Prevention of Alzheimer’s Disease: Lessons Learned and Applied

2017 Aug 2. doi: 10.1111/jgs.14997. [Epub ahead of print]

Prevention of Alzheimer’s Disease: Lessons Learned and Applied.

Galvin JE1.


Alzheimer’s disease (AD) affects more than 5 million Americans, with substantial consequences for individuals with AD, families, and society in terms of morbidity, mortality, and healthcare costs. With disease-modifying treatment trials unsuccessful at the present time and only medications to treat symptoms available, an emerging approach is prevention. Advances in diagnostic criteria, biomarker development, and greater understanding of the biophysiological basis of AD make these initiatives feasible.

Ongoing pharmacological trials using anti-amyloid therapies are underway in sporadic and genetic forms of AD, although a large number of modifiable risk factors for AD have been identified in observational studies, many of which do not appear to exert effects through amyloid or tau. This suggests that prevention studies focusing on risk reduction and lifestyle modification may offer additional benefits. Rather than relying solely on large-sample, long-duration, randomized clinical trial designs, a precision medicine approach using N-of-1 trials may provide more-rapid information on whether personalized prevention plans can improve person-centered outcomes.

Because there appear to be multiple pathways to developing AD, there may also be multiple ways to prevent or delay the onset of AD. Even if these precision approaches alone are not successful in preventing AD, they may greatly improve the likelihood of amyloid- or tau-specific therapies to reach their endpoints by reducing comorbidities. Keeping this in mind, dementia may be a disorder that develops over a lifetime, with individualized ways to build a better brain as we age.

Alzheimer’s disease (AD) affects more than 5 million Americans, with substantial costs to individuals with AD, families, and society.[1] Projections by the Alzheimer Association are that, if nothing is done, by 2050, there will be more than 16 million people with AD in the United States and more than 60 million worldwide. Over the past 25 years, only five symptomatic medications have met their primary clinical trial endpoints in Phase III clinical trials and successfully come to market; of these, four are still available.

Since 2003, every symptom- and disease-modifying agent has failed in Phase II or III trials because of challenges with safety or efficacy. This led to a bold initiative put forth in the National Alzheimer Plan Act to develop a disease-modifying treatment (DMT) by 2025.[2] Two important concepts are associated with success to reach this target date. First, only medications that have already entered Phase II testing can make it to market by 2025.[2] Second, if a DMT were available by 2025, then the 2050 projection of 16 million Americans with AD would be reduced by 5.7 million cases, with societal savings of $367 billion.[1]

To complement efforts to develop a DMT for individuals with symptomatic AD, a concerted effort is underway to initiate preventive measures in asymptomatic individuals. Such efforts are also consistent with the 2025 target goal.An important question, is whether AD can be prevented. A large number of modifiable (e.g., exposures, lifestyle and social habits) and nonmodifiable (e.g., age, sex, genetics) risk factors have been identified (Table 1). Recent revisions to the clinical criteria for AD[3] and mild cognitive impairment (MCI)[4] helped clarify the role of biomarkers in defining the pathological cascade, and the addition of research criteria for presymptomatic disease[5] sets the stage for better modeling of the preclinical and prodromal stages of disease.[6]

Efforts developing and validating fluid (blood and cerebrospinal fluid) and imaging biomarkers make it possible to explore underlying pathological changes in amyloid, tau, dopamine transport, inflammation, signaling pathways, and in the future, alpha-synuclein and TDP-43 in symptomatic, prodromal, and presymptomatic individuals. Advances in genetic, epigenetic, and “omic” (e.g., proteomic, lipidomic, metabolomic) approaches will permit the modeling of transcriptional, translational, and posttranslational changes. Furthermore, precision medicine approaches with demonstrable benefits in oncology are being applied to neurodegenerative disorders. Thus, the platform is in place to begin prevention initiatives.

Table 1. Alzheimer’s Disease Risk and Protective Factors

  1. Apolipoprotein E is the major risk gene; a number of other minor risk genes have been identified.
Risk Factors (Nonmodifiable)
Family history
Apolipoprotein E ε4 allele*
Risk Factors (Modifiable)
Diabetes mellitus and insulin resistance
Metabolic syndrome
Cerebrovascular disease
Psychological and physiological stress
Traumatic brain injury
Sleep disordered breathing
Alcohol abuse
Protective Factors (Modifiable)
Cognitive reserve and mental activity
Educational attainment and lifelong learning
Cognitive leisure activities
Physical activity and exercise
Social engagement
Mindfulness and wellness activities
Optimism and purpose in life
Omega-3 intake


These efforts have a great potential for pharmaceutical and nonpharmacological approaches, with earlier identification of at-risk individuals, expanding opportunities for faster and earlier diagnoses, better stratification of at-risk individuals, higher enrollment into randomized clinical trials (RCTs) by reducing screen failure rates, and eventually more-effective treatments. Although the focus of this article is on AD, the principles discussed are relevant to related neurodegenerative disorders (e.g., Parkinson’s disease, Lewy body dementia, vascular dementia). Assuming that curing AD remains a substantial unsolved challenge, preventing or simply delaying the onset of dementia could significantly change the face of disease.

Pharmacological Approaches to Prevention

A number of prevention studies are ongoing in sporadic and autosomal-dominant forms of AD. Results are still pending, but a review of[7] provides important details about these studies. One such study is the Anti-Amyloid Treatment in Asymptomatic Alzheimer’s study, which is recruiting individuals aged 65 to 86 with normal cognitive function but positive AD biomarkers (amyloid deposition according to positron emission tomography). These individuals are followed for 3 years of treatment with an anti-amyloid monoclonal antibody. The Alzheimer Prevention Initiative will study individuals aged 60 to 75 with normal cognition who have two copies of the apolipoprotein E e4 allele, putting them at high risk of AD. These individuals are followed for 2 years of treatment with a different anti-amyloid monoclonal antibody. The TOMMORROW Study will enroll individuals aged 65 to 83 with normal cognition who have a polymorphism of the TOMM40 gene that is associated with greater risk of AD. Individuals are followed for 5 years of treatment with pioglitazone, an anti-diabetes drug. In parallel to these trials in sporadic cases with enhanced risk, trials are ongoing in individuals with autosomal-dominant forms of AD, including the Alzheimer Prevention Initiative and Dominantly Inherited Alzheimer Network—Treatment Unit using monoclonal antibodies. A particular advantage of the trials in familial AD is that age of onset is predictable so that, if a DMT effect exists, it is more likely to be detected. A potential disadvantage of these trials is that the results may not be generalizable to the much more common sporadic forms in which risk factors other than genetics may predominate.

Similar to prevention trials, there are number of prodromal studies with a major focus on individuals with MCI. Results are pending for many of these trials, but at least one Phase II study had promising results. Aducanumab immunotherapy[8] against the amyloid protein not only demonstrated changes in cerebrospinal fluid amyloid over 54 weeks in a dose-dependent manner, but also demonstrated significant changes in cognitive and functional measures. These findings support the concept that there may be a critical window for these agents to be administered to increase the likelihood of success.

Lifestyle Modification Interventions

Modifiable Lifestyle Factors

A number of modifiable risk and preventive factors for AD have been described in observational studies (Table 1). Risk factors that have been found in several studies include diabetes mellitus, hypertension, renal dysfunction, alcohol and smoking patterns, high cholesterol, coronary heart disease, depression, sedentary life style, low cognitive activity, and diet. These factors combined account for more than half of the attributable risk for AD.[9] The most difficult of these factors to address is diet because it is highly dependent on income and access to fresh foods.[10] In a 16-year observational study of 949 individuals using the Lifestyle for Brain Health (LIBRA) measure of modifiable risk factors, a 1-point increase in LIBRA score was associated with a 19% greater risk of dementia.[11] In a metaanalyses of 19 studies, cognitive leisure activities,[12] including crossword puzzles, card games, computer use, arts and crafts, life-long learning, group discussions, and music, had a protective effect (odds ratio (OR) = 0.58). In addition, physical activities may lead to a 20% to 65% risk reduction depending on the type and intensity of activity through mechanisms involving lower vascular disease risk, better respiratory function, stimulation of trophic factors, and lower oxidative stress and inflammation.[13] Objective measurement of midlife vascular risk factors demonstrated greater risk of dementia in late life.[14] In a study of 2,000 individuals aged 71 to 78, work-related stress increased the risk of MCI (OR = 1.38), dementia (OR = 1.53), and AD (OR = 1.55).[15]

Despite the differences between countries of origin, culture, language, educational attainment, and ages studied, the aforementioned studies and many others are convergent for a short list of risk factors that seem to play a critical role in the development or prevention of AD and related disorders. This consistency has led to the implementation of a number of dementia prevention initiatives to modify these risk factors, most of which cannot be directly linked to amyloid or tau deposition.

Ongoing Prevention Initiatives

The European Prevention of AD[16] initiative is recruiting participants to examine whether alteration of risk factors for AD that occur in early and mid-life potentiate pathophysiological changes decades before dementia onset. The Innovative Midlife Intervention for Dementia Deterrence trial[9] is examining 11 identified risk factors (e.g., diabetes, hypertension, renal) that account for half of the attributable risk and has enrolled 600 individuals to participate in an on-line education intervention. The largest initiative to date is the FINGER study,[17] enrolling 1,260 individuals in an educational intervention that includes modules in diet, exercise, cognitive training, and vascular risk reduction. Overall between-group differences were statistically significant for global cognition, executive function, and processing speed but not episodic memory.[17] These results suggest that lifestyle modification may offer some benefit in cognitive function, albeit with a small effect size.

Precision Medicine Approaches to Prevention

Although age is the single greatest risk factor for AD, AD is not inevitable. The best estimates suggest that, at age 85, there is 42% risk of developing AD,[1] which means that 58% of older adults do not develop dementia, even if amyloid can be detected in the brain. The reasons are unknown, but may be explained in part by a host of modifiable and nonmodifiable risk factors (Table 1). Up to 30% of AD cases may be preventable through modification of risk factors and behavioral changes to mitigate the effect of those risk factors that are not modifiable.[18] There is an ongoing debate as to whether the current evidence base is sufficient to initiate prevention programs because it is difficult to prove causation from observational studies, and it is difficult to pool multiple RCTs because of differences in study design, measurements used, and anticipated outcomes.[18] Although a well-balanced, healthy lifestyle may be the cornerstone of disease prevention and brain health, each risk factor (vascular, lifestyle choices, psychosocial) may both act independently and potentiate the effects of each other.[19, 20] Therefore, a prevention initiative needs to be multimodal and tailored to address individual risks.

These requirements lead to a number of design and analytical challenges. Many prevention RCTs use time-to-event analytical strategies to demonstrate a DMT effect. Such designs are optimal when anticipated treatment effects remain constant over time, but in the case of dementia prevention, this is unknown. Thus, time-to-event analyses such hazard ratios may not be the best way to model effects, particularly if the detectable signal is a late effect of the intervention.[21] Because AD is heterogeneous in terms of risk factors, age of onset, presentation, progression, and pathology burden, designing a study to treat individuals as a homogenous population requires large sample sizes (thousands to tens of thousands) to be followed for long periods of time (years to decades).[22] This results in large study costs, staff burden, and participant burden. In the absence of robust biomarkers that mark disease onset and progression, rather than just the presence of pathology, RCT design will remain a challenge.[23] Barriers to prevention studies include limited understanding of the real relationship between dementia risk factors and the effect of risk reduction; complexity of the effect of life course on dementia risk factors; lack of standardization of study design, definitions, and outcomes; difficulty translating RCT findings into real-world practice; cultural and social barriers to implementation; lack of research capacity to enroll large research cohorts for long periods of time; and pervasive social stigma associated with AD.[18]

Because effective prevention strategies are elusive despite significant advances in understanding of the biology and pathophysiology of AD, an alternative approach would be to take advantage of precision medicine designs used in oncology trials to tailor interventions to an individual’s phenotypic and genotypic expression. With better classification and phenotyping of individuals with AD, trial-ready cohorts can be targeted more appropriately with interventions (pharmacological and nonpharmacological) designed to ameliorate specific pathological mechanisms, based on specific biomarkers and individual characteristics.[23] Future trials could then be created to determine efficacy and safety (fast to fail) more quickly by moving away from one-size-fits-all approaches to person-specific precision treatments.

This would also require rethinking trial design for prevention measures, moving to more N-of-1 designs. N-of-1 trials consider the individual as the sole unit of observation to study the efficacy and adverse effects of an intervention[22] and are guided by objective data-driven criteria while leveraging the study designs and statistical techniques common to RCTs.[22] Because risk and molecular profiles of AD vary widely by person, grouping individuals into single entities (placebo vs treatment arm) may mix “super-responders” with “nonresponders,” washing out treatment effects that only become apparent in post hoc analyses.[24] Instead, comparing time-to-disease progression of an individual using a novel therapeutic approach to the time-to-disease progression for that same individuals for the immediately preceding treatment paradigm may be preferable.[25] N-of-1 trials may be less bound by threats to generalizability of large RCTs due to recruitment delays and challenges to translate significant p-values in large treatment groups to the care of an individual, which is the ultimate goal of clinical practice. In a metaanalyses of 70 N-of-1 trials, 50 of 57 completed trials provided definitive clinical or statistical answers, with 39% prompting physicians to change the plan of care.[26] Another metaanalysis examined 108 trials involving 2,154 participants and found that 54% of participants had subsequent treatment decisions changed based on the results.[27] To create the platform for such trials, several conditions must be met (Table 2). Participants must be deeply phenotyped with characterization of sociodemographic, psychological, clinical, cognitive, functional, biomarker, and genetic traits. Ideally, these individuals would agree to be followed longitudinally, have samples banked for future analyses, and consent to autopsy to provide confirmation of diagnosis and treatment effects on brain pathology. Statistical considerations may take advantage of alternative time-series analyses and within- and between-subject comparisons. Lastly, an important concern is the cost of care when a high-cost intervention is planned—for example, off-label treatment with an expensive medication.[22]

Table 2. Basic Principles of a Dementia Prevention Program

Establish a longitudinal cohort of individuals without memory impairment and with prodromal disease
Develop a protocol that can measure person-centered and health-economic outcomes
Evaluate clinical, cognitive, functional, and behavioral features annually
Collect and bank biomarkers: blood, spinal fluid, deoxyribonucleic acid, cell lines, magnetic resonance imaging, positron emission tomography
Encourage autopsy participation
Perform deep phenotyping of individuals with near total participation in all biomarker collection protocols by all participants
Apply precision medicine–type interventions to match treatment to individual characteristics
Test customized N-of-1 interventions over a designated period to determine whether protocols alter biophysiological profiles, disease-relevant biomarkers, and outcome measures
Develop a statistical plan to incorporate immediate, intermediate, and long-term time-to-event analyses
Create a trial-ready cohort for large-scale pharmacological and nonpharmacological interventions


Example of a Personalized Medicine Approach to Dementia Prevention

As an example of how this may applied in a pragmatic sense, N-of-1 trials are being developed to personalize dementia prevention using an evidence-base derived from an extensive literature review and results of a National Institutes of Health–funded project to conduct dementia screening in multicultural communities (R01 AG0402–11-A1, study design reviewed[28, 29]). In addition to screening for cognitive impairment, broader medical screening for diabetes mellitus, hypertension, vascular risk factors, obesity, mobility, physical performance, frailty, and depression was incorporated into a “healthy body, healthy mind” approach to make the concept of dementia screening more acceptable and to understand the effect of comorbid disease on cognitive performance. This cross-sectional study confirmed many findings of observational studies regarding the association between cognitive performance and diabetes mellitus, hypertension, obesity, vascular risk factors, and depression while providing novel findings linking cognitive performance to sarcopenia[28] and mobility.[29] These collective findings were prospectively applied to develop N-of-1 trials. For example, a 68-year-old college-educated woman (although an actual case, some features were altered to preserve anonymity) presented with a 1-year history of subjective memory complaints (misplacing car keys, forgetting conversations, defensiveness about memory issues) but with independent functioning in everyday activities. Her relevant past history was significant for hypertension and hypercholesterolemia. Physical examination findings included mild hypertension (blood pressure 132/92 mmHg) but normal cardiac and peripheral vascular examinations. Pertinent neurological findings included mild symmetric weakness, poor vibration sensation in the lower extremities, and mild postural instability. Cognitive testing revealed global deficits (Montreal Cognitive Assessment score 19/30), working memory and executive dysfunction, and episodic memory deficits on list learning that disappeared with cued recognition. Physical and functional testing revealed low lean muscle mass, at-risk nutritional status, modest daily physical activity, mild deficits in physical functionality (Short Physical Performance Battery score 7/12), and mild frailty (Fried Frailty Phenotype Score 3/5). Gait and balance testing revealed slowed gait speed (1.07 m/s), marked slowing (22.9%) with a dual-task challenge (walking while talking), and postural sway with eyes closed. Blood-based biomarkers revealed an abnormal lipid profile (high total, low-density lipoprotein (LDL), and non–high-density lipoprotein (HDL) cholesterol; high LDL and small LDL particles; and low HDL particles), high inflammatory markers (high-sensitivity C-reactive protein and myeloperoxidase), apolipoprotein (Apo)E 3/4 genotype, and evidence of insulin resistance (high fasting glucose, glycosylated hemoglobin, and estimated average glucose). Quantitative magnetic resonance imaging (MRI) revealed normal hippocampal size and lateral ventricle volume but confluent white matter hyperintensities with frontal lobe predominance. Auditory event–related potentials demonstrated slow median reaction times and low amplitude at the N200 peak (linked to impaired attention and executive function) and left–right asymmetry with frontal predominance (linked to vascular injury) but normal amplitudes and latencies at all other peaks, including the P50 peak associated with amyloid deposition.[30] This deep phenotypic evaluation provided findings in cognitive testing (executive and working memory deficits with cued episodic memory improvements supporting intact hippocampal circuitry), physical testing (sarcopenia, at-risk nutritional status, poor physical functionality, and early frailty), gait testing (slowed gait speed, impaired dual tasks, postural instability with eyes closed), biomarker testing (lipid profile, inflammation, insulin resistance, ApoE4 genotype suggesting poor response to statins), MRI (preservation of hippocampal and cortical volume, extensive white matter disease), and electroencephalography (executive dysfunction and evidence of vascular injury) that could be treated and supported a diagnosis of vascular cognitive impairment. A personalized treatment plan was then developed focusing on dietary counseling (Mediterranean-DASH Intervention for Neurodegenerative Delay diet, high-protein snacks, glycemic control); physical therapy for gait, balance, strengthening, and conditioning; referral to a personal trainer for aerobic, resistance, and flexibility training; mindfulness (yoga, meditation) for stress reduction; cognitive exercise focusing on problem-solving skills; omega-3 supplementation and possible resin therapy for cholesterol lowering; better blood pressure monitoring; and initiation of low-dose aspirin to improve blood flow. Longitudinal follow-up is needed to monitor adherence to recommendations and for evidence of improvement in outcomes. Such a trial could provide a direct estimate of individual treatment effects, fine-tune personalized care plans, enhance precision of future treatment decisions, improve person-centered outcomes, and if successful, reduce long-term healthcare costs.[25]


There is increasing evidence that multiple medical conditions increase the risk of neurodegeneration and subsequent development of dementia (Figure 1). It is also becoming clear that the majority of these risk factors act in amyloid- and tau-independent ways. Trials testing the amyloid hypothesis (β- and γ-secretase inhibitors, antiaggregation medications, mono- and polyclonal antibody approaches), antiinflammatory agents, and early-phase anti-tau therapies have failed to meet outcomes or have been discontinued because of safety concerns. While we wait for successful pharmacotherapy, these multiple pathways leading to AD can be taken advantage of to test hypotheses regarding risk reduction and mitigation. In all likelihood, efforts to prevent cognitive decline and development of dementia may be more successful when they are multimodal and directed to at-risk individuals based on their personal health profile, rather than using “one-size-fits-all” approaches. The detection of and interventions addressing root causes may offer novel approaches to diagnosing, treating, curing, or preventing AD. AD offers a large array of potentially modifiable risk factors (lifestyle, exposure, environment, comorbid disease) that are excellent targets to personalize the approach to medical care. Precision medicine approaches specifically target the heterogeneity of AD by identifying person-specific risk factors and applying a customized intervention directed against this risk profile. Even if these precision approaches do not cure or prevent AD, removing other pathways to neurodegeneration may greatly improve the likelihood that amyloid- or tau-specific therapies reach their endpoints. Perhaps it is time to abandon generalized approaches to AD and consider neurodegenerative disorders as diseases of a lifetime and that there may be individualized ways to build a better brain as we age.

Figure 1. Model of a dementia-prevention initiative. (A) Hypothetical model of the development of clinical dementia. Before diagnosis, there would be evidence of cognitive decline, which reflects neurodegenerative changes including cellular dysfunction and loss, synaptic dysfunction, and loss of connectivity. Presumably, accumulation of one or more pathologies cause these downstream changes. If more than one pathology is present, each should have its own risk factor or factors (e.g., Risk Factor A causes Pathology A, Risk Factor B causes Pathology B). (B) Application of model in N-of-1 trial (described in detail in text). Six risk factors were identified during the clinical evaluation, presumably working through different pathways, with the end result being neurodegeneration, cognitive decline, and if unchecked, eventually clinical dementia. A personalized prevention plan directed at root causes of impairment, if successful, would prevent the conversion of cognitive decline to dementia (marked by X). Because root causes may interact or potentiate each other’s effect on neurodegeneration (connecting arrows on the left side), multimodal approaches are more likely to have an effect than single approaches.




Conflict of Interest: Dr. Galvin serves as a scientific advisor for Axovant, Biogen, Eisai, and Eli Lilly; receives licensing fees from Pfizer, Lilly, Axovant, and Quintiles; and conducts on-going clinical trials funded by Biogen, Axovant, and Janssen. Dr. Galvin is funded by grants from NIH (R01 AG0402–11-A1, U01 NS100610, and R01 NS088040–01), the Florida Department of Health, the Harry T. Mangurian Foundation, and the Association for Frontotemporal Degeneration. He is on the editorial boards of Neurodegenerative Disease Management, Alzheimer’s Disease and Associated Disorders, and Acta Neuropathologica.

Author Contributions: Dr. Galvin was responsible for the study design, statistical analyses and interpretation, drafting, and revising and submitting the manuscript.

Sponsor’s Role: None.

© 2017, Copyright the Author Journal compilation © 2017, The American Geriatrics Society.