Neural Mechanisms of Motor Dysfunction in Mild Cognitive Impairment and Alzheimer’s Disease: A Systematic Review

Search methods

PubMed was searched on August 26, 2021 for articles in the English language. A search query was built to extract articles that combine research on a) MCI and/or AD (keywords: MCI, “mild cognitive impairment”, and Alzheimer); b) motor behavioral assessments (keywords: “muscle strength”, “grip strength”, “isokinetic strength”, “vertical jump”, “motor speed”, “tapping speed”, “motor coordination”, dexterity, bimanual, “finger tapping”, “foot tapping”, “motor skill”, “motor function”, “motor performance”, “motor adaptation”, “Purdue pegboard”, “grooved pegboard”, “Archimedes spiral”, gait, walk, “timed up and go”, TUG, “walking while talking”, “dual task”, mobility, balance, Romberg, “sensory organization”, “sensory organisation”, vestibular, “motor sequence learning”, “motor learning”, “serial reaction time task”, stride); and c) neuroimaging (keywords: PET, “positron emission tomography”, “magnetic resonance imaging”, MRI, fMRI, fcMRI, “functional connectivity”, “diffusion weighted imaging”, DWI, “diffusion tensor imaging”, DTI, “near infrared spectroscopy”, NIRS, fNIRS, “brain structure”, “brain function”, “brain volume”, “volumetric”). The exact search query can be found in Supplementary Material 1. Note that we focused only on objectively measured motor performance and not on self-report measures such as questionnaires on activities of daily living.

Study selection

Author VK reviewed all results from the PubMed query. Studies were selected that reported on group (i.e., MCI and/or AD versus control participants) by brain (measures of brain structure/function) interaction on motor behavior (see Fig. 1A). Describing these interactions can improve our understanding of the potential causal role of Alzheimer’s neuropathology such as hippocampal degeneration on motor dysfunction.

OPEN IN VIEWER

Fig. 1

Schematic overview of analysis models of studies for which motor behavioral data and brain structural or functional data were collected from individuals with AD pathology. Here we review studies that report on the interaction between group and brain structure/function on motor behavior (A). We also report on studies that collected data on an MCI/AD and a control group but report the association between brain volume on motor performance stratified by group (B), and studies that report the association between brain volume on motor performance in MCI/AD participants without including a control group (C). We do not include studies that collected data on an MCI/AD and a control group, but only report the association between brain volume on motor performance for the entire sample combined (D), or studies that report group differences in (E) motor behavior and (F) brain structure/function without describing their association.

During our literature search, we also encountered studies which report on brain-behavioral associations for MCI/AD and controls separately without statistically testing the interaction effect (Fig. 1B), and studies which did not include a control group and thus were not able to compare this association to that in individuals without AD pathology (Fig. 1C). We will report on these brain-behavioral associations because their outcomes can inform future studies on selecting regions of interest for brain-motor behavioral relationships in AD pathology. Studies that only reported brain-behavioral associations within a combined sample of MCI/AD and controls (Fig. 1D), and studies that reported on group differences in motor behavior (Fig. 1E) and group differences in brain structure or function (Fig. 1F), but which did not report on brain-behavioral associations were excluded from this review. For an overview of this metric of quality assessment see Supplementary Table 1.

With respect to MCI populations, we were particularly interested in amnestic MCI (aMCI) as compared to non-amnestic MCI (naMCI), because aMCI is more likely to progress into AD [18]. When studies did specify the MCI subtypes, only the aMCI results were included. When studies did not specify the MCI subtype (e.g., mix of aMCI and naMCI), all MCI results were included. However, since aMCI tends to be a more common subtype than naMCI [19], the mixed sample of all MCI participants likely includes mostly aMCI. With respect to dementia, we review only populations in which AD is the primary diagnosis for dementia, and not because of other types of dementia such as frontotemporal dementia, or dementia with Lewy bodies. Considering that AD commonly presents with additional neuropathology such as vascular changes [20], it is to be expected that this mixed pathology will also be reflected in the samples we review here.

Supplementary Table 1 provides an overview of the definition of the study population, the samples sizes, and the design of the studies that we included here. Studies with preferable designs (e.g., large samples, well defined experimental groups, longitudinal designs, and inclusion of a control group) have been highlighted. It is noteworthy that the studies with preferred study designs generally had smaller sample sizes than the studies that only reported correlational results within MCI or AD samples.

Findings to be reported

Here we review studies that report how brain differences between healthy control participants and participants with MCI/AD are related to motor behavioral differences between these groups. We report both the presence and absence of these relationships. Reporting the absence of associations between neural and motor behavioral measures will help with defining hypotheses for future research.

Results will be discussed in order of study design: studies reporting brain-behavioral interactions by group (Fig. 1A) will be reported first, followed by studies not reporting brain-behavioral interactions by group (e.g., Fig. 1B and 1C). Results will also be organized according to motor outcome measure, and subsequently by neural outcome (e.g., regional volume or WML volume), starting with more global measures (e.g., total gray matter volume) and ending with regional measures (e.g., hippocampal volume).

Studies in MCI/AD reporting brain by group interactions on motor performance

Studies evaluating gait and non-gait motor function in relation to neuroimaging outcome measures in participants with an MCI or AD diagnosis when compared to healthy controls. Gait speed refers to walking speed at either normal pace, or maximum pace, but without running. Within MCI and AD samples, the association between neural metrics and gait was investigated in 5 out of 9 studies (see Table 2), while non-gait measures were investigated in three studies.

OPEN IN VIEWER

Table 2

Aggregated results for studies reporting association between motor performance, neuroimaging participants with MCI or AD diagnosis when compared to healthy controls

Text inside the cells indicate the first few letters of the first author from the manuscript from which the result is extracted. If necessary, these letters are followed by two numbers indicating the year the study was published to differentiate between different publications that have the same first author. If no side (hemisphere; (L) or (R)) is indicated, studies looked at bilateral regions. = undifferentiated MCI; = MCI + dementia mix; = AD; [A] AD exhibit less activation than controls; [R] AD recruit additional regions in these areas that controls do not recruit; OTHER REGIONS 1 = 3rd ventricle, 4th ventricle, Accumbens-area, Amygdala, Banks of the Superior Temporal Sulcus, Brain stem, CSF, Caudal middle frontal gyrus, Cerebellum cortex, Cerebellum white matter, Choroid plexus, Cingulate cortex (caudal anterior, rostral anterior), Corpus callosum (Anterior, Central, Mid-anterior, Mid-posterior, Posterior, Total), Entorhinal cortex, Frontal pole, Fusiform gyrus, Insula, Inferior parietal lobule, Isthmus cingulate, Lateral occipital gyrus, Lateral ventricle, Lingual gyrus, Middle temporal gyrus, Optic chiasm, Orbitofrontal cortex (lateral, medial), Paracentral gyrus, Parahippocampal gyrus, Pars opercularis, Pars orbitalis, Pars triangularis, Pericalcarine gyrus, Precunues, Rostral middle frontal gyrus, Superior frontal gyrus, Superior parietal gyrus, Superior temporal gyrus, Supramarginal gyrus, Temporal pole, Thalamus proper, Transverse temporal, Ventral DC; OTHER REGIONS 2 = Banks of the Superior Temporal Sulcus, Caudal middle frontal gyrus, Cingulate cortex (caudal anterior, rostral anterior, posterior), Cuneus, Entorhinal cortex, Frontal pole, Fusiform gyrus, Insula, Isthmus cingulate, Lateral occipital gyrus, Lingual gyrus, Orbitofrontal cortex (lateral, medial), Paracentral gyrus, Parahippocampal gyrus, Pars opercularis, Pars orbitalis, Pars triangularis, Pericalcarine gyrus, Postcentral gyrus, Precentral gyrus, Precunues, Rostral middle frontal gyrus, Superior frontal gyrus, Superior parietal gyrus, Superior temporal gyrus, Supramarginal gyrus, Temporal pole, Transverse temporal; OTHER REGIONS 3 = Banks of the Superior Temporal Sulcus, Caudal middle frontal gyrus, Cingulate cortex (caudal anterior, rostral anterior, posterior), Cuneus, Entorhinal cortex, Frontal pole, Fusiform gyrus, Inferior temporal gyrus, Inferior parietal lobule, Insula, Isthmus cingulate, Lateral occipital gyrus, Lingual gyrus, Middle temporal gyrus, Orbitofrontal cortex (lateral, medial), Paracentral gyrus, Parahippocampal gyrus, Pars opercularis, Pars orbitalis, Pars triangularis, Pericalcarine gyrus, Postcentral gyrus, Precentral gyrus, Precunues, Rostral middle frontal gyrus, Superior frontal gyrus, Superior parietal gyrus, Superior temporal gyrus, Supramarginal gyrus, Temporal pole, Transverse temporal; OTHER CONNECTIONS = Connections between the LPFC-RPFC, LPFC-RMC, LPFC-LOL, RPFC-LMC, RPFC-RMC, RPFC-LOL, RPFC-ROL, LMC-RMC, LMC-LOL, LMC-ROL, RMC-LOL, RMC-ROL, and LOL-ROL (left prefrontal cortex (LPFC), right prefrontal cortex (RPFC), left motor cortex (LMC), right motor cortex (RMC), left occipital leaf cortex (LOL), and right occipital leaf cortex (ROL)

Gait

Here we discuss studies that report on measures of average gait speed, stride measures, functional mobility, and also walking while completing an independent cognitive task. Overall, these studies reveal that less hippocampal volume, a thinner temporal gyrus, larger lateral ventricles, more subcortical white matter degeneration, and activation changes in the prefrontal cortex explained differences in gait performance between MCI and AD (see the next paragraphs for more detail). This could indicate that changes in hippocampal-frontal networks could be involved in poorer motor function in MCI and AD [22].

Longitudinal research with 14 years of follow up showed that gait speed predicted development of MCI and unspecified dementia in initially cognitively normal individuals [23] (see Table 2). This association was partially explained by volume of the right hippocampus, suggesting that hippocampal volume acts as a shared neural substrate in gait speed and cognitive decline. Slower gait speed in AD is furthermore explained by presence of subcortical white matter lesions. Cognitively normal individuals with a low lesion load had a faster gait than individuals with AD with either a low or a high WML load [24, 25]. This gait speed difference was larger between control participants and participants from the AD group with a high lesion load than between control participants and participants from the AD group with a low lesion load. However, this difference was not statistically significant. Functional imaging using near-infrared spectroscopy (fNIRS) has revealed that brain activation in the prefrontal cortex showed a larger decline in MCI than controls over multiple walking trials [26].

Stride measures comprise aspects of gait including step length (the distance between the left and right toes in a step), stride length (the distance covered during a consecutive left and right step), stride time (the time elapsed between the first contact of two consecutive footsteps), swing time (the amount of time of the swing phase of one leg), step width (the horizontal distance between two steps), and cadence (full gait cycles per unit time, where a gait cycle is the sequence measured from one foot strike to the subsequent foot strike of the same foot). Although MCI was linked to smaller hippocampal volume and a higher normalized variance in stride time compared to cognitively normal participants, the smaller volume did not explain the difference in stride variance [27] (see Table 2).

WML load was not associated with stride measures in AD. Although AD participants with either low or high WML load had shorter stride length and slower cadence than cognitively normal participants with a low WML load, no effect of lesion load on gait between AD participants with low or high lesion load was reported [24]. Later work by the same group did not reveal associations between an overall gait score that combines information from multiple stride measures [28] and WML [25].

Mobility that is part of functional behavior, such as completing an obstacle course, or walking after rising from a chair and before sitting back down is labeled functional mobility. Functional mobility is crucial for successful independent living and is closely related to activities of daily living (ADL) functioning. Subcortical WML have been related to poorer performance of AD participants on a the timed up-and-go test (see Table 2); a test where participants have to stand up from a chair, walk three meters, turn around, and return to a seated position. AD participants with a high WML load were slower on this test than cognitively normal participants with a low WML load [24]. A more recent study from the same lab replicated these results, but also found that AD with a high WML load were slower than the cognitively normal participants with a high WML load [25].

Motor-cognitive dual-tasking cost is the decrease in motor performance under a dual-task condition (i.e., simultaneously performing a motor and a cognitive task), relative to a single-task condition (i.e., performing a motor task only). Dual tasking can be particularly sensitive to AD pathology as it requires both intact motor and cognitive skills [29]. The association between brain measures and performance on a dual task that required simultaneous elbow flexion and counting backwards by a random number were significantly different for MCI participants compared to controls [30] (see Table 2). Specifically, the association between dual task performance and volume and thickness of the inferior temporal gyrus (more gray matter equals better performance), volume of the inferior lateral ventricle (larger ventricles result in worse performance), and cortical thickness of the inferior and middle temporal gyrus (thicker gyri relate to better performance) were stronger in the MCI group than in the control group. Participants with AD with a high lesion load have shown a higher dual tasking cost for cadence (steps per unit time) than control participants with either a low or high WML lesion load during treadmill walking while concurrently completing an n-back working memory task [25]. fNIRS studies have linked dual tasking deficits in MCI to both reductions in brain activation [26] and stronger functional connectivity [21]. Compared to controls, MCI participants exhibited lower brain activation in the prefrontal cortex during a task that involved walking while reciting alternate letters of the alphabet. MCI participants also had a relatively smaller increase in brain activation from single to dual tasking. The task was repeated three times and analyses over trials showed that the decline in brain activation from the first to the second dual-task trial was attenuated in the MCI group, while the brain activation from the first to the third trial decreased in control participants, but increased over these time points in MCI. During other dual tasks, which comprised of walking while counting backwards or naming animals, MCI status was linked to stronger functional connectivity between the left prefrontal cortex and the right occipital cortex. For the animal naming dual task, this was also true for connectivity between the left prefrontal cortex and the left motor cortex.

Manual motor function

Manual motor function comprises strength measures, as well as dexterity measures of the hand and fingers. Functional neuroimaging studies have looked at both brain activation and brain connectivity related to motor function with AD pathology. During a hand squeeze task inside an MRI scanner, AD participants showed less brain activation than control participants in the left supplementary motor area/superior frontal gyrus and the left premotor cortex/middle frontal gyrus [31] (see Table 2). Psycho-physiological interaction analysis with the primary motor cortex as seed region was used to identify regions where brain activity correlates with brain activity of the primary motor cortex during the hand squeeze task. This analysis revealed that relative to the control group, the AD group recruited larger regions of the fusiform gyrus, the middle cingulate gyrus, the sensorimotor cortex, the anterior cerebellum (lobule I-IV), and the cuneus. No significant differences between controls and individuals with MCI were observed in a study on the association finger tapping speed with brain connectivity [32]. In this study, within and between network connectivity of the default mode network, the frontoparietal network, and the sensorimotor network was assessed and tapping of both the left and right hand were studied independently.

Symptoms of parkinsonism

Motor symptoms of parkinsonism in the studies discussed here were assessed using the motor subscale of the Unified Parkinson’s Disease Rating Scale [33]. This clinician-rated tool results in a sum score of assessments of speech, tremor, rigidity, posture, hand movement, gait, leg agility, functional mobility, hypokinesia and bradykinesia. AD participants with a high subcortical WML load had more symptoms of parkinsonism than controls with a low lesion load, controls with a high lesion load, and AD participants with a low lesion load [24] (see Table 2). These results were replicated by the same group in a more recent study [25].

Studies evaluating gait in relation to neuroimaging without healthy control comparison

The studies described below report on the association between brain outcome measures and motor function in MCI/AD, but don’t compare these associations to those in a control group (see Table 3). Although these associations may shed light on the relationship between brain metrics and motor function in MCI/AD, it is not possible to ascribe these associations to the neurodegeneration resulting from MCI/AD because they could be part of normal aging.

Gait speed

Gait measures and specifically gait speed were the most investigated motor behavioral measures. Outcomes of studies investigating gait measures are summarized in Table 4. Overall, slower gait speed was associated with less volume of temporal and frontal brain regions, larger lateral ventricle size, stronger functional connectivity with the supplementary motor area, and larger amyloid depositions throughout the brain (see for details below). This pattern corresponds with associations reported in the section ‘Studies in MCI/AD reporting brain by group interactions on motor performance’.

OPEN IN VIEWER

Table 4

Aggregated results for studies reporting associations between gait motor performance and neuroimaging in participants with MCI or AD diagnosis without healthy control comparison

Text numbers inside the cells indicate the citation number of the manuscript from which the result is extracted. If necessary, these letters are followed by two numbers indicating the year the study was published to differentiate between different publications that have the same first author. If no side (hemisphere; (L), (R), or (U) = unspecified) is indicated, studies looked at bilateral regions. = Amnestic MCI; = undifferentiated MCI; = undifferentiated MCI (aMCI +naMCI) or a clinical dementia rating scale score of 0.5 without dementia; = AD; * = For normal walking speed, but not for rapid walking speed; X = ROI study that only looked at one side (either left or right). For ROI studies that looked at left and right, both will be presented, but the X distinguishes from Voxel Wise studies that only report one side, but have looked whole brain; Suppl., supplementary.

Gray matter volume, cortical thickness, and atrophy

Within participants with MCI, comfortable self-selected gait speed has been positively associated with volume of the primary motor cortex [34] (see Table 4), regional gray matter volume of the right superior temporal gyrus and the left superior temporal pole, and a thinner left and right superior temporal cortex [35], regional volume spanning the left and right thalamus, left fusiform gyrus, and the right superior temporal gyrus [36], left putamen, left and right caudate nucleus, and left insula [37], and volume of the left orbitofrontal cortex [38] (but see [39, 40]). Two studies which did not directly correlate brain volume with performance, linked gait speed to gray matter volume of spatial maps derived based on covariance techniques. The first study used whole-brain multivariate covariance-based analysis; a technique where the first 12 spatial principal component maps were derived from the gray matter maps and then related to gait speed. Significant associations were observed for regions overlapping with the inferior, medial, and superior frontal gyri, fusiform gyrus, cingulate gyrus (both the anterior and posterior parts), the occipital and middle temporal gyri, supplementary motor cortex, primary motor cortex, and thalamus [41]. The second study calculated which brain regions’ gray matter volume was associated with gray matter volume of a cluster in the orbitofrontal cortex that directly correlated with gait speed [38]. Anatomical regions around the orbitofrontal seed region stretching from the left orbitofrontal cortex to the right medial frontal gyrus covaried with the seed region, and were therefore suggested to play a role in gait speed. Reductions in gray matter and increases in ventricular volume can both represent cerebral atrophy. Reduced gait speed has also been linked to larger lateral cerebral ventricle volume in individuals with MCI [42].

White matter lesions

Three studies related semi-quantitative measures of either subcortical or periventricular white matter lesions to gait speed in MCI participants [42–44] (see Table 4), and one study related WML of five subcortical regions to gait speed in AD participants [24]. None of these studies reported significant associations after adjusting for potentially confounding factors.

Functional network connectivity

Brain functional connectivity can be derived from fMRI resting state scans. Connectivity strength can be assessed between pre-selected regions of interest, within networks where a network is defined as two or more brain regions that show coactivation, or between networks. Two studies within MCI participants showed that faster gait speed was related to stronger connectivity between a) the default mode network and the supplementary motor area [45] (see Table 4); and b) the supplementary motor area with the frontoparietal network, bilateral frontal eye fields, and bilateral superior lateral occipital cortex [46]. The later study also reported that a stronger functional connection between the supplementary motor area and the bilateral ventral visual cortices was predictive of slower gait speed.

Amyloid deposition

Higher global amyloid burden has been related to slower gait speed in a mixed group of individuals with MCI and individuals with physical and cognitive frailty [47] (see Table 4). Subregion analysis showed that slower gait speed was also associated with a higher amyloid burden in the basal ganglia, hippocampus, precuneus/posterior cingulate cortex, parietal cortex, temporal cortex, and frontal cortex.

Stride measures

Five studies related a variety of stride measures to gray matter volume and cortical thickness, while one study reported on the association between stride measures and WML burden. This suggest that widespread gray and white matter degeneration affects multiple aspects of gait, leading to disrupted ambulation in MCI and AD. Step length was inversely related with cortical thickness of the right middle frontal caudal and rostral cortex in aMCI participants [35] (see Table 4). Step length variability was negatively related to regional volume of the right middle temporal gyrus, right middle, superior and inferior frontal gyri, right precentral gyrus, left supplementary motor cortex, precuneus, fusiform, putamen, lingual, parahippocampal gyri and left cerebellum in aMCI participants [35]. In AD participants, stride length negatively correlated with WML severity in subcortical frontal regions and in the basal ganglia [24]. In MCI participants, stride length was positively associated with volume of the right thalamus, right parahippocampal gyrus, and right superior temporal gyrus [36]. Stride time variability was inversely related to volume of the primary motor cortex in MCI participants [34], but was not linked to hippocampal volume in individuals with aMCI, non-amnestic MCI, and AD, although the association in aMCI was marginally significant [48].

Gait disorder

A study in individuals with probable AD used the Rating Scale for Gait Evaluation in Cognitive Deterioration [49] to assess gait dysfunction. It was reported that less gray matter volume in clusters in the bilateral primary motor cortices, bilateral medial cingulate gyrus, bilateral insula, and bilateral anterior cerebellum indicated poorer gait [50] (see Table 4). Diffusion imaging was used in this sample to measure fractional anisotropy (FA); a metric of directionality of water diffusion in tissue. Among others, deterioration of white matter tissue can lead to lower FA values. Gait dysfunction in the sample of probable AD participants was linked to lower FA values in clusters within the right corticospinal tract, bilateral cingulum, left inferior longitudinal fasciculus, bilateral inferior fronto-occipital fasciculus, left forceps minor, left superior longitudinal fasciculus, and right anterior thalamic radiation (anatomical labels for the coordinates reported by Olazaran et al., were obtained from the JHU White-Matter Tractography Atlas [51]). In a different sample of AD participants, expert-rated gait disorder was found to be significantly associated with both WML in the subcortical/centrum semiovale and periventricular WML [52]. However, in an independent study in AD participants, subcortical WML were not associated with gait performance that was measured using the Tinetti gait assessment tool [24, 28].

Functional mobility

The majority of reviewed studies have used the timed up-and-go (TUG) test as a measure of functional mobility. Certain studies differentiated between physically completing the TUG (realized TUG; rTUG) and mentally executing the TUG without physically performing the actions (imaginary TUG; iTUG). The iTUG test allows for assessing higher (brain) level of gait control without the interference of physical limitations [53]. The time difference between iTUG and rTUG (ΔTUG) is a measure of motor functioning corrected for cognitive control. Overall, poorer functional mobility in MCI was associated with less frontal, temporal, and cerebellar volume, periventricular WML, and regional amyloid deposition (see below for details).

In individuals with non-amnestic MCI, rTUG was associated with smaller total white matter, total gray matter, and left and right hippocampal volume [53] (see Table 4). Increased iTUG completion time was related to smaller total gray matter volume and smaller left premotor cortex volume. Increased ΔTUG was associated with greater left prefrontal cortex volumes. No significant correlations were observed in aMCI participants. In a mixed sample of non-amnestic MCI and aMCI, significant correlations were found between ΔTUG and volume of clusters in the entorhinal cortex, amygdala, parahippocampal gyrus, insula, and hippocampus of the right hemisphere [54]. Slower TUG performance has also been associated to smaller gray matter volume of the cerebellum lobule VIII (bilateral) and the left middle cingulate cortex [55]. Post-hoc covariance analysis showed that the gray matter of the this left cerebellar region covaried with gray matter of the bilateral cerebellar regions, right middle cingulate cortex, left middle frontal gyrus, right thalamus and precuneus.

Presence of periventricular WML in individuals with MCI resulted in slower performance on the TUG [43] (see Table 4). In one report, functional mobility was expressed as a composite measure of number of falls in the previous 6 months, the walking-while-talking test, the timed up-and-go test, gait speed, and static balance [56]. Within a group of MCI participants, this outcome was linked to periventricular lesions around the frontal horns, but not to cerebral ventricle size.

Larger amyloid-β accumulation within the basal ganglia and the precuneus/posterior cingulate cortex has been related to slower TUG performance in a combined sample of individuals with MCI and cognitive and physical frailty [47] (see Table 4).

Time to complete the TUG test was related to total volume of WML in MCI [57] (see Table 4). However, in this same sample, performance on a TUG test, during which the cognitive load was increased by adding a serial subtraction task, was not associated with WML load.

Gait and cognition dual tasking

Volumetric and functional connectivity studies suggest that degeneration of the primary motor cortex and temporal lobe, as well as degeneration and activation changes of the default mode network are implied in poorer gait-cognitive dual tasking in MCI. Dual tasking comprised of walking while counting backwards from 100 was related to gray matter volume of regions spanning the right middle temporal gyrus, left and right parahippocampal gyrus, right fusiform gyrus, left inferior temporal gyrus, left and right middle occipital gyrus, left cuneus, right posterior cingulate, left anterior cingulate, left and right precuneus, and left cingulate gyrus [41] (see Table 4). Dual task cost under two conditions (counting backwards from 100 by one; subtracting serial sevens from one hundred) was inversely associated with volume of the left hemispheric entorhinal cortex [39]. Gait velocity during dual tasking (subtracting serial sevens from one hundred) was positively associated with volume of the primary motor cortex [34]. Two studies tested for the associations between deep and subcortical or periventricular WML and dual tasking performance, but no statistically significant findings were observed [43, 44]. Gait speed during walking and simultaneously subtracting serial sevens, relative to normal walking speed, was inversely related to functional connectivity within the default mode network [45].

Studies on non-gait motor function in relation to neuroimaging without healthy control comparison

Balance and trunk stability

Balance and trunk stability refer to the ability to maintain a vertical position during walking and while standing still under different conditions (e.g., one-legged, or with feet in front of one another). The studies reviewed here indicate that increased functional connectivity and degeneration of white matter reflect balance problems in MCI. A composite score of bipedal standing balance performance that comprised measures of visual, vestibular, and proprioceptive contributions to balance, as well as number of falls during a unipedal balance test were related to nucleus accumbens volume in AD [58] (see Table 5). More postural sway was related to greater functional connectivity of the default mode network (DMN) with the supplementary motor area in MCI participants [45]. One legged stance balance performance was related to presence of periventricular WML in a sample of unspecified MCI [43], but was not related to atrophy of the medial temporal lobe in a sample of aMCI participants [40]. Smoothness and stability of trunk movements during a dual tasking walking test (counting back by 7 from 100) was inversely related with subcortical and deep or periventricular WML load in MCI participants [44]. No link was observed in MCI participants between total volume of WML and balance measured using the figure-eight walking test [57].

OPEN IN VIEWER

Table 5

Aggregated results for studies reporting associations between non-gait motor performance and neuroimaging in participants with MCI or AD diagnosis without healthy control comparison

Text inside the cells indicate the first few letters of the first author from the manuscript from which the result is extracted. If necessary, these letters are followed by two numbers indicating the year the study was published to differentiate between different publications that have the same first author. If no side (hemisphere; (L) or (R)) is indicated, studies looked at bilateral regions. = Amnestic MCI; = undifferentiated MCI; = undifferentiated MCI (aMCI +naMCI) or a clinical dementia rating scale score of 0.5 without dementia; = AD; * = On a hard surface floor, but not on a foam surface.

Manual motor performance

Lower grip force strength has been associated with higher global and regional (i.e., the frontal, temporal, parietal cortices, the precuneus/posterior cingulate cortex, the hippocampus, and basal ganglia) amyloid-β depositions in a sample of MCI [47] (see Table 5). Furthermore, slower manual reaction time speed (pressing a button when a stimulus is presented on a screen) in a sample of AD participants was linked to less dopamine receptor availability in the sensorimotor region of the striatum [59].

Performance on measures of tapping speed and manual dexterity (i.e., the ability to make coordinated hand and finger movements to grasp and manipulate objects [60]), have not been linked to neural measures in MCI and AD. In a group of AD participants, no significant association was observed between dopamine receptor availability and dominant hand performance on the grooved pegboard test [59] (see Table 5). Studies associating left and right hand finger tapping speed with whole brain WML in AD [61] or hand tapping frequency with whole brain WML in MCI [62] also did not report significant links.

Physical performance and knee extension strength

Physical performance refers to overall motor function spanning multiple motor domains. Cardio-respiratory fitness, as approximated by performance on a 6-minute walking test, has been inversely linked to medial temporal lobe atrophy in aMCI [40] (see Table 5). Another study defined physical fitness as a composite score of a balance measures, gait speed, and rise-from-chair performance. In a combined sample of participants with MCI or physical and cognitive frailty, better performance on this summary measure was linked to higher amyloid depositions in the basal ganglia and the precuneus/posterior cingulate cortex [47].

Left hippocampal volume has been positively related to isokinetic (resistance is placed on the muscle that makes a full range of motion) knee extension strength in AD [63] (see Table 5). Isometric (muscle activation without movement; joints stay static and the muscle does not change in size) knee extension strength on the other hand, did not show a significant relationship with volume of the medial temporal lobe in aMCI [40].

Symptoms of parkinsonism

Within a sample of participants with single-domain aMCI, those who scored positive on at least one motor category of the UPDRS (i.e., tremor, rigidity, bradykinesia, or gait/balance/axial) were more likely to have periventricular WML, lacunae in the basal ganglia (small CSF filled cavities [64]), small vessel disease, and cortical atrophy [65] (see Table 5). Within multi-domain aMCI participants, such symptoms of parkinsonism were associated with presence of deep/subcortical and periventricular white matter lesions, and cortical and subcortical atrophy.

Integrated summary

Out of the 37 studies we identified in our literature search, nine formally tested if associations between motor function and neural measures differed between a control group and participants with MCI and/or AD. These studies found that slower gait speed in individuals with MCI and AD was related to smaller hippocampi [23, 27]. They also found that individuals with MCI had a larger decline in prefrontal cortex activation during walking at normal pace [26]. Participants with MCI also showed less prefrontal brain activation but stronger connectivity of the left prefrontal cortex with the left motor cortex and the right occipital cortex when walking while performing a cognitive task [21, 26]. Gray matter of the inferior temporal lobule and inferior and middle temporal gyrus were strongly related to motor-cognitive dual tasking performance in MCI, while this association was weak to absent in control participants [30]. Also, the extent of increase in prefrontal brain activation that is seen in controls when adding a cognitive task while walking is smaller in MCI [30]. Furthermore, participants with MCI and AD have more activation in cerebellar, cingulate, cuneal, somatosensory, and fusiform brain regions when performing a hand squeezing task [31], and the relative excess of subcortical WML in AD is associated with more signs of parkinsonism, poorer performance during a task requiring simultaneous cognitive and motor performance, and poorer functional mobility [24, 25]. This suggests that motor dysfunction in MCI and AD is related to both the characteristic medial temporal lobe degeneration as well as small vessel disease. It furthermore indicates that motor deficits in MCI and AD are related to the activation and interconnection of an elaborate network of motor brain regions that span cortical, subcortical, and infratentorial regions. The differential association between gait and prefrontal cortical brain activation further reflects the cognitive component in gait and is in line with studies showing that gait relies on higher cognitive functions such as attention and executive functioning [66] and that slower gait has been linked to frontal gray matter volume reductions [67].

The majority of studies that we identified reported relationships between motor function and neural metrics stratified for MCI/AD and control samples, without reporting the neural predictor-by-group interaction effects. With such statistical designs, it is not possible to conclude that disease-related neurodegeneration is responsible for any of the observed motor impairments in MCI/AD. Nonetheless, in light of the limited studies that do report interaction effects, the outcomes of these studies provide some insight in the brain regions and degenerative processes potentially underlying motor impairments in MCI and AD. Together, these studies suggest that slower gait speed in MCI and AD is related to less gray matter volume in frontal [34, 41], temporal [35, 41], cingulum [41], and subcortical brain regions [36, 41], larger lateral ventricles [42], larger widespread white matter degeneration [24, 42–44], global build up of amyloid-β [47], and decreases in coherence between the sensorimotor network and the frontoparietal network [45, 46].

Other aspects of gait, such as stride time variability and step length, were found to be related to volume of the primary motor cortex [34] and precentral gyrus, supplementary motor cortex, cerebellum and parahippocampal gyrus respectively [35]. Motor tests that incorporate gait, such as gait-cognitive dual tasking, and the timed up-and-go test, were related to degeneration of the primary motor cortex and temporal lobe [34, 41], as well as degeneration and activation changes of the default mode network (dual tasking) [45], and less frontal, temporal, and cerebellar volume [55], periventricular WML [43, 56], and amyloid deposition in the precuneus and cingulum [47]. Gait disorders in AD were related to volume loss of motor brain regions such as the primary motor cortex and the cerebellum [49], subcortical and periventricular WML [52], and microstructural white matter changes of several white matter projections, including the corticospinal tract [50].

Other motor functions that were investigated in MCI and AD samples were much less often investigated than gait, which makes it difficult to draw firm conclusions. These non-gait studies are suggesting that balance deficits in MCI and AD are related to periventricular and subcortical WML [41, 43] and connectivity between the supplementary motor area and DMN [45]. There were some indications that grip weakness is related to global amyloid deposition [47], and that manual reaction time in AD is related to striatal dopamine binding [59]. Symptoms of parkinsonism in amnestic MCI were associated with subcortical vascular pathology, periventricular WML, and cortical atrophy [65], whereas overall physical performance was related to medial temporal lobe atrophy [40], amyloid deposition in the parietal lobe and basal ganglia [47]. It should be noted that lack of statistically significant associations is not proof for the absence of such associations. To the contrary, in light of the limited number of investigations and often small sample sizes, future studies should continue to investigate them to provide more conclusive evidence while Bayesian models can be used to formally test the absence of relationships between brain measures and motor function [68].

In summary, the reviewed studies indicate that gait and gait-related dysfunction in MCI and AD is related to atrophy of medial temporal lobe and motor brain regions, ventricular enlargement, global amyloid deposition, white matter damage, disturbed connectivity between motor and frontal regions, and prefrontal hypoactivation. There was limited evidence that dysfunction of non-gait motor functions in MCI and AD are affected by neurodegeneration specific to AD pathology, but this is mainly due to lack of investigations of non-gait motor behavior.

Confounding factors and limitations

Despite the body of work completed in this area, there are a number of noteworthy limitations that were revealed through this review. All but two studies had cross-sectional study designs, thereby limiting our understanding of the direction of the association and any indication of causation. Another limitation of the studies reviewed here is that many studies had small sample sizes, with 13 studies having at least one experimental group with fewer than 30 participants. While the sample sizes for studies reporting group-by-brain interactions ranged from 9–43 per experimental group, the remaining studies had experimental samples with a larger upper bound, ranging from 20 to 347. The negative findings of studies with small studies should thus be interpreted with caution.

The majority of studies included only MCI patients while a minority included AD patients. Only two studies included both MCI and AD participants. Studies that do include both groups and use designs to test differences in motor function and brain associations can improve our understanding on how these associations develop with the progression of AD pathology.

The majority of studies reviewed did not distinguish between amnestic and non-amnestic MCI. Whereas the greater field has not consistently and specifically characterized the subtypes of MCI examined, it may not be surprising that these motor/imaging studies have perpetuated this limitation. However, failing to more clearly characterize one’s sample may stymie true advancement. For example, a sample of poorly defined MCI may reflect AD, cerebrovascular disease, or dementia with Lewy bodies. This might be particularly important, as Lewy body disease is associated with slowness of gait [69] and other parkinsonian features [70]. On most motor behavioral measures Lewy body patients generally perform worse than individuals with MCI due to AD [71] and motor dysfunction in Lewy body disease has been associated with neural degeneration such as white matter microstructural changes and frontal and parietal cortical thinning [72] as well. Future studies should focus on participants with amnestic MCI when studying the neural mechanisms of motor dysfunction due to AD to ensure we are not in fact studying the mechanisms of motor dysfunction in MCI due to Lewy body disease or other non-AD etiologies.

Thirteen studies that we reviewed had experimental groups with fewer than 30 participants. These studies may have been underpowered to detect true associations between motor function and neural measures. Therefore, the negative findings of these small studies should be interpreted with caution.

Few studies have investigated the same motor behavior with the same neuroimaging outcome measures, as can be seen from the overview tables. Although we report convergence of results despite differences in methodology, mixed results in associations between motor measures and brain metrics may have been due to differences in study design, predictive, and outcome measures. More consistency in both the motor measures used and the neuroimaging modalities will help determine the robustness of their associations.

Lastly, the reviewed studies did not look at sex differences in the relationship between motor function and brain measures, but merely adjusted for its effect statistically. It is therefore uncertain if the relationship between motor differences and brain differences in MCI and AD differ by sex. The fact that shared risk factors for AD and brain disease, such as hypertension [73, 74], are also related to motor performance such as gait speed [75], and the fact that there are sex differences in motor performance [76], neuroanatomy [77], neurobiology [78], and age-related neuropathology [79] warrants studying sex differences. This idea is supported by reports showing sex differences in motor behavior in AD, such as relationship between cognitive function and muscle strength in AD males that is absent in AD females [80], and poorer balance performance in MCI and AD compared to cognitively normal controls in females, but not in males [6]. Nevertheless, the small sample sizes of the reviewed study may have prevented them to detect true sex differences, which further stress the necessity of larger sample sizes in future investigations.

Future perspectives

Need for integration of multi-modal imaging

Overall, motor dysfunction in MCI and AD appears to be related to multi-system neuropathology including atrophy, small vessel disease, and amyloid deposition. As such, it is important to study motor dysfunction in MCI and AD using multi-modal imaging to improve our understanding of the neural mechanisms of motor dysfunction in these conditions. Currently, most studies report on brain-motor behavioral associations of single imaging modalities. This prevents the development of models that consider these neural predictors to interact. Future studies should consider an integrated approach that includes multi-modal imaging predictors when studying the relationship between neural degeneration and motor dysfunction in MCI and AD [96].

Necessity of studying brain regions beyond hippocampal volume and frontal circuits

Gray matter volume, specifically that of the hippocampus, was the most studied neural predictor of motor dysfunction in MCI and AD. Brain volume of other regions that were most often related to gait function in MCI and AD included frontal (motor) regions, subcortical areas, and medial temporal lobe structures. Other regions that play a crucial role in motor function [97] and which have reported to be affected in AD [98–100], such as the cerebellum, were only investigated in five studies. Considering our current limited understanding of the neural underpinnings of motor dysfunction in MCI and AD, it is warranted to use both hypothesis free (e.g., voxel-wise analysis) as well as more powerful hypothesis driven (e.g., region of interest analysis) techniques.