Literature Search
The literature search yielded a total of 14,960 records. An additional 485 articles were retrieved via soft searching. After removal of duplicates, screening of the titles and abstracts was conducted on 8882 articles. Of these, 103 studies were identified as likely meeting inclusion criteria and were subject to full review. Of these, 76 failed to meet inclusion criteria, leaving 27 studies included in the narrative review. Given the heterogeneity in study design and reporting of outcomes, it was not possible to conduct a meta-analysis.
Description of Included Studies
Tables 2 and 3 summarise the key characteristics and main outcomes of the 27 included studies. For the purposes of reporting and analysis, studies were grouped according to major mental health constructs. In total, 37.0% (n = 10) of included studies reported on more than one mental health domain (i.e. anxiety and depression). Depression was the most frequently assessed mental health outcome, reported in 70.4% (n = 19) of the included studies, typically using self-report rating scales, including the Beck Depression Inventory 2 (BDI-II) (n = 7), Center for Epidemiologic Studies Depression Scale (CES-D) (n = 4), the Patient Health Questionnaire (PHQ-9) (n = 4) and the Hamilton Depression Rating Scale (n = 4). The next most frequently mentioned outcome was anxiety (25.9%, n = 7), followed by impulsivity (7.4%, n = 2) and attention deficit hyperactivity disorder (ADHD) (3.7%, n = 1). Five studies (18.5%) utilised multi-domain outcome measures. These studies used either the Profile of Mood States (POMS) (n = 3), the Patient-Reported Outcomes Measurement Information System (PROMIS) (n = 1), or the Veterans RAND 12-Item Health Survey (VR-12) (n = 1). Single studies reported on apathy (3.7%) and aggression (3.7%). Highlighting the rapid and growing interest in the field, most (67%) of the included studies were published in 2014 or later (i.e. ≤ 2012, n = 6; 2013, n = 3; 2014, n = 6; 2015, n = 6; 2016, n = 4, 2017, n = 2).
Table 2 Summary of included studies Table 3 Summary of findings
With the exception of one study conducted in France, all other studies were from North America (USA 85.2%, n = 23 studies; Canada 11.1%, n = 3). Almost half the included studies (44.4%, n = 12) reported on data from male-only samples. A variety of study designs were utilised, including cross-sectional, with and without a comparison group (55.6%, n = 15), prospective cohort (25.9%, n = 7), mixed longitudinal and cross-sectional (7.4%, n = 2), prospective case-control (7.4%, n = 2), longitudinal (7.4%, n = 2) and cohort studies (3.7%, n = 1). There were no randomised controlled trials within the included studies. The included studies examined athletes from a broad range of individual sports (e.g. swimming, diving, wrestling, boxing) and team-based sports (e.g. American Football, hockey, soccer, rugby), with many studies including elite athletes from a range of sports. The majority of studies were conducted with North American college athletes (66.7%, n = 18). The sample size of the included studies ranged from 21 to 3656. The follow-up period for the prospective and longitudinal studies ranged markedly from 2 days to 29 years.
Quality Appraisal
All 27 studies included in the review were assessed for methodological rigour using the Quality Assessment Tool for Observational Cohort and Cross Sectional Studies published by the US National Institutes of Health (2014) [42]. The tool comprises 14 criteria (see Electronic Supplementary Material Table S1). An additional criterion was added to determine whether the study included a statement regarding the sample being free of the outcome of interest (in this instance a mental health outcome) at exposure. A final score was calculated for each study as the percentage of criteria met. Percentages were categorised according to previously published guidelines [43, 44]. Each study was rated as excellent (75–100%; very low risk of bias), good (50–74%; most methodological criteria met, low risk of bias), fair (25–50%; some criteria met, possible risk of bias), or poor (0–25%; few criteria met, high risk of bias).
The quality of the included studies varied, with final scores ranging between 28% and 77%. The mean quality rating was 53.8%. In total, one study was rated as excellent (3.7%), 15 studies were rated as good (55.6%), 11 studies were rated as fair (40.7%), and no studies were rated as poor. All included studies reported clear research objectives and most reported clearly defined population (88.9%, n = 24) and mental health outcomes (96.3%, n = 26). However, the majority of studies included in the analysis did not measure or report on concussion history prior to the onset of any mental health outcomes, such as previous concussions that may have been experienced during childhood or adolescence (92.6%, n = 25), and only one fifth of studies (n = 6) reported that the sample was assessed as free of the mental health outcome at the time of exposure (concussion). Inference of a temporal association was not possible in half of the studies (51.9%, n = 14), and concussion was assessed at more than one time point in only one of the included studies (3.7%).
Main Findings
Study outcome variables and outcomes measures are summarised in Table 3. Findings are discussed below relative to mental health outcome and athlete population. Outcomes are reported separately for depression, anxiety, multi-domain symptom assessment, ADHD, and finally, impulsivity, apathy and aggression.
Depression
As indicated in Sect. 3.2, depression symptoms were the most frequently assessed mental health outcome of the included studies. The findings for depression symptoms are organised below according to (a) studies reporting on retired athlete populations, followed by (b) studies reporting on currently competing athletes.
Of the 19 studies reporting depression symptoms as an outcome, almost half (47.4%, n = 9) involved samples of retired athletes. The majority of these (n = 6 studies) involved retired American National Football League (NFL) players. In these studies, the reported rates of depression symptoms varied markedly relative to concussion exposure. For example, in two small studies that used self-report measures, rates of depression meeting clinical cut-off were 33% (total n = 45) [16] and 19.2% (total n = 26) [39], whereas in larger samples that assessed rates of previous clinician diagnosis of depression, the rates were 11.1% (total n = 2434) [20] and 10.2% (total n = 1044) [25] (community 12-month prevalence of diagnosed depression approximates 7.5% [45]). Of the included studies, Hart et al. [21] was the only one to use formal mental health diagnosis (via a clinical interview), with 24% (n = 8) meeting clinical diagnostic criteria (total n = 34). Another smaller study reported a dose-response relationship between cumulative head impact and later-life risk for depression symptoms (total n = 76) [33], with this response also observed in Didehbani et al.’s study [19], which reported a significant moderate correlation (r = 0.43) between number of concussions and self-reported depression (total n = 59).
This dose response between number of concussions and depression symptoms was also reported in the remaining two studies with retired (non-NFL) athletes. Self-reported depression severity increased with the number of reported concussions and was higher in those who had played rugby relative to other sports (total n = 377) [18]. In a large, mixed sports sample (involving basketball, baseball, wrestling), those reporting three or more concussions were 2.4 times more likely than those reporting no concussion to experience moderate-severe depression (total n = 797) [24].
The remaining ten studies were conducted in athletes in current competition. Of these, over half (60%; n = 6) included a baseline (i.e. pre-concussion) assessment, enabling a comparison of relative pre-morbid symptoms prior to recent concussion exposure, and four included a non-concussed control comparison. One study (n = 44) found that relative to healthy controls, concussed American college footballers exhibited partial, but not full, recovery of clinically assessed mood symptoms by 13 and 44 days [30]. Another study (total n = 263) in a mixed sports sample of athletes found no interaction between group (i.e. concussion, healthy control) and time of assessment (baseline, follow-up); however, this study included a long-term follow-up period (mean 283 days) between concussion exposure and depression symptoms [36]. The third study (total n = 84), also in a mixed sports sample of athletes, found that compared to baseline (11%), concussed athletes were significantly more likely to experience threshold symptoms of depression at follow-up (23%), while depression rates did not differ from baseline to follow-up for the control group [40]. Another study in a mixed sports sample (total n = 67) found that concussed athletes who reported symptoms of depression at baseline (e.g. pre-injury) were 4.59 times more likely to experience post-concussion depression symptoms than concussed athletes not reporting depression symptoms at baseline [41]. With the exception of the study employing the longer-term follow-up [36], relative to healthy controls, there is some evidence for elevated symptoms of depression post-concussion, although this may be confined to the short–medium term only, and may be influence by pre-morbid depression. The two small studies (both n = 21) reporting depression symptoms without a control comparison (both in mixed sports samples) indicated findings consistent with the controlled studies above. In one study, concussed athletes reported significantly elevated depression scores relative to their baseline, at 2, 7 and 14 days post-concussion [26], with the other study reporting a significant difference between baseline and assessment at week 1 post-concussion, but not at 1 or 3 months [37].
The remaining four studies reporting post-concussion depression symptoms in currently competing athletes did not include a baseline assessment, though two included a comparison with healthy controls. These two controlled studies reported similar results to the studies described above, where concussed athletes (mixed sports sample) showed improvement in mood symptoms at each time point, but had significantly higher (i.e. worse) mood scores than healthy athletes at 1 week and 1 month post-concussion (total n = 94) [32]; this mirrored the results for the small sample of American college footballers at 3 days, 1 week, and 1 month post-concussion (total n = 36) [38]. The remaining two studies reporting depression symptoms in currently competing athletes found that in a mixed sample of professional (NFL) and semi-professional American footballers, those with a history of three or more concussions reported significantly higher depression symptoms than those with two or fewer concussions (total n = 27) [35]. A separate sample of mixed sports athletes demonstrated a tendency for under-reporting of post-concussive mood symptoms when symptoms reported to coaching staff were compared to those reported on a confidential standardised depression scale (total n = 40) [31].
Anxiety
There was also variability in the outcomes for anxiety symptoms. Seven studies reported anxiety symptoms as an outcome, all of which were conducted in currently competing athletes. Of the three studies that included a baseline assessment, two included a healthy control comparison. One of these studies (n = 44) found significantly higher anxiety symptoms (assessed by structured clinical interview) in post-concussed American college footballers at 13 days (but not at 44 days) relative to healthy controls [30]. Consistent with this, the other study with a mixed sports sample (n = 263) with a longer-term follow-up (283 days) reported no significant interaction for anxiety symptoms between time of assessment and concussed versus healthy controls [36]. The third of these studies reported that in a mixed sports sample (n = 67), those with depression at baseline were 3.4 times more likely to experience state anxiety symptoms post-concussion than athletes without depression at baseline [41]. Two studies included healthy controls, but not a baseline assessment. These studies reported that relative to healthy controls, concussed American college footballers reported significantly higher anxiety symptoms at 3 days and at 1 week, though not at 1 month (n = 36) [38], and that in a mixed sports sample, concussed athletes reported higher anxiety scores at 1 day, 1 week and 1 month (n = 94) [32]. A further anxiety study in a mixed sports sample found equivalent rates of anxiety symptoms in athletes with recent concussion and orthopaedic injury and little difference between the groups in sources of social support sought during injury (n = 126) [17]. The final study (in a mixed sports sample) found that post-concussion, athletes under-reported anxiety symptoms to coaching staff relative to self-reported anxiety (using a standardised rating scale) (n = 40) [31].
Multi-domain Measures
Five studies reported on broad multiple domain outcomes for mental health. Four studies used the POMS, of which two were conducted in current athletes, and included a baseline assessment prior to recent concussion exposure. One of these studies also included a healthy control comparison, and reported that in a mixed sports sample (n = 51), athletes with either concussion or anterior cruciate ligament (ACL) injury reported significant increases in depression symptom scores post-injury compared with un-injured controls, though athletes with ACL injury reported higher levels of depression for a longer duration than athletes with concussion [27]. In a larger sample of mixed sports athletes (n = 369), a significant acute increase in depression symptoms, confusion, and total mood disturbance was seen at 4 days post-concussion, although these increases were transient and resolved by 14 days [28]. In the third study of mixed sports athletes (n = 53), concussion was associated with a short-term (2-week) profile characterised by elevated fatigue and decreased vigour, and short-term mood disturbance, whereas musculoskeletal injury was associated with anger [22]. The remaining two studies were cross-sectional, reporting multi-domain outcomes in samples of retired athletes from mixed sports. Respondents with a history of concussion self-reported worse outcomes on several measures, including positive affect, negative consequences of alcohol use, sleep disturbance, emotional and behavioural dyscontrol and fatigue (n = 3656) [29]. Similarly, those without a concussion history reported better composite mental health scores relative to those with either one to two concussions or three or more concussions (n = 797) [23].
Attention Deficit Hyperactivity Disorder
One study reporting on ADHD was identified. This study found that athletes with a self-reported history of concussion scored significantly higher on an ADHD checklist relative to athletes without a history of concussion (total n = 42) [34].
Impulsivity, Aggression, Apathy
The remaining mental health outcomes that were assessed in the included studies were impulsivity (n = 2), aggression (n = 1) and apathy (n = 1). In the studies reporting impulsivity, currently competing boxers (mean concussions = 0.4) and mixed martial arts athletes (mean concussions = 0.6) reported lower impulsiveness and self-control scores relative to healthy male controls (n = 131) [15], whereas a sample of retired athletes from mixed sports with two or more reported concussions had significantly higher mean impulsivity scores compared to those reporting no concussions (n = 797) [23]. This same study also found that those reporting three or more concussions had a significantly higher mean score for aggression, compared to those reporting no concussions [23]. The study reporting apathy was conducted in retired American college footballers and found that risk of higher apathy scores increased with cumulative head impacts (total n = 76) [33].
Control Groups Used
In total, ten of the included studies drew comparisons with a participant control group. In one study, the control condition included orthopaedic injury, and was matched according to sex, sport, and time lost due to injury [17]. In the remaining nine studies, the control condition was broadly characterised as ‘healthy’. Of these nine studies, three reported comparison data from a physically active uninjured group of similar age peers [22, 28, 36]. The remaining six reported some level of specific matching controls (i.e. age, sex, education) [15, 18, 19, 30, 32, 38]. One of these studies also matched on estimated IQ [19]. Of the seven longitudinal studies using a control group, four reported baseline comparison only (i.e. no control group longitudinal data reported) [15, 30, 32, 38]. For the remaining three studies, there was no observable change in mood symptoms over time for uninjured athletes [22, 28, 36].