[Meta-analysis of the accuracy of different intraocular lens calculation formulas in cataract patients with short axial length].

Abstract

Objective: To systematically evaluate the predictive accuracy of different intraocular lens (IOL) power calculation formulas in cataract patients with short axial length (≤22.5 mm). Methods: This is a meta-analysis. Databases including PubMed, Web of Science, Cochrane Library, Embase, and CNKI (from the establishment of the database to August 2023) were searched, and case-control studies comparing≥2 IOL formulas were included. The literature involves 12 intraocular lens (IOL) power calculation formulas, including Barrett Universal Ⅱ, EVO, Haigis, Hill RBF, Hoffer Q, Holladay 1, Kane, Olsen, PEARL DGS, SRK/T, T2, and VRF. The primary outcome measures were cases with prediction errors within the ranges of±0.25 D,±0.50 D, and ±1.00 D.Two researchers independently screened the literature, extracted data, and evaluated the risk of bias (using the QUADAS-2 tool). Traditional meta-analysis (fixed/random-effect model) was performed using RevMan 5.3, and network meta-analysis was carried out using Stata 14. The proportion of predictive errors within the ranges of ±0.50 D and ±1.00 D were the main indicators. The odds ratio (OR) and its 95% confidence interval (CI) were used to compare and evaluate the predictive error situations of different formulas. Results: A total of 8 studies (785 eyes) were included. Traditional meta-analysis showed that within the ±1.00 D range, Hoffer Q (93.4%; 648/694) and Kane (95.7%; 426/445) were significantly better than SRK/T (OR=1.50, 95%CI: 1.01-2.25, P=0.020; OR=2.00, 1.73-3.43, P=0.007). Kane was signifficantly better than T2 (OR=2.09, 1.15-3.80, P=0.020); within the ±0.50 D range, PEARL-DGS was better than SRK/T (OR=1.34, 1.00-1.80, P=0.050); within the ±0.25 D range, EVO (46.1%; 94/204) had the highest proportion and was significantly better than Hoffer Q (OR=0.62, 0.41-0.91, P=0.020). Network meta-analysis (ranked by the area under the cumulative ranking curve) indicated that EVO (ranked first within ±0.50 D) and Kane (ranked first within ±1.00 D) had the best overall performance, and SRK/T was the worst in all ranges (P<0.05). Conclusion: Among the traditional optical refraction formulas, Haigis and Hoffer Q have good accuracy, and SRK/T is not recommended; the new-generation formulas (EVO, Kane) have better prediction accuracy in patients with short axial length and are recommended for clinical priority selection.