Abstract / Summary
To develop and internally validate a risk prediction model for postoperative vault abnormalities following implantable collamer lens (ICL) surgery. Eyes with an observed 1‑month vault measurement were categorized into normal and abnormal vault groups based on a vault ≤ 250 or ≥ 750 μm. Penalized logistic regression framework under the EPV constraint was performed to identify potential risk factors. A patient-level random-intercept mixed-effects logistic regression model was then developed, and its stability was assessed through bootstrap resampling. The model’s performance was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC). Among 625 eyes from 313 patients undergoing ICL surgery, 410 eyes had 1-month vault data and were included in model development; 65 had abnormal vault. Vault declined most rapidly during the first postoperative week and subsequently stabilized, supporting the 1-month endpoint. The protected-HSTSL LASSO model retained six predictors: mesopic pupil size, anterior chamber angle, angle-to-angle distance, horizontal STSL lens rise, ICL size, and vertical ICL orientation. Apparent discrimination was moderate (AUC = 0.701), but optimism-adjusted and nested cross-validated AUCs were lower (0.594 and 0.554), indicating limited generalizability. Sensitivity analyses showed consistent apparent performance across analytic approaches and follow-up subgroups. Preoperative parameters, particularly horizontal STSL lens rise, ACA, ATA, and ICL size, predict 1-month vault abnormalities, although discrimination was modest after optimism correction. This nomogram may support preoperative risk assessment and lens selection as an adjunct to clinical judgement, rather than a stand-alone tool. External validation and recalibration are required before clinical application. Not applicable.