Search

ObjectivesThere is growing recognition that simultaneously assessing multiple exposures may reduce false positive discoveries and improve epidemiological effect estimates. We evaluated the performance of statistical methods for identifying exposure–outcome associations across various data structures typical of environmental and occupational epidemiology analyses.MethodsWe simulated a case–control study, generating 100 data sets for each of 270 different simulation scenarios; varying the number of exposure variables, the correlation between exposures, sample size, the number of effective exposures and the magnitude of effect estimates. We compared conventional analytical approaches, that is, univariable (with and without multiplicity adjustment), multivariable and stepwise logistic regression, with variable selection methods: sparse partial least squares discriminant analysis, boosting, and frequentist and Bayesian penalised regression approaches.ResultsThe variable selection methods consistently yielded more precise effect estimates and generally improved selection accuracy compared with conventional logistic regression methods, especially for scenarios with higher correlation levels. Penalised lasso and elastic net regression both seemed to perform particularly well, specifically when statistical inference based on a balanced weighting of high sensitivity and a low proportion of false discoveries is sought.ConclusionsIn this extensive simulation study with multicollinear data, we found that most variable selection methods consistently outperformed conventional approaches, and demonstrated how performance is influenced by the structure of the data and underlying model.

BackgroundOccupational allergic diseases are a major problem in some workplaces like in the baking industry. Diagnostic rules have been used in surveillance but not yet in the occupational respiratory clinic.ObjectiveTo develop diagnostic models predicting baker’s asthma and rhinitis among bakery workers at high risk of sensitisation to bakery allergens referred to a specialised clinic.MethodsAs part of a medical surveillance programme, clinical evaluation was performed on 436 referred Dutch bakery workers at high risk for sensitisation to bakery allergens. Multivariable logistic regression analyses were developed to identify the predictors of onset of baker’s asthma and rhinitis using a self-administered questionnaire and compared using a structured medical history. Performance of models was assessed by discrimination (area under the receiver operating characteristics curve) and calibration (Hosmer-Lemeshow test). Internal validity of the models was assessed by a bootstrapping procedure.ResultsThe prediction models included the predictors of work-related upper and lower respiratory symptoms, the presence of allergy and allergic symptoms, use of medication (last year), type of job, type of shift and working years with symptoms (≥10 years). The developed models derived from both self-administered questionnaire and the medical history showed a relatively good discrimination and calibration. The internal validity showed that the models developed had satisfactory discrimination. To improve calibrations of models, shrinkage factors were applied to model coefficients.ConclusionThe probability of allergic asthma and rhinitis in referred bakers could be estimated by diagnostic models based on both a self-administered questionnaire and by taking a structured medical history.

ObjectiveWithin the SYNERGY project, we evaluated the exposure-response relation between occupational exposure to respirable crystalline silica (RCS) and the risk of lung cancer in pooled analyses of community-based studies.MethodsRCS was estimated using a quantitative general population job-exposure matrix (‘SYNJEM’) for 16 786 lung cancer cases and 20 818 matched population or hospital controls. 34.2% of the men were ever exposed to RCS, while 8.6% of the women were ever exposed. Odds ratios (ORs) and 95% confidence intervals (95% CI) were estimated stratified by gender using unconditional logistic regression models adjusted for age, study centre, cigarette pack-years, time-since-quitting smoking, and ever employment in an occupation with known lung cancer risk.ResultsWe observed a monotonic increase in risk of lung cancer associated with occupational exposure to RCS among men (unexposed versus 4thquartile among exposed 1.45 (95%CI, 1.31–1.60). Result did not differ by smoking status and remained significantly elevated among non-smokers. The association was stronger in squamous cell carcinoma and small cell lung cancer as compared to adenocarcinoma of the lung especially among former and current smokers. The effect of RCS on lung cancer in women was not detectable likely related to small numbers.ConclusionsThe SYNERGY results show that occupational silica exposure is associated with an increased risks of all lung cancer types in a pooled analyses of community-based studies.

ObjectiveUsing data from the SYNERGY project, we evaluated the association between occupational exposure to polycyclic aromatic hydrocarbons (PAH) and the risk of lung cancer, adjusted for tobacco smoking. MethodsFor 16 901 lung cancer cases and 20 965 matched population or hospital controls, PAH exposure was estimated using a quantitative general population job-exposure matrix (‘SYNJEM’) based on five-digit ISCO-68 codes (4639 cases, 4713 controls ever exposed). Odds ratios (ORs) and 95% confidence intervals (95% CI) were estimated using unconditional logistic regression models adjusted for age, sex, study centre, smoking behaviour, and ever employment in an occupation with known lung cancer risk.ResultsWe observed a modest increased risk of lung cancer associated with occupational exposure to PAHs according to various exposure metrics (ever/never, duration, cumulative dose, time since last exposure). The ORs for ever exposure to PAH were 1.09 (95%CI, 1.04–1.15) overall, 1.08 (95%CI, 1.02–1.15) in men, 1.20 (95%CI, 1.05–1.38) in women, and 1.04 (95%CI, 0.88–1.22) in never smokers. These results are further supported by significant exposure response-relationships (p-value for trend <0.05 for years of employment and cumulative exposure [(BaP) µg/m3-years]). When stratified by histological subtype, increased risks and positive exposure response-relationships were apparent only for squamous cell carcinoma and small cell lung cancer.ConclusionsOur pooled analysis suggests that occupational exposure to PAH is associated with a modest increase in the risk of lung cancer, after adjustment for tobacco smoking and exposure to other occupational lung carcinogens.

ObjectiveThe use of measurement data in occupational exposure assessment allows more quantitative analyses of possible exposure–response relations. We describe a quantitative exposure assessment approach for the five lung carcinogens selected for the SYNERGY project, that is, asbestos, chromium-VI, nickel, polycyclic aromatic hydrocarbons (by its proxy benzo(a)pyrene (BaP)) and respirable crystalline silica. A quantitative job-exposure matrix (i.e. SYN-JEM) was developed based on statistical modelling of large quantities of personal measurements.MethodsEmpirical linear models were developed using personal occupational exposure measurements from Europe and Canada, as well as auxiliary information like job (industry), year of sampling, region, an a priori exposure rating of each job (none, low, and high exposed) and sampling duration. The model outcomes were used to create SYN-JEM with a quantitative estimate of the level of exposure by job, year, and region.ResultsDecreasing time trends were observed for all agents between the 1970s and 2009, ranging from −1.2% per year for personal BaP and nickel exposures to −10.7% for asbestos before a ban was implemented. Regional differences in exposure concentrations varied by agent, ranging from a factor 3.3 for chromium-VI up to a factor 10.5 for asbestos.ConclusionWe estimated time-, job-, and region-specific exposure levels for four (asbestos, chromium-VI, nickel, and RCS) out of the five considered lung carcinogens. Statistical modelling of large amounts of personal occupational exposure measurement data enabled the derivation of a quantitative general population JEM, which can be applied to the SYNERGY population.

BackgroundThere is limited evidence regarding the exposure-effect relationship of exposure to hexavalent chromium (CrVI) and nickel (Ni) with lung cancer. We estimated the cumulative exposure for CrVI and Ni and assessed exposure-effect relationships for lung cancer risk by sex, smoking status, and histological subtypes.MethodsFourteen case-control studies (1985–2010) from Europe and Canada were pooled, including 16 901 lung cancer cases (80% men) and 20 965 controls (78% men). Cumulative exposure to CrVI and Ni were estimated. Unconditional logistic regression models were fitted to estimate odds ratios (OR), 95% confidence intervals (CI), and exposure-effect trends adjusted for smoking and occupations with recognised lung cancer risk.ResultsThe OR for the highest quartile (>98.95 μg/m3-years) of cumulative CrVI-exposure was 1.33 (95% CI 1.20–1.48) in men and 1.04 (95% CI 0.48–2.23) in women. In never smokers, the OR for ever CrVI-exposure was 1.37 (95% CI 1.09–1.73) in men, and OR=1.09; 95% CI 0.70–1.69 in women.The OR for the highest quartile of cumulative Ni-exposure (>77.53 μg/m3-years) was 1.30 (95% CI 1.16–1.45) in men and 1.29 (95% CI 0.59–2.81) in women. The OR for ever Ni-exposure was 1.22 in never smokers for both sexes.ConclusionsOur results showed an exposure-dependent excess risk of lung cancer by occupational exposure to Ni in both sexes, and for CrVI in men. The pattern for CrVI in women was less clear. Analysis of an interaction between Cr- and Ni-exposure was impaired by a high correlation of these agents in metal fumes.

ObjectiveThe Synergy project derived quantitative exposure-response associations for five occupational lung carcinogens (asbestos, chromium-VI, nickel, polycyclic aromatic hydrocarbons, and respirable crystalline silica) and lung cancer in a pooled analysis of population based case-control studies. Considering a proportion of workers were exposed to more than one of these carcinogens, a joint effect on lung cancer risk is possible.MethodsWe estimated joint effects by including an interaction term between two occupational carcinogens in the logistic regression models that were developed for the Synergy project. Analyses were conducted with either both exposures dichotomized (ever vs. never exposed), or with one exposure on a continuous scale (cumulative exposure), and the other dichotomized. Analyses were conducted for all lung cancer subtypes combined and stratified by subtype. We applied a Bonferroni correction.ResultsWe observed a negative interaction between occupational exposure to nickel and asbestos. The interaction effect was largest for the subtype of squamous cell carcinoma: ratio of odds ratios: 0.76 (95% CI 0.65–0.88), odds ratio of the joint effect: 1.40 (95% CI 1.26–1.56). No other interaction effects were statistically significant after correction for multiple testing. Analyses in which one of the exposures was included on a continuous scale resulted in similar results.ConclusionWe observed little evidence for a statistical multiplicative interaction between most of the occupational carcinogens. The negative multiplicative interaction between asbestos and nickel was not explained by a high correlation between these exposures. Ignoring specific study specific matching criteria might have introduced some bias in the results.