A call for comparative effectiveness research to learn whether routine clinical care decisions can protect from dementia and cognitive decline

Common diseases like diabetes, hypertension, and atrial fibrillation are probable risk factors for dementia, suggesting that their treatments may influence the risk and rate of cognitive and functional decline. Moreover, specific therapies and medications may affect long-term brain health through mechanisms that are independent of their primary indication. While surgery, benzodiazepines, and anti-cholinergic drugs may accelerate decline or even raise the risk of dementia, other medications act directly on the brain to potentially slow the pathology that underlies Alzheimer’s and other dementia. In other words, the functional and cognitive decline in vulnerable patients may be influenced by the choice of treatments for other medical conditions. Despite the importance of these questions, very little research is available. The Alzheimer’s Drug Discovery Foundation convened an advisory panel to discuss the existing evidence and to recommend strategies to accelerate the development of comparative effectiveness research on how choices in the clinical care of common chronic diseases may protect from cognitive decline and dementia.


Background
Cognitive impairment diagnosed as dementia, mild cognitive impairment (MCI), and/or mild neurocognitive disorder (mNCD) represents one of the most feared conditions in the United States and one of the most common reasons to enter a nursing home. In North America alone, the financial cost of dementia was estimated at $270 billion for 2015 with $61 billion in direct medical costs [1]. The burden from MCI is more difficult to quantify [2] but profound effects on quality of life, productivity, and health are likely [3,4]. Cognitive aging itself, while not a diagnosed medical condition, can influence quality of life and can be protected against through a variety of steps outlined in a recent Institute of Medicine report [5]. The consequences of cognitive decline pervade other aspects of health. For example, patients with cognitive impairment are less likely to comply with prescribed treatments and more likely to require hospitalization or experience treatment-related adverse events.
The factors that contribute to the risk and progression of MCI and dementia are varied and uncertain even for Alzheimer's disease, the most extensively researched cause of dementia [6]. Several different trajectories appear to lead to the clinical phenotype of Alzheimer's disease, with distinct molecular causes and risk factors as well as different patterns of progression [7] as reflected by the inability to generate a single risk prediction model for population-based settings [8]. Combinations of neuropathology are common and probably contribute synergistically to cognitive impairment (e.g. [9]). For many patients, the comorbidities and the corresponding medical care may contribute to their longterm risk of cognitive and functional decline. These factors, in turn, may provide opportunities for precision medicine to tailor patient treatment according to their risk profile.
In 2015, the Alzheimer's Drug Discovery Foundation convened an advisory panel on "The Prevention of Dementia and Mild Cognitive Impairment as Variables to Consider In the Comparative Effectiveness of Treatments for Common Chronic Conditions." The panel included representatives of the American Diabetes Association and the American Heart Association as well as experts from industry, academic and government research institutes, and the Patient-Centered Outcomes Research Institute (PCORI).
The consensus perspective was that clinical care of chronic diseases and comorbidities can likely influence cognitive decline, particularly in high-risk patients, but the existing evidence base for specific treatment choices is weak and conflicted. The meeting discussions were formulated into five recommendations for researchers, clinicians, and other stakeholders to advance research on how clinical treatment options may influence long-term cognitive decline. The views expressed are those of the authors and not their organizations.

Prevalent diseases are risk factors for cognitive decline and dementia
Several major chronic diseases are risk factors for dementia, linked by both epidemiology and biological rationale, suggesting that their clinical management might influence cognitive decline.
Hypertension in mid-life associates with a higher risk of dementia (reviewed in [10]). In the SYST-EUR trial, nitrendipine reduced the risk of Alzheimer's or vascular dementia by about 50 % [11]. In the PROGRESS trial, an angiotensin converting enzyme (ACE) inhibitor reduced the risk of cognitive decline by 19 % largely because of a reduced risk of recurrent stroke [12]. Other trials have not reported such benefits [13], likely because of differences in study design and patient population (reviewed in [10]). In some cases, the null result might have been due to outcome ascertainment bias such as in the SHEP trial, in which patients assigned to placebo were 24-60 % more likely to miss their outcome assessment [14].
Atrial fibrillation has also been associated with dementia. This relationship is independent of stroke (reviewed by [15]) and stronger with longer durations of atrial fibrillation [16]. A causative relationship is possible through repetitive microemboli or microbleeds but research is needed to show whether choices in the treatment of atrial fibrillation can influence dementia risk [15].
Type 2 diabetes mellitus (T2DM) is also associated with a higher risk of Alzheimer's (reviewed in [17,18]). Insulin resistance and glucose dysmetabolism in the brain may directly drive Alzheimer's neuropathology and increase the risk of cerebral infarcts that can in turn contribute to the manifestation of dementia (reviewed in [17]). Whether the clinical management of diabetes can influence cognitive decline is not yet certain. In the ACCORD-MIND trial, intensive versus standard glycemic control failed to protect against cognitive decline over 40 months but did result in a higher total brain volume [19].
Other diseases linked to cognitive decline include sleep-disordered breathing [20], heart failure [21], and chronic obstructive pulmonary disease [22]. More research is needed to establish whether these relationships are causal and, most importantly, if and how their clinical management can influence cognitive decline. Frailty and the accumulation of health deficits with old age [23] have also been linked to a higher risk of dementia, serving as an important reminder that biological vulnerabilities caused by aging underlie a suite of serious and common age-related ailments.
Specific medical treatments may influence the risk of cognitive decline Several common medical treatments are suspected to increase the risk of long-term cognitive decline. Cumulative anti-cholinergic drug burden has been associated with a dose-dependent increase of incident dementia as high as 54 % [24]. Similar risks have been reported for benzodiazepines and proton pump inhibitors [25]. Surgery can cause delirium or cognitive dysfunction in vulnerable patients which, in turn, associates with a higher risk of poor functional outcomes including dementia (reviewed in [26]). An ongoing controversy is whether surgery can cause persistent cognitive decline or simply unmask an underlying neurodegenerative illness (e.g. [27]). Even if the latter is true, earlier manifestation of progressive dementia is a significant concern.
While some treatments may increase the risk or the rate of cognitive decline, others may reduce it. Roughly 2 million cases of Alzheimer's disease have been attributed to physical inactivity, smoking, and mid-life obesity [28], suggesting that lifestyle modifications for cardiometabolic disease may profoundly protect the brain (reviewed in [29]). In clinical trials, cognitive function in the elderly has been improved through diet [30] or a multi-faceted intervention of lifestyle, diet, and vascular risk management [31]. These studies must be followed with longer and larger studies to evaluate effects on dementia risk.
There is intense interest in the question of whether specific drugs can be repurposed to treat or prevent neurodegenerative disease. Table 1 lists clinical trials that are underway to test whether specific drugs might reduce the risk or slow the progression of dementia or MCI through direct effects in the brain that are independent of their approved indications for treating high blood pressure or blood glucose. If these drugs do indeed slow neurodegeneration, a promising strategy to reduce dementia risk would be to manage hypertension or diabetes in high-risk patients with a clinically appropriate drug that has additional beneficial effects on the brain. Examples of potential treatment comparisons are described in Table 2.
The examples provided are not intended to comprehensively review the available data but rather to highlight the many ways through which common clinical care decisions might slow cognitive decline or reduce the risk of dementia. More research is needed to guide clinical care. Below, we recommend five general approaches to enable or bolster research that could create an evidence base to guide clinical care to mitigate the risk of cognitive decline or dementia in vulnerable patients.

Main text
Recommendation 1: researchers should utilize complementary study designs that incorporate patients at high risk for cognitive decline The study of real-world populations is central to CER yet particularly challenging for research on cognitive decline. Patients with advanced age or comorbidities, who are highly vulnerable to cognitive decline, are rarely included in research (e.g. [32]). Even in geriatrics, cognitively impaired patients are rarely recruited for clinical research on other conditions [33] and those that develop impairment during the trial are likely to drop out before their outcome can be assessed [14]. Most of the clinical research that has been done on other health conditions Examples of clinical trials underway to evaluate whether a drug approved for hypertension or diabetes could be repurposed to treat or prevent Alzheimer's disease or cognitive impairment. Many other trials have already been completed. In all cases, the putative mechanism of action involves a direct effect on the brain rather than an indirect effect through treatment of the primary indication. Other repurposing efforts are underway with drugs approved for depression, epilepsy, and erectile dysfunction. Very few studies are designed for CER, i.e. to compare the cognitive outcomes from treatments that are clinically equivalent for their currently approved indication ACE angiotensin converting enzyme inhibitor, ARB angiotensin receptor blocker, MCI mild cognitive impairment cannot therefore accurately inform how those treatments affect cognitive decline and dementia risk even if those outcomes were recorded. Another challenge is the heterogeneous nature of dementia, as described above. The treatments that reduce the risk of decline and the trials capable of detecting a benefit may succeed in one population but not another. To overcome these challenges, a combination of study designs with complementary strengths and weaknesses that include patients at high risk of cognitive decline is needed.
Pragmatic Phase III randomized controlled trials (RCTs) are most likely to influence clinical care but there are limits on our ability to fund and carry out multiple randomized head-to-head comparisons with sufficient power and follow-up. Short-term RCTs rarely have sufficient power to detect clinically meaningful change in cognitive decline and related function but shorter trials can use biomarker endpoints to validate the putative disease-modifying effects of treatments.
For example, a pilot trial reported that hypertension management with nilvadipine versus amlodipine might improve cerebral blood flow in patients with MCI despite similar effects on blood pressure [34]. Currently, a trial is underway at the Sunnybrook Research Institute in Canada to compare hypertension management with telmisartan versus perindopril in patients with comorbid Alzheimer's, looking at global brain atrophy over one year (Table 1). Although these biomarkers are not validated as surrogate markers, these exploratory trials can raise confidence for larger and longer trials on patient-centered outcomes.
Observational study designs are essential tools for comparative effectiveness research (CER) [35] and dementia prevention research [36,37] that provide windows into real-world heterogeneous patient populations. These studies are at high risk of confounding by indication and other bias but robust associations with cognitive decline can still inform hypotheses and guide clinical trial design. Polymorphisms in ACE have been linked to Alzheimer's disease [60] but whether central ACE inhibition will protect or harm is unclear. ACEi but not ARBs might accelerate beta-amyloid pathology by blunting ACE activity on non-angiotensin pathways and by inhibiting AT2 and AT4 receptors. One ARB in particular, telmisartan, has additional activity on PPAR gamma that might protect against neurodegeneration [61] A network meta-analysis concluded that ARBs had more benefit on cognition than ACEi drugs (adjusted effect size 0.47 +/-0.17, p = 0.04) [62]. Yet hypertension management with centrally acting versus non-centrally acting ACEi was associated with 25 % slower functional decline in Alzheimer's patients [63]. A Phase II trial is underway to compare telmisartan to perindopril in patients with comorbid hypertension and Alzheimer's ( Most DHP CCBs are likely to penetrate the brain except for amlodipine. Nilvadipine and nitrendipine but not amlodipine decreased beta-amyloid accumulation and blunted apoptosis in a mouse model of Alzheimer's. DHP CCBs varied in their capacity to increase amyloid clearance from the brain [64]. Effects on the brain may vary depending on their selectivity for different calcium channels [65] In a small trial in hypertensive patients with MCI, nilvadipine versus amlodipine slowed cognitive decline and improved cerebral blood flow despite similar effects on blood pressure [34]. Nitrendipine and nimodipine, have clinical data to suggest utility for the prevention or treatment of dementia, respectively, while nifedipine was associated with an increased risk of cognitive decline (reviewed in [64] GLP-1 agonists have been shown to protect against hippocampal synapse loss, lower beta-amyloid pathology and related damage, reduce neuroinflammation, and promote neurogenesis. While exenatide, liraglutude, and lixisenatide cross the blood-brain barrier, albiglutide and dulaglutide are large proteins unlikely to reach the brain [18] Treatment with liraglutide blocked decline in cerebral glucose metabolism over 6 months in Alzheimer's patients in a Phase II trial [66]. Additional trials are underway to repurpose these drugs to treat cognitive impairment ( There are several examples of the use of electronic health records and related databases for exploratory questions. The use of angiotensin receptor blockers (ARBs) compared to other cardiovascular drugs was associated with a 24 % lower risk of incident dementia in the US Veteran Affairs database [38]. The use of proton pump inhibitors was associated with an increased risk of all-cause dementia (HR 1.33; 95 % CI 1.04-1.83) and a 44 % increased risk of Alzheimer's disease from a German database on primary care patients (HR 1.44; 95 % CI 1.01-2.06) [25]. These exploratory associations should be followed up with additional observational studies to confirm the association and inform the design of RCTs.

Recommendation 2: incorporate cognitive assessment of high-risk individuals into routine clinical evaluations and electronic health records
In order to learn whether specific clinical care decisions influence cognitive decline, better recognition and reporting of cognitive function is needed in clinical settings particularly for patients at risk of cognitive decline because of comorbidities, frailty, age, genetics, or family history. Between 27 % and 81 % of cases of cognitive impairment are not currently recognized in primary care [39]. When a patient is diagnosed, they often have fairly advanced impairment with little if any objective data showing the trajectory of development of their impairment over time.
Some groups have recommended widespread screening for MCI and/or dementia (e.g. [39][40][41]) while others have called for research to prove that such screening improves patient outcomes [42]. While this debate continues, a parallel consideration is annual cognitive testing that informs clinical care even if it not used for diagnostic screening.
Objectively measured cognitive ability has been recommended as an important variable to inform the clinical care of geriatric patients with diseases like diabetes [43], cancer [44], and heart failure [45]. Routine cognitive evaluations can identify patients who might need additional medication reminders or other treatments for anxiety, depression, insomnia, polypharmacy, alcoholism, or drug abuse. They can also create a longitudinal record to help inform prognosis in later years if the patient or their family reports subjective cognitive impairment. Importantly for this discussion, regular cognitive evaluations can create a longitudinal record of cognitive assessments in real-world patient populations which could inform CER on which treatments associate with better or worse cognitive trajectories in which patients. A variety of neuropsychological tests are suitable for general practice, with strengths and weaknesses reviewed elsewhere [39,40].
Recommendation 3: in comparative effectiveness trials, incorporate outcomes on dementia incidence, cognitive decline, or neurodegeneration One strategy to increase the feasibility of CER on dementia risk or cognitive decline is to embed secondary outcomes in trials designed for other conditions. Several examples demonstrate that this can be done successfully. A sub-study of the ACCORD T2DM trial compared the effects of intensive versus standard blood glucose control on cognitive decline and brain atrophy [19]. The SPRINT trial, comparing intensive versus less intensive control of systolic blood pressure, was stopped early due to a significant reduction in the primary cardiovascular outcome [46]. For modest incremental cost, the trial also included assessment of all-cause dementia incidence, global cognitive decline, and MRI-measured changes in brain structure [47], with results yet to be reported.
A major concern with this strategy is outcome ascertainment bias [14]. Unlike major cardiovascular endpoints, cognitive outcomes cannot be readily captured from medical databases for patients who fail to return to the research clinic. Trial methods can be adapted, however, as achieved in the Ginkgo Evaluation of Memory Study [48]. Another issue is that trials may be stopped early because of emergence of clear differences in the primary outcome (e.g. [46]). Extended follow-up may be needed to ascertain differences in cognitive outcomes that may take longer to emerge.
More recently in the Alzheimer's disease field, there has been a change in emphasis to "slow progression" of the disease to "prevent" or "delay" the onset of the disease itself. This implies the need to rethink traditional study designs, including whether cognitive endpoints are the most suitable ones for such studies. As such studies tend to be more like epidemiological studies, "time to event" may be more suitable than a cognitive test, particularly since cognitive changes in a "prevention" setting tend to be small/subtle (much less the functional changes, which are even more subtle), though it is not yet clear which outcomes are best for such studies.

Recommendation 4: develop, validate, and standardize practical and acceptable methods for frequent assessment of cognition and function
Cognitive function varies on a daily, even hourly basis due to sleep impairment, alcohol or nicotine use, emotional or physical status, and other variables. Similar variation exists for quality of life and activities of daily living. This variability weakens the power to detect meaningful change with infrequent, intermittent assessments.
Neuropsychiatric testing deployed on consumerdevices has the potential to lower costs and increase the capacity of frequent assessment, particularly with data passively collected from routine patient behavior [49].
For example, a patient's conversations might be recorded and processed with automated speech-analysis software to detect the changes in verbal fluency and syntactic complexity that may be symptomatic of MCI (reviewed in [49]). An unobtrusive home-monitoring system can measure parameters of activity such as gait deficits, stride variability, and functional activities of daily living [50]. A different approach adopted by Akili Interactive Labs, Inc. is to design assays that entertain patients and thereby encourage more frequent assessment.
These technologies may provide statistical power that is not possible with current clinical instruments for activities of daily living or cognitive ability. For example, an estimated 80 % fewer patients would be required to detect a change in behavior using the unobtrusive home monitoring system mentioned above compared to conventional annual neuropsychological testing [51]. It may also provide the basis for collecting the "real-world" evidence being sought by the 21st Century Cures Act.
The rapid evolution of technology creates opportunities but also challenges. The ways in which data are collected, anonymized, and aggregated will need to be standardized before becoming widely adopted in clinical research. Developers will need to define and demonstrate the context of use for a given assessment tool (e.g. prognosis, diagnosis, rate of decline) and the applicability of the test to different populations (e.g. the influence of language, culture, and disease state). As with other types of ever-expanding data collected on individual citizens, careful attention will be needed to guard against risks to privacy, individual rights, and ethical abuse.
Recommendation 5: explore innovations in policy and funding to encourage healthcare providers and pharmaceutical companies to engage in comparative effectiveness research The scope of CER on cognitive decline will be severely limited without investment and participation by organizations outside of academic and government settings. In theory, a drug would have a market advantage if its use was shown to yield a lower risk of long-term cognitive decline. In practice, however, companies have little commercial incentive to pursue either CER [52] or repurposing of products for high-risk indications like dementia [53]. Policy changes may help to incentive industry, for example with an increased duration of data exclusivity after approval of a new indication [53]. Alternatively, the need for industry investment might be reduced by strategies to lower the cost of CER and repurposing, for example with preliminary or staged approval for a new indication followed by post-approval pharmacovigilance to confirm safety and efficacy [53].
Healthcare providers may similarly lack incentive to participate in CER, which limits the potential to understand real-world patient care using electronic medical records and related databases [52]. Providers need resources, reimbursement, and training to evaluate patient cognition and to understand the potential relationships between clinical care and cognitive decline. They also need compelling evidence that cognitive assessment improves patient care without slowing provider productivity or raising costs. Improved communication between researchers and frontline healthcare providers could also improve the clinical relevance of CER experimental design [54].

Conclusions
Over the past decade, CER has received more and more attention in the United States and elsewhere. Patients, families, healthcare providers, and payers increasingly ask which treatments have the best benefit versus harm ratio for a given patient. As this research advances, we recommend increased assessment of cognitive decline and the risk of dementia or MCI in CER studies. These outcomes are central to patient interests, quality of life, medical care, and societal costs. They are also likely to be influenced in profound ways by routine clinical care decisions.
Many challenges exist in the design, implementation, and funding of such research. Nevertheless, creative solutions exist and resources for CER (e.g. PCORI) may be leveraged to meet this need. Stakeholders across the spectrum of patients, providers, payers, researchers, and regulators can all play a role in accelerating the development of an evidence base so that clinicians and patients can learn whether the management of their existing conditions can influence the risk of cognitive decline and dementia.
CER in older people is a particular challenge because the vulnerabilities caused by aging often lead to comorbid chronic and acute diseases and ailments such as frailty that impair quality of life and function. Indeed, frailty and the accumulation of health deficits with old age have themselves been linked to a higher risk of dementia [23,55]. In these vulnerable populations, it can be challenging if not impossible to compare the full breadth of clinical benefit versus harm of diverse treatment options. Some therapeutics, however, may alter specific aspects of aging biology either through direct effects [56] or indirect effects via the management of comorbidities linked to accelerated aging (e.g. [57][58][59]). In order to assess which therapies have the best harm versus benefit ratio in older people, there is a strong need for defined and validated independent measures of elements of aging that affect morbidity that could be evaluated more efficiently than the incidence of numerous single clinical diseases. The rate of decline in certain domains of cognitive function (e.g. language, executive function) could be one of several components of such a screen for independent measures of morbidity associated with aging.

Funding
This position paper and the corresponding advisory panel were funded by the Alzheimer's Drug Discovery Foundation. Support for authors' time was provided by the Canadian Consortium on Neurodegeneration in Aging (JJA); the Intramural Research Program, National Institute on Aging, NIH (LJL); NIH-NIA grants K23 AG028726 and R01AG047975 (AV). Takeda Pharmaceuticals provided some support for the advisory panel; the American Heart Association supported the cost of travel for their representative (AV).
Availability of data and materials Not applicable.
Authors' contributions PAD drafted the manuscript based on discussions during the advisory panel. HMF, JJA, SKB, AJC, AMG, MSK, LRK, LHK, LJL, SL, EM, PJN, KR, DWS, RGS, PV, AV, RAW, and JDW read the manuscript, provided feedback and edits, and approved the final manuscript. For the advisory panel, whose discussions served the basis for this manuscript, HMF and PAD chaired and conceived of the panel while all panelists examined the agenda in advance of the meeting, provided feedback, prepared relevant material, and discussed the available evidence and potential recommendations during the panel.
Competing interests JDW, KR, and other authors receive federal or non-profit research funding for example from the NIH and CIHR. AV has served on the Data Safety Monitoring Board for Roche Pharmaceuticals. KR is the President and Chief Science Officer of DGI Clinical in Canada, which has contracts with Baxalta, Neurodyn, Nutricia, and Roche. KR has given lectures at Otsuka-Lundbeck and Nutricia and served as a consultant on methodology with AbbVie, Amgen, MeraJJA. MSK has research support from NIDDK, Novo Nordisk, and Bayer. PJN is on the strategic advisory board with Axovant and has consulted with Lundbeck. The authors do not declare non-financial competing interests.

Consent for publication
Not applicable.
Ethics approval and consent to participate Not applicable.