Background Medical rescue therapy (MRT) is effective in intravenous corticosteroid refractory acute severe ulcerative colitis (ASUC). We recently described the ADMIT-ASC score to identify patients on admission at high risk of steroid failure. In this setting, our primary objective in this analysis is now to identify parameters at admission which could identify non-response to MRT, following corticosteroid failure. Co-primary objective is to develop and validate a model to identify non-responders to MRT.

Methods Two cohorts were studied, for discovery and validation. Retrospective analysis of 66 adult ASUC admissions receiving MRT between 2015-19 at Gold Coast University Hospital and Logan Hospital was first performed. Clinical, endoscopic and laboratory data were collected. Non-response was defined as undergoing colectomy during the index admission. Univariable and multivariable logistic regression were employed to identify predictors of non-response. A predictive model was developed based on thresholds obtained from ROC curves. The predictive model was validated in 99 patients admitted to Gold Coast University Hospital and Sunshine Coast University Hospital between 2020-23.

Results In the 2015-19 cohort, out of 66 episodes who received MRT, 11 episodes (16.7%) underwent colectomy during the same admission. UCEIS at admission was the strongest predictor of non-response to MRT [OR 3.03 (95%CI 1.47-6.22), p 0.033), AUROC 0.7552 (95%CI 0.6271 – 0.8832]. UCEIS score ≥6 had sensitivity 100%, specificity 41.7%, PPV 26.3%, NPV 100%, to predict non-response to MRT. CRP on day 3 after commencing rescue therapy (CRP Day R+3) was also a predictor for non-response to MRT (OR 1.02 (95%CI 1.00-1.05), p 0.015). A score comprising 2 points involving UCEIS ≥ 6 and CRP value of ≥ 22 mg/L on day R +3 was developed (1 point each). Score of 2 points had accuracy 79.6%, specificity 75%, sensitivity 100%, PPV 47.4%, NPV 100%, AUROC 0.8750 (95%CI 0.8070-0.9429) for non-response to MRT. In the validation cohort, 6 (6.1%) patients underwent colectomy during the same admission after failing MRT. On implementing the score, a score of 2 points had accuracy 86.7%, specificity 88.1%, sensitivity 66.7%, PPV 28.6%, NPV 97.4%, AUROC 0.8095 (95%CI 0.6423-0.9768) for predicting non-response to MRT.

Conclusion UCEIS score at admission and CRP on day 3 after commencing MRT are predictors of non-response to MRT. We have developed and validated a model to identify non-response to medical rescue therapy, and objective criteria to mandate surgical intervention.