Discordance among clinical studies addressing the same hypothesis is either due to differences in study design (so-called clinical heterogeneity) or to error. Error in clinical research can be categorized into two types.8 The first is random error, or error due to chance. Perhaps chance led to the inclusion of unusually healthy or unusually sick patients in some of the published studies of RBILD. The larger the study population, the less likely it is that random error will influence the results. By this measure, the cohort of Portnoy et al7 seems the least likely to be affected by random error, as its sample size is twice that of the next largest study. The second type of error in clinical research is systematic error, or error due to bias. To some degree, bias is present in all clinical investigation, although a rigorous study design can minimize its impact. A common source of bias in case series from highly specialized referral centers is called referral bias. Patients seen at these centers often differ substantially from the general population of patients with a given disorder. These differences may be obvious (eg, the patients are sicker, younger, have more aggressive disease, or are more likely to have tried and not responded to treatment) or more subtle (eg, the patients are better educated, more motivated, or better insured). It is likely that referral bias is present in many, if not all, published case series of RBILD, and this may further explain some of the variability in results.