Abstract
PURPOSE: We evaluated a large cohort of patients with longstanding ulcerative colitis in a colonoscopic surveillance program to determine predictors of colectomy. METHODS: We queried a retrospective database of patients who had symptoms of ulcerative colitis for seven years or more. Histologic inflammation in biopsies was graded on a validated four-point scale: absent, mild, moderate, severe. We performed a multivariate analysis of the inflammation scores and other variables to determine predictive factors for colectomy. Patients who underwent colectomy for neoplasia were censored at the time of surgery; those who did not undergo colectomy were censored at the time of last contact. RESULTS: A total of 561 patients were evaluated, with a median follow-up of 21.4 years since disease onset. A total of 97 patients (17.3 percent) underwent surgery; 25 (4.5 percent) for reasons other than dysplasia. These 25 constitute events for this analysis. For univariate analysis, mean inflammation (P < 0.001) and steroid use (P = 0.01) were predictors of colectomy. For multivariable proportional hazards analysis, mean inflammation (P < 0.001) and steroid use (P = 0.03) were predictors of colectomy, whereas salicylate use (P = 0.007) was protective. CONCLUSIONS: Higher median inflammation scores and corticosteroid use were predictors of colectomy in this patient population. The overall rate of colectomy during a long period of follow-up was low (<1 percent per year).