Combinations of physical and cognitive training for subcortical neurodegenerative diseases with physical, cognitive and behavioral symptoms: a systematic review

The present review is in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis methods (PRISMA) guideline [24]. We constructed this systematic review using the Evidence synthesis tool and CADIMA database, a free web tool [25]. We searched PubMed, BASE and the ACM Digital Library for English-language articles (see Table 4 in Appendix for search strategy). Two researchers (CC and JG) independently screened the titles to verify the eligibility of articles and then checked the fulltext versions. Any disagreements were settled by consensus.

Data extraction was done by one researcher (JG or CC) and checked by two (JG and CC). These data included population, pathology stage, characteristics of control group, number of participants, mean age, type of physical-cognitive training, component targeted, and training modalities (frequency, length of training program in weeks, mean session duration in minutes, total duration of sessions in hours, and presence or absence of supervision).

Risk of bias was assessed by two independent researchers (CC and JG), using the Cochrane risk-of-bias tool [26]. This tool assesses the risk of bias in five domains (randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome, and selection of the reported results). There are three types of results: high risk of bias, some concerns, or low risk of bias. In the case of disagreement, a consensus was reached.

The database search yielded a total of 333 articles. After duplicate removal, there were 252 articles, 165 of which were excluded after screening their titles and abstracts. A total of 87 full-text articles were therefore assessed for eligibility, and 65 were excluded because they failed to meet one or more inclusion criteria: population (n = 1), type of intervention (n = 54), control group (n = 1), outcome (n = 2), not RCT, feasibility or pilot study (n = 5), or publication or access (n = 8) (Fig. 1 and Supplementary materials 1).

Results of the risk of bias assessment are summarized in Fig. 2 (see Appendix). All studies had some risks of bias. First, there was considerable variability in the intended intervention and selection of reported results. Maintaining the blinding of participants and caregivers in rehabilitation studies was also difficult, again leading to possible methodological biases. Moreover, several studies diverged slightly from the study protocol, and several studies did not supply any preregistered or published analysis plan.

Our review suggests that improvements in cognitive functions are inconsistent, whatever the type of exercise. Inconsistency in rehabilitation studies seems to be linked to the type of exercise and its specificity, methodological bias, or study duration [4, 59, 60]. Some studies [31‐33, 42, 44, 45, 47, 53‐56] have only small numbers of participants. This either meant that significant results had to be nuanced, or made it impossible to report significant effects, owing to a lack of statistical power.

As the scientific corpus lacks data about the impact of combined training on NDs, we relied on data for healthy older adults or ones with mild cognitive impairment. Regarding duration and frequency, a meta-analysis revealed that larger cognitive gains could be expected in older participants who underwent combined interventions administered in a group fewer than five times per week [19]. When there were more than five sessions per week (intensive training), training was less efficient, as it could induce cognitive fatigue, excessive stress, and less adherence in intervention activities, especially for highly challenging training [61, 62]. It is therefore necessary to establish optimum difficulty, frequency, and duration for training, especially if participants have neurocognitive and/or behavioral disorders. In this paper, we focused on EFs which are the first to sustain damage in subcortical ND.

To be more specific, for inhibition, interventions should last between 12 and 24 h with one, two or three sessions a week to observe significant effects in older healthy adults [44, 63]. When the training is shorter and/or the sessions are less frequent, no significant difference is found [47]. For flexibility, a duration of 12–36 h with sessions once, twice or three times per week is needed to bring about significant results in healthy older adults [64]. Even so, only one of the three studies that met these criteria reported significant results, and then only with a small effect size [45] (accurate effect size not available for this study). One of the other two studies had the minimum number of hours needed to improve flexibility, which may explain the inconclusive results [47]. In this study, the same was true for attentional abilities, for which the recommended duration is between 12 and 104 h. Nevertheless, insufficient duration cannot explain the nonsignificant result for processing speed (TMT-A [40]) [47]. One study [duration = 18—30 h] found that psychomotor speed improved in both exergame and aerobic groups, compared with active controls, and effects could still be observed at 24 weeks [63]. Finally, dual-tasking costs were lower with interventions totaling either 12 h (one 60-min session per week) or 60 h (three 50-min sessions per week) [64]. A single study included in the present review assessed dual tasking and arrived at the same conclusion. The results were encouraging, but with inconsistencies, possibly because of variability in the cognitive and physical tests used, the sample size and statistical power, the function targeted and exercise content, or stage of the pathology [51].

The level of difficulty reached by participants should be considered. In one study, the non-superiority of the CMG can be explained by the inadapted difficulty of exercises to the participant’s level of education [35]. Thus, the cognitive exercises were not challenging enough for participants with a higher level education. And yet, this is an essential factor in cognitive training, the aim of which is to maintain efficient brain function. What is more, in another study, DT and SEQ training groups improved in equivalent proportions [50], but the SEQ training group achieved a higher level of difficulty than the DT training group. One possible explanation is that the SEQ training group could focus fully on their cognitive performance, instead of having to focus on their physical performance as well. As SEQ training is more intense than DT training, it would presumably yield identical results [50]. Imaging data could be relevant in this context, as these clinically identifiable results may be subtended by different mechanisms. While SEQ training may allow for better automatization of the task, when the latter resembles ADL, DT training may allow for more efficient integration of task-related networks [50]. Nevertheless, the authors reported good functional transfer ability, which may be of use when performing ADL.

Improvements in physical function were also inconsistent across the studies included in this systematic review, owing to considerable methodological heterogeneity. Disease stage was an important factor for training adherence and effects. In the studies included in the present review, disease stages fluctuated between early and severe. Thus, authors should consider disease stage as a key factor for interpreting their results.

In our review, interventions lasted between 4 and 24 weeks, but physical function effects did not seem to depend on intervention duration. As in previous reviews [5, 65], it was difficult to highlight an optimum duration for physical training, although one previous meta-analysis found that individuals with and without cognitive impairments only derived significant physical benefits from interventions lasting less than 12 weeks [4]. Concerning training frequency, in our review, two to three sessions per week were usually used to induce physical benefits. Similar recommendations are contained in other reviews on physical-cognitive training [5, 66].

Gait and balance training were the main physical components practiced and the main physical outcomes assessed. This is probably because gait and balance impairments are major symptoms in PD. Moreover, these abilities are required for functional autonomy [67]. However, results were disparate, with seven (balance) and twelve (walking abilities) studies out of 21 reporting significant improvements. This heterogeneity can be linked to parameters cited previously and to the different tools used to assess balance and walking abilities.

Several studies investigated the effects of combined physical and cognitive training on walking parameters during dual tasking [29, 31‐33, 44‐46, 50, 55]. The addition of cognitive or motor tasks has been found to affect walking parameters in PD [68]. In the present review, we highlighted improvements in gait parameters such as speed, step and stride length, cadence, and double support time during dual-task walking. These results were in line with a meta-analysis on dual-tasking performance among people with PD following combined physical and cognitive training [69].

Moreover, studies [28, 32, 51, 52] reported relative improvements in disease-specific motor impairment (UPDRS-II or III). Difficulty improving these scores can be explained by the natural progression of the pathology, and therefore of disease-specific motor impairments.

Two studies highlighted the benefit of exercise on the structural and functional aspects of the brain. The authors' hypothesis is as follows: in Parkinson's disease (PD), the prefrontal areas, associated with executive functions and multitasking activities, would be recruited to compensate for the alteration of neural networks related to walking. Indeed, a lower activation of the prefrontal areas would confirm the effectiveness of motor function training [55]. However, training via treadmill only does not stimulate the prefrontal cortex and thus does not improve the walking capacity. The deficient neural networks typically attributed to walking function will be compensated for by cortex activation. This study strengthens the idea that VR with a cognitive component provides specific benefit in motor symptomatic diseases such as PD by stimulating the prefrontal cortex [55]. In the same vein, another study explains the increased neural connections between the supplementary motor area and the pedunculopontine nucleus would be an adaptive response to PD symptoms [44]. This brain activity is decreased through physical and cognitive training but not through education. This study also demonstrates the relevance of cognitive-motor training on motor symptoms of PD [44]. These examples permit to support the need for evaluating cognitive function along with physical function when using motor-cognitive training. In the studies included, cognitive functions were not systematically assessed in studies (7 of 21 didn’t report cognitive measures) whereas the results about brain correlates lead us to think motor-cognitive training could improve cognitive function.

Studies [22, 70] highlighted the additive effects of cognitive stimulation and physical activity in healthy older participants when performing untrained dual tasks (i.e., good transfer of acquired skills) [70]. Taken separately, cognitive training and physical training each have beneficial effects, but the neurobiological effects of combining the two are probably greater. Each type of training seems to involve a different mechanism of brain plasticity, suggesting a potential effect of SYN training on cognition [70]. One theory is that physical activity facilitates neuroplasticity, while cognitive activity guides neuroplasticity [22]. Additionally given that the cognitive demand is incorporated into the motor task, SYN training could be more efficient than SEQ or DT training, owing to the similarity with everyday brain functioning [23]. Although this theory has yet to be proven, studies comparing these types of training in healthy older people are currently underway [71, 72].

In subcortical NDs, the effects of cognitive and physical training may differ according to the symptoms induced by the disease and their impact on performance during training. One task may be given more emphasis than another during DT training [50]. The studies included on the present review focused on PD, in which EFs are impaired at an early stage. Patients with PD may have difficulty dual tasking, owing to flexibility and/or attentional disorders. Furthermore, when attention is divided between two tasks, the quality of task performance is impacted. And, especially when participants have an ND, and when one of the tasks is supposed to be more difficult (subjectively or objectively) and is therefore allocated more attention. For instance, one review showed that patients with walking difficulties may focus more on this aspect of training than on the cognitive aspects [65]. This is exacerbated by the fear of falling. These different points may explain the inconsistency in the results and/or the lack of significance. Compared with healthy individuals, it may take longer to observe significant effects of combined training in neurological pathologies, and longer in subcortical NDs.

SYN training could prove relevant in the context of these pathologies, as it combines two types of stimulation to achieve a single goal but studies are needed to prove this point. More attention is therefore focused on the task, and there may be more benefits. Precautions must be taken in view of the early difficulties encountered in subcortical NDs. For example, patients must be accompanied during the exercises so that these are performed safely (prevention of falls). Similarly, on the cognitive level, working memory impairment can lead to patients forgetting the instructions, which must therefore be displayed throughout the exercise. More specifically, patients have to remember a set of movements, in order to perform them correctly when giving the desired responses. These movements therefore have to be displayed on the screen throughout the exercise.

The present systematic review has several limitations. First, all included studies concerned PD, possibly because PD is one of the most common NDs with early motor and cognitive symptoms. There was therefore a dearth of data on other motor and cognitive pathologies. Second, we chose to include only interventional studies focusing explicitly on motor-cognitive training that targets cognitive and physical functions. For example, we did not consider dance and tai-chi as part of motor-cognitive training because they do not explicitly result from a combination of cognitive and physical exercises. Additionally, we did not consider occupational therapy as a cognitive training on specific functions. This focus on a specific training methodology that may introduce a selection bias, which should not be overlooked.

Third we focused solely on the physical and cognitive aspects, even though we know that physical and cognitive activity can influence QoL and behavioral aspects (irritability, anxiety, etc.). Fourth, our review lacked follow-up data. Follow-up assessments were not systematically carried out in the studies we reviewed, even though it would have been interesting to know which types of training brought about the most lasting changes.

Structural and functional imaging data could be used to support and guide future studies. It would also be relevant to compare different types of SYN training according to their presentation modalities (e.g., more technological forms such as exergames vs. more ecological forms such as tai chi [5].

Combined training is feasible, tolerable, and seems promising in PD. The advantage of combining physical training and cognitive training, rather than using them separately, has not yet been demonstrated in subcortical NDs with early physical and cognitive symptoms. More studies are needed to show that combined training is relevant in these populations. Nevertheless, the present systematic review shows that the fun element of exergames can help patients stay motivated, with excellent rates of compliance [50, 52]. Future studies should focus on comparing the feasibility, tolerability, and effectiveness of physical and cognitive training, and specify which combinations to use [71]. Differences between DT and SYN training remain unclear, particularly regarding the role of cognitive load. It would therefore be interesting to examine the cognitive implications of each type of training.