To study how the leptokurtic shape of the refractive distribution can be derived from ocular biometry by means of a multivariate Gaussian model.
Autorefraction and optical biometry data (Scheimpflug and partial coherence interferometry) were obtained for 1136 right eyes of healthy white subjects recruited by various European ophthalmological centers participating in Project Gullstrand. These biometric data were fitted with linear combinations of multivariate Gaussians to create a Monte Carlo simulation of the biometry, from which the corresponding refraction was calculated. These simulated data were then compared with the original data by histogram analysis.
The distribution of the ocular refraction more closely resembled a bigaussian than a single Gaussian function (F test, p < 0.001). This also applied to the axial length, which caused the combined biometry data to be better represented by a linear combination of two multivariate Gaussians rather than by a single one (F test, p < 0.001). Corneal curvature, anterior chamber depth, and lens power, on the other hand, displayed a normal distribution. All distributions were found to gradually change with age. The statistical descriptors of these two subgroups were compared and found to differ significantly in average and SD for the refraction, axial length, and anterior chamber depth. All distributions were also found to change significantly with age.
The bigaussian model provides a more accurate description of the data from the original refractive distribution and suggests that the general population may be composed of two separate subgroups with different biometric properties.