Development and evaluation of a predictive model for central precocious puberty in girls.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
BACKGROUND: Early puberty, particularly central precocious puberty (CPP), is an increasingly common pediatric endocrine disorder affecting young girls. Existing CPP predictive models either rely on post-gonadotropin-releasing hormone (GnRH) stimulation data or lack comprehensive integration of clinical, laboratory, and imaging indices, limiting their utility for early screening. This study developed a single blood draw-based predictive model for CPP screening or diagnosis. METHODS: This retrospective study was conducted at a single center (The Second Affiliated Hospital of Shantou University Medical College) and included girls diagnosed with precocious puberty. The training cohort consisted of patients enrolled from April 2021 to April 2023, while the external validation cohort comprised patients enrolled from January to November 2025. Inclusion criteria included secondary sexual traits before age 8 and completion of the GnRH test; exclusion criteria were >30% missing data, organic lesions, and tumors. Clinical, laboratory, and imaging data were analyzed using least absolute shrinkage and selection operator (LASSO) regression and logistic model regression for variable selection and model construction. The gold standard for CPP diagnosis was the GnRH stimulation test [peak luteinizing hormone (LH) ≥5.0 IU/L and LH/follicle-stimulating hormone (FSH) ratio >0.6]. Model reliability was assessed using area under the receiver operating characteristic curve (AUC), decision curve analysis, Brier score, 10-fold cross-validation and external validation. RESULTS: This study enrolled 206 girls with precocious puberty who presented to our hospital from April 2021 to April 2023 as the training cohort. Of these, 88 (42.7%) had CPP with a mean age of 8.32±1.10 years, and 118 (57.3%) had non-CPP with a mean age of 6.98±1.18 years. The LASSO-logistic regression model identified six predictors for CPP: age, estradiol, FSH, LH, breast Tanner stage, and bone age. The model demonstrated good discriminative performance, with an AUC of 0.92 [95% confidence interval (CI): 0.889-0.959], the sensitivity of 0.886, specificity of 0.805 and a Brier score of 0.113. The calibration curve showed good agreement between predicted probabilities and observed values. Decision curve analysis indicated that the model could yield good clinical net benefit across a wide range of threshold probabilities. Ten-fold internal cross-validation also confirmed that the predictive model had good performance, with an AUC of 0.929 and the Brier score of 0.12690. For external validation, this study enrolled 97 girls with precocious puberty who presented to our hospital from January to November 2025. The results showed that the model maintained good performance [AUC: 0.924 (95% CI: 0.861-0.982)], sensitivity of 0.795, specificity of 0.966, Brier score =0.094). For clinical convenience, the prediction model has been deployed on a web-based dynamic nomogram platform. CONCLUSIONS: In this study, we developed a high-performing dynamic nomogram web page that accurately predicts individual CPP risk in girls, aiding screening and diagnosis.