The Impact of TBI on Executive Function

10.12.1 Verbal Fluency Measures

As predicted in Section 10.1, the performance of TBI sufferers was inferior to that of controls on the Phonemic Fluency task. The lack of a difference predicted in Section 10.1 however between TBI sufferers and controls for Semantic Fluency was unexpected as both measures have been previously shown to be sensitive to the impact of TBI (Belanger et al., 2005; Henry & Crawford, 2004; Goleburn & Golden, 2001; Goldstein et al., 2001; Rapoport

174 et al., 2006; Raskin & Rearick, 1996). The lack of significant differences on the Semantic Fluency task is further surprising given that Henry and Crawford (2004) noted a similitude of effect sizes between the two fluency paradigms in their meta-analysis of TBI patient data. It is also surprising given that Rapoport et al. (2006) were able to detect differences between patients and controls at 12 months post-injury in their study of adults 50 years and older, who, as per the current sample, had also suffered mild-to-moderate injuries. Perhaps in the more modestly sized current sample, it was simply that the Phonemic Fluency task, having a greater number of trials, was the more reliable and thus sensitive of the two paradigms. The number of trials has certainly been identified previously by Strauss et al., (2006) as a

procedural and interpretative concern when comparing the two tasks. Such an explanation is given more credence in light of the trend in the predicted direction, with TBI sufferers generated approximately four fewer words than controls (see Table 10.4). And while

discordance in the number of trials between the Phonemic and Semantic Fluency task can be viewed as a problematic, employing the standard forms facilitates both ease of comparison with the existing literature and generalisability to clinical practice.

Consistent with the prediction made in Section 10.1, TBI sufferers performed significantly worse than controls with respect to Ideational Fluency, as measured by total number of correct uses generated on the Alternate Uses (AU) test (Guilford, Christensen, Merrifield & Wilson, 1978). There were no significant differences between the two groups in terms of commission of errors. As to where the results fit with respect to existing TBI

research, there is little published data available for comparison. Milders, Fuchs and Crawford (2003) recorded a trend for poorer performance of TBI sufferers relative to controls using a small sample of young severely injured patients. Patients and controls in that study did not differ on the proportion of errors versus correct uses generated. Prior to the work of Milders et al., Crawford, Wright & Bate (1995) published a conference abstract suggesting that the

AU task was the most sensitive of verbal fluency measures employed within their TBI sample. As the AU test was shown to be adequately sensitive to TBI in the current study, greater discussion of the paradigm’s utility as a measure of executive function is held over until Chapter 11, where it can be considered in tandem with the findings from Study 1.

The dissociation between the Phonemic Fluency task’s invariance to age and its converse sensitivity to TBI was not anticipated when commencing the literature reviews for this thesis. The same disparity was apparent in the pattern of results returned by Studies 1 and 2, and has not received commentary within the literature to date. Further discussion of the issue takes place within the final Chapter. In terms of validity, it has been suggested that the Phonemic Fluency task merely taps lexical access (Fisk and Sharp, 2004; Shores,

Carstairs & Crawford, 2006; Turner, 1999) rather than strategic retrieval and goal-directed behaviour (Bryan, & Luszcz, 2000a; Strauss et al., 2006). As per Study 1, the current analyses, in lieu of more sophisticated techniques such as linear regression, factor analysis and structural equation modelling, cannot greatly clarify the issue (MacCallum, Widaman, Zhang & Hong, 1999; Tabachnick & Fidel, 2001). Of all the cognitive variables, the Phonemic Fluency task correlated significantly with TMT-B only, at r = .73; whereas Semantic Fluency and AU total correct correlated with one another at r = .59 (as per Section 11.9).

10.12.2 The Stroop Test

It was predicted in Section 10.1 that TBI would be deleterious to Stroop performance. Taking the Stroop index score only, there was a strong trend for inferior performance of TBI sufferers at p = .057. Using the index score is preferable to time taken to complete the incongruent trial due to the potential for baseline differences in processing speed to exert an extraneous influence (Strauss et al., 2006; Troyer, Leach, & Strauss, 2006). Nevertheless, when analysing results for time taken in seconds on the incongruent trial by way of further

176 exploring the above noted trend, the performance of the TBI patients was significantly poorer relative to normal controls. Further, there was no significant difference between the two groups on the control condition which goes some way to allay concerns that the result merely represents baseline speed differences between the two groups. Thus, in line with predictions, it can be concluded that there was a deleterious impact of TBI on the Stroop task for the older adult patients captured by the current study.

In terms of the literature reviewed previously, the Stroop was either employed infrequently or collapsed together with other variables to give a composite executive score making cross-study comparisons difficult. Within the non-ageing TBI literature, poorer performance was demonstrated in severe cases by Bate, Mathias and Crawford (2001) and in mild-to-moderate cases by Chan (2000), although the latter's subjects were recruited on the basis of subjective attentional complaint. Hennessy, Geffen, Pauley and Cutmore (2003) did not find differences between mTBI patients and controls at one month post-injury. In review Strauss et al., (2006) deemed the Stroop to be TBI-sensitive, but not at the milder end of the injury spectrum. However, that finding may not hold among older TBI sufferers given the differences detected in the current sample, especially given the preponderance of mild injuries and the non-acute interval between time of injury and testing. It is unlikely that the influence of age and education account for the differences recorded herein due to the close matching of the TBI and control samples (see Table 10.1). Perhaps in this age group the effects of normal ageing combined with insult from TBI assailed cognitive reserve beyond a critical point. Further replication among older adult TBI populations is warranted.

10.12.3 Trail Making Test Part B (TMT-B)

As noted in Section 10.3, control participants were not administered the TMT-B, so normative data was used for comparison purposes. In this instance, the prediction made in Section 10.1 that TBI would preferentially impact executive function was not supported. The

TBI patients performed on average at the 48th percentile, although the standard deviation was large (32.4; see Table 10.4). The result is consistent with the existent literature. In review Strauss et al., (2006) deemed the TMT to have questionable utility in mTBI populations. Brooks, Fos, Greve & Hammond (1999) found poor performance in mTBI patients relative to controls at only three days post-injury, while Hennessy and colleagues (2003) did not

demonstrate deficits for mTBI cases at one month post-injury. Chan (2000) failed to document poorer performance in mild-to-moderate TBI patients on the TMT-B, despite recruiting on the basis of attentional complaint. With respect to older TBI samples, Goldstein et al., (2001) recorded differences between moderate TBI sufferers and controls at 1 month post-injury on the TMT-B, while Ashman et al., (2008) did not when testing subjects several years post-injury. And while the current TMT result is consistent with the existing literature a caveat is warranted. Due to the longitudinal nature of the NTR protocol, participants had typically completed the TMT at least twice previously, thus practice may have advantaged the TBI cohort in comparison to the normative control data. That is, repeated measurement for TBI sufferers may have led to an underestimation of true difference, thus further testing of older adult TBI patients on the measure may be warranted. Additionally, Lezak et al., (2004) have previously argued that the large standard deviations on Part-B may obscure true

differences and thus contribute to negative findings.

101.2.4 Divided Attention – Telephone Search While Counting (TSC)

As predicted in Section 10.1, the performance of the TBI cohort was inferior to that of control subjects on Telephone Search while Counting (TSC; Robertson, Ward, Ridgeway, & Nimmo-Smith, 1994). At ² = .213, the effects size was noteworthy (Trusty, Thompson & Petrocelli, 2004). The result gives stronger support for Robertson and colleagues (1994) assertion of the index’s great sensitivity than either Hennessy et al. (2003), or Chan (2000). Hennessy and colleagues had previously documented differences between controls and mTBI

178 sufferers on this measure at only 1 month from injury, while Chan (2000) differentiated moderate and severe TBI sufferers from controls at a mean time of 14 months post-injury using a group recruited on the basis of subjective attentional complaint. Ziino and Ponsford (2006) failed to demonstrate differences in a sample of mixed TBI severity relative to controls, although great heterogeneity in time since injury and large SDs in the TBI group may have obscured such differences. Bate et al., (2001) also failed to detect differences between controls and severe TBI patients on this measure and suggested that Robertson et al., may have succeeded in doing so due to having a shorter post-injury interval, by not

controlling for the influence of IQ and due to possible sampling error given that there were only 15 patients sampled. The IQ criticism does not apply to the current study as the variable was covaried. As with the Stroop result, the impact of TBI coupled with a reduction in cognitive reserve via the normal ageing process, may account for differences detected on the TSC in comparison to the lack of differences that have been documented for younger patients either more severely or acutely injured. As no available published studies have used the TSC among an older TBI cohort, replication is merited.