Risk Of Delayed Intracranial Hemorrhage in Anticoagulated Patients With Mild Traumatic Brain Injury

Systematic Review and Meta-Analysis

Jean-Marc Chauny, MD, MSC; Martin Marquis, MSC; Francis Bernard, MD; David Williamson, BPHARM, PHD; Martin Albert, MD; Mathieu Laroche, MD, MSC; Raoul Daoust, MD, MSC


J Emerg Med. 2016;51(5):519-528. 

In This Article


Study Selection

The literature search was conducted in December 2014, with an update in September 2015 (Figure 2). This search yielded 892 unique citations, of which 89 full texts were reviewed by the authors, 12 studies were carefully considered, and 7 studies ended up meeting the final inclusion criteria, for a total of 1594 patients.[27–38]

Figure 2.

Flow diagram depicting the selection process for studies to be included into the systematic review.


From the 12 studies identified, 5 had to be excluded from our meta-analysis. In the first study, only 4 of 346 patients were on warfarin before injury; the focus of this study was clearly not on warfarin-anticoagulated patients.[34] Of note, none of the 4 patients had any significant findings on the second scan. Given the small sample size, we decided not to include these results in our analysis. In the second excluded study, the authors described a 1.19% (2 of 168) rate of delayed ICH.[35] However, almost all patients had a positive initial scan (166 of 168 or 98.8%), while the 2 remaining patients were found to be positive on the subsequent CT scan. No data were available for patients with a normal initial scan, preventing the estimation of proportion of patients with delayed ICH. The third was a systematic review on delayed ICH in anticoagulated patients without any meta-analysis calculation.[36] The last two studies were conference abstracts and were not retained for the following reasons: 2 of 130 patients had delayed ICH, however, no time frame was associated with the second CT scan;[37] 1 patient on warfarin developed a delayed ICH, however, data on prior anticoagulant use were available for that patient only and not for the remainder of the cohort.[38]

Selected Studies

The characteristics of included studies are reported in Table 1, while the details of the patient population are in Table 2. Four of the studies were retrospective and two involved more than one center.[28–30,32,33] The sample size of the studies ranged from 58 to 687 patients, with a median of 137 patients. The mean age was 76.9 years, 55.6% of the patients were female, the Glasgow Coma Score was 15 for almost all patients, mean INR was 2.8, and the most common cause of injury was falls (79.2%). The overall incidence (number of new ICH cases during the first 24 h) of delayed ICH as reported in the included studies ranged from 0 to 6% (or up to 11% when considering CI). This wide disparity in results generated moderate statistical heterogeneity (I2 = 42.8%). We chose to use the random effect model in our meta-analysis due to this apparent clinical heterogeneity in order to obtain a broader CI and err on the side of patients' safety.

Risk of Bias

The overall quality of the studies was moderate, mostly with respect to unclear or missing information, but none of the included studies contained biases that were deemed significant enough to justify exclusion. Two studies had a higher risk of bias. The main concern with these two papers was the short delay before the second CT scan, at 8.8 h and 6.0 h, respectively.[28,32] For each study, the quality of the individual components is presented in Table 3.


Data on delayed intracranial bleeding were extracted from the 7 selected studies to generate the Forest plot for the overall risk of delayed ICH presented in Figure 3. Regarding the primary outcome (risk of delayed ICH), we reported signs of intracranial bleeding within 24 h in 14 of 1,594 patients. The nature and localization of the bleeding are reported in Table 3. Thus, the pooled estimated overall risk of delayed bleeding was 0.6% (95% CI 0–1.2%). In a sensitivity analysis including data from excluded studies, the risk of delayed bleeding did not change significantly (pooled estimate = 0.5%, 95% CI 0.1–0.9%).[34,35,37,38] The patients with delayed ICH were mostly male (10 of 14 or 71%), with a mean age of 79.1 ± 8.5 years, and an INR of 2.71 ± 0.92 at presentation (Table 4). Because we only had the INR for a limited number of patients and never for patients without delayed ICH, it was impossible to determine the impact of this variable. Falls were the most common mechanism of injury (6 of 13) within the group of affected patients.

Figure 3.

Forest plot of associated risk of delayed intracranial hemorrhage after mild brain traumatic injury (24 h post normal brain scan) in patients anticoagulated with vitamin K antagonist.

Careful examination of the 13 patients with known outcomes revealed that only 1 required surgical intervention, while another died from his injuries (there was also one case of unrelated death). All other patients were discharged without any further clinical interventions. The resulting risk of clinically relevant poor outcomes (secondary outcomes of neurosurgical intervention and death) was thus 0.13% (95% CI 0.02–0.45%) (2 of 1,594).