Estimating axial length using refraction and keratometry

Paper title: Estimation of ocular axial length from conventional optometric measures

Authors: Morgan PB (1), McCullough SJ (2), Saunders KJ (2)

1. Eurolens Research, Division of Pharmacy and Optometry, University of Manchester, Dover Street, Manchester M13 9PL, United Kingdom
2. Centre for Optometry and Vision Science Research, Ulster University, Cromore Road, Coleraine BT52 1SA, United Kingdom

Date: Published online November 27, 2019

Reference: Morgan PB, McCullough SJ, Saunders KJ. Estimation of ocular axial length from conventional optometric measures. Cont Lens Anterior Eye. 2020 Feb;43(1):18-20. Erratum in: Cont Lens Anterior Eye. 2020 Aug;43(4):413.

[Link to paper]

Myopia is typically classified by refractive error, but axial length is more closely linked to long-term risks like retinal detachment and glaucoma. Since optical biometers are expensive and not widely available in primary care, the authors proposed a method of estimating axial length from simple optometric measurements. This study aimed to develop and validate a formula using refractive error and corneal curvature to estimate axial length.

The formula was: 

The study used data from two sources: a 3-year myopia control trial involving CooperVision MiSight lenses and the Northern Ireland Childhood Errors of Refraction (NICER) study. The first dataset included 144 children aged 8–12 years who underwent annual assessments including cycloplegic and non-cycloplegic refraction, corneal topography, and axial length measurement. A linear mixed model was developed to predict axial length from spherical equivalent refraction and mean anterior corneal curvature. The formula was validated using the NICER dataset of 1,046 participants aged 6–22 years.

Key findings were as follows.

• Axial length could be estimated within ±0.73 mm (±3.0%) of biometer readings using the formula.
• Inclusion of keratometry improved prediction accuracy; refractive error alone yielded a larger error margin of ±1.26 mm.
• Results were similar between cycloplegic and non-cycloplegic refractions.
• Validation on a separate dataset showed similar agreement (±0.86 mm), with calculated values averaging 0.13 mm longer than measured values.
• The method is not reliable for tracking axial elongation over time due to its error margin (3%), and requires more accurate measures.

For eye care practitioners, particularly those in primary care settings or without access to a dedicated biometer, this study offers a valuable tool for estimating a patient’s axial length. By combining routinely collected data—spherical equivalent refraction and corneal curvature—clinicians can apply a validated formula to estimate absolute axial length within approximately ±0.73 mm of true biometer values. This estimate can be useful when discussing myopia risk, particularly when referencing published risk categories based on axial length (e.g. <24 mm, 24–26 mm, etc.) to inform decisions about initiating myopia management.

Incorporating corneal curvature measurements alongside refraction significantly improves the estimate’s accuracy compared to using refractive error alone. This insight reinforces the clinical value of keratometry as more than a tool for contact lens fitting, highlighting its role in risk assessment for myopic pathology. Additionally, the formula demonstrated consistent results across both cycloplegic and non-cycloplegic datasets, and held up well even when validated with a separate population measured using different instruments.

However, it is important to note that the formula was not designed to detect small changes in axial length over time. Its 95% limits of agreement—approximately ±0.73 mm—are too large for monitoring myopia progression, where year-to-year axial elongation may be as small as 0.10–0.20 mm.¹ As a result, its utility is limited to estimating absolute axial length at a single point in time, rather than for longitudinal assessment. Furthermore, the practical accuracy of the formula is questionable. While the authors considered ±0.73 mm a small amount of variability; this corresponds to an error in refraction of 1.83 D, which is highly clinically significant.² 

This study has several limitations. First, the derived formula was based on data from a specific age group—children aged 8–12 years—and validated in a population aged 6–22 years. While this provides a useful starting point, the formula may not perform as accurately outside these age ranges. Second, although the method was validated using a separate dataset, the two studies employed different instruments and measurement protocols, which could introduce variability not fully accounted for in the analysis. The authors acknowledge that further work is needed to understand how the formula performs across different clinical populations and with alternative equipment.

While the method is promising as a low-cost alternative for risk stratification, it is not suitable for monitoring the progression of myopia. Future research should aim to refine the formula for higher precision, and robustness when validated against different age groups, instrumentation, and broader clinical settings.