The Effect of Three-Dimensional Ocular Shape on Longitudinal Biometric Changes: A Six-Year Follow-Up Study.

PURPOSE: To investigate progression rates of ocular biometry in various ocular shapes using three-dimensional magnetic resonance imaging (3D MRI). METHODS: Highly myopic participants (spherical power ≤ -6 diopters [D]) of the prospective cohort study underwent ocular biometry, cycloplegic refraction and MRI examinations. Ocular shapes were classified using 3D MRI as spheroidal, ellipsoidal, conical, nasally distorted, temporally distorted, or barrel shaped. Spheroidal and ellipsoidal shapes were further categorized into non-deformed forms and other shapes were deformed. Mixed-effects model was used to assess the association between the change rate of spherical equivalent (SE) and ocular biometry and ocular shape. Machine learning models were developed to predict annual AL and SE progression rates. RESULTS: A total of 150 eyes including 89 non-deformed eyes and 61 deformed eyes from 75 Participants (36 females [48.00%]) were followed up for 6.70 ± 2.66 years, with a mean age of 34.30 ± 13.43 years. Mean SE and axial length (AL) were -11.49 ± 3.76 D and 28.23 ± 1.72 mm. Eyes showed progressive myopic shift (-0.36 D/y; 95% confidence interval [CI], -0.45 to -0.26) and axial elongation (0.07 mm/y; 95% CI, 0.06-0.08). Deformed ocular (61 eyes) exhibited faster SE progression (β = -0.20, P < 0.001) and AL elongation (β = 0.05, P < 0.001) compared with non-deformed shapes (89 eyes). For annual AL progression, the random forest model achieved the lowest root mean square error (RMSE) (0.067; 95% CI, 0.048-0.087). For annual SE progression, the XGBoost model outperformed other algorithms with the lowest RMSE (0.644; 95% CI, 0.192-1.248). CONCLUSIONS: Ocular shapes continue to progress in SE and AL, with deformed ocular demonstrating higher rates of SE and AL than non-deformed shapes. TRANSLATIONAL RELEVANCE: A 3D MRI-based ocular shape identifies highly myopic eyes at increased risk of rapid biometric progression.