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


The methodological approach followed for this systematic review and meta-analysis was based upon the MOOSE (Meta-Analysis of Observational Studies in Epidemiology) methodology and the PRISMA-P (Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols) 2015 statement.[21,22]

Search Strategy

EMBASE, MEDLINE, and Cochrane Library were searched using controlled vocabulary and keywords without language or date limitations. Search terms were as follows for our MEDLINE strategy: ((((hemorrhage[MeSH Terms] OR intracranial hemorrhage[MeSH Terms] OR brain hemorrhage[MeSH Terms] OR delayed bleeding[Title/Abstract] OR delayed hemorrha*[Title/Abstract]))) AND ((tbi[Title/Abstract] OR traumatic brain injury[Title/Abstract] OR craniocerebral trauma[MeSH Terms] OR brain injury, chronic[MeSH Terms] OR brain injuries[MeSH Terms]))) AND ((coumadins[MeSH Terms] OR warfarin[MeSH Terms] OR anticoagulants[MeSH Terms])). A similar but specific strategy was used for EMBASE. We also searched the grey literature with the following databases using the same keywords: OpenSIGLE, New York Academy of Medicine (Grey Literature Collection), Greylit.org, and Google Scholar. All database searches were done under the guidance of an information specialist. Attempts were made to contact authors of all selected manuscripts to ensure data content and complete cases of missing information. References from eligible studies were also screened for other relevant publications.

Identification of Studies

Studies retained for our systematic review were: randomized controlled trials, prospective or retrospective cohort studies, case-control studies, or cross-sectional studies. To be included, studies had to focus mainly on patients with mild traumatic brain injury (see Figure 1 for definition by Carroll et al.), who used vitamin K antagonist before injury, who were scanned on hospital presentation, and whose initial scan was normal.[23] Delayed ICH was defined as any intracranial bleeding detected on the subsequent CT scan. Studies were included only if the secondary CT scan was performed within 24 h of initial investigation.

Figure 1.

Operational definition of mild traumatic brain injury (MTBI) as recommended by The World Health Organization (WHO) Collaborating Centre Task Force on Mild Traumatic Brain Injury in Carroll et al. (23).

Studies were independently screened for inclusion by three physicians (JMC, FB, and ML), with the requirement of a unanimous selection. All selected publications were then fully read by the two reviewers (JMC and MM) for inclusion. Disagreement was resolved by consensus among authors.


For this systematic review and meta-analysis, any sign of bleeding on the subsequent CT scan was considered as an event for the primary outcome. Neurosurgical intervention and death were considered as secondary outcomes.

Data Collection and Processing

Data from the selected publications were extracted by two authors (JMC and MM) using an adapted standardized data extraction form derived from Reljic et al..[24] All disagreements during this stage were resolved by consensus with the help of a third author (RD).

The following study characteristics were recorded in all cases: study design, patients characteristics (international normalized ratio [INR], age, sex), inclusion and exclusion criteria, antiplatelet therapy co-medication, mechanism of injury, number of patients, time of first and second head CT scans, bleeding description, and outcomes. Study authors were contacted for precisions when data could not be extracted from the original full text without ambiguity, or in cases of conference abstracts without full text publication. Studies with unresolved uncertainties were excluded.

Two authors independently assessed the study risk of bias (JMC, MM). Each study was evaluated using the Newcastle-Ottawa Scale to assess the quality of cohort studies in meta-analyses.[25] This tool was developed to assess the quality of non-randomized studies based on their design, content, and ease of use directed to the task of incorporating the quality assessments in the interpretation of meta-analytic results. With this system, studies are judged on three broad perspectives. In the selection category (maximum of four stars), studies received points if the cohort was representative (meaning the same type of patients who are typically kept for a 24-h observation period in the ED), if data were confirmed through secured records, and if there was a confirmation that the outcome was not present at the start of the study. There was also a point for a control cohort, which none of the studies included. A maximum of two stars was awarded in the comparability category. The first star was given if the study controlled for the most important factor, INR in our case. The second star could be awarded if significant secondary factors were considered (antiplatelet co-medication, age, mechanism of injury). Three stars were awarded in the outcome category if the outcome assessments were blinded or obtained from linked records, if the follow-up period was long enough (at least 24 h), and if all subjects were accounted for at the end of the study (<10% of lost to follow-up). All studies were kept regardless of their quality. Their evaluation is reported in the Results section.

Patients' Inclusion Criteria

In the selected studies, the data concerning a patient were extracted for analysis if: the patient was using vitamin K antagonist before injury, was victim of a mild traumatic brain injury, had a normal initial head scan on presentation, and was subsequently rescanned within 24 h. The 24 h cutoff is important because it represents the guidelines recommended observation period. Furthermore, looking at delayed ICH in patients for up to 1 month after the initial evaluation is not realistic or practical, plus it is impossible to exclude the possibility of a second traumatic incident causing a new ICH. Thus, our study population is exclusively composed of patients with an initially normal brain scan who were subsequently rescanned within 24 h.

Primary Data Analysis

The proportion of patients with delayed bleeding (positive subsequent scan) over the total number of patients undergoing repeat CT scans (our selected population) was used to conduct a single-arm meta-analysis. We used logit proportion transformation to estimate the pooled risk of presenting delayed bleeding. The pooled proportion was calculated as a back-transformation of the weighted mean of the transformed proportions (logit) using a random effect model.[26] Possible heterogeneity was tested with χ2 and I2 tests. Descriptive statistics were generated with SPSS software, version 22.0 (released 2013, IBM Corp, Armonk, NY) and meta-analyses were performed with OpenMeta-Analyst software (http://www.cebm.brown.edu/open_meta). CIs were calculated with an α of 0.05.