Fertility of workers. A comparison of logistic regression and indirect standardization.

Estimates of the effect of occupational exposure on the fertility of men employed at three chemical plants were obtained from data stored at the Chemical Industry Institute of Toxicology using logistic regression and indirect standardization. Logistic regression was explored as a possible alternative to indirect standardization because it 1) permits consideration of potential confounding variables, such as the relative spacing between consecutive births, that are not included in the characterization of the fertility of an external reference population, and 2) may enable the study of occupational cohorts for which fertility data from an appropriate external reference population are not available. In addition to the main effects of age, parity, and birth cohort, main effects of certain lag variables which characterize the timing of birth events during the five-year period preceding each person-year were found to be significant at all three plants. Interactions of some of these lag variables with age also emerged as significant. Despite this, both methods were consistent in accepting or rejecting the null hypothesis regarding the effect of exposure at two of the plants. At the third plant, however, the logistic regression yielded a significant interaction between exposure status and one of the lag variables. The net exposure effect was significantly increased fertility in exposed person-years for which no birth had occurred 2-3 years prior, and nonsignificantly decreased fertility in exposed person-years for which such a prior birth had occurred. While this may be a spurious finding, it suggests that both methods of analysis should continue to be explored until a larger body of similar comparative studies has accumulated.