Answers are grounded in this paper and cite their sources.
INTRODUCTION: To explore the potential for improving intraocular lens (IOL) power calculation accuracy, we performed stratified IOL constant optimization based on ocular biometric parameters and compared refractive prediction accuracy with overall optimization. METHODS: This retrospective study included 271 eyes of 271 patients who underwent cataract surgery between July 2022 and October 2025. Ocular biometric parameters were measured using the OA-2000 (Tomey). Eyes were stratified by axial length (AL), keratometry (K), anterior chamber depth (ACD), and lens thickness (LT), with 40 eyes selected at each extreme for subgroup-specific IOL constant optimization and exploration of trends in refractive prediction accuracy. The lens factor (LF) for Barrett Universal II (B-UII) and A constant for SRK/T were optimized for the overall cohort and within each stratum. For B-UII, prediction accuracy using the LF converted from the optimized SRK/T A constant was also evaluated. RESULTS: Stratified optimization tended to improve prediction accuracy in eyes with long AL for B-UII and in eyes with long AL, deep ACD, and flat K for SRK/T. Prediction accuracy was reduced in eyes with steep K for both formulas, with little change after stratified optimization. The LF (2.06) converted from the optimized SRK/T A constant (119.34) differed from the overall LF (2.16) optimized specifically for B-UII. When the converted LF was applied to B-UII, the mean prediction error was -0.13 D, and numerically greater mean absolute errors were observed than with the overall LF across all biometric subgroups. CONCLUSIONS: For B-UII, overall optimization of the LF appeared more useful than additional stratified optimization. Although the effect of stratified optimization on absolute prediction error was limited, the directional trends in refractive prediction identified across biometric subgroups may assist in choosing between adjacent IOL powers.
No takes yet. Share an insight, caveat, or question.
Miyamoto et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: