Derivation and Validation of a Major Toxicity Risk Score Among Nonsteroidal Antiinflammatory Drug Users Based on Data From a Randomized Controlled Trial.

OBJECTIVE: While nonsteroidal antiinflammatory drugs (NSAIDs) are commonly used in rheumatology, they can cause major toxicity. Improving the risk/benefit ratio requires a more precise understanding of risk. This study was undertaken to derive and validate a risk score for major toxicity among NSAID users enrolled in a randomized controlled trial.

METHODS: Patients enrolled in a randomized controlled trial who had known cardiovascular disease or risk factors as well as osteoarthritis or rheumatoid arthritis were divided into derivation and validation cohorts. Patients were randomized to receive celecoxib, naproxen, or ibuprofen at typical dosages. The risk score was designed to predict the 1-year occurrence of major toxicity among NSAID users, including major adverse cardiovascular events, acute kidney injury, significant gastrointestinal events, and mortality. Variables significantly associated with major toxicity were candidates for inclusion in the final regression model. After derived models were found to have a similar model fit in the validation set, the cohorts were combined, allowing calculation of a risk score.

RESULTS: In the derivation cohort, significant variables included age, male sex, history of cardiovascular disease, hypertension, diabetes mellitus, tobacco use, statin use, elevated serum creatinine level, hematocrit level, and type of arthritis. The C-index was 0.73 in the validation cohort and 0.71 in the total cohort; the model was well calibrated. Of the total population with complete data (n = 23,735), 1,080 participants (4.6%) had a predicted 1-year risk of major toxicity of <1%, 16,273 (68.6%) had a predicted risk of 1-4%, and 6,382 (26.9%) had a predicted risk of >4%.

CONCLUSION: The risk score accurately categorizes the 1-year risk of major toxicity among NSAID users and may be useful in identifying patients who can safely use these agents.