External validation of a prognostic model of preoperative risk factors for failure of restorative proctocolectomy

Aim: The Cleveland Clinic has proposed a prognostic model of preoperative risk factors for failure of restorative proctocolectomy (RPC) with ileal pouch–anal anastomosis. The model incorporates four predictive variables: completion proctectomy, handsewn anastomosis, diabetes mellitus and Crohn's disease. The aim of the present study was to perform an external validation of this model in a new cohort of patients who had RPC. Method: Validation was performed in a multicentre cohort of 747 consecutive patients who had an RPC between 1990 and 2015 in three tertiary-care facilities, using a Kaplan–Meier survival analysis and Cox regression analysis. The performance of the model was expressed using the Harrell concordance error rate. The primary outcome measure was pouch survival with maintenance of anal function. Results: During the study period, 45 (6.0%) patients experienced failure at a median interval of 31 months (interquartile range 9–82 months) from the original RPC. Multivariable analysis showed handsewn anastomosis to be the only significant independent predictor. The Harrell concordance error rate was 0.42, indicating poor performance. Anastomotic leakage and Crohn's disease of the pouch were strong postoperative predictors for pouch failure and showed a significant difference in pouch survival after 10 years (P < 0.001). Conclusion: The poor performance of the Cleveland Clinic prognostic model makes it unsuitable for daily clinical practice. Handsewn anastomosis was associated with pouch failure in our cohort with relatively few events. A prediction model for anastomotic leakage or Crohn's disease of the pouch may be a better solution since these variables are strongly associated with pouch failure.