Neuropsychological and clinical measures.
A compilation of all major neuropsychological and clinical indices is presented in . Results indicated an increase in emotional (t1,38 = −3.11; p < 0.05; mTBI > HC), cognitive (t1,38 = −4.20; p < 0.001), and somatic (t1,38 = −3.62; p < 0.005) complaints for mTBI patients compared to controls. Estimates of premorbid intellectual functioning were lower in mTBI patients (t1,37 = 2.09; p < 0.05) despite educational matching.
Table 2 Demographic and clinical measures for visit 1
A multivariate analysis of covariance (MANCOVA) examining differences in neuropsychological testing using premorbid intelligence as a covariate was not significant for group differences. However, effect sizes () in the domains of attention, executive functioning, and memory were of similar magnitude to those reported in recent meta-analyses on cognitive deficits in mTBI.
Structural imaging data.
Anatomic images were limited to T1- and T2-weighted images. These were found to be free of pathology for both groups of subjects by a board-certified neuroradiologist (i.e., all mTBI patients were classified as being noncomplicated).
Three MANCOVAs were conducted to examine group differences (mTBI patients vs matched controls) in FA values within the corpus callosum and left and right hemisphere ROI () with estimates of premorbid intellectual functioning as a covariate. Results indicated a multivariate effect of group for both the CC (F3,36 = 3.81; p < 0.05) and the left (F5,34 = 2.70; p < 0.05) but not right (p > 0.10) hemisphere. Follow-up univariate tests indicated that mTBI patients had higher FA within the genu (F1,38 = 7.52; p < 0.01, d = −0.91), left SCR (F1,38 = 5.54; p < 0.05, d = −0.77), left CR (F1,38 = 5.47; p < 0.05, d = −0.74), and left UF (F1,38 = 6.67; p < 0.05, d = −0.84). Trends were observed for the left IC (F1,38 = 3.69; p = 0.062, d = −0.62) and the splenium (F1,38 = 2.95; p = 0.094, d = −0.53) with mTBI patients again exhibiting higher FA values than HC (see figure e-1 for normalized FA histograms).
Figure 1 Fractional anisotropy (FA) values from all regions of interest (ROI)
HC were then compared with a larger normative sample. However, there were no multivariate effects of group for all multivariate analyses of variance (p > 0.10), suggesting that our control group was statistically similar to the larger normative sample in terms of FA.
Next, we compared axial diffusivity and RD values for the 6 ROI that exhibited significant or trend differences in FA using one-way analyses of covariance (). There were no significant differences between patients and controls in terms of axial diffusivity. In contrast, RD was lower in mTBI patients within the genu (F1,38 = 5.09; p < 0.05, d = 0.74), the left UF (F1,38 = 5.67; p < 0.05, d = 0.77), and the left CR (F1,38 = 4.42; p < 0.05, d = 0.66), with trends present in the left SCR (F1,38 = 3.58; p = 0.06, d = 0.59) and left IC (F1,38 =3.99; p = 0.053, d = 0.66). Histograms for the normalized RD data are presented in figure e-2.
Finally, a MANCOVA () comparing variability in FA measurements between right and left hemisphere homologue ROI (SFL, IC, UF, SCR, and CR) revealed a group effect (F5,34 = 4.53; p < 0.005), with univariate tests indicating increased variability in patients compared to controls for the SCR (F1,38 = 15.06; p < 0.001, d = −1.21), with a trend for the UF (F1,38 = 3.82, p = 0.058; d = −0.63).
Figure 2 Variability in mean fractional anisotropy (FA) between right and left hemisphere regions of interest (ROI)
DTI and clinical measures.
Hierarchical multiple regressions were performed on the 6 clinical measures with the largest effect sizes (attention, memory, executive functions, cognitive complaints, somatic complaints, and emotional complaints) using FA from the CC and right and left hemisphere ROI as the independent variables and premorbid intelligence as a covariate. Although premorbid intelligence accounted for significant variance in terms of both attentional and executive functioning, only FA levels in the right hemisphere (F2,18 = 6.84; p < 0.01) predicted variance in attentional deficits (positive relationship) for the mTBI group.
Next we determined which of our objective measures of deficits (FA or neuropsychological testing) would more accurately classify mTBI patients and HC using binary logistical regression. Estimates of premorbid intelligence were entered into both models as it discriminated (Wald = 4.16; p < 0.05) between HC (65% accuracy) and mTBI patients (66.7%) at slightly above chance levels. Traditional neuropsychological measures (attention, memory, and executive function) did not significantly improve classification accuracy in the first model (HC = 60%; mTBI = 71.4%). In contrast, results from the second model indicated that both the left (Wald = 7.73; p < 0.05) and right (Wald = 5.66; p < 0.05) hemisphere FA indices improved classification accuracy (HC = 70%; mTBI = 81%), with a trend being noted for the CC (Wald = 3.59; p = 0.059). A support vector machine analysis with the leave-one-out methodology confirmed the generality (HC = 65%; mTBI = 81%) of the classification findings.
Visit 2 data.
To date, 10 out of 17 (59%) eligible mTBI patients and 15 out of 16 (94%) eligible HC participants have returned for their 3- to 5-month follow-up visit (see appendix e-1). Intraclass correlation coefficient values for FA were highly reliable (all ROI = 0.65 < r >.93; p < 0.01) in the HC sample; however, reliability of homologue measures was much more variable (SCR r14 = 0.64, p < 0.01; SLF r14 = 0.81, p < 0.001; UF r14 = 0.22, p > 0.10; CR r14 = 0.71, p < 0.01; IC r14 = −0.26, p > 0.10).
There were no significant differences for all clinical measures for mTBI patients who returned and those who did not. Additionally, there were no significant differences in FA values between the 2 groups across the 3 sets of ROI (CC, right and left hemisphere).
Change scores in clinical measures were calculated (visit 2 − visit 1 data) for those measurements that were most suggestive (i.e., based on significance or effect sizes) of group differences at visit 1 (attention, memory, executive functions, emotional distress, somatic and cognitive complaints) using premorbid intelligence as a covariate. Although there were no significant group effects, effect sizes suggested that memory scores improved (d = −0.52) and cognitive complaints decreased (d = 0.79) for the returning mTBI group compared to their matched controls at visit 2.
Differences in visit 1 and 2 FA and RD measurements were compared separately across the 2 groups with paired samples t tests to maximize power (see ). Tests were again limited to those ROI that exhibited significant or trend differences in mean FA and RD (genu, splenium, left SCR, left IC, left UF, and left CR) at visit 1. In HC, there were no significant differences for either FA or RD across the 2 visits. In contrast, partial normalization (i.e., decrease) in FA values was evident in the splenium (t8 = 4.17, p < 0.005) and CR (t8 = 1.89, p = 0.09) at visit 2 for mTBI patients. Although none of the RD effects reached statistical levels of significance, visual examination of the data suggests that RD differences may have partially normalized at visit 2 as well.
Figure 3 Fractional anisotropy (FA) and radial diffusivity (RD) values at both visits