Comparison of the accuracy of nine intraocular lens power calculation formulas using partial coherence interferometry

Cataract surgery is the most performed procedure in the world. To achieve the target refraction, several intraocular lens (IOL) power calculation formulas have been developed to improve the accuracy of IOL power predictions. We compared the accuracy of 9 IOL power calculation formulas (SRK/T, Hoffer Q, Holladay 1, Haigis, Barrett Universal II, Kane, EVO 2.0, Ladas Super formula and Hill-RBF 3.0) using partial coherence interferometry (PCI). We collected data from patients who underwent uncomplicated cataract surgery with implantation of 1 of 3 IOL types currently used in our center. All preoperative biometric measurements were performed using PCI. Prediction errors (PE) were deduced from refractive outcomes evaluated 3 months after surgery. The mean prediction error (ME), mean absolute prediction error (MAE), median absolute prediction error (MedAE), and standard deviation of prediction error (SD) were calculated, as well as the percentage of eyes with a PE within ± 0.25, ± 0.50, ± 0.75 and ± 1.00D for each formula. We included 126 eyes of 126 patients. Kane achieved the lowest MAE and SD across the entire sample as well as the highest percentage of PE within ± 0.50D and was shown to be more accurate than Haigis and Hoffer Q (P<001). For an axial length of more than 26.0mm, EVO 2.0 and Barrett obtained the lowest MAEs, with EVO 2.0 and Kane showing a higher percentage of prediction at ±0.50D compared to old generation formulas except for SRK/T (P=04). All investigated formulas achieved good results; there was a tendency toward better outcomes with new generation formulas, especially in atypical eyes.