Statistical challenges exist when using diffusion tensor imaging (DTI) to assess traumatic axonal injury (TAI) in individual concussed athletes. The authors examined active professional American football players over a 6-year time period to study potential TAI after concussion and assess optimal methods to analyze DTI at the individual level.
Active American professional football players recruited prospectively were assessed with DTI, conventional MRI, and standard clinical workup. Subjects underwent an optional preseason baseline scan and were asked to undergo a scan within 5 days of concussion during gameplay. DTI from 25 age- and sex-matched controls were obtained. Both semiautomated region-of-interest analysis and fully automated tract-based spatial statistics (TBSS) were used to examine DTI at individual and group levels. Statistical differences were assessed comparing individual DTI data to baseline imaging versus a normative database. Group-level comparisons were also performed to determine if longer exposure to professional-level play or prior concussion cause white matter microstructural integrity changes.
Forty-nine active professional football players were recruited into the study. Of the 49 players, 7 were assessed at baseline during the preseason and after acute concussion. An additional 18 players were assessed after acute concussion only. An additional 24 players had only preseason baseline assessments. The results suggest DTI is more sensitive to suspected TAI than conventional MRI, given that 4 players demonstrated decreased fractional anisotropy (FA) in multiple tracts despite normal conventional MRI. Furthermore, the data suggest individual assessment of DTI data using baseline premorbid imaging is more sensitive than typical methods of comparing data to a normative control group. Among all subjects with baseline data, 1 reduced FA tract (± 2.5 standard deviations) was found using the typical normative database reference versus 10 statistically significant (p < 0.05) reduced FA tracts when referencing internal control baseline data. All group-level comparisons were statistically insignificant (p > 0.05).
Baseline premorbid DTI data for individual DTI analysis provides increased statistical sensitivity. Specificity using baseline imaging also increases because numerous potential etiologies for reduced FA may exist prior to a concussion. These data suggest that there is a high potential for false-positive and false-negative assessment of DTI data using typical methods of comparing an individual to normative groups given the variability of FA values in the normal population.