Development and validation of a model for predicting the adult height of girls with idiopathic central precocious puberty

The aim was to develop a model to predict the adult height (AH) of idiopathic central precocious puberty (ICPP) girls who underwent gonadotropin-releasing hormone analog (GnRHa) treatment. Data analysis included 258 girls with ICPP. Among them, 101 girls who reached final AH (FAH) with GnRHa treatment were analyzed to verify three previous prediction models and develop a unique model based on multiple linear regression. The control group consisted of 41 untreated ICPP girls. Moreover, 116 girls treated with GnRHa who almost attained FAH were included for external validation. Based on our cohorts, all of the three previously published models underestimated the FAH with an R of 0.667, 0.793, and 0.664. The AH prediction model was built as follows: Calculated AH (cm) = 1.89630 * Height SDS + 2.29927 * Height SDS for bone age + 0.40776 * Target height + 100.16684 (R-2 = 0.66 and adjusted R-2 = 0.65). Internal validation showed a mean root mean squared error (RMSE) of 2.16 cm and a mean absolute error (MAE) of 1.64 cm. External validation showed that a significant error (> 1 SD) appeared only in 7 of 116 girls (6.0%). The model is displayed on the website: http:// cpppr edict.shiny apps. io/ dynno mapp.Conclusion: A model for predicting the AH of girls with ICPP was developed incorporating the variables of height SDS, height SDS for bone age, and target height. The internal and external validation ensures an appropriate degree of discrimination and calibration of the prediction model.

Automated summary

The aim of this study was to develop a model for predicting the adult height (AH) of girls with idiopathic central precocious puberty (ICPP) who received gonadotropin-releasing hormone analog (GnRHa) treatment. A unique model based on multiple linear regression was developed and verified using data from 258 girls with ICPP. The model incorporated variables of height SDS, height SDS for bone age, and target height, and showed good discrimination and calibration in both internal and external validation.