Age- and sex-specific knee alignment patterns in symptomatic Korean patients: A cross-sectional deep learning analysis of 16,000 knees

Abstract

Objective: To analyze age-and sex-specific associations in knee alignment using a large dataset of long-leg radiographs with deep learning-based automated measurements. Methods: This retrospective cross-sectional study included 8014 radiographs (16,028 knees) from symptomatic adults who underwent long-leg radiography at a tertiary hospital in South Korea (2010-2023). Alignment parameters, including mechanical tibiofemoral angle (MTFA), lateral distal femoral angle (LDFA), medial proximal tibial angle (MPTA), and joint line convergence angle (JLCA), were automatically extracted using a deep learning model. Linear regression assessed age-related associations stratified by sex and age group (< 50 years versus >= 50 years), adjusting for body mass index, Kellgren-Lawrence grade, and clinical indication. Results: After adjustment, age-related associations in knee alignment differed by sex and age group. MTFA increased with age, most prominently in women aged >= 50 years (right/left beta = 0.053/0.040 degrees per year, 95% CI 0.038-0.068/0.025-0.054, whereas the association was minimal in men aged >= 50 years (right/left beta = -0.006/0.001, 95% CI-0.026-0.014/-0.018-0.021). LDFA was positively associated with age among participants < 50 years, but showed no clear association thereafter. MPTA decreased with age in younger men and women, while remaining near zero or slightly positive in older groups. JLCA increased with age in participants >= 50 years, particularly in women. Conclusion: Knee alignment demonstrated nonlinear associations with age and clear sex-related differences. These findings underscore the need to consider age-and sex-specific alignment characteristics in clinical evaluation and research on osteoarthritis risk. (c) 2025 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.