[Comparison of the accuracy of intraocular lens power calculation formulas based on the new swept-source optical...

Objective: To compare the accuracy of 6 intraocular lens power calculation formulas based on the new swept-source optical coherence tomography biometry and to analyze the prediction error. Methods: Retrospective case series study. Clinical data were collected from 599 patients (599 eyes) who had underwent uncomplicated phacoemulsification and the IOLMaster 700 examination at the Eye Hospital of Wenzhou Medical University between November 2018 and November 2019. Among the patients, there were 208 males and 391 females with an age of (69±10) years. According to the axial length (AL), eyes were divided into the short AL group (≤22.5 mm, n=100), the normal AL group (>22.5 mm and<25.5 mm, n=375); and the long AL group (≥25.5 mm, n=124). Eyes were also grouped based on the mean keratometry (Km) as flat (≤42.00 D, n=47), normal (>42.00 D to<46.00 D, n=461), and steep (≥46.00 D, n=91), and by anterior chamber depth (ACD) as shallow (≤2.5 mm, n=71), normal (>2.5 mm to<3.5 mm, n=436), and deep (≥3.5 mm, n=92). The median absolute errors (MedAEs) of the Barrett Universal Ⅱ, Haigis, Hoffer Q, Holladay Ⅰ, Holladay Ⅱ, and SRK/T formulas in different AL, Km, and ACD groups were compared using the Friedman test. Results: The differences in MedAE among the 6 formulas of 599 patients (599 eyes) were statistically significant (χ²=120.549, P<0.001). The MedAE of the Barrett Universal Ⅱ formula was smallest (0.35 D), followed by the SRK/T formula (0.36 D). There was no significant difference between the MedAEs of the Barrett universal Ⅱ and Haigis, SRK/T formula (all P=1.000), but there were statistically significant differences among the other formulas (all P<0.01). In different AL groups, the MedAE of each formula was statistically different (χ²=38.307, 38.779, 112.997; all P<0.01).The Barrett Universal Ⅱ formula resulted in the lowest MedAE in the short AL group (0.40 D) and the long AL group (0.31 D). The MedAE of the SRK/T in the normal AL group was lowest (0.35 D). The 6 formulas showed significant differences in MedAE values in different Km groups (χ²=12.284, 90.924, 39.387; all P<0.05).The Haigis formula achieved the lowest MedAE in the flat Km group (0.26 D) and the steep Km group (0.34 D). The Barrett UniversalⅡ formula achieved the lowest MedAE in the normal Km group (0.33 D). The differences in MedAE values of the 6 formulas in different ACD groups were statistically significant (χ²=37.389, 57.643, 52.845; all P<0.01), and the MedAE values of the Barrett Universal Ⅱ in different ACD groups were smallest (0.46, 0.33, 0.31 D). Conclusions: The Barrett Universal Ⅱ formula perform the best over the entire AL range, followed by the Haigis and SRK/T formulas. The Barrett Universal Ⅱ formula result in the lowest prediction error in the short AL group, the long AL group, and all ACD groups. The Haigis formula may be more accurate when the Km was ≤42.00 D or ≥46.00 D. (Chin J Ophthalmol, 2021, 57: 502-511).