Screening risk factors for carotid plaque formation based on LASSO regression model and constructing a prediction model in patients with prehypertension

SANG Yao; LYU Lei; JIA Yuqing; CHEN Xiaoqing; ZHANG Chen; ZHANG Ling; WU Yanmei

doi:10.3969/j.issn.1674-4055.2025.10.009

Abstract：Objective To screen factors for carotid plaque formation based on regression model of least absolute shrinkage and selection operator(LASSO),and construct a prediction model based on these factors in patients with prehypertension(PHT).Methods PHT patients(n=752)were selected from Heze Municipal Hospital from June 2022 to May 2023.According to the formation of carotid plaques,patients were divided into a plaque group(n=295)and a plaque free group(n=457).The baseline data was compared between 2 groups.The factors that may affect the formation of carotid plaques in PHT patients were screened by using LASSO regression model and analyzed by using binary Logistic regression analysis.Based on the analysis results,a nomogram model was constructed and internally validated,and the model's predictive value to carotid plaque formation was analyzed by using ROC curve in PHT patients.Results Age,body mass index(BMI),triglycerides(TG),uric acid(UA)and percentage of staying up late were higher(P<0.05),and high-density lipoprotein cholesterol(HDL-C)was lower in plaque group than those in plaque free group.There were 6 potential relative factors screened by LASSO regression model,including age,BMI,staying up late,TG,HDL-C and UA.The results of Logistic regression analysis showed that age and BMI were independent risk factors(OR>1,P<0.05),and HDL-C was a protective factor(OR<1,P<0.05)for carotid plaque formation in PHT patients.The nomogram model was validated by using Bootstrap internal validation method.The concordance index value was 0.702.The ROC curve showed that AUC for predicting carotid plaque formation in PHT patients by the nomogram was 0.702(95%CI:0.665～0.739,P<0.001),with a specificity of 0.595,sensitivity of 0.719,and Youden index of 0.314,indicating certain predictive value.The results after plotting the decision curve showed that within the threshold ranges of 0.000～0.619 and 0.834～1.000,the net benefit rate of combining age,BMI and HDL-C to predict carotid plaque formation was superior to that of any single predictor alone in PHT patients.Conclusion Age,BMI and HDL-C are influence factors for carotid plaque formation in PHT patients.The nomogram model,constructed based on above influence factors,can effectively predict risk of carotid plaque formation in PHT patients.