Adjustment for sparse data bias in odds ratios: Significance to appraisal of risk of diabetes due to occupational trichlorfon insecticide exposure.

Background: Bias away from the null in odds ratios (OR), aggravated by low power, is a well-known phenomenon in statistics (sparse data bias). Such bias increases in presence of selection of "significant" results on the basis of null hypothesis testing (effect size magnification, ESM). Objectives: We seek to illustrate these issues and adjust for suspected sparse data bias in the context of a reported more than doubling of the odds of new onset type 2 diabetes in presence of occupational trichlorfon insecticide exposure reported in the Agricultural Health Study. Methods: We performed ESM analysis on the crude ORs extracted from the contingency table in the published report, which is done by simulating selected OR given a posited true OR. Next, we applied easily accessible methods that adjust for sparse data bias to the extracted contingency tables, including data augmentation, bootstrap, Firth's regression, and Bayesian methods with weakly informative priors. Results: During the ESM analysis, we observed that there was a reasonable chance that a "statistically significant" OR of around 2.5-2.6 would be observed for true OR of 1.2. Adjustment for sparse data bias revealed that Bayesian methods outperformed alternative approaches in terms of yielding more precise inference, while not making unjustified distributional assumptions about estimates of OR. The OR in the original paper of about 2.5-2.6 was reduced on average to OR of 1.9 to 2.2, with 95% (Bayesian) credible intervals that included the null. Discussion: It is reasonable to adjust ORs for sparse data bias when the reported association has societal importance, because policy must be informed by the least biased estimates of the effect. We think that such adjustment would lead to a more appropriate evaluation of the extent of evidence on the contribution of occupational exposure to trichlorfon pesticide to risk of new onset diabetes.

Determinants of SARS-CoV-2 IgG response and decay in Canadian healthcare workers: A prospective cohort study.

INTRODUCTION: Healthcare workers (HCWs) from an interprovincial Canadian cohort gave serial blood samples to identify factors associated with anti-receptor binding domain (anti-RBD) IgG response to the SARS-CoV-2 virus. METHODS: Members of the HCW cohort donated blood samples four months after their first SARS-CoV-2 immunization and again at 7, 10 and 13 months. Date and type of immunizations and dates of SARS-CoV-2 infection were collected at each of four contacts, together with information on immunologically-compromising conditions and current therapies. Blood samples were analyzed centrally for anti-RBD IgG and anti-nucleocapsid IgG (Abbott Architect, Abbott Diagnostics). Records of immunization and SARS-CoV-2 testing from public health agencies were used to assess the impact of reporting errors on estimates from the random-effects multivariable model fitted to the data. RESULTS: 2752 of 4567 vaccinated cohort participants agreed to donate at least one blood sample. Modelling of anti-RBD IgG titer from 8903 samples showed an increase in IgG with each vaccine dose and with first infection. A decrease in IgG titer was found with the number of months since vaccination or infection, with the sharpest decline after the third dose. An immunization regime that included mRNA1273 (Moderna) resulted in higher anti-RBD IgG. Participants reporting multiple sclerosis, rheumatoid arthritis or taking selective immunosuppressants, tumor necrosis factor inhibitors, calcineurin inhibitors and antineoplastic agents had lower anti-RBD IgG. Supplementary analyses showed higher anti-RBD IgG in those reporting side-effects of vaccination, no relation of anti-RBD IgG to obesity and lower titers in women immunized in early or mid-pregnancy. Sensitivity analysis results suggested no important bias in the self-report data. CONCLUSION: Creation of a prospective cohort was central to the credibility of results presented here. Serial serology assessments, with longitudinal analysis, provided effect estimates with enhanced accuracy and a clearer understanding of medical and other factors affecting response to vaccination.

Parkinson's disease and occupational exposure to organic solvents in Finland: a nationwide case-control study.

OBJECTIVE: This study aimed to investigate the association between Parkinson's disease (PD) and occupational exposure to organic solvents generally and chlorinated hydrocarbons (CHC) in particular. METHODS: We assembled a Finland-wide case-control study for birth years 1930-1950 by identifying incident PD cases from the register of Reimbursement of Medical Costs and drawing two controls per case using incidence density sampling from the Population Information System, matched on sex, birth year, and residency in Finland in 1980-2014. Occupation and socioeconomic status (SES) were identified from national censuses. We assessed cumulative occupational exposures via FINJEM job-exposure matrix. Smoking was based on occupation-specific prevalence by sex from national surveys. We estimated confounder-adjusted PD incidence rate ratios (IRR) via logistic regression and evaluated their sensitivity to errors in FINJEM through probabilistic bias analysis (PBA). RESULTS: Among ever-employed, we identified 17 187 cases (16.0% potentially exposed to CHC) and 35 738 matched controls. Cases were more likely to not smoke and belong to higher SES. Cumulative exposure (CE) to CHC (per 100 ppm-years, 5-year lag) was associated with adjusted IRR 1.235 (95% confidence interval 0.986-1.547), with stronger associations among women and among persons who had more census records. Sensitivity analyses did not reveal notable associations, but stronger effects were seen in the younger birth cohort (1940-1950). PBA produced notably weaker associations, yielding a median IRR 1.097 (95% simulation interval 0.920-1.291) for CHC. CONCLUSION: Our findings imply that PD is unlikely to be related to typical occupational solvent exposure in Finland, but excess risk cannot be ruled out in some highly exposed occupations.

Smokers' utilization of quitting methods and vaping during pregnancy: an empirical cluster analysis of 2016-2018 Pregnancy Risk Assessment Monitoring System (PRAMS) data in seven US states.

BACKGROUND: Patterns of utilization of numerous smoking cessation methods among pregnant women amidst the increasing popularity of vaping (use of e-cigarettes) remains unknown. METHODS: This study included 3,154 mothers who self-reported smoking around the time of conception and delivered live births in 2016-2018 in seven US states. Latent class analysis was used to identify subgroups of smoking women based on their utilization of 10 surveyed quitting methods and vaping during pregnancy. RESULTS: We identified four subgroups of smoking mothers with different utilization patterns of quitting methods during pregnancy: 22.0% reported "not trying to quit"; 61.4% tried to "quit on my own" without any behavioral or pharmacological assistance; 3.7% belonged to the "vaping" subgroup; and 12.9% utilized "wide-ranging methods" with higher use rate of multiple approaches, such as quit line and nicotine patch. Compared to mothers "not trying to quit," the subgroup trying to "quit on my own" were more likely to be abstinent (adjusted OR 4.95, 95% CI 2.82-8.35) or to reduce the number of cigarettes smoked daily (adjusted OR 2.46, 95% CI 1.31-4.60) in late pregnancy, and these improvements lasted into early postpartum. We did not observe a measurable reduction in smoking among the "vaping" subgroup or women trying to quit with "wide-ranging methods". CONCLUSIONS: We identified four subgroups of smoking mothers with different utilization patterns of eleven quitting methods during pregnancy. Pre-pregnancy smokers who tried to "quit on my own" were most likely to be abstinent or to reduce smoking amount.

Exposure-response analysis of the association of maternal smoking and use of electronic cigarettes (vaping) in relation to preterm birth and small-for-gestational-age in a national US sample, 2016-2018.

Introduction: The US experienced a surge in use of e-cigarettes. Smoking women may consider e-cigarettes during pregnancy as an alternative to smoking. E-cigarettes typically contain nicotine, an established cause of reduction in fetal growth in animal studies. Methods: This cohort study included 99,201 mothers who delivered live singletons in 2016-2018 from the Pregnancy Risk Assessment Monitoring System. We created exposure categories based on self-reported number of cigarettes smoked per day and vaping frequency and evaluated their associations with preterm birth and small-for-gestational-age (SGA) birth (two established cigarette smoking-related risks). Results: Dual users in late pregnancy were a heterogeneous group: 29% lightly smoked and occasionally vaped; 19% lightly smoked and frequently vaped; 36% heavily smoked and occasionally vaped; and 15% heavily smoked and frequently vaped. While dual users who heavily smoked and occasionally vaped had the highest adjusted OR for SGA (3.4, 95% CI 2.0, 5.7), all the dual users had, on average, about twice the odds of having SGA than non-users. While the risks of preterm birth were higher among sole light smokers (adjusted OR 1.3, 95% CI 1.1, 1.5) and sole heavy smokers (adjusted OR 1.5. 95% CI 1.2, 1.8) than non-users, the adjusted odds of preterm birth for dual users were not noticeably higher than those of non-users. Conclusion: Relative to non-users, both smoking and vaping during pregnancy appear to increase risk of SGA, but excess risk of preterm birth appears to be primarily attributable to smoking alone. Higher levels of exposure tended to confer more risk.

Interactions between exposure to polycyclic aromatic hydrocarbons and xenobiotic metabolism genes, and risk of breast cancer.

PURPOSE: Polycyclic aromatic hydrocarbons (PAHs) are a group of environmental pollutants associated with multiple cancers, including female breast cancer. Several xenobiotic metabolism genes (XMGs), including the CYP450 family, play an important role in activating and detoxifying PAHs, and variations in the activity of the enzymes they encode can impact this process. This study aims to examine the association between XMGs and breast cancer, and to assess whether these variants modify the effects of PAH exposure on breast cancer risk. METHODS: In a case-control study in Vancouver, British Columbia, and Kingston, Ontario, 1037 breast cancer cases and 1046 controls had DNA extracted from blood or saliva and genotyped for 138 single nucleotide polymorphisms (SNPs) and tagSNPs in 27 candidate XMGs. Occupational PAH exposure was assessed using a measurement-based job-exposure matrix. RESULTS: An association between genetic variants and breast cancer was observed among six XMGs, including increased risk among the minor allele carriers of AKR1C3 variant rs12387 (OR 2.71, 95% CI 1.42-5.19) and AKR1C4 variant rs381267 (OR 2.50, 95% CI 1.23-5.07). Heterogeneous effects of occupational PAH exposure were observed among carriers of AKR1C3/4 variants, as well as the PTGS2 variant rs5275. CONCLUSION: Our findings support an association between SNPs of XMGs and female breast cancer, including novel genetic variants that modify the toxicity of PAH exposure. These results highlight the interplay between genetic and environmental factors, which can be helpful in understanding the modifiable risks of breast cancer and its complex etiology.

Smoking and use of electronic cigarettes (vaping) in relation to preterm birth and small-for-gestational-age in a 2016 U.S. national sample.

Women who smoke may be motivated to switch to vaping (use electronic cigarettes, e-cigs) around pregnancy in seeking to alleviate known hazards of smoking. E-cigs typically contain nicotine but either eliminates or greatly reduces exposure to the combustion products of tobacco. We studied a U.S.-wide representative sample of 31,973 live singleton births in 2016. In the three months before pregnancy, 5029 (14%) mothers exclusively smoked tobacco ("sole smokers") and 976 (3%) used both tobacco and e-cigs ("dual-users"). Among pre-pregnancy sole smokers, 44% continued to only smoke while 1% became dual-users in late pregnancy. Logistic regression models were used to assess the adjusted odds ratios (aOR) for preterm and small-for-gestational-age (SGA) by reported smoking or vaping in late pregnancy. Compared to women who used neither product ("non-users"), late-pregnancy sole smokers had increased risks for preterm birth (aOR 1.6, 95% CI 1.2-2.0) and SGA (aOR 2.4, 95% CI 1.8-2.9), after adjusting for their pre-pregnancy smoking or vaping status and other confounders. The adjusted models also showed that late-pregnancy sole vapers had similar risk of preterm birth as non-users (aOR 1.2, 95% CI 0.5-2.7). Late-pregnancy dual-users also had similar risk of preterm birth as non-users (aOR 1.3, 95% CI 0.8-2.3). However, late-pregnancy sole vapers and dual-users had increased risk of SGA compared to non-users (aOR 2.4, 95% CI 1.0-5.7 for sole vapers, and aOR 2.3 95% CI 1.3-4.1 for dual-users). These findings suggest that vapers during pregnancy had similar risk of preterm as non-users but still had elevated risk for restricted fetal growth.

What to Do When Accumulated Exposure Affects Health but Only Its Duration Was Measured? A Case of Linear Regression.

Background: We considered a problem of inference in epidemiology when cumulative exposure is the true dose metric for disease, but investigators are only able to measure its duration on each subject. Methods: We undertook theoretical analysis of the problem in the context of a continuous response caused by cumulative exposure, when duration and intensity of exposure follow log-normal distributions, such that analysis by linear regression is natural. We present a Bayesian method to adjust duration-only analysis to incorporate partial knowledge about the relationship between duration and intensity of exposure and illustrate this method in the context of association of smoking and lung function. Results: We derive equations that (a) describe under what circumstances bias arises when duration of exposure is used as a proxy of cumulative exposure, (b) quantify the degree of such bias and loss of precision, and (c) describe how knowledge about relationship of duration and intensity of exposure can be used to recover an estimate of the effect of cumulative exposure when only duration was observed on every subject. Conclusions: Under our assumptions, when duration and intensity of exposure are either independent or positively correlated, we can be more confident in qualitatively interpreting the direction of effects that arise from use of duration of exposure per se. We can use external information on the relationship between duration and intensity of exposure (namely: correlation and variance of intensity), even if intensity of exposure is not available at the individual level, to make reliable inferences about the magnitude of effect of cumulative exposure on the outcome.

Maternal smoking and gestational hypertension: Heterogeneous effect by timing of the exposure.

OBJECTIVES: The objective of this study was to examine whether the association between maternal smoking and gestational hypertension varies by the timing of exposure. STUDY DESIGN: Retrospective cohort study of women identified in 2015 US natality records for singleton births. MAIN OUTCOME MEASURES: Our outcome was whether a woman was diagnosed with gestational hypertension (GH) on the birth record, a category which includes preeclampsia. RESULTS: Women who smoked before and during pregnancy had a reduced risk for GH relative to non-smokers (adjusted RR 0.92, 95% CI 0.90-0.94). In contrast, women who apparently quit just before the start of pregnancy had higher risk than non-smokers (adjusted RR 1.02, 95% CI 1.00-1.05). When the trimester-specific effects were examined, only women who smoked before pregnancy and in all three trimesters had reduced risk for GH. Smoking mothers who quit just before the start of the 3rd trimester had an increased risk for GH compared to non-smokers (adjusted RR 1.08, 95% CI 1.02-1.16). CONCLUSION: In our analysis, women who smoked before pregnancy and in all three trimesters have reduced risk of GH compared to non-smokers, while smokers who reported quitting before pregnancy were at an increased risk. Our results offer new insights into the importance of timing of smoking in pregnancy on risk of GH, and challenge the notion that any smoking during pregnancy has a protective effect.

Women's occupational exposure to polycyclic aromatic hydrocarbons and risk of breast cancer.

OBJECTIVE: To estimate the association between occupational polycyclic aromatic hydrocarbon (PAH) exposure and female breast cancer. METHODS: Lifetime work histories for 1130 cases and 1169 controls from British Columbia and Ontario (Canada) were assessed for PAH exposure using a job-exposure matrix based on compliance measurements obtained during US Occupational Safety and Health Administration workplace safety inspections. RESULTS: Exposure to any level of PAHs was associated with an increased risk of breast cancer (OR=1.32, 95% CI: 1.10 to 1.59), as was duration at high PAH exposure (for >7.4 years: OR=1.45, 95% CI: 1.10 to 1.91; ptrend=0.01), compared with women who were never exposed. Increased risk of breast cancer was most strongly associated with prolonged duration at high occupational PAH exposure among women with a family history of breast cancer (for >7.4 years: OR=2.79, 95% CI: 1.25 to 6.24; ptrend<0.01). CONCLUSIONS: Our study suggests that prolonged occupational exposure to PAH may increase breast cancer risk, especially among women with a family history of breast cancer.

Bayesian Correction for Exposure Misclassification and Evolution of Evidence in Two Studies of the Association Between Maternal Occupational Exposure to Asthmagens and Risk of Autism Spectrum Disorder.

PURPOSE OF REVIEW: Inference in epidemiologic studies is plagued by exposure misclassification. Several methods exist to correct for misclassification error. One approach is to use point estimates for the sensitivity (Sn) and specificity (Sp) of the tool used for exposure assessment. Unfortunately, we typically do not know the Sn and Sp with certainty. Bayesian methods for exposure misclassification correction allow us to model this uncertainty via distributions for Sn and Sp. These methods have been applied in epidemiologic literature, but are not considered a mainstream approach, especially in occupational epidemiology. RECENT FINDINGS: Here, we illustrate an occupational epidemiology application of a Bayesian approach to correct for the differential misclassification error generated by estimating occupational exposures from job codes using a job exposure matrix (JEM). We argue that analyses accounting for exposure misclassification should become more commonplace in the literature.

Correction of odds ratios in case-control studies for exposure misclassification with partial knowledge of the degree of agreement among experts who assessed exposures.

OBJECTIVES: Estimates of association between exposures and diseases are often distorted by error in exposure classification. When the validity of exposure assessment is known, this can be used to adjust these estimates. When exposure is assessed by experts, even if validity is not known, we sometimes have information about interrater reliability. We present a Bayesian method for translating the knowledge of interrater reliability, which is often available, into knowledge about validity, which is often needed but not directly available, and applying this to correct odds ratios (OR). METHODS: The method allows for inclusion of observed potential confounders in the analysis, as is common in regression-based control for confounding. Our method uses a novel type of prior on sensitivity and specificity. The approach is illustrated with data from a case-control study of lung cancer risk and occupational exposure to diesel engine emissions, in which exposure assessment was made by detailed job history interviews with study subjects followed by expert judgement. RESULTS: Using interrater agreement measured by kappas (κ), we estimate sensitivity and specificity of exposure assessment and derive misclassification-corrected confounder-adjusted OR. Misclassification-corrected and confounder-adjusted OR obtained with the most defensible prior had a posterior distribution centre of 1.6 with 95% credible interval (Crl) 1.1 to 2.6. This was on average greater in magnitude than frequentist point estimate of 1.3 (95% Crl 1.0 to 1.7). CONCLUSIONS: The method yields insights into the degree of exposure misclassification and appears to reduce attenuation bias due to misclassification of exposure while the estimated uncertainty increased.

Bayesian Analysis of Occupational Exposure Data with Conjugate Priors.

Bayesian analysis is a flexible method that can yield insight into occupational exposures as the methods quantify plausible values for exposure parameters of interest, such as the mean, variance, and specific percentiles of the exposure distribution. We describe three Bayesian analysis methods for the analysis of normally distributed data (e.g. the logarithm of measurements of chemical hazards) that use conjugate prior distributions (normal for the mean, and inverse-χ2, inverse-Γ, or vague for the variance) to provide analytical expressions for the posterior distributions of the sufficient statistics of the normal distribution (e.g. the mean and variance). From these posterior distributions, the posterior distribution of any parameter of interest about the exposure distribution can be tabulated. The methods are illustrated using lead exposure data collected by the Occupational Safety and Health Administration at a copper foundry on multiple occasions. A unique feature of the normal-inverse-Γ method is that dependence of the mean and variance prior distributions is integrated out of the posterior distributions expressions, suggesting that a 'default' prior distribution on variance may be used: candidate default distributions are proposed based on the literature. Relative to other Bayesian analysis methods used in industrial hygiene, the methods described are flexible, and can be implemented without specialized software.

Visualizing the Heterogeneity of Effects in the Analysis of Associations of Multiple Myeloma with Glyphosate Use. Comments on Sorahan, T. Multiple Myeloma and Glyphosate Use: A Re-Analysis of US Agricultural Health Study (AHS) Data. Int. J. Environ. Res. Public Health 2015, 12, 1548-1559.

We address a methodological issue of the evaluation of the difference in effects in epidemiological studies that may arise, for example, from stratum-specific analyses or differences in analytical decisions during data analysis. We propose a new simulation-based method to quantify the plausible extent of such heterogeneity, rather than testing a hypothesis about its existence. We examine the contribution of the method to the debate surrounding risk of multiple myeloma and glyphosate use and propose that its application contributes to a more balanced weighting of evidence.

Maternal Exposure to Occupational Asthmagens During Pregnancy and Autism Spectrum Disorder in the Study to Explore Early Development.

Maternal immune activity has been linked to children with autism spectrum disorder (ASD). We examined maternal occupational exposure to asthma-causing agents during pregnancy in relation to ASD risk. Our sample included 463 ASD cases and 710 general population controls from the Study to Explore Early Development whose mothers reported at least one job during pregnancy. Asthmagen exposure was estimated from a published job-exposure matrix. The adjusted odds ratio for ASD comparing asthmagen-exposed to unexposed was 1.39 (95 % CI 0.96-2.02). Maternal workplace asthmagen exposure was not associated with ASD risk in this study, but this result does not exclude some involvement of maternal exposure to asthma-causing agents in ASD.

Polymorphisms of Insulin-Like Growth Factor 1 Pathway Genes and Breast Cancer Risk.

Genetic variants of insulin-like growth factor 1 (IGF1) pathway genes have been shown to be associated with breast density and IGF1 levels and, therefore, may also influence breast cancer risk via pro-survival signaling cascades. The aim of this study was to investigate associations between IGF1 pathway single nucleotide polymorphisms (SNPs) and breast cancer risk among European and East Asian women, and potential interactions with menopausal status and breast tumor subtype. Stratified analyses of 1,037 cases and 1,050 controls from a population-based case-control study were conducted to assess associations with breast cancer for 22 SNPs across 5 IGF1 pathway genes in European and East Asian women. Odds ratios were calculated using logistic regression in additive genetic models. Polytomous logistic regression was used to assess heterogeneity by breast tumor subtype. Two SNPs of the IGF1 gene (rs1019731 and rs12821878) were associated with breast cancer risk among European women. Four highly linked IGF1 SNPs (rs2288378, rs17727841, rs7136446, and rs7956547) were modified by menopausal status among East Asian women only and associated with postmenopausal breast cancers. The association between rs2288378 and breast cancer risk was also modified by breast tumor subtype among East Asian women. Several IGF1 polymorphisms were found to be associated with breast cancer risk and some of these associations were modified by menopausal status or breast tumor subtype. Such interactions should be considered when assessing the role of these variants in breast cancer etiology.

DNA repair variants and breast cancer risk.

A functional DNA repair system has been identified as important in the prevention of tumour development. Previous studies have hypothesized that common polymorphisms in DNA repair genes could play a role in breast cancer risk and also identified the potential for interactions between these polymorphisms and established breast cancer risk factors such as physical activity. Associations with breast cancer risk for 99 single nucleotide polymorphisms (SNPs) from genes in ten DNA repair pathways were examined in a case-control study including both Europeans (644 cases, 809 controls) and East Asians (299 cases, 160 controls). Odds ratios in both additive and dominant genetic models were calculated separately for participants of European and East Asian ancestry using multivariate logistic regression. The impact of multiple comparisons was assessed by correcting for the false discovery rate within each DNA repair pathway. Interactions between several breast cancer risk factors and DNA repair SNPs were also evaluated. One SNP (rs3213282) in the gene XRCC1 was associated with an increased risk of breast cancer in the dominant model of inheritance following adjustment for the false discovery rate (P < 0.05), although no associations were observed for other DNA repair SNPs. Interactions of six SNPs in multiple DNA repair pathways with physical activity were evident prior to correction for FDR, following which there was support for only one of the interaction terms (P < 0.05). No consistent associations between variants in DNA repair genes and breast cancer risk or their modification by breast cancer risk factors were observed.

Using an epiphytic moss to identify previously unknown sources of atmospheric cadmium pollution.

Urban networks of air-quality monitors are often too widely spaced to identify sources of air pollutants, especially if they do not disperse far from emission sources. The objectives of this study were to test the use of moss bio-indicators to develop a fine-scale map of atmospherically-derived cadmium and to identify the sources of cadmium in a complex urban setting. We collected 346 samples of the moss Orthotrichum lyellii from deciduous trees in December, 2013 using a modified randomized grid-based sampling strategy across Portland, Oregon. We estimated a spatial linear model of moss cadmium levels and predicted cadmium on a 50m grid across the city. Cadmium levels in moss were positively correlated with proximity to two stained-glass manufacturers, proximity to the Oregon-Washington border, and percent industrial land in a 500m buffer, and negatively correlated with percent residential land in a 500m buffer. The maps showed very high concentrations of cadmium around the two stained-glass manufacturers, neither of which were known to environmental regulators as cadmium emitters. In addition, in response to our findings, the Oregon Department of Environmental Quality placed an instrumental monitor 120m from the larger stained-glass manufacturer in October, 2015. The monthly average atmospheric cadmium concentration was 29.4ng/m(3), which is 49 times higher than Oregon's benchmark of 0.6ng/m(3), and high enough to pose a health risk from even short-term exposure. Both stained-glass manufacturers voluntarily stopped using cadmium after the monitoring results were made public, and the monthly average cadmium levels precipitously dropped to 1.1ng/m(3) for stained-glass manufacturer #1 and 0.67ng/m(3) for stained-glass manufacturer #2.

Cross-classified occupational exposure data.

We demonstrate the regression analysis of exposure determinants using cross-classified random effects in the context of lead exposures resulting from blasting surfaces in advance of painting. We had three specific objectives for analysis of the lead data, and observed: (1) high within-worker variability in personal lead exposures, explaining 79% of variability; (2) that the lead concentration outside of half-mask respirators was 2.4-fold higher than inside supplied-air blasting helmets, suggesting that the exposure reduction by blasting helmets may be lower than expected by the Assigned Protection Factor; and (3) that lead concentrations at fixed area locations in containment were not associated with personal lead exposures. In addition, we found that, on average, lead exposures among workers performing blasting and other activities was 40% lower than among workers performing only blasting. In the process of obtaining these analyses objectives, we determined that the data were non-hierarchical: repeated exposure measurements were collected for a worker while the worker was a member of several groups, or cross-classified among groups. Since the worker is a member of multiple groups, the exposure data do not adhere to the traditionally assumed hierarchical structure. Forcing a hierarchical structure on these data led to similar within-group and between-group variability, but decreased precision in the estimate of effect of work activity on lead exposure. We hope hygienists and exposure assessors will consider non-hierarchical models in the design and analysis of exposure assessments.