Maintenance treatment for type 1 autoimmune pancreatitis: effectiveness and development of the PrescrAIP relapse prediction model

Background & Aims Type 1 autoimmune pancreatitis (AIP) is a relapsing remitting disorder that often requires multiple treatment courses. Our aims were to assess the efficacy of maintenance treatment in preventing relapse and to develop the PrescrAIP risk score predicting relapse risk and the benefit of maintenance treatment. Methods We retrospectively analyzed patients meeting international diagnostic criteria for type 1 AIP who reached partial or complete remission after initial treatment. The primary outcome was disease relapse, defined as recurrence of symptoms and/or radiologic findings. We developed a multivariable prediction model using Cox proportional hazards regression, performed internal and internal-external validation, and built a predictive nomogram. Results We included 577 patients (68% male). During a median follow-up of 34 months (interquartile range, 13–69 months), we observed 154 relapses. The overall 3-year relapse risk was 28% (95% confidence interval [CI], 24%–32%), lower in patients receiving maintenance treatment than in those without (22% vs 35%; P < .001). The final PrescrAIP model incorporated protective factors (maintenance treatment, prior surgery, focal mass, female sex) and risk factors (biliary involvement, other organ involvement, IgG4 elevation, allergy, jaundice, acute pancreatitis). The model showed moderate discrimination (Concordance index, 0.69) and good calibration. Internal-external validation yielded Concordance index values ranging from 0.64 to 0.71. Maintenance treatment significantly reduced relapse in patients with a PrescrAIP score >155, but not in those with lower scores. Conclusions Maintenance therapy reduced relapse only in patients at high relapse risk. Once externally validated, the PrescrAIP score may guide personalized maintenance treatment decisions.