Scientific standards and the design of case-control research

A new controversy has emerged over the results of a case-control study alleging a causal relationship between certain non-steroidal anti-inflammatory agents (NSAIAs) and the risks of agranulocytosis and aplastic anemia. After describing the methods and results of the International Agranulocytosis and Aplastic Anemia (IAAA) study, we review the distinctive methodologic challenges of this study and the requirements for avoiding bias, and then reconcile the study results with the principles of case-control design. As a result of our analysis, we believe that the IAAA study's most important and reliable finding is its documentation of the infrequent occurrences of aplastic anemia and agranulocytosis with analgesic use. In contrast, a causal association between NSAIAs and blood dyscrasias has not been suitably established, and may well have resulted from several distinctive sources of bias. These include the effects of diagnostic-suspicion bias in case determination, of exclusion bias in choosing controls, of recall bias in determining exposure, and of publicity bias in both case selection and ascertainment of exposure. These problems could have been avoided and a more valid result obtained with closer attention to the experimental paradigm for case-control research.