Abstract
OBJECTIVES: To develop regression formulas for determining optimal Implantable Collamer Lens (ICL) size based on Anterion AS-OCT (anterior segment optical coherence tomography) biometric data. METHODS: In this retrospective interventional case series, 89 patients were included in the development phase and 57 patients were included in the (internal) validation phase. The study developed the SN (Salouti-Nowroozzadeh) formula, a multiple step-wise linear regression model, incorporating anterior chamber width (ACW), lens thickness, and anterior chamber volume (ACV) as predictors, to determine optimal ICL size (R-square=0.602). Another formula predicted ICL vault using a cubic non-linear model and the difference (Delta) between implanted and predicted ICL size as an explanatory variable (R-square = 0.599). Logistic regression determined the probability of achieving optimal vault or acceptable vault post-surgery. We also present findings from 73 consecutive prospective cases from a separate dataset, where the ICL was determined using the SN formula (external validation). RESULTS: Delta was the sole independent factor predicting postoperative ICL vault. A ∆ value of 0.3 mm corresponded to a 65% probability of achieving optimal vault, while a ∆ value of 0.15 resulted in an 80% probability. The SN formula matched with the STAAR nomogram in 69.9% of cases (79/113). Compared to the STAAR formula, the SN formula recommended a larger size in 10 eyes (8.8%) and a smaller size in 24 (21.2%). The 95% LoA for SN-predicted and actual ICL vault were (-339 to 518) µm. The mean absolute error was 191 µm (SD,139), and 72 eyes (63.7%) had a difference of ≤200 µm. The external validation analysis confirmed the primary findings. CONCLUSIONS: The developed regression formula provided acceptable predictions for proper ICL sizing in our patients. However, its performance may vary across different populations or measurement devices. This study also highlights the need for smaller increments of ICL sizes to improve postoperative outcomes with any given formula.