Efficacy of Extremity Nonunion Prediction: State of the Art

Abstract

Background: Nonunion remains a major orthopedic complication, occurring in 5–10% of fractures and drastically diminishing patients' quality of life. Traditional guidelines necessitate waiting up to nine months for a diagnosis, leading to therapeutic delays, prolonged disability, and substantial healthcare costs. This underscores the urgent need for early risk stratification.

Aim: This study aims to summarize and critically evaluate the efficacy of currently available and innovative tools for the early prediction of extremity nonunion.

Material and methods: A state-of-the-art narrative review was conducted. The analysis encompassed studies evaluating biological bone turnover markers, advanced imaging diagnostics, clinical scoring systems, machine learning (ML) algorithms, and smart implant technologies.

Results: The isolated use of single predictive tools has significant limitations. Systemic markers reflect early fracture biology but exhibit excessive inter-individual variability. Clinical heuristics effectively exclude complications (NPV up to 100%) but inadequately identify high-risk patients (PPV ~35.5%). While ML models provide superior analytics, their current PPV of ~55% indicates a continued challenge in accurately identifying patients for early intervention. Monitoring load transfer via smart implants represents a promising parallel alternative.

Conclusions: Predicting nonunion based on single modalities is insufficient. Effective prediction requires integrating multimodal diagnostics, combining comprehensive patient information with AI-driven analytics. This approach necessitates prospective, centralized databases for optimal and safe algorithm training.