Predicting Risk of Insulin Resistance in a Chinese Population with Polycystic Ovary Syndrome: Designing and Testing a New Predictive Nomogram

Abstract

Background. This research is aimed at establishing and internally validating the risk nomogram of insulin resistance (IR) in a Chinese population of patients with polycystic ovary syndrome (PCOS). Methods. We developed a predictive model based on a training dataset of 145 PCOS patients, and data were collected between March 2018 and May 2019. The least absolute shrinkage and selection operator regression model was used to optimize function selection for the insulin resistance risk model. Multivariable logistic regression analysis was used to construct a prediction model integrating the function selected in the regression model of the least absolute shrinkage and selection operator. The predicting model’s characteristics of prejudice, disease, and lifestyle were analyzed using the C-index, the calibration diagram, and the study of the decision curve. External validity was assessed using the validation of bootstrapping. Results. Predictors contained in the prediction nomogram included occupation, disease durations (years), BMI, current use of metformin, and activities. With a C-index of 0.739 (95 percent confidence interval: 0.644–0.830), the model showed good differentiation and proper calibration. In the interval validation, a high C-index value of 0.681 could still be achieved. Examination of the decision curve found that the IR nomogram was clinically useful when the intervention was determined at the 11 percent IR potential threshold. Conclusion. This novel IR nomogram incorporates occupation, disease durations (years), BMI, current use of metformin, and activities. This nomogram could be used to promote the estimation of individual IR risk in patients with PCOS.