One Size Doesn't Fit All Ages: A Comparative Study On The Influence Of Age-Specific Ocular Anatomical Differences On The Prediction Of Post-Icl Surgery Vault

To compare ICL vault prediction accuracy between a full-age model and a young cohort model; to quantify age-related anatomical differences in key posterior chamber parameters; and to identify anatomical predictors of dynamic vault change (ΔVault), thereby informing the need for age-specific prediction algorithms.

Age-related ocular changes may affect ICL vault prediction, yet current models are developed primarily in young cohorts. Clinically, patients aged ≥40 often exhibit unexpected vault outcomes, but the underlying anatomical drivers remain unquantified. This study evaluates how age-specific anatomical differences influence postoperative vault prediction and identifies key predictors of vault change (ΔVault).

This retrospective study analyzed 2107 eyes undergoing ICL V4c implantation. Preoperative UBM parameters included: STS-H/V, lens thickness (LT), anterior chamber depth (ACD), ciliary process length (CPL), and maximum ciliary body thickness (CBTmax). Machine learning algorithms (LightGBM, XGBoost, SVR, RF, GBDT) with 5-fold cross-validation constructed two models: full-age (18-50y, n=2107) and young (<40y, n=1788). Additionally, 426 eyes were stratified into three age groups: Young (18-30y, n=188), Middle (31-39y, n=134), Older (≥40y, n=104). ΔVault (6 months minus 1 day postop) was calculated as a dynamic stability metric. Model performance was evaluated using RMSE and R². Anatomical parameters and ΔVault were compared across groups.

The young cohort model significantly outperformed the full-age model (RMSE: 53.59 vs. 64.36 μm; R²: 0.896 vs. 0.881; P<0.01), confirming age-related prediction decay. Anatomically, Older eyes exhibited progressive LT increase (3.921±0.247 vs. 3.513±0.196 mm in Young, P<0.001) and ACD decrease (3.044±0.232 vs. 3.266±0.288 mm, P<0.001). Dynamic vault stability declined with age: ΔVault variability was highest in Older (SD=122.55 μm) vs. Young (67.62 μm). Feature importance identified CPL, CBTmax, and LT as top ΔVault predictors.

Including patients ≥40 years degrades universal vault prediction model performance. Older eyes exhibit distinct anatomy (increased LT, altered ciliary parameters) and poorer vault stability. These findings support developing age-specific prediction models incorporating ciliary body metrics. Clinically, preoperative attention to LT and ciliary process length in patients ≥40 may improve vault outcomes.