Adjustment for temporal confounders in a reanalysis of a case-control study of beryllium and lung cancer

Abstract

Objectives: To evaluate potential confounding of the association between beryllium and lung cancer in a reanalysis of data from a published case-control study of workers at a beryllium processing facility. Methods: The association of cumulative and average beryllium exposure with lung cancer among 142 cases and five age-match controls per case was reanalyzed using conditional logistic regression. Adjustment was made independently for potential confounders of hire age and birth year. Alternative adjustments to avoid taking the logarithm of zero were explored. Results: Adjustment for either birth cohort or hire age (two highly correlated factors) attenuated lung cancer risk associated with cumulative exposure; however, lung cancer risk was significantly associated with average exposure using a 10-year lag following adjustment. Stratification of analyses by birth cohort found greater lung cancer risk from cumulative and average exposure for workers born before 1900 than for workers born later. The magnitude of the association between lung cancer and average exposure was not reduced by modifying the method used to take the log of exposure. Conclusion: In this reanalysis, average—but not cumulative—beryllium exposure was related to lung cancer risk after adjustment for birth cohort. Confounding by birth cohort is likely related to differences in smoking patterns for workers born before 1900 and the tendency for workers hired during the World War II era to have been older at hire.