Previous meta-analyses of methylprednisolone (MPS) for patients with acute traumatic spinal cord injuries (TSCIs) have not addressed confidence in the quality of evidence used for pooled effect estimates, and new primary studies have been recently published. We aimed to determine whether MPS improves motor recovery and is associated with increased risks for adverse events. We searched MEDLINE, EMBASE, and The Cochrane Library, and two reviewers independently screened articles, extracted data, and evaluated risk of bias. We pooled outcomes from randomized, controlled trials (RCTs) and controlled observational studies separately and used the Grades of Recommendation, Assessment, Development, and Evaluation approach to evaluate confidence. We included four RCTs and 17 observational studies. MPS was not associated with an increase in long-term motor score recovery (two RCTs: 335 participants; mean difference [MD], -1.11; 95% confidence interval [CI], -4.75 to 2.53; p = 0.55, low confidence; two observational studies: 528 participants; MD, 1.37; 95% CI, -3.08 to 5.83; p = 0.55, very low confidence) or improvement by at least one motor grade (three observational studies: 383 participants; risk ratio [RR], 0.84; 95% CI, 0.53-1.33; p = 0.46, very low confidence). Evidence from two RCTs demonstrated superior short-term motor score improvement if MPS was administered within 8 h of injury (two RCTs: 250 participants; MD, 4.46; 95% CI, 0.97-7.94; p = 0.01, low confidence), but risk of bias and imprecision limit confidence in these findings. Observational studies demonstrated a significantly increased risk for gastrointestinal bleeding (nine studies: 2857 participants; RR, 2.18; 95% CI, 1.13-4.19; p = 0.02, very low confidence), but RCTs did not. Pooled evidence does not demonstrate a significant long-term benefit for MPS in patients with acute TSCIs and suggests it may be associated with increased gastrointestinal bleeding. These findings support current guidelines against routine use, but strong recommendations are not warranted because confidence in the effect estimates is limited.