Search

Objective Based on the data of patients with ulcerative colitis treated with Five-Flavor Sophora Flavescens Radix Enteric-coated Capsules, we discuss the feasibility and methodology of constructing a prediction model of the efficacy of traditional Chinese medicine. Methods Based on phase Ⅲ clinical trial data of Five-Flavor Sophora Flavescens Radix Enteric-coated Capsules, a total of 274 patients with active ulcerative colitis with complete population data after 8 weeks of treatment in the experimental group ( Five-Flavor Sophora Flavescens Radix Enteric-coated Capsules) were selected. R 4. 1. 0 was used for statistical analysis. First, all the variables to be screened were included to construct a multivariate Logistic regression model. Then, variables with P<0. 05 were screened for inclusion in the final model. The nomogram was plotted. The bootstrap method was used to resample 1 000 times for internal validation of the model, the C index was calculated, and the calibration chart was plotted to evaluate the discrimination and calibration of the model. The model was evaluated by the HosmerLemeshow goodness of fit test and external validation. The overall evaluation of the model was performed by the above methods. Results The prognostic factors for complete remission of ulcerative colitis were younger age (15-65 years old) (OR = 0. 96, 95%CI:0. 94-0. 99, P = 0. 002), non-chronic persistent (compared with primary type) (OR = 0. 36, 95%CI:0. 18-0. 74, P = 0. 005), no microscopic bleeding (OR = 0. 42, 95% CI:0. 22 - 0. 82, P = 0. 011), and decreased erythrocyte sedimentation rate (OR = 0. 97, 95%CI:0. 95-0. 99, P = 0. 006). We used the above variables to construct the final model and drew a nomogram. The corrected C index was 0. 735 ( 95% CI: 0. 672 - 0. 797), indicating that the model’s discrimination was acceptable. The calibration diagram shows that the calibration curve and the standard line roughly fit, indicating the calibration model is acceptable. The Hosmer-Lemeshow goodness of fit test (χ2 = 7. 07, P = 0. 53) showed that the model was well fitted. We conducted an external test of the prediction model. The C index and confidence interval were 0. 54 ( 95% CI: 0. 44 - 0. 65 ). Conclusion It is feasible to construct a prediction model by using data from randomized controlled trials. A prediction model based on the study data may provide an effective evaluation tool for clinical monitoring and targeted therapy after large sample validation. [ABSTRACT FROM AUTHOR]