Risk factors analysis and risk prediction model construction for central precocious puberty in children

ZHANG Xin-ran; WANG Yi-dan; YANG Yi-nan; XIE Yun

doi:10.13820/j.cnki.gdyx.20241973

Abstract：Objective To investigate the risk factors of central precocious puberty(CPP)in children and to de-velop a risk prediction model.Methods A total of 362 children diagnosed with precocious puberty in our hospital be-tween January 2019 and January 2022 were included.Using a 7∶3 ratio,participants were randomly assigned to a model-ing group(n=253)and a validation group(n=109).The modeling group was further divided into a CPP group(n=177)and a non-CPP group(n=76)based on diagnostic criteria.Multivariate logistic regression was used to identify in-dependent risk factors for CPP.A nomogram was constructed using R software,and the model's predictive performance was evaluated.Results There were statistically significant differences between the CPP and non-CPP groups in the fol-lowing variables:frequent high-protein diet,frequent intake of tonic foods,preference for sweets,watching romantic TV shows,family harmony,screen time(＞28 hours per week),and maternal menarche age(P＜0.05).Multivariate logis-tic regression analysis identified the following as independent risk factors for CPP:frequent high-protein diet,frequent intake of tonic foods,preference for sweets,screen time＞28 hours/week,watching romantic TV shows,family disharmo-ny,and maternal age at menarche ≤13 years(all P＜0.05).The area under the receiver operating characteristic(ROC)curve(AUC)for the modeling group was 0.812(95％CI:0.759-0.866),indicating good discrimination,with a Hos-mer-Lemeshow(H-L)test statistic of x2=8.650(P=0.373),showing good calibration.In the validation group,the AUC was 0.956(95％CI:0.923-0.989)and the H-L test x2=6.342(P=0.601),indicating excellent predictive performance and consistency.Conclusion Frequent high-protein diet,tonic food intake,sweet preference,excessive screen time,exposure to romantic media,family disharmony,and early maternal menarche are significant risk factors for CPP in children.The developed nomogram model demonstrates good discrimination and calibration and can be used as a visual and practical tool for predicting the risk of CPP.