Logistic regression withstands comparison with machine learning for thyroid nodule malignancy risk stratification under strict temporal validation
Logistische Regression hält dem Vergleich mit Machine Learning zur Risikostratifizierung von Schilddrüsenknotenmalignität unter strenger zeitlicher Validierung stand
PLOS Digital Health · 11.11.2026 · 5 (11) · e0001828
DOI: 10.1371/journal.pdig.0001828 ↗Auf einen Blick
In 861 Schilddrüsenektomien zeigte ein logistisches Regressionsmodell unter zeitlicher Validierung eine gute Leistung zur Malignitätsvorhersage. Es war den getesteten ML-Verfahren nicht überlegen unterlegen.
Fragestellung
Verbessern Machine-Learning-Algorithmen die Risikostratifizierung von Schilddrüsenknotenmalignität gegenüber logistischer Regression unter strenger zeitlicher Validierung?
Methode
Einzentrische retrospektive Kohortenstudie mit chronologischer Aufteilung; Training vor 2019, Test ab 2019. Verglichen wurden LR, Random Forest, XGBoost, Bagged CART und MLP.
Ergebnisse
LR: AUC 0,838 (95 %-KI 0,789–0,884). RF 0,843 (p = 0,76) und XGB 0,844 (p = 0,69) waren nicht verschieden von LR; Bagged CART 0,821, MLP 0,768 (p < 0,001).
Einschränkungen
Einzentrige chirurgische Kohorte; moderates Sample; mehrere Subgruppen waren unterpowert.
Schlussfolgerung der Autor:innen
Nach Ansicht der Autor:innen zeigte keiner der geprüften ML-Algorithmen eine bessere Leistung als das gut spezifizierte LR-Modell.
Kernaussagen
- LR: AUC 0,838 (95 %-KI 0,789–0,884)
- RF und XGB waren statistisch ähnlich zu LR
- Kein ML-Verfahren übertraf das LR-Modell
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Fachartikel in der Zeitschrift PLOS Digital Health, online erschienen am 11.11.2026. Verfasst von Juan Jesús Fernández Alba und 3 weiteren Autor:innen. Beteiligte Einrichtung: Research Group EM18 "Development and Innovation in Maternal-Fetal Medicine and Gynaecological Oncology", Instituto de Investigación e Innovación Biom…. Die Arbeit nennt 20 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Ultrasound risk-stratification systems (ACR TI-RADS, EU-TIRADS, ATA 2015) guide fine-needle aspiration decisions for thyroid nodules but do not integrate clinical or biochemical information. Multivariable logistic regression (LR) models incorporating these variables provide individualised malignancy probability estimates, and machine learning (ML) algorithms have been proposed as potential improvements, but rigorous comparative evidence under temporal validation is scarce. We conducted a single-centre retrospective cohort study of 861 thyroidectomies (2010–2023), split chronologically (training: before 2019, n = 432, 94 malignant; test: 2019 onwards, n = 429, 92 malignant). An LR model using the specification of a previously published and externally validated model — with coefficients re-estimated in the training cohort — was compared head-to-head with Random Forest (RF), XGBoost (XGB), Bagged CART, and a multilayer perceptron (MLP) trained on the same predictors. Performance was assessed by discrimination (ROC-AUC and precision-recall AUC), calibration (Brier score, calibration slope and intercept), and clinical utility (decision curve analysis), with operating characteristics reported at the pre-specified 9.55% probability threshold. LR achieved an AUC of 0.838 (95% CI 0.789–0.884), closely matching the published external validation of the original model (0.84). RF (AUC 0.843, p = 0.76) and XGB (AUC 0.844, p = 0.69) were statistically indistinguishable from LR; Bagged CART (0.821, p = 0.40) was non-significantly inferior; the MLP (0.768, p < 0.001) performed significantly worse. LR showed the highest precision-recall AUC (0.672) and good calibration without post-hoc adjustment (slope 1.12, 95% CI 0.91–1.43). LR, RF, XGB and Bagged CART discriminated significantly better than the ATA 2015 classification (AUC 0.750; DeLong p ≤ 0.002), whereas the MLP did not differ significantly from it (p = 0.45). In exploratory subgroup analyses by Bethesda category, nodule size, and age, no evidence of a consistent ML advantage was observed, although several subgroups were underpowered. SHAP importance rankings from RF and XGB showed broadly similar global orderings to the LR coefficients (ρ = 0.71 and 0.77). In this single-centre surgical cohort with a moderate sample size and routinely collected structured predictors, none of the evaluated ML algorithms demonstrated improved performance over a well-specified LR model.
Autor:innen und Institutionen
- Research Group EM18 "Development and Innovation in Maternal-Fetal Medicine and Gynaecological Oncology", Instituto de Investigación e Innovación Biomédica de Cádiz (INIBICA), Cádiz, Spain.
- Department of Obstetrics and Gynaecology, Hospital Universitario de Puerto Real, Cádiz, Spain.
- University of Cadiz, Cádiz, Spain.
- Department of Endocrinology and Nutrition, Hospital Universitario de Puerto Real, Cádiz, Spain.
ORCID-Profile: Juan Jesús Fernández Alba, Florentino Carral San Laureano, Ana Isabel Jiménez Millán, Carmen González Macías
Zitieren
APA
Fernández Alba, J. J., Carral San Laureano, F., Jiménez Millán, A. I., & González Macías, C. (2026). Logistic regression withstands comparison with machine learning for thyroid nodule malignancy risk stratification under strict temporal validation. PLOS Digital Health, 5(11), e0001828. https://doi.org/10.1371/journal.pdig.0001828
Vancouver
Fernández Alba JJ, Carral San Laureano F, Jiménez Millán AI, González Macías C. Logistic regression withstands comparison with machine learning for thyroid nodule malignancy risk stratification under strict temporal validation. PLOS Digital Health. 2026;5(11):e0001828. doi:10.1371/journal.pdig.0001828
BibTeX
@article{FernndezAlba2026,
author = {Fernández Alba, Juan Jesús and Carral San Laureano, Florentino and Jiménez Millán, Ana Isabel and González Macías, Carmen},
title = {Logistic regression withstands comparison with machine learning for thyroid nodule malignancy risk stratification under strict temporal validation},
journal = {PLOS Digital Health},
year = {2026},
volume = {5},
number = {11},
pages = {e0001828},
doi = {10.1371/journal.pdig.0001828}
}
EUVIMED