Abstract
BACKGROUND AND OBJECTIVE: Proximal interphalangeal joint (PIPJ) fractures present significant therapeutic challenges in hand surgery. This systematic review evaluated the comparative efficacy of dynamic external fixation against traditional treatment modalities, integrating machine learning analysis to enhance outcome prediction and treatment selection. METHODS: We systematically reviewed 43 clinical studies published between January 2014 and January 2024, including 26 dynamic external fixations, 6 traditional internal fixations, and 11 static external fixations. Studies were included if they reported quantitative outcomes of PIPJ fracture treatment, had a minimum follow-up of 4 weeks, and included at least 20 patients. Case series with fewer than 5 patients and non-English publications without available translations were excluded. The analysis focused on four key outcomes: range of motion (ROM), recovery time, complication rates, and functional results. We developed a neural network model to predict treatment outcomes, achieving 89.7% accuracy (95% CI 87.3-92.1%). Methodological quality was assessed using the Newcastle-Ottawa Scale and Cochrane Risk of Bias tool. RESULTS: Dynamic external fixation demonstrated superior outcomes across multiple domains. ROM analysis revealed a median of 86.12° (range: 70°-95°) for dynamic fixation compared to 72.30° (range: 56°-88°) for traditional approaches (mean difference: 13.82°, 95% CI 10.24-17.40°). Dynamic fixation significantly reduced recovery duration (9.68 weeks vs. 20.47 weeks, p < 0.001). Complication profiles favored dynamic fixation, with pin tract infection rates of 2.4% versus 3.8% for traditional fixation. Functional assessment using the Ishida scoring system showed favorable outcomes in the dynamic fixation group, with a mean score of 85.3 points and 78% of cases achieving scores above 80 points. DISCUSSION: This comprehensive systematic review provides evidence supporting the efficacy of dynamic external fixation for PIPJ fracture treatment. The findings demonstrate improved functional outcomes, accelerated rehabilitation, and reduced complication rates. The integration of machine learning analysis shows promise for optimizing patient-specific treatment selection. Further validation through large-scale, multicenter randomized controlled trials with extended follow-up periods is warranted.