Objectives: To establish an efficient and clinically applicable predictive model and scoring system for central precocious puberty (CPP) in girls, and to develop a diagnostic prediction application.
Methods: A total of 342 girls aged 4 to 9 years with precocious puberty were included, comprising 216 cases of CPP and 126 cases of isolated premature thelarche. Lasso regression was used to screen for predictive factors, and logistic regression was employed to establish the predictive model. Additionally, a scoring system was constructed using the evidence weight binning method. Data from 129 girls aged 4 to 9 years with precocious puberty were collected for external validation of the scoring system.
Results: The logistic regression model incorporated five predictive factors: age, insulin-like growth factor-1 (IGF-1), serum follicle-stimulating hormone (FSH), the luteinizing hormone (LH)/FSH baseline ratio, and uterine thickness. The calculation formula was: ln(P/1-P)=-8.439 + 0.216 × age (years) + 0.008 × IGF-1 (ng/mL) + 0.159 × FSH (mIU/mL) + 9.779 × LH/FSH baseline ratio + 0.284 × uterine thickness (mm). This model demonstrated good discriminative ability (area under the curve=0.892) and calibration (Hosmer-Lemeshow test P>0.05). The scoring system based on this logistic regression model showed good discrimination in both the prediction model and external validation datasets, with areas under the curve of 0.895 and 0.805, respectively. Based on scoring system scores, the population was stratified into three risk levels: high, medium, and low. In the high-risk group, the prevalence of CPP exceeded 90%, while the proportion was lower in the medium and low-risk groups.
Conclusions: The CPP diagnostic predictive model established for girls aged 4 to 9 years exhibits good diagnostic performance. The scoring system can effectively and rapidly stratify the risk of CPP, providing valuable reference for clinical decision-making.
目的: 建立高效和临床易用的女童中枢性性早熟（central precocious puberty, CPP）预测模型和评分卡，并建立诊断预测应用程序。方法: 纳入342例4~9岁性早熟女童，其中CPP患儿216例，单纯性乳房早发育患儿126例。使用Lasso回归筛选预测因子，利用逻辑回归建立预测模型，并借助证据权重分箱法构建评分卡。另外收集129例4~9岁性早熟女童数据对评分卡进行外部验证。结果: 逻辑回归模型纳入5项预测因子：年龄、胰岛素样生长因子-1（insulin-like growth factor 1, IGF-1）、血清促卵泡激素（follicle-stimulating hormone, FSH）、促黄体生成素（luteinizing hormone, LH）/FSH基础比值和子宫厚度，计算公式为：ln（P/1-P）=-8.439+0.216×年龄（岁）+0.008×IGF-1（ng/mL）+0.159×FSH（mIU/mL）+9.779×LH/FSH基础比值+0.284×子宫厚度（mm）。该模型表现出良好的区分度（曲线下面积=0.892）及校准度（Hosmer-Lemeshow检验P>0.05）。基于该逻辑回归模型构建的评分卡在预测模型和外部验证数据集上均有良好的区分度，曲线下面积分别为0.895和0.805。基于评分卡得分将人群划分为CPP高、中、低3个风险层。在高风险人群中，CPP患病率超过90%，而在中低风险人群中这一比例较低。结论: 该研究建立的4~9岁女童CPP诊断预测模型有良好的诊断性能。评分卡能够有效且快速简便地对CPP风险进行分层，为临床决策提供有价值的参考。.
Keywords: Central precocious puberty; Girl; Predictive model; Scoring system.