Abstract Background In Australia, the April 2014 listing of infliximab for acute severe ulcerative colitis (ASUC) transformed rescue and maintenance treatment and marked the biologic era, but the impact of biologic therapy on colectomy rates in real-world practice remains unclear. We compared emergency and cumulative colectomy risk in patients hospitalised with severe ulcerative colitis (UC) flares or ASUC in the pre-biologic and biologic eras at two Australian centres. Methods This multicentre retrospective cohort study included adults hospitalised with severe UC flares or ASUC between 2008 and 2023, grouped into a pre-biologic (2008–2014) and biologic eras (2014–2023). Primary outcomes were emergency colectomy during index admission and colectomy at 1 and 5 years post-admission. An extreme gradient boosting model, with hyperparameters optimised and SHAP (SHapley Additive exPlanations) for explainability, was trained to predict emergency and 1-year colectomy; features deemed redundant due to multicollinearity (such as Truelove and Witts criteria (TWC) and infliximab use for emergency colectomy prediction) were automatically excluded. Results We analysed 513 admissions (122 pre-biologic, 391 biologic era) with median follow-up 138 months IQR 63–166 and 36 months IQR 16–71, respectively. TWC for ASUC was met in 87.7% of pre-biologic and 59.8% of biologic-era admissions. Median age was 34 years IQR 26–48, with similar median disease durations in both groups (49 vs 47 months). Emergency colectomy was more frequent in the pre-biologic era (18.9% vs 6.4%, p 0.0001), as were colectomies at 1 and 5 years (32.0% vs 11.8% and 38.5% vs 17.9%, both p 0.0001). Similar colectomy rates were seen in the TWC-defined ASUC subgroup (emergency: 18.7% vs 6.7%; 1-year: 31.8% vs 12.4%; 5-year: 38.3% vs 19.1%; all p 0.001). Median time to colectomy was shorter in the pre-biologic era (25 days IQR 10–187 vs 101 days IQR 20–402, p = 0.019). Our ML model predicted emergency and 1-year colectomy with AUCs of 0.82 and 0.86, respectively. SHAP analysis identified antibiotic therapy during admission (OR 1.88, 95% CI 1.80–1.96), protein 62.4 g/L (OR 1.24, 95% CI 1.21–1.28) and higher CRP as the top predictors to emergency colectomy risk, and longer length of stay, disease duration and higher mean corpuscular haemoglobin as the main predictors of 1-year colectomy risk. Conclusion Since the introduction of biologic therapies in Australia, emergency and elective colectomy rates after admissions for ASUC and severe colitis have declined significantly, with biologic-era patients showing longer colectomy-free survival. An ML-based model showed moderate discrimination and identified simple, routinely available variables as key predictors of emergency and 1-year colectomy risk. Conflict of interest: Cao, Rena: No conflict of interest Baker, Lucy: No conflict of interest Juergens, Benjamin: No conflict of interest Cunningham, Juliet: No conflict of interest Ramesh, Mrday: No conflict of interest Leong, Rupert: advisory board: AbbVie, Aspen, BMS, Celgene, Celltrion, Chiesi, Ferring, Glutagen, Hospira, Janssen, Lilly, MSD, Novartis, Pfizer, Prometheus Biosciences, Takeda, Spyre, Roche research grants: Joanna Tiddy USYD, McCusker Charitable Foundation, Celltrion, Shire, Janssen, Takeda, Gastroenterological Society of Australia, NHMRC, Gutsy Group, Pfizer Corte, Crispin: No conflict of interest
Cao et al. (Thu,) studied this question.