Optimization of a Scoring System to Predict Microscopic Colitis in a Cohort of Patients With Chronic Diarrhea

Goals: Our aim was to develop a scoring system to predict risk of microscopic colitis (MC), to identify patients at low risk, potentially avoiding unnecessary biopsies. Background: Patients with chronic diarrhea often undergo colonoscopy with biopsy, but few have histologic abnormalities. Study: We conducted a retrospective study of patients with chronic diarrhea and a macroscopically normal colonoscopy at our institution over a 9-month period. Multivariable logistic regression assessed the association between predictors and the presence of biopsy-proven MC. Results: The derivation cohort included 617 patients. Median age was 55.1 (39.6 to 68.1) years; 397 (64.3%) were female and 81 (13.1%) had MC. Age ≥55 years, duration of diarrhea ≤6 months, ≥5 bowel movements per day, body mass index <30 kg/m², current smoking, and current use of selective serotonin reuptake inhibitors/serotonin-norepinephrine reuptake inhibitorss and non-steroidal anti-inflammatory drugs were independently associated with MC. A score of ≥10 points in our scoring system, yielded an area under the ROC curve (AUC) of 0.83 with a sensitivity of 93% and specificity of 49% in predicting which patients have MC. The negative predictive value (NPV) was 97.8% (95.0% to 99.1%). In the validation cohort, the scoring system performed similarly (AUC 0.79, sensitivity 91%, specificity 49%, NPV 97%). By avoiding biopsies in patients at low risk of having MC, costs associated with colon biopsies could be reduced by almost 43%. Conclusion: This scoring system including 7 clinical variables was able to identify patients unlikely to have MC, with excellent sensitivity, reasonable specificity, and a high NPV, translating into important potential cost savings.