IOL Calculation Formulas: What Should the Ophthalmologist Expect？

Modern cataract surgery has progressively evolved into a precise refractive procedure. To achieve optimal surgical outcomes, the preoperative calculation of intraocular lens (IOL) power is particularly crucial. The calculation of IOL power has a long history, having transitioned from computations based on linear regression formulas to the establishment of theoretical models, and further advancing to complex prediction systems integrated with artificial intelligence. Given the multitude of available formulas, how should clinicians make their selection? This article will systematically outline the technical characteristics of existing mainstream formulas and, combined with clinical data, provide a personalized selection strategy based on axial length.

Evolution of IOL power calculation formulas

First- and second-generation formulas: SRK I and SRK II (1, 2) were primarily based on simple regression analysis, establishing a linear relationship between axial length (AL), keratometry (K), and IOL power. Their simple structure resulted in poor predictive performance in eyes with extreme axial lengths.

Third-generation formulas: SRK/T, Hoffer Q, and Holladay (3, 4, 5) introduced theoretical models building upon linear regression. They calculate the effective lens position (ELP) using optical principles and incorporate AL and K values as predictive variables, significantly improving accuracy for eyes within the normal axial length range.

Fourth-generation formulas: Haigis and Holladay 2 (6, 7) further integrated additional preoperative biometric parameters, including anterior chamber depth (ACD), lens thickness (LT), and white-to-white distance (WTW), to optimize ELP prediction in a multivariable manner, enhancing calculation stability for atypical eyes.

New-generation formulas: Barrett Universal II, PEARL-DGS, and Hill-RBF (8, 9, 10, 11) are based on extensive databases of postoperative outcomes. They utilize machine learning algorithms or precise physical ray-tracing models, capable of identifying and learning complex, non-linear relationships between variables that traditional formulas cannot capture. Consequently, they demonstrate exceptional accuracy across various eye types, especially in complex cases. Seven of the newer formulas can be found on the ESCRS site (https://iolcalculator.escrs.org/).

Table 1 summarizes existing IOL power calculation formulas.

WEBUPLOAD INSERT TABLE 1 

IOL formula selection strategy based on axial length

Synthesizing analyses from numerous studies on modern IOL formulas, results indicate that a single formula cannot meet the diverse needs of eyes with complex anatomical variations. Therefore, modern precision refractive cataract surgery necessitates personalized selection. Below, we provide recommendations for IOL formula choice based on different axial length ranges:

Short axial length (AL < 22.0 mm)

For patients within the short axial length range, postoperative myopic shift is a common occurrence. Studies indicate that among traditional formulas, SRK/T demonstrates significantly lower accuracy in this range (MAE = 0.75 D) compared to the Hoffer Q formula, which is better suited for short eyes (MAE = 0.46 D) (12). In a recent study, ZEISS AI outperformed Barrett, Pearl-DGS, and Kane, and the K6 formula outperformed some formulas in selected parameters (13). AI-based formulas, such as PEARL-DGS and Hill-RBF, have also demonstrated exceptional accuracy in the short axial length subgroup, with MAEs as low as approximately 0.30–0.40 D, significantly outperforming several traditional formulas (14,15).

Normal axial length (22.0 mm ≤ AL ≤ 26.0 mm)

Most modern formulas achieve good results within this range. The SRK/T formula, due to its relatively lower parameter requirement and compatibility with a wide array of devices, remains one of the most accessible options. Its accuracy in normal eyes is considered good (MAE about 0.35–0.40 D) (12), though a performance gap persists compared to the top-performing formulas. More recently introduced formulas, such as Barrett Universal II and the AI-optimized Kane formula, can reduce the MAE to approximately 0.25–0.30 D, while also improving the proportion of eyes achieving a postoperative refractive error within ±0.50 D (16, 17). Therefore, in clinical practice, Kane or Barrett Universal II can be prioritized, as their predictions demonstrate high consistency.

Long axial length (AL > 26.0 mm)

For patients with long eyes, axial length measurement errors often occur due to poor fixation or posterior staphyloma, and there is a higher risk of postoperative hyperopic shift. The classic SRK/T formula performs relatively well in longer eyes (12). The Wang-Koch optimized version of the SRK/T formula can further mitigate the error associated with hyperopic shift as axial length increases, demonstrating a statistically significant reduction in MAE (from 0.47 D to 0.31 D) (18). Among newer-generation formulas, studies have validated that the Kane formula and the EVO 2.0 formula offer higher predictive accuracy and stability in patients with long axial lengths (16, 19).

Extremely long axial length (AL > 30.0 mm)

This represents a special minority case. The Wang-Koch optimized SRK/T formula remains a reliable choice. Simultaneously, the latest generation of AI-based formulas or the Barrett Universal II formula should be actively explored, and segmented axial length measurement should be strictly adopted to improve AL measurement precision. The recently introduced Zhu-Lu formula (20), specifically designed for highly myopic eyes, has shown higher predictive accuracy for IOL power compared to other formulas in initial studies, but its results require further validation.

Any new IOL power calculation formula is developed based on the continuous summarization and derivation from previous related formulas. However, as of now, there is no absolute gold standard for the use of various formulas. In the future, ophthalmologists may need to conduct larger-scale, broader studies to assist in the correct clinical selection of IOL power calculation formulas, thereby further improving the accuracy of IOL power calculation.