Artificial intelligence-based prognostic modeling in temporomandibular disorders and chronic orofacial pain: a critical conceptual review.

Abstract

This conceptual review (ⅰ) analyzes the outcomes predicted, data modalities, modeling approaches, validation strategies, and reporting quality of existing artificial intelligence (AI)-driven prognostic models in temporomandibular disorders (TMD) and chronic orofacial pain (OFP); (ⅱ) identifies enduring methodological and ethical constraints that hinder clinical translation; and (ⅲ) proposes a pragmatic research framework to guide the responsible development of clinically relevant prognostic tools for TMD and OFP. The review covers peer-reviewed and other relevant publications from the previous decade, emphasizing AI or machine-learning (ML) based models for prognosis, outcome prediction, or trajectory modeling in TMD and OFP populations. Established paradigms, including the Transparent Reporting of a multivariable prediction model for individual Prognosis Or Diagnosis plus Artificial Intelligence extension (TRIPOD + AI) and the Prediction model Risk Of Bias Assessment Tool (PROBAST), were used to assess the literature. Methodologies remain highly inconsistent, and current literature lacks the volume and rigor required for clinical translation. Most AI research has concentrated on diagnostic classification rather than prognostic modeling. Small sample sizes, short follow-up, single-center datasets, omission of psychosocial factors, and a general lack of external validation hamper the few prognostic studies that exist. Most model outputs are neither clinically actionable nor suitable for direct use in treatment decisions, limiting their value for clinicians and their potential impact on patient outcomes. Research applying AI to forecast TMD and OFP remains in its early stages. Without a prognosis-first research design, longitudinal data integration, inclusion of biopsychosocial predictors, and clinically significant outcome objectives, existing models are unlikely to influence clinical practice. Clear research objectives, reporting criteria, and ethical norms must be established before AI-based prognostic models can be confidently adopted in TMD and OFP clinical practice. Future objectives comprise establishing multicenter longitudinal cohorts, conducting trajectory-based modeling, employing federated learning for external validation, and initiating prospective clinical trials to demonstrate clear clinical benefit.