ESCRS IOL Power Calculator Accuracy

Measured by conventional discrete accuracy values such as mean and median absolute prediction error, the seven IOL power calculation formulas available from the ESCRS online calculator, plus three incorporated into the IOLMaster 700 (Carl Zeiss Meditec), showed statistically comparable accuracy in a test of normal-length Korean eyes. However, two stood out when evaluated using the Formula Performance Index (FPI), which derives a single total accuracy value by combining several weighted discrete measures, report Hun Lee MD, PhD and Nahyun Park MD. 

In a test involving 221 eyes in 221 patients who underwent cataract surgery at the Asan Medical Center in Seoul, Republic of Korea, Drs Lee and Park found that the Cooke K6 and PEARL-DGS formulas produced FPI scores exceeding 0.700, while the other eight ranged from 0.334 to 0.572. The eight formulas, in descending order of FPI, were Hill-RBF 3.0, Barrett Universal II (two versions from the ESCRS and IOLMaster 700 tested), Kane, EVO 2.0, SRK/T, Hoffer QST, and Holladay 2. Prediction errors were calculated six months after surgery.

Sources of prediction error

The slope of the correlation of prediction error to axial length is one factor incorporated into the FPI. First developed by Dr Wolfgang Haigis, other factors in the FPI are prediction error standard deviation, median absolute error, and the percentage of eyes with a prediction error within ±0.5 D. Because it is a composite measure, Dr Lee said they could not calculate the statistical significance of differences in the FPI.

Cooke K6 and PEARL-DGS were the only two formulas that did not show a significant correlation with axial length and lens tilt. Dr Park hypothesised this may be because these two formulas use a method developed by Dr David L Cooke for adjusting the axial length value to reduce skewing prediction errors in longer and shorter eyes. Testing this hypothesis is one goal of the team’s ongoing research into IOL formula accuracy with Dr Cooke, she added.

The study also found the SRK/T formula showed a significant keratometry correlation. Prediction error correlated with anterior chamber depth in the Barrett, SRK/T, Holladay 2, and Hoffer QST.

Identifying the source of prediction errors may help guide formula use, particularly in longer or shorter eyes or eyes with unusual anatomy, Dr Lee said. “If there is a correlation, and you are operating [on] extreme eyes, you might get a bigger prediction error.”

Broadening the research

Mean axial length of the eyes tested was 23.82 mm, ranging from about 21.50 mm to nearly 28.00 mm. Drs Lee and Park are extending their study to 1,000 eyes and hope to include a greater sample of shorter and longer eyes for a better idea of how the formulas perform in Asian eyes, which have high rates of myopia. The larger sample may reveal statistically significant differences, Dr Lee said.

The team is also collaborating with Dr Dante Buonsanti, who heads the ESCRS online calculator project, to see if there are any significant differences between Asian and European eye samples. “Comparison of many formulas is very important,” Dr Lee said, even in clinical practice. He routinely uses 10 different formulas to guide lens power selection.

Drs Lee and Park presented at ARVO 2024 in Seattle, US.

Hun Lee MD, PhD is an associate professor of ophthalmology at Asan Medical Center, University of Ulsan College of Medicine, Songpa-gu, Seoul, Republic of Korea. yhun777@gmail.com

Nahyun Park MD is a third-year ophthalmology resident at Asan Medical Center. laurenpark66@gmail.com