Objectives: Long-term neurological response to treatment after a severe traumatic brain injury (sTBI) is a dynamic process. Failure to capture individual heterogeneity in recovery may impact findings from single endpoint sTBI randomized controlled trials (RCT). The present study re-examined the efficacy of erythropoietin (Epo) and transfusion thresholds through longitudinal modeling of sTBI recovery as measured by the Disability Rating Scale (DRS). This study complements the report of primary outcomes in the Epo sTBI RCT, which failed to detect significant effects of acute treatment at 6 months post-injury. Methods: We implemented mixed effects models to characterize the recovery time-course and to examine treatment efficacy as a function of time post-injury and injury severity. Results: The inter-quartile range (25th–75th percentile) of DRS scores was 20–28 at week1; 8–24 at week 4; and 3–17 at 6 months. TBI severity group was found to significantly interact with Epo randomization group on mean DRS recovery curves. No significant differences in DRS recovery were found in transfusion threshold groups. Conclusions: This study demonstrated the value of taking a comprehensive view of recovery from sTBI in the Epo RCT as a temporally dynamic process that is shaped by both treatment and injury severity, and highlights the importance of the timing of primary outcome measurement. Effects of Epo treatment varied as a function of injury severity and time. Future studies are warranted to understand the possible moderating influence of injury severity on treatment effects pertaining to sTBI recovery. (JINS, 2019, 25, 293–301)