Comparative efficacy of twice-daily 0.01% atropine combined with Orthokeratology versus DIMS lenses in myopia control: an elastic net penalized regression analysis

The increasing prevalence of childhood myopia necessitates effective intervention strategies. This study compared the 1-year efficacy of twice-daily 0.01% atropine combined with either Orthokeratology (OKA) or Defocus Incorporated Multiple Segments (DIMSA) lenses in controlling axial length (AL) elongation, and utilized Elastic Net penalized regression to identify key predictive factors. This retrospective cohort study included 220 myopic children (220 eyes), divided into the DIMSA group (n=120) and the OKA group (n=100). Axial length changes (ΔAL) over 12 months were evaluated. Multivariable linear regression and subgroup analyses were conducted to evaluate the independent associations. Additionally, propensity score methods and E-value sensitivity analyses were employed to address potential treatment-selection bias and unmeasured confounding. An Elastic Net regression model with 5-fold cross-validation was employed to handle multi-collinearity and perform feature selection for potential predictors of ΔAL. The OKA group demonstrated significantly slower AL elongation than the DIMSA group (0.157 ± 0.154 mm vs. 0.313 ± 0.221 mm, P < 0.001). Multivariable regression indicated that the DIMSA regimen was associated with greater AL elongation compared with the OKA regimen (β= 0.144, P < 0.001). This association remained robust across all propensity score models, with high E-values (> 5.0) indicating strong resistance to unmeasured confounding. Subgroup analyses and interaction tests revealed a consistent association of OKA with less axial elongation across different age and myopia severity strata. Elastic Net regression identified shorter baseline AL and younger age as the primary predictors of faster progression. Notably, the Elastic Net models revealed distinct predictive architectures: the OKA group exhibited a parsimonious pattern primarily associated with age and baseline AL, whereas the DIMSA group displayed a "multi-driver" pattern associated with broader baseline covariates. The combination of twice-daily 0.01% atropine with OKA was associated with less axial elongation in this retrospective cohort compared to the DIMSA regimen. Elastic Net regression suggested distinct predictive structures for each treatment, providing a basis for hypothesis generation.