Anatomy-guided prediction of complications after hypospadias repair: development and multicenter validation of stratified machine learning models

Introduction

Hypospadias is one of the most common congenital anomalies of the male external genitalia and encompasses a broad spectrum of urethral, glanular, and ventral penile abnormalities (1,2). Although many repairs achieve satisfactory functional and cosmetic outcomes, postoperative complications and secondary procedures remain an important burden for children and families (3,4). In contemporary pediatric urology, the clinically relevant outcome is often not a minor postoperative finding, but a complication that requires another intervention, such as urethrocutaneous fistula repair, treatment of urethral stricture, diverticulectomy or urethroplasty, or correction of recurrent ventral curvature.

Anatomy and reconstruction pathway are central to this risk. Ventral curvature management follows a stepwise surgical logic, and persistent curvature may require urethral plate transection (UPT), after which a substitution or staged urethroplasty is usually needed (5,6). Outcomes in proximal or high-complexity repairs vary substantially across centers and techniques, and severity scores or traditional prediction models only partly capture this heterogeneity (7,8). Existing scoring systems such as Glans-Urethral Meatus-Shaft (GMS) improve structured description, while prior complication prediction work has shown the potential value of multivariable risk estimation, but the optimal way to account for the different risk structure of plate-preserving and plate-transecting repairs remains unclear (9,10).

Machine-learning methods can model nonlinear relationships between anatomical variables and outcomes, but prediction studies must be reported conservatively and transparently (11,12). We therefore developed and externally validated two stratified machine-learning models for postoperative complications requiring surgical intervention after primary hypospadias repair, one for non-UPT patients and one for UPT patients. We hypothesized that pathway-specific models would provide clinically interpretable risk estimates that could support perioperative counseling and follow-up planning. We present this article in accordance with the TRIPOD reporting checklist (available at https://tau.amegroups.com/article/view/10.21037/tau-2026-1-0176/rc).

Methods

Study design and participants

This cohort study used prospectively collected clinical data from boys undergoing primary hypospadias repair between December 2018 and December 2020 at 17 tertiary referral hospitals in China. The development cohort consisted of 584 patients treated at Beijing Children’s Hospital, National Center for Children’s Health. The external validation cohort consisted of 511 patients treated at 16 additional tertiary institutions. The available final modeling cohort contained 1,095 consecutive eligible records with complete candidate predictors and outcome ascertainment for model development and validation. Eligible patients were boys undergoing primary urethroplasty for hypospadias with complete anatomical measurements, UPT status, procedure type, and postoperative outcome assessment. Patients undergoing procedures outside the study’s standardized curvature-correction pathway, redo repairs, or records without sufficient predictor or outcome ascertainment were not part of the final modeling cohort.

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Beijing Children’s Hospital, Capital Medical University, National Center for Children’s Health (IEC-C-008-A08-V.05.1). Written informed consent was obtained from patients’ guardians. All participating hospitals were informed of and agreed to the study. The study was registered in the Chinese Clinical Trial Registry (ChiCTR1900023055).

Surgical pathway and UPT stratification

Ventral curvature was managed using a stepwise approach. After complete penile degloving and release of ventral tethering tissues, an artificial erection test was performed. When residual ventral curvature was <30°, curvature correction was completed with dorsal plication when needed. When clinically significant curvature persisted, the urethral plate (UP) was transected at the corona, and further dorsal plication was performed if residual curvature remained. Subsequent urethroplasty was performed according to local tissue conditions and surgeon judgment. Plate-preserving repairs included tubularized incised plate (TIP) urethroplasty and Onlay flap repair; repairs after UPT included transverse preputial island flap (TPIF) urethroplasty and two-stage urethroplasty. For model development, UPT status was used to assign patients to the non-UPT or UPT prediction pathway.

Predictors and outcome

Candidate predictors were selected a priori from routinely recorded anatomical and operative variables. Preoperative variables included age, penile length, glans length, glans width, UP width, preoperative meatal position, and preoperative ventral curvature. UP width was measured at the narrowest point after gentle horizontal traction. Meatal position was coded using a modified Duckett system from 1 to 9, ranging from normal/fossa navicularis to perineal position (13,14). Intraoperative variables included whether curvature was corrected after degloving, dorsal plication, meatal position after ventral curvature correction, length of deficient urethra, and primary urethroplasty type. The final predictor set for each model and the clinical time point at which each variable becomes available are summarized in Table S1.

The primary outcome was postoperative complications requiring surgical intervention, equivalent to Clavien-Dindo grade IIIb or higher (15). These included urethrocutaneous fistula requiring fistula repair or redo urethroplasty, urethral stricture requiring dilation under anesthesia, urethrotomy, urethrostomy, or urethroplasty, urethral diverticulum requiring diverticulectomy or urethroplasty, and recurrent ventral curvature >30° requiring operative correction. Minor nonoperative events and cosmetic concerns not requiring surgical treatment were not modeled. Outcomes were assessed through outpatient follow-up and telephone review according to each center’s postoperative follow-up protocol.

Model development, validation, and interpretation

Feature selection was performed separately in the non-UPT and UPT development cohorts. Variables with low variance were first removed using a variance-threshold screen, followed by recursive feature elimination with cross-validation. Eight algorithms were compared within each stratum: logistic regression, k-nearest neighbors, support vector machine, decision tree, random forest (RF), gradient boosting decision tree, adaptive boosting (AdaBoost), and light gradient boosting machine (LightGBM). Model selection was based primarily on area under the curve (AUC) in the development cohort, with accuracy, precision, recall/sensitivity, and F1-score used as supportive threshold-dependent metrics.

After the best-performing algorithm was selected for each stratum, hyperparameters were tuned by grid-search procedures guided by learning curves; the selected hyperparameters are reported in the Results. The finalized models were applied to the external validation cohort without updating. Calibration plots and decision curve analysis were used to evaluate agreement and potential clinical utility across threshold probabilities (16,17). SHapley Additive exPlanations (SHAP) was used for global and individual-level interpretation of the final models (18). Model performance was interpreted according to discrimination, calibration, and potential clinical usefulness rather than AUC alone (19).

Web calculators

Two web calculators were implemented using Streamlit, one for non-UPT patients and one for UPT patients. The calculators output predicted risk and SHAP-based explanations for individual patients. They are intended for perioperative counseling and follow-up planning after the surgical pathway is known. They are not intended to decide whether UPT should be performed.

Statistical analysis

Continuous variables are summarized as median [interquartile range] and categorical variables as number (percentage). Between-group comparisons used the Mann-Whitney U test for continuous variables and Chi-squared or Fisher exact tests for categorical variables. Analyses were performed using R and Python/scikit-learn. Two-sided P values <0.05 were considered statistically significant.

Results

Patient characteristics

The final modeling cohort included 1,095 boys: 584 in the development cohort and 511 in the external validation cohort. UPT was performed in 577 patients overall, including 333/584 (57.0%) in the development cohort and 244/511 (47.7%) in the external validation cohort. All candidate predictors in the final modeling cohort were complete. Baseline anatomical and operative characteristics by UPT status are shown in Table 1 and Table 2.

In the development cohort, complications requiring surgical intervention occurred in 78/251 (31.1%) non-UPT patients and 162/333 (48.6%) UPT patients (P<0.001). In the external validation cohort, the corresponding rates were 64/267 (24.0%) and 118/244 (48.4%) (P<0.001). The distribution of individual complications differed by stratum, particularly for urethral stricture and urethral diverticulum.

Feature selection and algorithm comparison

In the non-UPT stratum, feature selection retained six predictors: age, glans length, glans width, preoperative ventral curvature, length of deficient urethra, and UP width. In the UPT stratum, feature selection retained seven predictors: age, penile length, glans length, glans width, preoperative ventral curvature, length of deficient urethra, and meatal position after ventral curvature correction.

Among the eight algorithms, RF performed best in the non-UPT stratum before final tuning (AUC =0.741), while decision tree performed best in the UPT stratum (AUC =0.800) (Table 3). Comparative receiver operating characteristic (ROC) curves and Sankey diagrams of observed and predicted classes for all eight algorithms are shown in Figure 1. After hyperparameter optimization, the final non-UPT RF model used n_estimators =100, max_depth =10, min_samples_split =5, and min_samples_leaf =2. The final UPT decision tree model used max_depth =8, min_samples_split =10, and min_samples_leaf =5. Optimized final-model performance is summarized in Table S2.

Figure 1 Performance of machine-learning models for predicting postoperative complications requiring surgical intervention after hypospadias repair. (A) ROC curves for eight machine learning models in the non-UPT group. (B) ROC curves for eight machine learning models in the UPT group. (C) Sankey diagram for non-UPT group: the Sankey diagram displays the confusion matrices for various machine learning models in the non-UPT group. The left rectangular box represents the actual class, while the right box shows the predicted class. Correct classifications are highlighted in green, and incorrect classifications are shown in red. (D) Sankey diagram for UPT group: the Sankey diagram displays the confusion matrices for various machine learning models in the UPT group. The left rectangular box represents the actual class, while the right box shows the predicted class. Correct classifications are highlighted in green, and incorrect classifications are shown in red. AdaBoost, adaptive boosting; AUC, area under the curve; DT, decision tree; GBDT, gradient boosting decision tree; KNN, K-nearest neighbors; LightGBM/LGBM, light gradient boosting machine; LR, logistic regression; RF, random forest; ROC, receiver operating characteristic; SVM, support vector machine; UPT, urethral plate transection.

Model validation, interpretation, and web calculators

The optimized RF model for non-UPT patients achieved an AUC of 0.85 in external validation. Global SHAP importance ranked length of deficient urethra, age, glans length, preoperative ventral curvature, glans width, and UP width. The optimized decision tree model for UPT patients achieved an AUC of 0.84 in external validation. Global SHAP importance ranked age, meatal position after ventral curvature correction, length of deficient urethra, penile length, glans length, glans width, and preoperative ventral curvature. Calibration plots suggested acceptable agreement between predicted and observed risk, while decision curve analysis showed potential net benefit across clinically plausible threshold ranges. The ROC, calibration, decision-curve, and SHAP-based interpretation results for the final models are summarized in Figure 2.

Figure 2 Evaluation and interpretation of the final stratified machine-learning models. (A) ROC curves for the RF model predicting complications requiring surgical intervention in both the training and external validation cohorts. (B) ROC curves for the DT model predicting complications requiring surgical intervention in both the training and external validation cohorts. (C) Calibration curves for the RF and DT models, showing the relationship between model-predicted and observed probabilities of complications requiring surgical intervention. (D) DCA curves for the RF and DT models, illustrating the net benefit (y-axis) across a range of risk thresholds (x-axis) compared with intervention for all participants or no intervention. (E) Global model explanation via SHAP: SHAP summary beeswarm plots showing feature importance ranked by mean absolute SHAP value. Dot color represents the original feature value (red for higher values and blue for lower values), while horizontal position represents the direction and magnitude of each feature’s contribution to the model output. (F) Local model explanation via SHAP: SHAP waterfall plots showing the contribution of each feature for individual patients at low or high risk of postoperative complications requiring surgical intervention. Case 1 (left) shows a non-UPT patient with a higher predicted risk by the RF model, driven mainly by preoperative curvature and length of deficient urethra. Case 2 (right) shows a UPT patient with a lower predicted risk by the DT model, with glans width and age contributing most strongly toward the lower prediction. (G) SHAP force plots: contributions of individual features to the predictions for the illustrated patients. Features increasing the predicted probability are shown in red and those decreasing it in blue; longer bars indicate greater influence. DCA, decision curve analysis; DT, decision tree; RF, random forest; ROC, receiver operating characteristic; SHAP, SHapley Additive exPlanations; UPT, urethral plate transection.

Two Streamlit calculators were developed: RF model (non-UPT): https://bchzh-hypospadias-non-upt.streamlit.app/ and DT model (UPT): https://bchzh-hypospadias-upt.streamlit.app/. Because the current Streamlit implementation does not hard-code variable boundaries, users should restrict inputs to the observed ranges of the final modeling cohort to avoid extrapolation. For the non-UPT calculator, these ranges are age 3–180 months, glans length 5–22 mm, glans width 7–25 mm, preoperative curvature 0–105°, deficient urethra length 0.05–4.5 cm, and UP width 1–12 mm. For the UPT calculator, the corresponding ranges are age 8–134 months, penile length 2.0–6.8 cm, glans length 5–22 mm, glans width 8–25 mm, preoperative curvature 10–130°, deficient urethra length 1.5–7.0 cm, and post-correction meatal position codes 3–9 (Table S3).

Discussion

In this multicenter study, stratified machine-learning models estimated the risk of postoperative complications requiring surgical intervention after primary hypospadias repair. The final models were intentionally aligned with the two major reconstruction pathways: plate-preserving repair in non-UPT patients and substitution or staged reconstruction in UPT patients. This distinction is clinically relevant because the mechanisms leading to fistula, stricture, diverticulum, and recurrent curvature may differ by reconstruction pathway. The models achieved moderate discrimination and provided interpretable predictor profiles, supporting their use as adjuncts for counseling and follow-up planning. These predictions address reoperation-requiring complications and should be interpreted within the broader multidimensional assessment of hypospadias outcomes (20).

The rationale for stratification by UPT status is both anatomical and clinical. UPT is not merely a binary variable; it reflects a transition from plate-preserving urethroplasty toward repair pathways that require alternative tissue transfer or staging. In non-UPT repairs, risk may be driven by UP width, glans configuration, residual curvature, and the length over which the native plate must be tubularized or augmented. In UPT repairs, risk may be more closely related to the severity of curvature, urethral defect length, corrected meatal location, and the quality and vascularity of transferred tissue. This interpretation is consistent with the high complication burden described in proximal repairs and with contemporary discussion about staged reconstruction for complex cases (21-23).

The SHAP results were clinically plausible. Length of deficient urethra had the greatest global influence in the non-UPT model, consistent with the greater suture-line length, tissue-transfer requirement, and tension associated with longer reconstructions. In the UPT model, age and post-correction meatal position had the greatest global influence, while preoperative ventral curvature remained a retained anatomical predictor. In non-UPT patients, glans dimensions and UP width also contributed to risk. These findings complement earlier structured assessment work, including GMS and objective penile outcome scores, by translating multiple anatomical measures into individualized postoperative risk estimates (24,25).

This work also illustrates both the promise and the limitations of machine learning in pediatric urology. Image-based artificial intelligence has been used to improve hypospadias classification, and automated approaches may eventually reduce interobserver variation in anatomical assessment (26). However, risk prediction after hypospadias repair remains partly constrained by factors not fully captured in structured data, including surgeon experience, center volume, tissue vascularity, flap handling, suture technique, dressing and drainage practices, and family adherence to follow-up. Accordingly, the models should not be interpreted as removing the influence of technique or surgeon experience. These anatomy-guided variables are also consistent with prior evidence concerning urethral plate preservation during curvature correction and the prognostic relevance of urethral plate width (27,28).

The web calculators are intended for practical use after the likely reconstruction pathway is known. A low predicted risk may support routine counseling and standard follow-up, whereas a higher predicted risk may justify more detailed family counseling, closer early surveillance, and a lower threshold for evaluation of voiding symptoms or local swelling. The calculators are not designed to determine whether UPT should be performed, and they should not be used outside the observed variable ranges. The current Streamlit implementation therefore includes usage instructions rather than hard-coded range enforcement.

Several limitations should be noted. First, the study was conducted in Chinese tertiary referral centers, and the spectrum of anatomy, procedures, surgeon experience, and surveillance may differ from other health systems. Second, the outcome was limited to complications requiring surgical intervention. This endpoint is relevant for counseling about reoperation burden, but it does not capture nonoperative complications, urinary flow, cosmetic assessment, patient-reported outcomes, psychosocial adjustment, or sexual function (20). Third, although the final modeling cohort had complete predictor and outcome fields, record-level information was unavailable for cases not retained before assembly of the final modeling cohort. Consequently, we could not report variable-level missingness in nonretained records, compare included with nonretained patients, or perform imputation sensitivity analyses. This may introduce selection bias and limits assessment of the missing-data mechanism. Fourth, individual follow-up duration was unavailable for analysis, and pubertal outcomes were not evaluated. Longer surveillance is important because a substantial proportion of complications and reoperations may be detected beyond early follow-up (29,30). Finally, external validation was multicenter but contemporaneous, and future temporally independent validation is needed.

In summary, UPT-stratified models provide interpretable, anatomy-guided estimates of the risk of postoperative complications requiring surgical intervention after primary hypospadias repair. Their main role is to support structured counseling and follow-up planning. Prospective implementation studies, longer follow-up, and validation in independent international cohorts are needed before broad clinical adoption.

Conclusions

Stratified machine-learning models based on UPT status estimated the risk of postoperative complications requiring surgical intervention after primary hypospadias repair with moderate discrimination and interpretable predictor profiles. The non-UPT RF model and UPT decision tree model may support perioperative counseling and follow-up planning when used within the observed data ranges. They should complement, not replace, surgeon judgment, intraoperative assessment, and center-specific quality control.

Acknowledgments

The authors gratefully acknowledge the Urology Group of the Chinese Society of Pediatric Surgery, Chinese Medical Association, for providing nationwide collaborative support for the external multicentre cohort, including coordination among participating institutions and support for case accrual, data collection, and follow-up. This organizational effort was essential for assembling the high-quality external dataset used in this study. We are especially grateful to the following non-author collaborators: Yun-Man Tang, Department of Urology, Sichuan Academy of Medical Sciences–Sichuan Provincial People’s Hospital, Chengdu, China; Lu-Gang Huang, Department of Urology, West China Hospital of Sichuan University, Chengdu, China; Yi Yang, Department of Urology, Shengjing Hospital of China Medical University, Shenyang, China; Min Chao, Department of Urology, Anhui Children’s Hospital, Hefei, China; Hong Ma, Department of Urology, Affiliated Hospital of Zunyi Medical University (Guizhou Children’s Hospital), Zunyi, China; Jingti Zhang, Department of Urology, Xi’an Children’s Hospital, Xi’an, China; Xuhui Zhang, Department of Urology, Shanxi Children’s Hospital, Taiyuan, China; Shoulin Li, Department of Urology, Shenzhen Children’s Hospital, Shenzhen, China; Ning Li, Department of Pediatric Surgery, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China; Chao Chen, Department of Urology, The First Affiliated Hospital of Guangxi Medical University, Nanning, China; Dawei He, Department of Urology, Children’s Hospital of Chongqing Medical University, Chongqing, China; Wenbo Wu, Department of Urology, Children’s Hospital of Jiangxi Province, Nanchang, China; Hua Xie, Department of Urology, Shanghai Children’s Hospital, Shanghai, China; Yong Guan, Department of Urology, Tianjin Children’s Hospital, Tianjin, China; Yanfang Yang, Department of Urology, Henan Children’s Hospital, Zhengzhou, China; and Bin Yang, Department of Urology, Baoding Children’s Hospital, Baoding, China. Their efforts under demanding clinical conditions made possible the high-quality external dataset on which this multicentre study depends. Finally, we also thank the reviewers for constructive comments that improved the clarity and transparency of the manuscript.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tau.amegroups.com/article/view/10.21037/tau-2026-1-0176/rc

Data Sharing Statement: Available at https://tau.amegroups.com/article/view/10.21037/tau-2026-1-0176/dss

Peer Review File: Available at https://tau.amegroups.com/article/view/10.21037/tau-2026-1-0176/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tau.amegroups.com/article/view/10.21037/tau-2026-1-0176/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Beijing Children’s Hospital, Capital Medical University, National Center for Children’s Health (IEC-C-008-A08-V.05.1). Written informed consent was obtained from patients’ guardians. All participating hospitals were informed of and agreed to the study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Baskin LS. Hypospadias and urethral development. J Urol 2000;163:951-6.
2. van der Horst HJ, de Wall LL. Hypospadias, all there is to know. Eur J Pediatr 2017;176:435-41. [Crossref] [PubMed]
3. Prat D, Natasha A, Polak A, et al. Surgical outcome of different types of primary hypospadias repair during three decades in a single center. Urology 2012;79:1350-3. [Crossref] [PubMed]
4. Lucas J, Hightower T, Weiss DA, et al. Time to Complication Detection after Primary Pediatric Hypospadias Repair: A Large, Single Center, Retrospective Cohort Analysis. J Urol 2020;204:338-44. [Crossref] [PubMed]
5. Castagnetti M, El-Ghoneimi A. Surgical management of primary severe hypospadias in children: an update focusing on penile curvature. Nat Rev Urol 2022;19:147-60. [Crossref] [PubMed]
6. Castagnetti M, El-Ghoneimi A. Surgical management of primary severe hypospadias in children: systematic 20-year review. J Urol 2010;184:1469-74. [Crossref] [PubMed]
7. Pippi Salle JL, Sayed S, Salle A, et al. Proximal hypospadias: A persistent challenge. Single institution outcome analysis of three surgical techniques over a 10-year period. J Pediatr Urol 2016;12:28.e1-28.e7.
8. Snodgrass W, Bush N. Recurrent ventral curvature after proximal TIP hypospadias repair. J Pediatr Urol 2021;17:222.e1-5.
9. Merriman LS, Arlen AM, Broecker BH, et al. The GMS hypospadias score: assessment of inter-observer reliability and correlation with post-operative complications. J Pediatr Urol 2013;9:707-12. [Crossref] [PubMed]
10. Dokter EM, van der Zanden LF, Laumer SJ, et al. Development of a prediction model for postoperative complications after primary hypospadias correction. J Pediatr Surg 2020;55:2209-15. [Crossref] [PubMed]
11. Greener JG, Kandathil SM, Moffat L, et al. A guide to machine learning for biologists. Nat Rev Mol Cell Biol 2022;23:40-55. [Crossref] [PubMed]
12. Collins GS, Reitsma JB, Altman DG, et al. Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med 2015;162:55-63. [Crossref] [PubMed]
13. Duckett JW Jr. Hypospadias. Pediatr Rev 1989;11:37-42. [Crossref] [PubMed]
14. Sarhan O, Saad M, Helmy T, et al. Effect of suturing technique and urethral plate characteristics on complication rate following hypospadias repair: a prospective randomized study. J Urol 2009;182:682-5; discussion 685-6. [Crossref] [PubMed]
15. Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 2004;240:205-13. [Crossref] [PubMed]
16. Van Calster B, Wynants L, Verbeek JFM, et al. Reporting and Interpreting Decision Curve Analysis: A Guide for Investigators. Eur Urol 2018;74:796-804. [Crossref] [PubMed]
17. Van Calster B, McLernon DJ, van Smeden M, et al. Calibration: the Achilles heel of predictive analytics. BMC Med 2019;17:230. [Crossref] [PubMed]
18. Nohara Y, Matsumoto K, Soejima H, et al. Explanation of machine learning models using shapley additive explanation and application for real data in hospital. Comput Methods Programs Biomed 2022;214:106584. [Crossref] [PubMed]
19. Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models: a framework for traditional and novel measures. Epidemiology 2010;21:128-38. [Crossref] [PubMed]
20. Springer A. Assessment of outcome in hypospadias surgery - a review. Front Pediatr 2014;2:2. [Crossref] [PubMed]
21. Long CJ, Chu DI, Tenney RW, et al. Intermediate-Term Followup of Proximal Hypospadias Repair Reveals High Complication Rate. J Urol 2017;197:852-8. [Crossref] [PubMed]
22. Chan YY, D'Oro A, Yerkes EB, et al. Challenging proximal hypospadias repairs: an evolution of technique for two-stage repairs. J Pediatr Urol 2021;17:225.e1-225.e8. [Crossref] [PubMed]
23. Babu R, Chandrasekharam VVS. Meta-analysis comparing the outcomes of single-stage versus two-stage repair for proximal hypospadias in the last decade. J Pediatr Urol 2021;17:681-9. [Crossref] [PubMed]
24. Arlen AM, Kirsch AJ, Leong T, et al. Further analysis of the Glans-Urethral Meatus-Shaft (GMS) hypospadias score: correlation with postoperative complications. J Pediatr Urol 2015;11:71.e1-5. [Crossref] [PubMed]
25. Holland AJ, Smith GH, Ross FI, et al. HOSE: an objective scoring system for evaluating the results of hypospadias surgery. BJU Int 2001;88:255-8. [Crossref] [PubMed]
26. Fernandez N, Lorenzo AJ, Rickard M, et al. Digital pattern recognition for the identification and classification of hypospadias using artificial intelligence versus experienced pediatric urologist. Urology 2021;147:264-9. [Crossref] [PubMed]
27. Snodgrass W, Prieto J. Straightening ventral curvature while preserving the urethral plate in proximal hypospadias repair. J Urol 2009;182:1720-5. [Crossref] [PubMed]
28. Chukwubuike KE, Obianyo NEN, Ekenze SO, et al. Assessment of the effect of urethral plate width on outcome of hypospadias repair. J Pediatr Urol 2019;15:627.e1-6.
29. Anttila A, Lahdes-Vasama T, Pakkasjärvi N, et al. Cumulative re-operation rates during follow-up after hypospadias repair. BJU Int 2024;134:960-6. [Crossref] [PubMed]
30. Zhao J, Sun N, Zhang W, et al. Time to complication after primary pediatric hypospadias repair with transverse preputial island flap urethroplasty. Transl Androl Urol 2024;13:1364-71. [Crossref] [PubMed]