Abstract
OBJECTIVES: Disease activity assessment is important for Crohn's disease (CD) management, since it involves the initial and subsequent therapeutic schedule. The purpose of this study is to identify a portable and reliable indicator for assessing and predicting activity and severity of CD. DESIGN: A multicentre, cross-sectional, diagnosis-based study. Data were obtained retrospectively from clinical records. SETTING: Patient data for the development cohort and internal validation cohort were collected from the Department of Gastroenterology, Xiangya Hospital, Central South University between January 2017 and June 2021. Patient data for the external validation cohort were collected from the Department of Gastroenterology, Xiangtan Central Hospital between January 2022 and April 2024. PARTICIPANTS: Inpatients diagnosed with CD were potential participants, and those with CD who also had other autoimmune diseases, malignant tumours, pregnancy or lactation were excluded. We identified 224 patients in the development cohort, 96 patients in the internal validation cohort and 80 patients in the external validation cohort. OUTCOME MEASURE: Demographic data and laboratory examination results were collected and seven integrated indices were established. Mann-Whitney U test, Kruskal-Wallis K test, χ(2) test and multivariate logistic regression analysis were used to identify independent predictors. Receiver operating characteristics curve analysis was used to evaluate the performance of integrated indices in CD activity and severity staging, and Delong's test was used for comparison. RESULTS: In the development cohort, platelet-to-albumin ratio (PAR) had the largest area under the curve (AUC) in prediction of activity (AUC of 0.753 (0.687-0.819), sensitivity 73.0%, specificity 68.4%, cut-off value 8.02) and was the only possible alternative in prediction of severity (AUC of 0.770 (0.693-0.848), sensitivity 81.2%, specificity 68.4%, cut-off value 9.71). Moreover, PAR demonstrated coherence in the internal validation cohort, effectively predicting activity (AUC 0.737 (0.639-0.835), sensitivity 65.7%, specificity 79.3%, positive predictive value (PPV) 88.0%, negative predictive value (NPV) 50.0%) and severity (AUC 0.720 (0.591-0.848), sensitivity 88.5%, specificity 77.4%, PPV 76.7%, NPV 64.9%). Furthermore, it showed generalisability in the external validation cohort for predicting activity (AUC 0.661 (0.536-0.785), sensitivity 61.9%, specificity 70.6%, PPV 88.6%, NPV 33.3%) and severity (AUC 0.752 (0.619-0.884), sensitivity 73.1%, specificity 64.9%, PPV 59.4%, NPV 77.4%). CONCLUSIONS: PAR could be a portable index to assess the activity and severity of CD. Several limitations of this study, such as the limited sample size and potential biases, should be overcome in the future via more extensive validation.