Timing and patterns of exposures during pregnancy and their implications for study methods.

Patterns of exposure variability across pregnancy were examined for medical, lifestyle, residential, and occupational exposures in a population-based sample of 357 livebirths from 10 rural California counties. A new measure of variability, the ratio of overall prevalence to time-window-specific prevalence, is introduced. The higher the overall : time window (OTW) ratio, the greater the potential for misclassification when using anytime-during-pregnancy prevalence for an agent that exerts its effect in a smaller time window. Exposures to cigarettes, marijuana, saunas/hot tubs, factors related to location of residence, and some workplace substances tended to be of longer duration. Intertrimester concordance was high (kappa's > 0.8) for smoking, residential proximity to crops, and use of video display terminals; moderately high (kappa's between 0.4 and 0.8) for many occupational exposures; and low (kappa's < 0.4) for illnesses, which tended to be of short duration. The lowest OTW ratios were for smoking and some residential exposures (1.1-1.3), while OTW ratios were much higher for paint applications, influenza, vaginal infections, and ultrasound (reaching, e.g., 4-6). Use of anytime-during-pregnancy exposure indices can bias measures of association between risk factors and adverse pregnancy outcomes, particularly if the OTW ratio is high. Misclassification bias occurs if there is a vulnerable time window during which the exposure exerts its effect. The misclassification can be differential when the average length of gestation of cases is shorter than that of controls. For exposures that vary, investigations of pregnancy outcome should collect as much detail as feasible regarding timing.