Rethinking the optimal methods for vector analysis of astigmatism

Purpose: To evaluate the accuracy and usefulness of certain methods of analyzing astigmatic vectors. Setting: Cullen Eye Institute, Baylor College of Medicine, Houston, Texas. Design: Case samples. Methods: Using 2 sample cases for analysis of corneal surgically induced astigmatism and an actual toric intraocular lens (IOL) case, univariate analyses from the ASSORT program were compared to double-angle plots of preoperative and postoperative astigmatism and prediction errors. Results: Certain univariate figures for analyzing the 2 sample cases were misleading. For the toric IOL case, some of the key outcome vectors were inaccurate. Conclusions: ASSORT's univariate analysis of astigmatic vectors can be unpredictably erroneous and misleading. Recommended vector analyses should include double-angle plots with centroids and confidence ellipses of preoperative and postoperative astigmatism and the prediction errors, along with means and standard deviations of these vector magnitudes.