Objective To analyze the predictive value of patient-reported outcomes (PRO) combined with hematological indicators for moderate-to-severe endoscopic activity in patients with ulcerative colitis (UC).
Methods Clinical data of UC patients were retrospectively collected, including PRO, hematological parameters and endoscopic findings. Based on the Mayo endoscopic score, patients were divided into remission and mild activity group (< 2 points) and moderate-to-severe activity group (≥ 2 points). Independent influencing factors for moderate-to-severe endoscopic activity were screened through multivariate Logistic regression analysis, and a binary Logistic regression model was constructed. The receiver operating characteristic (ROC) curve was used to evaluate the discriminatory ability of the predictive model. The calibration of the model was assessed using calibration curves and the Spiegelhalter Z-test, and the model's performance was further validated in an external validation cohort.
Results Rectal bleeding (RB), C-reactive protein/albumin (CAR) and erythrocyte sedimentation rate (ESR) were independent influencing factors for moderate-to-severe endoscopic activity (P < 0.05). The area under the curve (AUC) of the nomogram predictive model constructed based on the logistic regression analysis results was 0.848 (95%CI, 0.798 to 0.900), with sensitivity of 76.6% and specificity of 79.6%. In the external validation cohort, the model's AUC was 0.778 (95%CI, 0.699 to 0.857), and there was no statistically significant difference compared with the AUC of the training cohort model (P>0.05). In both the training and validation cohorts, the Spiegelhalter Z-test results indicated that the model had good goodness-of-fit (P>0.05).
Conclusion RB, CAR and ESR are independent influencing factors for moderate-to-severe endoscopic activity in UC patients. The combination of RB, CAR and ESR has high predictive value for moderate-to-severe endoscopic activity in UC patients, with good discriminatory and calibration abilities.
DownLoad: