Cerebral microbleed incidence, relationship to amyloid burden

Abstract

Objective

To determine the incidence of cerebral microbleeds (CMBs) and the association of amyloid PET burden with incident CMBs.

Methods

A total of 651 participants, age ≥50 years (55% male), underwent 3T MRI scans with ≥2 separate T2*-weighted gradient recalled echo sequences from October 2011 to August 2017. Eighty-seven percent underwent ¹¹C Pittsburgh compound B (PiB) PET scans. Age-specific CMB incidence rates were calculated by using the piecewise exponential model. Using structural equation models (SEMs), we assessed the effect of amyloid load and baseline CMBs on future CMBs after considering the direct and indirect age, sex, vascular risk factors, and APOE effects.

Results

Participants' mean age (SD) was 69.8 (10.0) years at baseline MRI, and 111 participants (17%) had ≥1 baseline CMB. The mean (SD) of the time interval between scans was 2.7 (1.0) years. The overall population incidence rate for CMBs was 3.6/100 person-years and increased with age: from 1.5/100 new CMBs at age 50 to 11.6/100 person-years at age 90. Using the piecewise exponential model regression, the incidence rates increased with age and the presence of baseline CMBs. The SEMs showed that (1) increasing age at MRI or carrying an APOE4 allele was associated with more amyloid at baseline, and higher amyloid, particularly occipital amyloid load, in turn increased the risk of a new lobar CMB; and (2) the presence of CMBs at baseline increased the risk of a lobar CMB and had a larger effect size than amyloid load.

Conclusions

Age and APOE4 carrier status act through amyloid load to increase the risk of subsequent lobar CMBs, but the presence of baseline CMBs is the most important risk factor for future CMBs.

While cerebral microbleeds (CMBs) are a common cerebrovascular pathology of aging, few population-based studies have investigated the incidence of CMBs in the general population. Understanding predictors for developing new CMBs has become increasingly important because increased CMB burden is a risk factor for both hemorrhagic and ischemic stroke. In addition, clinical trials for Alzheimer disease use CMBs as an important exclusionary and outcome variable because CMBs can develop as a consequence of amyloid immunotherapy. Prior population-based studies have not investigated whether amyloid PET burden at baseline is associated with developing incident CMBs.^1,2

The goal of this study was to determine the population incidence of CMBs and determine the role of baseline amyloid burden on PET on risk of future CMBs. Since lobar CMBs correlate with underlying cerebral amyloid angiopathy (CAA) pathology and amyloid burden but deep CMBs do not,^3,4 we hypothesized that amyloid burden predicts subsequent development of lobar CMBs but not deep CMBs.

Methods

Study participants

Participants were enrolled in the longitudinal, population-based Mayo Clinic Study of Aging (MCSA), which randomly enumerated Olmsted County residents in 2010 for recruitment using the Rochester Epidemiology Project medical records linkage system stratified by age and sex and which had 3,845 participants enrolled between August 1, 2011, and September 30, 2017. The details of the MCSA study design have been published previously.^5,6 Participants are invited to undergo neuroimaging at baseline and every 15 months. In October 2011, T2* gradient-recalled echo (GRE) sequences were added to the MRI protocol. Participants imaged with at least 2 T2* GRE sequences between October 2011 and December 2017 were included in this study (n = 651).

Clinical data retrieval

Clinical data, including vascular risk factors, were abstracted by a nurse from the detailed medical records included in the medical records linkage system.⁶

MRI examination and identification of CMBs

Three 3T General Electric (Milwaukee, WI) scanners acquired images in this study. CMBs were evaluated on a T2* GRE (repetition time/echo time, 200/20 ms; flip angle, 12°; in-plane matrix, 256 × 224; phase field of view, 1.00; slice thickness, 3.3 mm; acquisition time, 5 minutes). CMBs were defined according to consensus criteria^4,7 as homogeneous hypointense lesions in gray or white matter, which are distinct from iron or calcium deposits and vessel-flow voids on T2* GRE images.⁸ The full details of CMB detection have been published previously.⁸ The location of each CMB was recorded in the coordinate system of the image on which it was made. A structural T1 magnetization-prepared rapid gradient echo image of the participant was registered and resampled into the GRE image. An in-house, modified automated anatomic-labeling atlas delineated lobar regions and deep/infratentorial gray and white matter regions.⁹ CMBs were classified as lobar with or without cerebellar CMBs, deep/infratentorial CMBs with or without cerebellar CMBs, or cerebellar-only CMBs. Mixed CMBs were those with at least 1 lobar and 1 deep CMB. For participants who underwent multiple MRI scans, we took the baseline scan to evaluate existing CMBs and then the first scan when a new CMB occurred; otherwise, the last T2* GRE scan was used for the participant.

¹¹C Pittsburgh compound B (PiB) PET acquisition

The acquisition and processing details of amyloid PET have been published previously.¹⁰ Amyloid PET imaging was performed with Pittsburgh compound B. CT was obtained for attenuation correction. PET images were analyzed with our in-house, fully automated image-processing pipeline, where image voxel values are extracted from automatically labeled regions of interest (ROIs) propagated from an MRI template. A PiB standardized uptake value ratio (SUVR) for each participant was calculated.¹⁰ We also investigated whether incident CMBs were related to baseline amyloid load in the occipital lobe, which may be related to vascular amyloid.¹¹

Statistical analysis

Demographic and clinical characteristics of the participants were summarized using means and SDs for continuous variables and counts and percentages for categorical variables. The distributions of the continuous variables were examined for approximate symmetry and normality using plots; PiB SUVR was subsequently log-transformed for statistical tests to reduce a positive skew. We conducted statistical analyses to answer 3 broad questions in a stepwise fashion:

1. Which factors influence the age-specific incidence of CMBs? We employed a piecewise exponential model (PEM)¹² to calculate integer age-specific incidence rates of CMBs and to evaluate the effect of baseline factors (sex, hypertension, amyloid PET) on the incidence rates. Like the Cox model, the PEM adjusts for covariates by prescribing the log hazard ratio to be linear in the covariates, but in addition specifies constant baseline hazards in prespecified intervals, allowing a complete specification of the likelihood and simultaneous estimation of both baseline hazards and hazard ratios. We used a stepwise procedure to derive a multivariable PEM including terms to maximize the Akaike information criteria. Model terms with 2-sided p values ≤0.05 in the final fit were regarded as statistically significant. The PEMs were fit with and without baseline CMBs. These were reported with and without baseline CMBs. We used stepwise regression to fit a multivariable model including terms to maximize the Akaike information criteria. Model terms with 2-sided p values ≤0.05 were regarded as statistically significant.

2. What is the population incidence of CMBs? Based upon the incidence rates found in the multivariable model of step 1, we derived standardized rates for our study population (Olmsted County residents in 2010). This was done in the usual manner by enumerating all combinations of predictors in the model of step 1 found in our study population, calculating the relative weight of each combination in our population, multiplying these weights and the corresponding incidence rate estimates from the PEM, and summing.

3. Finally, what is the causal ordering of events that effects the development of a future CMB? We developed structural equations models (SEMs) to determine if age at MRI (age), sex (male), APOE4 carrier status (APOE4), smoking (ever/never), hypertension, diabetes, dyslipidemia, or amyloid PET (log-transformed) could have direct or indirect effects on time to next CMB. We also developed models with occipital amyloid PET instead of global amyloid since occipital amyloid may be more associated with CAA. To that end, we tested discrete survival time SEMs (figure 1). Prior publications indicated that baseline CMBs might predict incident CMBs; therefore, we also tested SEMs including baseline CMBs. The “prior CMB” variable is the number of CMBs at baseline MRI. SEM diagrams contain 2 types of nodes: squares contain manifest (measured) variables, and circles contain latent (unmeasured but assumed to be present) variables. “Next CMB” is a latent variable representing time to next CMB, manifested by year 1–2 to year 4–5: 4 (for any CMB and lobar CMB) or 3 (for deep CMB) time periods (years) with either 0 for no CMBs or 1 for observed CMBs. We had almost no scans, and no observed CMBs, within 1 year of baseline. Arrows that join nodes are direct effects (e.g., APOE4→event), and arrows that join nodes but pass through intervening variables (e.g., APOE4→amyloid→event) are indirect effects. The intervening variables are mediators. Individual indirect effects can be added to produce a total indirect effect, and the indirect and direct effects can be added to produce a total effect. We measured the effects using regression models and reported the regression coefficient for each (either a linear-regression coefficient where the outcome is continuous, or a log-hazard ratio where the outcome is next CMB), the associated standard error, and the p value. We started with the full SEM and used backwards elimination to reduce the SEM to a parsimonious model. MPlus, version 8, was used to fit the models.¹³

Figure 1. Discrete survival time structural equation model.

Rectangles contain manifest (measured) variables and the circle contains the latent (unmeasured but assumed to be present) variable. “Next CMB” is a latent variable representing time to next cerebral microbleed (CMB), manifested by 4 (for any CMB and lobar CMB) or 3 (for deep CMB) time periods (years) with observed events. “Prior CMB” variable is number of CMBs at baseline.

Standard protocol approvals, registrations, and patient consents

The MCSA and associated studies were approved by the Mayo Clinic and Olmsted Medical Center institutional review boards, and written informed consent was obtained from all participants before they joined the study. This article does not include photographs, videos, or other information of any recognizable person.

Data availability

Data from the MCSA, including data from this study, are available on request.

Results

The baseline characteristics of participants with 2 or more GRE scans are reported in table 1. The mean (SD) interval between MRI scans was 2.7 (1.0) years. Participants having 2 scans numbered 519, 124 participants underwent 3 scans, and 8 participants underwent 4 scans. Sixty-five percent (n = 50) of those with lobar only CMBs, 73% (11) with deep only CMBs, and 86% (12) of those with mixed CMBs had hypertension. Seventy-two participants (11%) developed new CMBs, 54 (8%) developed new lobar CMBs, 16 (2%) developed new deep CMBs, and 2 (0.3%) developed new cerebellar CMBs. Thirty-seven were individuals who developed their first CMB and 35 individuals developed a new CMB with at least 1 baseline CMB already present. A total of 111 (17%) individuals had a CMB at the baseline scan (lobar n = 77, deep = 15, mixed n = 14, cerebellum only n = 5). The distribution of the baseline CMBs among those who developed a new CMB was baseline lobar n = 19, baseline deep/infratentorial = 4, baseline mixed n = 10, and cerebellum only n = 2. A total of 76 participants with at least one baseline CMB did not develop any new CMBs. Twenty-five individuals had multiple (2 or more) new CMBs between scans.

Table 1.

Baseline characteristics of participants with 2 or more MRI scans with gradient-recalled echo sequences

Table 1 compares participants who developed a new CMB to those participants who did not. Participants with new CMB were older (74.9 [SD 9.8] vs 69.2 [9.9] years) and more likely to be male (52 [72%] vs 304 [53%]). Participants with new CMBs were also more likely to have baseline CMBs: 35 (49%), compared to those without incident CMBs, 76 (13%). Four individuals developed a nontraumatic intracerebral hemorrhage at follow-up.

Table 2 reports the proportional hazards for likelihood of developing CMB in any location or in a lobar-only location. The final model included covariates for age, male sex, and baseline CMBs when estimating CMBs incidence in any location. Age, male sex, and baseline CMBs were associated with incident lobar CMBs. In a model (not shown) accounting for baseline lobar-only CMBs rather than any baseline CMB, increasing PiB SUVR was also associated with incident lobar-only CMBs: hazard ratio (95% confidence interval) 1.87 (1.02–3.44), p = 0.043.

Table 2.

Proportional hazards (95% confidence interval [CI]) for likelihood of developing cerebral microbleeds (CMBs) in any location where the final full model is derived using the piecewise exponential model

Population incidence of CMBs

To describe the incidence of CMBs in the population, we used the multivariable PEM in table 2 for incident CMB with predictors age, sex, and prior CMBs. The incidences are described in table 3 with confidence intervals, and visually in figure 2. Advancing age and male sex were both strongly associated with increased incidence of CMBs.

Table 3.

Rates per 100 person-years of incident cerebral microbleeds (CMBs) by age, sex, and baseline CMB status from piecewise exponential mode

Figure 2. Incidence rates of cerebral microbleeds (CMBs) by age according to baseline CMB status and sex.

Fitting a PEM for incidence of any CMB as predicted by age and sex, and then standardizing the estimated rates by the Olmsted County 2010 population, the estimated overall incidence rate for developing a new CMB in the population aged 50–89 years was 3.6 per 100 persons per year. The incidence rate increased with age from 1.5 new CMBs per 100 person-years at age 50–5.6 and 11.6 per 100 person-years at ages 70 and 90. To make sure the results were not biased by including the 4 participants with dementia, we repeated the analysis without the 4 patients with dementia and the results did not change significantly.

Structural equation models of CMBs and amyloid PET (without baseline CMBs)

Models for any incident CMB (deep or lobar)

In SEMs examining any incident CMB (deep or lobar), only age and male sex (model not shown) were associated with time to next CMB.

Models for incident deep CMB

None of the variables was associated with time to deep-only CMB.

Models for incident lobar CMB

The structural equation models examining time to incident lobar CMB are shown in figure 3. As expected, increasing age at MRI (estimate [standard error] 0.010 [0.001], p < 0.001) and APOE4 (estimate 0.116 [0.018], p < 0.001) were significant direct predictors of higher amyloid burden at baseline. There was an increased risk (hazard) of subsequent lobar CMBs through (1) a direct effect of higher amyloid burden (estimate 1.976 [0.637], p = 0.002); (2) indirect effects of older age at MRI (estimate 0.022 [0.006], p = 0.001) and APOE4 (estimate 0.228 [0.08], p = 0.002) through higher amyloid at baseline; (3) a direct effect of male sex (estimate 0.921 [0.357], p = 0.010); and (4) a direct effect of diabetes (estimate 0.804 [0.401], p = 0.045).

Figure 3. Model without prior cerebral microbleeds (CMBs): predicting incident lobar-only CMBs.

Rectangles contain manifest (measured) variables and the circle contains the latent (unmeasured but assumed to be present) variable. Predicting lobar CMB reduced to the following model: “event” is a latent variable representing time to next CMB.

Structural equation models of CMBs and amyloid PET including baseline CMBs

Models for any incident CMB (deep or lobar)

The structural equation models examining any incident CMB were repeated after adding baseline CMBs (figure 4A). These models revealed the following: (1) older age at MRI associated with more CMBs at baseline (estimate 0.054 [0.023], p = 0.020) and diabetes (estimate 0.003 [0.001], p = 0.020), and also associated with an increased risk (hazard) of a subsequent CMB (any location; total indirect estimate 0.025 [0.012], p = 0.033); (2) baseline CMBs also directly increased the risk of a subsequent CMB (estimate 0.415 [0.135], p = 0.002; and (3) diabetes directly increased the risk of a subsequent CMB (estimate 0.827 [0.315], p = 0.009) (figure 4A). Neither amyloid PET (p = 0.172) nor APOE4 (p = 0.992) were significant predictors of time to event in this model. Substituting occipital amyloid PET did not change the model.

Figure 4. Models with prior cerebral microbleeds (CMBs).

(A) Predicting incident CMBs (any location). (B) Predicting incident lobar CMBs with occipital amyloid load. Rectangles contain manifest (measured) variables and circles contain latent (unmeasured but assumed to be present) variables. “Next CMB” is a latent variable representing time to next CMB. “Prior CMB” variable is number of CMBs at baseline.

Models for incident deep CMB

The models for incident deep CMB were similar to figure 4A, with the exception of diabetes, which did not increase the risk of a subsequent deep CMB. Prior CMBs with a log hazard ratio of 0.293 (0.069) predicted a future CMB. Again, substituting occipital amyloid PET did not change the model.

Models for incident lobar CMB

At p < 0.05 significance level, amyloid and APOE4 did not significantly contribute to the model. We therefore had a model diagram similar to figure 4A with an estimate of 0.375 (0.141) for the log hazard ratio from prior CMBs to incident lobar CMBs with the exception that male sex (estimate 0.735 [0.343], p = 0.032) was also associated with incident lobar CMBs. By substituting occipital amyloid PET load for the meta-ROI amyloid PET, amyloid became a direct predictor of incident lobar CMBs (figure 4B). Age at MRI (estimate 0.007 [<0.001], p < 0.001), male sex (estimate −0.023 [0.009], p = 0.006), and APOE4 (estimate 0.043 [0.010], p < 0.001) were significant direct predictors of higher occipital amyloid burden at baseline. There was an increased risk (hazard) of subsequent lobar CMBs through (1) a direct effect of higher amyloid burden (occipital) (estimate 2.379 [1.098], p = 0.030); (2) indirect effects of older age at MRI through amyloid, diabetes, and prior CMBs (total indirect estimate 0.039 [0.012], p = 0.002) and APOE4 through amyloid (estimate 0.103 [0.050], p = 0.041); (3) a direct effect of male sex (estimate 0.814 [0.361], p = 0.024); (4) a direct effect of prior CMBs (estimate 0.344 [0.133], p = 0.010); and (5) a direct effect of diabetes (estimate 0.857 [0.378], p = 0.023).

Discussion

In this population-based study, the incidence of CMBs was 3.6 per 100 persons per year. Using the piecewise exponential model, the effect of age and the presence of baseline CMBs on the incidence of CMBs can be visualized. The majority of new CMBs were lobar. Increased age, male sex, and the presence of baseline CMBs were associated with developing a CMB in any location and in the lobar-only location. In addition, baseline amyloid load in the occipital lobe was associated with developing lobar-only CMBs even when taking into consideration baseline CMB status.

Prior studies have shown an association of age and APOE4 status on the prevalence of CMBs.¹⁴ In order to better characterize the direct and indirect effects of these variables on the risk of CMBs and the temporal ordering of events, SEMs were employed. Based on our models, older age and the presence of an APOE4 allele were associated with greater amyloid load and higher risk of a new lobar CMB. A novel finding from this work was that baseline CMBs were the strongest predictor of future new CMBs. Further, we found that the effect of baseline CMBs on the risk of future new CMBs was stronger than the effect of greater amyloid load at baseline, and this was true even when the analysis was restricted to new lobar CMBs.

CAA pathology has an occipital predilection.¹⁵ Importantly, when we investigated amyloid regionally, occipital amyloid load was associated with only incident lobar CMBs even after taking into account baseline CMBs. Amyloid PET studies have shown elevated occipital amyloid load in patients with CAA compared to patients with Alzheimer dementia.¹⁶ Since strictly lobar CMBs are associated with CAA pathophysiology, the relationship between occipital amyloid load and incident lobar CMBs suggests that CAA pathology may be contributing to the development of incident lobar CMBs. In patients with CAA, the relationship between amyloid load and the location of incident CMBs has been investigated. New lobar CMBs occurred in areas with higher baseline amyloid load.¹⁷

Our study extends the findings of the Rotterdam Scan Study and Age, Gene/Environment Susceptibility (AGES)–Reykjavik Study through 3 important findings: (1) we considered PiB-PET of participants at baseline to determine the role of amyloid load on incident CMBs; (2) we provided age-specific incident rates of new CMBs; and (3) we evaluated the causal ordering of events that influence the risk of future CMBs.

In the population-based Rotterdam Scan Study, over a 3.4-year time window, approximately 10% of participants who underwent 2 MRIs developed new CMBs.² In the AGES-Reykjavik Study, 18.4% developed new CMBs over a 5.2-year period.¹ In the present study, 11% developed new CMBs over a 2.7-year time window. In addition to differences in time interval between MRIs, differences in participants' characteristics and imaging techniques may explain the slightly different rates and associations with CMBs. The mean age was similar between the current study (69.8 years) and the Rotterdam Scan Study (68.5 years); participants in the AGES-Reykjavik Study were older (mean age 74.6 years). While there was a slightly higher prevalence of men in the current study (55%), participants in the AGES-Reykjavik Study were predominantly female (58.9%). The Rotterdam Scan Study and the AGES-Reykjavik Study participants underwent 1.5T MRI scans, compared to 3T MRI scans in the current study.

Prior cross-sectional studies have shown that APOE4 and atrophy in signature regions for Alzheimer disease on MRI are both associated with lobar CMBs.^18,19 Using structural equation models, this study demonstrated that increasing age and APOE4 carrier status indirectly increase the risk of lobar CMBs by increasing amyloid load. Therefore, prior associations of APOE4 and atrophy in Alzheimer disease with lobar CMBs may reflect that these lobar CMBs are indirect markers of CAA pathology, which is associated with increased amyloid load on PET, particularly in the occipital lobe.¹¹

Interestingly, in the SEM models we found a direct relationship between diabetes and incident CMBs. While some studies have not shown a relationship between diabetes and CMB prevalence,^14,18 in the AGES-Reykjavik Study, diabetes was associated with the presence of more than 1 CMB.²⁰ More recently, diabetes has been associated with hemorrhage risk among patients with atrial fibrillation treated with anticoagulation.²¹ Whether treating diabetes may influence these risks requires further investigation.

The presence of baseline CMBs has previously been shown to predict subsequent additional CMBs.^1,2 The inclusion of baseline CMBs in the SEM model overwhelmed the effect of global amyloid load on incident CMBs. Since there were only 54 (8%) incident lobar CMBs and a subset of 49 also had amyloid PET, we expect the relationship between amyloid load and incident lobar CMBs to be stronger in a larger cohort with more lobar CMBs. However, our findings indicate that CMBs are expressions of a progressive cerebrovascular pathology, and therefore the presence and number of CMBs on brain MRI can estimate future risk of progression. In individuals with lobar CMBs, the degree of amyloid load, particularly in the occipital lobe, on PET scan can also assist in risk stratification.

Deep/infratentorial CMBs are associated with cardiovascular risk factors, in particular hypertension.^3,14,19,22 In the Rotterdam Scan Study, vascular risk factors, lacunar infarcts, and white matter hyperintensities were associated with incident deep/infratentorial CMBs.² We have confirmed the relationship between hypertension and deep/infratentorial CMBs cross-sectionally,⁸ but we did not detect a relationship between hypertension and new deep/infratentorial CMBs over time in the present study, likely because only 16 incident deep/infratentorial CMBs occurred. Longer follow-up and more participants with hypertension will be necessary to demonstrate this potential relationship.

This article has several clinical implications. Individuals with CMBs may undergo longitudinal imaging within the context of their clinical care. This study provides estimates of how often incident CMBs may occur in the general population as well as the demographics such as male sex and baseline CMBs that increase the risk of a future CMB, which can inform patients and their clinicians. These estimates also have implications for clinical trial design as CMBs are now used as an outcome marker in clinical trials utilizing amyloid immunotherapy. For example, occipital amyloid load may be a candidate biomarker to predict which clinical trial participants develop a future CMB.

The key strengths of our study were the population-based nature of the study and the availability of PiB-PET scans for most of the study participants. In addition, we were able to estimate a true population incidence of CMBs by accounting for 2 types of biases: study nonparticipation and imaging nonparticipation. There were some limitations as well. It is possible that some CMBs were present at baseline but were missed due to slice thickness and may appear incidental. Similar to other population-based studies, we used T2* GRE sequences. The utility of susceptibility-weighted imaging in the future will be able to provide greater sensitivity for detecting CMBs. Though one may argue that the sample size was small, the ascertainment of amyloid burden and measurement of CMBs using longitudinal imaging data allowed us to investigate the effect of amyloid on future new CMBs. The length of follow-up was different among study participants, as expected for observational studies, but inclusion of time as a continuous variable accounted for differences in the length of follow-up as well as number of follow-ups. Finally, it remains uncertain how much vascular amyloid PiB binds to even in the occipital lobe and future studies with CSF may provide additional insight into the CMB amyloid relationship since β-amyloid 40 and 42 can be measured separately.

The incidence of CMBs increases steeply with age. While APOE4 and age act through amyloid to increase the risk of incident lobar CMBs, the presence of baseline CMBs was the most important predictor of future CMBs.

Glossary

AGES-Reykjavik Study

Age, Gene/Environment Susceptibility–Reykjavik Study

CAA

cerebral amyloid angiopathy

CMB

cerebral microbleed

GRE

gradient-recalled echo

MCSA

Mayo Clinic Study of Aging

PEM

piecewise exponential model

PiB

Pittsburgh compound B

ROI

region of interest

SEM

structural equations model

SUVR

standardized uptake value ratio

Appendix. Authors

Study funding

Research reported in this publication was supported by the National Institute on Aging of the NIH under award numbers K76AG057015, R01 AG041851, U01 AG006786, and R01 AG034676 (the Rochester Epidemiology Project); National Institute for Neurologic Disorders and Stroke NS097495; and the GHR Foundation. The funders had no role in the conception or preparation of this manuscript.

Disclosure

J. Graff-Radford reports grants from National Institute on Aging, receives funding from the Alzheimer's Treatment and Research Institute, and received an honorarium from the American Academy of Neurology for serving as a guest editor of Continuum. T. Lesnick reports no disclosures relevant to the manuscript. A. Rabinstein receives royalties from Elsevier and Oxford University Press and has received research support from DJO Global, Inc. J. Gunter, J. Aakre, S. Przybelski, and A. Spychalla report no disclosures relevant to the manuscript. J. Huston reports patents from the Mayo Foundation for Medical Education and Research; royalties from Resoundant, Inc.; and stock and stock options in Resoundant, Inc. He reports no competing financial interests to the present article. R. Brown reports no disclosures relevant to the manuscript. M. Mielke is a consultant for Eli Lilly and Co. and Lysosomal Therapeutics, Inc. She receives unrestricted research grants from Biogen, Lundbeck, and Roche Holding AG and receives research funding from the National Institute on Aging of the NIH and the US Department of Defense. V. Lowe serves on the scientific advisory boards for Piramal Imaging, Merck Research, Inc., and Bayer Schering Pharma. He reports receiving funding from the NIH. D. Knopman serves on a data safety monitoring board for the Dominantly Inherited Alzheimer Network study. He is an investigator in clinical trials sponsored by Eli Lilly Biogen and the Alzheimer's Treatment and Research Institute and receives research support from NIH. R. Petersen is a consultant for Roche Holding AG, [Q] Biogen, Merck & Co., Eli Lilly and Co., and Genentech. He receives publishing royalties from Mild Cognitive Impairment (Oxford University Press, 2003) and research support from NIH. C. Jack consults for Eli Lilly and serves on an independent data monitoring board for Roche Holding AG. He receives no personal compensation from any commercial entity. He receives research support from the National Institute on Aging of the NIH and the Alexander Family Professor of Alzheimer's Disease Research, Mayo Clinic. P. Vemuri reports grants from the National Institute for Neurologic Disorders and Stroke. W. Kremers reports grants from the National Institute on Aging and research funding from AstraZeneca, Biogen, and Roche. K. Kantarci serves on the data safety monitoring board for Takeda Global Research & Development Center, Inc.; receives research support from Avid Radiopharmaceuticals, Inc., and Eli Lilly and Co.; and receives funding from NIH and the Alzheimer's Drug Discovery Foundation. Go to Neurology.org/N for full disclosures.