Determination of low concentrations of benzene in urine using multi-dimensional gas chromatography

Abstract

A method for the determination of benzene in urine of occupationally or environmentally exposed persons was developed. The method was based on dynamic headspace, preconcentration on a solid sorbent, followed by thermal desorption and gas chromatographic determination. To achieve sufficient selectivity, we used multi-dimensional gas chromatography in combination with the inexpensive and robust flame ionisation detector. The limit of detection was 7 ng l⁻¹ and the limit of quantification was 23 ng l⁻¹. The linearity was good (correlation coefficient 0.999) in the range examined (20–4000 ng l⁻¹) and the repeatability was 9%. The average recovery at low concentrations (20–400 ng l⁻¹) was 86%. Analysis of a certified reference material of benzene in water, traceable to NIST, did not differ significantly from the certified value. Samples, frozen (−20 °C) in glass bottles sealed with Teflon–silicon septa, were stable for 1 year and refrigerated samples (4 °C) for at least 1 week. Loss of benzene during the collection and transfer of urine was investigated and found to be acceptable. The method is a cost effective and robust alternative to GC-MS and permits reliable quantification of occupational exposure and, in most cases, also of urine concentrations that can be expected from environmental exposure.

---

Introduction

Benzene is a haematotoxic substance that can induce acute myeloid leukaemia and has been classified as a carcinogen in humans by the IARC.¹ It is a widespread pollutant and the general population is more or less permanently exposed.² The major sources of benzene exposure are vaporisation and incomplete combustion of petrol, industrial processes, smoking and wood firing. Hence there is a need for methods for the assessment of both occupational and environmental exposure.

The metabolic pathways of benzene are complex, and the resulting metabolites offer many possible biological markers.^3–5 Urinary benzene is a promising biomarker because it is specific and reflects air exposure even at low levels.^6–8 The general analytical approach has been either static^9,10 or dynamic¹¹ headspace followed by gas chromatographic separation and photoionisation^9,10 or flame ionisation detection¹¹ (FID). In our laboratory, we previously developed a method based on dynamic headspace, preconcentration of the analyte on a solid sorbent and subsequent thermal desorption and gas chromatographic–mass spectrometric (GC-MS) analysis.¹² More recently, methods based on solid phase microextraction (SPME) in the headspace mode have been presented.^13,14

Benzene is a volatile compound with a low partition coefficient between water and air.^15,16 This suggests the possibility of losses during sampling and storage of urine. With few exceptions,¹² little attention has been paid to this problem, although it might explain in some part the conflicting results from different research groups.¹⁷ In this study we performed further experiments to clarify the possibility of losses during the collection and storage of urine samples.

It is essential in the determination of benzene in urine at low concentrations to have a method which offers both high selectivity and sensitivity. These goals can be achieved by using capillary gas chromatography connected to a mass selective detector but, compared with a flame ionisation detector, this detector is considerably more expensive and also less robust. In this study we demonstrate another strategy based on multi-dimensional gas chromatography to achieve sufficient selectivity and normal FID to offer the sensitivity needed.

Experimental

The determination of benzene in urine is based on dynamic headspace and preconcentration of the analyte on a solid sorbent, followed by thermal desorption and multi-dimensional gas chromatography with flame ionisation detection.

Equipment

The dynamic headspace equipment has been described elsewhere¹² and consisted of a 150 ml Erlenmeyer flask placed on a magnetic stirring device. The effluent gas was dried by a permeable membrane dryer (Perma Pure MD-125-12T; Perma Pure Products, Toms River, NJ, USA) in series with a glass tube filled with about 600 mg of magnesium perchlorate hydrate. The adsorbent tube was filled with about 240 mg of 2,6-diphenyl-p-phenyl oxide (Tenax TA, 60–80 mesh; Chrompack, Bergen op Zoom, The Netherlands).

For the analysis we used an automatic thermal desorber (ATD 400; Perkin-Elmer, Norwalk, CT, USA) coupled to a gas chromatograph (AutoSystem, Perkin-Elmer) equipped with a multi-dimensional column system and a flame ionisation detector. The configuration is shown in Fig. 1.

Fig. 1 Schematic diagram of the instrumentation.

After primary desorption from the sample tube, the analyte was cryofocused on a cold trap filled with 90 mg of Tenax TA. During the secondary desorption of the analyte from the cold trap, the gas flow was split before entering the gas chromatograph to reduce peak tailing. The analyte was transferred to column A, the outlet of which was attached to a two-position, six-port switching valve (Valco, Houston, TX, USA) placed in the GC oven. A restrictor column was used with a pressure drop equal to that of column B. During the heart cut, the carrier gas flow was switched from the restrictor column to column B. When the transfer of the analyte was complete, the valve was switched back to its initial position. The columns were placed in the same oven and subjected to the same temperature programming. Column A was of medium polarity with a polyethylene glycol stationary phase (CP-Wax 57CB, 50 m × 0.32 mm id, 1.2 μm film thickness; Chrompack) and column B was a porous layer (styrene–divinylbenzene polymer) open tubular column (Poraplot Q-HT, 25 m × 0.32 mm id, 10 μm film thickness; Chrompack). Helium was used as the carrier gas and the column flow rate was 2 ml min⁻¹.

Chemicals

Benzene, anhydrous sodium sulfate and magnesium perchlorate hydrate (analytical-reagent grade) were purchased from Merck (Darmstadt, Germany) and methanol (purge and trap grade) from Aldrich (Milwaukee, WI, USA). Nitrogen (AGA, Göteborg, Sweden) was further purified by a cartridge filled with a molecular sieve and activated charcoal.

The certified reference material used was BTEX in water (Lot 96027, Environmental Resource Associates, Arvada, CO, USA), containing benzene, toluene, ethylbenzene and xylene. The reference material is traceable to NIST SRM 1586-1.

Procedures

The urine samples were collected in 250 ml polyethylene bottles and then immediately transferred to 125 ml glass bottles (referred to below as sample bottles) sealed with Teflon–silicon septa (Supelco, Bellefonte, PA, USA). The bottles were kept in a refrigerator (+4 °C) if they were to be analysed within a few days or frozen (−20 °C ) if they were to be stored for longer periods. Before analysis, the content in the sample bottles was brought to ambient temperature.

Urine standards were prepared by spiking urine from non-smoking subjects with methanol solutions of benzene. The spiking concentrations ranged from 20 to 4000 ng l⁻¹. The urine used for the standard preparation was also analysed, and the calibration curve was corrected for its benzene content. The background levels were in the range 5–25 ng l⁻¹.

The sample work-up has been described in detail elsewhere.¹² After 20 g of anhydrous sodium sulfate and a magnetic stir bar had been added to the headspace vessel, 50 ml of urine were transferred from the sample bottle to the headspace vessel. Nitrogen was swept over the urine surface at a flow rate of 200 ml min⁻¹ for 15 min. The water-saturated headspace gas, containing benzene vapours, was dried in two steps before passing to the collecting adsorbent.

The adsorption tube was desorbed at 250 °C for 5 min with the cold trap kept at −30 °C. After completed desorption, the temperature of the cold trap was raised to 250 °C and the analyte was transferred to the gas chromatograph. The split flow during this operation was optimised at 4 ml min⁻¹ to give sufficient detector response but insignificant peak tailing, which facilitated a short heart cut with a complete transfer of the analyte. The temperature was kept at 70 °C for 11 min, then raised to 170 °C at 20 °C min⁻¹ and held at that temperature for 10 min. Finally, the temperature was raised to 200 °C at 20 °C min⁻¹ and held at that temperature for 5 min. The heart cut was made between 9.5 and 11.0 min. The detector temperature was 300 °C.

Statistical methods

Student’s t-test was used for the statistical analysis, except for the storage stability test where Dunnet’s multiple comparisons test¹⁸ was used.

Results and discussion

The dynamic headspace benzene transfer step

Breakthrough on the adsorption tube, which would lead to underestimation of the benzene concentrations, was investigated by connecting two tubes in series during the dynamic headspace procedure. The results showed the necessity for effective drying of the headspace gas before it entered the adsorbent tube. When a permeable membrane dryer with reduced drying capacity (final relative humidity 42% compared with <20% normally) was used, breakthrough was about 10% in the range 20–4000 ng l⁻¹ benzene. After the defective drying device has been replaced by one with normal capacity, no breakthrough could be observed. As a consequence, daily control of the drying capacity of the membrane dryer and replacement of the magnesium perchlorate were introduced.

Carryover in the headspace equipment was investigated by first analysing blank urine, then spiked urine samples with a high benzene content and then, immediately, blank samples again. All analyses were made in duplicate. The analysed blank values were approximately the same after the analyses of spiked samples with a benzene concentration of 500 ng l⁻¹ as before (average value 23.6 ng l⁻¹ compared with 22.6 ng l⁻¹ before). After two samples with a benzene concentration of 4000 ng l⁻¹, the blank value was significantly, but not seriously, higher (28.9 ng l⁻¹). We conclude that carryover will not be a problem in most real situations, where the concentration range of the analysed samples will be much narrower.

GC determination

Urine is a complex matrix, containing a large variety of compounds. Headspace techniques generally reduce the interference from non-volatile compounds, but there can still be problems of interference from other, volatile compounds evaporating from the urine.

Selectivity can be achieved by using mass selective detection,¹² but an alternative solution is the use of multi-dimensional gas chromatography in combination with the less selective but more robust and less expensive FID. Ideally, the resolving power of a two-dimensional separation system is the product of the peak capacities of the two columns combined in the system.^19,20 In comprehensive two-dimensional gas chromatography, the entire sample introduced into the primary column is submitted to the secondary column for a second independent separation.^21,22 This is close to the ideal solution and it can be argued that the term ‘two-dimensional GC’ should be restricted to these kind of systems. In a conventional multi-dimensional separation system based on heart cutting, the resolving power is the sum of the peak capacities of the columns. However, this is true only if the width of the fraction transferred to the second column is equal to the width of the plug injected on the first column and no correlation exists between the retention mechanisms on the two phases. However, the band broadening on the first column in an isothermal run will lead to broader peaks, which reduces the peak capacity of the second column. Furthermore, a real heart cut must always be broader than the peak width because the absolute retention time will fluctuate with changes in gas flow rate, column temperature and the amount of co-eluting components.²³ A too-narrow cut can result in incomplete transfer of the fraction of interest. A broader cut will transfer more, initially separated, components to the second column, where they may be remixed. Assuming that the resolving power of the second column is equal to that of the first, the selectivity of the second column for the components in the heart cut must be good to make it possible to resolve the components a second time. This makes it difficult to find a proper combination of columns, even if they meet the basic demand of having different selectivities. Our strategy was first to combine a non-polar column A with 5% phenyl–95% dimethylpolysiloxane phase with columns with phases of increasing polarity, but a satisfactory separation was not achieved until we used an aluminium oxide PLOT column (Al₂O₃/KCl PLOT; Chrompack) as column B. However, repeated analyses in this system resulted in increased baseline noise and drifting retention times. This was corrected by conditioning the column B at high temperature for ∼2 h, indicating that water from the headspace procedure caused the problem. Because of these problems, we changed our strategy and used a polar column A with a polyethylene glycol phase (CP-Wax 57CB), in which water was eluted to waste after the heart-cut procedure, and a non-polar column B, which was a water stable and low bleeding PLOT column (Poraplot Q-HT). This combination proved to be the best compromise in terms of separation efficiency, analysis time and long-term stability. During the first separation the temperature was kept isothermal at 70 °C to give an optimal separation. At that temperature benzene and co-eluting interferents had negligible or no migration on the PLOT column and were trapped on the top of column B. A possible separation of the transferred components on column A will therefore be lost, but instead the band broadening of the peaks will be less. Fig. 2 shows a GC-MS chromatogram using only column A. The retention times of benzene and also 3-methyl-2-butanone, which elutes just in front of benzene, are indicated. The heart cut will include this interferent as well as other minor compounds. As the concentration of the analyte decreases, the problems with interfering substances will be more pronounced.²⁴ However as shown in Fig. 3, benzene was well resolved from the interferents in the final separation.

Fig. 2 A GC-MS total ion chromatogram showing the separation of volatile compounds in a urine sample after the first column (column A) in a multi-dimensional chromatographic system. The retention time of benzene, the time interval of the transfer to column B and the main interferent are indicated.

Fig. 3 Chromatogram for the determination of benzene in urine with a multi-dimensional chromatographic system and flame ionisation detection. The benzene concentration was 24 ng l⁻¹.

Both columns were placed in the same GC oven. This is an inexpensive solution in comparison with the ideal solution, in which the columns are placed in different ovens, which permits optimisation of each set of separation parameters. To minimise band broadening and the risk of contamination, a valve configuration in which the analyte passed through the valve only once was chosen. This configuration does not permit backflushing of column A. However, during the separation on column B, occurring at a higher temperature, heavier compounds will be eluted from column A to waste, creating a chromatographic system without memory effects.

In summary, the selectivity of the multi-dimensional gas chromatographic system was high, even with non-selective FID. This is in agreement with the opinion expressed by Huber et al.,²⁵ who concluded that the information content of retention data from a multi-dimensional gas chromatographic system equals the information content from a low-resolution GC-MS system. A multi-dimensional gas chromatographic system can be laborious to set up and requires skilled operators, but it has the advantages of being less expensive and more robust than a GC-MS system. So far, we have analysed hundreds of samples with a minimum of problems.

Quantification

Linearity was assessed by duplicate analyses of eight spiked urine standards (n = 16) in the range 20–4000 ng l⁻¹. The linearity was very good in the concentration range studied, with a slope of 2211, an intercept of −786 and a correlation coefficient of 0.999. The 95% confidence interval (CI) of the intercept included the origin. The repeatability of duplicate analyses of urine from 12 smokers in the range 15–300 ng l⁻¹ was 9.0%. Samples in the range 15–100 ng l⁻¹ had approximately the same RSD as samples in the range 100–300 ng l⁻¹ (9.3 and 8.9%, respectively).

The recovery at low concentrations (50–400 ng l⁻¹) was investigated by analysing spiked urine samples together with adsorbent tubes spiked directly. The average recovery was 86%, with a slight tendency for increasing recovery over the interval from 50 to 400 ng l⁻¹. This is in agreement with previously reported results.^11,12

Determination of limit of detection (LOD) and limit of quantification (LOQ) was made according to a procedure proposed by NIOSH,²⁶ which includes both sample work-up and analysis. The LOD was 7.0 ng l⁻¹ and the LOQ 23 ng l⁻¹. These are in very close agreement with the performance of the GC-MS method described previously,¹² even if the instrumental LOD of MS is about an order of magnitude lower than the LOD of FID. The most probable explanation is that the headspace procedure, which is common to both methods, is the main source of variation.

The trueness of the method was investigated by analysing a certified reference sample of benzene in water, traceable to NIST SRM 1586-1. The analysis was performed using a calibration curve based on spiked urine. The result is shown in Table 1. The average recovery compared with calculated concentrations was 105%. This difference is not significantly different, because the 95% CI of the average recovery was 98–112%.

Table 1 Analysis of a certified reference material of BTEX in water, traceable to NIST SRM 1586-1

Calculated concentration^a/ng l^−1	Analysed concentration/ng l^−1	Recovery (%)
a The calculation is based on the certified reference value and the final dilution of the sample.
49.7	52.2	105
49.7	57.4	115
49.4	50.3	102
149.3	153.9	103
151.1	150.3	99

---

Method comparison

The proposed method was compared with a GC-MS method that was developed previously in our laboratory.¹² The dynamic headspace procedure is common to both methods. Urine from 10 smokers was collected, divided into four sample bottles and frozen. From these sets, two randomly chosen bottles were analysed by each method. The GC-MS analyses were made 3 weeks after the analyses using the proposed method. The result is shown in Fig. 4. The data points fit well to a regression line except for one outlier. With this outlier excluded, the regression equation was y = 0.95x + 12 and the correlation coefficient was 0.985. The 95% CI for the slope and intercept included 1 and the origin, respectively, which means that there is no significant difference between the two methods. The outlier cannot be explained, but it certainly does not depend on the repeatability of the methods, because the RSD of the duplicate analyses of this specific urine sample was <7% for both methods. Because the analyses were performed by two different chemists and at different times, the deviation might be a result of differences regarding sample handling before the analysis.

Fig. 4 Comparison between multi-dimensional gas chromatographic determination (ordinate) and GC-MS determination (abscissa) of benzene in urine. Each data point represents the average of two analyses. The filled triangle represents an outlier.

Sample collection and storage

It has been shown that loss of benzene during collection and initial transfer from the polyethylene bottles to sample bottles was 6% if the urine was transferred immediately and 12% if the transfer was made after 1.5 h.¹² However, these experiments were made using cold spiked urine, a situation different from the real situation, where warm urine is voided into plastic bottles. To investigate this problem further, a series of experiments were conducted with urine samples from two smokers with urinary benzene concentrations ranging from 20 to 500 ng l⁻¹. Urine voided into 250 ml polyethylene bottles was transferred to sample bottles immediately or after 4–8 h of storage.

In the first experiment, the subjects were asked first to void directly into a sample bottle and then to shift to a polyethylene bottle. The urine from the latter bottle was then immediately transferred to a sample bottle by the subjects. This experiment confirmed earlier findings that the loss of benzene during the transfer from the polyethylene bottle to the sample bottle was moderate (average 3.9%, s 12%, n = 8).

In the second experiment, the subjects voided consecutively into two different polyethylene bottles to avoid losses during repeated transfers to sample bottles. Urine from one of the polyethylene bottles was immediately transferred to the sample bottle and urine from the other polyethylene bottle was transferred after 4 or 8 h of storage. The losses after 4 h and 8 h of storage were of the same magnitude (average 26%, s 4.0%, n = 4 and average 20%, s 18%, n = 9, respectively). The polyethylene bottles were filled with approximately 140 ml of urine, leaving a headspace volume of 110 ml. The calculated loss to the headspace volume was estimated to be 16–21%, based on reported water–air partition coefficients.^15,16 This loss is of the same magnitude as the experimental values. However, the standard deviations were considerable in all experiments, indicating that transfers should always be made with great care.

The consequences of these results are two alternative sampling procedures. The convenient wide-necked polyethylene bottles can be used for the collection of urine without significant loss of benzene if the urine is transferred immediately to storage bottles of glass sealed with Teflon lined caps. If this is not feasible, voiding directly into glass sample bottles is preferred. We used 250 ml glass bottles (Schott Duran, KEBO, Spånga, Sweden) with a 45 mm thread (38 mm id) and Teflon lined caps for self-administered sampling in environmentally exposed subjects. To avoid losses to the headspace, storage bottles should be filled completely, if possible.

The general weakness of a thermal desorption step is that it permits only one analysis. The best way to compensate for this type of problem is to collect a sufficient volume of urine and to perform the whole analytical procedure in duplicate. The 125 ml sample bottles permitted duplicate analyses of two 50 ml aliquots. However, the results of the second analyses were usually lower (average 9.7%, s 10%, n = 37), because of loss of benzene to the increased headspace volume after the transfer of the first aliquot. As a consequence, we recommend the use of the result of the first analysis and the result of the second analysis as a check. Alternatively, the result of the second analysis could be corrected for the losses to headspace and the average of both analyses calculated.

The storage stability of spiked urine in sample bottles at a concentration of 875 ng l⁻¹ at 4 °C for 3 days and at −20 °C for 1 month has been shown to be good, with a loss of <2%.¹² We now investigated the stability at a lower benzene concentration by filling 24 sample bottles with urine spiked with benzene to a final concentration of about 40 ng l⁻¹. The bottles were randomly divided into sets of six. The first set was analysed immediately and the second set was analysed after being stored for 7 days at about 4 °C. The remaining sets were frozen and stored at−20 °C for 5 weeks and 12 months, respectively, and then analysed. The results are shown in Table 2. The differences were not statistically significant.

Table 2 Storage stability of urine samples spiked with benzene in 125 ml glass sample bottles, sealed with Teflon lined caps under different conditions and for different periods

Storage conditions	Average concentration/ ng l^−1	s/ng l^−1	RSD	n
Direct analysis	37	4.3	0.12	5
7 d at 4 °C	39	4.5	0.12	6
5 weeks at −20 °C	37	6.6	0.18	6
12 months at −20 °C	40	4.1	0.10	6

---

Acknowledgements

This work was financially supported by the Swedish Environmental Protection Agency and by the Foundation for Strategic Environmental Research (MISTRA). The authors also thank Associate Professor Lars Barregård for his contributions during the investigations and for his valuable comments on the manuscript.

References

1. IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, International Agency for Research on Cancer, Lyon, 1982, vol. 29, pp. 93–148..
2. L. Wallace, Environ. Health Perspect., 1996, 104, Suppl. 6, 1129..
3. C.-N. Ong and B.-L. Lee, J. Chromatogr., B, 1994, 660, 1.
4. K. H. Schaller, in Update on Benzene. Advances in Occupational Medicine and Rehabilitation, ed. M. Imbriani, S. Ghittori, G. Pezzagno and E. Capodaglio, PI-ME Press, Pavia, 1995, vol. 1, No. 2, pp. 135–143..
5. A. M. Medeiros, M. G. Bird and G. Witz, J. Toxicol. Environ. Health, 1997, 51, 519.
6. C. N. Ong, P. W. Kok, B. L. Lee, C. Y. Shi, H. Y. Ong, K. S. Chia, C. S. Lee and X. W. Luo, Occup. Environ. Med., 1995, 52, 528.
7. C. N. Ong, P. W. Kok, H. Y. Ong, C. Y. Shi, B. L. Lee, W. H. Phoon and K. T. Tan, Occup. Environ. Med., 1996, 53, 328.
8. F. Dor, W. Dab, P. Empereur-Bissonnet and D. Zmirou, Crit. Rev. Toxicol., 1999, 29, 129.
9. K. Pekari, M.-L. Riekkola and A. Aito, J. Chromatogr., B, 1989, 491, 309.
10. P. W. Kok and C. N. Ong, Int. Arch. Occup. Environ. Health, 1994, 66, 195.
11. S. Ghittori, M. L. Fiorentino, L. Maestri, G. Cordioli and M. Imbriani, J. Toxicol. Environ. Health, 1993, 38, 233.
12. G. Ljungkvist, M. Lärstad and L. Mathiasson, J. Chromatogr., B, 1999, 721, 39.
13. S. Fustinoni, R. Giampiccolo, S. Pulvirenti, M. Buratti and A. Colombi, J. Chromatogr., B, 1999, 723, 105.
14. R. Andreoli, P. Manini, E. Bergamaschi, A. Brustolin and A. Mutti, Chromatographia, 1999, 50, 167.
15. T. Lindqvist, The Partition Coefficients Blood/Air and Water/Air for Some Common Solvents, Arbete och Hälsa 1977∶8, Arbetarskyddsverket, Stockholm, 1977..
16. A. Sato and T. Nakajima, Br. J. Ind. Med., 1979, 36, 2.
17. L. Perbellini, M. Buratti, M. L. Fiorentino, S. Fustinoni, F. Pasini and S. Magnaghi, J. Chromatogr., B, 1999, 724, 257.
18. D. C. Montgomery, Design and Analysis of Experiments, Wiley, New York, 2nd edn., 1984..
19. J. C. Giddings, in Multidimensional Chromatography Techniques and Applications, ed. H. J. Cortes, Marcel Dekker, New York, 1990, pp. 1–27..
20. W. Bertsch, J. High Resolut. Chromatogr., 1999, 22, 647.
21. J. B. Phillips and J. Beens, J. Chromatogr., A, 1999, 856, 331.
22. W. Bertsch, J. High Resolut. Chromatogr., 2000, 23, 167.
23. K. Himberg, E. Sippola, F. David and P. Sandra, J. High Resolut. Chromatogr., 1993, 16, 645.
24. J. W. Jorgensen, in Proceedings of the 16th International Symposium on Capillary Chromatography, Riva del Garda, 1994, ed. P. Sandra and G. Devos, Hüthig, Heidelberg, 1994, p. xxxix, and personal communication..
25. J. F. K. Huber, E. Kenndler and G. Reich, J. Chromatogr., 1979, 172, 15.
26. E. R. Kennedy, T. J. Fischbein, R. Song, P. M. Eller and S. A. Shulman, Guidelines for Air Sampling and Analytical Method Development and Evaluation, DHHS (NIOSH) Publication No. 95-117, NIOSH, Cincinnati, OH, 1995..

---