Photo Credit: Ogtay Mammadov

The following is a summary of “Intraocular Lens Power Calculation – Comparing Big Data Approaches to Established Formulas,” published in the February 2025 issue of American Journal of Ophthalmology by Redden et al. 

Researchers conducted a retrospective study to compare the predictive performance of traditional intraocular lens (IOL) power calculation formulas (e.g., SRK/T, Haigis, Hoffer Q, Holladay I) with advanced regression models, including classical linear models, regression splines, and random forest regression, in estimating postoperative refraction after cataract surgery.  

They used optical biometry (IOLMaster 700) to obtain biometric measurements and assessed postoperative refraction at least 4 weeks after surgery. Formula constants for 5 IOL formulas (SRK/T, Haigis, Hoffer Q, Holladay I, and Castrop V1) were optimized using root mean squared error (RMSE). Regression models, including classical linear models, regression splines, and random forest regression, were trained on 4 datasets categorized by axial length (AL) as normal, short, long, and random. Model performance was assessed by mean absolute error (MAE), RMSE, and prediction error variance for both in-sample and out-of-sample predictions.  

The results showed the regression models had lower in-sample prediction errors than traditional IOL formulas. Linear regression models had similar out-of-sample prediction errors to traditional formulas. The lowest out-of-sample prediction error (MAE = 0.279, RMSE = 0.359) was achieved using a model where certain covariates (R2, AL, CCT) were modeled nonlinearly with regression splines, outperforming all traditional formulas. The Castrop formula had the lowest errors among traditional formulas (MAE = 0.284, RMSE = 0.359). Random forest regression had strong in-sample performance but showed poor out-of-sample generalizability due to overfitting.  

Investigators concluded the regression models integrating nonlinear effects, such as regression splines, offered an alternative to traditional IOL formulas for predicting postoperative refraction, though the clinical utility remained limited by out-of-sample performance. 

Source: ajo.com/article/S0002-9394(25)00071-6/fulltext