Autism Spectrum Disorder, 2

Hundreds of genes and more than a hundred environmental exposures have been identified as potential causes, mediators, or markers of risk for autism spectrum disorder (ASD). The findings for the environmental exposures, almost all occurring during pregnancy, have emerged from regression analyses in observational studies. The risk estimates are most often presented as odds ratios (ORs), sometimes as hazard ratios (HRs), and rarely as relative risks. This article uses gestational exposure to antidepressant drugs and risk of ASD in offspring as a background to explain how estimates of ASD risk in observational studies are commonly interpreted and why and when the usual interpretations are wrong, often very wrong. The article provides discussions on crude and adjusted estimates, ORs and HRs, individual studies and meta-analyses, strategies that help address confounding by unmeasured and unknown variables, and a detailed discussion on the imprecision of the numerical value of the adjusted estimate. The article explains how the value of an OR is not set in stone; different procedures and approaches in analysis of the same data result in different OR values. The article also explains how to evaluate an individual patient's risk when multiple risk factors are present that may or may not be independent of each other. Finally, the article suggests the presence of an elephant in the room: risk factors that, though independent, may saturate mechanisms that mediate the outcome; so, when simultaneously present, their individual ORs may suggest falsely lower values of risk. This suggestion could explain why ASD is uncommon in the population although the risk factors for ASD are common and many. It is important to be aware of the issues considered in this article when attempting to understand the field, counsel patients, communicate research findings to peers and the public, and frame policy.