Rehabilitation of executive function in chronic paediatric brain injury: a randomized controlled trial

This is an evaluator-blinded, parallel-RCT, with a previously published trial protocol [46], set at two paediatric hospitals in Norway, St. Olavs hospital, Trondheim University Hospital and Oslo University Hospital, Rikshospitalet. The sample size was initially calculated using the Global Executive Composite (GEC) from the BRIEF (parent report) [47], assuming an effect size of .70, significance level of .05, 80% power, and a t-test for difference in means, yielding samples of 32 individuals per group. Later, however, the Behavioral Regulation (BRI) and Metacognition (MI) indexes from BRIEF (that form the GEC), were considered to be more appropriate as co-primary outcomes. For practical and logistical reasons, it was not feasible to recalculate and increase the sample size of the study. Therefore, this study should not be considered as confirmatory as it is likely to be underpowered to find differences between the two groups, if they exist.

Participants were randomized to either pGMT or pBHW in a 1:1 ratio, by blocks of 4, and stratified by hospital site (Trondheim vs Oslo) and age at intervention (10–13 years vs 14–17 years). A computer-based algorithm was set up for randomization by an independent allocator. The following blinding procedures were applied to reduce systematic bias: (i) Treatment allocation was not disclosed to participants and families, and they were encouraged not to discuss course content outside their group. (ii) “Brain training” was used consistently in both groups. (iii) Independent, trained, and blinded test technicians conducted assessments. (iv) Group therapists were blinded to all assessments. (v) Blinded data monitoring and statistical analysis were implemented. Finally, interventions were arranged in random order.

Written informed consent was provided from primary caregivers of participants (< 16 years) and from participants and their caregivers (> 16 years). Study procedures and monitoring according to Norwegian Clinical Studies Infrastructure Network procedures.

Participants eligible for a written invitation were identified based on discharge diagnosis and medical records (Fig. 1) [46]. Further eligibility screening was done in an initial telephone contact and more thoroughly in a semi-structured screening interview [see Additional file 1] with primary caregivers (or participants surpassing 16 years). Successively, after written informed consent, eligible individuals were designated a study number and randomized.

Eligible children were aged 10 to 17 years, diagnosed with pABI (TBI, brain tumour, cerebrovascular accident, hypoxia/anoxia, or brain infection/inflammation), at intervention > 12 months post-insult or > 12 months stable brain tumour situation/ended cancer therapy, with reported executive function difficulties in daily life as described in the screening interview by parents (or participants). Exclusion criteria included pABI before the age of 2 years; cognitive (including memory), physical, or language impairments affecting the capacity to follow educational goals of peers and attend regular classroom teaching; pre-injury neurological disease; severe psychiatric disorder and/or on stimulant medication; recently detected brain tumour relapse; or not fluent in Norwegian. To our knowledge, none of the participants had prior to the study received specific cognitive rehabilitation. Recruitment lasted from November 2017 until May 2019 and was ended at achieved sample size.

The two group-based interventions compared in this study were the metacognitive pGMT and an active control (pBHW) involving the use of educational materials and lifestyle topics, similar to many other psychoeducational ABI rehabilitation programmes [48]. While the pBHW programme was not specifically aimed at remediating EF, the pBHW programme was tailored to keep non-specific factors constant (i.e., structure and intensity/duration of training, between session-assignments, involvement of parents and teachers, professional attention, group dynamics).

The manualized GMT®, adult protocol has previously been translated into Norwegian [49, 50]. A preliminary paediatric protocol was developed and piloted by our research group [29] and then further refined based on the feedback for the present study. The pGMT promotes inhibitory control and metacognitive strategies and facilitation of goal-achievements, in addition to mindfulness training [51]. The protocol consists of 7 sessions, each of 2-h duration (Fig. 2). Original GMT materials were made more age-appropriate (e.g., examples and discussions more related to school and leisure activities, children portrayed in illustrations instead of adults, language simplified, more visual materials, and movie clips were included).

The original adult Brain Health Workshop (BHW) protocol is a manualized psychoeducational intervention developed to match GMT for non-specific factors [20, 24], and adapted into a paediatric version in a similar manner as described for pGMT. pBHW represents an active control intervention including educational materials addressing brain injury and (dys)function, plasticity, memory and learning, EF, fatigue, and lifestyle issues (e.g., stress, sleep, exercise and nutrition) [20, 24] (Fig. 2).

The two group-based interventions were corresponding regarding structure, intensity, duration, therapist contact, and access to distributed materials. Treatment fidelity was ensured by two trained therapists at each site and standardization of materials to warrant consistent and reliable implementation. Accompanying all sessions, pre-prepared manualized PowerPoint slides were used to standardize content across groups and sites. Additionally, participant workbook inserts specific for each session were handed out at the start of each session. Both interventions included discussions of participants’ real-life experiences and attention deficits, in-session practice, between-sessions exercises, monitoring of activity, education of parents and teachers, and periodic alertness cueing following the fourth session (SMS; stating the key term “STOP” in pGMT and the more general “Brain training” in pBHW) to facilitate transfer to daily living [52, 53]. If a participant was prevented from attending a group session, an individual session of the missed module was offered before the following to ensure progress. Parents received a 1-h review after every paediatric session with group therapists allowing questions and exchange of experience. In the 4 weeks following sessions, telephone counselling was conducted to promote use of the interventions. All designated teachers received written information about the study and a 1-h telephone counselling post sessions that included a brief review and encouragement to implement intervention principles [30]. Last follow-up was conducted in November 2019.

Medical characteristics at the time of insult were extracted from the medical records (Table 1). At start of intervention (T₁), systematic assessments of baseline sociodemographic factors and medical history as well as a standardized physical examination was performed for all individuals by a study nurse and a paediatrician. Fatigue was measured with the Pediatric Quality of Life Inventory-Multidimensional Fatigue Scale (PedsQL MFS, parent report) [54]. Parents rated each item according to their child’s function the prior month (e.g., “Feels too tired to spend time with his/her friends”). Presented are reversed total score, linearly transformed to a 0–100 scale. A higher score indicates fewer fatigue symptoms. General intellectual ability was measured by full scale intelligence quotient (FSIQ) from Wechsler Intelligence Scale for Children-Fifth edition (WISC-V), with normalized IQ distribution (M = 100, SD = 15). Blinded baseline assessment was performed the day prior to initiation of the interventions as several participants had a long commute to the research site and reported fatigue, thus avoiding the burden of an additional travel just for the assessment.

An accurate and valid evaluation of EF benefits from assessments at different levels to obtain clinically useful functional ability information [38, 55]. Since pGMT addresses all three ICF- levels [41], EF outcome measures pertaining the three levels were included by questionnaire (activity level, primary outcome measure), performance-based tests (impairment level, secondary outcome measure), and a complex naturalistic task (participation level, secondary outcome measure) [38, 43, 56]. Investigators blinded to the intervention performed assessments at pre-intervention (baseline, T₁), post-intervention (T₂), and at 6-month follow-up (T₃).

The BRIEF parent report [47] at T₃ was employed as the primary EF outcome measures using the two indexes, the BRI and the MI as co-primary endpoints (raw scores BRIEF_BRI and BRIEF_MI). Since the BRIEF has been used as a primary outcome in adult GMT studies [49] results can potentially be compared. The BRIEF is an 86-item standardized questionnaire designed to capture parent perceptions of a child’s everyday EF. Each item’s frequency of occurrence is rated on a 3-point Likert scale from 1 (never) to 3 (often), higher scores indicate greater dysfunctions. The BRIEF has proven applicable to several clinical groups [57] and has demonstrated adequate internal consistency, inter-rater and test-retest reliability [47, 58] for parent report, in addition to self- and teacher reports (secondary outcomes).

Secondary outcome assessments of EF at the impairment level were conducted with standardized neuropsychological tests pertaining updating (digit span, total score, raw score (Wechsler Intelligence Scale for Children-Fifth edition, WISC-V)) [59]; shifting (Trail Making Test 4 (TMT 4); total time; raw scores from the Delis–Kaplan Executive Function System (D-KEFS) [60]; inhibition (Conners’ Continuous Performance Test, 3ed (CPT-III) Commissions, T-scores) [61]; and executive attention (Color Word Interference Test 4 (CWIT 4, D-KEFS), total time, raw scores) [60]. All tests are paper and pencil tests, except CPT-III, which is a computerized assessment of different aspects of attention. Higher test scores on CPT-III and digit span indicated better performance, whereas higher scores on TMT 4 and CWIT 4 indicated poorer performance.

Finally, at the participation level, we included a complex open-ended task, the Children’s Cooking Task (CCT), total errors, and raw scores [45, 62]. The task (cooking) required multitasking in a naturalistic setting (kitchen) with aspects of novelty and distractions and, hence, less external control and structure. The CCT has demonstrated sensitivity to executive dysfunction in pABI [63].

Categorical data were summarized by counts and percentages. Continuous data were summarized by mean and standard deviation (SD) or median and interquartile range (IQR), presented for each group. Any baseline differences between the groups were viewed as incidental [64] due to the randomization, but noted. A statistical analysis plan (SAP) was outlined in advance and registered at ClinicalTrials.​gov prior to un-blinding of the data. The primary analysis was carried out blind of allocation by a statistician. The within and between differences of the two co-primary endpoints constituting BRIEF (raw scores BRIEF_BRI and BRIEF_MI) were estimated by linear mixed models [65] in a full analysis set (FAS). All randomized individuals with post-baseline outcome data (ICH E9) [66] were included, representing an intention-to-treat approach. The model included baseline raw scores as a covariate [67], and treatment group, time (post-treatment (T₂) and at 6 months follow-up (T₃)), interaction between time and treatment, and randomization strata (1) site (Trondheim or Oslo) and (2) age at intervention (10–13 or 14–17 years) as fixed factors. Time was included as categorical with unstructured covariance as the starting point. According to the SAP, the per-protocol population included participants who had missed a maximum of 2 out of the 7 group sessions. Since all participants who completed the allocated intervention (n = 73) underwent all 7 modules the main analysis (FAS) and per-protocol were identical. The two co-primary endpoints from BRIEF (_BRI and _MI) were analysed using the Hochberg procedure to control for the type I error rate [68]. The global null hypothesis of no difference between groups was rejected if the test of either endpoint was statistically significant at the two-sided 0.025 level, or if both tests were significant at the two-sided 0.05 level. Differences between treatment groups were accordingly estimated with 95% confidence intervals. Exploratory subgroup analyses of the primary outcome measures were performed according to strata and sensitivity analyses were performed by exploring the effect of additional covariates thought to be of prognostic importance [69]: type of injury (aetiology), age at injury (years), time since injury (years), and reported fatigue at intervention. Secondary outcomes of self and teacher reports (raw scores BRIEF_BRI and BRIEF_MI) were also analysed by the linear mixed model. Secondary outcomes pertaining performance-based neuropsychological tests at 6-month follow-up were analysed by linear regression, with baseline raw scores, treatment group, and strata as covariates. For secondary endpoints, the number of outcomes were theory-driven, pre-planned and limited (only one outcome per function) to avoid an inflation of type I error. Estimates of treatment differences were presented with 95% confidence intervals, and tests were performed using a two-sided significance level of 0.05. No adjustment of multiplicity was planned. Distributional assumptions were checked by visual inspection of residual plots. No imputation of missing scores was made. Main analysis was conducted March 2020 using SAS v9.4. IBM-SPSS version 26 and Stata16 were used for secondary analyses.

Seventy-three participants (96%) completed the allocated intervention (Fig. 1). There was no attrition post-intervention, and only two participants (one in each group) failed to attend at the 6-month follow-up due to cancer recurrence, resulting in 93% attendance.

The results show no significant difference between the two intervention groups for the primary outcome (changes in parent-reported BRIEF raw scores from baseline (T₁) to 6 months follow-up (T₃)). Thus, estimated between-group differences in the linear mixed model for BRIEF_BRI was − 2.3 (95% CI − 5.1 to 0.6), p = .12, and the corresponding result for BRIEF_MI was − 1.4 (95% CI − 8.5 to 5.8), p = .71 (Fig. 3). Exploratory stratified subgroup analyses and sensitivity analyses did not differ meaningfully from the main result [see Additional file 2, 3]. Although no difference in effect for the two interventions was demonstrated, the results show significant decrease in parent-reported executive dysfunction for both interventions. In linear mixed models, the estimated changes from baseline (T₁) to 6-month follow-up (T₃) with 95% CIs and p values for BRIEF_BRI were − 3.8 (− 5.6, − 1.9, p < .001) for pGMT and − 5.7 (− 8.5, − 2.9, p < .01) for pBHW. The corresponding results for BRIEF_MI were − 6.9 (− 10.9, − 2.9, p = .001) for pGMT and − 8.9 (− 13.2, − 3.9, p = .001) for pBHW.

As presented in Fig. 3, comparing pGMT and pBHW at T₃ mixed modelling revealed no difference in either self- or teacher reports for BRIEF_BRI or BRIEF_MI, which supports the primary results of no difference between interventions.

Results from performance-based tests and the complex naturalistic task are presented in Table 2 and demonstrates that at 6-month follow-up, the pGMT group had improved inhibition scores (difference in T-score at T₃, 5.2 (95% CI 1.6 to 8.7); p = .005) and executive attention (difference in raw score at T₃, 9.9 (95% CI 2.2 to 17.6); p = .01), but not updating (difference in raw score at T₃, − 0.2 (95% CI − 1.7 to 1.4); p = .83) or shifting (difference in raw score at T₃, − 4.5 (95% CI − 18.9 to 9.9), p = .54). The pBHW enhanced performance on the complex naturalistic task (CCT) (difference in raw score at T₃, − 4.3 (95% CI − 8.4 to − 0.3), p = .04).

One psychological reaction was reported in the pGMT group at the end of intervention. It was assessed and handled according to study protocol and good clinical practice [70]. Although symptoms originated before study inclusion, it was recorded as an adverse event (moderate) with uncertain connection to the intervention.