Development of a simplified prediction model for diagnosing progressive central precocious puberty using clinical and pelvic ultrasound parameters

Abstract

This study aimed to explore the predictive value of clinical and pelvic ultrasound parameters for diagnosing central precocious puberty (CPP) and to establish a clinically useful simplified prediction model to differentiate progressive CPP (P-CP) from nonprogressive precocious puberty (N-PP). Girls aged <9 years with secondary sexual development who underwent a gonadotropin-releasing hormone stimulation test and pelvic ultrasound between September 2020 and November 2023 were retrospectively included and divided into the P-CP and N-PP groups. Logistic regression analysis was used to determine the significant parameters and develop prediction models. The diagnostic performance of the models was compared using the area under the receiver operating characteristic curve (AUC) analysis and the Delong method. The continuous net reclassification improvement (cNRI) and absolute integrated discrimination improvement (IDI) were used to determine the additive effects of ultrasound parameters. A nomogram scoring system was constructed based on a simplified model to predict the probability of developing P-CP. A total of 109 girls were included, with 64 (58.7%) in the P-CP group. Age, bone age, height, height minus midparental height, basal luteinizing hormone (LH), follicle-stimulating hormone, estradiol, insulin-like growth factor-I, Tanner stage, and cervical and fundus width were significant parameters for the diagnosis of P-CP. The models with ultrasound parameters yielded significantly higher cNRI and IDI values than the models without ultrasound parameters. The simplified model was composed of basal LH, estradiol, and fundus width that showed an AUC value of 0.93 (95% confidence interval: 0.88-0.98) with a cutoff value of 16. In conclusion, adding pelvic ultrasound parameters to traditional clinical results has an additive effect on P-CP screening. A simplified predictive model is effective for CPP screening in real-world clinics. These findings highlight the potential of the prediction model to overcome the limitations of the classical diagnostic approach for CPP in children.