PURPOSE: To determine whether imaging features derived from fundus photographs contain 3D eye shape information beyond that available from spherical equivalent refraction (SER). METHODS: We analysed 99 eyes of 68 normal adults in the UK Biobank. An ellipsoid was fitted to the entire volume of each posterior eye (vitreous chamber without the lens)-segmented from magnetic resonance imaging of the brain. Asphericity was computed based on the semidiameters of the ellipsoid's axes to describe posterior eye shape along the horizontal (temporal-nasal) and vertical (superior-inferior) meridians, while volume was calculated as the total number of foreground voxels. Mixed-effects linear regression models were used to test the association of SER with asphericity and volume, controlling for age and sex. Then, the association between various fundus features and asphericity was tested-both before and after controlling for SER, age and sex. RESULTS: Posterior eyes were generally oblate (asphericity >
0), but the degree of oblateness reduced as SER decreased, with the shape tending towards prolateness in high myopia. Neither sex nor age influenced asphericity. However, males had larger posterior eyes on average (this difference disappeared after height was additionally controlled for). Optic disc (OD) orientation, OD-fovea angle, vessel tortuosity, vessel fractal dimension and central retinal arteriolar or venular equivalent (CRAE or CRVE) showed significant univariable associations with asphericity along at least one meridian. After controlling for SER, age and sex, a more negative OD-fovea angle (larger OD-fovea angular separation) remained significantly associated with reduced horizontal oblateness (p = 0.01). Similarly, decreasing CRAE (narrower arterioles) remained significantly associated with reduced oblateness along both the horizontal (p = 0.04) and vertical (p <
0.01) meridians. CONCLUSIONS: Variations in OD-fovea angle and CRAE are associated with differences in ocular asphericity-even in eyes with similar SER-suggesting that fundus imaging provides eye shape information beyond what is available from refractive error alone.