Data on retrospective compensation claims for injuries caused by pharmaceutical drugs are prone to selection and reporting biases. Nevertheless, this case study of the antidiabetic drug benfluorex shows that such data can be used to estimate the cumulative incidence of drug-related injury, and to provide insights into its epidemiology. To this end, we develop a modelling framework for under-reporting of retrospective claims for compensation arising from drug damage. The model involves a longitudinal component related to attrition of cases over time, and a cross-sectional component related to incomplete reporting. We apply this model to cardiac valve surgery necessitated by exposure to benfluorex. Benfluorex was marketed in France between 1976 and 2009, when it was withdrawn because it caused valvular heart disease. A scandal erupted in 2010 over the scale of the damage caused by the drug. Since then, no further estimates of cumulative incidence have been published, though thousands of claims for compensation have been processed. The analysis combines compensation claims data and sociological survey data on benfluorex users, together with data on benfluorex sales and duration of treatment. We find a threshold of toxicity at about 6 months' exposure, and that at least 1690 individuals (95% CI 1290 to 2320) needed heart surgery to replace or repair valves damaged by exposure to benfluorex in France: a cumulative incidence of 3.68 per 10,000 (95% CI 2.68 to 5.34) benfluorex users or 3.22 per 10,000 (95% CI 2.48 to 4.39) person-years at risk above the exposure threshold. While these findings are tentative, they are consistent with those obtained previously using very different methods.
Keywords: benfluorex; compensation claim; cumulative incidence; pharmacoepidemiology; under‐reporting; valvular heart disease.
© 2024 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.