A comparative analysis of 12 intraocular lens power formulas

To evaluate the accuracy of 12 intraocular lens (IOL) power calculations: Barrett Universal II, EVO, Haigis, Hill-RBF version 2.0, Hoffer Q, Holladay 1, Holladay 2, Kane, Olsen, SRK/T, Super Formula and T2. In this retrospective consecutive case series, cataract extraction and IOL implantation cases in Toronto, Canada, were recruited between 2017 and 2019. Refractive predictions were compared to the observed 1-month postoperative spherical equivalent to determine the refractive error for each formula cohort. Subgroup analysis stratified eyes into short (≤ 22.5 mm)-, intermediate (22.5 mm-25.5 mm)- and long (≥ 25.5 mm)-axial length (AL) cohorts. The primary outcome was the percentage of cases within ± 0.50D of refractive error. Overall, 764 cataract cases were analyzed. Formulas with the highest percentage of eyes within ± 0.50D of refractive error, in decreasing order, were: Kane (77.7%), Barrett Universal II (77.4%), EVO (76.6%), T2 (76.4%), Super (75.9%), Holladay 1 (75.4%), Hill-RBF 2.0 (74.7%), SRK/T (72.6%), Hoffer Q (72.5%), Haigis (71.7%), Olsen (67.4%) and Holladay 2 (67.3%). For short-AL eyes, the Holladay 1 formula was most accurate (n = 69, 78.3% within ± 0.50D), and for long-AL eyes, the Barrett Universal II formula was most accurate (n = 116, 76.7% within ± 0.50D). Kane, Barrett, EVO, T2 and Super formulas led to a significantly lower mean absolute error compared to the open-source calculations with optimized lens constants (p-value