Across 271 short eyes, EVO and Holladay 1 had the lowest overall mean error, while Haigis, K6, Kane and PEARL-DGS led in particular axial-length ranges.
In a 2,550-eye retrospective multicenter dataset, Moussa had the lowest mean absolute error, but its margin over the closest formulas was measured in hundredths of a diopter.
In 164 eyes measuring 26.04–29.72 mm, Holladay 1 NLR and SRK/T WK2 achieved an RMSAE of 0.393 D; some adjusted older formulas underperformed newer methods.
In 255 long eyes, Hoffer QST had the lowest absolute error at 0.33 D, yet only about 70% of results from it and the strongest alternatives fell within ±0.50 D.