Value of Epidemiologic Studies in Determining the True Incidence of Adverse Events

Abstract

Evidence supporting differential toxic effects of nonsteroidal anti-inflammatory drugs (NSAIDs) often is derived from spontaneous reports of adverse events to the US Food and Drug Administration (FDA). These reports represent observations from an undefined, exposed population that are not collected in a standardized manner and therefore are subject to reporting biases. Epidemiologic studies, in which the numbers of patients experiencing an adverse event and exposed to the drug are known, provide more reliable measures of risk and can place spontaneous reports in perspective. To compare both data sources with regard to NSAID-associated gastrointestinal, liver, and skin events. We obtained spontaneous reports of these adverse events for diclofenac, nabumetone, naproxen, and piroxicam. Published epidemiologic studies of these events were reviewed. Spontaneous reports did not mirror reliably the results of epidemiologic studies. Spontaneous reports showed higher associations of gastrointestinal and skin events with nabumetone and piroxicam and hepatic events with diclofenac. Epidemiologic studies generally did not show differential risk among these NSAIDs. When the 4 NSAIDs are compared in epidemiologic studies, there is no quantitative basis for identifying 1 as more or less toxic than the others, underlining the hazard of deriving quantitative conclusions from spontaneous reports. Spontaneous reports are an unreliable measure of risk; rather, they may provide evidence of the relative awareness of specific toxic effects among physicians.