Risk-stratified monitoring for sulfasalazine toxicity: prognostic model development and validation

Background: Sulfasalazine-induced cytopenia, nephrotoxicity and hepatotoxicity is uncommon during long-term treatment. Some guidelines recommend 3 monthly monitoring blood tests indefinitely during long-term treatment while others recommend stopping monitoring after 1 year. To rationalise monitoring, we developed and validated a prognostic model for clinically significant blood, liver or kidney toxicity during established sulfasalazine treatment. Design: Retrospective cohort study. Setting: UK primary care. Data from Clinical Practice Research Datalink Gold and Aurum formed independent development and validation cohorts. Participants: Age ≥18 years, new diagnosis of an inflammatory condition and sulfasalazine prescription. Study period: 1 January 2007 to 31 December 2019. Outcome: Sulfasalazine discontinuation with abnormal monitoring blood-test result. Analysis: Patients were followed up from 6 months after first primary care prescription to the earliest of outcome, drug discontinuation, death, 5 years or 31 December 2019. Penalised Cox regression was performed to develop the risk equation. Multiple imputation handled missing predictor data. Model performance was assessed in terms of calibration and discrimination. Results: 8936 participants were included in the development cohort (473 events, 23 299 person-years) and 5203 participants were included in the validation cohort (280 events, 12 867 person-years). Nine candidate predictors were included. The optimism adjusted R2D and Royston D statistic in the development data were 0.13 and 0.79, respectively. The calibration slope (95% CI) and Royston D statistic (95% CI) in validation cohort was 1.19 (0.96 to 1.43) and 0.87 (0.67 to 1.07), respectively. Conclusion: This prognostic model for sulfasalazine toxicity uses readily available data and should be used to risk-stratify blood-test monitoring during established sulfasalazine treatment.