Contribution of Brain Imaging to the Understanding Of Gait Disorders in Alzheimer’s Disease

Abstract

Although gait disorders are common in Alzheimer’s disease (AD), determining which brain structures and related lesions are specifically involved is a goal yet to be reached. Our objective was to systematically review all published data that examined associations between gait disorders and brain imaging in AD. Of 486 selected studies, 4 observational studies met the selection criteria. The number of participants ranged from 2 to 61 community dwellers (29%-100% female) with prodromal or dementia-stage AD. Quantitative gait disorders (ie, slower gait velocity explained by shorter stride length) were associated with white matter lesions, mainly in the medial frontal lobes and basal ganglia. The nigrostriatal dopamine system was unaffected. Qualitative gait disorders (ie, higher stride length variability) correlated with lower hippocampal volume and function. Gait disorders in AD could be explained by a high burden of age-related subcortical hyperintensities on the frontal–subcortical circuits (nonspecific) together with hippocampal atrophy and hypometabolism (specific).

Introduction

Gait disorders are common throughout the course of Alzheimer’s disease (AD), with a prevalence increasing with the stages of AD. ^1,2 Interestingly, it has been suggested that AD-related gait disorders (ADRGDs) could be not only a companion event of AD but also rather a specific sign of AD-related cognitive decline. ¹ The study of ADRGDs is attractive in the sense that they predict adverse outcomes including falls, loss of independency, institutionalization, hospitalization, and death, ^3,4 and also because they are thought to be useful for the early diagnosis of AD. For instance, it was recently reported that gait disorders illustrated by increased gait variability could be used as a clinical marker of prodromal AD, also called amnestic mild cognitive impairment (aMCI). ^{5 –8} The ADRGDs may even appear first in the course of AD, before memory impairments. Two longitudinal studies ^9,10 conducted on healthy elderly participants found that reduced gait velocity could be observed prior to the development of cognitive impairment in the oldest old. Also, Stark et al ¹¹ recently reported more falls, in most cases resulting from gait disorders, during an 8-month follow-up in 125 cognitively healthy seniors (mean age 75 years) when positron emission tomography scans showed large deposits of amyloid-β plaques indicating preclinical AD. Unfortunately, the authors did not specify which brain structures were most affected by the amyloid-β deposits among fallers.

Highlighting that gait disorders may help early detection of AD is informative; however, understanding the mechanisms responsible for gait disorders in AD seems even more important to develop new strategies for the early management of the disease and to avoid further mobility decline and falls. ¹² In functional terms, ADRGDs have been related to the impairment of higher levels of gait control at subcortical and cortical levels. ^{13 –15} Yet, it remains unclear which brain structures and related lesions are specifically involved in AD and could explain ADRGDs. This question has not received yet a structured critical evaluation, unlike gait disorders in vascular dementias or in subcortical degenerative dementias and Parkinsonism, whose links with, respectively, vascular ischemic lesions ¹⁶ and disorders of the basal ganglia ¹⁷ are well described. Our purpose was to systematically review all published literature that examined the relationships between gait disorders and morphological, metabolic, or functional brain imaging in patients with AD.

Methods

Literature Search

A systematic Medline literature search was conducted in July 2011, without limit of date and language restriction, using the Medical Subject Heading terms “Gait” OR “Gait Disorders, Neurologic” OR “Walking” combined with “Brain Mapping” OR “Magnetic Resonance Imaging” OR “Tomography, X-Ray Computed” OR “Tomography, Emission-Computed, Single-Photon” OR “Positron-Emission Tomography” OR “Nuclear Medicine” OR “Brain” combined with “Alzheimer disease” OR “Dementia” OR “Mild cognitive impairment.” The search also included the Cochrane Library (Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effectiveness, and Cochrane Controlled Trials). An iterative process was used to ensure all relevant articles had been obtained. A further hand search of bibliographic references of extracted articles and existing reviews was also conducted to identify potential studies not captured in the electronic database searches.

Study Selection and Analysis

One member of the team (C.A.) screened abstracts from the initial search and obtained articles deemed potentially relevant. Initial screening criteria for the abstracts were (1) observation studies (ie, case report, case series, cross-sectional, case–control, and cohort studies); (2) intervention studies; (3) data collection of gait as outcome; (4) data collection of brain mapping (whether morphological, metabolic, or functional) as outcome; (5) diagnosis of degenerative cortical dementia and/or white matter hyperintensities (WMHs); (6) involvement of adult human participants aged 18 years and older; and (7) article written in Latin alphabet. If a study met the initial selection criteria or its eligibility could not be determined from the title and abstract (or abstract not provided), the full text was retrieved. Two reviewers (C.A. and O.B.) then independently assessed the full text for inclusion status. Disagreements were resolved by a third reviewer (M.M.O.). The full articles were screened using the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) checklist which describes items that should be included in reports of cohort studies ¹⁸ and the Consolidated Standards of Reporting Trials (CONSORT) statement for clinical trials. ¹⁹ Final selection criteria were, therefore, applied when gait assessment and brain mapping were provided and when the participants presented with AD (ie, preclinical AD, prodromal AD, or dementia-stage AD). The study selection is shown on a flow diagram (Figure 1).

Figure 1.

Flow diagram of selection of studies focusing on gait disorders and brain mapping in Alzheimer’s disease.

Of the 486 originally identified abstracts, 39 met the initial inclusion criteria. ^{20 –58} Following thorough examination, we excluded 35 of those 39 studies because no diagnosis of AD was made (vascular lesions as single outcome, n = 22; other nondegenerative dementias, n = 6; other degenerative dementias, n = 4; no cognitive disorder, n = 2) or because no data were available for brain imaging (n = 1). The remaining 4 studies were included in this review. ^{21 –24} Articles selected for the full review had the following information extracted: authors, date of publication, study design, settings and study population, diagnosis criteria for dementia-stage AD or prodromal AD, brain imaging techniques, gait assessment methods, and description of brain imaging and gait parameters.

Results

Table 1 summarizes the 4 observational studies included in this review. ^{21 –24} The number of participants with AD ranged from 2 to 61. ^21,23 Women represented 28.6% to 100% of participants with AD. ^21,23 Participants were more likely to be older adults, with a mean age calculated at 74.95 years. Data collection was based on 1 case series ²¹ and 3 cross-sectional studies. ^{22 –24} No clinical trials were found. Two studies involved patients with dementia-stage AD^21,22 and the other 2 focused on prodromal AD. ^23,24 The cross-sectional studies also included either cognitively healthy controls ^22,24 or people with nonamnestic MCI. ²³ Regarding brain mapping, all studies used magnetic resonance imaging (MRI). Rossor et al ²¹ also utilized other brain imaging techniques (ie, computed tomography scan [CT scan], electroencephalography, and positron emission tomography [PET]). Zimmerman et al ²⁴ also used proton magnetic resonance spectroscopy (H ¹ -MRS). Studies focused either on brain morphology ^{21 –23} or vascular burden, ^{21 –23} or brain metabolism. ^21,24 With the exception of the case series that evaluated gait only clinically, ²¹ cross-sectional studies assessed gait disturbances with the use of functional tests (gait velocity at normal or fast pace, Timed Up and Go test (TUG), Walking while talking test, and Tinetti test) ^{22 –24} or computerized treadmill with embedded pressure sensors (GAITRite, CIR Systems, Pennsylvania). ^22,24 All parameters were measured at steady state walking and were averaged over 2 or 3 walking tests ^22,24 . The parameters studied reflect gait either quantitatively (ie, gait velocity in cm/s, ^{22 –24} stride length in cm, ^22,24 cadence in steps/min, ²² stride width in cm ²² and swing time in seconds ²⁴ ) or qualitatively (ie, dynamic stability assessed with stride length variability in standard deviation ²⁴ ). Due to its design, the case series only described the brain abnormalities observed in AD cases with gait disorders, ²¹ while cross-sectional studies used simple comparisons and models of correlation and regression to examine the relationships between gait and brain. ^{22 –24} Due to the heterogeneity of the methods, the results could not be meta-analyzed.

Table 1.

Main Characteristics of the 4 Studies Exploring the Association Between Gait Disorders and Brain Imaging in Patients With Alzheimer’s Disease

Study			Outcomes				Results
References	Design	Settings/population	Brain imaging technique	Brain outcome	Gait assessment	Gait outcome	Relationships between gait parameters and brain imaging
Rossor et al 21	Case series	Location: United Kingdom Community-dwelling N = 2 cases with profound gait disturbances Case 1: Female 70 years old at the beginning of health problems Cognitive performance: WAIS verbal IQ = 78 WAIS performance IQ = 72 Case 2: Female 62 years old at the beginning of health problems Cognitive performance: WAIS verbal IQ = 100 WAIS performance IQ = 106 Diagnosis of AD based on necropsy (senile plaques, neurofibrillary tangles and large swollen eosinophilic neurons)	CT scan (cases 1 + 2) EEG (cases 1 + 2) MRI (case 2 only) 15 O2 steady state PET (T1/931/08/12 scanner, CTI Knoxville, TE, USA; cases 1 + 2) 18 F-labelled fluorodopa uptake (cases 1 + 2)	CT scan: Brain size Vascular lesions EEG: Size, frequency an location of waves 15 O2 PET: Regional values of CMRO2 for each of the 18 slices of an anatomical atlas 18 F-dopa: Ki for 18 F-dopa in the nigrostriatal dopamine system	Clinical examination: ability to get up and walk	Inability to walk without assistance and confinement to wheelchair, with a quick installation in 2 (case 1) or 3 years (case 2)	CT scan: Case 1: mild diffuse atrophy Case 2: normal except a doubt on a small ischemic area in white matter EEG: Case 1: modest intermediate slow waves frequencies in the temporal region Case 2: low amplitude but otherwise normal MRI: Case 2: minimally increased signal around anterior parts of the lateral ventricles 15 O2 PET: Cases 1 and 2: symmetric areas of hypometabolism in the medial frontal lobes bilaterally Marked reduction in the anterior frontal gyrus and cingulated gyrus Significant difference with a normal group for medial frontal cortical areas Nonsignificant trend for a difference with a normal group for some temporal and parietal areas 18 F-Dopa: Ki within the normal ranges in caudate and putamen
Nadkarni et al 22 Sunnybrook Dementia Study	Cross- sectional	Location: Canada Community-dwelling N = 42 mild AD visiting a university memory clinic Able to walk independently for 15 minutes without discomfort 60% female Mean age 74 ± 8 years Cognitive performance: MMSE ≥ 20 (mean MMSE score = 25 ± 3) Diagnosis of AD based on NINCDS/ADRDA criteria, after exclusion of major depression, history of other neurological disorders and stroke, sedative medication or neuroleptic use, or alcohol dependence N = 22 normal controls 47% female Mean age 73 ± 8 years	MRI 1.5-Tesla Signa MR imager (GE Medical systems, Milwaukee, Wisconsin) T1-weighted, proton density and T2-weighted images	Total and regional subcortical WMH scored on the 4-point ARWMC visual rating scale 5 areas: frontal area, parieto-occipital area, temporal area, infratentorial area (ie, brainstem and cerebellum), and basal ganglia (ie, striatum, insula, globus pallidus, thalamus, internal, and external capsules) Total ARWMC score = sum of degree of subcortical WMH in the 5 areas Binarized total ARWMC score, according to the median split of the score (ie, low versus high score)	Walk test at most comfortable pace on a computerized 12 × 2 feet long walkway mat with embedded pressure sensors (GAITRite, CIR Systems) TUG test Tinetti test	Parameters (steady-state walking; mean values of 3 walk tests): Gait velocity (cm/s) Stride length (cm) Cadence (steps/min) Step width (cm) Time to perform TUG (s)	Among participants with AD: No significant difference for gait velocity, stride length, cadence, step width, TUG score and Tinetti gait score between high and low total ARWMC score (P value not shown) The total ARWMC score correlated with stride length (r = −.4, P = .01) and showed a trend toward significance for gait velocity (r = −.2, P = .07) The stride length correlated significantly with ARWMC score in the frontal area (r = −.4, P < .05) and basal ganglia (r = −.4, P = .01), but not in the parieto-occipital, temporal and infratentorial areas (P values not shown) Comparison of participants with AD with a low total ARWMC score compared to normal controls with a low total ARWMC score: Significantly worse gait velocity (P < .001), step length (P = .04), cadence (P < .01) in AD Nonsignificant trend for a worse TUG score in AD (P = .05) No significant difference for the step width and the Tinetti gait score (P value not shown) Comparison of participants with AD with a high total ARWMC score compared to normal controls with a high total ARWMC score: No significant difference for the gait velocity, step length, cadence, step width, TUG score and Tinetti gait score (P value not shown)
Onen et al 23 MRI CODE study	Cross-sectional	Location: France Community-dwelling N = 61 MCI 54.1% female Mean age 71.70 ± 5.52 years N = 8 amnestic MCI N = 11 with periventricular WMH Diagnosis of mild memory impairments based on (1) memory complaints, (2) low memory (MMSE 3-word list recall) and/or attention performance (IST), (3) relatively preserved general cognitive functioning (MMSE) (4) relatively intact activities of daily living (IADL), (5) not demented (DSM-III).	MRI 1.5-Tesla imaging system 3D T1-weighted coronal fast SPGR echo acquisition sequence T2-weighted coronal fast FLAIR sequence	Presence of periventricular WMH (defined as WMH directly contiguous to the ventricles, with thickness > 5 mm)	TUG Walking while talking test 14-m walk test at fast pace	Poor TUG (time to perform TUG > 10seconds) Stop walking when starting a conversation Time to walk 14 m at fast pace categorized as follows: <1.25 m/s; 1.25-1.50 m/s; >1.50 m/s	Comparison between participants with periventricular WMH and participants without periventriculat WMH: Participants with periventricular WMH more often underperformed the TUG (P = .0018) No between-group significant difference for the Walking while talking test and for the categories of gait velocity (P values not shown) Among the whole cohort: Positive associations of periventricular WMH with poor TUG, longer time to walk 14 m and altered Walking while talking test (P values not shown) Positive associations of periventricular WMH with gait disturbances disappeared after adjustment for age and ventriculomegaly, except with poor TUG (P = .0022)
Zimmerman et al 24 Einstein Aging Study	Cross-sectional	Location: USA Community-dwelling N = 14 amnestic MCI 28.6% female Mean age 83.14 ± 4.20 years N = 2 with abnormal clinical gait Diagnosis of mild memory impairments based on the CDR, BIMC and FCSRT-IR scores From a cohort of 48 nondemented older adults 47.9% female Mean age 81.18 ± 5.47 years N = 7 with abnormal clinical gait	MRI Varian-Magnex 4-Tesla imaging system T1-weighted images H 1 -MRS Varian INOVA console and quadrature head coil using a modified LASER sequence	Midsagittal area as a surrogate measure of the whole brain volume Hippocampal volume (combination of left and right volumes) Hippocampal NAA/Cr (mean value of left and right ratios)	Walk test at most comfortable pace on a computerized 180 × 35.5 inches long walkway mat with embedded pressure sensors (GAITRite, CIR Systems)	Parameters (steady-state walking; mean values of 2 walk tests): Gait velocity (cm/s) Stride length (cm) Stride length variability (SD) Swing time (s)	Among participants with aMCI: Stride length correlated with hippocampal NAA/Cr (r = .56, P = .04) but not with hippocampal volume (P value not shown) Stride length variability correlated with hippocampal NAA/Cr (r = −.56, P = .04), but not with hippocampal volume (P value not shown)
							Among the whole cohort: Gait velocity Gait velocity correlated with hippocampal volume (r = .50, P < .01), but not with hippocampal NAA/Cr (P value not shown) Stride length Stride length not correlated with midsagittal area (r = .14, P = .39) Stride length correlated with hippocampal volume (r = .45, P = .01) after controlling gender and midsaggital area Stride length associated with hippocampal volume after adjustment for age and midsagittal area (β = .36, P = .03) and also after adjustment for gender and weight (P values not shown), as well as after adjustment for NAA/Cr (β = .42, P = .02), but not after adjustment for gait velocity (β = .01, P = .90) Stride length not associated with hippocampal NAA/Cr (β = .09, P = .48) Stride length variability Stride length variability not correlated with midsagittal area (r = .01, P = .97) Stride length variability correlated with hippocampal NAA/Cr (r = −.36, P = .02), but not with hippocampal volume (P value not shown) Stride length variability associated with hippocampal NAA/Cr after adjustment for age (β = −.38, P < .01), and also after adjustment for weight and gait velocity (P values not shown), and showed a trend towards significance for hippocampal volume (β = −.33, P = .08) Swing time Swing time not correlated or associated to any neuroimaging measures (P value not shown)

Abbreviations: AD, Alzheimer’s disease; ARWMC, age-related white matter change; BIMC, Blessed Information-Memory-Concentration test; CDR, Clinical Dementia Rating scale; CMRO₂, cerebral metabolic rate for oxygen; CT scan, computed tomography scan; EEG, electroencephalography; FCSRT-IR, Free and Cued Selective Reminding Test with Immediate Recall; IADL, Instrumental Activities of Daily Living; IST, Isaac Set Test; Ki, influx constants; MMSE, Mini-Mental State Examination; MCI, Mild Cognitive Impairment; MRI, magnetic resonance imaging; H ¹ -MRS, proton magnetic resonance spectroscopy; MRI CODE study, Magnetic Resonance Imaging of COgnitive DEcline study; NAA/Cr, N-acetylaspartate to creatine ratio; NINCDS/ADRDA, National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer Disease and Related Disorders Association; PET, positron emission tomography; SD, standard deviation; SPGR, spoiled gradient recalled; TUG, Timed Up and Go; WAIS, Wechsler Adult Intelligence Scale; WMH, white matter hyperintensities; IQ, intelligent quotient; DSM III, Diagnostic and Statistical Manual of Mental Disorder (Third Edition); 3D, three-dimensional.

From the morphological point of view, 3 studies focused on the subcortical WMH ^{21 –23} and the third one tested the midsagittal area and hippocampal volume on morphological MRI. ²⁴ Regarding the WMH, Rossor et al ²¹ found, among patients with poor gait function, increased signal intensity around the anterior parts of the lateral ventricles only. In line with this, Onen et al ²³ found that participants with periventricular WMH underperformed the TUG and the Walking while talking test and walked slower than those without periventricular WMH. These associations were yet weakened by adjustment for age and brain atrophy (Table 1). In the study by Nadkarni et al, ²² the WMHs were more precisely rated with the use of the 4-point Age-Related White Matter Change (ARWMC) visual scale. The authors found that the gait parameter most related to the ARWMC score on the whole brain was stride length (r = −.4, P = .01; Table 1). More precisely, they showed that the stride length correlated with the ARWMC score in the frontal area (r = −.4, P < .05) and the basal ganglia (r = −.4, P = .01), but not in other brain areas. Nevertheless, when comparing participants with AD with a low total ARWMC score to cognitively healthy controls with the same low total ARWMC score, the authors highlighted that participants with AD still had slower gait velocity (P < .001), shorter stride length (P = .04), and lower cadence (P < .01) than controls. ²² Regarding the level of brain atrophy, Zimmerman et al ²⁴ found that neither the stride length nor the stride length variability correlated significantly with MRI-derived midsagittal area measurement, a substitute for whole brain volume (P = .39 and P = .97, respectively; Table 1). Then, focusing on the hippocampus—the brain area more specifically affected in AD—the same authors showed that the hippocampal volume correlated with gait velocity (r = .5, P < .01), was associated with stride length (β = .36, P = .03), and showed a trend toward significance for stride length variability (β = −.33, P = .08). ²⁴

From the functional point of view, ¹⁵ O₂ PET metabolism measurements made by Rossor et al ²¹ from the brain were divided into 5 areas and the nigrostriatal system specifically. Both cases exhibited symmetric areas of low cerebral metabolic rate for oxygen in the medial frontal lobes bilaterally, more particularly in the anterior frontal gyrus and the cingulate gyrus (Table 1). There was also a trend for hypometabolism in some temporal and parietal areas compared with a control group. Conversely, the metabolism of the nigrostriatal dopamine system was unaffected in AD cases with gait disorders as the influx constants for ¹⁸ F-Dopa were normal in caudate and putamen. ¹⁹ In another study using H ¹ -MRS, Zimmerman et al ²⁴ found that the stride length and the stride length variability correlated with hippocampal metabolism measured by a decreased N-acetylaspartate to creatine ratio in the case of short stride length (r = .56, P = .04) or high stride length variability (r = −.56, P = .04; Table 1).

Discussion

Although this systematic review was able to select a limited number of studies, the highlighted findings remained consistent within them and suggested subcortical white matter as well as hippocampus, but not nigrostriatal system, involvement in ADRGDs.

First, Rossor et al, ²¹ Nadkarni et al, ²² and Onen et al ²³ reported the presence of WMH among the patients with AD with quantitative ADRGDs relative to motor power and propulsion abilities, such as slower gait velocity or shorter stride length. WMH are usually considered as evidence of ischemic disease while aging with changes predominantly affecting small vessels and parenchyma in the deep white matter, thereby resulting in the disruption in tracts connecting different parts of the brain. ^32,59 The WMH typically begins proximal to the lateral ventricles and then grows radially outward. ³² Based on studies cited in this review, it appears that quantitative ADRGDs are associated with WMH specifically in the frontal area and the basal ganglia, which are both part of the frontal–subcortical circuits. Disruption in these neuronal circuits may explain the finding of symmetric areas of hypometabolism in the medial frontal lobes of patients with AD unable to walk in Rossor et al’s series, ²¹ and is likely to explain ADRGD of slowdown type as these circuits are involved in upright posture, ⁶⁰ spatial information, ⁶¹ movement initiation, and planning. ⁶² Increasing epidemiological, neuroimaging, and clinical studies suggest that the vascular burden plays a key role in the onset and progression of AD, which is primarily a neurodegenerative disease. ⁶³ It is, therefore, not surprising that the vascular component of AD appears to be involved in gait disorders among patients with AD. Yet, it is interesting to note that, while adjusting for age and brain atrophy, the association of WMH with gait disorders was weakened (Table 1). Moreover, while matching participants for the level of vascular burden, participants with AD still had worse performance on gait velocity (P < .001), stride length (P = .04), and cadence (P < .01) compared with controls (Table 1). These observations are in line with previous studies reporting that the WMH correlated with locomotor performance even among high-functioning elderly individuals. ^42,45,59 Therefore, we suggest that the finding of quantitative gait disorders explained by WMH in AD may be fortuitous and not specific of AD but rather explained by the older age of participants with AD.

Besides this propulsion aspect, our review also provided compelling evidence that qualitative ADRGDs relative to gait dynamic stability and regularity, illustrated by increased stride length variability, may be explained by AD-related cortical misprocessing. Gait variability is generally thought to be a surrogate marker of the efficiency of higher-level motor control supported by gray matter. ¹ In this perspective, Zimmerman et al ²⁴ demonstrated that ADRGDs could not be explained by the atrophy of the brain as a whole, justifying further investigations of specific gray matter areas. In particular, the study by Rossor et al ²¹ showed that ADRGDs may not be attributed to a dysfunction of the deep dopaminergic system since the ¹⁸ F-Dopa uptake into the caudate and putamen was normal in both reported cases. Conversely, Zimmerman et al ²⁴ reported that, in AD, higher stride length variability was associated with lower metabolism in hippocampal cortex, which is precisely the first cortical region damaged in AD. ⁶³ Of note, a prior PET study in healthy younger adults consistently showed an involvement of hippocampus in gait performance and reported an association between higher activity in hippocampal region and increasing complexity of gait (eg, walking and avoiding obstacles). ⁶⁴ Animal models were also consistent with this finding as it has been shown that hippocampus lesions generated memory disorders as well as limb coordination impairments illustrated by higher gait variability. ⁶⁵ Two main explanations may be proposed for this finding. First, the hippocampus has a functional relationship with the prefrontal cortex through the entorhinal cortex and the nigrostriatal system. ⁶⁶ Degeneration of the hippocampus thus causes a disintegration of visual, vestibular, and proprioceptive sensory and contextual information into spatial maps, which is a necessary cognitive resource to maintain normal walking pattern since walking in the real world is a complex task requiring paying attention to environmental features to avoid stumbles and falls. ¹³ Second, the hippocampus is involved in memory of past strategies and retrieval of complex foot movement sequences. ^67,68 As a consequence, learning and memory disorders arising from hippocampus lesion in the course of AD may also explain uncoordinated movements and qualitative ADRGDs.

Gait disorders are common across AD spectrum. Their gravity is related to their high prevalence and their adverse consequences including disability and mortality. ^3,4 Thus, understanding the determinants of ADRGDs, and in particular which part of the brain is involved and by which mechanism, may help elucidate how to maintain function late in life and prevent adverse health events as long as possible in the course of AD. The present systematic review emphasizes that gait disorders in AD are explained by age-related vascular ischemic burden independent of AD, as well as by AD-related cortical misprocessing. As a consequence, strategies for ADRGDs management should fix each of these domains—for example, by combining cerebrovascular prevention ⁶⁹ with pharmaceutical or nonpharmaceutical cognitive enhancement ^70,71 —or both at once. In this perspective, cross-domains therapies such as physical training, which limits vascular risk and improves performance in physical and cognitive domains at once, ^72,73 appear of particular interest.

The findings of our systematic review need to be tempered by a number of limitations. First, this is a relatively new and emerging area of research, and few studies have been conducted to date, which narrowed the amount of studies to be included in this systematic review. Therefore, our conclusions need to be confirmed in larger studies. The heterogeneity of the studied populations must also be considered. For instance, the stage of AD differed depending on the study; Onen et al ²³ and Zimmerman et al ²⁴ included prodromal participants with AD while Nadkarni et al ²² examined participants with mild AD, and Rossor et al ²¹ reported 2 severe atypical AD—corticobasal degeneration cases. Moreover, the 4 studies included in this literature review did not assess the same morphological measurements and did not use the same imaging techniques, with the exception of MRI. Harmonization of outcome measures seems desirable.

Conclusions

In conclusion, the present systematic review shows that ADRGD relative to propulsion appear to be explained by age-related vascular burden, while ADRGD relative to dynamic stability are more likely explained by AD-related cortical misprocessing in hippocampus. These associations should be considered with caution due to the limited number of studies on this topic. The role of the dopaminergic system and determining whether hippocampus and/or vascular lesions are specifically responsible for ADRGDs should be assessed in bigger samples. Future studies should include a clear description of the population stratified by AD severity, standardized protocols for quantitative gait analysis, and robust structural, functional and metabolic imaging techniques. Understanding neuroanatomical correlates of ADRGDs may offer a powerful mechanism to act on the health care needs of older adults with AD and to maintain function late in life.