With the accelerating global aging trend, healthcare priorities for the elderly have shifted from single-disease management to comprehensive interventions addressing multisystem comorbidities. In this context, falls, as a critical manifestation of geriatric syndromes, have progressively become a hot topic in terms of their underlying mechanisms and preventive strategies (1). Although falls are prevalent among the elderly population, patients with neurodegenerative diseases face compounded risk due to their unique pathophysiologic features (2). Alzheimer’s disease (AD), a common age-associated neurodegenerative disease, is classically defined by progressive cognitive decline. However, owing to shared neuroanatomical substrates for cognitive and motor functions, a substantial proportion of patients exhibit motor abnormalities, including reduced gait velocity and increased gait variability, which can culminate in falls (3). Recent, falls are increasingly recognized as events reflecting cognitive-motor interactive imbalance, and the risk factors leading to falls of AD patients have not been explicated.

Existing studies indicate that older adults with dementia experience falls at eight times the rate of cognitively intact peers, with significantly higher incidence and severity of fall-related complication (4). Multiple fall risk assessment scales have been developed, including the widely utilized Morse Fall Scale and Fall Risk Assessment for Older Adults; however, their sensitivity in cognitively impaired populations requires further validation. By contrast, international research has focused more extensively on sensory function, balance status, and gait characteristics, and other factors. Gait impairment, one of the determinants of motor dysfunction, is ubiquitous in dementia progression and parallels cognitive decline. Oliveira Silva et al. suggested that the Timed Up and Go (TUG) test could serve as a screening tool for fall risk by quantifying gait performance in individuals with cognitive impairment (5). Nevertheless, while TUG is a standard screening instrument, its predictive capacity for community-dwelling older adults is limited, and it should not be used in isolation (6). Zhang et al. (7) reported that gait variability during dual-task conditions predicted future falls in community-dwelling older adults over a 2-year follow-up, whereas gait velocity did not. Sheridan and Hausdorff (8) further established that stride time variability partially explains the propensity for falls in AD patients. Accumulating evidence supports gait parameters-particularly gait variability, which exhibits superior predictive value over conventional metrics such as gait velocity-are valuable screening tools for identifying high-risk fall individuals suffering AD.

At present, the potential mechanisms underlying fall risk in cognitive impairment remain incompletely elucidated. There is contradictory evidence of the association between cognitive impairment and falls. It was reported that memory deficit was associated with increased fall risk in the elderly with aMCI and mild AD (9). It was also reported that cognition was not associated with the fall outcome in older adults (10). Our study preliminarily compared the cognitive domain differences between falls and non falls in AD through detailed neuropsychological assessment. Moreover, altered brain structure may contribute to explain the risk of falls in AD patients. Studies have demonstrated that AD fallers exhibited more severe periventricular white matter hyperintensity (11) and smaller hippocampal volumes (12) compared with non-fallers. Furthermore, several studies have demonstrated that reduced volumes in multiple brain regions correlated with heightened fall risk in MCI and AD patients (9, 13).

Strengths of this study include the prospective ascertainment of incident falls over 6–12 months, strengthening causal inference regarding predictors, and the novel application of VBM to identify structural brain alterations associated with falls in individuals with aMCI and AD, providing new insights into potential neural mechanisms. Besides, to clarify the relationship between cognitive impairment and fall risk, our study preliminarily compared the cognitive domain differences between falls and non falls in AD. In summary, our study preliminarily investigated fall risk factors in a limited sample AD queue through integrating neuropsychological assessments, quantitative gait analysis with structural neuroimaging data. We aimed to investigate whether gait parameters (particularly stride length variability) were the independent predictors of falls and validate the hypothesis that reduced gray matter volume in specific brain region may be associated with increased risk of falls in AD patients.

Falls represent a major threat to the physical and mental health of older adults. Patients with AD exhibit a significantly higher incidence of falls than age-matched cognitively healthy controls, which is attributable to AD-specific neuropathology (3). In this prospective nested case–control study, we identified incident fallers within a cohort of individuals with aMCI and AD. Each faller was matched 1:2 with non-fallers based on sex, age (±5 years), and MMSE score (±3 points). This design helped mitigate confounding by global cognitive function and key demographics, allowing us to focus on the association of baseline domain-specific cognition, gait parameters, and gray matter volume with incident falls. BMI, stride length variability, and medication burden were found to be independent predictors of incident falls by univariate and multivariate analyses. Furthermore, VBM analysis suggested significantly reduced gray matter volume in fallers compared to non-fallers within the left cerebellar Crus I, lobule VI, and fusiform gyrus; right cerebellar lobule VI, fusiform gyrus, and lobule IV–V; and right superior and middle temporal gyrus. These findings suggest that atrophy in these specific brain regions may contribute to increased fall risk in individuals on the AD spectrum.

Extensive evidence indicates that older adults with cognitive impairment face heightened fall risk. Impairments in specific cognitive domains have been implicated as contributing factors. Casemiro et al. (17) established executive function as a critical cognitive resource for safe ambulation, with its impairment directly elevating fall risk. During adaptive walking, executive abilities enable: path planning, rapid decision-making (e.g., obstacle avoidance), working memory (e.g., environmental mapping), response inhibition (e.g., distraction suppression). Consequently, executive impairment may precipitate falls through dysregulation of these processes. In contrast, Beauchet et al. (18) associated high fall risk with memory decline in dementia-free community dwellers, while another study using the Auditory Verbal Learning Test (AVLT) linked poor memory-particularly short-delay recall-to increased fall incidence in aMCI and mild AD (9). Conversely, Chantanachai et al. (19) found no prospective association between cognitive measures and falls. Given these inconsistent findings, our study matched fallers and non-fallers on MMSE scores (±3 points), thereby controlling for global cognition to better isolate potential domain-specific associations. Univariate analysis revealed a significant association between executive dysfunction (longer TMT B-A time) and faller status. However, this association was attenuated in the multivariate model, suggesting that executive dysfunction may influence fall risk indirectly, potentially mediated through its effects on physiological factors such as gait stability, rather than exerting a direct effect. Future large-scale prospective studies should quantify domain-specific cognitive contributions to fall pathogenesis or examine interaction effects between cognitive deficits and biomechanical risk factors.

Gait impairments belong to the strongest biomechanical correlates of falls. Contemporary analysis techniques enable precise measurement of parameters such as gait velocity-a global functional indicator-and gait variability, an emerging metric for dynamic instability assessment. While univariate analyses implicated both slower stride velocity and higher stride length variability, multivariate logistic regression prompted only higher baseline stride length variability as an independent predictor of incident falls. Consistent with our results, dual-task gait variability was found to predict incident falls in community-dwelling older adults, whereas gait velocity did not (7). Besides, a cross-sectional study found that gait variability and braking force were associated with the risk of falls in AD patients (20). Gonçalves et al. (21) reported that increased stride velocity dual-task cost (DTC) predicted falls in individuals with MCI but not in those with AD dementia. However, our study showed no significant group differences in stride velocity DTC (for either calculation or verbal fluency tasks), possibly due to simpler cognitive tasks. Seavey and Walters (22) demonstrated that gait deterioration increases with dual-task complexity, observing greater impairment during serial-7 subtraction than during category fluency tasks. In our study, which employed serial-3 subtraction and animal naming tasks, the participants’ median MMSE scores (22, 23) may have provided sufficient cognitive reserve to mitigate dual-task effects. This suggests that dual-task paradigms should be calibrated to participants’ cognitive reserve levels, as differential task demands may identify distinct high-risk subgroups. Absence of cognitive-task accuracy data may also affect the stability of the results. A Systematic Review suggested no dual-task tests predicted prospective falling in people with Alzheimer’s or Parkinson’s disease and complex dual tasks seemed to be more predictive of fall risk than simpler dual tasks (23). Furthermore, Makizako et al. (24) identified impaired balance as a critical factor elevating fall risk in MCI patients, potentially exceeding the impact of gait variability. Another study in AD populations confirmed this, showing that functional reach test (FRT), single-leg stance duration and TUG test are significantly related to falls (25). Notably, our study did not address balance metrics. Future investigations should incorporate quantitative balance assessments integrated with multivariate modeling to delineate key determinants of falls within this population.

Increasing evidence supports that lower BMI increases the risk of AD onset (26), but the association between BMI and fall risk in AD patients has not been investigated. Our findings suggested lower BMI maybe an independent predictor of falls of AD patients. Lower BMI has been considered to be associated with physical frailty, a geriatric syndrome characterized by declined physiologic reserve and heightened vulnerability to adverse outcomes-and muscle loss, which are also recognized correlates of falls in AD patients (27, 28). Implement nutritional interventions to optimize the BMI level may be prevent from fall of AD patients.

Patients with cognitive impairment, particularly moderate-to-severe AD, frequently exhibit neuropsychiatric symptoms and sleep disturbances necessitating polypharmacy. While pharmacotherapy offers benefits, polypharmacy increases the risk of adverse events, including falls. Medications particularly associated with fall risk include antidepressants, antipsychotics, and sedative-hypnotics. This study quantified the total medication burden for AD-specific cognitive treatments (ChEIs, NMDA receptor antagonists), antidepressants, antipsychotics, and benzodiazepines. The results suggested total medication count might independently predicted incident falls, aligning with findings from Cox et al. (29). In previous literature, ChEIs may improve gait and balance control by enhancing attention and executive function, potentially reducing fall risk (30). However, case reports also link ChEIs (e.g., donepezil) to syncope (31) and NMDA receptor antagonists (e.g., memantine) to abnormal cardiac rate (32). These findings underscore the necessity to carefully weigh the potential benefits of pharmacotherapy against the risk of serious adverse events (including fall risk), when considering treatment efficacy.

Accumulating neuroimaging evidence confirms brain-physical function linkages, yet the neuropathological mechanisms specifically underlying falls remain poorly elucidated. To our knowledge, this study provides the first evidence linking specific patterns of gray matter volume reduction to falls within the AD spectrum. Whole-brain VBM analysis showed compared to matched non-fallers, gray matter volumes within the left cerebellar Crus I, lobule VI, and fusiform gyrus; the right cerebellar lobule VI, fusiform gyrus, and lobule IV–V; and the right superior and middle temporal gyrus were significantly reduced in fallers. The results of VBM analysis suggested that reduced gray matter volume in specific brain region might be associated with increased risk of falls of AD patients. In agreement with our results, a resting state functional MRI found that Cerebello-cortical functional connectivity might regulate impaired reactive balance control in older adults with mild cognitive impairment, which may be associated with increased fall risk (33). A small sample case–control study collecting 14 elderly patients with falls and 20 controls found elderly patients with falls exhibited lower cerebellar volumes in the posterior cerebellum, lobules V, VI, VIIB, VIIIA, VIIIB, and Crus II, compared to control group, suggesting cerebellar atrophy contributes to the pathophysiology of fall risk in elderly fallers (34). It was reported that cerebellar cortical volume reduced persisting from early to late clinical stages of AD (35). Notably, atrophy in anterior (lobules I-V) and posterior (lobule VI) cerebellar lobes has been observed in the aMCI stage, with later involvement of the hemispheric portions of the posterior lobe (VI) and Crus I in AD (36). Neuropathological studies confirm the presence of AD pathology in the cerebellum, including Aβ deposition (though typically less dense than in the medial temporal cortex) (37) and p-Tau accumulation (38). Functional neuroimaging and clinical studies indicate that the cerebellum participates in diverse cognitive functions, including working memory, learning, attention, and executive control (39, 40). Specifically, Crus I and Crus II appear to modulate executive control and visuospatial memory (41). The findings of VBM were preliminary and observational based on the small sample size. The specific patterns of gray matter volume reduction may contribute to the increased fall risk in patients with AD, which need to be further validated in larger cohorts.

In generally, the preliminary findings of this study indicate that BMI, stride length variability, and medication burden may be associated with fall risk and reduced gray matter volume in specific brain region (cerebellum and the right superior and middle temporal gyrus) may be a potential neuroanatomical basis for increased risk of falls in AD patients. Several limitations should be acknowledged: (1) the small sample size limits the robustness of the predictive model and generalizability; findings require validation in larger, independent cohorts; (2) the exclusive focus on gray matter volume warrants future integration with fMRI and more detailed behavioral phenotyping to elucidate mechanisms; (3) the contributions of key subcortical nuclei involved in gait regulation are not examined and represent a critical target for future research.