2020, 219, 121162.
117. C. F. G. Frois, W. Boschetti, A. S. dos Passos, M. L. Potes, M. G. R. Vale and M. M. Silva, Microchem. J., 2020, 157, 105029.
118. R. M. Menezes, W. N. L. Santos, U. N. Silva and V. A. Lemos, Water, Air, Soil Pollut., 2020, 231(9), 473.
119. Ç. Büyükpınar, S. Bodur, E. Yazıcı, Z. Tekin, N. San, O. T. Tank Komesli and S. Bakirdere, Measurement, 2021, 176, 109192.
120. J. Machado, P. M. Carvalho, A. Felix, D. Doutel, J. P. Santos, M. L. Carvalho and S. Pessanha, J. Anal. At. Spectrom., 2020, 35(12), 2920–2927.
121. P. M. S. Carvalho, S. Pessanha, J. Machado, A. L. Silva, J. Veloso, D. Casal, D. Pais and J. P. Santos, Spectrochim. Acta, Part B, 2020, 174, 105991.
122. M. Ghidotti, Y. Fiamegos, C. Dumitrascu and M. B. de la Calle, Food Chem., 2021, 342, 128350.
123. J. S. Almeida, L. A. Meira, M. S. Oliveira and L. S. G. Teixeira, X-Ray Spectrom., 2021, 50(1), 2–8.
124. A. R. Chowdhury, N. Maheshwari, J. Soni, M. Kapil, T. Mehta and A. Mukharya, J. Pharm. Biomed. Anal., 2020, 189, 113292.
125. E. Margui, J. Dumic, I. Queralt, L. Bakovic and J. Jablan, Anal. Methods, 2020, 12(40), 4899–4905.
126. A. C. Poletto, P. Singi, R. M. Barri, A. A. Casanova, C. C. D. Garbelini, C. C. D. Silva and E. J. Venancio, J. Trace Elem. Med. Biol., 2021, 63, 126663.
127. B. G. Beltrán, V. Ramos-Sanchez, D. Chávez-Flores, R. Rodríguez-Maese, E. Palacio and A. Espinosa Ferao, J. Chem., 2020, 2020, 1–9.
128. X. Zhang, A. J. Specht, E. Wells, M. G. Weisskopf, J. Weuve and L. H. Nie, Sci. Total Environ., 2021, 753, 142351.
129. M. R. Gherase, B. Serna and S. Kroeker, Physiol. Meas., 2021, 42(4), 045011.
130. D. E. B. Fleming, S. L. Crook, C. T. Evans, M. N. Nader, M. Atia, J. M. T. Hicks, E. Sweeney, C. R. McFarlane, J. S. Kim, E. Keltie and A. Adisesh, J. Trace Elem. Med. Biol., 2020, 62, 126603.
131. K. Nelson, B. Hesse, O. Addison, A. P. Morrell, C. Gross, A. Lagrange, V. I. Suarez, R. Kohal and T. Fretwurst, Anal. Chem., 2020, 92(21), 14432–14443.
132. Z. Y. Zhang, H. L. Lin, C. X. Ma, L. Z. Zhang, J. Zelevinsky, M. Xia, Y. C. Xu, J. C. White, J. Tyson and L. L. He, Food Control, 2021, 121, 107588.
133. H. Cho, K. B. Dasari, S. Cui, S. H. Choi and Y. H. Yim, Int. J. Environ. Anal. Chem., 2020, 1–18.
134. S. De Vos, N. Waegeneers, E. Verleysen, K. Smeets and J. Mast, Food Addit. Contam., Part A, 2020, 37(11), 1831–1846.
135. A. Grasso, M. Ferrante, G. Arena, R. Salemi, P. Zuccarello, M. Fiore and C. Copat, Int. J. Environ. Res. Public Health, 2021, 18(8), 4076.
136. A. Grasso, M. Ferrante, P. Zuccarello, T. Filippini, G. Arena, M. Fiore, A. Cristaldi, G. O. Conti and C. Copat, Int. J. Environ. Res. Public Health, 2020, 17(24), 9547.
137. B. Li, S. L. Chua, D. Yu, S. H. Chan and A. Li, J. Chromatogr. A, 2021, 1643, 462059.
138. B. Bocca, B. Battistini and F. Petrucci, Talanta, 2020, 220, 121404.
139. I. Abad-Alvaro, D. Leite, D. Bartczak, S. Cuello-Nunez, B. Gomez-Gomez, Y. Madrid, M. Aramendia, M. Resano and H. Goenaga-Infante, J. Anal. At. Spectrom., 2021, 36(6), 1180–1192.
140. S. Fernandez-Trujillo, M. Jimenez-Moreno, A. Rios and R. D. R. Martin-Doimeadios, Talanta, 2021, 231, 122370.
141. L. Cid-Barrio, D. Bouzas-Ramos, A. Salinas-Castillo, Y. Ogra, J. R. Encinar and J. M. Costa-Fernandez, J. Anal. At. Spectrom., 2020, 35(9), 1979–1988.
142. A. Guillard, E. Gaultier, C. Cartier, L. Devoille, J. Noireaux, L. Chevalier, M. Morin, F. Grandin, M. Z. Lacroix, C. Comera, A. Cazanave, A. de Place, V. Gayrard, V. Bach, K. Chardon, N. Bekhti, K. Adel-Patient, C. Vayssiere, P. Fisicaro, N. Feltin, F. de la Farge, N. Picard-Hagen, B. Lamas and E. Houdeau, Part. Fibre Toxicol., 2020, 17(1), 51.
143. M. Roman, C. Rigo, H. Castillo-Michel, D. S. Urgast, J. Feldmann, I. Munivrana, V. Vindigni, I. Micetic, F. Benetti, C. Barbante and W. R. L. Cairns, Analyst, 2020, 145(20), 6456–6469.
144. F. Grønbæk-Thorsen, R. H. Hansen, J. Ostergaard, B. Gammelgaard and L. H. Moller, Anal. Bioanal. Chem., 2021, 413(8), 2247–2255.
145. S. Salou, D. Lariviere, C. M. Cirtiu and N. Fleury, Anal. Bioanal. Chem., 2021, 413(1), 171–181.
146. X. L. He, H. T. Zhang, H. L. Shi, W. Y. Liu and E. Sahle-Demessie, J. Am. Soc. Mass Spectrom., 2020, 31(10), 2180–2190.
147. M. V. Taboada-Lopez, D. Bartczak, S. Cuello-Nunez, H. Goenaga-Infante, P. Bermejo-Barrera and A. Moreda-Pineiro, Talanta, 2021, 232, 122504.
148. X. Men, X. Wei, X. Zhang, X. Y. Wang, M. L. Chen, T. Yang and J. H. Wang, Talanta, 2020, 218, 121065.
149. M. Bergant, J. Ščančar and R. Milačič, Talanta, 2020, 218, 121199.
150. R. Pechancova, J. Gallo, D. Milde and T. Pluhacek, Talanta, 2020, 218, 121150.
151. G. Alchoubassi, K. Kinska, K. Bierla, R. Lobinski and J. Szpunar, Food Chem., 2021, 339, 127680.
152. A. Schweikert, S. Theiner, D. Wernitznig, A. Schoeberl, M. Schaier, S. Neumayer, B. K. Keppler and G. Koellensperger, Anal. Bioanal. Chem., 2022, 414(1), 485–495.
153. K. Billimoria, D. N. Douglas, G. Huelga-Suarez, J. F. Collingwood and H. Goenaga-Infante, J. Anal. At. Spectrom., 2021, 36(5), 1047–1054.
154. C. Arnaudguilhem, M. Larroque, O. Sgarbura, D. Michau, F. Quenet, S. Carrere, B. Bouyssiere and S. Mounicou, Talanta, 2021, 222, 121537.
155. D. P. Bishop, M. T. Westerhausen, F. Barthelemy, T. Lockwood, N. Cole, E. M. Gibbs, R. H. Crosbie, S. F. Nelson, M. C. Miceli, P. A. Doble and J. Wanagat, Sci. Rep., 2021, 11(1), 1128.
156. P. Kim, C. C. Zhang, S. Thoroe-Boveleth, S. Weiskirchen, N. T. Gaisa, E. M. Buhl, W. Stremmel, U. Merle and R. Weiskirchen, Biomedicines, 2020, 8(9), 356.
157. X. Gao, H. Pan, Y. Han, L. Feng, J. Xiong, S. Luo and H. Li, Anal. Chim. Acta, 2021, 1148, 238197.
158. A. Olszewska and A. Hanc, J. Environ. Health Sci. Eng., 2021, 19(1), 379–388.
159. O. M. Voloaca, C. J. Greenhalgh, L. M. Cole, M. R. Clench, A. J. Managh and S. L. Haywood-Small, Rapid Commun. Mass Spectrom., 2020, 34(21), e8906.
160. C. J. Greenhalgh, O. M. Voloaca, P. Shaw, A. Donard, L. M. Cole, M. R. Clench, A. J. Managh and S. L. Haywood-Small, J. Anal. At. Spectrom., 2020, 35(10), 2231–2238.
161. P. Bucker, H. Richter, A. Radbruch, M. Sperling, M. Brand, M. Holling, V. Van Marck, W. Paulus, A. Jeibmann and U. Karst, J. Trace Elem. Med. Biol., 2021, 63, 126665.
162. S. Theiner, A. Schweikert, C. Haberler, A. Peyrl and G. Koellensperger, Metallomics, 2020, 12(8), 1246–1252.
163. M. J. Pushie, A. Hollings, J. Reinhardt, S. M. Webb, V. Lam, R. Takechi, J. C. Mamo, P. G. Paterson, M. E. Kelly, G. N. George, I. J. Pickering and M. J. Hackett, J. Anal. At. Spectrom., 2020, 35(11), 2498–2508.
164. S. Nakanishi, R. Kamezono, M. Nakatani and M. Denda, Skin Res. Technol., 2021, 27(4), 576–581.
165. A. Turyanskaya, S. Smetaczek, V. Pichler, M. Rauwolf, L. Perneczky, A. Roschger, P. Roschger, P. Wobrauschek, A. Limbeck and C. Streli, J. Anal. At. Spectrom., 2021, 36(7), 1512–1523.
166. A. Turyanskaya, M. Rauwolf, V. Pichler, R. Simon, M. Burghammer, O. J. L. Fox, K. Sawhney, J. G. Hofstaetter, A. Roschger, P. Roschger, P. Wobrauschek and C. Streli, Sci. Rep., 2020, 10(1), 6301.
167. S. Genoud, M. W. M. Jones, B. G. Trist, J. J. Deng, S. Chen, D. J. Hare and K. L. Double, Chem. Sci., 2020, 11(33), 8919–8927.
168. S. P. Shi, H. Wang, Z. H. Chen, X. H. Li, S. X. Liu and M. Q. Zhang, Nucl. Sci. Tech., 2020, 31(10), 96.
169. R. Kondo, T. Yamato, A. Munoz-Noval, S. Honda, Y. Nishiwaki, K. Komaguchi and S. Hayakawa, J. Anal. At. Spectrom., 2021, 36(5), 1041–1046.
170. Y. Gu, P. Wang, S. Zhang, J. Dai, H. P. Chen, E. Lombi, D. L. Howard, A. van der Ent, F. J. Zhao and P. M. Kopittke, Environ. Sci. Technol., 2020, 54(19), 12072–12080.
171. K. Kiss, A. Šindelářová, L. Krbal, V. Stejskal, K. Mrázová, J. Vrábel, M. Kaška, P. Modlitbová, P. Pořízka and J. Kaiser, J. Anal. At. Spectrom., 2021, 36(5), 909–916.
172. A. Sindelarova, P. Porizka, P. Modlitbova, L. Vrlikova, K. Kiss, M. Kaska, D. Prochazka, J. Vrabel, M. Buchtova and J. Kaiser, Sensors, 2021, 21(3), 900.
173. E. Baser, D. A. Kirmizi, V. A. Turksoy, T. Onat, M. D. Caltekin, M. Kara and E. S. Yalvac, Z. Geburtshilfe Neonatol., 2020, 224(6), 339–347.
174. H. Zhao, J. Tang, Q. Zhu, H. He, S. Li, L. Jin, X. Zhang, L. Zhu, J. Guo, D. Zhang, Q. Luo and G. Chen, Sci. Total Environ., 2020, 742, 140613.
175. C. Osorio-Yanez, M. Sanchez-Guerra, M. Solano, A. Baccarelli, R. Wright, A. P. Sanders, M. M. Tellez-Rojo and M. Tamayo-Ortiz, Environ. Pollut., 2021, 282, 116962.
176. S. Z. Wu, H. Y. Xu, Y. Chen, Y. Chen, Q. L. Zhu, M. H. Tan and M. M. Zhang, Environ. Res., 2021, 195, 110822.
177. M. T. Aung, K. M. Bakulski, J. I. Feinberg, J. F. Dou, J. D. Meeker, B. Mukherjee, R. Loch-Caruso, C. Ladd-Acosta, H. E. Volk, L. A. Croen, I. Hertz-Picciotto, C. J. Newschaffer and M. D. Fallin, Epigenetics, 2021, 1–16.
178. M. Jiang, Z. L. Huang, X. Zhao, R. Liu and Y. Lv, J. Anal. At. Spectrom., 2021, 36(2), 279–284.
179. W. Jiang, G. Sun, W. Cui, S. Men, M. Jing, D. Pu, S. Zhang, X. Yuan, X. Zhang and C. Wang, Molecules, 2020, 25(22), 5370.
180. D. N. Pu, W. C. Jiang, G. H. Guo, X. Y. Wen, W. B. Cui, M. Jing, C. B. Wang and P. Li, Clin. Lab., 2020, 66(11), 2159–2171.
181. J. Zhang, R. Zhou, Y. Jin and N. Cheng, Microchem. J., 2021, 160, 105541.
182. X. Yin, B. B. Chen, M. He and B. Hu, Anal. Chem., 2020, 92(15), 10308–10315.
183. D. F. Jiang, C. Huang, L. J. Shao, X. L. Wang, Y. N. Jiao, W. Li, J. D. Chen and X. W. Xu, Anal. Chim. Acta, 2020, 1127, 182–189.
184. Y. Q. Xing, J. Han, X. Wu, D. T. Pierce and J. X. J. Zhao, Analyst, 2020, 145(24), 7932–7940.
185. J. Mu, L. L. Yu and T. E. Wellems, Front. Cell. Infect. Microbiol., 2021, 10, 620419.
186. F. Zhang, L. Yong, X. Hua, F. You, B. Wang, Y. L. Feng and L. Mao, Analyst, 2021, 146(6), 2074–2082.
187. Y. L. Zhu, Y. M. Lian, J. K. Wang, Z. P. Chen and R. Q. Yu, Anal. Chem., 2021, 93(14), 5839–5848.
188. Q. Kang, M. He, B. B. Chen, G. Y. Xiao and B. Hu, Anal. Chem., 2021, 93(2), 737–744.
189. X. Yin, B. B. Chen, M. He and B. Hu, Anal. Chem., 2021, 93(11), 4952–4959.
190. M. Vlcnovska, A. Stossova, M. Kuchynka, V. Dillingerova, H. Polanska, M. Masarik, R. Hrstka, V. Adam, V. Kanicky, T. Vaculovic and M. Vaculovicova, Molecules, 2021, 26(3), 630.
191. F. Parant, M. Bost, E. Nohuz, J. Prost, E. Cerruto, S. Moret, J. Maurin and G. Chene, Eur. J. Obstet. Gynecol. Reprod. Biol., 2020, 252, 434–438.
192. E. Martynko, E. Oleneva, E. Andreev, S. Savinov, S. Solovieva, V. Protoshchak, E. Karpushchenko, A. Sleptsov, V. Panchuk, A. Legin and D. Kirsanov, Microchem. J., 2020, 159, 105464.
193. M. Ohanian, P. Telouk, S. Kornblau, F. Albarede, P. Ruvolo, R. S. S. Tidwell, A. Plesa, R. Kanagal-Shamanna, E. L. Matera, J. Cortes, A. Carson and C. Dumontet, Am. J. Hematol., 2020, 95(4), 422–434.
194. S. Peters, K. Broberg, V. Gallo, M. Levi, M. Kippler, P. Vineis, J. Veldink, L. van den Berg, L. Middleton, R. C. Travis, M. M. Bergmann, D. Palli, S. Grioni, R. Tumino, A. Elbaz, T. Vlaar, F. Mancini, T. Kuhn, V. Katzke, A. Agudo, F. Goni, J. H. Gomez, M. Rodriguez-Barranco, S. Merino, A. Barricarte, A. Trichopoulou, M. Jenab, E. Weiderpass and R. Vermeulen, Ann. Neurol., 2021, 89(1), 125–133.
195. F. Patti, M. Fiore, C. G. Chisari, E. D’Amico, S. Lo Fermo, S. Toscano, C. Copat, M. Ferrante and M. Zappia, Environ. Res., 2020, 188, 109820.
196. A. L. Galusha, L. J. Howard, P. C. Kruger, T. Marks and P. J. Parsons, JPEN, J. Parenter. Enteral Nutr., 2021, 45(1), 175–182.
197. I. Legaz, E. Barrera-Perez, G. Prieto-Bonete, C. Perez-Martinez, A. Sibon, A. Maurandi-Lopez and M. D. Perez-Carceles, Forensic Sci. Int., 2021, 323, 110815.
198. B. A. Joda and N. I. Ward, J. Trace Elem. Med. Biol., 2021, 65, 126733.
199. A. Sajnog, M. Tkaczyk, M. Stanczyk, K. Szaflik, J. Suliburska, R. Kocylowski and D. Baralkiewicz, Talanta, 2021, 222, 121672.
200. A. S. Oliveira, E. A. C. Costa, E. C. Pereira, M. A. S. Freitas, B. M. Freire, B. L. Batista, M. S. Luz and K. P. K. Olympio, Sci. Total Environ., 2021, 758, 143583.
201. B. Arriaza, D. Blumenstiel, D. Amarasiriwardena, V. G. Standen and A. Vizcarra, Am. J. Phys. Anthropol., 2021, 174(2), 254–267.
202. A. Knoop, P. Planitz, B. Wust and M. Thevis, Drug Test. Anal., 2020, 12(11–12), 1666–1672.
203. C. D. Quarles Jr, M. Macke, B. Michalke, H. Zischka, U. Karst, P. Sullivan and M. P. Field, Metallomics, 2020, 12(9), 1348–1355.
204. B. Toubhans, A. T. Gourlan, P. Telouk, K. Lutchman-Singh, L. W. Francis, R. S. Conlan, L. Margarit, D. Gonzalez and L. Charlet, J. Trace Elem. Med. Biol., 2020, 62, 126611.
205. B. R. Duncan, M. Hasegawa, D. A. Marshall, L. F. Gonzalez-Cuyar, M. Paulsen, M. Kobayashi, K. R. Maravilla and C. D. Simpson, Anal. Bioanal. Chem., 2021, 413(6), 1571–1582.
206. K. A. Layne, K. Raja, P. I. Dargan and D. M. Wood, Invest. Radiol., 2021, 56(7), 458–464.
207. E. Pinto, P. Ramos, C. Vital, A. Santos and A. Almeida, J. Trace Elem. Med. Biol., 2020, 62, 126579.
208. J. Zhang, J. Li, L. Zhang, Z. B. Wang, S. L. Sun and Z. Y. Luo, Rapid Commun. Mass Spectrom., 2020, 34(7), e8658.
209. B. Liu, G. Xu, W. Yang, L. Strathearn, L. G. Snetselaar and W. Bao, Nutr. Neurosci., 2021, 1–7.
210. H. Wang, J. Cui, X. Hao, L. Guo, J. Zhao, R. Wang and H. Liu, Int. Arch. Occup. Environ. Health, 2021, 94(7), 1513–1522.
211. A. Gawor, A. Konopka, A. Ruszczynska, L. Paczek, Z. Gajewski and E. Bulska, J. Anal. At. Spectrom., 2020, 35(9), 1840–1847.
212. H. Ahmad, L. Zhao, C. Liu, C. Cai and F. Ma, Food Chem., 2021, 338, 128028.
213. Y. C. Chen and S. J. Jiang, J. Anal. At. Spectrom., 2021, 36(5), 938–945.
214. M. J. Clemente, S. Serrano, V. Devesa and D. Velez, Food Chem., 2021, 336, 127587.
215. J. Milinovic, C. Rodrigues, M. Diniz and J. P. Noronha, Algal Res., 2021, 53, 102149.
216. I. Aakre, D. D. Solli, M. W. Markhus, H. K. Maehre, L. Dahl, S. Henjum, J. Alexander, P. A. Korneliussen, L. Madsen and M. Kjellevold, Food Nutr. Res., 2021, 65, 7584.
217. H. S. Kainth, T. Singh, D. Khandelwal, G. Singh and S. Puri, Spectrosc. Lett., 2021, 54, 266.
218. I. Sarvan, A. E. Kolbaum, U. Pabel, T. Buhrke, M. Greiner and O. Lindtner, Food Chem. Toxicol., 2021, 149, 112005.
219. J. Yoshinaga and T. Narukawa, Regul. Toxicol. Pharmacol., 2020, 116, 104745.
220. M. M. Oliveira, T. M. B. Trevilato, S. I. Segura-Muñoz, D. C. Aragon, L. G. Alves, M. Nadal, M. Marques, J. L. Domingo, J. Sierra and J. S. Camelo Jr, Environ. Res., 2020, 188, 109733.
221. F. Masotti, S. Cattaneo, M. Stuknyte, V. Pica and I. De Noni, Trends Food Sci. Technol., 2020, 103, 348–360.
222. S. Z. Chen, Y. X. Liu, J. T. Yan, C. L. Wang and D. B. Lu, J. AOAC Int., 2021, 104(1), 129–136.
223. M. L. Astolfi, D. Marotta, V. Cammalleri, E. Marconi, A. Antonucci, P. Avino, S. Canepari, M. Vitali and C. Protano, Int. J. Environ. Res. Public Health, 2021, 18(10), 5073.
224. A. M. Pereira, D. H. Bonemann, S. H. Scherdien, B. P. Avila, I. F. Antunes, A. S. Ribeiro and M. A. Gularte, J. Food Compos. Anal., 2020, 94, 103622.
225. M. Filippini, A. Baldisserotto, S. Menotta, G. Fedrizzi, S. Rubini, D. Gigliotti, G. Valpiani, R. Buzzi, S. Manfredini and S. Vertuani, Chemosphere, 2021, 263, 127983.
226. A. S. Alnaimat, M. C. Barciela-Alonso, P. Herbello-Hermelo, R. Dominguez-Gonzalez and P. Bermejo-Barrera, Talanta, 2021, 225, 122083.
227. C. Cerutti, R. Sanchez, C. Sanchez, F. Ardini, M. Grotti and J. L. Todoli, Molecules, 2020, 25(23), 5602.
228. M. Esteki, E. Heydari, J. Simal-Gandara, Z. Shahsavari and M. Mohammadlou, Food Control, 2021, 124, 107889.
229. N. P. Kalogiouri, N. Manousi, D. Klaoudatos, T. Spanos, V. Topi and G. A. Zachariadis, Foods, 2021, 10(2), 349.
230. Y. Fiamegos, C. Dumitrascu, S. Papoci and M. B. de la Calle, Food Control, 2021, 120, 107496.
231. M. Foschi, A. A. D’Archivio and L. Rossi, Food Control, 2020, 118, 107438.
232. J. M. Muñoz-Redondo, D. Bertoldi, A. Tonon, L. Ziller, F. Camin and J. M. Moreno-Rojas, Food Control, 2021, 125, 107961.
233. L. X. Kuang, J. Y. Nie, J. Y. Zhang, G. F. Xu, J. Li, Y. Cheng, Y. M. Shen and F. Saqi, At. Spectrosc., 2021, 42(2), 91–98.
234. N. P. Rodrigues, E. Rodrigues, P. G. Celso, A. Kahmann, G. H. Yamashita, M. J. Anzanello, V. Manfroi and P. F. Hertz, LWT–Food Sci. Technol., 2020, 131, 109760.
235. T. Segelke, K. von Wuthenau, A. Kuschnereit, M. S. Muller and M. Fischer, Foods, 2020, 9(11), 1708.
236. M. Arif, G. Chilvers, S. Day, S. A. Naveed, M. Woolfe, O. Y. Rodionova, A. L. Pomerantsev, O. Kracht, C. Brodie, A. Mihailova, A. Abrahim, A. Cannavan and S. D. Kelly, Food Control, 2021, 123, 107827.
237. M. Perez-Rodriguez, P. M. Dirchwolf, Z. Rodriguez-Negrin and R. G. Pellerano, Food Chem., 2021, 339, 128125.
238. H. L. Liu, Y. T. Zeng, X. Zhao, Y. L. Ye, B. Wang and H. R. Tong, Food Res. Int., 2020, 136, 109483.
239. H. L. Liu, Q. Meng, X. Zhao, Y. L. Ye and H. R. Tong, Food Control, 2021, 123, 107735.
240. E. F. Fiorentini, M. B. Botella and R. G. Wuilloud, J. Food Compos. Anal., 2021, 95, 103661.
241. H. Shirkhanloo, M. Khaleghi Abbasabadi, F. Hosseini and A. Faghihi Zarandi, Food Chem., 2021, 347, 129042.
242. Y. Yang, M. He, B. Chen and B. Hu, Talanta, 2021, 232, 122425.
243. Y. Ji, M. Zhao, A. Li and L. Zhao, Microchem. J., 2021, 164, 105974.
244. A. Elik, M. Tuzen, B. Hazer, S. Kaya, K. P. Katin and N. Altunay, Sci. Rep., 2021, 11(1), 5481.
245. N. Manousi, E. Deliyanni and G. Zachariadis, Appl. Sci., 2020, 10(23), 8722.
246. S. M. Sorouraddin, M. A. Farajzadeh and T. Okhravi, J. Food Compos. Anal., 2020, 93, 103590.
247. N. A. Kasa, B. T. Zaman and S. Bakirdere, J. Anal. At. Spectrom., 2020, 35(11), 2565–2572.
248. E. M. Hafez, R. El Sheikh, A. A. Sayqal, N. AlMasoud and A. A. Gouda, Curr. Anal. Chem., 2020, 16(8), 1022–1031.
249. F. Tokay, R. Gunaydin and S. Bagdat, Talanta, 2021, 230, 122312.
250. V. M. Nik, E. Konoz, A. Feizbakhsh and A. A. M. Sharif, Microchem. J., 2020, 159, 105588.
251. D. S. C. Neves, A. S. Souza and L. R. de Lemos, Microchem. J., 2020, 159, 105458.
252. V. J. Ferreira, M. S. de Jesus, M. C. dos Santos, W. N. Guedes, V. A. Lemos, C. G. Novaes, F. S. Costa, C. S. V. Pacheco, E. G. P. da Silva and F. A. C. Amorim, Anal. Methods, 2021, 13(2), 267–273.
253. E. Bozorgzadeh, A. Pasdaran and H. Ebrahimi-Najafabadi, Food Chem., 2021, 346, 128916.
254. N. Altunay and A. Elik, Microchem. J., 2021, 164, 105958.
255. M. K. Abbasabadi and H. Shirkhanloo, J. Pharm. Biomed. Anal., 2020, 189, 113455.
256. S. G. Elci, Instrum. Sci. Technol., 2021, 49(3), 258–275.
257. S. G. Elci, Int. J. Environ. Anal. Chem., 2020, 1–15.
258. A. Elik, D. Bingol and N. Altunay, Food Chem., 2021, 350, 129237.
259. T. Borahan, B. T. Zaman, B. S. A. Polat, E. G. Bakirdere and S. Bakirdere, Anal. Methods, 2021, 13(5), 703–711.
260. K. Abdi, M. Ezoddin and N. Pirooznia, Int. J. Environ. Anal. Chem., 2021, 1–14.
261. M. Tuzen, A. Elik and N. Altunay, J. Mol. Liq., 2021, 329, 115556.
262. Y. Pourbakhshi, M. Heidari, E. Yahaei, S. Ghiyasi, H. Ebrahimi-Najafabadi and E. Bozorgzadeh, J. Anal. Chem., 2021, 76(6), 714–720.
263. S. Ozdemir, E. Kilinc, F. Sen and M. Soylak, Anal. Biochem., 2021, 617, 114122.
264. Y. Camci, A. T. Bisgin, Y. Surme, M. Ucan and I. Narin, J. Anal. Chem., 2020, 75(9), 1131–1136.
265. G. S. Tomasin, W. R. Silva, B. E. dos Santos Costa and N. M. M. Coelho, Microchem. J., 2021, 161, 105749.
266. S. Rais, A. Islam, I. Ahmad, S. Kumar, A. Chauhan and H. Javed, Food Chem., 2021, 334, 127563.
267. K. K. Jinadasa, P. Herbello-Hermelo, E. Pena-Vazquez, P. Bermejo-Barrera and A. Moreda-Pineiro, Talanta, 2021, 224, 121841.
268. A. B. S. Silva, I. P. Sá, J. M. Higuera and A. R. A. Nogueira, Food Anal. Methods, 2021, 14(3), 496–502.
269. A. F. Erulas, M. Saylan, S. Topal, B. T. Zaman, E. G. Bakirdere and S. Bakirdere, J. Food Compos. Anal., 2020, 94, 103623.
270. M. T. Mirzaee, S. Seidi and R. Alizadeh, Anal. Biochem., 2021, 612, 113949.
271. P. Chaikhan, Y. Udnan, R. J. Ampiah-Bonney and W. C. Chaiyasith, Microchem. J., 2021, 162, 105828.
272. M. Abdolhosseini, F. Shemirani and S. M. Yousefi, Microchim. Acta, 2020, 187(11), 602.
273. M. C. L. do Nascimento, F. N. Ferreira, A. P. Benevides, D. D. da Silva, D. V. Cesar, A. S. Luna and J. S. de Gois, J. Anal. At. Spectrom., 2021, 36(5), 900–908.
274. C. Demir, E. O. Er, B. Kartoğlu, N. Atsever, Ö. Yağci, M. Öner and S. Bakirdere, Chem. Pap., 2021, 75(8), 4261–4267.
275. R. Daymond, S. L. Curtis, V. Mishra and N. B. Roberts, Scand. J. Clin. Lab. Invest., 2020, 80(8), 630–639.
276. F. A. Alotaibi, R. J. Cornett and M. N. Herod, J. Environ. Radioact., 2021, 228, 106528.
277. R. Grigoryan, M. Costas-Rodriguez, P. Santens and F. Vanhaecke, Anal. Chem., 2020, 92(24), 15975–15981.
278. G. Xiao and R. L. Jones, J. Radioanal. Nucl. Chem., 2021, 328(1), 277–287.
279. A. A. Reinhard, J. D. Inglis, R. E. Steiner, S. P. LaMont, A. R. Cardon, A. M. Wende, D. L. Lopez and J. L. Roach, J. Radioanal. Nucl. Chem., 2020, 326(1), 611–619.
280. C. Hackethal, J. F. Kopp, I. Sarvan, T. Schwerdtle and O. Lindtner, Food Chem., 2021, 346, 128913.
281. R. Monteiro, C. Vale, N. Ferreira, P. Silva, E. Pereira and P. Vaz-Pires, Food Chem., 2021, 343, 128438.
282. M. A. Habila, N. AlMasoud, T. S. Alomar, Z. A. AlOthman, E. Yilmaz and M. Soylak, Molecules, 2020, 25, 4794.

---