Perpetuating fears: Bias against the null hypothesis in fetal safety of drugs as expressed in scientific citations

Background Bias against negative studies (i.e., those showing no issues with fetal safety of drugs) may cause distorted interpretation with apparently safe drugs being labeled as teratogenic, causing women to terminate pregnancy or not to treat serious medical conditions. Objective To investigate whether "positive" studies, claiming teratogenic effects of drugs, which were later shown to be safe, have been cited more often than "negative" studies on the same topic. Methods We reviewed published studies on the fetal safety of 6 drugs, which were the focus of appreciable controversy over the last 5 decades (oral contraceptives, bendectin^® benzodiazepines, paroxetine, ACE inhibitors and statins). While initial highly publicized papers claimed teratogenic effects, these were subsequently contradicted by large numbers of "negative" studies. We compared medical citation patterns of the "positive" vs. "negative" papers related to these 6 drugs. Results "Positive" papers were 70% more likely to be cited than "negative" articles (median 39 vs. 23, p=0.04). In multivariate linear regression, "positivity" of results (p=0.04), the number of years since publication (p=0.01) and journal citation impact (p<0.001) all independently predicted the total number of medical citations. Conclusions We documented bias against the null hypothesis in medical citations of fetal drug safety. Acknowledging this source of bias is critical in trying to avert the distortion of the medical knowledge created by it.