Accuracy of Kane, Hill-RBF, and PEARL-DGS versus traditional IOL formulas in highly myopic eyes: a systematic review and meta-analysis.

purposeTo evaluate the accuracy of Kane, Hill-RBF, and PEARL-DGS compared with traditional intraocular lens (IOL) formulas in highly myopic eyes undergoing cataract surgery.

methodsA systematic review registered in PROSPERO (CRD420261372807) and conducted in accordance with PRISMA 2020 evaluated studies comparing Kane, Hill-RBF, and/or PEARL-DGS with conventional IOL formulas in highly myopic eyes. PubMed, Embase, and the Cochrane Library were searched from inception to April 8, 2026 using reconstructed database-specific strategies combining concepts for high myopia or long axial length, cataract surgery and IOL power calculation, and the formula names Kane, Hill-RBF, and PEARL-DGS. Two independent reviewers performed screening, data extraction, and adapted PROBAST-style risk-of-bias assessment, with disagreements resolved by a third reviewer. Eligible studies included original human clinical studies reporting postoperative refractive prediction outcomes in highly myopic eyes, preferentially defined as axial length (AL) > = 26.0 mm, or in separately extractable high-myopia subgroups. Random-effects quantitative synthesis was performed for the most comparable threshold endpoint when exact percentages and denominators were available. Primary outcomes included the proportion of eyes within +/-0.50 D of prediction error and mean absolute error (MAE), when reported using sufficiently comparable definitions. Secondary outcomes included MedAE, ME, SD of prediction error, RMSAE, and the proportions of eyes within +/-0.25 D and +/-1.00 D.

resultsThe final included evidence set comprised 11 original comparative studies published online or in print from 2020 to 2026. Five studies were threshold-relevant in the narrative layer, 4 had numeric +/-0.50 D rows represented in the extraction seed, and 3 entered the pooled target-formula model. Across the included studies, Kane, Hill-RBF, and PEARL-DGS generally showed favorable refractive prediction performance in highly myopic eyes, but rankings varied by axial-length subgroup, formula version, biometry inputs, constant optimization, mean error zeroing, IOL design, and outcome family. Verified findings included a MedAE of 0.26 D for Kane in Cheng 2021, an MAE of 0.40 +/- 0.39 D with 71.44% within +/-0.50 D for Hill-RBF 2.0 in Chen 2021, and in Jiang 2025 a lowest overall MAE of 0.50 D for Hill-RBF 3.0 with the highest overall +/-0.50 D proportion of 69.57% for PEARL-DGS. For the exploratory cross-formula target benchmark, three independent studies contributed exact overall threshold rows, yielding a pooled proportion within +/-0.50 D of 70.90% (95% CI, 68.34% to 73.33%; I2 = 0.0%). A secondary Hill-RBF overall synthesis from two studies yielded 71.03% (95% CI, 68.35% to 73.58%; I2 = 0.0%). Because the main pooled model contains only 3 studies and combines architecture-distinct target formulas, I2 should be interpreted as a descriptive statistic with limited power rather than as evidence of biological or algorithmic homogeneity.

conclusionsCurrent evidence supports Kane, Hill-RBF, and PEARL-DGS as reasonable front-line options for highly myopic eyes, and the threshold synthesis provides a cautious benchmark for currently extractable overall target-formula threshold performance. However, the evidence does not establish a single formula as uniformly best across all long-eye settings, nor does the 3-study pooled model constitute a formula ranking. Interpretation should remain anchored to threshold-based accuracy and to study-level differences in population definition, formula implementation, IOL design, and analytical handling. REGISTRATION: This review was registered in PROSPERO under CRD420261372807.