BDNF Val66Met predicts cognitive decline in the Wisconsin Registry for Alzheimer's Prevention

Abstract

Objective:

To examine the influence of the brain-derived neurotrophic factor (BDNF) Val66Met polymorphism on longitudinal cognitive trajectories in a large, cognitively healthy cohort enriched for Alzheimer disease (AD) risk and to understand whether β-amyloid (Aβ) burden plays a moderating role in this relationship.

Methods:

One thousand twenty-three adults (baseline age 54.94 ± 6.41 years) enrolled in the Wisconsin Registry for Alzheimer's Prevention underwent BDNF genotyping and cognitive assessment at up to 5 time points (average follow-up 6.92 ± 3.22 years). A subset (n = 140) underwent ¹¹C-Pittsburgh compound B (PiB) scanning. Covariate-adjusted mixed-effects regression models were used to elucidate the effect of BDNF on cognitive trajectories in 4 cognitive domains, including verbal learning and memory, speed and flexibility, working memory, and immediate memory. Secondary mixed-effects regression models were conducted to examine whether Aβ burden, indexed by composite PiB load, modified any observed BDNF-related cognitive trajectories.

Results:

Compared to BDNF Val/Val homozygotes, Met carriers showed steeper decline in verbal learning and memory (p = 0.002) and speed and flexibility (p = 0.017). In addition, Aβ burden moderated the relationship between BDNF and verbal learning and memory such that Met carriers with greater Aβ burden showed even steeper cognitive decline (p = 0.033).

Conclusions:

In a middle-aged cohort with AD risk, carriage of the BDNF Met allele was associated with steeper decline in episodic memory and executive function. This decline was exacerbated by greater Aβ burden. These results suggest that the BDNF Val66Met polymorphism may play an important role in cognitive decline and could be considered as a target for novel AD therapeutics.

Preclinical Alzheimer disease (AD) is thought to be a critical period for intervention therapies that could potentially delay or prevent AD onset.¹ Considerable focus has been placed on genetic and environmental risk factors that may play a role in progression to AD and are possibly targetable for intervention, including APOE ε4,^2,3 physical activity,^4,5 and cognitive reserve.^6,7 Increasing evidence suggests that brain-derived neurotrophic factor (BDNF) may be a genetic risk factor for AD. BDNF is a neurotrophin known to play roles in synaptic plasticity, neurogenesis, neuronal survival, and cognitive health.^8–10 Additional research suggests that BDNF may moderate β-amyloid (Aβ) accumulation, a hallmark feature of AD,¹ by reducing Aβ-mediated cell death,¹¹ decreasing Aβ formation,¹² and repairing Aβ-induced damage.¹³

A polymorphism within the BDNF gene (rs6265) causes a valine (Val) to methionine (Met) substitution at codon 66 (Val66Met). Carriage of 1 or 2 Met alleles is associated with lower BDNF production,¹⁴ decreased hippocampal volume,¹⁵ and cognitive decline.^16–19 However, null^20,21 and opposite findings^22,23 have been documented, making the relationship between BDNF and cognition in aging populations unclear.

Our primary objective was to investigate whether BDNF is associated with longitudinal cognitive trajectories within a large cohort of middle-aged, cognitively healthy individuals enriched for AD risk, a target population for interventional therapies. Our secondary objective was to determine whether Aβ burden moderates the aforementioned relationship. We hypothesized that BDNF Met carriers would exhibit comparatively steeper cognitive decline in all cognitive domains and that Aβ burden would exacerbate this cognitive vulnerability.

METHODS

Standard protocol approvals, registrations, and patient consents.

The University of Wisconsin Institutional Review Board approved all study procedures, and all participants provided signed informed consent before participation.

Participants.

Study participants were enrolled in the Wisconsin Registry for Alzheimer's Prevention (WRAP), a longitudinal study of persons 40 to 65 years of age and cognitively healthy at study entry. Details about WRAP have been previously described.²⁴ For this study, 1,410 participants were selected on the basis of available BDNF and APOE data. One hundred five were subsequently excluded because of self-reported neurologic diagnosis (multiple sclerosis, Parkinson disease, etc). To preclude the influence of sibling clusters, only the first enrolled sibling from a family was included, excluding 277 individuals. Five participants were excluded because of missing covariate data. Thus, 1,023 individuals were included in the study. This sample is enriched for AD risk, with 64.2% of participants having at least one parent with AD as defined by National Institute of Neurological and Communicative Diseases and Stroke/Alzheimer's Disease and Related Disorders Association research criteria and 38.0% being APOE ε4 positive. Table 1 gives participant characteristics.

Table 1.

Participant characteristics split by BDNF Val66Met polymorphism in the full sample and in the PiB subsample

DNA collection, genotyping, and quality assurance.

DNA was extracted from whole-blood samples with the PUREGENE DNA Isolation Kit (Gentra Systems, Inc, Minneapolis, MN). DNA concentrations were quantified with ultraviolet spectrophotometry (DU 530 Spectrophotometer; Beckman Coulter, Fullerton, CA). Single nucleotide polymorphisms (SNPs) for BDNF (rs6265) and APOE (rs429358, rs7412) were genotyped by LGC Genomics (Beverly, MA) using competitive allele-specific PCR-based KASP genotyping assays. For quality assurance, duplicate quality control samples from 102 individuals were placed randomly throughout each of the 96-well plates. Further quality assurance was conducted with PLINK version 1.07.²⁵ BDNF and APOE SNPs did not deviate from Hardy-Weinberg equilibrium with the use of a Bonferroni-adjusted global significance level of p = 0.05, with accordant allele call rates ≥95%.

Cognitive evaluation.

The WRAP neuropsychological test battery²⁴ comprises measures that assess multiple cognitive domains. Prior factor analyses of these measures indicated they map onto 6 cognitive factors; details of this method have been previously described.²⁶ Four of these factor scores were included in the present study because of their representation of cognitive abilities implicated in AD.²⁷ These factors and their constituent tests are as follows: immediate memory—Rey Auditory Verbal Learning Test (RAVLT) Learning Trials 1 and 2; verbal learning and memory—RAVLT Learning Trials 3 through 5 and Delayed Recall; working memory—Digit Span and Letter-Number Sequencing subtests from the Wechsler Adult Intelligence Scale, third edition; and speed and flexibility—Stroop Color-Word Test Interference Trial and Trail-Making Tests A and B. Participants undergo cognitive evaluation at each study visit, with up to 5 visits completed and a maximum of 13.12 years of follow-up at the time of these analyses (table 1).

¹¹C-Pittsburgh compound B-PET neuroimaging protocol.

A subset of participants (n = 140) underwent 3-dimensional ¹¹C-Pittsburgh compound B (PiB)-PET scanning on a Siemens EXACT HR+ scanner (Siemens AG, Erlangen, Germany) through participation in a WRAP-affiliated study. Detailed methods of the acquisition and postprocessing of PiB-PET data have been described previously.⁴ Briefly, imaging entailed a 70-minute dynamic scan on bolus injection and a 6-minute transmission scan. Image postprocessing used an in-house automated pipeline.²⁸ We created distribution volume ratio (DVR) maps of ¹¹C-PiB binding using the time-activity curve in the gray matter of the cerebellum as a reference region.²⁹ Then, using an anatomic atlas,³⁰ we extracted quantitative DVR data from 8 bilateral regions of interest sensitive to Aβ accumulation, including the precuneus, posterior cingulate, orbitofrontal cortex, anterior cingulate, angular gyrus, supramarginal gyrus, middle temporal gyrus, and superior temporal gyrus. DVR data from these regions of interest were combined to form a continuous, composite measure of Aβ accumulation.³¹ The mean time interval between PiB-PET scan and baseline WRAP visit was 6.08 ± 1.68 years (table 1).

Statistical analyses.

We used linear mixed models to investigate differences in the 4 cognitive trajectories as a function of BDNF Val66Met polymorphism. Models were fit by use of the maximum likelihood estimation and were constructed in the following progressive steps (assuming normality in the random effects): random intercept only, random intercept and random slope (for time, measured in years since baseline visit) without correlation between the 2 (i.e., variance components covariance matrix), and random intercept and slope with correlation between the 2 (i.e., unstructured covariance matrix). The last was selected for all subsequent analyses on the basis of superior model fit (Akaike information criterion). After the covariance structure was decided, we then added fixed-effects covariates, which were determined a priori from known associations with cognition or AD. They included age at baseline visit, sex, years of education, APOE ε4 status (0 = no ε4 allele, 1 = 1 or 2 ε4 alleles), parental family history of AD (0 = negative parental history, 1 = positive parental history), time, and BDNF (0 = Val/Val homozygotes, 1 = Met carriers; table 1). Our term of interest was BDNF × time because it would indicate whether rates of decline in cognition across time differ between BDNF Val/Val and Met carriers. Primary analyses were adjusted for multiple comparisons with a false discovery rate correction.³² In addition, raw data visualization of each primary analysis confirmed that linearity was a reasonable representation of the data (figure e-1 at Neurology.org).

Given a significant BDNF × time term, follow-up analyses were conducted to elucidate which cognitive tests contributing to the factor scores drove the observed associations. We fitted the same mixed-effects models as above, except outcomes were the individual cognitive test scores rather than factor scores. Our term of interest remained the BDNF × time interaction.

In addition, we performed secondary mixed-effects models to determine whether Aβ burden, as indexed by the continuous composite PiB-PET measure, modified any BDNF × time effect observed in our primary analyses. For this purpose, we incorporated terms for Aβ, BDNF × Aβ, Aβ × time, and Aβ × BDNF × time into our original models, with Aβ × BDNF × time being the term of interest. All follow-up analyses to our primary analyses were not corrected for multiple comparisons because of their exploratory nature. Unless specified above, SPSS default mixed-model specifications were used for all analyses. Details on mixed-modeling procedures and assumptions are described elsewhere.³³ Only findings with values of p ≤ 0.05 (2-tailed) were considered significant. Analyses were conducted with IBM SPSS, version 21.0.

RESULTS

Background characteristics.

Participant background characteristics are described in table 1. The mean ± SD age at baseline was 54.94 ± 6.41 years; women made up 69.7% of the sample. The group was highly educated (mean ± SD education 16.36 ± 2.82 years) and primarily white (88.7%, self-reported). In our sample, BDNF Met carriers encompassed 32.3% of the sample, which is lower than the reported white population frequency of 37.0% (http://www.alzgene.org). In a comparison of BDNF Val/Val homozygotes and BDNF Met carriers, there were no differences in demographic characteristics except for race/ethnicity (p < 0.001). There were no differences in demographic characteristics in the subset of 140 individuals with PiB-PET data.

Influence of BDNF on cognitive trajectories.

Results of the linear mixed-effects models showed a BDNF × time interaction in the cognitive domains of verbal learning and memory (p = 0.002) and speed and flexibility (p = 0.017) such that BDNF Met carriers declined more steeply over time compared with BDNF Val/Val homozygotes (table 2 and figure 1). Differences in cognitive trajectories between Met carriers and Val/Val homozygotes were not seen in the cognitive domains of working memory (p = 0.590) and immediate memory (p = 0.931; table 2 and figure 1). Because of the difference in race/ethnicity between BDNF Met carriers and Val/Val homozygotes (table 1), we repeated the analyses while additionally including race/ethnicity as a covariate. Results were substantively unchanged.

Table 2.

Trajectories of change in cognition as a function of the BDNF Val66Met polymorphism

Figure 1. BDNF Met carriage is associated with decline in memory and executive function.

(A–D) Estimated trajectories of change in BDNF Val/Val homozygotes (blue) vs BDNF Met carriers (green) in cognitive domains of (A) verbal learning and memory, (B) speed and flexibility, (C) immediate memory, and (D) working memory. All models were adjusted for age at baseline, sex, years of education, parental history of Alzheimer disease, APOE ε4 allele carriage, BDNF, and time in years from baseline visit. Trajectories were plotted by calculating the regression equation lines for BDNF Val/Val homozygotes and Met carriers using the mean values for each of the covariates. Shaded regions represent standard errors. BDNF = brain-derived neurotrophic factor; Met = methionine; Val = valine.

Follow-up analyses were conducted to determine which cognitive tests in verbal learning and memory and speed and flexibility drove the associations noted above. For verbal learning and memory, both measures, RAVLT Learning Trials 3 through 5 and Delayed Recall, showed associations wherein BDNF Met carriers declined more steeply over time compared with BDNF Val/Val homozygotes (p = 0.005 and 0.007, respectively; table 2). For speed and flexibility, none of the 3 constituent tests showed associations with BDNF.

Influence of Aβ and BDNF on cognitive trajectories.

Given the BDNF × time interactions observed in verbal learning and memory and speed and flexibility in our primary analyses, we performed secondary analyses investigating whether the relationships observed were modified by Aβ burden. Results showed that there was an Aβ × BDNF × time interaction for verbal learning and memory (β [SE] = −0.218 [0.101], t = −2.164, p = 0.033) but not for speed and flexibility (β [SE] = −0.007 [0.086], t = −0.078, p = 0.938). When race/ethnicity also was included as a covariate, results remained substantively unchanged. A frequency distribution of DVRs is depicted in figure 2, with previously determined Aβ positivity cut points indicated.^34,35 Figure 3 depicts the change trajectories in verbal learning and memory, while accounting for covariates, across the following 4 prototypical groups: Val/Val Aβ+, Met carrier Aβ+, Val/Val Aβ−, and Met carrier Aβ−. Although Aβ burden, indexed with the PiB-PET composite DVR, was included in the analysis as a continuous variable, for graphing purposes, we chose 2 anchor points, the minimum (1.002) and maximum (2.073) DVR values, to represent Aβ+ vs Aβ−. As depicted in figure 3, Aβ+ Met carriers showed steeper cognitive decline over time compared with Aβ+ Val/Val homozygotes, who exhibited normal cognitive performance over time. These results indicate that the adverse influence of Met carriage on cognitive trajectory is further exacerbated by Aβ burden, whereas Val/Val genotype protects cognitive function even in the context of coexisting Aβ burden.

Figure 2. Frequency distribution of PiB-PET DVR values.

Histogram depicts the DVR distribution for the subset of participants with PiB-PET data (n = 140). Three cut points for β-amyloid positivity are indicated: (1) a low-threshold DVR cut point of 1.08 (orange),³⁵ (2) a high-threshold DVR cut point of 1.20 (red),³⁵ and (3) an in-house cut point of 1.19 (blue).³⁴ DVR = distribution volume ratio; PiB = ¹¹C-Pittsburgh compound B.

Figure 3. Aβ modifies BDNF-related trajectories of change in memory.

Shown are estimated trajectories of change in the verbal learning and memory cognitive factor when both the BDNF Val66Met polymorphism and Aβ positivity as measured by PiB-PET imaging are examined. This analysis was adjusted for age at baseline, sex, years of education, parental history of Alzheimer disease, APOE ε4 allele carriage, BDNF, Aβ, and time in years from baseline visit. Trajectories were plotted by calculating the regression equation lines for the 4 groups (Val/Val Aβ+, Met carrier Aβ+, Val/Val Aβ−, and Met carrier Aβ−) using the mean values for each of the covariates. Although Aβ burden, as indexed with PiB-PET composite DVR, was included in the analysis as a continuous variable, for ease of display, we chose 2 anchor points, the minimum (1.002) and maximum (2.073) DVR values, to represent Aβ+ vs Aβ−. Shaded regions represent standard errors. Aβ = β-amyloid; BDNF = brain-derived neurotrophic factor; DVR = distribution volume ratio; Met = methionine; PiB = ¹¹C-Pittsburgh compound B; Val = valine.

Of note, refitting our original BDNF × time analyses within this PiB subsample did not yield significant results, although effects were in the same direction as in the full sample (verbal learning and memory: β [SE] = −0.023 [0.017], t = −1.330, p = 0.186; speed and flexibility: β [SE] = −0.006 [0.015], t = −0.390, p = 0.697).

Exploratory analyses.

An “additive” genetic model (i.e., in which Val/Val = 0, Val/Met = 1, and Met/Met = 2, as opposed to our original “dominant” genetic model, in which Val/Val = 0, Met carriers = 1) was also considered for the cognitive analyses. Results from this additive model were similar to results of the original model, with Met/Met homozygotes exhibiting the steepest cognitive decline. However, because of the limited number of Met/Met homozygotes (n = 34), these results may have poor stability.

In addition, we examined whether APOE ε4 status moderated the relationship between BDNF and cognitive trajectories in verbal learning and memory and speed and flexibility. The findings were nonsignificant (p = 0.359 and 0.686, respectively).

DISCUSSION

This study found that the BDNF Val66Met polymorphism is associated with cognitive decline in a large cohort of individuals with increased risk for AD. Specifically, compared with Val/Val homozygotes, Met carriers exhibited steeper decline in the cognitive domains of verbal learning and memory and speed and flexibility. In addition, we showed that Aβ accumulation adversely moderated the relationship between BDNF and verbal learning and memory such that Met carriers with greater Aβ burden had steeper memory decline compared to those with lesser Aβ burden. This is one of few longitudinal studies to report the association between the BDNF Val66Met polymorphism and cognitive decline over time. Our study also expands on previous findings regarding the moderating relationship between Aβ accumulation and Met carriership on cognitive decline.¹⁶ Our research was conducted in a cognitively healthy cohort with risk factors for AD, highlighting the potential for early detection of cognitive decline and subsequent implementation of interventional therapies³⁶ during this preclinical AD phase.

Our study adds longitudinal evidence to a growing body of literature on the relationship between the BDNF Val66Met polymorphism and cognitive health. Initial cross-sectional studies on this polymorphism found that in younger adults, Met carriership was associated with worse episodic memory as measured by the Wechsler Memory Scale and an fMRI declarative memory paradigm.^14,19 More recently, a study in an aging population¹⁸ (mean age 56 years) demonstrated cross-sectionally that Met carriers performed worse in both item memory and prospective memory with advancing age compared with Val/Val homozygotes. Item memory was assessed with the California Verbal Learning Test,¹⁸ a psychometric measure similar to the RAVLT, which comprises our verbal learning and memory measure. Another cross-sectional study¹⁷ showed that in addition to memory, processing speed was reduced in Met carriers compared to Val/Val homozygotes (mean age 63 years). Our study adds evidence to these findings by demonstrating that in a longitudinal cohort at risk for AD, Met carriership was associated with steeper decline in the domains of verbal learning and memory and speed and flexibility over an average time period of 7 years.

However, other work examining the relationship between the BDNF Val66Met polymorphism and cognitive health has failed to find a relationship or has even observed findings that are opposite to the aforementioned. In one study, Val/Val genotype occurred with higher frequency in those with AD compared to healthy controls²²; another study showed lower scores on the Frontal Assessment Battery in Val/Val homozygote patients with mild AD.²³ These discrepancies could possibly be explained by genetic differences between samples (i.e., lack of Hardy-Weinberg equilibrium), differences in disease stage at the time of analysis (probable AD vs preclinical AD), or cross-sectional design. In fact, in the present study, Met carriers performed better at baseline in the domains of verbal learning and memory (p = 0.013) and speed and flexibility (p = 0.002) (table 2) yet declined more steeply over time, emphasizing the importance of longitudinal design in studies of BDNF and cognition.

Other studies have failed to find a relationship between the Val66Met polymorphism and cognitive outcomes or AD incidence.^15,20,21 However, in one of these studies, the authors reported relationships between greater BDNF expression in the dorsolateral prefrontal cortex and slower cognitive decline.²¹ Another group did not find associations with the polymorphism but noted that greater serum BDNF levels resulted in reduced dementia and AD incidence.²⁰ Finally, a third report did not find associations with the Val66Met polymorphism but did with other BDNF SNPs, specifically rs1157659, rs11030094, and rs11030108.¹⁵ These discrepancies underscore the need for further understanding of the intricacies in the relationship between BDNF polymorphisms, BDNF expression, and cognitive function.

We found that greater Aβ accumulation, as measured by PiB load in 8 bilateral regions of interest, moderated the relationship between BDNF and verbal learning and memory such that in Met carriers, those with greater Aβ accumulation had steeper cognitive decline. This finding aligns with a similar report¹⁶ that found that in cognitively healthy older adults (mean age 71 years), Met carriers with high levels of Aβ exhibited steeper cognitive decline in episodic memory, executive function, and language over 36 months. The same group also observed that among persons with amnestic mild cognitive impairment, Met carriers with high levels of Aβ had worse episodic memory.³⁷ Other groups have noted similar moderating effects of Aβ and BDNF on cognition in older adults.³⁸

This interaction between Aβ and the BDNF Val66Met polymorphism adds further support for the potential role the BDNF protein has in moderating Aβ production, thus protecting cognitive function. Animal studies have indicated that BDNF regulates sorting protein-related receptor with A-type repeats, a known modulator of Aβ precursor protein trafficking and processing, via extracellular signal-regulated protein kinase stimulation, suggesting that BDNF could deter amyloid production.¹² Other studies have found that BDNF delivery rescues memory impairment in mice injected with Aβ_1-42,³⁹ prevents neuronal loss,¹¹ and rescues cells from degeneration due to Aβ toxicity.¹³ It is critical for future studies to further investigate the role that the BDNF gene and BDNF protein may have in Aβ accumulation because it could be a potential target for intervention against Aβ toxicity.

A major strength of this study is its scale. With 1,023 individuals and up to 13 years of follow-up, it is the one of the largest studies investigating the BDNF Val66Met polymorphism. In addition, conducting the study within the WRAP cohort provides further evidence that subtle cognitive changes can be detected early in the AD cascade, and intervention during this time period could be critical for delaying or preventing onset of AD. However, this study is not without limitations. Our sample consists of predominantly highly educated, white individuals, reducing the generalizability of our results. Generalizability might also be reduced by the discord between Met carriers in our sample and the white population frequency. Furthermore, the sample size of individuals with PiB-PET data was limited. In addition, we were able to conduct analyses using only one BDNF polymorphism, and no serum BDNF data were available for this research. Ongoing work in our group will further examine the interplay between BDNF polymorphisms, systemic and central BDNF levels, and neuroimaging and cognitive outcomes. We find this research to be of particular importance, especially given the body of research suggesting that BDNF levels can be increased with physical activity⁵ and other modifiable lifestyle factors.⁴⁰

This study provides evidence that the BDNF Val66Met polymorphism is associated with cognitive decline in a large sample of individuals at risk for AD and that Aβ plays a moderating role in this relationship. Our findings suggest that Met carriership could accelerate cognitive decline throughout the preclinical phase of AD, emphasizing the importance of investigating BDNF as a potential target for novel AD therapeutics in the future.³⁶

GLOSSARY

Aβ

β-amyloid

AD

Alzheimer disease

BDNF

brain-derived neurotrophic factor

DVR

distribution volume ratio

Met

methionine

PiB

¹¹C-Pittsburgh compound B

RAVLT

Rey Auditory Verbal Learning Test

SNP

single nucleotide polymorphism

Val

valine

WRAP

Wisconsin Registry for Alzheimer's Prevention

AUTHOR CONTRIBUTIONS

Study concept or design: Boots, Okonkwo. Acquisition, analysis, or interpretation of data: Boots, Schultz, Clark, Racine, Darst, Koscik, Carlsson, Gallagher, Hogan, Bendlin, Asthana, Sager, Hermann, Christian, Dubal, Engelman, Johnson, Okonkwo. Drafting of manuscript: Boots, Okonkwo. Critical revision of manuscript: Boots, Schultz, Clark, Racine, Darst, Koscik, Carlsson, Gallagher, Hogan, Bendlin, Asthana, Sager, Hermann, Christian, Dubal, Engelman, Johnson, Okonkwo. Statistical analysis: Boots, Okonkwo. Obtaining funding: Carlsson, Asthana, Dubal, Johnson, Okonkwo. Administrative, technical, or material support: Boots, Schultz, Clark, Racine, Darst, Koscik, Gallagher, Hogan, Bendlin, Sager, Hermann, Christian, Engelman. Supervision: Okonkwo.

STUDY FUNDING

This study was supported by funds from the NIH, Veterans Administration, Alzheimer's Association, Wisconsin Alumni Research Foundation, Helen Bader Foundation, Northwestern Mutual Foundation, Extendicare Foundation, and Wisconsin Alzheimer's Institute Lou Holland Research Fund. The funders' role was limited to providing funding for the study. This work was supported by the National Institute on Aging grants K23 AG045957 (O.C.O.), R21 AG051858 (O.C.O), R01 AG021155 (S.C.J.), R01 AG027161 (S.C.J.), P50 AG033514 (S.A.), and T32 AG000213 (S.A.) and a Clinical and Translational Science Award (UL1RR025011) to the University of Wisconsin-Madison. Portions of this research were supported by the Alzheimer's Association, Wisconsin Alumni Research Foundation, Helen Badger Foundation, Northwestern Mutual Foundation, Extendicare Foundation, Wisconsin Alzheimer's Institute Lou Holland Research Fund, and Veterans Administration, including facilities and resources at the Geriatric Research Education and Clinical Center of the William S. Middleton Memorial Veterans Hospital, Madison, WI.

DISCLOSURE

The authors report no disclosures relevant to the manuscript. Go to Neurology.org for full disclosures.