What does this research mean for the field?

Machine learning models applied to simulated prosthodontic data can identify clinically plausible risk factors for prosthetic complications, though their overall predictive accuracy remains modest due to the multifactorial etiology of prosthetic failure. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

The aim is to evaluate how occlusion, bruxism, and materials impact prosthetic complications using simulation and machine learning.

March 27, 2026Open Access

Data-Driven Modeling of Prosthetic Complications: A Literature Based Simulation and Machine Learning Approach in Prosthodontics

Key Points

The aim is to evaluate how occlusion, bruxism, and materials impact prosthetic complications using simulation and machine learning.
Conducted a retrospective computational analysis with a synthetic dataset of 1,000 patients.
Incorporated demographic, behavioral, and prosthetic parameters based on clinical literature.
Generated complication outcomes using a logistic risk model informed by literature.
Trained five supervised machine learning algorithms to predict complications.
Evaluated model performance through various metrics such as accuracy and ROC AUC.
Identified key factors like sleep bruxism and PMMA material linked to higher complication risks.
Logistic Regression provided the highest predictive performance with a ROC AUC of 0.625.
Overall, model performances were modest, indicating the complexity of prosthetic failure.
SVM failed to detect complications likely due to dataset imbalances.
Findings support the potential of using machine learning in clinical decision support.

Abstract

Objective:To assess how occlusion type, bruxism, and material characteristics influence prosthetic complications using a simulated dataset and machine learning models.Methods:A retrospective computational analysis was performed using a synthetic dataset of 1,000 simulated patients. Variables were based on clinical prevalence and prosthodontic literature, incorporating demographic, behavioral, and prosthetic parameters. Complication outcomes were generated using a literature-informed logistic risk model. Five supervised machine learning algorithms—Logistic Regression, Random Forest, Gradient Boosting, Support Vector Machine (SVM), and AdaBoost—were trained to predict prosthetic complications. Performance was evaluated using accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (ROC AUC).Results:Key factors associated with increased complication risk included sleep bruxism, implant-opposing dentition, use of PMMA material, and a greater number of prosthetic units. Logistic Regression yielded the highest overall predictive performance (ROC AUC = 0.625), though performance across models remained modest. The SVM model failed to detect complications, likely due to the imbalanced nature of the dataset. The limited predictive power of all models underscores the multifactorial etiology and complexity of prosthetic failure.Conclusion:This study demonstrates that machine learning, when applied to a realistically simulated dataset, can identify clinically plausible risk patterns for prosthetic complications. While predictive accuracy remains moderate, this approach offers a foundation for developing future clinical decision support tools. Incorporating synthetic modeling and machine learning may enhance risk stratification and personalized treatment planning in prosthodontics.Keywords: Bruxism; Occlusion; Prosthetic Complications; Machine Learning; Simulated Data; Prosthodontics; Risk Modeling

Perguntar à IA

Bookmark

View Full Paper