Stopyra Wiktor, Voytsekhivskyy Oleksiy, Grzybowski Andrzej
MW-med Eye Center, Krakow 31-416, Poland; University of Applied Sciences, Nowy Targ 34-400, Poland.
Kyiv Clinical Ophthalmology Hospital Eye Microsurgery Center, Kyiv 03680, Ukraine.
Asia Pac J Ophthalmol (Phila). 2025 May-Jun;14(3):100198. doi: 10.1016/j.apjo.2025.100198. Epub 2025 Apr 10.
To compare the accuracy of twenty intraocular lens (IOL) power calculation formulas in long eyes (26.00-29.99 mm) DESIGN: Retrospective accuracy and validity analysis METHODS: The data of patients who underwent uneventful phacoemulsification between August 2019 and July 2024 were reviewed. Pre-surgery IOL power was calculated using Haigis, Holladay 2, SRK/T and Barrett Universal II formulas. Three months after cataract surgery, refraction was measured. Post-surgery IOL power calculations were performed utilizing the following formulas: Castrop, EVO 2.0, Hill-RBF 3.0, Hoffer Q, Holladay 1, Hoffer QST, K6, Kane, Karmona, Ladas Super Formula AI (LSF AI), Naeser 2, Olsen (standalone), PEARL-DGS, T2, VRF and VRF-G. Root mean square absolute error (RMSAE), median absolute error (MedAE) and percentage of eyes with prediction error (PE) within ± 0.25 D, ± 0.50 D, ± 0.75 D and ± 1.00 were calculated.
One hundred fifty-three eyes with axial length ranging between 26.00 mm and 29.47 mm, were studied. The SRK/T formula yielded the lowest RMSAE (0.349) with statistical superiority over Holladay 1 (0.519, P = 0.000), Olsen (standalone) (0.505, P = 0.000) and Hoffer Q (0.497, P = 0.000). In terms of MedAE, the lowest outcome was obtained by VRF-G (0.256) followed by Barrett Universal II, Hill-RBF 3.0 and LSF AI (all equally of 0.260). The highest percentage of eyes with prediction error within ± 0.50 D was achieved by SRK/T (84.97 %) statistically significant vs Holladay 1 (60.78 %), Hoffer Q (64.05 %) and Olsen(standalone) (68.63 %).
SRK/T has delivered highly accurate results in long eyes. Barrett Universal II, Hill-RBF 3.0, VRF-G and LSF AI are also very true.
比较20种人工晶状体（IOL）屈光度计算公式在长眼（眼轴长度26.00 - 29.99毫米）中的准确性。
回顾性准确性和有效性分析。
回顾2019年8月至2024年7月期间接受顺利白内障超声乳化手术患者的数据。术前使用Haigis、Holladay 2、SRK/T和Barrett通用II公式计算IOL屈光度。白内障手术后三个月测量屈光度数。术后IOL屈光度计算采用以下公式：Castrop、EVO 2.0、Hill - RBF 3.0、Hoffer Q、Holladay 1、Hoffer QST、K6、Kane、Karmona、Ladas超级公式AI（LSF AI）、Naeser 2、Olsen（独立）、PEARL - DGS、T2、VRF和VRF - G。计算均方根绝对误差（RMSAE）、中位数绝对误差（MedAE）以及预测误差（PE）在±0.25 D、±0.50 D、±0.75 D和±1.00 D范围内的眼的百分比。
研究了153只眼轴长度在26.00毫米至29.47毫米之间的眼睛。SRK/T公式产生的RMSAE最低（0.349），在统计学上优于Holladay 1（0.519，P = 0.000）、Olsen（独立）（0.505，P = 0.000）和Hoffer Q（0.497，P = 0.000）。就MedAE而言，VRF - G获得最低结果（0.256），其次是Barrett通用II、Hill - RBF 3.0和LSF AI（均为0.260）。SRK/T在±0.50 D范围内预测误差的眼的百分比最高（84.97%），在统计学上显著高于Holladay 1（60.78%）、Hoffer Q（64.05%）和Olsen（独立）（68.63%）。
SRK/T公式在长眼中给出了高度准确的结果。Barrett通用II、Hill - RBF 3.0、VRF - G和LSF AI也非常准确。
