Precision prevention of Alzheimer's and other dementias: Anticipating future needs in the control of risk factors and implementation of disease‐modifying therapies

Empirical evidence suggests that a fair proportion of dementia cases are preventable, that some preventive actions can be taken immediately, and others may soon be implemented. Primary prevention may target cognitively normal persons with modifiable risk factors through lifestyle and multiple domain interventions (including general cardiovascular health). While the effect on individuals may be modest, it might have a large societal impact by decreasing overall dementia incidence by up to 35%. Secondary prevention will target cognitively normal persons at high risk of dementia due to Alzheimer's disease pathology with future anti‐amyloid, anti‐tau, or other drugs. This approach is likely to have major benefits to both individuals and society. Memory clinics will need structural and functional changes to adapt to novel technologies and increased patients’ demands, and brand‐new services may need to be developed with specific skills on risk profiling, risk communication, and personalized risk reduction plans.


INTRODUCTION
Few medical conditions raise as much controversy and as many contradictions among physicians, scientists, and the society at large as dementia. Dementia is a syndrome encompassing, among others, a number of neurodegenerative and cerebrovascular diseases, often presenting in combination, the most frequent of which is Alzheimer's disease (AD). The general mechanism of neurodegeneration has been identified as protein misfolding and aggregation followed by neurotoxicity; some of the molecular culprits have been identified (beta amyloid, hyper-phosphorylated tau, and other neurotoxic proteins) as well as their mode of diffusion to and into the brain. 1 Symptomatic drugs are available for dementia, and proved to be effective at the group level in a number of clinical trials. 2  This entails tailoring risk reduction to the clinical, genetic, and biological characteristics of each patient. Primary prevention aims to reduce disease incidence, either through addressing disease mechanisms or increasing resistance to disease, by targeting persons in the population at a time when they do not yet bear either disease markers or clinical impairment. Secondary prevention aims to detect and target clinically normal individuals with biomarker evidence of disease to delay or prevent symptom onset. Tertiary prevention aims to target patients with clinical impairment to reduce the impact of progressive symptomatic decline. In the following sections, we will address the rationale for primary and secondary dementia prevention with a precision medicine approach; the role of risk factors and their contribution to dementia; future scenarios of primary and secondary prevention of dementia; and foreseeable clinical, scientific, and organizational needs. Tertiary prevention included pharmacological and non-pharmacological interventions routinely carried out in memory clinics and will not be addressed in this article.

DEMENTIA PREVENTION IS ALREADY IN ACTION: THE SECULAR TREND TO AGE-SPECIFIC INCIDENCE REDUCTION
According to the findings of the world's largest survey on people's attitudes to dementia (almost 70,000 people across 155 countries and territories), summarized by the recently published "World Alzheimer Report 2019: Attitudes to dementia," even though 54% of people think that lifestyle plays a role in developing dementia, 25% of them think, paradoxically, that there is nothing we can do to prevent it. 4 Despite skepticism in the general public about the possibility of preventing and/or curing AD and other dementing disorders, 5 epidemiological observations suggest that dementia prevention has been taking place, albeit unintended, for the past 20 years. While the worldwide prevalence of dementia is expected to grow in the next 30 years by 92% in highand 176% in low-to-middle-income countries, epidemiological studies have shown decreasing incidence or age-specific prevalence rate of dementia in recent decades in Western countries. [6][7][8][9][10][11][12] In Bronx County, New York, 75-to 79-year-old persons born in 1915 had an incidence rate of dementia approximately 2.5 cases per 100 person/year, while persons of the same age born in 1935 had an incidence rate approximately 0.3 cases per 100 person/year 7 ( Figure 1). Consistently, the Cognitive Function and Ageing Study I and II reported a 20% drop in dementia incidence in the same geographical areas in the UK over two decades, from 1991 to 2011. 12 Moreover, a decreased prevalence of brain amyloid pathology, a key AD marker, has been found in 1599 brain autopsies from 1972 to 2006. 13 This phenomenon has not yet been observed in low-to-middle-income countries.
The determinants of the age-specific incidence reduction cannot be ascertained definitely in a post-hoc manner. However, it is likely that improvements in lifestyles (eg, more physical activity, longer formal education, healthier nutrition, and reduced smoking) and medical advances (eg, better control of vascular risk factors) in the general population may have played, and are still playing, a major role 6,11 perhaps via a strengthened neural reserve. Indeed, cerebrovascular F I G U R E 1 Dementia prevention in action: secular trend of decreased incidence of dementia as a function of date of birth and age in the Einstein Aging Study. Figure taken from: Derby et al. 7 disease, usually in the form of microangiopathy, often accompanies neurodegeneration and AD pathology, 14 and cerebrovascular health has recently been improved in the general population through a better control of cardiovascular risk factors.
These observations suggest that dementia prevention is not only possible but is already in action. While this beneficial effect is likely to have come as an unintended byproduct of the secular trend to greater wealth and healthier lifestyles in higher-income societies, the future incremental improvements required to counteract the demographic pressure of an aging world will need to be deliberated, planned, and equitable. Moreover, the rise in rates of obesity and diabetes in recent years may offset the gains made in existing cohorts entering the peak incidence period for dementia in the future. For such programs to be effective and efficient, sound knowledge is paramount about which risk factors are at work, at which stage of the life course, and their strength at both the individual and population levels.

RISK FACTORS AND THEIR CONTRIBUTION TO DEMENTIA
What constitutes a risk factor for a given disease depends on the definition of the disease itself. The definition and the very concept of AD as a disease are rapidly evolving. Consequently, what should be regarded as a risk factor for the disease is also changing. 5 For example, the most recent diagnostic criteria and research framework for AD stipulate that combined amyloidosis and tauopathy define the disease irrespective of cognitive symptoms and impairment, 15,16 while isolated amyloidosis is considered either a risk factor for AD (Preclinical Expert Consensus, 2016 15 ) or "Alzheimer's pathologic change" along the "Alzheimer's continuum" (National Institute on Aging-Alzheimer's Association, NIA-AA, Research Framework, 2018 16 ). While not challenging the NIA-AA's current conceptualization of AD as the association of brain amyloidosis and tauopathy, we believe that these conditions can also be regarded as strong risk factors for AD. 5 In 2018, the Lancet Commission on Dementia Prevention, Intervention, and Care estimated that approximately 35% of dementia is attributable to nine modifiable risk factors ( Figure 2). 17 Of these, five are general vascular disease risk factors (hypertension, obesity, smoking, physical inactivity, and diabetes), while four are more specific to dementia (low education, hearing loss, depression, and social isolation). While their prevalence varies from 3% to 40%, what is common to all is the relatively low relative risk for dementia, ranging between 1.4 and 1.9 ( Figure 2). Specifically, vascular risk factors have been vigorously tacked over the years because of their association with severe events other than dementia with significant impact on health (eg, stroke and myocardial infarction). On the contrary, psychosocial (or dementia-specific) risk factors are usually not associated with other severe events by themselves (except for depression, which may lead to suicide) and, thus, are considered less dangerous and have received less attention from a public health perspective. Nevertheless, we strongly recommend the study of psychosocial risk factors. This is further supported by the notion that dementia is the major adverse health outcome associated with these risk factors. Future personalized risk reduction protocols need to adopt a multidomain approach by targeting subject-specific risk factors, whether vascular or psychosocial.
Current evidence indicates that brain amyloidosis and tauopathy account for a large part of the remaining dementia risk (although the interplay among risk factors is still not clear and should be further elucidated). Indeed, brain amyloidosis with or without tauopathy and brain amyloidosis plus tauopathy are associated with a much greater risk for dementia (hazard ratio of 13.0 and 23.6, respectively; Figure 2 18 that might be the basis for F I G U R E 2 Risk factors for dementia and Alzheimer's disease and their corresponding population attributable fraction (PAF, the proportion of cases that might be spared by full control of the risk factor). PAF figures are unadjusted for communality (the variance in observed variables accounted for by common factors) for a fair comparison among risk factors based on available literature data. Risk for incident dementia: relative risk for lifestyle risk factors, subdistribution hazard ratios for biological, and hazard ratio for genetic risk factors. PAF = P(RR-1) / (1+P(RR-1)) where P is the prevalence of the risk factor in the population, and RR is the unadjusted relative risk for dementia associated with the risk factor. 17 See "Commentary", in the Supplementary material, for further information about PAF computation the development and implementation of evidence-based interventions.

PRIMARY PREVENTION ON MODIFIABLE VASCULAR AND NON-VASCULAR RISK FACTORS
To date, encouraging evidence suggests that a precision prevention approach (personalized prevention plans) can maximize the benefit of risk reduction programs based on lifestyle and pharmacologic interventions. When implemented, the impact at the societal level could be significant.

Lifestyle risk factors and multidomain interventions
Dementia is a syndrome often resulting from a combination of several factors, and recent evidence suggests that multidomain interventions should yield greater impact than interventions on individual factors.
To date, only four large multidomain trials for all-cause dementia prevention and three on general cardiovascular risk reductionhave been completed (Table 1;   cognitive domains was greater than that of the control intervention in apolipoprotein E (APOE) ε4 carriers but not in non-carriers. 26 In MAPT, the effect of the intervention was more beneficial in the participants with elevated Cardiovascular Risk Factors, Aging, and Incidence of Dementia (CAIDE) risk score or with positive amyloid positron emission tomography (PET) but not in those with low CAIDE risk score or negative amyloid PET 23 (Table 1 and Figure 3). SPRINT MIND showed a significant reduction of incident mild cognitive impairment (MCI; and in the composite outcome of MCI or probable dementia) cases in the group treated with a more "aggressive" antihypertensive strategy, 27 consistent with a previous clinical trial showing that low-dose statins reduce cognitive decline in patients with hypertension. 22 While other studies have shown the potential danger to the brain of too-low blood pressure, 28 it will be critical to identify patients who can benefit from aggressive blood pressure control from those for whom it will be detrimental based on individual features (eg, by taking age into account).
This points the way toward a precision prevention approach in dementia risk reduction, in which preventive interventions can be concen-  Table 1). The effect of intervention was significant in FINGER apolipoprotein E (APOE) ε4 carriers and MAPT amyloid-positive participants, but not in FINGER APOE ε4 non-carriers nor MAPT amyloid-negative participants. The interaction of intervention by APOE ε4 carrier status in FINGER was not statistically significant, while amyloid positivity in MAPT was significant. Bars represent 95% confidence intervals. NTB: neuropsychological test battery to be detected. Moreover, the implementation of a precision prevention approach requires the identification of a specific target population (eg, based on APOE genotype, and vascular and lifestyle risk factors).
Finally, while changing lifestyles in adults and older persons is already a significant challenge, maintaining healthy behaviors will be an even more challenging endeavor. Specifically, the FINGER and MAPT studies showed that participants' compliance to the intervention decreased with increasing complexity, and that some factors can enhance adherence (eg, face-to-face contact). 29 Moreover, health-care professionals will play a key role in this context. These are only some issues related to the study design of prevention trials, and future prospective randomized interventions-which are still strongly needed-should take them into account.

Synergies between current programs on vascular risk reduction and future programs on dementia
In 2011, a statement of the American Heart Association (AHA) on the contribution of cardiovascular risk factors to vascular cognitive impairment and AD dementia emphasized communalities that can be developed to improve the benefit of prevention programs. 30 More recently, the AHA recognized "optimal brain health" as the "foundation for a new strategic direction going forward in cardiovascular health promo-tion and disease prevention." 31 The opportunity for synergies between dementia prevention and cardiovascular prevention programs is obvious.
Cardiovascular disease prevention programs are ongoing in most industrialized countries, usually established by national health organizations, such as the National Health Service in the UK and the AHA in the United States. The experience gained in more than 50 years of cardiovascular prevention programs as well as some conceptual analogies and practical overlaps can be used today in the planning of dementia prevention action plans ( Table 2). Dementia prevention programs may more generally benefit from the experience and infrastructure of cardiovascular action plans that may run on a statewide scale, with comparable target population and at a lower cost.
A risk of integrating dementia with cardiovascular prevention programs is to blur the information that the patient receives. In persons with low health literacy and initial cognitive impairment, this might raise confusion and adversely affect compliance to prevention interventions. Information delivered to the population on cardiovascular health via television, newspaper advertisements, and general practitioners will need to be re-tuned to include clear and unequivocal dementia-specific messages. The involvement of current memory clinics in educational programs will ensure harmonization to the latest scientific advancements. While using "dementia prevention" as the hook to modify behavior in, for example, mid-life may be unlikely to prove relevant to people so many years from dementia onset, catching TA B L E 2 Analogies between prevention targets for coronary heart disease and Alzheimer's disease

Primary prevention Secondary prevention
Pathophysiology

Coronary heart disease
Induction: interaction between genetic and traditional modifiable risk factors (eg, hypercholesterolemia, hypertension, diabetes) leading to inflammation of the artery wall.

Latency: atherosclerosis
Alzheimer's disease Induction: interaction between genetic factors and environmental exposures to promote Aβ42 aggregation and tau phosphorylation.
Deposition of Aβ42 in cortical plaques and/or deposition of hyper-phosphorylated tau in neurofibrillary tangles.

Target population Coronary heart disease
Myocardial infarction-free and atherosclerosis-free patients.
High-risk persons (older age, multiple cardiovascular risk factors)

Alzheimer's disease
Cognitively normal persons at high risk (APOE ε4 carriers, family history) who are amyloid and/or tau negative, and neurodegeneration negative.
Cognitively normal persons who are amyloid and/or tau positive.

Coronary heart disease
Screening for cardio-vascular risk factors (hypertension, diabetes, obesity, smoking, physical inactivity, etc.), estimation of individual cumulative risk estimate, risk-lowering interventions to prevent atherosclerosis.
If symptomatic, identification of atherosclerosis and cardiac dysfunction on echocardiography, stress scintigraphy, cardiac magnetic resonance imaging, carotid ultrasound.

Alzheimer's disease
Screening for vascular and dementia-specific risk factors (hearing loss, depression, social isolation).
Interventions on non-pathophysiologic and specific risk factors. Interventions to increase resilience to pathology (eg, specific nutrients).
Detection of brain amyloidosis and brain tauopathy. Amyloid and tau lowering agents to prevent cognitive impairment.
Interventions to increase resistance to pathology.
message tags such as the one recently suggested by the European Academy of Neurology might be instrumental ("what is good for the heart is good for the brain"). 32

SECONDARY PREVENTION WITH DISEASE MODIFYING THERAPIES IN COGNITIVELY NORMAL PERSONS WITH AD PATHOLOGY
The failure of DMT trials on patients with MCI or dementia does not invalidate amyloid as a treatment target. Current DMT trials implicitly assume that amyloid is a deterministic cause of neurodegeneration in the context of the amyloid cascade hypothesis (Figure 2). However, if brain amyloid is rather a probabilistic risk factor, removing brain amyloid when neurodegeneration is already established (as is the case of mild to moderate AD dementia) is unlikely to affect it, as well as, by analogy, treating hypercholesterolemia in patients with stroke may only marginally alter its natural history. In the probabilistic risk factor scenario, prevention should take place when neurodegeneration is absent or very mild. Indeed, trials have been designed and implemented in CN persons at high risk of incident cognitive impairment and dementia due to AD pathology. In 2019, there were six DMTs in phase III trials on CN participants, 3 and their results are eagerly awaited. However, if hopes for efficacy are fulfilled, clinical and logistic challenges will be paramount.
More persons now seek advice in memory clinics complaining of declining cognitive performance although scoring normal on formal cognitive testing ("subjective cognitive decline"). Others do not report cognitive problems and are just worried about future cognitive decline due to family history, or simply concerned about preserving their memory and general cognitive abilities (sometimes referred to as "worried well" patients). This group of patients is enriched with some high-risk individuals 33 and accounts for 20% to 30% 34,35 of all new patients seeking help in memory clinics. The availability of DMTs will inevitably lead to more of these individuals seeking medical advice.
Current memory clinics are designed for the biopsychosocial needs of people with cognitive impairment and variable degrees of functional limitation that are likely to lead to a diagnosis of dementia but were never designed and are ill-equipped to deal with this new population of patients presenting with different concerns, requests, expectations, and hopes.
A new generation of brain health services (variably referred to as dementia prevention services, brain health clinics, dementia prevention clinics, etc.) is being developed along with specific expertise, working tools, and organizational models. Early pilot experiences of brain health services, so far in the research and development stage, are ongoing in Barcelona, Edinburgh, Cologne, and Geneva. To a greater or lesser degree, they adhere to three guiding principles: (1) risk profiling, (2) risk communication, and (3) implementation of personalized prevention plans. Their current health offer targets non-demented persons (CN and MCI) and includes a personalized approach of disclosing dementia risk estimates and precision risk control programs in the context of phase II and III clinical trials of experimental DMTs. This approach is a paradigm shift from the current diagnostic approach in memory clinics, where, for example, APOE genotyping is not recommended exactly for the reason that, being a risk factor, its diagnostic sensitivity and specificity are relatively poor. 36 Indeed, pilot evidence suggests that APOE genotyping might be a critical variable in the context of brain health services to estimate the personalized risk for dementia ( Figure 2) and because it was shown to mediate the effect of risk reduction interventions (subgroups analysis of the FIN-GER trial [see Figure 3], in which the effect of intervention was significant in APOE ε4 carriers but not in non-carriers). However, further research and ad hoc clinical trials selecting the target population based on APOE genotype are needed to prove and support its potential utility, and eventually recommend APOE testing in brain health services. The approach has also been expressed clearly in the UK-led Edinburgh Consensus on preparing for the advent of disease-modifying therapies for AD, which envisaged the establishment of new brain health services to complement existing memory clinics. In brain health services, disease detection, risk profiling, and prevention of dementia would be the overriding objective. 37 However, a number of challenges will need to be met before the health offer of these and similar brain health services can enter the production stage ( Table 3).
The first challenge will be efficient screening of persons at high risk. Performing amyloid PET, tau PET, and lumbar puncture in all persons would clearly be prohibitively expensive. Advances in blood-based biomarkers of amyloidosis, 38 tauopathy, 39 and neurodegeneration 40 promise to be of great value, and others based on retinal imaging and saliva are being studied. If peripheral biomarkers prove to be reliable and effective, a large-scale screening for neurodegenerative diseases will be possible and the general population will be reached, 41 and those at risk for dementia (ie, biomarker positive) will ultimately be addressed to brain health services. Large longitudinal studies will be needed to accurately estimate the effect on screening efficiency of key risk modifiers such as age, family history, and APOE genotype.
To accurately estimate the risk for AD dementia of those screening positive, large population-based studies including imaging and cerebrospinal fluid biomarkers, genetics, and accounting for "traditional" risk factors will be required. In due course, protocols that can be delivered at scale, cost, and with a high degree of precision will be developed.
Risk communication will be burdened with the known challenges of communicating the very concept of risk to the general public. New tools, protocols, and skills for structured risk disclosure will be needed, such as the one developed by the A4 trial for amyloid PET. 42 Special attention will need to be devoted to matching messaging and disclosure to people from a variety of cultural, socio-economic, and societal backgrounds where belief systems and expectations may vary substantially.
In the clinical pipeline of brain health services, risk profiling and communications will be followed by individualized risk-reduction interventions (aka personalized prevention plans). Lifestyle and pharmacologic interventions may be prioritized based on the individual cumulative risk of dementia (so-called risk-stratified care management), analogous to what is currently done for the treatment of hypertension, diabetes, and hypercholesterolemia. In times when effective interventions directed to pathophysiological risk factors are lacking, high-risk, pathophysiological-biomarker-positive persons will be referred to prevention clinical trials with putative DMT (eg, via registries 43 ). In these TA B L E 3 Preparatory actions in view of establishing and running brain health services

Screening
Finalizing analytical validation and carrying out clinical validation of blood-based and other peripheral biomarkers. Estimating the effect of age, family history, and APOE on false positive and negative rates of biomarkers.
patients, the uncertainty of treatment group allocation (active drug or placebo) and drug effectiveness may suggest a proactive approach to can also be partially integrated to existing prevention programs of other chronic diseases such as cardiovascular diseases, stroke, and diabetes given that these disorders share several common risk factors. Non-conventional preventive interventions targeting more innovative pathophysiological hypotheses of neurodegeneration may in due course be integrated into brain health services once shown effective; such approaches might include novel drugs, behavioral modifications, neurostimulation, and nutritional principles. 47,48 Ethical issues will arise that require being specifically addressed such as ensuring informative and respectful communication of risk and ensuring that disclosure of risk does not have adverse consequences, for example, for insurance. Educational activities will need to be set up in the general population, general medical practices, and current memory clinics. The financially sustainable business model that will be uptaken for brain health services will largely depend on the availability of approved new drugs. Registries for subjects at very high risk will allow us to set up a point-of-care registry system to ensure coordination of care. 49 Finally, the research community should strictly recommend the adoption of rigorous methods and the implementation of evidence-based prevention plans only. At the same time, the research community should be able to educate the general population and therefore avoid the proliferation of pseudo-medicine services 5,50 taking advantage of people's concerns.

CONCLUSIONS
Dementia and neurodegenerative diseases research and care is in a dynamic state of evolution and in need of increased synergy among public health, general practice, and specialist care. Independent of the availability of DMTs, precision dementia prevention and personalized care should liaise with resources in the vascular prevention field in public health programs and in general practice. If disease modifiers prove effective to delay adverse outcomes in CN persons, a new model of brain health services may need to be developed, which will encounter a number of novel clinical, ethical, and organizational challenges.
While much evidence still needs to be collected before scientifically sound brain health services can be launched in production mode, we believe that it is urgent to set up research brain health services that may pilot the model and start designing and testing the weapons that society needs in order to fight the battle against the rising dementia prevalence.