Background and aim of the study
Traumatic brain injury (TBI) is one of the most common reasons for visits to the Emergency Department (ED) [
1,
2]. More than 90% of patients who have suffered from head trauma present with a mild traumatic brain injury (MTBI), which is usually identified by a Glasgow Coma Scale (GCS) score ≥ 13 [
3‐
5]. MTBI is generally benign, without any sequelae; however, in almost 10% of cases, it is associated with acute intracranial hemorrhage (ICH), especially in elderly people who, for several pathophysiological reasons, are more susceptible to intracranial complications [
6‐
9]. Although less than 1% of patients with complications require neurosurgical intervention, considering the potentially life-threatening condition of some patients, it is of pivotal importance to identify risk factors that can help clinicians select such cases and avoid useless imaging in low-risk patients. This is particularly important for reducing the observation time in the ED or the length of hospitalization, which is well known to increase the risk of complications such as infection or delirium, especially in the elderly population.
While there is agreement among several guidelines recommending a computed tomography (CT) scan after MTBI in patients administered an anticoagulant, there is no consensus on how to monitor patients on antiplatelet therapy. Apart from neurological deterioration, there is even more confusion on indications for repeating a CT scan, resulting in each institute having its own protocol; many institutes recommend a routine repeated CT scan and observation for 12–24 h in patients receiving an anticoagulant even though several studies have demonstrated a low risk of delayed bleeding (DB) in this subgroup of patients [
10‐
12].
However, most of the studies evaluating the risk of DB in MTBI patients have not distinguished between patients on anticoagulation vs. antiplatelet therapy. Moreover, some authors have found that antithrombotic therapy is not a significant risk factor for DB when adjusted for age, with an old age being the most powerful risk factor for DB [
13]. Considering the ageing population and the increasing prescription of antiplatelet and dual antiplatelet therapy (DAPT), it is important to have clear indications for this particular subgroup of patients. For this reason, we conducted a systematic review and meta-analysis to evaluate the risk of DB after MTBI in patients on antiplatelet therapy.
Methods
Search strategy and study selection
A systematic review and meta-analysis were conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement [
14] and the Meta-analysis of Observational Studies in Epidemiology guidelines [
15]. The study protocol was designed and validated by all of the investigators before the search was started.
A systematic search on MEDLINE and EMBASE was performed from database inception to December 8, 2020. Combinations of the following terms were used: (head trauma OR brain injury OR cerebral injury OR brain trauma OR cerebral trauma OR brain contusion OR cerebral contusion OR concussion OR craniocerebral trauma) AND (antithrombotic OR platelet aggregation inhibitor OR carbasalate calcium OR aspirin OR lysine acetylsalicylate OR clopidogrel OR ticagrelor OR dipyridamole OR prasugrel OR ticlopidine OR indobufen OR thienopyridine OR antiplatelet OR acetylsalicylic acid OR salicyl*). Both prospective and retrospective studies published in English were included. Because of the high risk of bias, case reports and case series were excluded.
Inclusion criteria for studies were recruitment of patients ≥ 16 years on antiplatelet therapy (dipyridamole alone was not considered an antiplatelet therapy), with MTBI at ED presentation (defined according to the study definition), a first negative head CT scan, and the availability of a second CT scan or clinical follow-up. If data on MTBI could not be separated from moderate or severe TBI, the information was included in the study. It was preferred to increase the sensitivity and overestimate the rate of DB rather than underestimate it; therefore, a small portion of non-mild TBI cases were enrolled. Sensitivity analysis without the moderate or severe TBI studies was eventually performed, and those studies including only moderate or severe TBI were excluded.
ICH was defined as any type of intracranial bleeding (epidural, subdural, subarachnoid, or intraparenchymal hemorrhage) found at head CT scan. First CT scan was defined as the first CT scan performed in the ED, irrespective of the time lag between the trauma and CT acquisition, according to the study definition. The repeated head CT scan was defined as the second CT scan performed according to the study protocol, irrespective of the time lag between the two scans, meaning that the second CT scan could have been done routinely or as clinically indicated. The bleeding was defined as delayed if it was found after a first negative head CT scan, without any time restriction.
Two reviewers (G.C. and A.J.) independently screened all of the titles and abstracts of the retrieved articles to detect potentially eligible studies and to remove irrelevant reports. If the reviewers disagreed on a given study, it was initially included to increase the search sensitivity. Full texts of the selected articles were then obtained. Four reviewers (G.C., A.J., M.B., and E.M.F.) extracted data on the study design, inclusion and exclusion criteria, sample size, clinical characteristics of the patients, mechanism of injury, antiplatelet medication, and outcomes of interest using a predefined data extraction form. For each original study, the outcome data of only those patients corresponding to our inclusion criteria were extrapolated. Disagreements were discussed by all reviewers until a consensus was obtained. If the data could not be retrieved from the selected studies, the corresponding authors were contacted for clarification.
Study outcomes
The primary outcome of this study was to evaluate the absolute risk of DB in MTBI patients on antiplatelet therapy. The secondary outcome was the absolute risk of clinically significant DB, defined as DB leading to death or to neurosurgical intervention after a first negative CT scan. Exploratory analyses were also performed to detect any differences in DB related to the age of the patients, to the severity of the trauma, and to the type of antiplatelet medication.
Risk of bias assessment
Two reviewers (G.C. and A.J.) independently assessed the methodological quality of the selected articles using an adapted version of the Quality Assessment of Diagnostic Accuracy study-2 (QUADAS-2) [
16]; disagreements were discussed by all reviewers until a consensus was reached. QUADAS-2 was originally created to assess the quality of diagnostic studies; an adapted version of QUADAS-2 was used as it was more suitable for our review and review question. Indeed, the PRISMA protocol is based on the patient, intervention, comparison, outcome, and study query model, which is the basis for QUADAS-2 quality assessment as well. QUADAS-2 guides the author in assessing the risk of bias and the concern regarding applicability of the study to the review question along four domains: “patient selection,” “index test,” “reference standard,” and “flow and timing.” In our modified version, the domain regarding the index test was removed, and the other three domains were slightly adapted as presented in the supplementary material. Overall, studies that received only a low risk of bias evaluation were rated as having a “low risk of bias;” studies that received at least a high risk of bias evaluation were considered as having a “high risk of bias;” if no high risk of bias was detected but “unclear” was answered to at least one question, the study was considered at “unclear risk of bias.” The same concept was applied for the applicability of studies to the review question (Additional file
1 for adapted QUADAS-2 version template).
Data analysis
Categorical data were reported as counts and percentages. Continuous variables were presented as the mean ± standard deviation or as the median and interquartile range, based on the original study reports. For each primary study, the incidences of ICH, mortality, and neurosurgery as the proportion of events in the included patients were calculated with the 95% confidence interval (CI). Due to expected clinical heterogeneity among the included studies, all meta-analyses were performed using the DerSimonian and Laird random-effects model, after transforming the individual study proportions according to the Freeman-Tukey double arcsine transformation. The pooled estimates obtained from the meta-analyses were then back-transformed, and the results were expressed as pooled proportions. The chi-square test was used to assess differences of the proportions between some of the predefined subgroups (p < 0.05, two sided), as reported below. Due to the large number of small studies with zero events, for the secondary outcome, the meta-analysis was performed using an approximated fixed-effect approach, weighting single study proportions for the study sample size.
The chi-square test was used to assess the statistical heterogeneity (with p < 0.1), which was quantified using the inconsistency index (I2). Heterogeneity was considered relevant when I2 > 50%. Stata software (version 16, StataCorp LLC) was used for the data analysis.
Subgroup and sensitivity analyses
An important amount of heterogeneity was expected between the original studies, so prespecified subgroup analyses were performed in order to reduce the heterogeneity. The following factors were considered for subgroup analysis: different types of antiplatelet medication (aspirin, clopidogrel, prasugrel, ticagrelor, or DAPT); different ages considering the inclusion criteria of the original studies (< 60 years old vs. ≥ 60 years old); different outcome detection methods (follow-up vs. routine repeated CT scan).
Sensitivity analysis without the studies including moderate or severe TBI was eventually performed, considering two different definitions of mild TBI (TBI and GCS > 13 vs. GCS ≥ 13). Sensitivity analysis considering all patients lost at follow-up and all unexplained deaths as DB was also carried out.
Discussion
The main findings of this systematic review and meta-analysis were as follows: i) There is a low risk of DB in MTBI patients administered APAs; ii) There is an even lower risk of clinically relevant DB among these patients; iii) A routine repeated CT scan detects more DB compared to a clinical follow-up assessment only, but most of these bleedings are not associated with major events such as neurosurgical intervention or death; and iv) DAPT patients are at a higher risk of DB compared to ASA-only patients.
We aimed to perform this systematic review and meta-analysis to quantify the risk of DB after MTBI in patients on antiplatelet therapy in order to compare the possible diagnostic yield of a repeated head CT or a short-term clinical follow-up to a single head CT scan in MTBI patients administered APAs.
We found that the risk of DB in MTBI patients on antiplatelet therapy was as low as 0.77% (95% CI 0.2–1.5%) and that the absolute risk of clinically relevant DB was 0.17% (95% CI 0.06–0.4%). Furthermore, our review showed that a clinical follow-up is far less sensitive at detecting DB compared to a routine repeated head CT scan (follow-up: 0.22%, 95% CI 0.00–1.05 vs. routine repeated CT scan: 1.29%, 95% CI 0.6–2.18%.). We are aware that a repeated CT scan is the reference standard for DB detection after MTBI; however, in everyday clinical practice, waiting for a repeated CT scan exposes the patient to the risks of a prolonged ED stay, often without any relevant clinical benefit. Our review shows that even if the CT scan detects more DB when compared to an observation-only strategy, it seldom leads to major events such as neurosurgical intervention or death. Indeed, we found that the risk of clinically relevant DB is very similar to the risk of DB when assessed through a clinical follow-up. We speculate that a routine repeated head CT scan could reveal more minor DB than a clinical observation or follow-up, without leading to any relevant change in patient management.
When considering guideline recommendations for CT scan execution in MTBI patients on antiplatelet therapy, the topic is only covered by the Scandinavian and NICE guidelines [
32,
33]. The former only recommends a first head CT (or in-hospital observation ≥ 12 h after injury) for patients ≥ 65 years and on antiplatelet medication, while imaging is not required for younger subjects. The National Institute for Health and Care Excellence guidelines of the UK report that scientific data are insufficient to provide recommendations on the management of these patients. A recent meta-analysis has demonstrated a small increased risk of immediate ICH in MTBI patients on antiplatelet therapy, especially if concomitant with another risk factor for ICH such as GCS < 15 or age > 65 years [
34]. No guideline recommendations exist for routine repeated CT scans in MTBI patients on antiplatelet therapy, since the literature on this topic is very scant. We think that the results of our work could be useful in case of MTBI guideline revisions. The most valuable results that should be included in future guideline revisions are as follows: the low overall risk of DB in patients with MTBI and administered APAs, the low risk of DB in patients with a routine repeated head CT scan, and patients on DAPT seem to be at a higher risk of DB compared to patients taking only ASA.
Only one recent meta-analysis has analyzed the risk of DB in MTBI patients on APAs [
11]. Huang and colleagues have shown that in MTBI patients on antithrombotic therapy, repeat scans should be discretionarily based on neurologic assessments and that routine repeated CT may identify a larger proportion of minor delayed ICH. In addition, they found a slightly higher risk of DB compared to our results. Their meta-analysis included all types of antithrombotic agents, but a pooled estimate of DB for APAs only is not available. Even when considering the only comparable subgroup—ASA-only patients—we believe that the findings are not completely comparable to our work. The pooled results reported by Huang and colleagues included both APAs and anticoagulant agents, and the meta-analysis lacks some of the original studies that are included in our work [
17,
18,
21,
30]. Moreover, their meta-analysis, even the “ASA-only” patient subgroup, did not include any study that used a clinical follow-up as the primary outcome assessment. Furthermore, the work by Huang et al. does not show any data about DB and DAPT. In regard to DB risk according to different antiplatelet therapy subgroups, our work highlights that while DB seems to be comparable between patients taking clopidogrel vs. ASA, it is far more elevated in patients on DAPT. Due to the elevated risk of DB in MTBI patients on DAPT, we suggest a case-by-case evaluation of the need for a second CT scan (at least 6 h after the first CT scan) before discharge in this subgroup of patients in order not to miss any clinically relevant DB. Of note, in the original studies included in our systematic review, there were very few data on newer APAs, such as ticagrelor or prasugrel, which do not allow any further consideration on newer antiplatelet agents.
Substantial heterogeneity was present in the risk of DB according to the original studies. In fact, DB risk for MTBI patients on APAs ranges from 0% in some studies [
18,
20,
23‐
25,
31] to 4% in the study by Tauber et al. [
29]. Even if we tried to analyze and reduce the heterogeneity through prespecified subgroup analyses, many precautions still must be taken before our findings can be generalized to every MTBI patient on antiplatelet therapy (refer to Tables
3,
4 and to table B in Additional file
2 for details on heterogeneity).
Our systematic review has some limitations that must be addressed. First, the original studies were very biased, so our data should be interpreted with caution. We rated most of the original studies as “low quality.” Most of the biases were detected in the “flow and timing” section of the adapted QUADAS-2 quality assessment tool. Indeed, most of the studies had a retrospective design; thus, a common systematic error was to not report the MTBI patients with a negative first CT scan who did not undergo a repeated head CT scan before discharge in the flow diagram of the original study [
11,
17‐
20,
23,
24,
27,
28,
31]. It is likely that this bias could have increased the proportion of patients with DB in the original studies, since those patients who were deemed to be at low risk did not have their second CT scan performed due to medical decisions. Nevertheless, retrospective studies enroll patients according to local MTBI management protocols, and they are closer to everyday clinical practice. Second, another common source of bias was the way of outcome assessment by means of a clinical follow-up: some of the original studies had a loose or poor quality follow-up, and some had lost patients at follow-up [19, 21, 22,]. To face this issue, we performed a sensitivity analysis considering all patients lost at follow-up or all unexplained deaths as events. Even in this worst-case scenario, we detected a mean DB incidence as low as 1.7% (95% CI 0.93–2.67%). Third, some studies had an unclear concern for applicability to the review question because the rate of patients with non-mild TBI was not specified in the original studies [
18‐
21,
23,
31]. Indeed, we preferred to increase the sensitivity and overestimate the rate of DB rather than underestimate it; therefore, some non-mild TBI patients were enrolled.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.