Researchers often study the effectiveness of new treatments, with assessment of harms (side effects) as secondary outcomes. In this context, randomized controlled trials are generally regarded as the gold standard for assessing effectiveness and safety of new interventions. However, there are circumstances where some evidence reports an association between an intervention and patient harm, but researchers wish to further evaluate or quantitate the association. In this situation, it would be unethical to randomize individuals to treatment groups to determine which do and do not receive the possibly harmful intervention. Instead, observational studies, including case-control or cohort studies are often used, where an intervention is not intentionally assigned. Case-control studies identify cases with and without the outcome of interest, and then evaluate the association between intervention exposure versus outcome. Cohort studies follow patients either retrospectively or prospectively throughout a defined period of time to determine whether they develop the outcome of interest based on an exposure that has already occurred. Because harmful effects often occur infrequently, these studies tend to include a large number of individuals. In both types of studies, since participants are not randomized, it is important to maintain rigor in the study in other ways. Thus, the study groups need to be similar with the exception of the exposure of concern, follow-up must be complete, and, potential temporal and dose–response gradients should be considered. For the reader, who wants to know whether his patients will be harmed by the intervention, the magnitude and precision of the risk is important. Furthermore, presenting the results as the number needed to harm (NNH) may be of clinical value to the reader. Unlike the number needed to treat, a high number for the NNH is preferred.