A diagnostic model built from high-density lipoprotein cholesterol and four routine clinical measures distinguished ulcerative colitis from Crohn’s disease across participating centers, offering potential assistance for early patient triage. Differentiating ulcerative colitis from Crohn’s disease early in the diagnostic pathway remains difficult when a full clinical assessment is not yet available, prompting researchers to assess whether lipid measurements combined with standard laboratory indices could aid clinical distinction.
The multicenter retrospective cohort study evaluated a total of 783 patients treated across three participating clinical centers. For model training and verification, the investigators assigned 502 patients to the development cohort and distributed 281 patients across two external validation cohorts to assess the transportability and performance of the predictive model.
To identify the most informative diagnostic markers, candidate variables were systematically narrowed using recursive feature elimination with cross-validation. This selection procedure established a final five-variable panel comprising patient age, daily stool frequency, white blood cell count, platelet count, and high-density lipoprotein cholesterol. These variables were incorporated into a support vector machine classifier that was assessed for discrimination, calibration, and clinical net benefit.
In the pooled external validation cohorts, the support vector machine classifier achieved a discrimination of 0.834 as measured by the area under the receiver operating characteristic curve. The classifier demonstrated an overall diagnostic accuracy of 0.769, a sensitivity of 0.525, and a high specificity of 0.950 when identifying ulcerative colitis and Crohn’s disease in these independent patient populations.
Model feature contributions were interpreted using Shapley additive explanations, which demonstrated consistent directional influences on the classifier output. Increasing daily stool frequency, advancing age, higher white blood cell counts, and increased high-density lipoprotein cholesterol generally shifted predictions toward ulcerative colitis, whereas higher platelet counts shifted predictions toward Crohn’s disease. High-density lipoprotein cholesterol contributed meaningful diagnostic information beyond that provided by the other four variables.
The study authors highlighted that the high-density lipoprotein cholesterol component may reflect metabolic and inflammatory differences between the two conditions. However, the authors emphasized that the predictive model is intended strictly as an adjunct to early clinical triage rather than a substitute for definitive endoscopy, histopathology, and diagnostic imaging. To allow researchers and clinicians to test the model, an online calculator was deployed at https://uc-vs-cd-predictor-a3fq3ocovi7pagqpwjwh4m.streamlit.app/.
Subject of Research: Machine learning differentiation of ulcerative colitis and Crohn’s disease using routine laboratory indices
Article Title: Association of high-density lipoprotein cholesterol and routine laboratory indices with early differentiation of ulcerative colitis and Crohn’s disease: a multicenter cohort study
Article References: Liu, W., Wei, L., Zhao, R., Zhong, Q., Liu, W., Zhang, M., Jin, L., Yang, Z. X., Sun, N., & Xue, G. (2026). Association of high-density lipoprotein cholesterol and routine laboratory indices with early differentiation of ulcerative colitis and Crohn’s disease: a multicenter cohort study. Lipids in Health and Disease. https://doi.org/10.1186/s12944-026-03090-2
Image Credits: AI Generated
DOI: 10.1186/s12944-026-03090-2
Keywords: ulcerative colitis, Crohn’s disease, high-density lipoprotein cholesterol, HDL-C, support vector machine, SVM classifier, recursive feature elimination, cross-validation, Shapley additive explanations, SHAP values, white blood cell count, platelet count, daily stool frequency, patient age, multicenter cohort study, retrospective study, external validation, receiver operating characteristic, early triage
News Source: Ophelia Keating. (November 18, 2026). Five Clinical Measures Help Distinguish Ulcerative Colitis From Crohn’s Disease. Scienmag.



