Choice of control population in studies of adverse reproductive effects of occupational exposures and its effect on risk estimates.

Cohort studies of adverse reproductive effects of occupational exposures that compare the outcome of exposed pregnancies with non-exposed pregnancies among the same group of mothers (internal controls) and among a different group of mothers (external controls) have been reviewed. Confounding of the hypothesised associations by parental characteristics, the outcome of previous pregnancies, and other exposures is not adequately assessed in these studies. The risk of an adverse outcome appears to be consistently higher among the external control group of pregnancies than among the internal controls. If the differences between the two control groups are real, comparisons based on external controls underestimate the relative or attributable risk. Alternative explanation for this phenomenon are discussed. A likely explanation is a misclassification of exposure or a high prevalence of risk factors other than the exposure of interest among the external controls. Non-exposed pregnancies among exposed mothers may be used for comparison in preference to pregnancies among non-exposed mothers.