Can we accurately estimate axial length?

Paper title: Estimation of ocular axial length with optometric parameters is not accurate

Authors: Galvis, V (1,2,3), Tello, A (1,2,3), Rey, Juan J (4), Gomez, Sergio Serrano (4), Prada, A M (1,3)

1. Centro Oftalmologico Virgilio Galvis, Floridablanca, Colombia; 
2. Fundacion Oftalmologica de Santander, Floridablanca, Colombia; 
3. Department of Ophthalmology, Universidad Autonoma de Bucaramanga, Floridablanca, Colombia.
4. School of Medicine, Universidad Autonoma de Bucaramanga, Bucaramanga, Colombia.

Date: Jun 2022

Reference: Galvis V, Tello A, Rey JJ, Serrano Gomez S, Prada AM. Estimation of ocular axial length with optometric parameters is not accurate. Cont Lens Anterior Eye. 2022 Jun;45(3):101448

[Link to abstract]

With the increasing importance to monitor myopia progression in children closely, some eyecare practitioners are using optical biometers to measure axial length (AL) accurately. However, biometers can be cost prohibitive for some. Methods of predicting axial elongation without biometers have been suggested as an alternative.

One such method described by Morgan et al used data collected at routine eye examinations (keratometry and spherical equivalent error at the spectacle plane) to calculate a predicted axial length (PAL).¹ They suggested the AL predictions could be used to further assign patients to ‘risk groups,’ classified by Tideman et al, according to their axial lengths.² These groups reflect the risk of visual impairment in myopic patients by axial length values: less than 24mm, 24-26mm, 26 to less than 28mm, 28 to less than 30mm and 30mm or longer. Tideman et al calculated percentage risks of visual loss for each category according to estimated myopia prevalence rates in the respective countries.

This study evaluated the accuracy of adopting the PAL formula to assess AL in Colombian children aged 8-17yrs from the MIOPUR study. Tideman’s classification of visual loss risk was applied to both PAL and AL measurements.³ All children underwent non-cycloplegic ocular examination which included auto-keratometry, anterior chamber depth and AL measured with AL-Scan optical biometer (NIDEK, Japan).

The results for 2129 eyes showed that PAL over-estimated AL by an average of 0.516mm for 82.15% of all eyes. Myopic eyes (n = 325) had an average 0.426mm over-estimation of AL. Differences in AL using the PAL formula were recorded in myopia eyes as being shorter by at least 0.20mm (17.12%), longer by at least 0.20mm (64.26%), longer by at least 1mm (21.92%) or longer by at least 2mm (0.45%). When Tideman’s classification was applied to the estimated PAL and the measured AL values, 15.03% of all eyes and 29.81% of myopic eyes were found to be misclassified by using the PAL estimation formula.

Using a formula method to calculate or predict axial length, rather than objectively measuring with biometers, may lead to inaccurate estimations in monitoring myopia progression, but could be useful as a starting point for broadly determining absolute axial length and hence categorising the risk of visual loss according to axial length classifications. Since the risk of visual impairment in myopia is correlated more closely with higher axial lengths than level of myopia,² an estimate of absolute axial length derived from readily available optometric data can be highly valuable in determining 'risk groups' and level of proactivity in myopia management.¹ 

The authors of this current study found some misclassification of the estimation dataset into risk groups when compared to direct axial length measures, and noted the wide categories (2mm steps) as suggested in the well known Tideman et al paper.² The authors suggest the differences found may be due to variations in crystalline lens power, not accounted for within the estimation, but also not typically measurable in a clinical setting. There is agreement that accurate monitoring of axial length changes in myopia progression is best undertaken with direct axial length measurement rather than estimations. 

The authors concluded that the likely reason for the formula giving some inaccurate predictions and classifications for axial length was not accounting for crystalline lens power - further research could tell us if this is possible from use of standard optometric clinical data. While estimations do not appear likely to replace the accuracy of an optical biometer for measuring myopia progression, further research could narrow down the patients for whom estimations provide the most accurate estimations of absolute axial length for risk classification.

Title: Estimation of ocular axial length with optometric parameters is not accurate

Authors: Galvis, V; Tello, A; Rey, Juan J; Gomez, Sergio Serrano; Prada, A M

Purpose: Myopia is a worldwide major public concern, aside from the visual disturbance needing optical correction, myopia may be associated with open angle glaucoma, retinal detachment and myopic maculopathy. The higher the myopia the higher the risk for retinal associated comorbidities, and the axial length is the more important measure to estimate risk of visual impairment. Recently a formula to predict axial length using spherical equivalent and keratometry was proposed, with the intention of categorizing the risk of visual impairment with Tideman et al. classification.

To evaluate the accuracy of an axial length prediction formula in a Colombian population 8-17 years old.

Methods: Children from MIOPUR study with optical biometer axial length measure (AL), manifest refraction and keratometry were included in the analysis. Predicted axial length (PAL) was calculated with the prediction formula. A Bland-Altman assessment was conducted, and the concordance correlation coefficient was measured. Proposed classification of AL to establish risk of visual loss was used with measured AL and with PAL. The percentage of eyes misclassified was then established.

Results: A total of 2129 eyes were included in the analysis. Mean difference of axial length (actual AL minus PAL) was -0.516 mm (-1.559 mm - 0.528 mm). Concordance correlation coefficient (CCC) of 0.656 (IC95 0.636-0.675) was found between the real AL and PAL. PAL differed from measured AL by 1 mm or more in 16.58 %, and by 2 mm or more, in 0.61 % of the eyes. In myopic eyes, PAL was in average 0.426 mm longer than the AL actually measured with CCC of 0.714 (IC95 0.666-0.761). PAL differed from measured AL by 1 mm or more in 21.92 %, and by 2 mm or more, in 0.45 % of the myopic eyes. The study revealed that 15.03 % of all eyes, and 29.81 % of myopic eyes, were misclassified when PAL was used.

Conclusions: The proposed axial length prediction formula was not accurate, and it did not adequately classify risk of visual impairment in myopic eyes in a group of Colombian children. We consider that it is not possible to predict the axial length based only on optometric data, such as the corneal radius of curvature and the spherical equivalent. This is very possibly related to the variability of crystalline lens power within a population.

[Link to abstract]