Background: Randomized clinical trials (RCTs) in progressive multiple sclerosis (MS) often revealed non-significant treatment effects on disability progression. Objectives: To investigate whether the failure to detect a significant benefit from treatment may be motivated by a delay in treatment effect, possibly related to baseline characteristics. Methods: We re-analyzed data from two RCTs testing interferon-beta and glatiramer-acetate versus placebo in progressive MS with no significant effect on EDSS progression. We first designed a time-dependent Cox model with no treatment effect up to time = t₀, and constant hazard ratio (HR) after time = t₀. We selected the best-fitting t₀ from 0 (standard Cox model) to 2.5 years. Furthermore, we modeled the delay as a function of baseline EDSS and fitted the resulting Cox model to the merged dataset. Results: The time-dependent Cox model revealed a significant benefit of treatment delayed by t₀ = 2.5 years for the SPECTRIMS study (HR = 0.65 (0.43–0.98), p = 0.041), and delayed by t₀ = 2 years for the PROMISE study (HR = 0.65, (0.42–0.99), p = 0.044). In the merged dataset, the HR for the EDSS-dependent delayed effect was 0.68 (0.56, 0.82), p < 0.001. Conclusion: The assumption of a delayed treatment effect improved the fit to the data of the two examined RCTs, uncovering a significant, although shifted, benefit of treatment.