Multivariate analysis versus vector analysis to assess surgically induced astigmatism

To present a multivariate probability computation method for assessing surgically induced astigmatism.

Department of Ophthalmology, Rennes University Hospital, Rennes School of Medicine, France.

The multivariate method was used to evaluate 100 patients who had cataract surgery by phacoemulsification. Keratometry was recorded on the day before and 12 days after surgery. Surgically induced astigmatism was assessed by the multivariate method as well as by most of the published vector analysis methods.

The mean surgical induced astigmatism (+/- SD) with the multivariate analysis was 1.18 +/- 0.36 diopters (D) for the cylinder power and 25.00 +/- 5.50 degrees for the cylinder axis. With the vector analysis, the mean surgically induced astigmatism was 1.67 +/- 0.54 D with the Naylor, Jaffe, and Holladay methods; 0.45 +/- 0.30 D with the latest Naeser method; 1.34 +/- 0.38 D with the Cravy method (Cravy's vector); and -0.05 +/- 0.42 D for Cravy's delta K.

Although vector methods constitute interesting geometrical models, their number and their lack of linearity and explicit expression of results make them unsuitable for statistical analysis. Instead, a reliable, easily programmable method that uses existing software is recommended.