Bias toward the null hypothesis in pregnancy drug studies that do not include data on medical terminations of pregnancy: the folic acid antagonists.

Most studies on safety/risk of drugs in pregnancy consider the proportion of births (but not pregnancy terminations) affected by the drug from all exposed infants. Lack of data on pregnancy terminations could bias results. A computerized database for medications dispensed to pregnant women in southern Israel was linked with records from the district hospital; 84 823 deliveries and 998 medical pregnancy terminations took place; 571 of the women were exposed to folic acid antagonists in the first trimester. When only births were examined, there was no association between folic acid antagonists and fetal malformations. When data on pregnancy terminations were examined and births and pregnancy terminations were combined, there was a significant risk (neural tube defects: odds ratio 18.83, 95% confidence interval 9.24-38.37; cardiovascular defects: odds ratio 3.86, 95% confidence interval 1.67-8.88; and neural tube defects: odds ratio 6.30, 95% confidence interval 3.34-9.15; cardiovascular defects: odds ratio 1.76, 95% confidence interval 1.05-2.92, respectively). Inclusion of only birth data in observational studies of drugs in pregnancy constitutes a source of bias toward the null hypothesis.