Associations between age and ocular biometric parameters in individuals with short axial length: a large-scale cross-sectional study

Short axial length is an important ocular biometric characteristic associated with distinct anterior segment anatomy. However, large-scale evidence describing age-associated patterns of ocular biometric parameters in individuals with short axial length remains limited, and potential sex-specific patterns have not been fully characterized. This study aimed to investigate age-associated patterns and potential sex-specific variation in ocular biometric parameters in a large population of individuals with short axial length. This retrospective cross-sectional study included biometric data from patients who underwent ocular biometry at Shenzhen Eye Hospital between September 2021 and February 2024. Individuals aged ≥ 45 years with AL < 23 mm were included, while those with a history of refractive surgery, lens extraction, or vitrectomy were excluded. A total of 20,444 eyes were analyzed. Ocular biometric parameters included axial length (AL), aqueous depth (AQD), lens thickness (LT), and white-to-white corneal diameter (W2W). Segmented regression analysis was used to examine cross-sectional associations between age and each biometric parameter, and model fit was assessed using the likelihood ratio test. Sex-stratified analyses were performed to characterize sex-specific patterns. Generalized estimating equation (GEE) models adjusted for sex were used in sensitivity analyses to account for potential correlations between fellow eyes from the same participant. Cluster-adjusted piecewise GEE models were additionally used to assess the robustness of the piecewise association patterns, and formal sex interaction analyses were performed using age × sex and hinge × sex interaction terms. Statistical analyses were conducted using R software version 4.2.3. Segmented regression analysis identified nonlinear cross-sectional associations between age and all ocular biometric parameters. Estimated breakpoints were identified at 61.2 years for AL, 63.0 years for AQD, 61.0 years for LT, and 60.1 years for W2W. Segmented regression models provided a significantly better fit than the corresponding linear models for all four parameters (all P < 0.001). Sex-stratified analyses showed different estimated breakpoint patterns between males and females. Earlier estimated breakpoints in the associations between age and AL and LT were observed in females (56.8 and 60.6 years, respectively) than in males (63.0 and 65.3 years, respectively). However, formal interaction analyses provided evidence of sex modification only for the piecewise age–AL association (age × sex, P = 0.001; hinge × sex, P = 0.005), whereas no significant sex interactions were observed for AQD, LT, or W2W. GEE sensitivity analyses accounting for inter-eye correlation supported the overall and piecewise age-associated patterns. This study identified nonlinear cross-sectional associations between age and ocular biometric parameters among individuals with short axial length, with estimated changes in association slopes occurring between approximately 60 and 63 years. Sex-stratified analyses showed different estimated breakpoint patterns, while formal interaction analyses supported sex modification of the piecewise age–AL association only. These estimated breakpoints represent statistical features of cross-sectional associations and should not be interpreted as within-individual anatomical transitions with aging.