Update on the efficacy of cognitive rehabilitation following moderate to severe traumatic brain injury: a scoping review

Objectives: To identify, categorize and analyze the methodological issues of cognitive rehabilitation of patients with moderate to severe traumatic brain injury and

Cognitive disorders after a traumatic brain injury (TBI) have been well described over the last decades. Long-term memory, attention, processing speed, executive functions and self-awareness disorders are frequent and related to the high frequency of temporal and frontal lesions 1 . Cognitive sequelae commonly persist several years after a moderate to severe TBI 2,3 , impacting vocational integration and quality of life 4,5 . Cognitive rehabilitation aims to decrease acquired neurocognitive impairment and disability using various and complementary approaches 6 .
Interventions could aim to train or strengthen impaired cognitive functions and/or to implement compensatory mechanisms in addition to external aids 6 . Metacognitive strategies are also trained in order to facilitate the transfer to different environmental contexts [7][8][9] .
Over the last years, the literature has provided quantitative data about cognitive rehabilitation after TBI, leading to a better understanding of the underlying cerebral mechanisms and the development of new interventions. Results were reported across reviews, systematic reviews, meta-analyses and scoping reviews. The most consequent systematic review was conducted by the Cognitive Rehabilitation Task cognitive reserve and the presence of associated psychiatric comorbidities. They also recommended including the frequency and intensity of cognitive rehabilitation as covariates in statistical models. Furthermore, several scoping reviews addressed complementary aspects of TBI, such as societal dimensions [13][14][15] , neurological and neuropsychological patterns [16][17][18]  The literature about cognitive rehabilitation following TBI is vast. Reviews on this subject usually analyze the content of rehabilitation to derive recommendations for clinical practice. Here, we chose to focus on methodological criteria to determine the level of scientific evidence of these studies. The most recent substantial systematic review on this subject includes published articles up to 2014 10 . In this paper, we aimed to review the scope of interventions in cognitive rehabilitation since 2015.
Moreover, we chose to select studies including only patients with TBI and to exclude the stroke population in order to limit the heterogeneity of the underlying physiopathology of cognitive disorders. We also excluded the mild TBI population because the functional and cognitive outcomes differ from moderate to severe TBI 26 .
Scoping review was an appropriate approach to map the scope and nature of research in cognitive rehabilitation after TBI, summarize research findings and identify gaps in the existing literature. In order to guide our search, we addressed four main questions: (i) Which cognitive domains does cognitive rehabilitation focus on? (ii) What are the characteristics of interventions in cognitive rehabilitation? (iii) What are the outcome measures used by authors? (iv) What is the efficacy of cognitive rehabilitation?

Methods
The scoping review was based on the framework developed by Arksey & O'Malley 27 including the successive stages described below.

Search strategy
A systematic search of publications listed in the Pubmed (via Medline) and PsycINFO databases was conducted in August 2021 using the keywords "cognitive rehabilitation" (OR "cognitive remediation," "cognitive intervention," "cognitive training," "cognitive treatment") AND "traumatic brain injury." The following terms were excluded from the systematic search: "children," "pediatric," "concussion," "mild" and "animal." The scope of the search went from January 1, 2015, to July 31, 2021.

Inclusion and exclusion criteria
Inclusion criteria were: (i) studies including adults or adolescents, no younger than 15 years old, with moderate to severe TBI. The Mayo Classification System criteria were used to define moderate to severe TBI: loss of consciousness lasting 30 minutes or more and/or post-traumatic anterograde amnesia lasting 24 hours or more and/or worst Glasgow Coma Scale score less than 13 in the first 24 hours and/or imaging evidence of intracranial pathology (intracerebral hematoma, subarachnoid hemorrhage, cerebral contusion, etc…) 28 . We also reported for each article if brain lesions were identified by authors through computed tomography / magnetic resonance scanning (Table 1). In a context of mixed samples including several acquired brain injuries, moderate to severe TBI should be the most represented group; (ii) patients had to be included at least 3 months after the onset; (iii) interventions had to investigate the rehabilitation of cognitive functions; (iv) effects of cognitive rehabilitation had to be documented by quantitative or qualitative comparisons throughout follow-up; (v) interventions had to be conducted in a rehabilitation center, ambulatory care or at home.
Reviews and study protocols were excluded from this research, as were those not written in the English language. Then, for all citations, two authors (AJ, ML) conducted an abstract review and excluded articles that did not meet the eligibility criteria. All remaining citations underwent a full text review.

Data analysis
For each of the four research questions, criteria of analysis were defined and collected in order to classify the characteristics and level of evidence of the reviewed studies.

Cognitive domains targeted by cognitive rehabilitation
All cognitive functions targeted by rehabilitation were listed. When several cognitive functions were trained, we registered all of them. We consider interventions to be "global training" interventions when they focused on three or more cognitive functions, or when the aim was defined with the generic term "cognitive skills."

Characteristics of cognitive rehabilitation
Types of cognitive rehabilitation were divided into three categories of interventions.
Cognitive training was defined as repetitive exercises without any explicit mention of metacognitive strategy training. Integrative cognitive intervention referred to interventions that explicitly combined the training of cognitive functions and metacognitive strategies. Finally, external aids training corresponded to the use of external compensatory mechanisms such as notebooks, cell phone applications and alarms.
We also identified combined approaches, which referred to cognitive rehabilitation associated with other interventions like pharmacotherapy or non-invasive brain stimulation (NIBS).
Three other parameters of cognitive interventions were analyzed: the delivery mode including group versus individual sessions, the length and the intensity. Length was studied by distinguishing very short (1 week or less), short (1 week to 1 month), moderate (1 to 3 months) and high (more than 3 months) duration. Intensity was classified as low (1 session per week), moderate (2 sessions per week) or high (3 or more sessions per week).

Behavioral examination and neuroimaging as outcome measures
Concerning behavioral outcome measures, four types of assessment were distinguished: (i) neuropsychological examination including standardized neuropsychological tests; (ii) ecological neuropsychological examination including standardized tests and/or experimental ecological tasks with reference to daily life situations; (iii) self-reporting of cognitive complaints, social participation in everyday activities and quality of life; (iv) relative-reporting of patient's difficulties in daily life.
We also counted the number of these types of assessment for each study in order to attest to the exhaustiveness of the assessment.
Neuroimaging outcome measures were classified as structural and/or functional imaging and/or electroencephalography (EEG).

Efficacy of cognitive rehabilitation
The efficacy of cognitive rehabilitation was analyzed according to three main criteria and associated sub-criteria detailed below. A coding grill was used for the extraction of these methodological criteria.
The outcome measures were the first criteria. We first pointed out the results showing a significant improvement in at least one of the outcome measures defined by the authors. Quantitative and qualitative improvements were coded when collected. Second, if a significant and/or clinically relevant change was reported, we distinguished whether it was in the primary or secondary outcome measures.
The internal validity of reviewed studies was assessed as secondary criteria, based on the classification used by Cicerone et al. in systematic reviews 6 . According to this classification, studies were classified as class I when they were well designed, prospective, randomized controlled trials. Class II referred to prospective, nonrandomized cohort studies, retrospective, non-randomized case-control studies or multiple baseline studies that allowed a direct comparison between treatment conditions. Class III included clinical series without concurrent controls or singlesubjects designs. In a second step, we also detailed the control group design, distinguishing active, passive or no control group. We considered it an active control group when patients participated in usual care or unspecific activities. A passive control group referred to a waiting list or a no-treatment phase.
The statistical analysis was the third criteria. As proposed by Cicerone et al. 29 , comparisons of between-group treatment conditions were considered as a higher level of methodological quality compared to within-group comparisons. We also identified whether or not the authors applied an intention-to-treat (ITT) analysis.
Finally, we analyzed whether the effect size and measures of variability such as confidence intervals were reported.

Charting the data
In accordance with the PRISMA-Scr guidelines 30 , a flow diagram was used in order to illustrate study selection ( Figure 1). The level of evidence for the efficacy of cognitive rehabilitation was also charted ( Figure 2). Figure 2 details the number of studies that met each pre-cited methodological criterion and associated sub-criteria.
For each study, the key characteristics of TBI participants, cognitive rehabilitation, experimental design, intervention, neuropsychological outcome measures and significant main results were collected and summarized in Table 1.

Results
Between January 2015 and July 2021, 458 studies were published in the Pubmed (via Medline) and PsycINFO databases. We found 31 duplicates across the two databases and removed them (Figure 1). Four hundred twenty-seven records were reviewed by title and abstract and 330 were excluded based on the inclusion and exclusion criteria. Ninety-seven articles were assessed by full text review. In the end, 46 studies were included in the scoping review.

Type of cognitive interventions
In this review, integrative cognitive interventions concerned 48% of studies (n=22), cognitive training was reported in 37% of studies (n=17) and external aids training was described in 11% (n=5). Two studies (4%) did not detail the type of intervention 31,32 . The effects of combined interventions were examined in 4 studies 33-36 , in which cognitive rehabilitation was associated with pharmacotherapy 36 , repetitive transcranial magnetic stimulation (rTMS) 33 or transcranial direct current stimulation (tDCS) 34,35 .

Methodological parameters of cognitive rehabilitation
Among the 46 reviewed studies, we showed that individual sessions were used in 83% of the studies (n= 38), whereas group sessions were only used in 11% (n=5).
Six percent (n=3) of the studies combined individual and group sessions.
Furthermore, the length of interventions was heterogeneous, ranging from 5 days 34 to 15 months 37 . Fifty-four percent of studies (n=25) proposed an intervention which lasted between 1 and 3 months. Shorter interventions lasting 1 week to 1 month were found in 24% of studies (n=11). Finally, cognitive rehabilitation interventions including a duration of less than 1 week or longer than 3 months were found in 1 (2%) and 8 studies (18%), respectively. In one study (2%), this methodological feature was not detailed 38 .
Concerning the intensity of interventions, 26% of the reviewed studies (n=12) proposed two sessions per week and 44% (n=20) proposed three or more sessions per week. Conversely, 11% (n=5) included only one session per week. In one study, the intensity was variable and progressively decreased among each phase of cognitive rehabilitation 39 . Finally, 17% (n=8) did not describe this methodological point.
Our results indicated that 9 out of the 46 studies (20%) did not detail both the length and intensity of the interventions. Among studies that detailed length and intensity, the most common design combined 3 or more sessions per week over 1 to 3 months and was found, in this scoping review, in 10 studies.

Behavioral examination and neuroimaging as outcome measures
The effects of cognitive rehabilitation were mostly measured with standardized neuropsychological tests in 41 out of the 46 studies (89%). Ecological neuropsychological examination was used in 35% of studies (n=16). Fifty percent (n=23) included a self-report questionnaire, whereby cognitive complaint was assessed in 16 studies (70%) and quality of life was measured in 7 studies (30%).
Finally, reporting by relatives was used in 35% of studies (n=16).
Thirty-seven percent of studies (n=17) used one of these 4 types of measures, 28% of studies (n= 13) used 2 types of measures and 24% (n=11) used 3 types of measures. In contrast, 11% of studies (n=5) proposed an exhaustive evaluation with these 4 types of measures.
Neuroimaging outcome measures used as brain markers of cognitive rehabilitation were reported in 20% (n=9) of studies, whereas EEG was performed in only two studies 38,40 . More specifically, resting state functional magnetic resonance imaging (fMRI) 41 , regional cerebral blood flow 33 and brain activation during an fMRI cognitive task 34,42 were analyzed in 4 studies. Structural MRI data were reported in 4 studies 43-46 . Only one paper combined diffusion tensor imaging, attention-related fMRI and resting-state fMRI sequences 47 .

Efficacy of cognitive rehabilitation
According to Cicerone's criteria for evidence-based classes 6  Ninety-three percent of studies reported significant cognitive improvement on at least one outcome measure, among which 19 studies described clinical improvement on the primary outcome independently of statistical change (Figure 2). Within these studies, 10 were classified as class I and involved an active control group 34,36,42,44,48,49,[52][53][54]56 . Then, with regard to statistical analysis, these 10 studies applied between-groups comparisons to assess the efficacy of treatment, among which 7 used an intention-to-treat analysis (15% of reviewed studies) 34,42,44,49,52,53,56 . Medium to large effect sizes were reported in 5 out of these 7 studies 44,49,52,53,56 and the confidence interval was reported in only one out of these 7 studies 49 .

Discussion
This scoping review was conducted starting with 2015, after the most recent systematic review 10 , in order to identify and characterize studies evaluating cognitive rehabilitation following a moderate to severe TBI, to summarize the cognitive approach used and the domains investigated and to analyze their efficacy.
Memory, attention and executive functions were most often targeted in individual sessions adopting an integrative cognitive approach. Cognitive interventions were mainly temporally distributed with 3 or more sessions per week over 1 to 3 months.
One or two behavioral outcome measures were mostly preferred by authors to assess the efficacy of intervention, while neuroimaging outcome measures were rarely used. The review found clinically significant effects of cognitive rehabilitation after a moderate to severe TBI in a very large part of reviewed studies (93%), among which 41% described an improvement on the primary outcome measure. The high number of positive published results could be the sign of a publication bias according to Dwan and colleagues' conclusions in 2013 77 . Nevertheless, when methodological criteria for the level of evidence were controlled (outcome measures, internal validity and statistical analysis) a significant decrease was observed, from 93% to 15%. This significant decrease is unsatisfactory and highlights the methodological requirements for future studies. Challenges in TBI rehabilitation imply that cognitive interventions must be based on a robust experimental design to prove their efficacy and to replicate the findings on which recommendations for clinical practice could be finally derived. Therefore, this scoping review provides a complementary approach to prior systematic reviews 6,10-12 by identifying five key methodological points.

Specific experimental designs for cognitive rehabilitation of TBI patients
In this scoping review, 41% of reviewed studies were classified as class I. This result highlights a continuing upward trend of randomized controlled trials in cognitive rehabilitation. Indeed, Cicerone et al. reported a percentage of class I studies ranged from 17% 6,12 to 20% 11 until 2008, which increased to 36% between 2009 and 2014.
RCTs were crucial for evidence-based studies but not always relevant in rehabilitation practice, where double blind was sometimes not feasible 11 because the therapist was systematically aware of the hypothesis underlying the contents of intervention. Furthermore, experimental and control groups have to share common methodological parameters such as delivery mode, length and intensity of rehabilitation to allow between-groups comparisons 78 . A major advance in the literature is the presence of an active control group to attest to the specificity of the experimental intervention and to rule out the nonspecific effects of global cognitive stimulation, such as treatment effect, motivational or novelty effect and Hawthorne effect 78 . Statistically, the efficacy of interventions cannot be only demonstrated using within-group analysis. Improvements must be specific to the experimental intervention and thereby confirmed with between-group comparisons. Effect sizes, rarely presented in reviewed studies, are also a supplementary indicator of the efficacy of cognitive interventions and should be systematically added in the future.
All these methodological points were controlled in one study 49 , in which the authors investigated the added effects of psychoeducation and metacognitive strategy training in an experimental group compared to an active control group with cognitive rehabilitation including non-training-oriented tasks, with a positive effect for patients.
Finally, a challenge for further group studies may be the individualization of the cognitive intervention regarding cognitive profiles and complaints in order to compensate for the clinical heterogeneity of TBI. Two main solutions could be proposed for greater methodological relevance. The first is to constitute toolboxes for each cognitive domain, including standardized exercises with increasing levels of difficulty, like those developed by Visch-Brink et al. 79 and Van Rijn et al. 80 in aphasia therapy 81 . For a single cognitive function rehabilitated, the therapist will be able to The second level of combination was between group and individual sessions. Even if, in this scoping review, results showed that individual interventions remained the majority (83%), a combined approach of these two delivery modes was proposed in 3 studies, but its specific benefits were not analyzed 61,73,74 .
The third level concerned the use of combined interventions. Only 4 studies proposed combined rehabilitation with pharmacotherapy 36 or NIBS 33-35 . The heterogeneous designs and the low statistical power of these studies call for replication.

Specific effects of length and intensity of cognitive rehabilitation
The main temporality reported by this scoping review included a moderate duration (i.e. ranging between 1 and 3 months) with a high intensity (i.e. 3 or more sessions per week). This choice seems related to clinical relevance and feasibility in clinical research protocols. As mentioned by Cicerone et al. 10 , the intensity and length of the cognitive interventions must be studied in order to determine their respective contribution to the efficacy of the rehabilitation and thus have to be integrated into statistical models. None of these two parameters were analyzed across all reviewed studies. Furthermore, Chiaravalotti et al. 49 have investigated the use of monthly booster sessions proposed over 5 months, after memory training with 10 sessions over 5 weeks. These focused on applying trained memory strategies in daily life.
Although the authors reported no effect of these booster sessions during follow-up, it seems very useful to check the implementation and efficacy of trained cognitive strategies in daily living.
In addition to length and intensity parameters, future studies should investigate the severity of cognitive impairment at inclusion, the delay from the injury or fatigability as contributing variables in determining the dynamic of the intervention.

Selection of outcome measures as a key experimental point
The choice of outcome measures is a key methodological point as well as the categorization into classes I to III for evidence-based medicine. Assessment using standardized neuropsychological examination was the most frequently reported (89%), followed by self-report questionnaires (50%), ecological neuropsychological assessment (35%) and relative-report questionnaires (35%). An exhaustive neuropsychological examination of all cognitive domains could contribute to demonstrating the benefits of therapy on trained as well as on untrained functions.
Moreover, after a wash-out period, a follow-up assessment may show maintained benefits of rehabilitation. However, it has been well described that standardized pencil-paper neuropsychological performance test could not exactly reflect those obtained in daily contexts, especially in executive functions assessment 85 . In this way, an ecological cognitive assessment could be a sensitive measure to predict real-life performance 86 . Ecological tests such as the Test of Everyday Attention 87 or the Rivermead Behavioral Memory Test 88 were frequently proposed in the reviewed studies but remained in non-ecological environments and encompassed a restricted representation of daily life tasks. Conversely, the Multiple Errands Test (MET) 85 , which was not reported here, implies daily life activities, takes place outside of the rehabilitation sites and offers a more sensitive image of executive disorders 85 . MET should be combined with person-centered assessment to improve the clinical relevance of the evaluation. The Goal Attainment Scaling (GAS) 89 , derived from occupational therapy, makes it possible to set personalized goals with the patient as well as 5 levels of predicted attainment for a sensitive evaluation of progress 90  Nevertheless, no details were given about cognitive scores on task-related functional activation.
Brain imaging constituted a promising method but further research is needed to identify potential contributions. Structural and functional MRI continue to contribute to a better understanding of TBI physiopathology. These techniques illustrate the brain reorganization and the dynamics of plasticity mechanisms that could be associated with short and long-term cognitive changes.
Brain imaging may also participate in the identification of potential modulators of recovery trajectories after TBI 92 such as brain reserve, including measures of specific patterns of gray matter volume, cortical thickness, synaptic integrity or white matter microstructural properties. Neuroimaging could make multiple contributions but at this time its use as a measure of the efficacy of an intervention should be done in combination with cognitive measures.

Study limitations
A few main limitations were identified in the scoping review. The first concerned the search strategy, which focused on only two databases and did not include the gray literature. As reported, the impact of publication biases could contribute to an inaccurate picture of the literature on cognitive rehabilitation. Second, only one reviewer performed data extraction and analysis. While we made efforts to define criteria precisely to assess the methodological quality of the reviewed studies, there may be subjective interpretation involved in this process.

Conclusions
This scoping review highlights the persistent and growing interest in cognitive rehabilitation with major methodological improvements in the design of studies for moderate to severe TBI since 2015. In consequence, this led to higher number of studies that show an improvement in the primary outcome measures after cognitive rehabilitation. Our findings make it possible to identify three methodological criteria and sub-criteria for determining the level of evidence of cognitive interventions and could be used in future studies. Our approach is complementary to the prior systematic reviews 6,10-12 which were mainly focused on the content of interventions.
Methodological efforts must be continued, and combined interventions studies must be proposed. Individualized cognitive rehabilitation also remains a challenge.
Outcome measures must be well selected, including neuropsychological tests in ecological and non-ecological environments, patient-and relative-reports.
Rehabilitation of social cognition and emotion regulation should be better investigated. The results of this scoping review now need to be confronted with systematic reviews and meta-analyses.    Score on executive function tests.
Self-report quality of life.