A comprehensive literature search was conducted on AMED, Embase, MEDLINE, Ovid, PubMed, Scopus, and Web of Science for all relevant articles reporting on the use of DTI in subjects who developed PCS post-mTBI, through 20 May, 2016. The databases were searched with the following search phrase using the Boolean logic operators “OR” and “AND”: “(DTI OR diffusion tensor imaging OR diffusion tractography) AND (mTBI OR mild traumatic brain injury OR concussion) AND (postconcussive syndrome OR post-concussive syndrome OR post concussive syndrome OR postconcussion syndrome OR post-concussion syndrome OR post concussion syndrome) AND human.” To ensure maximal article capture, these search terms also encompassed the following Medical Subject Headings (MeSH) terms: “diffusion tensor imaging,” “brain injuries,” and “post-concussion syndrome.” Manual searching of relevant journals and reference lists of studies found in the above search provided additional articles.

The search terms identified above yielded 205 studies. For this review, the PCS population was defined as patients who experience persistent symptoms for 3 months or longer post-injury. Inclusion criteria were studies published in English; use of human participants; and studies that analyzed changes in measured DTI parameters in patients with PCS following an mTBI. Exclusion criteria were studies that did not report method of diagnosing PCS; studies that did not report method of diagnosing mTBI; studies that were not original research; and studies with fewer than six participants. Two readers independently screened all 205 studies, removing 79 duplicates, 16 conference abstracts, and 1 foreign language paper. Six additional studies were identified from relevant reference lists and journals. Of the 115 studies remaining, 92 were excluded for containing one or more exclusion criterion. A further 13 studies were removed after rescreening because they included patients who were assessed for PCS symptoms within 3 months of injury. This resulted in 10 studies that were included in this review. The flow diagram for the paper selection process is presented in Figure 1.

Each study was assessed by two reviewers for quality using STROBE guidelines (41). A point was awarded for each required criterion that was met. The two reviewers performed individual assessments, and scores were compared, with discrepancies resolved by discussion. The two reviewers also performed data extraction for the population studied, patient source, number of PCS patients, patient demographics (age and gender), mechanism of injury, mTBI diagnosis method, PCS diagnosis method, time interval between injury and PCS diagnosis, time interval between injury and imaging, control group characteristics (screening process, number, and matching process), DTI values measured, DTI analysis method [voxel-wise or region of interest (ROI)], significant findings, conclusions (correlations between DTI and PCS), and study limitations for all nine studies. DTI studies most often do not quantify changes in DTI parameters because there are no established values for healthy patients, so only qualitative changes (increases and decreases) in these parameters were extracted for this review. Results of data extraction were compared between the reviewers, with discrepancies resolved with help from a third reader.

Decreased FA and increased MD and RD were reported in PCS patients compared to healthy controls or PCS-absent trauma patients. Abnormal DTI values were also reported to be correlated with an increase in number and severity of PCS symptoms, suggesting that greater axonal damage causes more severe PCS symptoms. Decreased FA is a well-documented finding in brain injury (42–46), as is increased MD (42, 44, 45) and RD (43). In addition, regions within the corpus callosum were most often found to be affected in PCS patients. Two additional articles reporting on case studies also found decreased FA, specifically in PCS patients (47, 48). The corpus callosum is involved in inter-hemispheric integration of motor, sensory, and cognitive information, and damage to this area might lead to extensive behavioral, emotional, and cognitive impairments, as observed in the PCS groups. Although the effects of damage to the corpus callosum are not fully understood, it is reasonable to expect that these effects would include the symptoms of PCS patients.

Studies included in this review assessed different patient populations, including pediatric, adult, and military. Although age- and education-related differences in structural and functional neuroanatomy have been documented (49, 50), all of the studies had a case–control design. Patients were matched to controls by age, gender, and education level, therefore ensuring that changes detected in DTI parameters were not a result of comparisons made between inherently distinct populations. However, consistent findings in FA, MD, and RD across the studies suggest that the microstructural white matter damage detected is consistent in all patient groups. This possibility increases the value of DTI as a universal biomarker of PCS. In addition, 8 of the 10 studies (31–35, 38–40) excluded patients with abnormal CT or routine clinical MRI findings. This exclusion criterion indicates that there is damage identified by DTI, which is not detected via more commonly used modalities, further emphasizing the utility and importance of DTI in the clinical setting. It is often acknowledged that the more widely available structural neuroimaging modalities have limited value in the mTBI and PCS populations (8, 51, 52).

Post-concussion syndrome has been found to be more prevalent in the mTBI population than the moderate or severe TBI populations (19, 20). However, it has been noted that axonal injury in these more severely injured groups usually presents with focal lesions and so is detectable via clinical MRI (53). Therefore, not all axonal injury is indicative of PCS. This evidence suggests that generalized axonal injury does not correlate well with PCS, but diffuse axonal injury (DAI), which presents as widespread axonal injury limited to the microstructure scale, may be more specific to PCS. This is also supported by a recent study using magnetic resonance spectroscopy (MRS), which found significant correlations between DAI and PCS (54). However, PCS may not only be a result of neurological damage but may also develop due to psychological distress (55–57). For this reason, a patient who has not sustained an mTBI may still experience PCS because it is not unique to the mTBI or TBI population. Although patients with DAI may develop PCS, not all PCS patients have DAI.

Two studies (37, 39) that applied a voxel-wise analysis of the whole brain reported no significant findings in the corpus callosum. Maruta et al. (37) reported no significant findings in any brain region but was the only study to have no significant findings in all four DTI parameters in a whole brain analysis. Miller et al. (39) did find a decreased FA in several white matter clusters in the whole brain but did not identify specific brain regions where the clusters were located. Voxel-wise analysis requires intersubject registration of subjects’ brains and normalization to standard atlases. Both of these processes involve smoothing and hence may result in masking small differences between subjects or groups. All studies that used ROI analysis found significant changes in DTI parameters, suggesting that ROI analysis might be more proficient at identifying changes in specific brain regions. However, ROI analysis has some limitations: it requires a brain structure to be predefined for analysis, and manual selection of the region to draw an ROI may lead to intersubject differences in ROI location. Despite these limitations, the success of ROI analysis in the studies reviewed is encouraging; once a particular region has been identified as most commonly damaged in PCS, diagnosis via DTI will be faster than if a voxel-wise approach was necessary.

Although DTI can detect axonal damage and possibly predict PCS onset or be used to diagnose PCS, there is insufficient evidence supporting the observed results to validate any parameter in a specific brain region as a biomarker for PCS. The trends in the parameters are a decrease in FA and an increase in MD and RD, but there is no uniformity in the brain areas investigated for these changes. These findings suggest that there may be more than one DTI biomarker for PCS and that axonal damage does contribute to PCS symptoms. The subjective nature of PCS and the possibility of a large number of brain regions being involved in PCS may have led to the indefinite results. Each patient has a unique illness experience due to their baseline for many of the symptoms, such as fatigue, feelings of depression, feelings of frustration, forgetfulness, poor concentration, and restlessness. Patients have varying pain tolerances and emotional fortitudes that could either magnify or diminish the severity of PCS symptoms being reported. In addition, a researcher may focus on one or more brain regions or on the entire brain, depending on his or her research interests and image processing preferences.

Other imaging modalities have been used to study PCS. In addition to the poor value of CT and MRI in the PCS population, studies have shown conflicting results in positron emission tomography (PET) in the PCS population, with some finding correlations between PET results and PCS (58–60), while others do not (51). A more promising imaging modality may be MRS, which has been used to detect DAI that is significantly correlated with PCS (54), although there is unsubstantial evidence that it is a reliable PCS biomarker. Evoked potential (EP) studies have concluded that significant results in the EP data correlate with PCS (61, 62), but there is not enough concrete evidence to support these measurements as a PCS biomarker. Biochemical markers of PCS may also be viable. A review paper (63) identified S100 proteins, neuron-specific enolase (NSE), and cleaved Tau protein (CTP) as potential serum biochemical markers for predicting PCS after an mTBI. The review concluded that none of the three compounds are well-validated for use, although S100 was most widely studied in the mTBI population and appears to be the most promising. It is possible that a combination of DTI, clinical factors, and biochemical markers may be needed to accurately and objectively diagnose PCS or predict its development after an mTBI.

Further research into this topic is necessary. All of the studies included in this review are cross-sectional. Future studies should consider a longitudinal cohort study design to track changes in DTI parameters during PCS progression and resolution, which would provide more concrete evidence of a specific biomarker for PCS. Recruiting a larger PCS patient population is required to reduce sample size biases. Research on DTI parameters in the brain is also required to establish “normal” values for FA, MD, RD, and AD, so that significant differences are based on a universal standard as opposed to being derived from each study’s controls. This will ensure that significant results are not misrepresented. Future case–control studies should try to use an orthopedic or other non-head trauma control group to help remove error in PCS reporting due to factors other than the sustained mTBI. Case–control studies may also consider comparing PCS-present mTBI patients to PCS-absent mTBI patients to eliminate most external confounding factors. These studies would be more likely to observe changes in DTI parameters specifically due to PCS. The large range of time intervals between injury and imaging reported in the ten reviewed articles is also a concern. In addition, 13 articles were excluded from this review because researchers conducted PCS assessment within 3 months of injury. It is suggested that future studies include patients whose PCS symptoms persist for 3 months or longer, as the literature supports PCS diagnosis when symptoms persist for this length of time. Finally, a standard PCS assessment should be administered to ensure consistency across all future studies. Consulting the Common Data Elements identified by the National Institutes of Health would help establish assessment standards.