==== Front Ind HealthInd HealthINDHEALTHIndustrial Health0019-83661880-8026National Institute of Occupational Safety and Health, Japan 2017-022510.2486/indhealth.2017-0225Original ArticleComparison of personal air benzene and urine t,t-muconic acid as a benzene exposure surrogate during turnaround maintenance in petrochemical plants KOH Dong-Hee 1*LEE Mi-Young 2CHUNG Eun-Kyo 2JANG Jae-Kil 2PARK Dong-Uk 31 Department of Occupational and Environmental Medicine, International St. Mary’s Hospital, Catholic Kwandong University, Korea2 Occupational Safety and Health Research Institute, Korea Occupational Safety and Health Agency, Korea3 Department of Environmental Health, Korea National Open University, Korea* To whom correspondence should be addressed. E-mail: koh.donghee@gmail.com12 4 2018 7 2018 56 4 346 355 30 12 2017 09 4 2018 ©2018 National Institute of Occupational Safety and Health2018This is an open-access article distributed under the terms of the Creative Commons Attribution Non-Commercial No Derivatives (by-nc-nd) License. (CC-BY-NC-ND 4.0: https://creativecommons.org/licenses/by-nc-nd/4.0/)Previous studies have shown that biomarkers of chemicals with long half-lives may be better surrogates of exposure for epidemiological analyses, leading to less attenuation of the exposure-disease association, than personal air samples. However, chemicals with short half-lives have shown inconsistent results. In the present study, we compared pairs of personal air benzene and its short-half-life urinary metabolite trans,trans-muconic acid (t,t-MA), and predicted attenuation bias of theoretical exposure-disease association. Total 669 pairs of personal air benzene and urine t,t-MA samples were taken from 474 male workers during turnaround maintenance operations held in seven petrochemical plants. Maintenance jobs were classified into 13 groups. Variance components were calculated for personal air benzene and urine t,t-MA separately to estimate the attenuation of the theoretical exposure-disease association. Personal air benzene and urine t,t-MA showed similar attenuation of the theoretical exposure-disease association. Analyses for repeated measurements showed similar results, while in analyses for values above the limits of detection (LODs), urine t,t-MA showed less attenuation of the theoretical exposure-disease association than personal air benzene. Our findings suggest that there may be no significant difference in attenuation bias when personal air benzene or urine t,t-MA is used as a surrogate for benzene exposure. Benzenet,t-muconic acidAttenuationBiomarkerExposure ==== Body Introduction Benzene is a potent carcinogen that causes acute non-lymphocytic leukemias and is suspected to be associated with lymphocytic leukemias, multiple myeloma and non-Hodgkin’s lymphoma1). It also suppresses bone marrow, resulting in serious hematological adverse effects such as pancytopenia and aplastic anemia2,3,4). Benzene exposure has occurred across a wide range of industries5). In assessing occupational and environmental exposures, personal air samples and biomarkers have been used. Personal air sampling is used to measure external exposures, but it usually does not reflect moderating factors such as respirator use or dermal absorption6). Biomarkers are used to measure internally absorbed doses accounting for the moderating factors, but these are also affected by personal factors such as cigarette smoking and food intake7). In epidemiological analyses, both personal air samples and biomarkers have been used to evaluate associations between exposures and diseases. Many studies have sought better indicators with which to examine exposure-disease association6, 8,9,10). Some studies that analyzed multiple datasets containing both personal air samples and biomarkers have shown that biomarkers may be a less-biasing surrogate than personal air samples in terms of attenuation of theoretical exposure-disease associations, especially in chemicals with long half-lives, such as lead6, 8). The attenuation bias represents an underestimation of the true exposure-disease association due to measurement error11). However, the notion that biomarker is a less biasing surrogate is inconsistent for chemicals with short half-lives. t,t-MA is a urinary benzene metabolite with a short half-life (5 h)12). Urine t,t-MA has been used as a surrogate of benzene exposure in epidemiological studies, but it has not yet been tested for whether personal air benzene samples or urine t,t-MA is a better surrogate of benzene exposure in terms of attenuation of the exposure-disease association. In the present study, we estimated attenuation of the theoretical exposure-disease association for personal air benzene and its urinary metabolite t,t-MA to determine which is a better surrogate of benzene exposure for use in future epidemiological analyses. Methods Study subjects The study subjects consisted of 474 male workers who participated in turnaround maintenance operations held in seven petrochemical plants13, 14). During turnaround maintenance, workers can be exposed to various carcinogenic chemicals such as benzene and 1,3-butadiene15). The plants are located in a refinery/petrochemical complex in Korea and have produced or used benzene. The turnaround maintenance operations were held in manufacturing processes in which benzene exposure can occur: styrene monomer (SM; one plant), cumen (two plants), mono-nitrobenzene (MNB; two plants), and aromatic processes (BTX; two plants). Study subjects were workers who participated in the benzene-exposed turnaround maintenance operations in the seven petrochemical plants. The study subjects can be largely classified into temporary maintenance workers and employees of petrochemical plants. Temporary maintenance workers were hired on fixed-term contracts and engaged in petrochemical plant maintenance operations16). Jobs of temporary maintenance workers were categorized into eight jobs: insulation fitter, plumber, mechanical engineer, scaffold worker, drain fitter, construction craftsman, electrical fitter, welder, painter, and tank fitter. Jobs of employees of petrochemical plants were categorized into three jobs: board man, field man, and maintenance engineer. Overall the jobs of the study subjects were categorized into 13 groups a priori. The major activity of each job is described in detail elsewhere13). The study subjects were supposed to be directly or indirectly exposed to benzene during turnaround maintenance operations. During turnaround maintenance operations, workers were randomly selected and asked to wear personal air samplers and provide urine samples, which were used to evaluate their benzene exposure levels. Shift-long personal air benzene sampling and post-shift urine sampling started in October of 2007 and ended in January of 2009. Personal air samples and urine samples were collected simultaneously at the work sites immediately after the work has finished. Overall, 669 pairs of personal air benzene and urine t,t-MA samples were collected from 474 workers. Among them, 122 workers wore personal air samplers and provided urine samples two or more times on different days. The time differences between the first and last measurements for the 122 workers were skewed with a median of 2 d (interquartile range (IQR): 1–24 d, max 215 d). The average number of repeated measurements was 2.6 times (max: 6). Personal air benzene measurements Shift-long passive sampling by organic vapor monitors (OVM 3500, 3M, USA) was employed to measure personal air benzene concentrations during turnaround operations. Personal air benzene samples were collected immediately after the work has finished at the work sites from the workers. The mean sampling time and standard deviation (SD) were 441 min and 78 min, respectively, with a range of 198 to 693 min. All personal air benzene samples were analyzed according to the NIOSH Manual of Analytic Methods 1500 and 1501 using GC/FID (Gas chromatography/flame ionization detector) in a laboratory at the Occupational Safety and Health Research Institute (OSHRI). OSHRI is an accredited laboratory from PAT (Proficiency Analytic Testing) program of the AIHA (American Industrial Hygiene Association). To validate the performance of the passive sampler, active sampling using charcoal tubes (SKC 26–01, SKC Inc., USA) was carried out in a pilot study. A pump (Gillian LFS-113, USA) calibrated to <0.2 l/min was used to estimate external inhalation benzene exposure during turnaround operations. These two sampling methods showed a strong correlation (r=0.95, p<0.001)13). Details of the sampling and analytical procedures used in this study are described in the literature13). t,t-Muconic Acid measurements Post-shift urine samples were collected in high-density polyethylene (HDPE) containers at the work sites immediately after the work has finished, and stored in a refrigerator at 4°C. All urine t,t-MA samples were analyzed in a laboratory of the OSHRI. Urine t,t-MA analyses were performed by the following processes. Collected samples and standard solutions were purified by solid phase extraction using Bond Elut SAX (500 mg, Varian, USA) as an anion exchange cartridge. After applying 1,000 µl of sample or standard solution mixed with 5 µl of 10 mM sodium hydroxide solution, the cartridges were washed with 2 ml of 1% aqueous acetic acid twice. t,t-MA was eluted with 3 ml of 10% aqueous acetic acid and collected into 5-ml measuring test tubes, which were then filled with an eluent of water washed through the cartridge. An aliquot (80 µl) of eluent filtered through a 0.22-μm membrane filter was directly injected into an HPLC system for determination of t,t-MA levels. A Capcell Pak MF Ph-1 SG80 column (150 × −4.6 mm I.D., 5 μm, Shisheido, Japan) was employed for the separation process. The mobile phase used was 10 mM KH2PO4 containing 0.1% H3PO4 at a flow rate of 0.5 ml/min. The column temperature was held constant at 40°C, and UV detection was performed at 259 nm. G-EQUAS urine samples of round robin 26, 29 and 32 were analyzed as a reference sample for it, t-MA. The accuracy was calculated from the recovery of t,t-MA, which was 82−129% of reference samples in the range of 0.41–2.91 mg/l. The linearity was good with the correlation constant of 0.9999 in the range of 0.1−5.0 g/l. Limit of detection (LOD) was calculated from calibration data of standard solutions of t,t-MA in the range of 0.5–5.0 mg/l17). In a pilot study, pairs of pre-shift and post-shift urine samples from 15 workers were analyzed, and post-shift t,t-MA levels were significantly higher than pre-shift t,t-MA levels (mean difference 1.18 mg/g creatinine, p<0.001). Statistical analysis To examine which is a less-biasing surrogate of benzene exposure for epidemiological analyses, we computed attenuation of the theoretical exposure-disease association both for personal air benzene and its urinary metabolite t,t-MA18, 19). Generation of complete data sets using a multiple imputation technique Of the personal air benzene samples, 34.1% were below the 0.012 ppm of LOD. Of the urine t,t-MA samples, 10.9% were below the 0.06 mg/l of LOD. In the first step, we generated complete data sets for subsequent analyses using a multiple imputation technique20). We created a bootstrap data using random sampling with replacement. Then, Tobit regression was performed to obtain distribution parameters of measurements. We imputed values below the LOD assuming that measurements were log-normally distributed defined by estimates of distribution parameters from the Tobit regression. This imputation procedure was repeated five times to obtain five complete data sets, consisting of imputed values for measurements below the LOD and real values above the LOD. The imputation process was conducted for personal air benzene and urine t,t-MA separately. Tobit regression was performed using ‘AER’ package21) of statistical software R22). Calculation of variance components In the second step, we calculated the pooled variance components across five imputed data sets to be used for estimating attenuation of the theoretical exposure-disease association. Personal air benzene and urine t,t-MA levels in the five complete data sets were natural log-transformed (Ln(Y)) to approximate normal distributions. Mixed-effects models were developed separately to calculate variance components for personal air benzene and urine t,t-MA. Mixed-effects models for group-based and individual-based exposure assessment were developed separately23). For group-based analyses in which a group mean is assigned to workers within the group as an exposure estimate in epidemiological analyses, we developed the following mixed-effects model: Ln(Yijk)=α0 + Jobi + Workerij + εijk In this equation, α0 represents the intercept of the model or the mean exposure. Jobi is a random job group effect [Jobi to N (0, σ2BG)] (between-group variance), Workerij is a random effect for the jth worker in the ith job group [Workerij to N (0, σ2WG)] (within-group variance), and εijk is a random effect for the kth measurement of the jth worker in the ith job group [εijk to N (0, σ2WW)] (within-worker variance). We assumed that Jobi , Workerj, and εijk were independent of one another. For individual-based analyses in which individual means are assigned to workers as an exposure estimate in epidemiological analyses, we developed the following mixed-effects model across job groups: Ln(Ykn)=α0 + Workerj + εjk Here, Workerj is a random effect for the jth worker [Workerj to N (0, σ2WG)] and εjk is a random effect for the kth measurement of the jth worker [εjk to N (0, σ2WW)]. We assumed that Workerj and εjk were independent of one another. We fitted each model for five imputed data sets with a restricted maximum likelihood estimation method using ‘lme4’ package24) of statistical software R. Then, we combined the variance components across five complete data sets to obtain pooled variance component estimates, using ‘mitml’ package of statistical software R. Estimation of attenuation of the theoretical exposure-disease association In the third step, we estimated the attenuation of the theoretical exposure-disease associations both for personal air benzene and urine t,t-MA, using pooled estimates of variance components19). We calculated estimates separately for the group-based analysis and the individual-based analysis. For group-based analysis, the expected value of the true exposure-disease coefficient (β^1) can be calculated as follows: E(β^1)=(σ2BG+σ2WGkσ2BG+σ2WGk+σ2WWkn)β1 Here, k and n represent the number of workers for a job group and the number of repeated measurements for a worker, respectively. β1 represents true exposure-disease coefficient. For individual-based analysis, the expected value of β^1 can be calculated as: E(β^1)=(σ2WGσ2WG+σ2WWn)β1 In addition, we repeated above calculations, restricting the analysis to repeated measurements and/or values above LODs of personal air benzene and urine t,t-MA. Correlation between personal air benzene and urine t,t-MA We calculated correlation coefficients between natural log-transformed personal air benzene and natural log-transformed t,t-MA from the five complete data sets. Fisher’s Z-transformation was used to estimate the pooled correlation coefficient because the distribution of a sample correlation coefficient is known to be skewed25). Pooled correlation coefficients were calculated for complete data sets and subset consisting of repeated measurements. In addition, correlation coefficients were further calculated for other subsets, consisting of values above LODs of personal air benzene and urine t,t-MA. Results Distributions of personal air benzene and urine t,t-MA by job and manufacturing process were presented in Table 1Table 1. Summary statistics of personal air benzene and urine t,t-muconic acid by job and manufacturing process N Personal air benzene (ppm) Urine t,t-muconic acid (mg/g creatinine) N LODs (N=395) k 23.0 - 23.0 - n 1.3 1.3 1.3 1.3 G 13 299 13 299 σ2BG 0.14 - 0.07 - σ2WG 0.27 0.23 0.23 0.30 σ2WW 2.20 2.35 0.52 0.53 Attenuation 0.68 0.11 0.82 0.45 Repeated measurements & > LODs (N=155) k 7.4 - 7.4 - n 2.6 2.6 2.6 2.6 G 8 59 8 59 σ2BG 0.21 - 0.10 - σ2WG 0.38 0.29 0.22 0.31 σ2WW 2.49 2.70 0.52 0.53 Attenuation 0.67 0.22 0.83 0.60 k: Average number of workers per job group ; n: average number of measurements per worker; G: the number of groups for group-based analysis or the total number of workers for individual-based analysis; σ2BG: between-group variance; σ2WG: within-group variance; σ2WW: within-worker variance; LODs: limits of detection for personal air benzene and urine t,t-MA. . Personal air benzene showed a similar degree of attenuation with urine t,t-MA in both group-based (0.91 vs. 0.87) and individual-based (0.40 vs. 0.48) analyses. Group-based analyses showed less attenuation than individual-based analyses. In analyses for subset consisting of only repeated measurements, in general, similar patterns were found with complete data set. In analyses for subset consisting of values above LODs of personal air benzene and urine t,t-MA, attenuation was substantially increased for personal air benzene (group-based analysis 0.68; individual-based analysis 0.11), while a relatively small increase was found in urine t,t-MA (group-based analysis 0.82; individual-based analysis 0.45). When further restricting analyses to repeated measurements, similar patterns were found with the subset consisting of values above LODs of personal air benzene and urine t,t-MA. The pooled correlation between personal air benzene and urine t,t-MA showed weak correlations (all measurement 0.24; repeated measurements 0.26) (Table 3Table 3. Correlation between log-transformed personal air benzene and log-transformed urine t,t-muconic acid Dataset N r SE All measurements 669 0.24 0.04 Repeated measurements 317 0.26 0.05 All measurements, personal air benzene>LOD & urine t,t-muconic acid >LOD 395 0.37 0.05 Repeated measurements, personal air benzene>LOD & urine t,t-muconic acid >LOD 155 0.43 0.05 N: number of measurements; r: pooled correlation coefficient; SE: standard error; LOD: limit of detection. ). The correlation became stronger when the analyses were restricted to values above LODs of personal air benzene and urine t,t-MA (all measurement 0.37; repeated measurements 0.43). Discussion In this study, we estimated attenuation of the theoretical exposure-disease association for personal air benzene and its urinary metabolite t,t-MA to examine which is a better surrogate for benzene exposure in epidemiological analyses from a viewpoint of exposure variability. Previous studies6, 8) examining measurement errors of personal air samples and biomarkers showed that, in general, biomarkers of chemicals with long half-lives present less attenuation than air samples. However, because chemicals with short half-lives have shown inconsistent results, and furthermore t,t-MA has not been examined before, here we attempted to compare the theoretical attenuation using both personal air benzene and its short-half-life urinary metabolite t,t-MA19, 28). According to our results, personal air benzene and urine t,t-MA showed a similar degree of attenuation of the theoretical exposure-disease association, which suggests that there may be no significant difference in attenuation bias when personal air benzene or urine t,t-MA is used as a surrogate for benzene exposure in epidemioloigical studies. Group-based analyses showed less attenuation than individual-based analyses, which is consistent with the results of previous studies examining measurement errors19, 23). During a maintenance operation, workers held different jobs work together around benzene exposure sources, but some workers may work far outside the exposure sources. Only including workers near exposure sources may lose variabilities among jobs and among workers, which may explain the decrease in between-group and within-group variance components in analyses for subsets consisting of values above LODs of personal air benzene and urine t,t-MA, thus resulting in an increase of attenuation. Our findings may suggest that assigning mean of only high exposure values for a job or a worker would lead to an increase of attenuation bias. The correlation coefficient between personal air benzene and urine t,t-MA, excluding values below LODs was 0.37, but the correlation using all data resulted in a decreased correlation (0.24). The decrease might be caused mainly by the introduction of the multiple imputation technique, because the imputation process was performed independently for personal air benzene and for urine t,t-MA; thus, imputed values for personal air benzene and urine t,t-MA bear no correlation. Including uncorrelated imputed values in the analysis might lead to decrease in the correlation coefficient. One of the causes of the low correlations may be due to the low benzene exposures. In previous studies, correlation coefficients between personal air benzene and urine t,t-MA reported ranged from 0.4 to 0.912, 29,30,31,32,33,34,35,36,37,38), and these studies usually examined correlations in high-exposure environments. Personal air benzene and urine t,t-MA are well correlated in high-exposure circumstances; by contrast, if personal air benzene concentrations are low; i.e., <1 ppm, urine t,t-MA may show less-strong correlations36, 39,40,41,42) or not be linearly correlated with air benzene concentration, instead showing a constant level43). Other studies reported that urine concentrations of t,t-MA were consistently elevated at or above an air benzene concentration of 0.2 ppm43), or were useful at levels as low as 0.5 ppm44, 45). In our data set, only 9% of personal air benzene samples exceeded 1 ppm; such a high proportion of low concentrations might lead to a poor correlation between personal air benzene and urine t,t-MA. When excluding values below LODs of personal air benzene and urine t,t-MA, the correlation coefficient improved to 0.37, but not substantial. The low correlation might also be explained by other factors, such as smoking, food intake, and genetic polymorphisms39, 46). Smokers usually show higher urine t,t-MA levels than non-smokers, especially in low-level exposure circumstances, which might confound the correlation between air benzene and urine t,t-MA39, 47, 48). Regarding food intake, t,t-MA levels can be substantially influenced by food preservatives and sorbic acid intake, especially at lower air benzene levels (<1 ppm)7, 30, 49, 50). However, in the present study, relevant information was not available. For monitoring of extremely low benzene exposure environments, recent studies reported unmetabolized urine benzene would be a more reliable biomarker than S-phenylmercapturic acid and t,t-MA39, 41, 42). Other causes of the low correlation might be associated with working conditions, such as respirator use and dermal exposure. Most of the workers who participated in the maintenance operations wore respirators. However, in general, highly exposed workers are more likely to wear personal protective equipment properly than less or indirectly exposed workers, which might lead to low urine t,t-MA levels in highly exposed workers and, thus, a low correlation. Our study had several limitations. First, the equations used for estimating attenuation can be applied when the data are balanced; i.e., each job group consists of an equal number of workers and the workers are measured on the same number of occasions51). Our measurement data were highly unbalanced in terms of the number of workers per job group and the number of repeated measurements per worker. However, because it was reported that the equations for attenuation are relatively robust despite the violation of the above assumption52), we assume that our estimates of attenuation are not substantially distorted. Second, the equations used to calculate the attenuation are based on the assumption that no other factors affect the corresponding health outcomes, except for the exposure53). This type of assumption may be unrealistic for most health outcomes, considering the multi-factorial nature of the causes of disease. Therefore, the attenuation formulas may be used to determine which surrogate is better, but they should be used with caution for adjusting the exposure-disease association a posteriori51). Third, we assumed that the first and subsequent measurements are independent, as the half-life of t,t-MA is short (5 h). In previous studies with other chemicals, however, various degrees of serial correlation was observed8, 52). In general, the strength of the correlation is inversely proportional to the time differences between the first and subsequent measurements8, 52). The time differences in our repeated measurement data varied widely. Therefore, we assumed that the correlation structure had compound symmetry. However, considering that introduction of a serial correlation increased attenuation in a previous study8), our estimates of attenuation can be changed to some degree by adopting different correlation structures. Finally, we could not obtain information on important factors, such respirator use, food intake and smoking, which might affect correlations between personal air benzene and urine t,t-MA. In summary, we aimed to evaluate whether personal air benzene or urine t,t-MA is a better surrogate for benzene exposure in terms of attenuation of the theoretical exposure-disease association in epidemiological analysis. Our results suggest that there would be no definitive difference in attenuation bias when personal air benzene or urine t,t-MA is used as a surrogate for benzene exposure. Our prediction of attenuation of theoretical exposure-disease association may aid exposure assessment efforts examining the associations between benzene exposure and health outcomes, but our results should also be tested in future epidemiological studies. Author Contributions DHK designed the study and collected urine samples. MYL performed the analysis of urine t,t-MA. EKC and JKJ performed air benzene sampling and analysis. DUP contributed to the analyses and their interpretation. Ethics approval This study protocol was reviewed by the Institutional Review Board of the OSHRI. All participants gave signed consent to provide their urine samples. Conflict of Interests The authors declare that they have no actual or potential conflict of interest. This study was supported by the Occupational Safety and Health Research Institute (OSHRI), Korea Occupational Safety and Health Agency (KOSHA). ==== Refs References 1 Loomis D Guyton KZ Grosse Y El Ghissassi F Bouvard V Benbrahim-Tallaa L Guha N Vilahur N Mattock H Straif K , International Agency for Research on Cancer Monograph Working Group (2017 ) Carcinogenicity of benzene . Lancet Oncol 18 , 1574 –5 . 29107678 2 Kuang S Liang W (2005 ) Clinical analysis of 43 cases of chronic benzene poisoning . Chem Biol Interact 153-154 , 129 –35 . 15935809 3 Tunsaringkarn T Soogarun S Palasuwan A (2013 ) Occupational exposure to benzene and changes in hematological parameters and urinary trans, trans-muconic acid . Int J Occup Environ Med 4 , 45 –9 . 23279797 4 Koh DH Jeon HK Lee SG Ryu HW (2015 ) The relationship between low-level benzene exposure and blood cell counts in Korean workers . Occup Environ Med 72 , 421 –7 . 25575529 5 Park D Choi S Ha K Jung H Yoon C Koh DH Ryu S Kim S Kang D Yoo K (2015 ) Estimating benzene exposure level over time and by industry type through a review of literature on Korea . Saf Health Work 6 , 174 –83 . 26929825 6 Lin YS Kupper LL Rappaport SM (2005 ) Air samples versus biomarkers for epidemiology . Occup Environ Med 62 , 750 –60 . 16234400 7 Panev T Popov T Georgieva T Chohadjieva D (2002 ) Assessment of the correlation between exposure to benzene and urinary excretion of t, t-muconic acid in workers from a petrochemical plant . Int Arch Occup Environ Health 75 Suppl, S97 –100 . 12397418 8 Symanski E Greeson NMH Chan W (2007 ) Evaluating measurement error in estimates of worker exposure assessed in parallel by personal and biological monitoring . Am J Ind Med 50 , 112 –21 . 17238141 9 Jakubowski M (2012 ) Biological monitoring versus air monitoring strategies in assessing environmental-occupational exposure . J Environ Monit 14 , 348 –52 . 22130625 10 Truchon G Tardif R Charest-Tardif G de Batz A Droz PO (2013 ) Evaluation of occupational exposure: comparison of biological and environmental variabilities using physiologically based toxicokinetic modeling . Int Arch Occup Environ Health 86 , 157 –65 . 22411213 11 White E Armstrong BK Saracci R (2008 ) Principles of exposure measurement in epidemiology, 2nd ed. Oxford University Press, Oxford. 12 Boogaard PJ van Sittert NJ (1995 ) Biological monitoring of exposure to benzene: a comparison between S-phenylmercapturic acid, trans,trans-muconic acid, and phenol . Occup Environ Med 52 , 611 –20 . 7550802 13 Chung EK Shin JA Lee BK Kwon J Lee N Chung KJ Lee JH Lee IS Kang SK Jang JK (2010 ) Characteristics of occupational exposure to benzene during turnaround in the petrochemical industries . Saf Health Work 1 , 51 –60 . 22953163 14 Chung EK Jang JK Koh DH (2017 ) A comparison of benzene exposures in maintenance and regular works at Korean petrochemical plants . J Chem Health Saf 24 , 21 –6 . 15 Koh DH Kim TW Yoon YH Shin KS Yoo SW (2011 ) Lymphohematopoietic cancer mortality and morbidity of workers in a refinery/petrochemical complex in Korea . Saf Health Work 2 , 26 –33 . 22953184 16 Koh DH Chung EK Jang JK Lee HE Ryu HW Yoo KM Kim EA Kim KS (2014 ) Cancer incidence and mortality among temporary maintenance workers in a refinery/petrochemical complex in Korea . Int J Occup Environ Health 20 , 141 –5 . 24999849 17 Miller JC Miller JN (1988 ) Statistics for analytical chemistry, 2nd ed. Ellis Horwood Limited, Chichester. 18 Nieuwenhuijsen MJ (1997 ) Exposure assessment in occupational epidemiology: measuring present exposures with an example of a study of occupational asthma . Int Arch Occup Environ Health 70 , 295 –308 . 9352332 19 Tielemans E Kupper LL Kromhout H Heederik D Houba R (1998 ) Individual-based and group-based occupational exposure assessment: some equations to evaluate different strategies . Ann Occup Hyg 42 , 115 –9 . 9559571 20 Lubin JH Colt JS Camann D Davis S Cerhan JR Severson RK Bernstein L Hartge P (2004 ) Epidemiologic evaluation of measurement data in the presence of detection limits . Environ Health Perspect 112 , 1691 –6 . 15579415 21 Leiber C Zeileis A (2008 ) Applied econometrics with R. Springer-Verlag, New York. 22 R Core Team (2016 ) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://ww.R-project.org/. 23 Seixas NS Sheppard L (1996 ) Maximizing accuracy and precision using individual and grouped exposure assessments . Scand J Work Environ Health 22 , 94 –101 . 8738886 24 Bates D Mächler M Bolker B Walker S (2015 ) Fitting linear mixed-effects models using lme4 . J Stat Softw 67 , 1 –48 . 25 SAS The MIANALYZE Procedure (2008 ) SAS/STAT 9.2 User’s Guide. SAS Institute Inc., Cary. 26 American Conference of Governmental Industrial Hygienists (2001 ) Benzene. In: ACGIH, ed. Documentation of Threshold Limit Values and Biological Exposure Indices, 7th ed. ACGIH, Cincinnati. 27 American Conference of Governmental Industrial Hygienists (2001 ) Benzene BEI. In: ACGIH, ed. Documentation of Threshold Limit Values and Biological Exposure Indices, 7th ed. ACGIH, Cincinnati. 28 Liu K Stamler J Dyer A McKeever J McKeever P (1978 ) Statistical methods to assess and minimize the role of intra-individual variability in obscuring the relationship between dietary lipids and serum cholesterol . J Chronic Dis 31 , 399 –418 . 711832 29 Inoue O Seiji K Nakatsuka H Watanabe T Yin SN Li GL Cai SX Jin C Ikeda M (1989 ) Urinary t,t-muconic acid as an indicator of exposure to benzene . Br J Ind Med 46 , 122 –7 . 2923822 30 Ducos P Gaudin R Bel J Maire C Francin JM Robert A Wild P (1992 ) trans,trans-Muconic acid, a reliable biological indicator for the detection of individual benzene exposure down to the ppm level . Int Arch Occup Environ Health 64 , 309 –13 . 1487326 31 Lauwerys RR Buchet JP Andrien F (1994 ) Muconic acid in urine: a reliable indicator of occupational exposure to benzene . Am J Ind Med 25 , 297 –300 . 8147402 32 Lee BL New AL Kok PW Ong HY Shi CY Ong CN (1993 ) Urinary trans,trans-muconic acid determined by liquid chromatography: application in biological monitoring of benzene exposure . Clin Chem 39 , 1788 –92 . 8375048 33 Ghittori S Maestri L Fiorentino ML Imbriani M (1995 ) Evaluation of occupational exposure to benzene by urinalysis . Int Arch Occup Environ Health 67 , 195 –200 . 7591178 34 Ghittori S Maestri L Rolandi L Lodola L Fiorentino ML Imbriani M (1996 ) The determination of trans, trans-muconic acid in urine as an indicator of occupational exposure to benzene . Appl Occup Environ Hyg 11 , 187 –91 . 35 Ong CN Kok PW Lee BL Shi CY Ong HY Chia KS Lee CS Luo XW (1995 ) Evaluation of biomarkers for occupational exposure to benzene . Occup Environ Med 52 , 528 –33 . 7663638 36 Ong CN Kok PW Ong HY Shi CY Lee BL Phoon WH Tan KT (1996 ) Biomarkers of exposure to low concentrations of benzene: a field assessment . Occup Environ Med 53 , 328 –33 . 8673180 37 Hotz P Carbonnelle P Haufroid V Tschopp A Buchet JP Lauwerys R (1997 ) Biological monitoring of vehicle mechanics and other workers exposed to low concentrations of benzene . Int Arch Occup Environ Health 70 , 29 –40 . 9258705 38 Javelaud B Vian L Molle R Allain P Allemand B André B Barbier F Churet AM Dupuis J Galand M Millet F Talmon J Touron C Vaissière M Vechambre D Vieules M Viver D (1998 ) Benzene exposure in car mechanics and road tanker drivers . Int Arch Occup Environ Health 71 , 277 –83 . 9638485 39 Fustinoni S Consonni D Campo L Buratti M Colombi A Pesatori AC Bonzini M Bertazzi PA Foà V Garte S Farmer PB Levy LS Pala M Valerio F Fontana V Desideri A Merlo DF (2005 ) Monitoring low benzene exposure: comparative evaluation of urinary biomarkers, influence of cigarette smoking, and genetic polymorphisms . Cancer Epidemiol Biomarkers Prev 14 , 2237 –44 . 16172237 40 Carrieri M Bonfiglio E Scapellato ML Maccà I Tranfo G Faranda P Paci E Bartolucci GB (2006 ) Comparison of exposure assessment methods in occupational exposure to benzene in gasoline filling-station attendants . Toxicol Lett 162 , 146 –52 . 16289653 41 Campagna M Satta G Campo L Flore V Ibba A Meloni M Tocco MG Avataneo G Flore C Fustinoni S Cocco P (2012 ) Biological monitoring of low-level exposure to benzene . Med Lav 103 , 338 –46 . 23077794 42 Lovreglio P D’Errico MN Fustinoni S Drago I Barbieri A Sabatini L Carrieri M Apostoli P Soleo L (2011 ) Biomarkers of internal dose for the assessment of environmental exposure to benzene . J Environ Monit 13 , 2921 –8 . 21909569 43 Kim S Vermeulen R Waidyanatha S Johnson BA Lan Q Rothman N Smith MT Zhang L Li G Shen M Yin S Rappaport SM (2006 ) Using urinary biomarkers to elucidate dose-related patterns of human benzene metabolism . Carcinogenesis 27 , 772 –81 . 16339183 44 Pezzagno G Maestri L (1997 ) The specificity of trans,trans-muconic acid as a biological indicator for low levels of environmental benzene . Indoor Built Environ 6 , 12 –8 . 45 Weaver VM Buckley T Groopman JD (2000 ) Lack of specificity of trans,trans-muconic acid as a benzene biomarker after ingestion of sorbic acid-preserved foods . Cancer Epidemiol Biomarkers Prev 9 , 749 –55 . 10919747 46 Gobba F Rovesti S Borella P Vivoli R Caselgrandi E Vivoli G (1997 ) Inter-individual variability of benzene metabolism to trans,trans-muconic acid and its implications in the biological monitoring of occupational exposure . Sci Total Environ 199 , 41 –8 . 9200846 47 Kivistö H Pekari K Peltonen K Svinhufvud J Veidebaum T Sorsa M Aitio A (1997 ) Biological monitoring of exposure to benzene in the production of benzene and in a cokery . Sci Total Environ 199 , 49 –63 . 9200847 48 Cocco P Tocco MG Ibba A Scano L Ennas MG Flore C Randaccio FS (2003 ) trans,trans-Muconic acid excretion in relation to environmental exposure to benzene . Int Arch Occup Environ Health 76 , 456 –60 . 12684810 49 Boogaard PJ van Sittert NJ (1996 ) Suitability of S-phenyl mercapturic acid and trans-trans-muconic acid as biomarkers for exposure to low concentrations of benzene . Environ Health Perspect 104 Suppl 6, 1151 –7 . 9118886 50 Ruppert T Scherer G Tricker AR Adlkofer F (1997 ) trans,trans-muconic acid as a biomarker of non-occupational environmental exposure to benzene . Int Arch Occup Environ Health 69 , 247 –51 . 9137998 51 van Tongeren MJ Kromhout H Gardiner K Calvert IA Harrington JM (1999 ) Assessment of the sensitivity of the relation between current exposure to carbon black and lung function parameters when using different grouping schemes . Am J Ind Med 36 , 548 –56 . 10506737 52 Van Tongeren M Burstyn I Kromhout H Gardiner K (2006 ) Are variance components of exposure heterogeneous between time periods and factories in the European carbon black industry? Ann Occup Hyg 50 , 55 –64 . 16126761 53 van Tongeren M Gardiner K Calvert I Kromhout H Harrington JM (1997 ) Efficiency of different grouping schemes for dust exposure in the European carbon black respiratory morbidity study . Occup Environ Med 54 , 714 –9 . 9404318