Abstract
Purpose :
Calculating intraocular lens (IOL) power for cataract surgery is more complex in patients with prior corneal refractive surgery. We performed a retrospective, comparative analysis of patients who had surgically treated cataracts with a history of corneal refractive surgery, assessing the accuracy of IOL outcome predictions generated by the ASCRS calculator separately and in conjunction with other devices in multiple combinations.
Methods :
The study was performed at a single site. Data from all patients with a history of uncomplicated cataract surgery, prior myopic corneal refractive surgery, and multi-platform analysis was included. Biometric measurements included Zeiss IOLMaster® 700, Zeiss ATLAS® 9000, and Oculus Pentacam®. Eyes with significant additional ophthalmic pathology affecting vision or without one-month post-operative checks were excluded. Outcomes included manifest refraction at post-operative month one, power of implanted IOL, and predicted IOL power from each group. IOL prediction error was found by subtracting the power of implanted IOL from the power of the predicted IOL. A one sample t-test to identify prediction errors significantly different from zero, an ANOVA to determine differences in IOL prediction error between groups, an F test to determine variance of IOL prediction error for each group, and a chi-square test to analyze the difference in percentage of eyes within +0.50D and +1.00D error between groups. A p-value of ≤ 0.05 was considered statistically significant.
Results :
Twenty-eight eyes met inclusion criteria. Biometry alone (IOLMaster® 700) predicted significantly higher IOL power as compared to implanted IOL power (mean=0.45D). No significant difference in IOL power prediction performance, variance of IOL prediction error, or percentage of eyes within +0.50D and +1.00D of refractive prediction error was found between groups.
Conclusions :
IOLMaster® 700 and Atlas® 9000 group yielded prediction error closest to zero with the highest percentage of eyes within 0.50D and 1.00D, though this result was not statistically significant. Adding parametric data achieved only 36-56% accuracy to within +0.50D.
This is a 2021 ARVO Annual Meeting abstract.