Determine whether the early total-response ridge outperforms the longitudinal mixed-effects model

Determine whether the early total-response ridge model provides superior prediction of future ADAS-Cog change compared with the longitudinal mixed-effects model when evaluated beyond the internally validated two-site rTMS trial dataset.

Background

The study compares several approaches for forecasting ADAS-Cog change at 8, 16, and 24 weeks after treatment using information available by the Week-5 landmark. The early total-response ridge model uses aggregate Week-3 and Week-5 ADAS-Cog changes, whereas the longitudinal mixed-effects model uses grouped cognitive composites and explicitly represents repeated outcomes within participants.

Although the early total-response ridge had the lowest numerical overall MAE among all evaluated models, the difference from the mixed-effects model was small and statistically uncertain. Resolving whether either approach is genuinely superior requires larger samples and independent external validation beyond the single internally analyzed trial dataset.