Diagnostic Yield of Random Colon Biopsies for Microscopic Colitis: A Community-Based Assessment

Abstract

Introduction: Microscopic colitis(MC) is a diagnosis frequently considered in patients undergoing colonoscopy for diarrhea. Recent studies identified variables to predict likelihood of finding MC on colonoscopy, including: 1) weight loss 2) BMI 3) nocturnal stools 4) autoimmune disease and 5) medications introduced three months prior to diarrhea onset. These studies were performed at academic medical centers and thus could be influenced by referral bias. No community-based studies have evaluated these predictive variables for diagnostic yield of MC in the general population. Methods: A national pathology database was queried for colon biopsies conducted at a community-based ambulatory surgery center from January through December 2014. Diagnostic codes utilized included diarrhea, change in bowel habits, nonspecific colitis. Specimens with MC diagnosis, either collagenous colitis (CC) or lymphocytic colitis (LC), were recorded as cases. Biopsies with normal colon histology served as controls. A retrospective chart review was performed on all samples using the predictor variables listed above, excluding nocturnal stools. Results: We identified 413 colonoscopies with biopsies obtained at random locations in normal appearing colon mucosa. There were 38 cases (17CC, 21LC), 375 controls with a diagnostic yield for MC of 10.1%. Univariable analyses identified age and weight loss as positive predictors of MC. BMI was a negative predictor(Table 1). Medications or autoimmune disease were not significant on univariable analysis. Multivariable analysis revealed BMI was no longer a significant negative predictor, while age, modeled as a continuous variable, remained a significant positive predictor (Table 2). Weight loss was not included in multivariable analysis due to missing data upon chart review. However, a subgroup regression analysis of patients with non-missing weight loss data suggested age and BMI as predictors of MC.Table: Table. Univariable Logistic Regression ModelsTable: Table. Multivariable Logistic Regression ModelsConclusion: To our knowledge, this is the first community-based study evaluating the diagnostic yield of MC on random colon biopsies in patients with diarrheal presentations. The yield of MC is low (10%), raising questions on the utility of obtaining biopsies without considering predictive variables. In our study, multivariable regression analysis did not clearly establish predictive factors for MC. However independently, an older patient, a patient with weight loss, or a patient with low BMI who presents with diarrhea could heighten suspicions for MC and justify colonoscopy with biopsy.