Related Experiment Video
Updated: Jun 22, 2025
Signal Acquisition, Score Interpretation, and Economics of a Non-Invasive Point-of-Care Test for Coronary Artery Disease
Published on: August 9, 2024
Analysis of the ESCRS calculator's prediction accuracy
Nir Sorkin1, Rotem Zadok, Hashem Totah
1From the Department of Ophthalmology, Tel Aviv Medical Center, Tel Aviv, Israel (Sorkin); Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel (Sorkin, Zadok); Department of Ophthalmology, Shaare Zedek Medical Center and the Hebrew University-Hadassah Medical School, Jerusalem, Israel (Totah, Abulafia); IRCCS Bietti Foundation, Rome, Italy (Savini); Departement of Ophthalmology, Hospital da Luz de Lisboa, Lisbon University, Lisbon, Portugal (Ribeiro); Vienna Institute for Research in Ocular Surgery (VIROS), Department of Ophthalmology, Hanusch Hospital, Vienna, Austria (Findl); Centro Buonsanti, Buenos Aires, Argentina (Buonsanti); Department of Ophthalmology, Unidade Local de Saúde de Coimbra, Coimbra, Portugal (Raimundo).
Purpose:
To evaluate prediction accuracy of formulas included in the ESCRS online intraocular lens (IOL) calculator using standard keratometry (K) or total keratometry (TK).
Setting:
Hospital-based academic practice.
Design:
Retrospective case-series.
Methods:
523 patients with cataract (523 eyes) were included in this study. The outcome measures were trimmed means of the spherical equivalent prediction error (SEQ-PE, trueness), precision and absolute SEQ-PE (accuracy) of all 7 formulas available on the ESCRS online IOL calculator, as well as mean (Mean-All) and median (Median-All) of the predicted SEQ refraction of all formulas. Subgroup analyses evaluated the effect of axial length on formula accuracy.
Results:
Trimmed-mean SEQ-PE range of all formulas varied from -0.075 to +0.071 diopters (D) for K-based and from -0.003 to +0.147 D for TK-based calculations, with TK-based being more hyperopic in all formulas ( P < .001). Precision ranged from 0.210 to 0.244 D for both K-based and TK-based calculations. Absolute SEQ-PE ranged from 0.211 to 0.239 D for K-based and from 0.218 to 0.255 D for TK-based calculations. All formulas, including Mean-All and Median-All, showed high accuracy, with 84% to 90% of eyes having SEQ-PEs within 0.50 D.Myopic trimmed-mean SEQ-PEs significantly different from zero were observed in long eyes for Pearl DGS (-0.110 D, P = .005), Hill RBF (-0.120 D, P < .001), and Hoffer QST (-0.143 D, P = .001) and in short eyes for EVO 2.0 (-0.252 D, P = .001), Kane (-0.264 D, P = .001), Hoffer QST (-0.302 D, P < .001), Mean-All (-0.122 D, P = .038), and Median-All (-0.125 D, P = .043).
Conclusions:
Prediction accuracy of all ESCRS IOL calculator formulas was high and globally comparable. TK-based calculations did not increase prediction accuracy and tended toward hyperopia. Observations indicating formula superiority in long and short eyes merit further evaluation.
