Development of predictive models for adverse outcomes in children with tethered spinal cord syndrome associated with spinal dysraphism and assessment of their effectiveness

Objective. To develop and evaluate a model for predicting adverse outcomes of microsurgical untethering and vertebral shortening osteotomy in children with tethered spinal cord syndrome associated with spinal dysraphism.

Material and Methods. A retrospective cohort of 120 children with spinal dysraphism and tethered cord syndrome was randomly divided into a training set (Group 1: operated patients) and a test set (Group 0: non-operated children with follow-up data). A comparison of the training and test sets revealed no statistically significant differences in key characteristics, confirming the comparability of the groups. Demographic, clinical-neurological, and neuroimaging features were analyzed. The predictive performance of three models was evaluated: LASSO regression, decision tree, and eXtreme Gradient Boosting (XGB). Class balancing was performed using the SMOTE technique. Model quality was assessed using ROC-AUC, sensitivity/specificity, accuracy, and Brier score (with 95% confidence intervals); clinical utility was quantified using Decision Curve Analysis (DCA); and interpretation was based on feature importance and SHAP values for  the XGB model.

Results. LASSO yielded an ROC-AUC of 0.82 and an accuracy of 80.6 % on the test set, whereas the decision tree showed an ROC-AUC of 0.64 and an accuracy of 69.4 %. The best performance was achieved by XGB (ROC-AUC 0.97; accuracy 86.1 %), indicating an excellent level of discrimination. Key predictors of an adverse outcome included trophic disturbances, lower leg deformity, shortening vertebrotomy, filum terminale thickening ≥2 mm, and pelvic function parameters; DCA confirmed the model’s positive net benefit across a wide range of threshold probabilities.

Conclusion. The results align with current understanding regarding treatment outcomes for tethered spinal cord syndrome and the applicability of machine learning approaches in pediatric neurosurgery. The XGB ensemble model demonstrates high accuracy and clinical utility in predicting adverse outcomes in children with tethered spinal cord syndrome, making it a promising tool for risk stratification; it could be integrated into decision-support systems to help select the safest and most effective surgical strategy for spinal dysraphism.

1. Liu M, Deng W, Lu YY, He YZ, Huang LY, Du H. Surgical treatment of tethered cord syndrome showed promising outcome in young children with short duration. Eur Rev Med Pharmacol Sci. 2023;27:1831–1836. DOI: https://doi.org/10.26355/eurrev_202303_31545

2. Хачатрян В.А., Сысоев К.В. Об актуальных проблемах патогенеза, диагностики и лечения синдрома фиксированного спинного мозга (аналитический обзор). Нейрохирургия и неврология детского возраста. 2014;3:76–78. [Khachatryan VA, Sysoev KV. Current issues of pathogenesis, diagnostics and treatment of the tethered spinal cord syndrome (analytical review). Pediatric Neurosurgery and Neurology. 2014;3:76–78]. EDN: https://elibrary.ru/SYOZVH

3. Рябых С.О., Горчаков С.А., Калашников А.А. Синдром фиксированного спинного мозга при spina bifida: клинико-лучевая характеристика и показания к оперативному вмешательству (систематический обзор литературы). Хирургия позвоночника. 2024;21(1):27–34. [Ryabykh SO, Gorchakov SA, Kalashnikov AA. Tethered spinal cord syndrome associated with spina bifida: clinical and radiological characteristics and indications for surgery (systematic review of the literature). Russian Journal of Spine Surgery (Khirurgiya Pozvonochnika). 2024;21(1):27–34]. DOI: https://doi.org/10.14531/ss2024.1.27-34 EDN: https://elibrary.ru/HHGVIT

4. Сысоев К.В., Тадевосян А.Р., Назинкина Ю.В., Хачатрян В.А. Результаты хирургического лечения детей с синдромом фиксированного спинного мозга. Прогноз на основании данных спинальной 3Тл МРТ-трактографии. Вопросы нейрохирургии» имени Н.Н. Бурденко. 2016;80(3):66–73. [Sysoev KV, Tadevosyan AR, Nazinkina YuV, Khachatryan VA. Surgical treatment outcomes in children with tethered spinal cord syndrome. A prognosis on the basis of spinal 3T MRI tractography. Burdenko’s Journal of Neurosurgery. 2016;80(3)66–73]. DOI: https://doi.org/10.17116/neiro201680366-73 EDN: https://elibrary.ru/WDOXAH

5. Hsieh P, Apaydin E, Briggs RG, Al-Amodi D, Aleman A, Dubel K, Sardano A, Saint-Val J, Sysawang K, Zhang D, Yagyu S, Motala A, Tolentino D, Hempel S. Diagnosis and treatment of tethered spinal cord: a systematic review. Pediatrics. 2024;154:e2024068270. DOI: https://doi.org/10.1542/peds.2024-068270

6. Annaperesov NV, Musa G, Lobankin PV, Kuznetsova AA, Livshits M, Nikolaevich UV, Levov AV, Kolcheva MA, Konfetova N, Maier A, Kim AV, Bupe MM, Kislyakov AN, Chmutin GE, Chaurasia B. Pediatric spinal lipomas: insights from a single institution’s experience in surgical management of 55 cases. J Pediatr Neurosci. 2024;19:97–102. DOI: https://doi.org/10.4103/jpn.jpn_44_24

7. Vickers AJ, van Calster B, Steyerberg EW. A simple, step-by-step guide to interpreting decision curve analysis. Diagn Progn Res. 2019;3:18. DOI: https://doi.org/10.1186/s41512-019-0064-7

8. Zhang Z, Rousson V, Lee WC, Ferdynus C, Chen M, Qian X, Guo Y. Decision curve analysis: a technical note. Ann Transl Med. 2018;6:308. DOI: https://doi.org/10.21037/atm.2018.07.02

9. Wang R, Zhang J, Shan B, He M, Xu J. XGBoost machine learning algorithm for prediction of outcome in aneurysmal subarachnoid hemorrhage. Neuropsychiatr Dis Treat. 2022;18:659–667. DOI: https://doi.org/10.2147/NDT.S349956

10. Yan Z, Wang J, Dong Q, Zhu L, Lin W, Jiang X. XGBoost algorithm and logistic regression to predict the postoperative 5-year outcome in patients with glioma. Ann Transl Med. 2022;10:860. DOI: https://doi.org/10.21037/atm-22-3384

11. Trapp B, de Andrade Lourenção Freddi T, de Oliveira Morais Hans M, Fonseca Teixeira Lemos Calixto I, Fujino E, Alves Rojas LC, Burlin S, Cerqueira Costa DM, Carrete Junior H, Abdala N, Tobaru Tibana LA, Takehara ET, Dalul Gomez G. A practical approach to diagnosis of spinal dysraphism. Radiographics. 2021;41:559–575. DOI: https://doi.org/10.1148/rg.2021200103

12. Van Calster B, Wynants L, Verbeek JFM, Verbakel JY, Christodoulou E, Vickers AJ, Roobol MJ, Steyerberg EW. Reporting and interpreting decision curve analysis: a guide for investigators. Eur Urol. 2018;74:796–804. DOI: https://doi.org/10.1016/j.eururo.2018.08.038

13. Findlay MC, Tenhoeve S, Terry SA, Iyer RR, Brockmeyer DL, Kelly MP, Kestle JRW, Gonda D, Ravindra VM. Disparities in indications and outcomes reporting for pediatric tethered cord surgery: The need for a standardized outcome assessment tool. Childs Nerv Syst. 2024;40:1111–1120. DOI: https://doi.org/10.1007/s00381-023-06246-y

14. Иванов С.В., Кенис В.М., Щедрина А.Ю., Онуфрийчук О.Н., Ходоровская А.М., Осипов И.Б., Сарычев С.А. Spina bifida: мультидисциплинарная проблема (обзор литературы). Российский вестник детской хирургии, анестезиологии и реаниматологии. 2021;11(2):201–213. [Ivanov SV, Kenis VM, Shchedrina AYu, Onufriichuk ON, Khodorovskaya AM, Osipov IB, Sarychev SA. Spina bifida: a multidisciplinary problem (a literature review). Russian Journal of Pediatric Surgery, Anesthesia and Intensive Care. 2021;11(2):201–213]. DOI: https://doi.org/10.17816/psaic958 EDN: https://elibrary.ru/TRMLIX

15. Warman A, Kalluri AL, Azad TD. Machine learning predictive models in neurosurgery: an appraisal based on the TRIPOD guidelines. Systematic review. Neurosurg Focus. 2023;54:E8. DOI: https://doi.org/10.3171/2023.3.FOCUS2386