Abstract
Purpose:
To evaluate the validity of the FullMonte (FM) intraocular lens (IOL) power computational process and compare its accuracy to four popular formulas.
Methods:
A retrospective analysis of the first operated eye from 160 patients who underwent cataract surgery with an Acrysof SN60WF IOL (Alcon, Inc) implantation by a single surgeon (JL) was conducted. Biometric data and keratometry obtained by IOLMaster (Carl Zeiss Meditec) as well as pre- and postoperative visual acuities and refractive errors were recorded. Data from the first 80 patients were used to optimize the FM calculator, which was used to get the predicted refraction for the next 80 patients. The absolute difference between predicted refraction and postoperative refraction (spherical equivalent (SE)) was calculated as the prediction error (PE). The PE for Full Monte (FM) was compared to the PE for the Haigis, HofferQ, Holliday1, and SKR/T in the same set of 80 patients. Subset analysis was conducted for long (<26 mm, n=5) and short (<22 mm, n=6) eyes. Repeatability of the FM computing process was also assessed using the intraclass correlation coefficient (ICC).
Results:
The average preoperative visual acuity and absolute SE was 20/62 and 2.27 diopters. Postoperatively, the average visual acuity was 20/33 and the SE was 0.49 diopters. The average PE for the optimized FM was 0.48 +/- 0.43, for Haigis 0.60 +/- 0.45, for HofferQ 0.52 +/- 0.42, for Holladay 0.50 +/- 0.43, and for SRK/T 0.48 +/- 0.43. For short eyes, the average PE for FM was 0.31 +/- 0.21, for HofferQ 0.44 +/- 0.32, for Holladay 0.52 +/- 0.45, for SRK/T 0.40 +/- 0.42. For long eyes, the average PE for FM was 0.91 +/- 0.53, for HofferQ 0.92 +/- 0.44, for Holladay 0.78 +/- 0.42, for SRK/T 0.46 +/- 0.18. The ICC for Full Monte was 0.69.
Conclusions:
Markov Chain Monte Carlo estimations are based on raw data simulations of results, and not on formulaic IOL calculations. They are optimized to a particular surgeon's outcomes and predict future results directly from the data. This study shows that the PE calculated from the Full Monte calculator was not statistically different from current formulas, suggesting that it is at least as accurate as existing popular formulas used in IOL power selection. Subgroup analysis shows a trend toward better predictions by the Full Monte in short eyes and therefore analysis of a larger sample size may prove to be even more useful.