P671 Calprotectin does not add accuracy to PRO2 for prediction of continuous clinical response in moderate-to-severe ulcerative colitis patients treated with golimumab

Abstract

Background: Stool frequency and rectal bleeding are major patientreported outcomes (PRO2) in ulcerative colitis (UC). The STRIDE consensus recommends monitoring of UC activity by PRO2 and endoscopy, while faecal calprotectin (CP) is not a treatment target.1 We investigated the association of PRO2 alone and in combination with CP with continuous clinical response (CCR) in the PURSUITmaintenance (-M) trial.2 Methods: This analysis included 456 patients who responded to golimumab induction therapies, and were randomised in the PURSUIT-M trial through week 54. Logistic regression was used to assess the association of PRO2 and faecal CP concentration with CCR. The accuracy of predictors of CCR was assessed through area under the receiveroperating characteristic curve (AUC), which ranges from 0 (low) to 1 (high accuracy). The ability to predict CCR was assessed by comparing AUC between a model with and without the predictor. PRO2 was assessed every 4 weeks and CP was assessed at weeks 0, 30, and 54. From time of loss of response (LOR), the baseline PRO2 (week 0 of induction) was imputed, assigning non-CCR status to patients with LOR. CP values at week 0 were carried forward through week 26, and from week 30 through week 50. The actual 4-weekly (partial) Mayo clinical response (PMS) was introduced in the model as ±1 if the PRO2 assessment date was earlier/later than the date of LOR. The model thus was: CCR = PRO2 + log10 (CP) ± PMS. Results: The CCR prediction model that included PRO2, CP, and PMS was highly associated with CCR, and accuracy (AUC) increased through week 54 (Table). The model was mostly driven by PRO2, as noted when assessing the accuracy (AUC) of PRO2 alone. In contrast, CP alone and the prediction model without PRO2 had a lower accuracy for prediction of CCR. Finally, the model without CP had similar accuracy (AUC) for prediction of CCR compared with the complete model (that included CP). (Table presented) Conclusions: This PURSUIT-M post hoc analysis suggests that CP does not provide added predictive value beyond that of PRO2 alone in predicting CCR in UC patients.