A validated method to measure benzo[ a ]pyrene concentrations in tobacco by high-performance liquid chromatography-ﬂ uorescence detection

This publication describes a validated method suitable for the quanti ﬁ cation of Benzo[ a ]Pyrene (B[ a ]P) in tobacco blend and smokeless tobacco products by High-Performance Liquid Chromatography-Fluorescence Detection (HPLC-FLD). Samples were hydrated and extracted with a mixture of hexane and acetone. For the quanti ﬁ cation of levels of B[ a ]P the sample extracts were subjected to adsorption chromatography using base-modi ﬁ ed silica to remove co-extracted substances. The concentrated ﬁ nal extract was re-dissolved in acetonitrile and analysed by HPLC with ﬂ uorescence detection (FLD). B[ a ]P and the internal standard, deuterated B[ a ]P (D 12 -B[ a ]P) were resolved chromatographically. The method was validated and determined to be ﬁ t for purpose for the quanti ﬁ cation of B[ a ]P in tobacco from 3R4F Kentucky reference cigarettes, a ﬂ ue-cured Virginia cigarette tobacco blend containing 10% air-cured Burley tobacco, and a Tanzanian dark ﬁ re-cured cigarette blend. The method was also validated for smokeless tobacco products including commercially available dry snu ﬀ , soft pellet tobacco and pouched snus products over the concentration range of 0.38 ng g (cid:1) 1 to 150 ng g (cid:1) 1 , based on extraction of a 0.5 – 2 g aliquot of sample. The measurement uncertainty at a con ﬁ dence interval of approximately 95% was estimated from data generated by three analysts using two instruments on three separate occasions using matrix (pouched snus) forti ﬁ cation experiments. The expanded uncertainty of the method was (cid:3) 21.3% of the mean B[ a ]P concentration.


Introduction
There are a number of current developments in the area of tobacco product regulation that require validated analytical methods for quantication of toxicants in tobacco and cigarette smoke.For example, the World Health Organisation (WHO) Study Group on Tobacco Product Regulation (TobReg) has recommended limits on the benzo[a]pyrene (B[a]P) and TSNA content of smokeless tobacco products (STPs). 1 The same group has also recommended disclosure of the mainstream smoke emissions of 18 toxicants, and mandated lowering of the emissions of 9 of these. 2 In the USA the Food and Drug Administration (FDA) has identied over 90 Harmful and Potentially Harmful Constituents (HPHCs) in tobacco products and tobacco smoke, 3 and currently requires annual reporting of a subset of these in tobacco products and cigarette smoke emissions. 4B[a]P is one of the priority compounds highlighted by TobReg 1 and the FDA, 3 and TobReg has recommended a regulatory limit of 5 ng g À1 dry weight of smokeless tobacco, 1 based on approximate levels of quantication of currently available analytical methods.Benzo [a]pyrene is also a toxicant identied in smokeless tobacco manufacturer standards, Gothiatek® (a quality established by Swedish Match), 5 and the levels established by ESTOC. 6Validated analytical methods are an essential foundation for the effective measurement and potential future regulation of toxicant levels in tobacco products.
B[a]P is a Polycyclic Aromatic Hydrocarbon (PAH) and is mainly formed by the incomplete combustion or pyrolysis of organic compounds. 7Approximately 100 PAHs have been detected in tobacco 8 and they usually occur as complex mixtures rather than individual compounds.PAH compounds are predominately formed when organic materials are burned at temperatures in the range 500-700 C, as in the combustion of fossil fuels and cigarettes. 7PAHs may also be introduced into tobacco during leaf growth from environmental sources and the curing process. 9STPs contain variable levels of PAH compounds depending on the type of tobacco used in the product.For instance re-cured tobaccos, which may contain elevated levels of PAHs, are commonly used in snuffs and certain pellet tobacco products. 9AHs consist of condensed aromatic ring structures, are lipophilic compounds and are readily absorbed by inhalation, ingestion and dermal exposure. 7The most widely studied PAH is B[a]P, as it is the only PAH identied by IARC as a "known human carcinogen". 10Data obtained from rat tissue samples have indicated that B[a]P is rapidly distributed throughout the body and is metabolised by the enzyme cytochrome P450 to a reactive epoxide, B[a]P-7,8-diol-9,10epoxide.The epoxide metabolite is thought to be responsible for the carcinogenic properties of B[a]P. 11,12The structures of the target compound, B[a]P and its deuterated analogue B[a]P (D 12 -B[a]P), used as internal standard in this study, are shown in Fig. 1.
There are a number of published methods for the quantication of B[a]P and other PAHs including their measurement in narghile waterpipe tobacco smoke 13 and in mainstream cigarette smoke, using Gas Chromatography-Mass Spectrometry (GC-MS); [14][15][16] and of these, CRM 58 (ref.16) is the only method that has been subjected to extensive inter-laboratory testing and assessment of repeatability and reproducibility.A method for quantication of PAHs (including B[a]P) in mainstream tobacco smoke has been developed using Gas Chromatography-High Resolution Mass Spectrometry (GC-HRMS). 17Alternative techniques in the literature for quantication of PAHs in mainstream cigarette smoke include HPLC-FLD [18][19][20][21] and HPLC-APPI-MS/MS methods. 22n contrast to tobacco smoke, there are far fewer published methods for B[a]P or PAHs in smokeless tobacco products.Some smokeless tobacco product manufacturers have been operating to Gothiatek® or ESTOC standards for a number of years and therefore unpublished analytical methods are presumably available for this analyte in these matrices.Published methodologies for the determination of 23 PAH compounds in STPs 23 and for 21 PAH compounds in STPs 9 focus on GC-MS approaches.
HPLC-FLD has been used to determine PAH levels in a similar matrix, tea, 24 with levels of sensitivity allowing an LOQ of 0.35 ng g À1 for B[a]P to be achieved.The LOQ of 0.38 ng g À1 for B[a]P in tobacco is considerably below the commonly available LOQ of 5 ng g À1 discussed by TobReg, 1 and potentially represents a signicant step forward in analytical capability over some of the currently employed methods.The current paper therefore describes the development and validation of a high-performance liquid chromatography-uorescence detection method for the quantication of B[a]P in both tobacco blends and smokeless tobacco products.The method has been validated in accordance with international standards and guidelines (ICH, [25][26][27] IUPAC, 28,29 ISO, 30 DG SANCO 31 and FDA 32 ).
The validation process included a series of experiments designed to test the performance of the method against dened acceptance criteria including precision, expressed as residual standard deviation (RSD) at the LOQ of <20%. 31,32The matrices selected for validation included cigarette tobaccos and STPs with low endogenous B[a]P levels such as commercially available pouched snus, a Swedish-sourced so pellet STP, and a USstyle cigarette tobacco blend containing re-cured, air-cured and oriental blends.STPs and cigarette tobaccos with high endogenous B[a]P concentration such as commercially available dry snuff product and dark Tanzanian re-cured tobacco were also examined.Validation of the method for a range of tobacco products ensured method robustness and reproducibility for matrices containing low to high endogenous B[a]P levels.

Samples
Reference cigarette tobacco blend.Tobacco from a Kentucky reference (3R4F) cigarette was selected as a low PAH content test sample.It has traditional US-style tobacco blend, with a composition of 32.5% ue-cured Virginia tobacco, 19.9% Burley, 1.2% Maryland and 11.1% Oriental tobacco, 27.2% reconstituted tobacco (Schweitzer Process), glycerin 2.8% and sugar 5.3%. 33odied Virginia cigarette blend.An internal reference blend comprising 90% ue-cured Virginia cigarette tobacco, with 10% air-cured Burley tobacco was also selected as a midlevel PAH test sample.
Tanzanian dark re-cured tobacco blend.Tanzanian dark re-cured tobacco, of the type used in pipe tobacco blends and dry snuff was chosen as a high PAH test-sample. 34Dark recured tobaccos are produced by curing tobacco in ventilated barns with open res allowing smoke to come into contact with the tobacco during the drying processthis leads to relatively high PAH contents.

Smokeless tobacco products (STP)
So tobacco pellets.A commercial product used for method validation was Oliver Twist; it is a tobacco pellet comprising a  cylinder of tobacco leaf and avourings with a moisture content of 5-20%. 9The sample selected was sourced from Scandinavia, in contrast to the US-sourced product analysed previously. 9nus.Commercial portioned snus (Granit White and Lucky Strike Original) were used for method validation purposes.Snus is a smokeless tobacco product used in Scandinavia and is manufactured from heated and processed tobacco.Typical moisture content is greater than 40%. 9ry snuff.The commercial product Square was used in the method validation.It is a light brown powder with a typical moisture content of less than 10%. 9As Dry snuff contains a signicant proportion of re-cured tobacco it is relatively high in PAH content.Samples were selected in order to demonstrate method suitability for a range of tobacco blends and STPs.

Reagents and materials
A B[a]P solution with a certied concentration of 1000 mg mL À1 in acetone (QMX Laboratories Ltd, Thaxted, UK) and D 12 -B[a]P ($98% isotopic purity) was purchased from Sigma Aldrich (Gillingham, UK).Silica gel 60 Å chromatography grade (70-200 mm) and potassium hydroxide pellets (Fisher Scientic; Loughborough, UK) were used for the preparation of base-modied silica used during the method validation process.Methanol (high-performance liquid chromatography (HPLC) grade) and acetonitrile (HPLC uorescence grade) with a purity of >99.9%, were purchased from Fisher Scientic (Loughborough, UK) and used in the preparation of matrix samples and for the HPLC mobile phase.Water from an Elga Process Water (High Wycombe, UK) deionised water generator (minimum quality 18.2 MU cm À1 ) was used.

Equipment
Cigarette tobacco blends were ground to 1 mm using a Retsch ZM200 mill and the so pellet tobacco (Oliver Twist) cryomilled prior to extraction using a Retsch Cryomill (Retsch UK Ltd.; Hope, UK).All liquid transfers were made with calibrated pipettes and grade B volumetric glassware.All laboratory consumables were supplied by Fisher Scientic (Loughborough, UK).Sample extraction was carried out using a at-bed shaker (IKA®-Werke; Staufen, Germany) and a Rotanta 460 centrifuge (Hettich; Tuttlingen, Germany).For production of the base-modied silica a rotary evaporator (Buchi, Oldham, UK) was used.Isolute 70 mL reservoir pack, Isolute 27 mm frits and prepacked base-modied silica 70 mL 10 g À1 cartridges (Biotage; Uppsala, Sweden).
A custom made manifold was used to collect eluent aer sample clean up.The manifold was constructed using Lexan® polycarbonate (Gilbert Curry Industrial Plastics; Coventry, UK) (Fig. 2) to enable it to be placed over a Nalgene™ 24 Â 30 mm tube rack holder containing 60 mL glass vials (Fisher Scientic; Loughborough, UK).Promega™ One-Way Luer-Lok™ stopcocks (Promega™ UK; Southampton, UK) were inserted into the drilled holes in the manifold to support the 70 mL 10 g À1 potassium silicate cartridge and to control eluent ow.
Turbovap sample concentrators (Biotage; Uppsala, Sweden) with a tube holder for 60 mL vials were used to concentrate the sample extracts.Two Agilent 1200 series HPLC systems (instruments 1 and 2) were coupled to Agilent 1200 Innity uorescence detectors and chromatographic separation performed using a Zorbax Eclipse Plus PAH column (250 mm Â 2.1 mm with a 5 mm particle size; Agilent Technologies).Data were processed with Agilent Chemstation soware (version B.04.03).A Cary 5 double beam spectrophotometer was used for the determination of water in tobacco by near infra-red spectroscopy (Agilent Technologies; Wokingham, UK).

Base-modied silica cartridges
To a 3000 mL round bottomed ask, 500 mL of methanol and 168 g potassium hydroxide were added and the ask attached to a rotary evaporator.The ask was submerged in a water bath and cooled to ambient temperature as the reaction is exothermic.When the potassium hydroxide pellets had dissolved in the methanol, 300 g of silica gel were added with continuous stirring followed by a further addition of 500 mL of methanol.The temperature of the water bath was increased to 40 C and le to mix for 30 minutes.The methanol was decanted from the ask and further 500 mL methanol added and then le to mix for a further 30 minutes; this was repeated for a further two methanol washes.The excess methanol was decanted off and the slurry poured into a chemical resistant tray and le to dry overnight in a fume hood.Free owing base-modied silica powder was obtained.A frit was placed in the bottom of a 70 mL reservoir cartridge followed by addition of 10 AE 0.3 g base-modied silica and a frit placed on top of the silica.For routine laboratory use custom made base-modied silica cartridges were produced by Biotage (Uppsala, Sweden).High-performance liquid chromatography-uorescence detection conditions.HPLC separation was performed at a temperature of 35 C and run time of 35 min.The injection volume was 10 mL and the eluent ow rate 0.5 mL min À1 .Mobile phases used were deionised water (eluent A) and acetonitrile (eluent B).Gradient proles are shown in Table 1.Excitation wavelength was 290 AE 3 nm and emission wavelength was 440 AE 3 nm.

Standard solutions
For the internal standard stock solution, 10 AE 0.1 mg of D 12 -B[a]P were weighed into a 20 mL amber glass vial.Aer addition of 5-7 mL acetonitrile, the vial was capped and sonicated for 5 min at 30 AE 5 C. The solution was cooled to room temperature, quantitatively transferred to an amber glass 10 mL volumetric ask and made up to volume with acetonitrile (nal D 12 -B[a]P stock solution concentration 1000 ng mL À1 ).
The working-standard solutions were prepared by transferring 100 mL D 12 -B[a]P stock solution to a 100 mL or a 1000 mL volumetric ask and adding acetonitrile to make up to volume (concentrations 1000 ng mL À1 (solution 1) and 100 ng mL À1 (solution 2).
For the B[a]P stock standard solution, 0.25 mL of 1000 mg mL À1 B[a]P in acetone was transferred from an ampoule to a 100 mL volumetric ask and made up to volume with acetonitrile (concentration 2500 ng mL À1 ; solution 3).A volume of 10 mL B[a]P stock solution was transferred to a 50 mL volumetric ask and made up to volume with acetonitrile (concentration 500 ng mL À1 ) to make the working standard stock solution (solution 4).Standard solutions were stored at À20 AE 2 C.

Calibration solutions
Nine calibration standard solutions in the concentration range of 0.25 to 25 ng mL À1 were prepared from D 12 -B[a]P (solution 1) and B[a]P (solution 4).A volume of 250 mL of D 12 -B[a]P (solution 1) was transferred to 50 mL volumetric asks to give a nal

Linearity and statistical evaluation
First order linear models using ANOVA were performed using Minitab (version 16) statistical soware (Minitab Inc., State College, Pennsylvania, USA).The calibration linearity was evaluated for the B[a]P working standard solutions over the entire calibration range.Tests for statistical signicance were set at the 95% condence level.Samples were analysed in triplicate on three separate occasions.A linear correlation t was applied to the response ratio (the ratio of the B[a]P peak area to the internal-standard peak area), as a function of the analyte concentration ratio (the ratio of B[a]P concentration to the internal-standard concentration).The linear regressions of the scatter plots for the replicates were assessed.The calibration range for the method was determined by the upper and lower concentrations of B[a]P in solvent at which acceptable precision, accuracy and linearity were achieved.The linearity of results for standard fortied matrix extracts were compared with a calibration curve of standards in pure solvent for the assessment of potential matrix effects.
In the absence of a blank matrix, the detection limit (LOD) of the method was estimated from calibration curves as LOD ¼ (3.3s)/slope,where s ¼ standard deviation of the response. 27he lower limit of quantication of the method was determined by completing six replicate injections of the lowest level of B[a]P standard.The target precision at the LOQ was set at a relative standard deviation of 20% in accordance with DG SANCO and FDA guidelines. 31,32

Stability
The stability of B[a]P working-standard solutions was assessed by comparison of the measured concentration of B[a]P in freshly prepared solutions with solutions stored at ambient conditions for 4 days.In addition, the stability of standards in solvent stored at À20 AE 2 C was assessed periodically over 15 weeks.The stability of modied-Virginia blend sample extracts (n ¼ 5) stored at À20 AE 2 C was assessed on day 0 and on day 30 by comparison with a freshly made standard solution.

Tobacco hydration
A pre-extraction hydration step of a similar dry leaf matrix (tea) was reported to improve the extraction efficiency of pesticide residues from black and green tea samples. 35Experiments were therefore carried out using tobacco from 3R4F reference cigarettes to determine the effect of hydration on B

Sample extraction
Because the method was applied to a wide range of tobacco products with B[a]P content ranging from 0.38 ng g À1 to 150 ng g À1 and in order to avoid potential non-linearity of response for samples with very high B[a]P content, the mass of sample extracted was matched to sample type.For snus samples a mass  of 2 g was extracted.For samples containing high proportions of re-cured tobacco, 0.5 g was extracted and the extracts were diluted prior to analysis.For other samples a mass of 1 g was extracted.Extraction was conducted in three stages in order to ensure exhaustive recovery of the incurred B[a]P and thus quantitative determination.The mass of sample was varied according to the matrix (Table 2).For STPs with low expected endogenous B[a]P concentration (for example Granit White, Lucky Strike Original snus) two pouches equivalent to a mass of 2.0 g were extracted.
For tobacco blend and STPs in which high levels of B[a]P were expected (for example products containing a high proportion of re-cured tobacco, such as Square dry snuff) a lower sample mass was extracted (0.5 g) to ensure that B[a]P responses were within the method calibration limits.For all other samples 1 g of sample was extracted.Samples were hydrated at a ratio of 1 : 1 water : tobacco and then refrigerated overnight (18 hours) to equilibrate.Granit White and Lucky Strike Original snus were stored in the freezer and defrosted prior to extraction.Optimal extraction efficiency of B[a]P was obtained for samples with a nal moisture content of 42-57%.Hydration was required for matrices with low moisture content, but additional hydration of Granit White and Lucky Strike Original snus was determined to be unnecessary due to high intrinsic moisture levels in the product (>40%).
In order to remain within the calibration range, samples containing very high B[a]P levels, including the Tanzanian dark re-cured tobacco and Square dry snuff samples, were fortied with 150 mL of solution 1, equivalent to 150 ng D 12 -B[a]P.A tenfold dilution was applied at the end of the extraction step.For all other matrices (including unknown samples) with expected low to medium B[a]P concentrations, 15 ng of D 12 -B[a]P was added (150 mL of solution 2) and samples were analysed without dilution.The mass of unknown samples was adjusted where necessary to ensure that the B[a]P concentration in the nal extract was within the calibration limits.
Two extraction approaches were adopted.For mixtures with low intrinsic moisture content the following approach was adopted.Aer the addition of D 12 -B[a]P, samples were equilibrated for 30 minutes at room temperature prior to extraction.The samples were extracted with 25 mL 90 : 10 (v/v) hexane : acetone for 30 min on a reciprocating shaker at 180 rpm and centrifuged at 4600 rpm for 5 min.The supernatant was decanted into a 60 mL amber vial.The process was repeated two more times to collect the cumulative sample extract.The extract volume was reduced between extraction steps using a Turbovap sample concentrator to obtain a nal extract volume of 1-5 mL.
Two snus pouches were fortied with D 12 -B[a]P and le to equilibrate at ambient temperature for 30 min.Due to the high intrinsic moisture content (>40%) a higher proportion of polar aprotic solvent was required for the initial extraction of pouched snus to ensure optimal extraction efficiency.The extraction efficiencies for B[a]P in pouched snus were 10% higher when a 50 : 50 (v/v) hexane : acetone solvent system was used for the rst extraction step.For blend and the other STPs, extraction efficiencies when using a higher proportion of polar aprotic solvent for initial extraction were reduced with a 14% reduction in extraction efficiency observed for modied Virginia blend).Therefore, two whole snus pouches were rst extracted with 50 : 50 (v/v) hexane : acetone for 30 min on a reciprocating shaker at 180 rpm before centrifugation and concentration, this was then followed by two further extractions with 25 mL 90 : 10 (v/v) hexane : acetone as described for the other tobacco samples.

Sample clean-up
Each supernatant was transferred to a hexane pre-conditioned base-modied 70 mL 10 g À1 potassium silicate cartridge.The eluent was collected in a 60 mL vial.B[a]P was further eluted from the cartridge with an additional two volumes of 25 mL hexane.The total eluent volume (50-55 mL) was reduced by careful evaporation of the solvent under a ow of nitrogen to just before the point of dryness.The sample was reconstituted in 3 mL acetonitrile and sonicated for 40 min at 25 C. Dry snuff and re-cured tobacco were diluted ten-fold prior to analysis because screening measurements indicated that these samples contained B[a]P levels above the upper calibration range.

Standard fortied matrix calibration
Fortication standard solutions in acetonitrile were prepared at three levels: low, medium and high (Table 3).Calibration solutions in matrix were prepared as follows: for the low and medium-level fortication solutions, different volumes of 500 ng mL À1 B[a]P working standard stock (solution 4) and for the high fortication standard solution a B[a]P stock of 2500 ng mL À1 (solution 3) was used (Table 3).Standard fortied matrix samples were prepared (Table 4) for all tobacco blends and STPs.The nal volume of fortied B[a]P matrix samples was 500 mL aer the addition of 400 mL of matrix extract.
The matrix extract for each sample was produced aer drying ten extracts each with a volume of 50-55 mL aer clean-up to just the point of dryness and reconstitution in 3 mL acetonitrile.The ten samples therefore gave a total extract volume of 30 mL (10 Â 3 mL) which was used to produce the standard fortied matrix at each calibration level (n ¼ 4) with two calibration curves produced on each HPLC system (instrument 1 and 2).Unfortied matrix samples were assessed to determine endogenous concentrations of B[a]P as a control.Accuracy Accuracy (i.e.closeness of agreement between the result of a measurement and the true value) of the method was assessed by calculation of the ratio between the determined and nominal concentrations.The accuracy was determined for standards in solvent and for each of the fortied matrix extracts aer subtraction of the endogenous level.Acceptable accuracies were 50-120% for an analyte concentration of <1 ng g À1 , 70-110%, for the range of 1-10 ng g À1 and 80-110% for B[a]P concentration >10 ng g À1 . 37epeat extractions of unfortied Granit White snus matrix were analysed (n ¼ 16) to assess the accuracy of the method for samples close to the LOQ (<1 ng g À1 ).The mean measured endogenous concentration of 2.20 ng g À1 DWB, (precision of 12.2% (n ¼ 16)) was subtracted from B[a]P fortied Granit White snus samples used to determine the accuracy at the three levels of matrix fortication (low, medium and high) with six replicates per level.

Precision
The precision of the method (i.e. the closeness of replicate results) was calculated as the ratio of the standard deviation to the mean corrected concentration of six samples at each forti-cation level (n ¼ 18 samples) i.e. the relative standard deviation.

Intermediate precision
The intermediate precision (i.e.within-laboratory variations arising from different analysts and instruments) was calculated from the measurements made by three different operators using two instruments (instruments 1 and 2) on three separate occasions.The medium fortication standard solution was prepared in solvent (acetonitrile) (Table 4) and used to fortify Granit White extracts.Twenty four Granit White pouches were fortied with 15 ng of internal standard D 12 -B[a]P (150 mL of solution 2) prior to extraction.To six replicates of each fortication level (unfortied, low, medium and high) the following volumes of B[a]P medium fortication standard 100 ng mL À1 (Table 3) were added: 0 mL, 30 mL, 210 mL and 450 mL equivalent to 0 ng, 3 ng, 21 ng and 45 ng respectively.Unfortied samples were assessed to determine endogenous concentrations of B[a]P as a control.

Uncertainty of measurement
Standard uncertainty of measurement (u) (i.e. the parameter characterising the dispersion of the values, was estimated from the mean ( x) and standard deviation (s) of repeated measurements (n)).The mean of the repeated readings, x was calculated using the formula (1) and estimated standard deviation (2) From these values the estimated standard uncertainty of the mean, u was calculated (3) and then the combined standard uncertainty (u c ) calculated by squaring the measurement The expanded uncertainty (U) (i.e. the quantity dening an interval around a measured value encompassing a large fraction of the distributed values was calculated using the formula where k is the coverage factor.When k ¼ 2, the values are expected to fall within two standard deviations of the mean to provide an approximate level of condence of 95%. 32,38sults and discussion

Tobacco hydration
Extraction efficiency for B[a]P was 11% higher for 3R4F cigarette blend and 9.4% higher for STP Oliver Twist with tobacco matrix hydration at a ratio of 1 : 1 water : tobacco (18 hours/ overnight in a refrigerator) compared to non-hydrated matrix (Fig. 3).Findings are consistent with published data for pesticide residue extraction in tea. 35

Stability
Standard solutions were stable at ambient temperature for a week at all calibration levels, with B[a]P concentrations measured as 100-103% of those in fresh solutions.Aer storage of the standards at À20 C for 15 weeks, the measured B[a]P concentrations for solutions at calibration levels 3-9 were 95.2-105%.The accuracy of calibration standard levels 1 and 2 were in the range 101-109% at all time-points.For matrix samples, the measured mean concentration of B[a]P in modied Virginia blend extracts (n ¼ 5) reduced by 14.6% aer 1 month of storage at À20 C. The precision of measurement (% RSD) was consistent for freshly prepared extracts and aer storage at À20 C for 15 weeks with values of 8.2% and 7.6% respectively.

Linearity and statistical evaluation
The correlation coefficient for calibration standards 1-9 in pure solvent (acetonitrile) was high (R 2 > 0.995).The p-value of >0.05 indicated no statistically signicant difference at the 95% condence level between calibration curves for standards in solvent, n ¼ 3.For matrix-matched standards, regression analysis of all calibration curves showed a strong correlation (R 2 > 0.996).
When the averaged calibration curve for Tanzanian dark recured tobacco was compared with that for solvent, the slopes were not signicantly different at the 95% condence level.However, a statistically signicant difference was seen between the intercepts (p < 0.05) at the 95% condence level.This difference corresponds to the high endogenous level of B[a]P observed in unfortied Tanzanian dark re-cured tobacco control samples.Signicant differences in intercepts between matrix matched standards and standards in solvent were also seen for Square dry snuff, Kentucky reference cigarette, 3R4F (Table 5) which is consistent with the endogenous levels of B[a]P in these matrices.The endogenous levels of B[a]P quantied in unfortied matrix samples are presented in Table 6.
The p-values for the slopes in all matrices except Oliver Twist pellet tobacco were >0.05 (Table 5), indicating no signicant differences between slope gradients for matrix matched standards and standards in solvent at the 95% condence level.There was a small difference in the magnitude of the slope for B[a]P levels in fortied Oliver Twist matrix compared to standards in solvent.Differences are attributed to the Oliver Twist extract containing endogenous levels of B[a]P close to the LOQ (Table 6) as slightly higher variability was observed at concentrations near the LOQ.Measurements were however within the recommended acceptance limits for the precision of measurement, RSD <20%. 31,32he Oliver Twist unfortied blank matrix (0.48 ng g À1 B[a]P DWB, RSD ¼ 11.1% n ¼ 3) was subtracted from the other fortication levels for statistical analysis which led to a slight difference between the slopes.Although there was a slight difference between slope gradients for matrix and solvent for Oliver Twist pellet tobacco, the other validation parameters were unaffected.Averaged data generated for modied Virginia blend indicated a signicant statistical difference between slopes at the 95% condence level with a p-value ¼ 0 for the slope of averaged matrix matched curves.The observed statistical differences in the slopes for modied Virginia blend may be attributed to instrumental variation as data used to generate the rst order linear models using ANOVA were averaged from both instruments 1 and 2. Comparison of calibration curves from a single instrument for modied Virginia blend indicated no signicant differences between slopes, with a p-value >0.05 (Table 5).For samples with low endogenous B[a]P concentrations in the unfortied matrix (Oliver Twist chewing tobacco, Granit White snus and LS Original snus) there was no statistically signicant difference between intercepts of calibration curves for standards in solvent and fortied matrix at the 95% condence level.In the absence of blank matrix, the method LOD for B[a]P was estimated as 59 pg from the calibration curves of standards in solvent. 27The LOQ for the analysis based on extraction of Granit White pouched snus matrix was 0.38 ng g À1 B[a]P (WWB) and was within the SANCO 31 and FDA 32 dened acceptance criteria.

Accuracy and precision
Calibration standards 1-9 in solvent (acetonitrile) exhibited accuracy values of 96.7-114% and precision values of 0.142-2.06%expressed as residual standard deviation (RSD) (Table 7).The accuracy of measurement of the calibration standards 1-9 in solvent (n ¼ 6) ranged from 96.7% to 114% across all calibration levels with a maximum precision (RSD) of 2.06% which met the specied acceptance criteria. 15,16,21The accuracy for calibration standard 1 (114%) and the precision (RSD) was 1.62% which was within the acceptance criteria for concentrations of analyte <1 ng g À1 with accuracy limits of 50-120% and precision of <20%.
In all fortied matrices (corrected for endogenous B[a]P concentrations), the accuracy of measurements close to the LOQ ranged from 70% to 120%, with a relative standard deviation <20%.For unfortied Granit White matrix (endogenous B[a]P concentrations close to the LOQ) the precision of measurement (RSD) was 12.2% (n ¼ 16).The average recovery values for D 12 -B[a]P in Granit White snus were 72.0-83.7% and the accuracy of B[a]P measurement aer correction for endogenous B[a]P ranged between 101% and 107%.Therefore, even at calibration level 1, B[a]P could be measured with acceptable accuracy. 31,32he intermediate precision for mass normalised samples at the low level of B[a]P fortication was 1.19 ng g À1 aer subtraction of the unfortied B[a]P concentration for Granit White (n ¼ 16).The pooled precision for the low fortication level was 23%.Variability was observed as concentrations were near the LOQ.However the precision (RSD) was #20% when data for individual analysts were assessed (range 7.80-14.7%).For medium-level and high-level fortication, the pooled relative standard deviation performance standard of #15% was met by all analysts when data were pooled, therefore meeting SANCO and FDA acceptance criteria. 31,32

Uncertainty of measurement
The expanded uncertainty for the determination of approximately 95% condence interval limits was calculated from data generated by the three analysts using instrument 1 and 2 on separate days (Table 8).The percentage-expanded uncertainty was AE22.2% of the mean for low level B[a]P fortication in Granit White samples, AE18.0% for the medium level and AE23.8% for the high level, with an average of AE21.3%.The main factors inuencing the uncertainty of measurement was some slight differences in uorescent detector sensitivity observed from the two different instrumental systems used for method validation.However the main inuencing factor on the uncertainty of measurement was the analyst skill (for example possible differences in B[a]P and D 12 -B[a]P fortication, variability in the pipettes used for fortication and peak integration differences).The uncertainty from each source therefore contributed to the overall expanded uncertainty of measurement.

Conclusion
A method has been developed and validated for the determination of B[a]P in Kentucky reference 3R4F cigarette tobacco, a lightly modied Virginia blend cigarette tobacco, Tanzanian dark re-cured tobacco blend, for STPs including commercially available Square dry snuff, Oliver Twist chewing tobacco and Granit White and Lucky Strike Original snus.The method applies to a concentration range of 0.38 ng g À1 to 150 ng g À1 based on extraction of a 0.5-2 g aliquot of sample.The expanded uncertainty of measurement was AE21.3% of the mean.In the absence of blank matrix, the method LOD for B[a]P was estimated as 59 pg from the calibration curves of standards in solvent 27 and the LOQ for the analysis based on extraction of Granit White pouched snus was 0.38 ng g À1 B[a]P (WWB), satisfying the DG SANCO 31 and FDA 32 dened acceptance criteria.

Fig. 1
Fig. 1 Chemical structures of B[a]P and its deuterated analogue.

Fig. 2
Fig. 2 Dimensions of the custom-made manifold.

Fig. 3
Fig. 3 Boxplot of B[a]P concentrations in 3R4F reference tobacco extract expressed on a Wet Weight Basis (WWB) in ng g À1 (1) without hydration (2) 1 : 1 ratio of water : tobacco hydration and equilibration at 5 AE 2 C overnight (3) 4 : 1 water : tobacco hydration with equilibration for one hour at 22 AE 2 C.
[a]P extraction efficiencies.A 3R4F cigarette tobacco was extracted (1) without hydration; (2) at a ratio of 1 : 1 water : tobacco for 18 hours/ overnight at in at 5 AE 2 C, and (3) at a ratio of 4 : 1 water : tobacco with an hour hydration at ambient temperature.In addition, the efficiency of extraction of B[a]P from STP Oliver Twist was evaluated without hydration and with hydration at a ratio of 1 : 1 water : tobacco for 18 hours/overnight at 5 AE 2 C.

Fig. 4
Fig. 4 HPLC fluorescence chromatograms of B[a]P and D 12 -B[a]P in standard stock solutions (A) and in Tanzanian fire-cured tobacco (B), Granit White snus (C) and 3R4F reference tobacco after fortification with 5 ng mL À1 D 12 -B[a]P.t r , retention time (min).
At a hydration ratio of 4 : 1 water : tobacco and equilibration time of one hour, the extraction efficiency of B[a]P was 9.6% lower than the B[a]P concentration determined in non-hydrated samples and 20% lower than in extracts obtained from 3R4F matrix hydrated at a 1 : 1 water : tobacco ratio.Preliminary data indicated that over hydration at a ratio of 4 : 1 water : tobacco signicantly reduced the extraction efficiency for 3R4F compared to a 1 : 1 water : tobacco ratio.It is however worth noting the differences in the equilibration time will also inuence the extraction efficiency.Sample selectivityB[a]P and D 12 -B[a]P were chromatographically separated (Fig.4) and baseline resolved.Assessment of sample matrices with and without fortication with D 12 -B[a]P enabled the optimization of chromatographic conditions and analytical selectivity.

Table 1
High-performance liquid chromatography gradient profiles

Table 2
36dration optimisation conditions and approximate moisture content determined by near infrared spectroscopy36

Table 3 B
[a]P fortification standard solutions in acetonitrile

Table 4
Levels of fortification of matrix with B[a]P standard

Table 5
Comparison of linearity of calibration between solvent and standard fortified matrix samples a Data from instrument 1. b Data from instrument 2.

Table 6
B[a]P concentrations obtained by HPLC-FLD a a WWB, wet weight basis; DWB, dry weight basis.

Table 7
Accuracy and precision of B[a]P solvent-based calibration standards a a RSD, relative standard deviation.This journal is © The Royal Society of Chemistry 2015 Anal.Methods, 2015, 7, 1590-1599 | 1595 Paper Analytical Methods Open Access Article.Published on 07 January 2015.Downloaded on 8/25/2024 2:02:18 PM.This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

Table 8
Uncertainty of measurement for B[a]P in fortified Granit White snus