A Clinical Prediction Model for Early Colectomy in Patients With Severe Ulcerative Colitis Treated With Tacrolimus

Tacrolimus is an effective treatment option for refractory ulcerative colitis; however, some patients still require colectomy due to insufficient response. Early assessment of surgical risk is clinically important, as delayed decision-making may worsen the patient's condition and increase the risk of postoperative complications. This study aimed to identify predictors of colectomy within 3 months of initiating tacrolimus therapy and to develop a clinically applicable prediction model.

We conducted a retrospective analysis of hospitalized patients with severe ulcerative colitis treated with tacrolimus between 2011 and 2025. Fourteen clinical background variables were evaluated using LASSO-penalized logistic regression with cross-validation to construct the prediction model.

Among 114 patients, 24 (21.1%) underwent colectomy, including 16 (14.0%) within 3 months of treatment initiation. The LASSO regression identified three predictive variables: serum albumin level, hemoglobin level, and age at tacrolimus initiation. The resulting model demonstrated good discriminative performance, with an area under the curve of 0.78. Using a cutoff value of logit(p), the model achieved a sensitivity of 87.5% and a specificity of 63.4%. Kaplan-Meier analysis revealed a significantly higher cumulative colectomy rate in the high-risk group (p < 0.001), supporting the model's predictive utility.

We developed a clinical prediction model that accurately estimates the risk of early colectomy based on baseline clinical factors at the start of tacrolimus therapy. This model may serve as a practical tool to guide decision-making regarding surgical timing and overall treatment strategy.