Methodological Quality of Prognostic Models for Periodontitis: A Systematic Review and Critical Appraisal

Area of Science:

• Periodontology
• Biostatistics
• Health Informatics

Background:

• Prognostic models are crucial for predicting periodontitis progression.
• Understanding their methodological quality and evolution is essential for clinical application.
• Previous appraisals have not comprehensively covered the historical development and quality of these models.

Purpose of the Study:

• To critically appraise the methodological quality of prognostic models for periodontitis.
• To analyze the historical evolution of these prognostic models.
• To compare the risk of bias across different types of prognostic models.

Main Methods:

• Systematic search across three databases.
• Classification of eligible studies into conceptual tools, category-based tools, and data-driven models.
• Evaluation of category-based and data-driven models using the PROBAST tool, assessing risk of bias across Participants, Predictors, Outcome, and Analysis domains.

Main Results:

• Twenty-seven studies were included, showing an evolution from qualitative categories to quantitative, individualized probability estimates.
• Conceptual tools were not evaluable by PROBAST; category-based tools had limitations in predictor handling and validation.
• Data-driven models showed refinement over time, with recent models reporting more validation metrics, but all studies except one had high risk of bias.
• Strengths were noted in the Participants domain, but persistent limitations existed in the Analysis domain, with modest improvements post-TRIPOD guidelines.

Conclusions:

• Prognostic models for periodontitis exhibit diverse methodological profiles based on design type.
• Conceptual and category-based frameworks hold historical clinical relevance.
• Data-driven models represent the current standard but still possess significant methodological gaps requiring attention.