# Development of a predictive model for high-risk infertility based on TCM constitution and syndrome: a secondary analysis

## TL;DR

This study evaluates how Traditional Chinese Medicine (TCM) variables can help predict high-risk infertility when combined with Western medical data.

## Contribution

The study introduces a predictive model integrating TCM and Western medical factors for infertility risk stratification.

## Key findings

- The integrated model (Model 3) showed slight improvement in discrimination but exhibited miscalibration after optimism correction.
- The modern medicine model (Model 2) demonstrated more robust predictive performance compared to the integrated model.
- TCM variables contributed limited discriminative ability on their own (AUC = 0.606).

## Abstract

To evaluate the informational contribution of Traditional Chinese Medicine (TCM) variables to high-risk infertility stratification and to develop predictive models integrating TCM and Western medical factors.

Data from 409 women with infertility were analyzed. High-risk infertility was defined as infertility duration >3 years or the presence of ≥2 documented infertility-related causes. Demographic, reproductive, menstrual, and Traditional Chinese Medicine (TCM) constitution and syndrome variables were collected. Three predictive models— a TCM model (Model 1), a modern medicine model (Model 2), and an integrated model (Model 3)—were developed using independent candidate variable sets for each model. Variable selection was performed separately for each model using Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression with 10-fold cross-validation. Multivariable logistic regression models were then fitted based on the selected predictors. Model performance was assessed using Receiver Operating Characteristic (ROC) analysis, calibration curves, and decision curve analysis (DCA). Internal validation and optimism correction were conducted using bootstrap resampling with an out-of-bag strategy.

Among the participants, 163 women (39.9%) were classified as high-risk infertility. The high-risk group was older and differed significantly in reproductive history and clinical characteristics. The Model 3 showed a slight improvement in apparent discrimination compared with the Model 2 [area under the curve (AUC) = 0.790 vs. 0.771]. However, after optimism correction, the Model 2 demonstrated more robust discrimination, calibration, and overall predictive accuracy. The Model 3 exhibited miscalibration after correction, indicating the need for further shrinkage and recalibration. The Model 1 showed limited discriminative ability (AUC = 0.606).

After internal validation, the modern medicine model showed the most stable predictive performance. The integrative model mainly contributed to exploring the potential informational value of TCM variables in infertility risk stratification rather than improving predictive accuracy. These findings provide methodological support for infertility risk assessment within an integrative Chinese and Western medicine framework.