Machine Learning-Based Prediction of Hypocalcaemia Following Total Thyroidectomy: A Systematic Review
Cureus · 08.11.2026
DOI: 10.7759/cureus.119319 ↗Kurz gefasst
Systematische Übersicht in der Zeitschrift Cureus, erschienen am 08.11.2026. Verfasst von Mohammed A Alhazmi und 11 weiteren Autor:innen. Die Arbeit nennt 13 Literaturangaben.
Abstract (Originalsprache)
Machine learning (ML) has increasingly been applied to predict hypocalcaemia following total thyroidectomy; however, its predictive performance and potential advantages over conventional statistical approaches remain unclear.This review evaluated studies developing or validating ML models for predicting postoperative hypocalcaemia or hypoparathyroidism after total or completion thyroidectomy.Six studies involving 20,845 patients from multiple countries were included.All studies focused on the prediction of early or transient hypocalcaemia, while none addressed permanent hypoparathyroidism.The performance of the best-performing ML models varied considerably, with area under the receiver operating characteristic curve (AUC) values ranging from 0.65 to 0.98.In studies directly comparing ML models with logistic regression using the same datasets, differences in discrimination were minimal, indicating no consistent advantage of ML over conventional methods.Models incorporating perioperative parathyroid hormone (PTH) measurements generally achieved the highest predictive performance, whereas those relying solely on clinical variables demonstrated more modest discrimination.Despite promising results in some studies, all models exhibited important methodological limitations, lacked external validation, and provided insufficient assessment of calibration.Overall, current evidence suggests that ML models can predict postoperative hypocalcaemia with variable accuracy but have not consistently outperformed conventional logistic regression.Predictive performance appears to depend largely on the inclusion of perioperative PTH measurements.Given the limitations of the available evidence, these models cannot yet be recommended for routine clinical use.Future research should focus on rigorous external validation, comprehensive calibration assessment, and evaluation in independent patient populations before clinical implementation.
Zitieren
APA
Alhazmi, M. A., Musawi, R. Y., Omer, R. A., Alruwaili, R. T., Oraybi, R. A., Nahari, Y. Y., Ibrahim, A. M., Alahmari, S. K., Althobaiti, A. A., Ahmed, T. M., Alqaed, M. H., & Badri, A. A. (2026). Machine Learning-Based Prediction of Hypocalcaemia Following Total Thyroidectomy: A Systematic Review. Cureus. https://doi.org/10.7759/cureus.119319
Vancouver
Alhazmi MA, Musawi RY, Omer RA, Alruwaili RT, Oraybi RA, Nahari YY, et al. Machine Learning-Based Prediction of Hypocalcaemia Following Total Thyroidectomy: A Systematic Review. Cureus. 2026. doi:10.7759/cureus.119319
BibTeX
@article{Alhazmi2026,
author = {Alhazmi, Mohammed A and Musawi, Raum Y and Omer, Rahaf A and Alruwaili, Reham T and Oraybi, Rayim A and Nahari, Yazan Y and Ibrahim, Amani M and Alahmari, Saud K and Althobaiti, Aryam A and Ahmed, Tahani M and Alqaed, May H and Badri, Amal A},
title = {Machine Learning-Based Prediction of Hypocalcaemia Following Total Thyroidectomy: A Systematic Review},
journal = {Cureus},
year = {2026},
doi = {10.7759/cureus.119319}
}
EUVIMED