Archives of Physical Medicine and Rehabilitation
Volume 89, Issue 12, Supplement , Pages S69-S76, December 2008

Long-Term Cognitive Outcome in Moderate to Severe Traumatic Brain Injury: A Meta-Analysis Examining Timed and Untimed Tests at 1 and 4.5 or More Years After Injury

  • Lesley Ruttan, PhD, CPsych

      Affiliations

    • Toronto Rehabilitation Institute, Toronto, ON, Canada
    • Corresponding Author InformationReprint requests to Lesley Ruttan, PhD, Toronto Rehabilitation Institute, 345 Rumsey Rd, Toronto, ON, M4G 1R7, Canada
    • ,
  • Krystle Martin, MA

      Affiliations

    • Ontario Institute for Studies in Education, University of Toronto, Toronto, ON, Canada
    • ,
  • Anita Liu, HBSc

      Affiliations

    • Toronto Rehabilitation Institute, Toronto, ON, Canada
    • ,
  • Brenda Colella, MA

      Affiliations

    • Toronto Rehabilitation Institute, Toronto, ON, Canada
    • ,
  • Robin E. Green, PhD, CPsych

      Affiliations

    • Toronto Rehabilitation Institute, Toronto, ON, Canada
    • Graduate Department of Rehabilitation Sciences, University of Toronto, Toronto, ON, Canada

Article Outline

Abstract 

Ruttan L, Martin K, Liu A, Colella B, Green RE. Long-term cognitive outcome in moderate to severe traumatic brain injury: a meta-analysis examining timed and untimed tests at 1 and 4.5 or more years after injury.

Objectives

To examine long-term outcome of moderate to severe traumatic brain injury (TBI) on timed and untimed cognitive tests using meta-analysis.

Design

Meta-analysis examining outcome at 2 epochs, 6 to 18 months postinjury (epoch 1) and 4.5 to 11 years postinjury (epoch 2).

Setting

Data source was published articles (1966–2007) identified through electronic and manual search.

Participants

A total of 1380 subjects with moderate to severe TBI participated in the 16 studies meeting inclusion criteria.

Interventions

Not applicable.

Main Outcome Measures

Timed and untimed neuropsychologic tests with quantitative results (means, SDs, t, and df tests) from studies containing a healthy comparison group and a mean time since injury falling within 1 of the 2 epochs.

Results

Patient versus control weighted effect sizes were medium to large at epoch 1 for both untimed tasks (r=−.46; confidence interval [CI], −.32 to −.65) and timed tasks (r=−.46; CI, −.35 to −.59). At epoch 2, effect sizes were slightly smaller for untimed tasks (r=−.38; CI, −.25 to −.60) and timed tasks (r=−.40; CI, −.32 to −.62).

Conclusions

Patients showed robust, persisting impairments on both timed and untimed tests at recovery plateau (ie, 6–18mo postinjury) and many years later. These findings converge with previous studies, though using an alternative approach that obviates some of the methodologic problems of longitudinal studies, such as selective attrition.

Key Words: Brain injuries, Neuropsychology, Rehabilitation

List of Abbreviations: CI, confidence interval, ES, effect size, GCS, Glasgow Coma Scale, LOC, loss of consciousness, PASAT, Paced Auditory Serial Addition Test, PTA, posttraumatic amnesia, TBI, traumatic brain injury, TSI, time since injury

 

ACCORDING TO THE World Health Organization, TBI will surpass many diseases as the major cause of disability (and death) by the year 2020, with approximately 10 million people affected annually worldwide.1 The magnitude of the problem is caused in large part by the persistence of cognitive impairments,2, 3, 4, 5, 6, 7, 8, 9, 10 which may cause decades of disability as a result of the young demographic that TBI afflicts.11, 12, 13, 14 In moderate and severe TBI, cognitive deficits are associated strongly with the ongoing disruption of social networks, employability, and return to work.15, 16, 17 Therefore, gaining an understanding of the long-term course of cognitive functioning is of marked clinical importance: knowing whether some or all cognitive functions show persisting deficits into the longer term would allow better prognostication, planning, and the likelihood of developing targeted prophylactic treatments; the last is particularly important given the small but growing number of findings that some people even show a regression of some cognitive functions over time.8, 18, 19, 20 Because some research has demonstrated that information processing impairments may mediate the relationship between TBI severity and post-TBI adaptive functioning,21 in the present study, we compared performance at 2 subacute time points on timed versus untimed tests.

Prospective studies examining early cognitive recovery after moderate and severe injury2, 22, 23 have generally found that most recovery occurs within the first 6 to 18 months of injury2, 23 and that some improvement may continue thereafter at a slower pace.8, 24, 25 There has been considerable variability in findings on the rate and pattern of recovery of different functions,22, 26 with the exception of speed of processing, which shows the most consistent findings after moderate to severe TBI.27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37 Studies have shown a higher prevalence of compromise36, 38, 39 and slower speed and extent of recovery40, 41 on timed versus untimed tests, suggesting that speed of processing is uniquely vulnerable to the effects of moderate and severe brain injury.

With regard to longer-term outcomes, prospective studies of long-term cognitive recovery are relatively few. As pointed out by Dikmen et al,42 longitudinal cognitive recovery studies have also contained significant methodologic limitations, with a follow-up period rarely exceeding 2 years, biased samples (eg, those referred for clinical care),19 and even in prospective, well defined samples, very high attrition rates.8, 19, 20 As well, test batteries have often included only 118 or 2 tests.42 As such, other approaches to studying long-term outcome are warranted.

Overall, long-term recovery studies have not yet shown clear-cut patterns of findings across tests,8, 18, 19 although persisting impairments to speed of processing seem to be a common thread across studies. Dikmen et al,42 who maintained an 80% retention rate (but with 60% of the sample in the mild TBI range), prospectively examined long-term outcome at 3 to 5 years postinjury on a test of verbal learning and recall, and on the PASAT, a challenging timed test of sustained attention requiring serial calculations.43 Clinically impaired performances (2–3 SDs below average) were observed on the PASAT43 at all degrees of severity. However, for verbal learning, clinically significant impairment was observed only at the most severe level of injury, and impairment was still only 1.35 deviations below average. Moreover, prior neurologic disorders were present in some patients, and therefore, even this level of impairment may have been an overestimate. Thus, definitive persisting deficits on the timed PASAT test were observed, while the persisting deficits in verbal learning were milder and more circumscribed.

Salmond et al44 reported persistence of deficits across 2 time points a minimum of 6 months apart, ranging from 6 months postinjury to 3 years postinjury. While not explicitly mentioned, 6 of the 8 variables showing significant differences between patients and controls were timed variables, and of the variables showing no between-group differences, only 1 of 6 was a timed variable. These results, too, suggest that performance on timed variables may be more significantly affected in the longer term. These results are consistent with Spikman et al,45 who examined 51 patients with TBI several years postinjury and whose findings underscored the persistence of impairments in speed of processing.

A small number of long-term outcome studies have also shown a regression in functioning over time. The finding does not appear to be attributable to age-related decline given the use of age-scaled normative data or age-matched control groups in studies. Again, clear patterns in the data have yet to emerge in this branch of research; also, methodologic weaknesses limit the inferences that can be drawn. In a 30-year follow-up study of 210 patients with TBI, Himanen et al19 observed slight cognitive decline in 56% of patients on learning, memory, visuospatial construction, and arithmetic, and improvement only in semantic memory. However, attrition was 70%, and the study sample was composed of patients specifically referred for neuropsychologic or neurologic follow-up and thus likely showed disproportionate impairments. Salmond44 found a decline in their battery in spatial recognition only; however, interestingly, it was the latency (ie, timed) measure of spatial recognition on which the decline was observed. Finally, in a study of 182 people with mild to severe TBI carried out by Millis et al,8 15.2% showed a decline, 22.1% showed improvement, and 62.2% were unchanged over a 5-year period. Although the authors reported that most improvement was observed on measures of cognitive speed, visuoconstruction, and verbal memory, it is of interest to note that 2 of the 3 tests showing the highest decline (grooved pegboard and Trail-Making Test Part B) were also timed tests. However, this study, too, had very high attrition (>90%), rendering the observations regarding patterns across tests somewhat tenuous.

Taken together, previous studies suggest that cognitive deficits—particularly speed of processing—may persist or even worsen in the longer term, but that a clear pattern of findings across functions has yet to be identified. One explanation for variability in findings, as Mathias and Wheaton27 have pointed out, is that past studies of recovery may be confounded by the conflation of timed with untimed tests,46 whereby speed of processing deficits compromise performance on tests that are purportedly measuring other cognitive functions (eg, spatial recognition), but nonetheless contain a timed dependent measure.

In sum, cognitive processing is consistently impaired by moderate and severe TBI, and there is evidence that deficits persist into the longer term. However, the patterns of persisting deficits are uncertain to date, because of methodologic limitations of the relatively few studies undertaken, including the confounding of untimed cognitive processes (eg, memory) with timed dependent variables. Therefore, further research is warranted to disentangle the contribution of speed of processing from untimed tests as well as to examine speed of processing on its own.

The current study employed meta-analysis to examine the long-term cognitive recovery of timed and untimed processes at approximately 1 year postinjury and at greater than 4.5 years postinjury. We were interested in how patients performed toward the end of the early recovery period versus how they fared many years postinjury, long after therapies would have stopped for most patients47 and after a return to prior role would have been attempted. Meta-analysis has the advantage of allowing integration of a collection of studies on a particular topic for the purpose of synthesizing the findings. The method calculates standardized mean differences or ESs (ie, the difference between 2 means relative to the pooled SD, resulting in a z score) to provide a common metric on which to compare studies, with higher ES reflecting stronger differences between groups than lower ES. Researchers have noted that meta-analysis can minimize errors and heighten understanding of neuropsychologically meaningful patterns in data by integrating the findings from multiple studies.48, 49 To our knowledge, there are no published meta-analytic studies in moderate to severe TBI specifically examining the performances of people with TBI versus healthy controls on timed and untimed neuropsychologic tasks at 2 postinjury epochs.

Back to Article Outline

Methods 

Literature Search and Study Selection 

A comprehensive search of PsycINFO and Medline electronic databases was undertaken in order to identify published studies examining cognition and TBI. Key words used in the search included brain injury, traumatic brain injury, TBI, and head injury. These words were combined with cognition, neuropsychology, and neurocognition. Search constraints included articles published in English and those printed after 1966 (after the inception of the Medline database) and indexed up until June 2007. For all articles identified as relevant, a manual search of the articles' reference lists was conducted to see whether there were any cited works that were missed by the search engines.

Original studies were included in the meta-analysis if they met the following criteria: (1) neuropsychologic tasks were administered to patients with TBI who were reported to have sustained either moderate or severe nonpenetrating brain injuries (ie, GCS score <13, LOC >1h, and/or duration of PTA >1h); (2) the mean TSI was reported for patients with TBI that allowed classification of studies into 2 nonoverlapping epochs (epoch 1: mean TSI, 6–18mo; epoch 2: mean TSI, ≥4.5y); (3) the research design included a comparison group composed of healthy, age-matched control subjects; (4) study statistics amenable for conversion to conventional ESs that reflect group differences were available (these measures included group means and SDs, t, and df); and (5) the cognitive tasks included in the study were adequately described so that a determination could be made about whether the cognitive task was timed or untimed. A test was classified as timed if the dependent variable was timed or if exposure to test material was timed, necessitating speeded processing. All classifications were agreed on by 3 experts in the field of neuropsychology.

The titles of articles and abstract content obtained in the literature search were perused for indications that the study met the inclusion criteria. Copies of the full text of these studies were obtained and the articles were reviewed to determine whether they met the inclusion criteria for the study. Each article was examined by a minimum of 2 of the coauthors to determine its eligibility for the meta-analysis, and a total of 16 appropriate studies were identified.

Data Collection 

For each study included in the meta-analysis (table 1), the journal name, title, authors, and date of publication were recorded. Several sample characteristics were also noted, including sample size, mean age, mean education, and sex for patient and control samples when available; plus, severity of brain injury, as measured by GCS, duration of PTA, and length of LOC for the patient group were recorded, as was TSI at testing.

Table 1. Demographic Details for the TBI and Control Groups: Injury Data for the TBI Group
StudiesTBI and Controls (n)TBI (n)Controls (n)Mean Age TBI (y)Mean Age Controls (y)Mean YOE TBIMean YOE ControlsMean TSI (mo)Injury Severity (tests used for classification)
Untimed—Epoch 1
Ariza et al5840.020.020.025.625.411.412.06.0Mod/Sev (GCS)
Dikmen et al22395.0274.0121.028.931.212.012.012.0Mod/Sev (GCS)
Formisano et al5545.025.020.022.3MatchedNANA11.2Sev (GCS)
Mathias et al5250.025.025.028.628.411.812.07.0Sev (GCS, LOC)
Salmond et al4441.021.020.033.036.0NANA9.8Sev (NA)
Mean untimed epoch 1114.273.041.227.728.711.712.09.2
Untimed—Epoch 2
Larson et al8350.026.024.040.735.913.913.8110.0Mod/Sev (GCS, LOC, PTA)
Schmitter-Edgecombe et al5940.020.020.029.529.314.114.564.8Sev (LOC)
Schmitter-Edgecombe et al5454.027.027.032.432.614.014.460.0Sev (GCS, LOC)
Schmitter-Edgecombe et al5048.024.024.034.435.414.114.2126.0Sev (GCS, PTA)
Simpson et al5340.020.020.032.432.214.314.872.0Sev (GCS)
Vickery et al5640.020.020.034.138.712.512.764.8Sev (GCS, LOC)
Mean untimed epoch 245.322.822.533.934.013.814.163.4
Mean epoch 1 and 276.645.631.031.131.613.113.438.8
Timed—Epoch 1
Ariza et al5840.020.020.025.625.411.412.06.0Mod/Sev (GCS)
Dikman et al22395.0274.0121.028.931.212.012.012.0Mod/Sev (GCS)
Formisano et al5545.025.020.022.3MatchedNANA11.2Sev (GCS)
Mathias et al5250.025.025.028.628.411.812.07.0Sev (GCS, LOC)
Periáñez et al64313.090.0223.034.638.912.913.312.5Sev (GCS, PTA)
Spikman et al65104.044.060.029.828.5NANA12.0Sev (GCS, PTA)
Vakil et al7352.025.027.027.025.4NANA4.9Sev (GCS, LOC)
Mean timed epoch 1142.771.970.928.130.312.012.39.4
Timed—Epoch 2
O'Keeffe et al6236.018.018.031.332.9NANA38.6Sev (GCS, PTA)
Park et al6324.012.012.053.251.515.315.962.4Sev (NA)
Schmitter-Edgecombe et al5940.020.020.029.329.514.114.564.8Sev (PTA)
Schmitter-Edgecombe et al5048.024.024.034.435.414.114.2126.0Sev (GCS, PTA)
Simpson et al5340.020.020.032.332.214.314.872.0Sev (GCS)
Vickery et al5648.023.025.034.138.712.512.761.7Sev (GCS, LOC)
Mean timed epoch 239.319.519.835.836.714.014.451.4
Mean epoch 1 and 295.047.747.331.733.213.113.528.8

NOTE. Severity ratings are based on self-description of mean severity level provided by author of article.

Abbreviations: LOC, loss of consciousness; Mod, moderate; NA, not available; YOE, years of education; Sev, severe.

For some studies, adjustments to the published data were required in order to allow inclusion of the data in the meta-analysis. For example, for longitudinal studies of recovery, in which patients were tested more than once, only data from the first occasion of testing were included (to avoid contaminating the results with practice effects). Also, if more than 1 neuropsychologic test was used to measure a cognitive domain within a study, the mean ES was used so that each study had only 1 ES estimate per grouping of tasks, timed or untimed. Also, in the case of separate articles conducted by the same authors in which participants overlapped substantially, the findings were combined and treated as 1 study.50, 51

All cognitive tasks and test variables were included in the meta-analytic database. Pooling of the numerous and various cognitive measures for purposes of analysis can raise methodologic and theoretical challenges because of the heterogeneity of test batteries across studies. For the purpose of this study, we were interested only in examining the global difference between timed and untimed neuropsychologic tests, so the various cognitive measures were simply divided into these 2 categories based on the researchers' specific knowledge of neurocognitive tests and/or as per the description provided in the original articles. A description of the individual cognitive variables that make up these 2 groups of tasks can be found in table 2. The studies were then further divided into the 2 nonoverlapping epochs. Epoch 1 was chosen to allow an examination of performances after the end of the acute recovery period and when most recovery would be expected to have occurred based on previous studies. Epoch 2 was selected to be nonoverlapping, and representative of chronic TBI. We selected a time point of 4 years postinjury as the earliest time point. Studies that met inclusion criteria ranged from a TSI of 4.5 to 10.6 years (see table 1 for studies included in each cognitive domain and epoch).

Table 2. Summary of Individual Untimed and Timed Cognitive Tests by Domain and Epoch
Untimed TestsE 1E 2Timed TestsE 1E 2
Learning and Recall Fluency
AVLT Immediate Recall52X Controlled Oral Word Association58, 60X
AVLT Delayed Recall52X Design Fluency52, 53, 55, 56, 57, 61X
California Verbal Learning Test—Total53 XScanning
California Verbal Learning Test—Trial 154 XDigital cancellation ms53 X
California Verbal Learning Test—Trial 554 XDigit cancellation hits53 X
California Verbal Learning Test Short-Delay54 XPsychomotor
Long Delay Free Recall53, 54 XFinger tapping—dominant22, 58, 59X
Prose Delayed Recall55X Grooved pegboard—dominant hand56 X
Related-Word Delayed Recall55X Finger tapping—nondominant22, 56, 58, 59XX
Unrelated-Word Delayed Recall55X Name writing—dominant22X
WMS—Logical Memory22X Name writing—nondominant22, 58X
WMS—Visual Reproduction 22X Attention
WMS—Logical Memory22, 51XXDual Attention to Response Task62 X
WMS—Visual Reproduction22, 51XXPASAT (adapted)1st Third63 X
WMS-III Word List Immediate Recall56 XPASAT (adapted) 2nd Third63 X
WMS-III Word List Delayed Recall56 XPASAT (adapted) 3rd Third63 X
WMS-III Word List Recognition56 XSustained Attention to Response Task—Fixed62 X
Selective Reminding Total Recall22X Sustained Attention to Response Task—Random62 X
Selective Reminding Total Delayed Recall22X Trail-Making Test Part A22, 52, 53, 58, 64X
Total Recall after 4-hour delay22X Word Reading/Naming
Prospective Memory Focal Cue Hits51 XStroop—color reading22X
Prospective Memory Peripheral Cue Hits51 XStroop—word reading22X
Executive Function Tests Stroop—color/word reading53, 65XX
Raven's Progressive Matrices55X Executive
WCST—Categories53, 54 XTrail-Making Test Part B22, 52, 53, 64X
WCST—Errors52X Trail-Making Test Part B – A51, 64X
WCST—Perseverations53, 54 XTrail-Making Test Part B:A Ratio64X
Category Test22, 55, 57X Trail-Making Test Part B-A/Trails A64X
Mental Efficiency
Alphabet Span Test51 X
Seashore Rhythm Test22X
Visuospatial Skills
Facial Recognition58X
Rey Figure Copy58X
Other
Tactual Performance Test22X
WAIS—Full Scale Intelligence Quotient51, 53, 59 X

Abbreviations: AVLT, Auditory Verbal Learning Test; WAIS, Wechsler Adult Intelligence Test; WCST, Wisconsin Card Sort Test; WMS, Wechsler Memory Scale.

Magnitude of effect 

A meta-analysis assumes that each study estimates real differences between groups. Combining several estimations thus yields a more veridical estimation of the real effect. Standardized ESs were initially computed for each variable with adequate quantitative information (eg, mean ± SD). ESs were then averaged across studies. While the d statistic has been the most conventional meta-index previously used, a recent preference for correlational ES measures is notable in the literature.66 As such, we used r values in our report. The d and r indices are interchangeable according to the relation r=[d2/(d2 + 4)]1/2 (assuming relatively equal sample sizes).49 Whereas d is interpreted in terms of SD units (ie, mean difference/pooled SD), mean r values are interpreted in the same manner as typical correlation coefficients. In Cohen's49 qualitative terms, mean r's of .10, .20, and .50 can be considered small, medium, and large ESs, respectively, in behavioral-science research. The r values (weighted and unweighted) and CIs were calculated by first obtaining the Fisher z-transformed coefficients from the individual test comparisons within each study.67

Weighted r values reported are corrected by sample size, thereby giving greatest weight to the studies with the most reliably estimated study ESs—those with the largest sample sizes.68 Each mean weighted r2 and SD was then used to compute 95% CIs for both the timed and untimed groups. Finally, the r2 values were reconverted to r values.66, 67 The 95% CI around the weighted r reflects the variability across studies and, if encompassing 0, indicates nonsignificant (P>.05) deviation of the mean ES from this null hypothesis value.

Note that, given the relatively small number of studies published in this area, the use of weighted r could disproportionately bias results based on a single study. Therefore, unweighted r values are also presented in order to provide the most comprehensive review possible.

In addition to the CIs, coefficients of robustness were calculated to compare the observed effects for robustness and homogeneity. Viewed as a “second-order effect size,”69 the coefficients of robustness provides a common metric in order to compare average group ESs as they are standardized by their SDs.

Because of the potential for meta-analyses to be subject to publication biases—that is, the selective publication of significant results—fail-safe calculations were also conducted. A number of different methods have been proposed to deal with this issue. Consistent with our choice of the r measure of ES, we applied the Orwin method70 that calculates the number of additional studies of null effect needed to reduce the average ES to a minimal size (ie, r=.10).49, 71

Back to Article Outline

Results 

Participants 

A total of 1380 participants from 16 studies were included in this meta-analysis: 694 people with moderate to severe TBI and 686 control participants. Demographic details for participants in the studies included in this analysis are summarized in table 1. Participants were similar in terms of their mean age and years of education. When the ages and educational levels were compared via t test, the TBI and control groups did not differ significantly (education, t15=.55, P>.05; age, t15=.76, P>.05), suggesting that these groups were overall well matched on these variables. Furthermore, participants in studies across both epochs, and across timed versus untimed tests, did not appear to differ in their demographic data and injury characteristics (see table 1). From among these publications, 13 studies had either similar numbers or a matched percentage of men in both TBI and control groups, 2 studies did not report the sex of participants in either group,63, 72 1 study reported only the sex of participants in the TBI group,73 and 4 studies reported unequal numbers of men between the TBI and control groups. The overall percentage of men in the TBI group was 75.8, while overall percentage of men in the control group was 69.4. In terms of injury severity, the overall average GCS score was 7.25, while average length of PTA was 53.23 days. Thus, participants were, overall, within the severe range of severity.

Untimed versus timed tests 

Similar numbers of studies were used in each epoch for untimed and timed tests (untimed=5 studies in epoch 1, and 6 studies in epoch 2; timed=7 studies in epoch 1, and 6 studies in epoch 2).

Table 3 lists the ES differences (weighted r, unweighted r, CIs, coefficients of robustness, and fail-safe n) between patients with TBI and healthy controls for the 2 epochs organized by timed and untimed neuropsychologic tests.

Table 3. ES Summary Statistics for TBI Versus Control Groups on Untimed and Timed Cognitive Tasks at Epochs 1 and 2
GroupsknTotal Weightedr ScoreCIsTotal Unweightedr ScoreFail-Safe nCoefficient of Robustness
Untimed cognitive tasks12895−0.45−0.36to−0.57 34−2.87
Overall −0.47
Epoch 15323−0.46−0.32to−0.65−0.5036−1.83
Epoch 27273−0.38−0.25to−0.60−0.4470−2.26
Timed cognitive tasks131295−0.45−0.40to−0.56−0.4988−2.11
Overall −0.52
Epoch 171015−0.46−0.35to−0.59−0.4854−1.60
Epoch 26280−0.40−0.32to−0.62−0.4941−1.44

The overall weighted r for each domain collapsed across the 2 epochs shows that patients with TBI performed worse than controls on both types of tests (table 3; untimed: weighted r=−.45, CI, −.36 to −.57; timed: weighted r=−.45, CI, −.40 to −.56). Examination of unweighted r yielded similar results.

For untimed tests, a moderate to large negative ES difference (weighted r) in both epochs (epoch 1: weighted r=−.46, CI, −.32 to −.65; epoch 2: weighted r=−.38 CI, −.25 to −.60) was observed. The negative ESs show that patients with TBI perform consistently worse than controls on untimed tests across these 2 points in the chronic TBI spectrum. Examination of unweighted r values yielded similar results. The overall CIs suggest that the results are reliable; the coefficient of robustness suggests the method of analysis is relatively robust to outliers and small departures from underlying assumptions.

Similarly, for timed tests, a moderate to large negative ES difference (weighted r) in both epochs (epoch 1: weighted r=−.46, CI, −.35 to −.59; epoch 2: weighted r=−.40; CI, −.32 to −.62) was noted. Like untimed cognitive tasks, the findings show that patients with TBI consistently performed worse than controls. Examination of unweighted r yielded a very similar pattern of findings from epoch 1 to 2.

Back to Article Outline

Discussion 

To our knowledge, this is the first research synthesis that has examined untimed versus timed cognitive processing during 2 subacute periods after moderate to severe TBI. Very few longitudinal studies of recovery have examined recovery at this late postinjury phase; of these, most were plagued by significant attrition and likely sample bias. Consequently, a clear and definitive clinical picture of long-term outcome from moderate and severe TBI has remained elusive.

We observed robust residual cognitive deficits at both epochs, for both timed and untimed processing. In a meta-analytic study of subjects with moderate to severe TBI, Schretlen and Shapiro74 noted large ESs when they explored earlier epochs than in our study (<6mo postinjury; >24mo postinjury). Our findings were compatible with these findings in that significant persisting cognitive impairments were observed during the chronic stages of recovery.

In the current study, we further characterized this picture by separately examining untimed versus timed cognitive processing across epochs, thus avoiding the potential confounding of untimed performances by impaired speed of processing. We observed moderate to large ESs at each epoch for untimed and timed tasks, indicating that significant cognitive impairments persist even when the dimension of speed of processing is parceled out from neuropsychologic tests encompassing verbal learning/recall, visuospatial learning/recall, prospective memory, executive functioning, mental efficiency, and visuospatial skills. These findings are consistent with those of a recent study38 finding that deficits in divided attention, executive functions, and long-term memory were better explained by a primary deficit in working memory than a deficit in speed of processing per se at approximately 1.5 years after brain injury.

That cognitive deficits persist even in the absence of any speed of processing requirement suggests that patients with TBI, their families, and the health care professionals that work with them should be prepared to implement compensatory strategies and accommodations for a wide variety of deficits into the long term in order to facilitate functioning and improve quality of life. These efforts are of particular import given the negative impact of persistent cognitive deficits post-TBI (eg, economic toll, increased psychologic distress such as depression/suicide, psychosocial dysfunction, marital/family/social network breakdown) and may be guided by findings linking specific cognitive deficits to functional outcome (eg, Rassovsky et al21 found that speed of information processing but not verbal memory mediated the relationship between TBI severity and post-TBI adaptive functioning).21

Further, more fine-grained research is needed to characterize primary, underlying persisting deficits and possible differences across neuropathologic, demographic, or psychosocial subgroups. It is also needed in order to uncover more subtle chronic changes that may exist, such as the clinically and scientifically important possibility of postrecovery cognitive decline. Animal75, 76 and human studies77, 78 have suggested that the brain may atrophy after initial injury and that cognitive functioning, too, may show a decline for some people in some domains after an initial normalized period of recovery.8, 19, 20 For example, Till et al20 found that approximately 1 (27%) in 4 adults with moderate to severe TBI demonstrated some degree of cognitive decline between the first year of recovery and 2 to 5 postinjury. Ng et al79 found visible interval changes on structural magnetic resonance imaging (read by expert neuroradiologists) over the same time period, including increased gliosis, increased white matter signal intensities, and tissue loss.

Persisting cognitive impairments may also leave some people vulnerable to a downward spiral of neuroplasticity with negative consequences analogous to that proposed by Mahncke et al80, 81 in association with aging. Turner and Green82 have argued that chronic TBI, with its endogenous and exogenous changes similar to those observed in normative aging (eg, cognitive impairment, loss of employment, reduced opportunities for social contact), fosters conditions for negative plasticity.

Study Limitations 

It should be noted that our literature search was limited to articles published in English, and that it is possible that unpublished studies with nonsignificant findings were excluded from the current meta-analysis. It is also possible that the search terms used were not exhaustive, resulting in the omission of relevant articles. Recruitment methods of the included studies were not assessed (eg, subjects may have be enrolled because they were symptomatic or had experienced particularly good or poor recoveries), and this too could potentially contribute to outcome heterogeneity.

The ultimate quality of a meta-analysis is guided by the quality of included primary studies. Mathias and Wheaton27 describe a number of factors that must be included within TBI studies in order to allow appropriate evaluation of individual studies and integration of studies in the form of meta-analyses. Future authors are advised to include those factors cited and to examine neuropsychologic recovery course separately for timed and untimed tasks. One limitation of the primary studies—which in turn produced limitations in the current meta-analysis—is that they did not quantify the percentage of variance accounted for by speed of processing versus other mental processes. Because the tests used in the studies eligible for inclusion at epoch 1 differed from those in epoch 2, direct comparisons across the epochs should be made with caution.

Another reason comparisons across epochs must be made conservatively is that there may be differences in the samples that would influence the results. For example, clinical or demographic factors (eg, persisting symptoms, socioeconomic status, education) could carry greater influences on the probability of participating in a long-term follow-up study than an earlier study; studies conducted early postinjury are often incorporated into routine clinical care, and thus factors such as family support, socioeconomic status, and severity of injury are less likely to influence participation. In this regard, we found that years of education were slightly higher in epoch 2 than epoch 1 in the current study, suggesting that those available for or interested in participating in a neuropsychologic study many years postinjury might be better educated than those who did not participate. A mitigating factor for this concern, however, is that the disparity in education from epoch 1 to epoch 2 was the same for timed and untimed tests. Therefore, any biases as a function of TSI may have influenced the 2 test types similarly. Another mitigating factor is that severity of injury was highly similar across test type and epochs, with most studies including patients in the severe range of injury on average, and a few made up of patients in the moderate-to-severe range.

Additional factors that may be worthwhile to include in future studies and were not evaluated in our study include the presence of chronic pain symptoms, affective status, and symptom validity performance. Issues such as chronic pain and affective status are additional factors associated with TBI that could influence an individual's neuropsychologic test performance, resulting in potential decline and/or fluctuation over time. Future work incorporating these factors would enrich our current knowledge regarding TBI recovery course and possible moderators.

Back to Article Outline

Conclusions 

People with moderate to severe TBI exhibit wide-ranging neuropsychologic deficits in both the acute and chronic phases of recovery. Relative to healthy controls, people with moderate to severe TBI were impaired on both timed and untimed neuropsychologic tasks at 2 chronic postinjury epochs: at or shortly after recovery plateau, and much later, at 4.5 or more years postinjury. These results, using meta-analysis, converge with and provide further specificity to the relatively small number of previous longitudinal studies of cognitive deficits in chronic TBI.

Back to Article Outline

Acknowledgments 

We thank Kadeen Johns, BA, Joel Verwegan, and Areeba Adnan for their valuable contributions to the preparation of this article. We are also grateful to Todd Girard, PhD, for his thoughtful and insightful input.

The authors acknowledge the support of Toronto Rehabilitation Institute who receives funding under the Provincial Rehabilitation Research Program from the Ministry of Health and Long-Term Care in Ontario. The views expressed do not necessarily reflect those of the Ministry.

Back to Article Outline

References 

  1. Hyder AA, Wunderlich CA, Puvanachandra P, Gururaj G, Kobusingye OC. The impact of traumatic brain injuries: a global perspective. NeuroRehabilitation. 2007;22:341–353
  2. Levin HS. Neurobehavioral outcome of closed-head injury: implications for clinical-trials. J Neurotrauma. 1995;12:601–610
  3. Paniak CE, Shore DL, Rourke BP. Recovery of memory after severe closed head injury: dissociations in recovery of memory parameters and predictors of outcome. J Clin Exp Neuropsychol. 1989;11:631–644
  4. Masson F, Thicoipe M, Aye P, et al. Epidemiology of severe brain injuries: a prospective population-based study. J Trauma. 2001;51:481–489
  5. Gray C, Cantagallo A, Della Sala S, Basaglia N. Bradykinesia and bradyphrenia revisited: patterns of subclinical deficit in motor speed and cognitive functioning in head-injured patients with good recovery. Brain Inj. 1998;12:429–441
  6. Haaland KY, Temkin N, Randahl G, Dikmen S. Recovery of simple motor skills after head injury. J Clin Exp Neuropsychol. 1994;16:448–458
  7. Ponsford JL, Olver JH, Curran C. A profile of outcome: 2 years after traumatic brain injury. Brain Inj. 1995;9:1–10
  8. Millis SR, Rosenthal MP, Novack TA, et al. Long-term neuropsychological outcome after traumatic brain injury. J Head Trauma Rehabil. 2001;16:343–355
  9. Karol RL. The duration of seeking help following traumatic brain injury: the persistence of nonneurological symptoms. Clin Neuropsychol. 1989;3:244–249
  10. Lezak MD, O'Brian KP. Longitudinal study of emotional, social, and physical changes after traumatic brain injury. J Learn Disabil. 1990;21:456–463
  11. Collins JG. Types of injuries by selected characteristics: United States, 1985-1987. Vital and Health Statistics. Series 10: data from the National Health Survey, DHHS Publication No. PHS-91-1053. In: Hyansville: U.S. Department of Health & Human Services; 1990;p. 1–68175
  12. Grady MS, McIntosh TK. Head trauma. In:  Asbury AK editors. Diseases of the nervous system. 3rd ed.. New York: Cambridge Univ Pr; 2002;p. 1793–1804
  13. Thurman DJ, Alverson C, Browne D, et al. Report to Congress on mild traumatic brain injury in the United States: steps to prevent a serious public health problem. Atlanta: Centers for Disease Control and Prevention, U.S. Department of Health and Human Services; 2003;
  14. Jager TE, Weiss HB, Coben JH, Pepe PE. Traumatic brain injuries evaluated in U.S. emergency departments, 1992-1994. Acad Emerg Med. 2000;7:134–140
  15. Hannay HJ, Howieson DB, Loring DW, Fischer JS, Lezak MD. Neuropathology for neuropsychologists. In:  Lezak MD,  Howieson DB,  Loring DW editor. Neuropsychological assessment. New York: Oxford Univ Pr; 2004;p. 157–285
  16. Satz P, Zaucha K, Forney DL, et al. Neuropsychological, psychosocial and vocational correlates of the Glasgow Outcome Scale at 6 months post injury: a study of moderate to severe traumatic brain injury patients. Brain Inj. 1998;12:555–567
  17. Kalechstein AD, Newton TF, van Gorp WG. Neurocognitive functioning is associated with employment status: a quantitative review. J Clin Exp Neuropsychol. 2003;25:1186–1191
  18. Ruff RM, Young D, Gautille T, et al. Verbal learning deficits following severe head injury: heterogeneity in recovery over 1 year. J Neurosurg. 1991;75:S50–S58
  19. Himanen L, Portin R, Isoniemi H, Helenius H, Kurki T, Tenovuo O. Longitudinal cognitive changes in traumatic brain injury: a 30-year follow-up study. Neurology. 2006;66:187–192
  20. Till C, Colella B, Green R. Post recovery cognitive decline in adults with traumatic brain injury. Arch Phys Med Rehabil. In press.
  21. Rassovsky Y, Satz P, Alfano MS, et al. Functional outcome in TBI II: verbal memory and information processing speed mediators. J Clin Exp Neuropsychol. 2006;28:581–591
  22. Dikmen SS, Machamer JE, Winn HR, Temkin NR. Neuropsychological outcome at 1-year post head injury. Neuropsychology. 1995;9:80–90
  23. Dikmen S, Reitan RM, Temkin NR. Neuropsychological recovery in head injury. Arch Neurol. 1983;40:333–338
  24. Van Zomeren A, Deelman B. Long-term recovery of visual reaction time after closed head injury. J Neurol Neurosurg Psychiatry. 1978;41:452–457
  25. Olver JH, Ponsford JL, Curran CA. Outcome following traumatic brain injury: a comparison between 2 and 5 years after injury. Brain Inj. 1996;10:841–848
  26. Howieson DB, Loring DW, Hannay HJ. Neurobehavioral variables and diagnostic issues. In:  Lezak MD,  Howieson DB,  Loring DW editor. Neuropsychological assessment. New York: Oxford Univ Pr; 2004;p. 286–334
  27. Mathias JL, Wheaton P. Changes in attention and information-processing speed following severe traumatic brain injury: a meta-analytic review. Neuropsychology. 2007;21:212–223
  28. Felmingham KL, Baguley IJ, Green AM. Effects of diffuse axonal injury on speed of information processing following severe traumatic brain injury. Neuropsychology. 2004;18:564–571
  29. Van Zomeren AH, Brouwer WH. Head injury and concepts of attention. In:  Levin HS,  Grafman J,  Eisenberg HM editor. Neurobehavioural recovery from head injury. New York: Oxford Univ Pr; 1987;p. 398–415
  30. Stuss D, Stethem L, Hugenholtz H, Picton T, Pivik J, Richard M. Reaction time after head injury: fatigue, divided and focused attention, and consistency of performance. J Neurol Neurosurg Psychiatry. 1989;52:742–748
  31. Ferraro FR. Cognitive slowing in closed-head injury. Brain Cogn. 1996;32:429–440
  32. Veltman JC, Brouwer WH, van Zomeren AH, van Wolffelaar PC. Central executive aspects of attention in subacute severe and very severe closed head injury patients: planning, inhibition, flexibility, and divided attention. Neuropsychology. 1996;10:357–367
  33. Zahn TP, Mirsky AF. Reaction time indicators of attention deficits in closed head injury. J Clin Exp Neuropsychol. 1999;21:352–367
  34. Madigan NK, DeLuca JP, Diamond BJ, Tramontano G, Averill A. Speed of information processing in traumatic brain injury: modality-specific factors. J Head Trauma Rehabil. 2000;15:943–956
  35. Bashore TR, Ridderinkhof KR. Older age, traumatic brain injury, and cognitive slowing: some convergent and divergent findings. Psychol Bull. 2002;128:151–198
  36. Hanks RA, Ricker JH, Millis SR. Empirical evidence regarding the neuropsychological assessment of moderate and severe traumatic brain injury. In: Differential diagnosis in adult neuropsychological assessment. New York: Springer; 2004;p. 218–242
  37. Mathias JL, Bigler ED, Jones NR, et al. Neuropsychological and information processing performance and its relationship to white matter changes following moderate and severe traumatic brain injury: a preliminary study. Appl Neuropsychol. 2004;11:134–152
  38. Ciaramelli E, Serino A, Di Santantonio A, Ladavas E. Central executive system impairment in traumatic brain injury. Brain Cogn. 2006;60:198–199
  39. Levin HS. Cognitive function outcomes after traumatic brain injury. Curr Opin Neurobiol. 1998;11:643–646
  40. Zwaagstra R, Schmidt I, Vanier M. Recovery of speed of information processing in closed-head-injury patients. J Clin Exp Neuropsychol. 1996;18:383–393
  41. Van Zomeren AH, Brouwer WH. Clinical neuropsychology of attention. In: New York: Oxford Univ Pr; 1994;p. 250
  42. Dikmen SS, Machamer JE, Powell JM, Temkin NR. Outcome 3 to 5 years after moderate to severe traumatic brain injury. Arch Phys Med Rehabil. 2003;84:1449–1457
  43. Gronwall DM. Paced auditory serial-addition task: a measure of recovery from concussion (PASAT). Percept Mot Skills. 1977;44:367–373
  44. Salmond CH, Menon DK, Chatfield DA, Pickard JD, Sahakian BJ. Changes over time in cognitive and structural profiles of head injury survivors. Neuropsychologia. 2006;44:1995–1998
  45. Spikman JM, Deelman BG, van Zomeren AH. Executive functioning, attention and frontal lesions in patients with chronic CHI. J Clin Exp Neuropsychol. 2000;22:325–338
  46. Wong PP, Monette G, Weiner NI. Mathematical models of cognitive recovery. Brain Inj. 2001;15:519–530
  47. Kolb B, Whishaw IQ. Recovery of function, in fundamentals of human neuropsychology. In:  Kolb B editors. New York: WH Freeman and Co, Worth Publishers; 1996;p. 539–565
  48. Cohen J. Statistical power analysis for the behavioral sciences. Hillsdale: Lawrence Erlbaum; 1988;
  49. Greiffenstein M, Lees-Haley P. Neuropsychological correlates of manganese exposure: a meta-analysis. J Clin Exp Neuropsychol. 2007;29:113–126
  50. Schmitter-Edgecombe M, Marks W, Wright MJ, Ventura M. Retrieval inhibition in directed forgetting following severe closed-head injury. Neuropsychology. 2004;18:104–114
  51. Schmitter-Edgecombe M, Wright MJ. Event-based prospective memory following severe closed-head injury. Neuropsychology. 2004;18:353–361
  52. Mathias JL, Mansfield KM. Prospective and declarative memory problems following moderate and severe traumatic brain injury. Brain Inj. 2005;19:271–282
  53. Simpson A, Schmitter-Edgecombe M. Intactness of inhibitory attentional mechanisms following severe closed-head injury. Neuropsychology. 2000;14:310–319
  54. Schmitter-Edgecombe M. Effects of divided attention on implicit and explicit memory performance following severe closed head injury. Neuropsychology. 1996;10:155–167
  55. Formisano R, Carlesimo GA, Sabbadini M, et al. Clinical predictors and neuropsychological outcome in severe traumatic brain injury patients. Acta Neurochir (Wien). 2004;146:457–462
  56. Vickery CD, Berry DT, Dearth CS, et al. Head injury and the ability to feign neuropsychological deficits. Arch Clin Neuropsychol. 2004;19:37–48
  57. Goldstein F, Levin HS, Goldman WP, Clark AN, Altonen TK. Cognitive and neurobehavioral functioning after mild versus moderate traumatic brain injury in older adults. J Int Neuropsychol Soc. 2001;7:373–383
  58. Ariza M, Junqué C, Mataró M, et al. Neuropsychological correlates of basal ganglia and medial temporal lobe NAA/Cho reductions in traumatic brain injury. Arch Neurol. 2004;61:541–544
  59. Schmitter-Edgecombe ME, Marks W, Fahy JF, Long CJ. Effects of severe closed-head injury on three stages of information processing. J Clin Exp Neuropsychol. 1992;14:717–737
  60. Azouv P. Neuroimaging correlates of cognitive and functional outcome after traumatic brain injury. Curr Opin Neurol. 2000;13:665–669
  61. Henry JD, Phillips LH, Crawford JR, Theodorou G, Summers F. Cognitive and psychosocial correlates of alexithymia following traumatic brain injury. Neuropsychologia. 2006;44:62–72
  62. O'Keeffe F, Dockree P, Moloney P, Carton S, Robertson I. Characterising error-awareness of attentional lapses and inhibitory control failures in patients with traumatic brain injury. Exp Brain Res. 2007;180:59–67
  63. Park NW, Moscovitch M, Robertson IH. Divided attention impairments after traumatic brain injury. Neuropsychologia. 1999;37:1119–1133
  64. Periáñez JA, Ríos-Lago M, Rodríguez-Sánchez JM, et al. Trail Making Test in traumatic brain injury, schizophrenia, and normal ageing: sample comparisons and normative data. Arch Clin Neuropsychol. 2007;22:433–447
  65. Spikman JA, Van Der Naalt JJ, Van Weerden TW, Van Zomeren AH. Indices of slowness of information processing in head injury patients: tests for selective attention related to ERP latencies. J Int Neuropsychol Soc. 2004;10:851–861
  66. Rosenthal R, DiMatteo MR. Meta-analysis: recent developments in quantitative methods for literature reviews. Annu Rev Psychol. 2001;52:59–82
  67. Howell DC. Statistical methods for psychology. 3rd ed.. Belmont: Duxberry Pr; 1992;
  68. Hedges LV, Olkin I. Statistical methods for meta-analysis. New York: New York Academic Pr; 1985;
  69. Rosenthal R. Writing meta-analytic reviews. Psychol Bull. 1995;118:183–192
  70. Orwin RG. A fail-safe N for effect size in meta-analysis. J Educ Stat. 1983;8:157–159
  71. Van Der Werf E. Lack's clutch size hypothesis: an examination of the evidence using meta-analysis. Ecology. 1992;73:1699–1705
  72. Brooks DN, Aughton ME. Cognitive recovery during the first year after severe blunt head injury. Int Rehabil Med. 1979;1:166–172
  73. Vakil E, Weisz H, Jedwab L, Groswasser Z, Aberbuch S. Stroop color-word task as a measure of selective attention: efficiency in closed-head-injured patients. J Clin Exp Neuropsychol. 1995;17:335–342
  74. Schretlen DJ, Shapiro AM. A quantitative review of the effects of traumatic brain injury on cognitive functioning. Int Rev Psychiatry. 2003;15:341–349
  75. Bramlett H, Dietrich D. Quantitative structural changes in white and gray matter 1 year following traumatic brain injury in rats. Acta Neuropathol. 2002;103:607–614
  76. Dixon CE, Kochanek PM, Yan HQ, et al. One-year study of spatial memory performance, brain morphology, and cholinergic markers after moderate controlled cortical impact in rats. J Neurotrauma. 1999;16:109–122
  77. Wilde EA, Bigler ED, Pedroza C, Ryser DK. Post-traumatic amnesia predicts long-term cerebral atrophy in traumatic brain injury. Brain Inj. 2006;20:695–699
  78. Trivedi MA, Ward MA, Hess TM, et al. Longitudinal changes in global brain volume between 79 and 409 days after traumatic brain injury: relationship with duration of coma. J Neurotrauma. 2007;24:766–771
  79. Ng K, Mikulis DJ, Glazer J, Kabani N, Till C, Greenberg G, et al. Magnetic resonance imaging evidence of progression of subacute brain atrophy in moderate to severe traumatic brain injury. Arch Phys Med Rehabil. 2008;89(12 Suppl 2):S35–S44
  80. Mahncke HW, Bronstone A, Merzenich MM. Brain plasticity and functional losses in the aged: scientific bases for a novel intervention. Prog Brain Res. 2006;157:81–109
  81. Mahncke H, Connor B, Appelman J, et al. Memory enhancement in healthy older adults using a brain plasticity-based training program: a randomized, controlled study. Proc Natl Acad Sci U S A. 2006;103:12523–12528
  82. Evans J, Bateman A, Turner GR, Green RE. Cognitive remediation in aging and ABI: a challenge of negative plasticity? (A review of Mahncke, Bronstone and Merzenich). Neuropsychological Rehabilitation, 18:372-84.
  83. Larson MJ, Perlstein WM, Demery JA, Stigge-Kaufman DA. Cognitive control impairments in traumatic brain injury. J Clin Exp Neuropsychol. 2006;28:968–986

 No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit on the authors or on any organization with which the authors are associated.

PII: S0003-9993(08)01476-7

doi:10.1016/j.apmr.2008.07.007

Archives of Physical Medicine and Rehabilitation
Volume 89, Issue 12, Supplement , Pages S69-S76, December 2008

Article Tools