A Further Understanding of Posterior Corneal Astigmatism: An Investigation of Constant and Keratometry-Scaled Components

Posterior corneal astigmatism is conventionally treated as a fixed against-the-rule vector. This study tested whether it is better described as two components, a constant Base Astigmatism and a component scaling with anterior corneal astigmatism. This study also evaluated a prediction model built on that idea. Retrospective cohort study. Posterior astigmatism was back-calculated from preoperative anterior corneal astigmatism and postoperative refractive astigmatism in 39,752 eyes following cataract surgery with a non-toric intraocular lens, after imputing a value for surgically induced astigmatism. Base Astigmatism was computed as a laterality-specific vector mean in eyes with 0.25 to 1.99 diopters of anterior astigmatism; the residual, termed Keratometry-Informed Astigmatism, was fitted against anterior astigmatism. The resulting model, K6, was compared with Abulafia-Koch, Barrett, PEARL-DGS, Kane, EVO 2.0 and unadjusted keratometry in 24,615 independent eyes, using double-angle plots, centroids and mean absolute errors. Base Astigmatism was −0.446 diopters at 92 degrees in right eyes and −0.453 at 90 degrees in left eyes. Keratometry-Informed Astigmatism equaled 0.299 of anterior corneal astigmatism. It was relatively consistent in magnitude across meridians, and aligned closely with the anterior steep axis at all magnitudes. The resulting model, termed K6, predicts postoperative astigmatism as 0.701 × anterior corneal astigmatism at its steep meridian, plus the laterality-specific Base Astigmatism, plus surgically induced astigmatism, all combined as vectors. K6 gave the lowest mean absolute error: 0.48 versus 0.50 to 0.70 diopters (D), and the highest proportion of eyes within 0.50 D of prediction, 63% versus 35% to 59%. Its advantage was largest in with-the-rule eyes and at higher magnitudes; Kane performed better in against-the-rule eyes below 2 diopters. Posterior corneal astigmatism is well described by a constant component and a component scaled to keratometry. This simplifies astigmatism modeling and improves prediction using keratometry alone.