Aims: To compare methods available to correct the magnification of images that result from the optics of the eye and identify errors, and source of error, of the methods.
Methods: 11 methods were applied to ocular biometry data from three independent cohorts. Each method was compared with the method of Bennett, which uses most biometric data. The difference between each method and Bennett's is the "error" of the method. The relation between the error and axial length, ametropia, and keratometry was explored by linear regression analysis.
Results: Methods using axial length had the lowest mean (+0.5 to +2.6%) and standard deviation (0.6 to 1.2%) of errors. Of methods using keratometry and ametropia only, the lowest mean (-1.4% to +4.4%) and standard deviation (2.9 to 4.3%) of errors was found for a new method described in this paper, and that used by the Heidelberg retina tomograph (HRT). The highest mean error (+2.2 to +7.1%) was found for Littmann's method. Littmann's correction was larger than the HRT's by 3.5 to 3.7%. The mean difference between the new and HRT methods and the "abbreviated axial length" method of Bennett is -1.3 to +2.0%. The error of the "keratometry and ametropia" methods is related to axial length.
Conclusions: Methods using axial length are most accurate. The abbreviated axial length method of Bennett differs little from more detailed calculations and is appreciably more accurate than methods using keratometry and ametropia alone. If axial length is unknown, the new and the HRT methods give results closest to the abbreviated axial length method.