ESCRS - FPM08.02 - Evaluation Of The Prediction Accuracy Of The Hill-Rbf 3.0 Formula

Evaluation Of The Prediction Accuracy Of The Hill-Rbf 3.0 Formula

Published 2022 - 40th Congress of the ESCRS

Reference: FPM08.02 | Type: Free paper | DOI: 10.82333/2afv-yt04

Authors: Daniel Romero Valero* 1 , Jose Juan Martínez Toldos 1 , Jaime Escolano Serrano 1 , Germán Castilla Martínez 1 , Rubén Toledano Martos 1 , Sandra Pardo López 1

1Ophthalmology,General University Hospital of Elche,Elche,Spain

Purpose

To evaluate the prediction accuracy of the Hill-RBF 3.0 formula and to compare its results with those provided by other calculation formulas.

Setting

General University Hospital of Elche (Alicante, Spain).

Methods

In this retrospective interventional case series, optical biometry measurements of the Lenstar LS 900 were automatically entered with the aid of a custom-made computer program into the following online calculators: Barrett Universal II (BUII), Pearl-DGS, RBF 3.0, Emetropia Verifying Optical Formula (EVO) and Kane. SRK-T, Hoffer Q, and Haigis formula were also calculated. Mean absolute errors (MAE), standard deviation of the prediction error (PE) and the percentage of eyes in ± 0.5D were evaluated.

Results

Two hundred ninety-two eyes of 292 patients were evaluated. Mean age was 72.52±7.33 years; mean average keratometry was 44.06±1.55D; mean AL was 23.37±1.16 mm; mean ACD was 3.09±0.35 mm; mean WTW was 11.93±0.42 mm; mean LT was 4.57±0.36; mean CCT was 542.33±34.13 µm and mean IOL power was 22.05±3.24 D. In terms of number of eyes in ± 0.5D of PE, the ranking of the formulas was: Hill-RBF 3.0 (80.82%), EVO (79.79%), PEARL (79.45%), Kane (78.08%), BUII (78.42%), SRK-T (74.31%), Hoffer Q (69.87%) and Haigis (68.49%). The MAE of the different formulas was 0.437D for PEARL, 0.438 for EVO, 0.451 for Hill-RBF 3.0, 0.451 for Kane, 0.458 for BUII, 0.491 for SRK-T, 0.497 for Hoffer Q and 0.516 for Haigis.

Conclusions

The Hill-RBF 3.0 was more accurate when compared with older generation formulas and had similar prediction accuracy compared with new generation formulas.  In terms of number of eyes in ± 0.5D of PE the Hill-RBF 3.0 showed the better results in this case series. In terms of MAE, PEARL and EVO performed slightly better than Hill-RBF 3.0.