Predictive modeling of intraoperative lens touch during pars plana vitrectomy using ocular biometry geometric simulation study

초록

Purpose : To develop a two-dimensional geometric eye model that uses ocular biometry to predict potential intraoperative lens touch during pars plana vitrectomy (PPV). Methods : The following biometric data were collected from 2,223 eyes using the IOLMaster 700: axial length (AL), central corneal thickness (CCT), anterior chamber depth (ACD), lens thickness (LT), keratometry (K), white-to-white distance (WTW), and vitreous chamber depth (VCD). The model simulated instrument insertion through the pars plana at three locations: the equator, 1 mm posterior to the equator, and 2 mm posterior to the equator. Risk of intraoperative lens touch was calculated. Results : The predicted probability of lens touch was 46.4% for an approach at the equator, 12.7% at 1 mm posterior, and 1.7% at 2 mm posterior to the equator. According to ROC curve analysis, ACD + LT demonstrated the strongest predictive performance, with AUCs of 0.820 (95% CI: 0.803–0.837), 0.854 (95% CI: 0.832–0.876), and 0.886 (95% CI: 0.837–0.935) for predicting lens touch at three locations, respectively (P < 0.001 for all). The optimal cut-off value of ACD + LT for predicting intraoperative lens touch was 7.58 mm (82% sensitivity, 66% specificity), 7.79 mm (82%, 75%), and 7.87 mm (89%, 76%) at three locations, respectively. Conclusions : A two-dimensional geometric eye model using biometric data can differentiate eyes at risk of lens touch during PPV, with ACD + LT identified as the strongest predictor. These findings may provide a practical tool for preoperative planning and reduce the risk of lens touch during PPV. This abstract was presented at the 2026 ARVO Annual Meeting, held in Denver, CO, May 3-7, 2026.