Prediction of late-onset pouch failure in patients with restorative proctocolectomy with a nomogram.

BACKGROUND: A proportion of UC patients with restorative proctocolectomy and IPAA develop pouch failure. Accurate risk assessment is critical for making proper evaluation and treatment. Information on factors that may reliably predict pouch failure for the patients requiring referral to a specialized care unit is minimal.

AIM: We sought to develop and internally validate a nomogram for the prediction of late-onset pouch failure.

METHODS: The study cohort included all eligible UC patients with restorative proctocolectomy and IPAA at the subspecialty Pouchitis Clinic from 2002 to 2009. Inclusion criteria were patients having: 1) inflammatory bowel disease; 2) ileal pouches; and 3) regular follow-up at the Pouchitis Clinic. Demographic and clinical variables were prospectively collected. Multivariable accelerated failure time regression model was developed to predict pouch failure defined as pouch excision or permanent diversion. Discrimination and calibration of the model were assessed following bootstrapping methods for correcting optimism, and the model was presented as a nomogram.

RESULTS: A total of 921 patients were included for the model. The mean age for this cohort was 45.5 years old. The mean follow-up at the Pouchitis Clinic was 5.8 years. Kaplan-Meier analysis showed that the probabilities for pouch retention are 0.939, 0.916 and 0.907 at 3, 5 and 7 years, respectively. The predictor variables which were included in the nomogram were smoking, duration of the pouch, baseline pouch diagnosis, and pre- and post-op use of biologics. The concordance index was 0.824. The nomogram seemed well calibrated based on the calibration curve.

CONCLUSIONS: The nomogram model appeared to predict late-onset pouch failure reasonably well with satisfactory concordance index and calibration curve. The nomogram is readily applicable for clinical practice in pouch patients.