Abstract
Purpose: :
Norrby (J Cataract Refract Surg 2008; 34:368-376) has shown that about 1/3 of the variance in IOL power calculation is due to the estimation of postop IOL position, based on a standard deviation (SD) of 0.31 mm in this estimate. In this study we investigate how new and current algorithms for estimation of postop IOL position influence the prediction of refractive outcome after cataract surgery.
Methods: :
In a pilot study we measured axial length (ALi), anterior chamber depth (ACDi) and mean corneal radius (CRi) with Zeiss IOLMaster ("i"), and ACDa, lens thickness (LTa), and white-to-white (WTWa) with Zeiss ACMaster ("a") preop on patients scheduled for routine cataract surgery. Either an AMO CeeOn 911A IOL (n=9) or an AMO TECNIS Z9000 IOL (n=13) was implanted. These models are identical, except for the shape of the anterior surface, and should therefore end up at the same axial position in a given eye. Postop ACDay was measured with ACMaster ("y" denoting result). By partial least squares regression (PCL) we tried various combinations of the preop parameters to predict ACDay. The usefulness of the fits was judged in terms of P-value obtained by CV-ANOVA (Cross-Validation of predictive residuals). Among current algorithms we used ALi and Ki = 337.5/CRi to calculate effective lens position (ELP) with the SRK/T, Hoffer Q and Holladay 1 formulas, and ALi and ACDi with the Haigis formula. For each formula, ELP was transformed to ACD by subtraction of a constant that made the difference with ACDay on average zero, for both IOL models.
Results: :
The best fit was found with a linear combination of ALi, ACDi and CRi, which predicted ACDay with a standard deviation (SD) of 0.20 mm and P=0.031. Using ALi and Ki, as in the SRK/T, Hoffer Q and Holladay 1 formulas, P=1 was obtained. With ALi and ACDi, as in the Haigis formula, P=0.27 was found. The prediction algorithms in those formulas all gave SD close to 0.30 mm in prediction of ACDay, which is higher than the SD of ACDay itself (0.25 mm). With our prediction algorithm adapted to the Haigis formula, we found a mean absolute error (MAE) of 0.49 D. The formulas had MAE ranging from 0.51 D to 0.62 D.
Conclusions: :
We found an algorithm that could predict actually measured postop IOL position with statistical significance. When implemented in the Haigis formula it gave lower MAE than any of the current formulas. Further we found that the parameter combinations used in current formulas are without predictive power for the IOL position. In fact all formulas had a prediction error with larger SD than the SD of the actually measured IOL position. However, our sample size is small. A larger study is ongoing.
Keywords: intraocular lens • cataract