Comparing the strengths and weaknesses of observational and experimental studies using a postmarketing surveillance...

A recent prospective, observational study in rheumatoid arthritis patients indicated that the addition of hydroxychloroquine to either aspirin or methotrexate therapy decreased the incidence of hepatic enzyme abnormalities. This interesting finding is of potential clinical importance, but its validity needs to be examined in terms of the potential confounders inherent in observational studies. Although one of the study's strengths is its derivation from "real-life" data, some potential confounders that might weaken the data include a need to examine whether any scientific rationale can be discerned for the observation; examination of control-case matching (issues of randomization and baseline disease characteristics); the potential for attribution bias; data-collection methods (prospective versus retrospective, uniform versus chart review); and equivalency of treatment protocols, dosing regimens, and concomitant medications. Potential scientific rationale exists for the observed interaction, and data collection is both uniform and prospective. These strengths are confounded by the inevitable lack of randomization in observational studies, the potential for differences in baseline disease characteristics, attribution bias, a lack of controlled dosing regimens and treatment protocols, and an assumption that all nonsteroid antiinflammatory drugs are alike (which is not true). On balance, the hypothesis generated by these data is compelling enough to deserve further testing in both observational and experimental settings.