Traumatic brain injury (TBI) remains a global health challenge, with computed tomography serving as the primary diagnostic tool for initial evaluation. However, significant variability exists in repeat computed tomography (CT) scanning protocols, ranging from routine scheduled imaging to selective approaches based on clinical deterioration. This systematic review synthesized evidence from 1247 initially identified records, ultimately including 26 studies that met inclusion criteria, to determine optimal timing strategies for repeat CT scanning in patients with TBI. The analysis revealed dramatic heterogeneity in hemorrhagic progression rates (0.4-65%) and intervention requirements across studies, largely explained by differences in TBI severity. Patients with mild TBI (Glasgow Coma Scale GCS 13-15) demonstrated consistently lower progression rates (0.4-42%), intervention rates (0.13-0.9%), and mortality (0.13-1.2%) compared with moderate-severe TBI cohorts, which exhibited progression rates of 42.3-61%, intervention rates of 8.9-24%, and mortality of 13-18%. Critical temporal patterns emerged, with Fletcher-Sandersjöö demonstrating that 94% of hematomas ceased progressing within 24 h postinjury, establishing a crucial surveillance window. Multiple predictors of progression were identified, including concomitant intracranial lesions (subarachnoid hemorrhage odds ratio OR 3.28, subdural hemorrhage OR 4.35), advanced age, and antiplatelet therapy. Notably, patients undergoing initial CT scanning within 2-3 h postinjury showed higher rates of subsequent progression, suggesting that early scans warrant scheduled follow-up regardless of clinical status. These findings support severity-stratified approaches to repeat imaging, with routine protocols potentially justified in moderate-severe TBI, while selective strategies may be appropriate for patients with stable mild TBI. The evidence emphasizes balancing diagnostic yield against radiation exposure concerns, advocating for personalized protocols based on individual risk factors rather than universal approaches.
Taddei et al. (2025) studied this question.