Data reduction and analysis
Index scores for the alerting, orienting, and executive components of the ANT were calculated to assess the efficiency of individual attention networks by subtracting the mean response times between conditions; alerting (no cue–double cue), orienting (central cue–spatial cue), and executive attention (incongruent–congruent). Responses to the BDS test of working memory were scored for both individual trials for each string length and aggregated. For individual string lengths (values specifically pertaining to trials in which string lengths of 4, 5, 6, 7, or 8 digits were presented), proportions of correct responses were determined by scoring the correct responses for each trial and averaged. Each string length contained six trials. Aggregate scores for all phases were calculated by summing the proportion scores from all string lengths of the task.
The data from 44 participants were included in the final analysis after removing three from the initial collection set due to incomplete data. Of the final participants, 21 (47.7%) had an active clinical diagnosis of PTSD while 23 (52.3%) served as controls. Among the final set, 34 participants were male and 10 were female. The female participants were disproportionately distributed between our diagnosis groups, with eight females (34.8%) included in the control group and two (9.5%) included in the PTSD group, χ2 (1, n= 44) = 3.99, P < 0.05, OR = 5.07. The number of participants reporting sleep difficulties in the control group (13%) was disproportionately lower than what was reported by the PTSD group (90%), χ2 (1, n= 44) = 26.33, P < 0.05, OR = 63.33. Eleven PTSD-diagnosed participants reported taking prescribed antidepressant medication (52.4%) as opposed to only two participants from the control group (8.7%), χ2 (1, n= 44) = 10.06, P < 0.05, OR = 11.55. The descriptive statistics for all other independent variables are presented in .
Independent variables: descriptive statistics
A series of t-tests were conducted to explore observed group differences between PTSD-diagnosed participants and the control group with regard to independent variables (). Bonferroni adjustments to the alpha levels were made in order to correct for familywise error rate. As expected, PTSD-diagnosed participants reported experiencing more PTSD symptoms (M = 58.62, SD = 8.95) than did participants in the control group (M = 24.17, SD = 6.10), t (42) = 9.70, P < 0.0071. PTSD-diagnosed participants also reported higher levels of combat exposure (M = 24.76, SD = 9.40) than did participants in the control group (M = 7.09, SD = 7.92), t (42) = 6.77, P < 0.0071. PTSD-diagnosed participants reported higher levels of both depression, t (42) = 9.70, P < 0.0071, (M = 22.71, SD = 7.84) and anxiety (M = 24.43, SD = 9.53) than control participants (M = 4.74, SD = 4.00; M = 2.17, SD = 2.57, respectively), t (42) = 10.79, P < 0.0071. No group differences were observed regarding age, years of education, or alcohol consumption.
Independent variables: comparisons of means between PTSD and control group (t-tests)
A multivariate analysis of variance (MANOVA) was conducted to explore group behavioral differences on BDS scores and alerting, orienting, and executive efficiency index scores of the ANT. Neither the assumption of homoscedasticity nor equal group variances were violated (P
> 0.05). PTSD participants had significantly lower scores (M = 1.41, SD = 0.91) on the BDS task than did controls (M = 2.54, SD = 1.11), F
(1, 42) = 13.35, P
. No other group differences were observed, P
> 0.05 ().
Dependent variables: comparisons of means between PTSD and control group
Working memory: aggregate scores
An analysis of covariance (ANCOVA) was conducted to explore the relationship between PTSD diagnosis and aggregate working memory (BDS) scores, while entering depression, anxiety, and combat exposure as covariates. A preliminary analysis evaluating the homogeneity-of-regression (slopes) assumption indicated that the relationship between the covariates and the dependent variable did not differ significantly as a function of the independent variable, P
> 0.05. After adjusting for covariates, the relationship between PTSD and working memory observed in earlier tests was no longer significant, F
(1, 42) = 1.77, P
. None of the three covariates were significantly related to working memory performance.
A series of mediation analyses were conducted to test if anxiety, depression, and combat exposure scores each served as total or partial mediators of the relationship between PTSD diagnosis and impaired working memory. Using the Freedman and Schatzkin (1992)
test for partial mediation, neither depression, t
(42) = 0.29, P
> 0.05, anxiety, t
(42) = 0.28, P
> 0.05, nor combat exposure, t
(42) = 0.31, P
> 0.05, proved to meet the criteria for partial mediators of the relationship between PTSD diagnosis and working memory scores.
The effects of antidepressant use upon working memory as measured by collapsed BDI scores were examined. Due to the low numbers of participants in the control group who reported taking some form of antidepressant medication at the time of data collection (n= 2), only data from participants diagnosed with PTSD were included for analysis. Among PTSD-diagnosed participants, there was no difference with regard to working memory as measured by collapsed BDS scores between those participants having reported using antidepressant medication (n= 11) at the time of data collection and those participants not reporting the use of antidepressant medication (n= 10), t (19) = 0.65, P > 0.05.
Prior concussions and working memory
To assess possible additive effects of having a prior concussion, individuals in the PTSD group were placed into one of two subgroups based on reporting having a LOC from a head injury: (1) PTSD with a LOC (PTSD + LOC; n= 9) from a prior head injury and (2) PTSD with no LOC (PTSD – LOC; n= 12). BDS scores from both of these subgroups were compared against each other and with the control group. The results indicated that, although, both the PTSD + LOC group and PTSD – LOC group were significantly different from the control group, t (30) = 3.46, P < 0.05; t (33) = 2.41, P < 0.05; respectively; the difference between the PTSD + LOC and PTSD – LOC groups was not significant, t (19) =–1.16, P > 0.05. In as such, the results do not suggest there was a significant contribution of prior concussions on working memory function above and beyond that of PTSD alone. However, one of the limitations is the small sample size of the subgroups. Therefore, future studies should continue to explore the main effects and interactions of PTSD and concussion comorbidity on neurocognitive functioning.
Working memory: individual phases
A series of t-tests were conducted to determine if a relationship existed between PTSD and working memory on each of the individual string lengths (4, 5, 6, 7, and 8 digits) for reverse recall. The Holm procedure was used to correct for familywise error rate. For all string lengths, participants diagnosed with PTSD exhibited poorer working memory than participants in the control group ().
Figure 1 Group differences in working memory performance qualified as average percent correct by string length on the Backward Digit Span task. [Correction added after first online publication on 04 May 2012: The P values have been amended to **p < .01 (more ...)