Association Between Brain β-Amyloid and Frailty in Older Adults

Abstract

Background

We sought to determine whether cortical and regional β-amyloid (Aβ) were cross-sectionally and prospectively associated with change in frailty status in older adults.

Methods

We used data from 269 community-dwelling participants from the Multidomain Alzheimer’s Preventive Trial (MAPT) who were assessed for brain Aβ using positron-emission tomography scan. Regional and cortical-to-cerebellar standardized uptake value ratios were obtained. Frailty was assessed by a frailty index composed of 19 items not directly linked to cognition and Alzheimer’s disease.

Results

A significant and positive cross-sectional and prospective relationship was found for Aβ in the anterior putamen (cross-sectional: β = 0.11 [0.02–0.20], p = .02; prospective: β = 0.11 [0.03–0.19], p = .007), posterior putamen (cross-sectional: β = 0.12 [0.009–0.23], p = .03; prospective: β = 0.11 [0.02–0.21], p = .02), and precuneus regions (cross-sectional: β = 0.07 [0.01–0.12], p = .01; prospective: β = 0.07 [0.01–0.12], p = .01) with increasing frailty.

Conclusions

This study has found new information regarding cross-sectional and prospective positive associations between region-specific brain Aβ deposits and worsening frailty. The potential mechanisms involved require further investigation.

Frailty is defined as a multidimensional condition that reduces physiological reserves, which makes older adults vulnerable and at an increased risk of adverse health events (1), such as disability, hospitalizations, and death (2,3). Frailty leads to increased health care costs (4), representing a major public health concern (5).

Frailty was found to be associated with cognitive decline (6,7) and even incident Alzheimer’s disease (AD) and dementia in older adults (8,9). Indeed, Song and colleagues showed that frailty, as measured by the accumulation of 19 deficits not related to AD, predicted the risk of developing future AD and dementia (8). However, very few studies have investigated the associations of frailty with biomarkers (apolipoprotein E ε4 [ApoE ε4] (10) and insulin resistance (11)) of AD and dementia, and as far as we know, only one cross-sectional study has found significant associations between components of frailty and β-amyloid (Aβ) load in the human brain of older adult populations (12). Because Aβ load is known to be associated with cognitive decline and constitutes a hallmark of the pathophysiology of AD, it is possible that the associations between frailty and cognitive decline would be partly dependent on Aβ deposition. Moreover, brain Aβ accumulation has been found to be associated with Aβ deposition in muscle fibers (13), which could promote muscle weakness and motor impairment, further worsening frailty. Indeed, Aβ load starts to accumulate in the brain decades before the clinical signs of cognitive decline (14); thus, it is possible that Aβ would contribute to increase frailty severity.

The objectives of this study were to examine the cross-sectional and prospective associations between cortical and regional brain Aβ with frailty in community-dwelling older adults with spontaneous memory complaints. The second objective of this study was to assess the association of cerebral region-specific Aβ with the incidence of frailty. We hypothesized that Aβ would be associated with higher severity and incidence of frailty severity in this population.

Methods

We used data from the Multidomain Alzheimer Preventive Trial (MAPT; registration: NCT00672685), a large multicenter, phase III randomized placebo-controlled trial. The trial had a four-arm design that comprised a placebo group, omega-3 fatty acid (n-3 polyunsaturated fatty acid) supplementation, multidomain intervention (physical activity and nutritional counseling and cognitive training) and the multidomain intervention combined with n-3 polyunsaturated fatty acid supplementation. The participants (N = 1,680) signed an informed consent, and the study protocol was approved by the French Ethics Committee located in Toulouse (CPP SOOM II). The study protocol has been published elsewhere (15). The main findings showed that MAPT interventions had no significant effects on cognitive function when multiplicity was taken into account (16).

Participants

The participants were community-dwelling men and women, aged 70 years and older, without dementia and meeting at least one of the following criteria: slow gait speed (<0.8 m/s), limitation in executing one or more Instrumental Activity of Daily Living and spontaneous memory complaints. Of the 1,680 participants, a total of 271 had data on cortical Aβ load. Two participants were excluded because they had dementia (Clinical Dementia Rating ≥ 1) at the clinical visit closest to amyloid assessment. Individuals who had a Clinical Dementia Rating of less than 1 and with an assessment of brain amyloid load (n = 269) were the participants of this study.

Main Outcome Measure

Frailty was evaluated at baseline, 6-month, and 1-, 2-, and 3-year waves of data collection using a frailty index (FI) composed of 19 items (see Supplementary Table S1 for details on the elaboration of the FI) built following standard procedures (17). The FI approach is based on the deficits accumulation hypothesis, that is, the more deficits, the more frail. The FI has been largely used in the literature and has proven its validity in predicting adverse health events (2,4,18), including higher use of health care services and mortality. Therefore, the FI for this study was calculated as the sum of 19 deficits divided by the total number of deficits, giving a continuous variable ranging from 0 (no deficits) to 1 (all deficits present). For example, an individual who has five deficits would have a FI of 0.26 (ie, 5/19). To avoid the influence of cognition, our FI was based on all of the subjective and objective measures that were not directly related to cognitive function (ie, data from neuropsychological tests were not included). The cutoff for defining frailty was greater than or equal to 0.25, as previously suggested (8,19–21).

Florbetapir F-18 Positron-Emission Tomography

Positron-emission tomography (PET) scans as a measure of cerebral Aβ load were performed using [18F]-florbetapir as previously described (15,22). PET data acquisitions commenced 50 minutes after injection of a mean of 4 MBq/kg weight of [18F]-florbetapir. Radiochemical purity of [18F]-florbetapir was superior to 99.5%. Regional standard uptake value ratios were generated from semiautomated quantitative analysis with the whole cerebellum used as the reference region. The PET sinograms were reconstructed with an iterative algorithm that corrected for randomness, scatter, photon attenuation, and decay, which produced images with an isotropic voxel of 2 × 2 × 2 mm³ and a spatial resolution of approximately 5-mm full width at half maximum at the field of view center. The brain regions of interest used were defined using the Avid and Montreal Neurological Institute templates: temporal cortex, parietal cortex, medial orbitofrontal cortex, occipital cortex, precuneus, anterior and posterior cingulate, anterior and posterior putamen, caudate, semioval center, and pons. The pons region was included as a negative control. Six predefined cortical regions of interest (frontal, temporal, parietal, precuneus, anterior cingulate, and posterior cingulate) were used to calculate the mean cortical-to-cerebellar Aβ load. A quality control procedure was carried out using a semiquantification-based method. PET scans were performed throughout the 3-year period of MAPT: the mean (SD) was 544.0 ± 267.5 days after study baseline.

Timing of Assessments

Because the PET assessment was measured at various time points throughout MAPT (544 ± 267.5 days after the study baseline), we considered the starting point of follow-up for each participant at the clinical evaluation closest to the PET scan evaluation. The average number of days between the PET assessment and the clinical evaluation closest to the PET was 76.3 (SD: 40.6) days. The average length of follow-up after PET scan assessment was 558 ± 245 days. For the cross-sectional associations between Aβ and FI, all clinical, behavioral, and performance-based data, including all items used to calculate the FI used in this study, were obtained from the clinical visit closest to the PET scan assessment. Changes in the FI over time were calculated using the closest to PET scan assessment, and all subsequent FIs (until the end of the study) were used.

Confounders

We selected confounding variables on the basis of data availability and the literature on AD: age, gender, educational level, cognitive status measured closest to PET scan (Mini-Mental State Examination), MAPT group allocation, and (ApoE ε4) genotype (carriers of at least one ε4 allele vs. noncarriers).

Statistical Analysis

Descriptive statistics were presented as means (SD) for continuous variables and absolute numbers and percentages for categorical variables. Multiple linear regressions adjusted for all the confounders were performed to examine the cross-sectional associations of the mean cortical-to-cerebellar standardized uptake value ratio and each specific brain region with the FI. For the over-time analysis, mixed-effect regressions (with a random effect at the participant’s level) adjusted for confounders were performed to test the associations of cortical and regional Aβ load with change in FI. For the latter analyses, the FI was calculated from data obtained at the clinical visit closest to PET, which was set as the baseline FI, and all subsequent FI assessments were entered into the model. Finally, for the incidence of frailty, discrete-time Cox proportional hazard models were run in participants having at least two assessments of frailty and who were not rated as frail at baseline. For all Cox models, the association of Aβ load with the risk of having the event (FI ≥ 0.25) was adjusted by age and MAPT groups; proportional hazards were systematically checked (estat phtest command; p > .05 was considered as nonviolation of the proportional hazards assumption). Participants were censored at follow-up if they did not have the event. Time-to-event was calculated as a discrete variable according to MAPT waves of data collection (ie, 6 months and 1, 2, or 3 years).

Because MAPT interventions had no significant effects on the FI used in this study (see Supplementary Table S2), we decided not to restrict our analyses to the placebo group and use the whole population with available data (n = 269); MAPT group allocation was entered as a confounding factor. We performed a sensitivity analysis on the prospective associations of cortical and regional Aβ with frailty severity only in the placebo group with the same confounders as the primary analyses. All statistical analyses were performed with Stata v14 (College Station, TX). Significance was set by a p value of less than .05. There was no correction for multiple testing due to the exploratory nature of this study.

Results

Baseline characteristics are given in Table 1. Participants had a mean FI score of 0.17 ± 0.10, which meant an average of around 3.0 ± 1.9 deficits out of 19.

In Table 2, multiple linear regressions showed a nonsignificant cross-sectional association between cortical-to-cerebellar standardized uptake value ratio and FI (p = .14). When focusing on regional standardized uptake value ratio, significant and positive relationships with the FI were found for both the anterior putamen (adjusted model: β = 0.11 [0.02–0.20], p = .02) and the posterior putamen (β = 0.12 [0.009–0.23], p = .03), meaning that more Aβ deposition was cross-sectionally associated with a higher FI. All the other cross-sectional associations were statistically nonsignificant.

Table 3 displays the over time associations of cortical and regional Aβ accumulation with change in FI (p = .12). Total cortical Aβ load was not significantly associated with change in FI. Similar to the cross-sectional analysis, the anterior putamen (adjusted model: β = 0.11 [0.03–0.19]; p = .007) and the posterior putamen (β = 0.11 [0.02–0.21]; p = .02) were positively associated with increasing FI over time. Moreover, the precuneus region (adjusted model: β = 0.07 [0.01–0.12]; p = .01) was also found to be positively and significantly associated with FI (higher Aβ load associated with higher FI). All the other prospective associations were statistically nonsignificant. Prospective associations among the placebo group showed that cortical Aβ load and 7 of the 13 regions were found to be associated with higher FI (Supplementary Table S3).

Regarding the incident events, frailty occurred in 40/201 (16.6%). Discrete-time Cox model adjusted for age and MAPT groups found no significant associations of all of cortical and region-specific Aβ load with the risk of frailty onset (Table 4).

Discussion

Here, we have shown that mean cortical Aβ load was not associated with a FI elaborated from noncognitive measures; however, Aβ in both the anterior and posterior putamen were associated with FI severity using both cross-sectional and prospective approaches in community-dwelling older adults. Moreover, Aβ accumulation in the precuneus region was associated with an increasing frailty severity over time. Neither cortical nor regional Aβ was associated with incident frailty. Our findings show for the first time that cerebral Aβ accumulation may determine frailty trajectory over time in older participants.

To the best of our knowledge, this is the first prospective study to show associations between Aβ load in specific brain regions and frailty. By showing that specific brain regions predicted frailty evolution, our study extends the cross-sectional findings of Yoon and colleagues (12), who found that elements composing the Fried frailty phenotype (eg, handgrip strength and gait speed) were associated with amyloid load. Moreover, our study goes beyond the study by Yoon and colleagues because we have a larger sample size composed of both mild cognitive impairment individuals and people without objective cognitive decline, whereas the study by Yoon and colleagues was restricted to people with mild cognitive impairment. Because our FI is a multidimensional measurement, it is difficult to explain precisely why significant associations were found only with the anterior putamen, the posterior putamen, and the precuneus, but not with other brain regions. A possible explanation is that the anterior putamen, posterior putamen, and precuneus are known to play a role in coordination and motor function. The precuneus is connected to and plays a multimodal function role with the visual, motor, and sensorimotor cortices (23). The anterior putamen is partially responsible for postural sway in walking (24). For the posterior putamen, it has been shown to be interlinked with motor function and the modulation of motor circuits (25). Therefore, it is possible that excessive Aβ deposition, and thus Aβ toxicity, in these brain regions might disrupt motor corticostriatal circuits, decreasing the ability of the individual to plan and execute motor-related tasks and activities; this, in turn, would be associated with increasing frailty. It is possible that the putamen is particularly vulnerable region in the brain for amyloid deposition and toxicity; previous work using MAPT data have also found an association between amyloid deposition in this brain area and motor function (22). Further studies are needed to confirm this hypothesis. It is important to highlight that our FI was mainly composed of motor-related items (objectively measured, such as gait speed, chair rise, and balance, and subjectively measured, such as the ability to perform daily tasks). Although such a hypothesis is interesting, we have previously found contradicting results regarding the associations of cerebral Aβ load with motor function using data from the MAPT study: Whereas one study showed positive associations of several brain regions (notably the posterior putamen and the precuneus) with impaired gait speed (22), another study found no associations of cortical and regional Aβ deposits with chair rise performance (26). Moreover, it is also possible that cerebral Aβ deposition would constitute a sign of increased Aβ accumulation in muscle, which may lead to muscle weakness (a hallmark of frailty), as suggested by an animal study (27); of note, Aβ load in both the brain and the skeletal muscle was found to be higher in AD patients than in people without dementia according to an autopsy study (13), suggesting that Aβ in the brain and muscle are related (28). In skeletal muscle, there is a high potential of Aβ proliferation in the neuromuscular junction, especially in people with myositis (29). Thus, the presence of amyloid in muscle could potentially induce fatigue or pain in older adults, thus reducing function. Another possibility is that Aβ toxicity would impair the functioning of the limbic system networks (because AD patients seem to have impaired limbic system networks) (30) that are involved in emotional responses and controlling appetite. However, this requires further investigations, despite the fact that our FI is composed of items related with weight loss and includes some subjective questions about participants’ feelings. Finally, another potential explanation of the accumulation deficits over time is the “brain muscle loop” hypothesis (26,27); because cognitive dysfunction may influence muscle mass and physical function, it is plausible to think that amyloid accumulation, which is linked to cognitive dysfunction, would influence frailty.

The nonsignificant associations between Aβ load and the incidence of frailty (FI ≥ 0.25) may be potentially explained by the slow progression of Aβ accumulation in the brain (31) and/or our small sample size. Thus, to provide evidence of an association between brain Aβ and the onset of the frailty condition, a larger observational study with a longer follow-up length would be needed. Such a study would shed light on the possible clinical implications of high Aβ load in the field of frailty.

Our results add important information about the associations between frailty and AD. It is possible that these conditions, that is, frailty and AD, are associated over time in a kind of vicious cycle, with worsening in the physiopathology of AD (eg, Aβ accumulation) being associated with increasing frailty severity, which, in turn, would be associated with an increased risk of developing AD; the predictive value of deficits accumulation to determine the onset of AD has already been evidenced (8). The associations between specific brain regions and frailty require further investigation. Indeed, the results of our sensitivity analysis restricted to the MAPT placebo group showed that Aβ load in most brain regions, including mean cortical load, were associated with increasing frailty, suggesting that the Aβ–frailty relationship might be a matter of the whole brain and not necessarily of specific brain regions. This hypothesis may stem on the foundation that frailty, as measured by the accumulation of deficits, is a multidimensional condition that may, in theory, affect and be affected by multiple factors and pathways.

This study has several strengths: First, this study is in the first to measure the association between cerebral Aβ and frailty in older adults. The sample size is relatively large for brain Aβ load studies, and we were able to investigate baseline and prospective associations with frailty over time. Nonetheless, several limitations must be mentioned. First, PET imaging was assessed at different time points across the 3-year study, imaging was not performed at enrollment, and there was an average gap of 76 days between PET scan and clinical assessments. This limits the interpretation of our cross-sectional analyses and the power of our prospective analyses (reduced the follow-up length). Second, these were secondary analyses using data from a randomized controlled trial that tested the effects of three interventions on cognitive function. However, because MAPT interventions had no significant effects on our FI, the importance of this potential bias was probably decreased. Third, caution must be taken in interpreting the results because the magnitude of associations between regional amyloid load and frailty evolution over time was relatively weak. Also, we used a 19-item FI instead of Fried’s frailty phenotype, in which is difficult to interpret in a clinical manner. Last, when analyzing only the placebo group, the results differ from our main analyses with most of the regional amyloid load with frailty evolution over time. Although we have not found any effect of the MAPT interventions on the main outcome (Supplementary Table S2), we cannot exclude that MAPT interventions had some effect on amyloid deposition.

To conclude, this study provides new information regarding the associations between frailty and AD pathology by showing that Aβ deposits in the anterior putamen, posterior putamen, and precuneus were associated with change in frailty severity in older adults. A larger observational longitudinal study with a longer follow-up including at least two time-point measurements of both brain Aβ load and frailty is needed to examine the temporal associations between these outcomes. Also, studies examining the neuronal pathways between regional Aβ load, motor function, and limbic system networks would shed light on the potential mechanisms linking AD pathology and frailty.

Funding

The MAPT study was supported by grants from the French Ministry of Health (PHRC 2008, 2009), the Gérontopôle of Toulouse, the Pierre Fabre Research Institute (manufacturer of the omega-3 supplement), Exhonit Therapeutics SA, Avid Radiopharmaceuticals and in part by a grant from the French National Agency for Research called “Investissements d’Avenir” no. ANR-11-LABX-0018-01. Study promotion was supported by the University Hospital Center of Toulouse. The data sharing activity was supported by the Association Monegasque pour la Recherche sur la maladie d’Alzheimer (AMPA) and the UMR 1027 Unit INSERM-University of Toulouse III.

Conflict of Interest

None reported.

Acknowledgments

M.M. performed the analyses and drafted the manuscript. P.S.B. interpreted the data and critically revised the draft for important intellectual content. C.H. and Y.R. interpreted the data and critically revised the draft for important intellectual content. P.P. designed the imaging protocol, collected data, and critically revised the draft for important intellectual content. B.V. conceived the MAPT study, interpreted the data, and critically revised the draft for important intellectual content. All authors agreed with the final version to be submitted. The promotion of this study was supported by the University Hospital Center of Toulouse.

References

Author notes