A systematic screening approach in fulfillment of the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) and the Revised Assessment of Multiple Systematic Reviews (R-AMSTAR) guidelines was implemented (26). Potentially eligible studies were identified by systematically searching the databases PUBMED, EMBASE and MEDLINE. The searches were limited to literature published from January 2014 to December 12, 2021. Studies published prior to January 2014 were discussed in our previous study (18). The search strategy was developed using combinations of the following MeSH terms: (“mild traumatic brain injur*” OR “concussion”) AND (“neuroinflammat*” OR “cytokine*”). A secondary manual search, using Google Scholar, was also conducted to ensure that all relevant articles were captured.

Citations were uploaded into Covidence for title, abstract, and full-text screening as well as duplicate removal. Two independent reviewers conducted the study screening in duplicate, from title to full-text screening stages (SM and OA). Disagreements regarding article inclusion were settled by mutual consensus after discussing the disputed articles together. Any further discrepancies were discussed with other team members and eventually resolved by the principal investigator (TG and MR).

Only studies providing information on inflammatory cytokines in the CSF, blood, plasma, or serum of the patients with mTBI were considered for review. The research question and inclusion/exclusion criteria were established a priori. Inclusion criteria were defined as: (1) concussions or mild traumatic brain injuries, (2) neuroinflammation, (3) inflammatory cytokines, and (4) articles published in English. Exclusion criteria were defined as: (1) complicated mild and more severe forms of traumatic brain injuries with Glasgow Coma Scale (GCS) < 13, (2) no blood or CSF cytokines, (3) no comparison healthy or trauma controls or baseline (pre-mTBI) groups, (4) review articles, abstracts or letter to editors, (5) cadaver/non-human studies, and (6) articles that included TBIs but did not make distinctions between the various types of TBIs. The reference lists of related studies were also searched for additional reports.

Two independent reviewers (SM and MM) abstracted the relevant data from the included articles on an Excel sheet. The characteristics extracted from each study included the author, publication year, study design, sample size, mTBI setting, mTBI diagnostic criteria, control type, and patient demographics (e.g., age, sex, etc.). Details about the number of previous mTBIs, time since last mTBI, and GCS data were also recorded. Furthermore, data regarding methods of cytokine measurement, type of biospecimen analyzed, the time interval between injury and cytokine measurement, cytokine levels (both mTBI and control groups), along with any relevant p-values, were recorded. In addition, any acute or chronic functional outcomes associated with a particular cytokine were noted. This included the presence of persistent symptoms, reduced or lack of return to normal activities (work, school, and sports), abnormal neurocognitive function, and Glasgow Outcome Scores (GOSE).

To conduct the meta-analysis, the mean cytokine concentrations, and their standard deviations (SD) for case and control groups were extracted at each follow-up visit. The timing of cytokine concentration measurement varied from <24 h to >1 month. Other descriptive statistics such as medians and measures of variance (e.g., 95% confidence intervals [CI] and range) were also extracted. Efforts were made to contact the authors of studies that presented their data as either medians, or in a log-transformed format only, to ensure that all possible mean values were available to conduct a thorough meta-analysis. Data from studies that had overlapping populations was extracted but a distinction was made in the analysis.

Risk of bias and study quality were evaluated using the Newcastle-Ottawa Quality Assessment Scale (NOS) (27). A score of 0 or 1 was given for each category/criterion on the NOS scale, where the maximum possible score of 8/8 could be achieved (maximum score of 1 for each category). The total scores were categorized according to the methodological quality of each study. Potential confounding factors (including type of biospecimen, assay type, time since mTBI, type of mTBI population and control for confounding inflammatory variables etc.) were also considered for a more detailed bias and quality analysis of studies.

Review Manager (RevMan) Version 5.4 (The Cochrane Collaboration, 2020) was used for the meta-analysis. Meta-analyses were conducted whenever the mean values of an inflammatory cytokine were available in at least three or more studies, with a minimum of 30 participants in each study. For the studies that did not report SD, it was calculated from SEM (Standard error of mean) and CI 95% (Confidence Interval 95%) using the following formulas:

Due to a limited number of studies that qualified for the meta-analysis, only four analyses were conducted comparing mean TNF-α, IL-6, IL-1β, and MCP-1/CCL2 levels in serum/plasma/blood between patients with mTBI and healthy control groups. A high level of heterogeneity was expected due to the utilization of different assay methods (i.e., Multiplex and ELISA), time of cytokine measurement, control for inflammatory variables, and different blood fractions and dilutions used. To account for this heterogeneity, a random effects model and inverse variance approach were utilized to estimate the pooled standardized (Std.) mean differences, their corresponding 95% confidence interval, and p-values. The Std. mean difference is used when the included studies measure the same outcome in different ways. It standardizes the differences in the measurement of the same outcome before pooling the means. It does not however remove the heterogeneity among the study population. Random effects models are preferable if significant heterogeneity is expected as this model accounts for both within-study variability and between-study variability. Heterogeneity was tested using Cochrane’s Q test with the p-value set at 0.1 for significance and quantified using the I² statistic (I² > 40% as low, 40–60% as moderate, and > 60% as substantial heterogeneity). The sources of heterogeneity were evaluated and the risk of bias across studies, publication bias, and selective reporting were assessed. Sensitivity analyses were conducted by excluding the studies in which mean and standard deviation were estimated (from reported standard deviations and range values) to assess the consistency of the estimated mean differences. If multiple studies conducted by the same group had overlapping populations, only the most recent study with the largest mTBI population size was included in the meta-analysis. If cytokines were measured at different time points, the time-point with the largest population size was used for the total cytokine analysis. For acute and chronic cytokine analysis, the most time-appropriate data consistent with the timings of the other studies was used.

A total of 5,138 studies were yielded across the three databases Embase, 139 from Medline, 662 articles from PubMed, and 7 from other sources. After removing duplicates, a systematic screening process was conducted as shown in Figure 1, yielding a total of 26 articles that met the selection criteria (Figure 1). Out of the 26 studies, 25 compared blood cytokine levels between patients with mTBI and healthy controls and 3 studies compared the levels between patients with mTBI and trauma controls (some studies had both control types). Only one study compared cytokine level differences in the CSF. This CSF data was not included in the qualitative or quantitative analysis, but the results are available in Table 1. The characteristics of the 26 included studies are described in Table 1.

A total of 3,248 participants were included across all studies, where 1,746 of these patients had at least one mTBI. There were 1,502 controls, 1,431 of which were healthy controls, and 71 were trauma controls. The mean sample size of patients with at least one mTBI was 67.15 ± 50.11 whereas, the mean sample size for controls was 57.24 ± 39.02. The mean age of patients with at least one mTBI was 27.37 years, and the mean age of control patients was 27.14 years. 74.5% of mTBI patients were male. Sport-related injuries were the most common source of mTBI in the majority of included studies (42%), followed by general trauma (31%), and military-related injuries (27%). The diagnostic criteria used for mTBI diagnosis was heterogenous.

Most studies in this review have a level of evidence of IV (N = 14; 53.8%). There was substantial agreement between the two reviewers at the title/abstract screening stage (κ = 0.80 [95%CI, 0.70–0.90]) and the full-text screening stage (κ = 0.79 [95%CI, 0.60–0.90]). The mean NOS score for the included studies was 6.38 ± 1.19, which indicates a fair quality of evidence for non-randomized studies. The areas of best performance based on the NOS checklist were the case definition (N = 25; 96%) and the definition of controls (N = 25; 96%). The area of worst performance was the non-response rate (the number of patients that were lost to follow-up), which was not provided in any of the included studies.

The most common blood inflammatory cytokines assessed in the included studies were IL-6, TNF-α, IL-10, IL-1β, Interleukin-8 (IL-8), IFN-γ, Interleukin-1 Receptor Antagonist (IL-1RA), Interleukin 4 (IL-4), and MCP-1/CCL2 (Figure 2). It should be noted that MCP-1 is also referred to as CCL-2 and both these terms are used interchangeably. Most of the included studies extracted peripheral inflammatory cytokine specimens from plasma (46%, n = 12). The remaining studies extracted inflammatory cytokines from either serum (38%, n = 10) or whole blood (15%, n = 4). About 42% of the studies (n = 11), assessed cytokine levels within 24 hrs of injury. However, most studies (53.8%, n = 14) assessed cytokine levels 30 days or later following a mTBI. The systematic review found elevated levels of IL-6 (time points: <24 h, 1–7 days and ≥ 30 days), TNF-α (≥30 days), IL-1β (1–7 days), IL-8 (time points: <24 h and ≥ 30 days), IFN-γ (<24 h), IL-1RA (<24 h), and MCP-1/CCL2 (time points: 1–7 days and ≥ 30 days) in patients with mTBI, compared with healthy controls, where any significant findings were replicated in at least two studies. The evidence was particularly strong for IL-6, IFN-γ, IL-1RA levels (at <24 h), and MCP-1/CCL2 (between 1 and 7 days), where ≥60% of the studies found significant elevated levels in patients with mTBI compared to healthy controls at these time points.

There are four groups of research articles that had subject overlap, where the participant pools were utilized multiple times by researchers in the same group. These groups are labeled as Groups A, B, C, and D. Group A published four of the included studies (28–31), Group B published two studies (32, 33), Group C published four studies (34–37), and finally Group D published three articles (38–40). For this review, clear distinctions were made if two or more studies had overlapping populations at a certain time point. This is to avoid any potential impact on the statistical analysis and the results of this review, caused by falsely giving undue weight to a certain study population.

Eleven studies (12 cohorts) involving 987 participants with mTBI were utilized to conduct the meta-analyses for IL-6, TNF-α, IL-1β, and MCP-1/CCL2 levels in the blood. The results show significantly higher circulating levels of IL-6, IL-1β, and MCP-1/CCL2 in the mTBI population in the acute stages (within 1 week), compared to healthy controls. No differences were observed for any inflammatory cytokine in the chronic stages.

Only three studies compared inflammatory cytokine levels in the blood between the patients with mTBI and trauma controls (38, 47, 53).

Shan et al. found no significant differences in the TNF-α, IL-1β, and the chemokines (CXCL1, CXCL8, and MCP-1/CCL2) levels in acute stages (<24 h) between the two groups (47). Vedantam et al. observed significantly elevated IL-2 and IL-6 levels in patients with mTBI in the acute stages (<24 h); however, in the chronic stages only IL-6 levels remained elevated (53). Battista et al. found that athletes with sport-related concussion had higher levels of the chemokines’ monocyte chemoattractant protein-4 (MCP-4) (p < 0.001) and macrophage inflammatory protein-1β (MIP-1β) (p = 0.001) compared to healthy athletes (within 1 week of injury). At medical clearance, there were no significant biomarker contributions towards the class separation between athletes with SRC vs. healthy athletes (39).

The relationship between inflammatory cytokines in the blood and mTBI prognosis was analyzed in 13 studies (Table 1).

For this analysis, the population was considered to have poor functional outcomes if they had any of the following conditions:

Our findings must be interpreted with caution.

First, this review shows that there is considerable heterogeneity in the data, leading to difficulties in pooling and analyzing the data to formulate a meaningful conclusion. Heterogeneity was caused by many reasons, some of which include differences in the time elapsed between the initial mTBI and blood sample collection, cytokine analysis technique, blood fraction used for analysis, confounding variable control, mTBI diagnostic criteria, data reporting and functional outcomes measured. Hence, there is a need for a standardized approach in acquiring and reporting data to allow for comparisons.

Secondly, the results of this review show that about 76% of mTBI patients were male. This is because most studies are conducted in the military (27%) and sports populations (42%), which happen to be male-dominant settings. Since females are more at-risk for poor recovery and develop persistent symptoms more frequently compared to males (65–67), we could not assess prognosis accurately based on this data. Future studies, especially those assessing prognosis in mTBI patients, may want to incorporate more female participants in their studies.

Thirdly, the cytokine alterations observed in the mTBI population do not necessarily reflect a pathophysiology associated with head injury alone. There are other variables that should be considered while measuring cytokine levels as they are known to cause considerable fluctuations. These include time of blood collection, sex differences, other injuries (orthopedic, whiplash, muscle strains, etc.) at the time of mTBI, acute and chronic illnesses, co-existing psychiatric conditions, and medication intake amongst others. While we attempted to take some of these limitations into account, the results should still be interpreted with caution due to the factors mentioned above.

Lastly, most studies reported their results using medians or log-transformed means. Although this way of reporting leads to more consistent results, a thorough meta-analysis cannot be conducted using medians as we have found in this review. Future studies should follow a more standardized methodology so that the data reported using means and SD is not as heterogeneous and is more consistent to allow for a more thorough analysis.

This systematic review and meta-analysis has several strengths. An exhaustive effort was made to capture the data by searching a variety of databases and acquiring the missing data by directly contacting the authors, extracting data from graphs and tables, or using standardized estimation methods for calculating the mean (SD). Although the latter two strategies may not be accurate, they provide estimates that are closer to the real value and are frequently employed in meta-analyses.

A systematic review with a similar aim was recently published (68). However, our study stands out because it included 15 additional studies. Hence, due to the availability of more data, we were able to conduct a meta-analysis that has not been accomplished before. Visser et al., just like our study, was able to identify IL-6 as a promising biomarker for brain injury (68). It was, however, unable to establish an association between other cytokines (such as MCP-1/CCL2 and IL-1β) and mTBI as we did.

Overall, we found substantial evidence of increased inflammatory cytokine levels in patients with mTBI. The evidence was particularly strong for IL-6, IL-1β, and MCP-1/CCL2. The results of this study were however limited by low study numbers as well as methodological heterogeneity between the studies.