Improving the prediction accuracy of toric intraocular lens power calculations

Abstract

Toric intraocular lenses (IOLs) enable the correction of corneal astigmatism at the time of cataract surgery, minimising postoperative refractive astigmatism and improving visual outcomes. Nevertheless, residual refractive astigmatism persists in a significant proportion of eyes due to inherent limitations in current toric IOL power calculation models. The accuracy of these models is affected by several factors, including surgically induced astigmatism (SIA), posterior corneal astigmatism (PCA), and altered corneal power in eyes with a history of myopic laser vision correction (LVC). This thesis investigates these three major sources of error to identify opportunities for refinement of contemporary toric IOL calculation methods.

A series of retrospective observational cohort studies were undertaken, analysing data from eyes that underwent routine phacoemulsification. Biometric data were obtained using swept-source optical coherence tomography. Vector analysis was employed to quantify corneal astigmatism, SIA, and refractive astigmatism prediction error. For each study, modern toric IOL formulas (Barrett Toric, EVO Toric, Holladay I with Abulafia-Koch regression, Barrett True K Toric and EVO TK Toric) were analysed, with specific attention to the effects of standard keratometric and posterior keratometric SIA, measured versus predicted PCA, and the role of total keratometry in post-LVC eyes.

Surgically-induced astigmatism accounted for approximately 15% of refractive astigmatism prediction errors greater than 0.50 dioptres when using the Barrett Toric formula. Adjusting for posterior corneal SIA did not improve predictive accuracy. A decision-tree regression algorithm demonstrated superior performance in estimating SIA for individual eyes compared with population-based centroid values. Regarding PCA, toric IOL power formulas that incorporate predicted PCA effectively compensate for the typical against-the-rule refractive effect, but residual error persists in approximately 9% of eyes where the posterior corneal steep meridian deviates from the vertical orientation. In these eyes, measured PCA should be incorporated to improve accuracy. In eyes with previous myopic LVC, PCA magnitude and orientation were comparable to those of virgin eyes. Use of total keratometry improved spherical equivalent prediction accuracy for both the Barrett True-K and EVO formulas, but reduced refractive astigmatism prediction accuracy for Barrett True-K, indicating that total keratometry should not be applied for toric cylinder power or axis determination with this formula.

This body of work quantifies and characterises important factors in refractive prediction error in toric IOL power calculation. Surgically induced astigmatism remains the most variable and individually dependent factor, while posterior corneal astigmatism exerts a smaller but clinically significant influence. In post-LVC eyes, total keratometry refines spherical but not astigmatic outcomes. Collectively, these findings support a move towards personalised, data-driven models of toric IOL planning. Specifically, using individualised prediction of SIA based on preoperative corneal astigmatism, and the selective use of measured posterior corneal astigmatism, again based on preoperative corneal and patient characteristics. While this approach requires the collection and analysis of preoperative and postoperative refractive data, rather than using default toric IOL calculator settings, the computational tools to allow this are readily available and the additional clinical burden is modest. Future refinements will be critical to achieving consistently precise refractive outcomes and optimising patient satisfaction following cataract surgery.

Thesis is embargoed until 31 July 2027.

Date of Award	Jul 2026
Original language	English
Awarding Institution	Queen's University Belfast
Supervisor	Salissou Moutari (Supervisor) & Augusto Azuara-Blanco (Supervisor)