Computer-Based Cognitive Training in Children with Primary Brain Tumours: A Systematic Review

Simple Summary Brain tumour survivors are often burdened by late sequelae, especially neurocognitive deficits, ultimately affecting their quality of life. For many years, treatments for neurocognitive impairments have been limited to educational, pharmacological, home-based interventions, or clinic-based cognitive rehabilitation, but these treatment modalities showed several limits. More recently, cognitive rehabilitation through digital tools to increase cognitive performance through exercises and games is spreading in experimental clinical settings. However, since these are innovative interventions, there is a need to further investigate their effects on cognitive outcomes and quality of life for children with brain tumours. Therefore, in this systematic review, we analyse the current evidence and trends regarding computer-based cognitive rehabilitation in paediatric patients diagnosed with, or survivors of, brain tumours. To our knowledge, this is the first systematic review investigating these new approaches to cognitive rehabilitation in children with brain tumours. Abstract Background: Late neurocognitive sequelae are common among long-term brain tumour survivors, resulting in significantly worse quality of life. Cognitive rehabilitation through specific APP/software for PC/tablets represents an innovative intervention spreading in recent years. In this study, we aim to review the current evidence and trends regarding these innovative approaches. Methods: A systematic literature review was performed. Inclusion criteria were: (i) Studies recruiting patients diagnosed with any brain tumour before 21 years of age; (ii) studies assessing the role of digital interventions on cognitive outcomes. Case reports, case series, reviews, letters, conference proceedings, abstracts, and editorials were excluded. Results: Overall, nine studies were included; 152 patients (67.8% males) with brain tumours underwent a digital intervention. The mean age at diagnosis and the intervention enrolment ranged from 4.9 to 9.4 years and 11.1 to 13.3 years, respectively. The computer-based software interventions employed were: Cogmed, Captain’s Log, Fast ForWord, and Nintendo Wii. Most of these studies assessed the effects of cognitive training on working memory, attention, and performance in daily living activities. Conclusions: The studies suggest that this type of intervention improves cognitive functions, such as working memory, attention, and processing speed. However, some studies revealed only transient positive effects with a significant number of dropouts during follow-up. Trials with greater sample sizes are warranted. Motivating families and children to complete cognitive interventions could significantly improve cognitive outcomes and quality of life.

on children and young adults in experimental clinical settings [9]. Several reviews and meta-analyses on CCTP have been conducted on cognitive training with Cogmed Working Memory Training in patients with neurocognitive disorders. However, most of these works included only/also adult patients, non-computerised cognitive training tools, and no distinction between hospital-and home-based training sections [10][11][12][13]. For these reasons, robust results are still missing and controversial [14].
Furthermore, among several reviews and meta-analyses analysing CCTP with or without game-like design, there is a lack of evidence on the effects of CTTP on the cognitive outcomes and quality of life of children with BTs [13].
This study aimed to review the current evidence and trends regarding computer-based cognitive rehabilitation in paediatric patients diagnosed with, or survivors of, BTs.

Materials and Methods
This systematic literature review was performed according to the methodology described in the Cochrane handbook for systematic reviews [15] and was reported based on the PRISMA statement for reporting systematic reviews and meta-analyses (Table S1) [16]. Additionally, the study protocol was registered in the International Prospective Register for Systematic Reviews (PROSPERO) and was ascribed the CRD42022344520 registration code.
All literature published to July 2021 (updated to June 2022) was retrieved by searching the databases "PubMed", "Cochrane", "APA PsycInfo", and "CINAHL" using the following search terms: (kinect* OR nintendo* OR "gaming" OR exergam* OR exer-game* OR "virtual reality" OR virtual-reality OR "augmented reality" OR softwar* OR "app" OR tablet* OR smartphone* OR smart-phone* OR *game* OR "machine learning" OR machine-learning OR "artificial intelligence" OR artificial-intelligence OR "deep learning" OR deep-learning OR "convolutional neural" OR convolutional-neural* OR "computer vision" OR computer-vision) AND (child* OR infant* OR adolescent* OR pediatri* OR paediatr* OR poediatr*) AND (cancer* OR tumour* OR neopla* OR malignan* OR *irradiation* OR *proton* OR chemother* OR radiother* OR radio-ther* OR chemother*) AND (cereb* OR brain* OR *crani* OR cns OR spin* OR medull*).
No limitations in the search strategy were applied to the publication date, study design, or language. References of considered studies were also searched to identify any further relevant data.
The records identified by the search were uploaded on "Rayyan" [17] to organise the study selection in a more efficient way. The titles and abstracts of the identified records were initially screened and selected by three groups composed of two independent and blinded reviewers (G.R. + A.C., L.T. + D.M., F.S. + G.P.) based on their pertinence to the review topic. Conflicts and disagreements were resolved by consensus.
The following set of pre-defined inclusion criteria were then individually applied to the selected articles in their full-text version: (i) Studies recruiting patients diagnosed with any brain tumour before 21 years of age; (ii) assessing the role of digital interventions based on apps, video games, augmented reality, or any other type of software based on cognitive outcomes. Case reports, case series, reviews, letters, conference proceedings, abstracts, and editorials were excluded. Articles not published in English were removed. Systematic reviews were considered separately to check for the consistency of the data. Data extraction was performed by three reviewers (L.T., G.R. and D.M.). The extracted data are reported in Table 1. From baseline to post-intervention assessment, the intervention group showed greater improvement than the control group on: Attention and working memory (WISC-IV spatial span forward, digit span backward, working memory index, p < 0.05; WISC-IV spatial span backward p < 0.001). Improvements also in executive functioning and attention. Patients with high-risk disease (p = 0.0042) and younger age at diagnosis (p < 0.0001) had more declines in reading during the follow-up. Older patients at diagnosis date had less decline in reading (p = 0.0008) and decoding (p = 0.0367).

Results
Bibliographic searches on literature databases yielded 2033 records. After a first screening, 13 records were selected. Of these, four were further excluded, as they did not meet the inclusion criteria. Overall, nine studies were included. Two of the nine were open-label studies and seven were randomised controlled trials (RCT). All included studies were published in the interval 2011-2021. The flow diagram of included studies is reported in Figure 1. A high agreement (>90%) concerning the inclusion of the records was reported by the reviewers involved in the study selection process (L.T.; G.R.; D.M.) and conflicts in the screening process were resolved by inter-personal discussion. The summary of the characteristics of the included studies are reported in Table 1.

Quality Assessment of the Studies
Quality of the included studies was assessed through the Cochrane Risk of Bias (RoB) tool [18] and reported in Figure 2.
The main reasons associated with a higher quality were: The randomisation of participants before the intervention (reported in 7/9 studies, 78%) and a clear blinding of participants and personnel before and during the outcomes assessment (described in 4/9 studies, 44%). On the other hand, the principal reasons associated with a poorer quality were represented by: (i) Incomplete data regarding participants related to significant drop-outs during the follow-up, and (ii) the absence of randomisation of the participants before the digital intervention.

Population
Overall, 152 patients (67.8% males) with BTs were included and underwent digital intervention. The mean age at diagnosis (not reported in three studies) and at the intervention enrolment ranged from 4.9 to 9.4 years and 11.1 to 13.3 years, respectively. The principal tumour types included were: Medulloblastomas, ependymomas, primate neuroectodermal tumours, astrocytomas, and germ-cell tumours. Five studies also enrolled a control population whose mean age at the intervention ranged from 10.7 to 13.2 years.
In most of the studies, all patients were examined after an average of 5-year intervals at the end of chemo-radiation therapy. In the study by Palmer et al. [22], patients underwent computer-based cognitive training after diagnosis of a primary BT in parallel to appropriate standards of care (SOC), including surgery, cranial and spinal radiation, and/or chemotherapy. In another study [21], authors enrolled patients who underwent cognitive training after surgery (Cases: 12.11 ± 16.97; Controls: 8.72 ± 13.92 months after surgery) in parallel to radiation or medical SOC.
Four studies enrolled mixed populations, including patients diagnosed with primary brain tumours (BT) or acute lymphocytic leukaemia (ALL). When miscellaneous cases were included in intervention studies, the primary justification reported by the authors was the high prevalence of neurocognitive and psychosocial deficits in these populations of patients undergoing chemo-radiation as SOC. However, some authors clarified how patients diagnosed with BTs are more likely to present more extensive and multi-domain cognitive deficits than ALL. Descriptive statistics of mixed populations have been reported, addressing patients diagnosed with BTs only in this section, while all outcomes described are intended as overall findings reported in the original investigations (additional information is available in Table 1).

Cognitive Training Interventions and Outcomes Assessment
All of the studies, except for two [19,20], were designed as randomised controlled investigations, in which a computer-based cognitive training was assigned to intervention or control groups. In four studies [5,22,23,26], the participants were randomised to experimental cognitive training, according to the study protocol or the waiting list group. In comparison, in three studies [21,24,25], the enrolled patients were assigned to adaptiveor not-adaptive computer-based training. The latter differed for active or absent progressive difficulty adjustment during cognitive tasks. Drop-out was higher in intervention programs, especially in groups undergoing adaptive computer-based training.
The majority of studies investigated the effects of cognitive training on working memory, attention, and performance in daily living activities (with peers and family). The primary outcome measures were: Wechsler Intelligence Scale for Children, Fourth edition [31], Working Memory Index (WMI), National Institute of Health Toolbox Cognition Battery (NTCB) [32], Conners' Parent Rating Scale (CPRS) [33], and a customised feasibility questionnaire. The intervention duration varied from 5 to 12 weeks among studies. All of the enrolled patients underwent baseline and post-intervention, or control cognitive assessment with a variable follow-up interval (3 months to 5 years).

Qualitative Summary of the Studies
Most of the studies (6/9, 67%) trained participants with BTs through the administration of Cogmed. Cogmed represents digital training for improving working memory and attention. In the included studies, the training period ranged from 5 to 12 weeks. All of the studies, except for one [20], asked participants to complete 25 at-home training sessions, while Carlson-Green et al. [20] reported 35 training sessions.
Five of six studies were RCTs. Three of them [21,24,25] randomly assigned participants to two different groups: Adaptive vs. not-adaptive Cogmed versions. In the adaptive version, the difficulty of the tasks was adjusted throughout each of the 25 training sessions, in order that as the child became more skilled, the exercises became more difficult. Conversely, in the not-adaptive condition, the computer program consisted of the same activities, with the level of difficulty that never increased. Therefore, the not-adaptive condition was intended to provide a non-therapeutic "dose" of the intervention.
Overall, results reported significant improvement in visual working memory skills in those participants who were compliant with the adaptive training program compared with those who completed the not-adaptive version. In the study by Siciliano et al. [21], the efficacy of the interventions diminished over time on WISC-IV WMI [F (2, 28) = 3.28, p = 0.05] and the NTCB Fluid Cognition Composite scores [F (3, 33) = 8.45, p < 0.001]. However, in this latter study there was no evidence of stable improvement over time, thus favouring the adaptive to the not-adaptive version of Cogmed. Furthermore, in one study [25], training had no significant effects on attention or verbal working memory tasks. Differences in parent ratings between the adaptive and not-adaptive groups were no longer significant at the 3-month follow-up evaluation [F = 3.65, p = 0.08].
Two studies [5,23] randomised participants to intervention (n = 30) vs. waiting list (n = 32). In one study [5], authors reported that the intervention group demonstrated greater short-term improvement than the control group (p = 0.02) for the primary outcome measure (spatial span backward). Moreover, the intervention group demonstrated greater shortterm improvement than the control group on secondary measures of attention (p = 0.01), working memory (p = 0.02), and processing speed (p = 0.02). In the other study [23], neither the control nor the intervention group showed acute or long-term decline in peer relation problems through the 6-month post-assessment. The waitlist group exhibited significantly elevated self-reported problems at baseline and showed an acute decline in family relation problems at immediate post-waitlist assessment, while the intervention group reported no significant acute change. Neither group experienced a decline in family relations over the 6 months.
Moreover, the intervention group significantly showed more remarkable improvement on several measures of working memory (p < 0.05) and attention (p < 0.01) than the control group. Parents of participants in the intervention group described more significant reductions in attention and executive functioning problems than parents of controls (Conners 3 Inattention and Executive Function Scales, p < 0.05). However, the authors admitted that parents' reports might have been affected by observer bias in investigation vs. control and adaptive-training vs. not-adaptive training version by guessing the intervention type.
The study by Palmer et al. [22] employed the intervention "Fast ForWord" to improve reading ability in patients (n = 81) with BTs. In addition to receiving the SOC, patients were randomly assigned to the intervention group. Seventeen of the 43 patients (39.5%) randomised to the intervention group were able to complete the target of 30 intervention sessions. Authors reported no change in reading ability over time (5 years) between control and intervention group (p = 0.62).
In another trial, Sabel et al. [26] showed that motor and process skills of participants (n = 7) improved significantly after active video gaming (Nintendo Wii). However, authors did not report significant changes in cognitive performances (general working memory, verbal and visuo-spatial learning) although trends for improvement in sustained attention and selective attention were reported.
The long-term results of the approaches hereby discussed are still controversial. Most of the authors reported that further studies are mandatory to explore the effects on additional cognitive domains other than working memory and the duration of improvement or reduction of cognitive decline in patients undergoing SOC for BTs, or those who are BT survivors and had previously undergone radiation-and chemo-therapies. Extended data are reported in Table 1.

Discussion
In this systematic review, we summarised the available evidence of the effects of computer-based interventions on cognitive outcomes in patients diagnosed with, or survivors of, paediatric BTs.
Overall, the studies suggest that this intervention improves cognitive functions, such as working memory and attention. Specifically, in those studies characterised by intervention group vs. waiting list [5,23], participants in the intervention group exhibited more significant cognitive improvements than the control group.
Furthermore, in two studies [21,25], the adaptive version of the training (Cogmed) led to higher improvements in working memory skills compared with participants who completed the not-adaptive version of the program. These results indicate that computerised cognitive training may represent a productive alternative to pre-existing interventions, contributing to a significant advancement in the management of cognitive functions. However, some trials depicted a scenario in which there was no significant improvement in cognitive tests over time. Several factors may explain these different results.
The first important variable to be considered is represented by the type of intervention employed in the studies. Cogmed has led to more significant results in improving cognitive functions (working memory and attention) compared with the other digital interventions (Nintendo Wii, Captain's Log, and Fast ForWord). Cogmed showed significant advantages compared with other computerised cognitive training programs for the survivor population. Its use has been associated with significant efficacy in a number of controlled trials with children and adolescents affected by ADHD and other pathologies associated with poorer performances in the domains of working memory and attention [34][35][36]. Moreover, Cogmed specifically targets working memory skills, which have been proposed to underlie the changes in intelligence and academic performance frequently seen in survivors with cognitive late effects [37]. Finally, this computer-based training consists of a fixed "dose" of training (i.e., 25 sessions) and an active control version of the program, making it ideally suited for empirical study in a randomised controlled design.
Another important factor associated with the different results which emerged in the studies may be represented by the timing of the intervention. Specifically, most of the studies enrolled patients after a mean of 5 years post-SOC. However, in some studies [21,22], patients were enrolled in parallel with the radiation therapy or chemotherapy, showing a lower completion of the digital intervention. In this regard, most of the participants were not able to complete the predefined "acceptable dose" of Cogmed sessions, indicating that patients were potentially overburdened by study procedures [21]. On the other hand, those studies conducting interventions after a minimum of 2 years, and on average 5 years after completion of all treatment, when completing an at-home online training program showed more feasibility and acceptance. These findings highlight that the delivery of interventions closer to diagnosis and treatment may be more stressful for families leading to higher number of dropouts and a lower efficacy of the intervention.
Feasibility and acceptance represent other essential variables to take into account. Evaluation of feasibility and acceptance plays a crucial role in validating computer-based interventions. Indeed, in those studies reporting higher rates of these two variables, there were significant improvements in cognitive and social outcomes compared with those studies indicating a poorer feasibility of the intervention. The study by Hocking et al. [24] investigated the acceptance and feasibility of Cogmed, showing significant feasibility difficulties in using the intervention, in which only half of the sample completed at least 20 sessions. Despite incentives for completing sessions, 18.5% of the sample did not conduct any training sessions and 48.1% did not complete the intervention, indicating a failure to engage paediatric patients with BTs in Cogmed. These findings contrast with other studies employing Cogmed that showed a good feasibility. To this regard, the contrasting results may be due to methodological differences, as the other studies [5,23,25] were largely also comprised of ALL survivors who may have less severe neurocognitive deficits than patients with BTs, thus being able to complete a higher number of training sessions. Overall, most of the trials reported a good degree of feasibility and acceptance in participants with BTs, despite the three studies [21,22,24] showing significant dropouts during the follow-up.
In addition to the aforementioned differences, it should be emphasised that the included studies enrolled small sample sizes (5/9 had a number of participants < 30). For this reason, despite reporting some significant results, most of the studies were not strong enough to evaluate the efficacy of these interventions. Moreover, the limitation of the samples is due to the fact that BTs in paediatric age represent a rare pathology and the recruitment of patients is challenging. In this regard, few studies assessed cognitive digital interventions in the paediatric population with BTs. In this review, only four studies evaluated the computer-based intervention exclusively in a population with BTs; one was not an RCT, thus representing a limitation regarding the quality of the results. To achieve more encouraging results, a higher number of trials with larger sample sizes will be needed.
As mentioned above, nearly 80% of the included studies assessed patients after the end of the standard treatments for BTs. The mean age of participants at the time of the assessment was 12.24 ± 0.59 years, with a mean of 7 years after the diagnosis (5.82 ± 1.49 years). Although children diagnosed with a BT can experience significant cognitive effects years after treatment, there is a need to start therapies closer to the diagnosis. Recent evidence suggests that deficits may emerge soon after or even prior to surgery, indicating the potential importance of delivering interventions as early as possible [38]. Therefore, early interventions could lead to a lower risk for subsequent cognitive decline and social impairments. However, the potential difficulties due to lower feasibility of computer-based interventions during the first phases after the diagnosis, as reported in the study by Siciliano et al. [21], represent a problem to be addressed. Utilization of at-home computerised training may be challenging for children while still actively dealing with the stress of their diagnosis. Accurate psychoeducation and understanding of cognitive late effects may be valuable in motivating families to complete the interventions [39]. In addition, strong collaboration between psychologists and medical providers will be beneficial in encouraging families and children to complete cognitive interventions, regardless of the timing of when they are delivered [21].
Despite some limitations and differences across all the studies, evidence from this review supports the idea that computer-based interventions are promising in cognitive rehabilitation. These results show similarities to evidence outlined in a systematic review [40] that assessed the efficacy of remote technology-based training programs (TP) for children with acquired brain injury (ABI). Based on the review process, 16 of the 18 studies on cognitive TP were found to be effective in improving cognitive outcomes immediately after training, on both near-and far-transfer measures. Moreover, 8 of the 12 studies on cognitive TP that considered also outcomes on functioning in daily life found positive gains. Moreover, another systematic review regarding cognitive rehabilitation on children and adolescents with ABI suggested that these types of interventions are promising, as they can lead to improvements in both cognitive and psychosocial functioning of children with ABI [41].
However, it is important to outline, that digital interventions should always follow crucial steps to achieve their aims. First, a study on the usability/feasibility of the digital tool should be conducted. Then, a trial with an adequate sample size should be designed to assess the device's safety and efficacy. By following these fundamental phases, the intervention may improve to reach final approval to become a digital therapy. Digital therapeutics are not simple applications concerning health, nor simple systems offered by pharmaceutical companies that help patients manage their diseases, instead they represent real curative interventions, enabling the improvement of clinical outcomes, such as a pharmacological treatment [42]. In 2012, the European Commission released guidance (updated in 2016) on the qualification and classification of stand-alone software used in the health care setting as a medical device [43]. Under this guidance, "mobile apps are considered medical devices if used specifically for diagnostic and/or therapeutic purposes, including the diagnosis, prevention, monitoring, treatment, or alleviation of disease".
There are significant differences in the structure and organization regarding medical products and software registration between the European authorities and the US Food and Drug Administration (FDA). The EMA and the FDA are responsible for pharmaceutical regulation, but only the FDA has the control on pharmaceuticals and medical devices. In the EU, only the European Commission (no single agency) is responsible for regulating digital therapeutics [43]. To this regard, a clear set of regulations should be defined at national and international levels to support interoperability and allow for safe and effective data exchange between different information and communication technology systems [42].
By improving these clinical and normative aspects, digital therapeutics could be safer and more effective for patients. It will indeed lead to an improvement of the quality of life of the patients themselves and their families.

Conclusions
This paper aimed to systematically review the evidence of computer-based cognitive interventions in paediatric patients diagnosed with BTs. Overall, the included studies showed a positive trend of these interventions on cognitive outcomes. Working memory and attention were the most trained and improved cognitive functions by the digital treatments. Nevertheless, some studies revealed only transient positive effects on these neurocognitive domains with a significant number of dropouts during the follow-up. More trials with larger sample sizes are warranted. Moreover, there is a need for trials to enrol patients closer to the time of diagnosis, even though patients and their families may experience a poorer feasibility and acceptance of the treatment. To this regard, strong collaboration between medical providers and psychologists could be beneficial in motivating families and children to complete cognitive interventions, thus leading to potentially significant improvements of cognitive outcomes and quality of life.
Author Contributions: Supervision and conceptualization, F.S. and N.V.; writing-original draft preparation, F.S., L.T. and G.R.; data curation, F.S., L.T., G.R., D.M., A.C. and G.P.; methodology, E.L. and N.V.; review and editing by all authors. All authors have read and agreed to the published version of the manuscript.
Funding: This work was funded by Onlus, Il Laboratorio Di Chiara, Rome, Italy.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available in the article "Computerbased cognitive training in children with primary brain tumours: a systematic review".