Risk Factor Analysis and Nomogram Predictive Model Construction for Neonatal Hypospadias: A Large-Scale Multicenter Study in East China

Five obstetric factors — gestational hypertension, abnormal placenta, and preterm birth among them — predicted hypospadias with AUC 0.857 in temporal validation

Journal: The Journal of Maternal-Fetal & Neonatal Medicine | Published: 2026-04-27 | Type: Multicenter Retrospective Study | PMID: 42045102 Authors: Hu Wen, Zhao Jian, Chen Cheng, et al. (Women's Hospital, Zhejiang University School of Medicine, Hangzhou; Xinchang MCH Hospital; Quzhou MCH Hospital) Funding/COI: Not disclosed

Summary

A retrospective study of 42,809 male births across three East China hospitals (January 2019–December 2021) found a hypospadias incidence of 0.325% and identified five independent obstetric risk factors: premature birth, multiple pregnancies, gestational hypertension, fetal distress, and abnormal placental morphology. The authors assembled these into a nomogram predictive model that achieved AUC 0.857 on temporal validation — a reasonable performance for a rare congenital condition. The model was never tested outside Zhejiang Province.

Claims

• Hypospadias incidence: 0.325% (139 cases among 42,809 male births; 121 cases included in final analysis)
• Five independent risk factors identified: premature birth, multiple pregnancies, gestational hypertension, fetal distress, abnormal placental morphology
• Nomogram internal validation: AUC 0.834 (95% CI 0.769–0.900), specificity 85.6%, sensitivity 68.9%
• Nomogram temporal validation: AUC 0.857 (95% CI 0.795–0.919), specificity 74.2%, sensitivity 85.0%
• Calibration curve slopes close to 1 in both validations, indicating good model fit

Study Quality

This is a retrospective cohort study with a large birth pool — 42,809 male deliveries — though the effective case count for modeling is only 121, after unexplained exclusions reduced the cohort from 42,809 to 35,517 newborns and from 139 to 121 hypospadias cases. The multicenter design across three regional hospitals is a genuine methodological strength. Using both internal and temporal validation (the latter drawing on a later time window from the same institutions) is more rigorous than leave-one-out cross-validation alone, though it is not the same as external validation.

The retrospective design means all identified "risk factors" are associations, not causes. The five predictors are all obstetric complications that may be downstream of unmeasured exposures — endocrine disruptors, maternal age, parity, or environmental factors — none of which appear to have been adjusted for. The abstract provides no odds ratios or relative risks for individual predictors, only the composite model performance, which makes it impossible to assess which factor carries the most weight.

Red Flags

• Funding and conflicts of interest not disclosed — a significant transparency gap for a university hospital study
• 18 cases dropped between total births (139) and analysis (121); exclusion criteria not explained in the abstract
• All hospitals are in Zhejiang Province — "multicenter" does not mean geographically diverse
• Temporal validation uses the same institutions as training; no external cohort
• No hypospadias severity grading (distal vs. proximal), which likely have different etiologies and may have been pooled inappropriately
• Incidence of 0.325% is at the lower end of global estimates (0.2–0.9%); possible underdiagnosis not discussed
• No individual predictor effect sizes reported — nomogram without odds ratios is opaque

Strengths

• Large birth cohort for a rare condition (42,809 deliveries)
• Multicenter design reduces single-site selection bias
• Both internal and temporal validation performed
• Calibration assessed beyond AUC via calibration curves
• Objective medical record data eliminates recall bias

Verdict

A regionally competent retrospective study that builds a modestly useful clinical nomogram from known obstetric risk factors. The AUC of 0.857 on temporal validation is a legitimate result. But the missing funding/COI disclosure, unexplained case exclusions, absence of individual predictor effect sizes, and Zhejiang-only data all limit how far this should travel. The nomogram may have value at the hospitals that built it; whether it generalizes to populations with different environmental exposures and genetic backgrounds is an open question this paper does not address.