Even with optimal pharmacological or surgical therapies, patients with Parkinson’s disease (PD) experience disabling symptoms (such as postural instability or freezing of gait) that do not respond to these treatments. Exercise is considered a possible, complementary treatment for PD (Clark et al., 1956) and recent studies showed benefits of different rehabilitative approaches on the progression of motor decay (Goodwin et al., 2008; Keus et al., 2009; Frazzitta et al., 2012, 2015b). Due to the loss of dopamine in the dorsolateral striatum, the basal ganglia region associated with the control of habitual motor behaviors, PD patients present an impaired ability to acquire and express automatic actions (Redgrave et al., 2010). Nevertheless, even if implicit learning mechanisms are defective in Parkinsonian patients in comparison to normal subjects, motor learning is feasible in PD (Nieuwboer et al., 2007) and this is crucial for rehabilitation. In this context, literature data indicate that training the patients to use cognitive strategies allows the execution of correct movements (Morris et al., 2009) under executive/volitional control.

Many PD patients present cognitive dysfunctions and the cognitive status is considered an important determinant of rehabilitation outcomes (Abbruzzese et al., 2016). However, in elderly patients with impaired cognition the effect of rehabilitation has been addressed before and several authors found that poor cognitive status does not hamper the outcomes of rehabilitative efforts (Diamond et al., 1996; Colombo et al., 2004; Heyn et al., 2004; Poynter et al., 2013). Poynter et al. (2013) in an observational study of 241 patients demonstrated that even patients with moderate cognitive impairment were able to make significant gains in grooming, dressing, toileting, transferring and mobility (as measured with Barthel Index Score) after rehabilitation treatment. Similarly, a study performed in a geriatric rehabilitation unit, on a large population, found no correlation between Mini Mental State Examination (MMSE) score and functional status on admission and demonstrated functional and clinical gains in Barthel Index Score for individuals through the full range of MMSE score (Colombo et al., 2004).

Even if this issue is very important in order to optimize the rehabilitative care, there are no data concerning the relation between cognition and rehabilitative outcome in PD.

Moving from the clinical practice with inpatient subjects with PD, in this retrospective study, we evaluated whether the cognitive decline affects the outcomes of an intensive, aerobic and goal-based rehabilitation program, based on motor-cognitive training and repetition.

Out of 485 PD patients, according to MMSE, 236 had normal cognition (mean ± SD: 28.7 ± 1.1), 208 had mild cognitive impairment (24.8 ± 1.6) and 41 were moderately or severely cognitively impaired (17.7 ± 3.1). Considering FAB, 255 patients had normal executive functions (16.0 ± 1.4) and 230 had pathological executive functions (10.4 ± 2.6). Correlation between MMSE and FAB values was good (r = 0.63, p < 0.0001). Considering categorized values, only one patient out of the 41 with MMSE ≤20 (moderately or severely cognitively impaired) had normal executive functions (FAB ≥13.8). MMSE scores were negatively related to age (R = −0.38, p < 0.0001) and H&Y (R = −0.24, p < 0.0001) and positively related to years of education (R = 0.15, p = 0.0007). Similarly, FAB scores were negatively related to age (R = −0.35, p < 0.0001) and H&Y (R = −0.26 p < 0.0001) and positively related to years of education (R = 0.14, p = 0.002).

This is the first study that addresses the issue of the relationships between cognitive impairment and rehabilitative outcome in PD patients. As opposed to what is normally believed (Abbruzzese et al., 2016), our data demonstrates that also individuals with cognitive impairment treated with a specific, intensive rehabilitative approach were able to gain comparable benefits in motor and functional outcomes as patients without cognitive impairment. Since there are several PD patients with cognitive deficits, it is important to re-examine the question of rehabilitative care also for these subjects. In our patients, age, severity of disease (evaluated with H&Y scale) and cognition (evaluated with MMSE and FAB) were correlated with each other: higher age and more advanced motor disturbances were associated with poorer cognitive performances. In accordance with literature data (Kuzis et al., 1999; Appollonio et al., 2005), MMSE and FAB scores showed a direct relation with years of education and an inverse relation with age. In all groups, no differences were found in the total amount of levodopa dose. This finding allows us to exclude a secondary, dose-dependent effect of dopaminergic therapy on rehabilitative outcomes. According to the patients’ subdivision on the basis of MMSE score, we observed that the three groups differ at the baseline in all scales. These differences have a relation with cognition: with the worsening of MMSE score, the performances get worse. This finding could be explained with the worsening of performances as age increases or with the existence of a direct relation between cognition and motor skills. However, even after correction for age, disease severity and years of education, cognitively impaired patients obtain significant benefits after rehabilitation.

Since PD patients frequently suffer from dysexecutive syndrome, we evaluated also the impact of this condition on rehabilitation subdividing patients on the basis of FAB score. Even in this case, in the group with pathological score the performances at baseline were worse and both groups improved in all scales after rehabilitation. Moreover, data concerning the improvement obtained in TUG and BBS score by patients with dysexecutive syndrome has a very important clinical relevance: these patients had pathological scores at admission and get normal performances after the treatment (Qutubuddin et al., 2005; Nocera et al., 2013).

We found that rehabilitation is effective in PD patients with cognitive impairment, as much as in those with normal cognition. Indeed, even if the percentage improvement in total UPDRS was significantly greater in patients with preserved cognition, all groups showed a significant improvement after rehabilitation.

Our findings could indicate that even in presence of cognitive decline, the ability to learn is not completely hindered. We argue that our intensive and goal-based rehabilitation treatment, entailing constant and repetitive exercises, exploits both the implicit and the explicit learning. Implicit learning is the acquisition of knowledge without awareness and volition (Reber, 1989) and differs from explicit learning, defined as a form of conscious and intentional process of knowledge acquisition (Cohen et al., 1985). Striatum plays a central role in implicit learning (Gamble et al., 2014), the anterior cingulate/mesial prefrontal cortex supports the explicit component of learning, whereas hippocampus is crucial for both explicit and implicit processes (Gamble et al., 2014). People with PD have degeneration of dopamine-producing neurons, leading to dopamine deficiency, striatal deficits and consequent impairment in implicit learning (Rieckmann et al., 2010; Gamble et al., 2014). Moreover, several studies demonstrated that patients suffering from PD with dementia have deficits on tasks of explicit learning as well as on some task of implicit learning (Bondi and Kaszniak, 1991; Kuzis et al., 1999).

It is conceivable that the proposed treatment is able to optimize motor learning both explicit, exploiting the acquisition of voluntary strategies, and implicit through incremental stimulus-response associations.

Therefore, MIRT probably works through the interaction between implicit and explicit learning mechanisms. This interaction was theorized and described previously: Destrebecqz et al. (2005) found that anterior cingulate/mesial prefrontal cortex exerts control on the activity of the striatum and suggested the existence of a functional interaction between different brain regions that subtend different computational objectives and in which information processing appears to be either accompanied by awareness (the anterior cingulate/mesial prefrontal cortex) or not (the striatum).

Probably, other mechanisms underlie the improvement we found after the treatment. In particular, with respect to walking, spinal mechanisms that are not influenced by cognition could be implicated. It has been postulated that treadmill training could provide adequate sensory inputs, which may stimulate the spinal locomotor circuitry (Van de Crommert et al., 1998). Certainly, treadmill-plus through the use of feedbacks and cues optimize learning. However, the belt of the treadmill itself forces the subject to step, through stretch facilitation of hip flexors and ankle plantar flexors and provides adequate sensory inputs which stimulates the spinal locomotor circuitry, stressing the functioning of the Central Pattern Generator (Bello and Fernandez-Del-Olmo, 2012). Beyond these mechanical concepts, Edgerton et al. (1991) provided evidence that different strategies, including treadmill training, could produce functional changes or “motor learning” properly in the spinal motor-generating circuitry.

Finally, we found better percentage of improvement in TUG and BBS in patients with pathological FAB score. Dysexecutive syndrome negatively affects those motor performances that entail a high degree of executive components, like TUG and BBS. Since any task that involves planning and executing a goal-directed action requires executive functions (Morris et al., 2001), we explain these results hypothesizing that MIRT leads to an improvement in these scales through the voluntary/attentive application of learned motor skills.

There are some limitations to this study: first, we did not perform a detailed assessment of neuropsychological functions. This would allow us to understand better the cognitive profile of patients and evaluate this in relation to the rehabilitative outcomes. We did not consider the aspect of the generalization of learned abilities. Further studies should explore how patients with and without cognitive decline transfer the effect of the treatment in daily life and the duration of the obtained benefits over time. Moreover, different rehabilitation approaches have been proposed for patients with PD: in this study we evaluated only the efficacy of an intensive and goal-based treatment. This is not enough to affirm that is possible to achieve the same results with other, different rehabilitation approaches. Further studies are needed to clarify this issue.

We demonstrated for the first time that PD patients in early, medium and advanced stages of disease with cognitive dysfunctions, who undergo a specific treatment exploiting implicit learning strategies, are able to gain benefit in rehabilitative outcomes. Our data allow challenging the belief that rehabilitation should be reserved only for patients with normal cognition.

DF wrote the text, provided substantial contributions to discussion of the content and edited the manuscript before submission; PO wrote the text, provided substantial contributions to discussion of the content and researched data for the article; RB researched data for the article; RM did the statistical analysis and generated tables and figures; NG, GP, MFG did a critical revision and provided substantial contributions to discussion of the content; GF wrote the text, provided substantial contributions to discussion of the content and did a critical revision.

DF, PO, RM, RB, MFG, GP, GF declare no disclosures or conflict of interests. NG—Prof. Giladi serves as a member of the Editorial Board for the Journal of Parkinson’s Disease. He serves as consultant to Teva-Lundbeck, Intec Pharma, NeuroDerm, Armon Neuromedical Ltd./Dexel, Monfort and Lysosomal Therapeutic Inc. He received payment for lectures at Teva-Lundbeck, Novartis, UCB, Abviee, Shaier and Genzyme. Prof. Giladi received research support from the Michael J Fox Foundation, the National Parkinson Foundation, the European Union 7th Framework Program and the Israel Science Foundation as well as from Teva NNE program, LTI, and Abviee and CHDI. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. There are no patents, products in development or marketed products to declare. This does not alter our adherence to all the Frontiers in Aging Neuroscience policies on sharing data and materials, as detailed online in the guide for authors.