Investigation of frailty as a moderator of the relationship between neuropathology and dementia in Alzheimer’s disease: Findings from the Rush Memory and Aging Project cohort study

Abstract

Background:

The neuropathological features of Alzheimer’s disease (AD) do not always correlate well with its clinical presentation; some people who show a high burden of AD pathology at autopsy demonstrate few characteristic clinical symptoms or signs, whereas others with very little AD pathology manifest Alzheimer’s dementia. Given how strongly dementia is associated with age, it is possible that frailty, which is associated with both age and dementia, impacts how individuals tolerate AD pathology.

Objective:

To examine whether frailty moderates the relationship between AD pathology and Alzheimer’s dementia.

Methods:

This was a cross-sectional analysis of data from the Rush Memory and Aging Project, a clinical-pathologic cohort study of older adults living in Illinois, USA. Participants underwent annual neuropsychological and clinical evaluations. We included 456 participants (mean age 89∙7±6∙1, 69∙3% females) who died and had autopsy. AD pathology was quantified by a summary measure of neurofibrillary tangles, neuritic and diffuse plaques. Clinical diagnosis of Alzheimer’s dementia was based on clinician consensus. Frailty was operationalized using the deficit accumulation approach (41-item frailty index). Logistic regression and moderation modeling assessed relationships between AD pathology, frailty and Alzheimer’s dementia.

Findings:

Frailty and AD pathology were independently associated with Alzheimer’s dementia (OR=1∙76, 95% CI 1∙54–2∙02, p<0∙0001; OR=4∙81, CI 3∙31–7∙01, p<0∙0001, respectively), adjusting for age, sex, and education. When frailty was added to the model with AD pathology, model fit improved (chi²(1)=86, p<0∙001). Frailty interacted with AD pathology (OR=0∙73, CI 0.57–0.94, p=0∙015); people with higher frailty had a weaker relationship between AD pathology and Alzheimer’s dementia.

Interpretation:

The degree of age-related deficit accumulation improved the relationships between AD pathology and Alzheimer’s dementia. That frailty is related to both odds of Alzheimer’s dementia and disease expression has implications for clinical management. Further research should assess trajectories of change in frailty and cognition to better elucidate this complex relationship.

INTRODUCTION

The neuropathological features of Alzheimer’s disease (AD; including amyloid-based plaques and neurofibrillary tangles) do not correlate well with the clinical expression of dementia (cognitive and functional decline)¹. Many people with a high burden of AD pathology at autopsy had few characteristic clinical symptoms or signs, whereas others with little AD pathology suffered from Alzheimer’s dementia. Together, this suggests that some latent factor influences who can ‘tolerate’ higher levels of AD pathology without suffering from dementia and who is more vulnerable to pathology. While some people who develop Alzheimer’s dementia are young and otherwise healthy (typically with familial AD), the vast majority of people who develop Alzheimer’s dementia are older with multiple other health problems. People with multiple, age-related health deficits are often considered frail.² Frailty may help explain the relationship between AD pathology and the clinical expression of dementia.

Frailty represents a state of decreased physiological reserve and increased vulnerability to adverse health outcomes (including hospitalization and death) compared to others of the same age.³ Frailty has been most frequently operationalized using a deficit accumulation approach or a phenotypic/syndromic approach.^4,5 Unfortunately, people who are frail are routinely excluded from clinical trials testing treatments aimed at improving their condition⁶.

Frailty is related to neuropathological features of AD as well as cognitive decline and dementia.^7–9 Frailty and Alzheimer’s dementia share many risk factors and clinical features, including age, inflammation, functional impairment, and atypical illness presentation.⁸ The link between frailty and Alzheimer’s dementia has been demonstrated both in clinical and epidemiological settings, but less so using neuropathological studies. To date, such studies assessing AD pathology have been restricted to the syndromic/phenotypic definition of frailty and none have examined the moderating effect of frailty.^9,10

We hypothesized frailty is a latent factor that moderates clinical dementia expression in relation to AD pathology. By reducing an individual’s ability to tolerate AD pathology, frailty could precipitate dementia disease expression where it might have remained asymptomatic in a non-frail individual. Our objective was to examine the impact of frailty on the relationship between AD pathology and expression of Alzheimer’s dementia.

METHODS

Overview

Data come from the Rush Memory and Aging Project (MAP), which has been described elsewhere.¹¹ MAP is a clinical-pathologic cohort study enrolling over 2100 older persons without dementia from about 40 residential facilities, senior and subsidized housing, church groups, and social service agencies in Northeastern Illinois, beginning in 1997. It aimed to identify factors associated with maintenance of cognitive health despite the accumulation of neuropathologic lesions. This cohort has been followed for 21 years with data collected annually, via home visits. Participants provided blood samples and underwent detailed clinical evaluation, including cognitive assessments. All participants signed the Anatomical Gift Act, agreeing to donate their brain, spinal cord, and other biospecimens at death (n=675). MAP was selected for our project as it is unique in combining comprehensive clinical and cognitive assessments and essential neuropathological characterization, with sufficient data to allow for operationalization of a frailty index (FI).

Dependent variables- Dementia status

Clinical diagnosis of cognitive status done at each annual assessment comprised three-steps: 1) computer scoring of a neuropsychological battery including 19 instruments (e.g. word list recall, category fluency, digit ordering, Stroop word reading);^12,13 2) clinical judgment by a neuropsychologist blinded to participant demographics; 3) diagnostic classification by a clinician (neurologist, geriatrician, geriatric nurse practitioner, or neuropsychologist).¹⁴ Clinical diagnoses of Alzheimer’s dementia are based on criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS/ADRDA) joint working group.¹⁰ For analyses presented here, people diagnosed with mild cognitive impairment or other forms of dementia were excluded; our dementia status variable refers to only Alzheimer’s dementia or no dementia.

Independent variables –Frailty Index (FI)

The FI is a health state measure designed to integrate multiple types of health information. The FI reflects vulnerability to adverse outcomes and proximity to death ³. It reliably predicts adverse outcomes in large health databases from many countries.³ An FI can be created using different routinely collected clinical information such as symptoms, signs, disabilities and diseases that meet standard criteria.¹⁵ The Frailty Index (FI) = (number of health deficits present) / (number of health variables measured). For example, a person with 10 of 40 potential deficits measured has an FI score of 10/40 = 0∙25. The FI takes advantage of the human organism’s high redundancy, making it replicable across different databases, even when different items, and numbers of items, are used.³

We created an FI from health variables obtained at each clinical evaluation. Candidate variables included symptoms, signs, comorbidities, and function. Variables were screened for inclusion based on standard procedures;¹⁵ variables strongly related to the outcome (cognitive variables) were excluded. Our 41-item FI can be found in Appendix B. Higher values of the FI indicate poorer health. For the purpose of interpretation, the FI value was multiplied by 10 so that odds ratios would represent the proportion change for an increased increment of 0∙1 of the FI (an additional 3–4 deficits).

Independent variables –neuropathological data

MAP includes data from post-mortem neuropathologic evaluation. Global AD pathology burden was quantified and described elsewhere.^9,11,13 It is a quantitative summary of three AD pathologies (neuritic plaques, diffuse plaques, and neurofibrillary tangles), as determined by microscopic examination of Bielschowsky silver-stained slides from five regions (midfrontal, midtemporal, inferior parietal, and entorhinal cortices, and the hippocampus) resulting in 15 separate counts. A summary score for each of the three pathologic findings is obtained by dividing the count for each index in each region by the corresponding standard deviation and then averaging the five scaled regional measures to obtain a scaled mean. The three scaled means are then averaged and the square root is calculated to obtain the measure of AD pathology.¹⁶ Additional information can be found in Appendix E. Higher values on the AD pathology measure indicate greater burden of pathology. For sensitivity analyses, counts of the three individual pathological measures were also used, as were mean percent volume of amyloid calculated over eight regions (hippocampus, entorhinal cortex, midfrontal cortex, inferior temporal, angular gyrus, calcarine cortex, anterior cingulate, superior frontal cortex), and Braak staging (a score between 1 and 6 of tangle pathology burden and spread), where higher scores are worse.¹⁷

Independent variables –confounding variables

Age, sex, education, and APOE status were evaluated as confounders with respect to the outcome. Age and education were measured in years, whereas sex was binary. Age was calculated from birth to date of cognitive assessment. Sex and education were self-reported. APOE status (presence of ≥1 e4 allele) was obtained via genotyping of brain tissue using high-throughput sequencing of codons 112 (position 3937) and 158 (4075) of exon 4 on the APOE gene on chromosome 19.¹⁸

Statistical analysis

We investigated characteristics and validated the FI using a standard approach including descriptive statistics and appropriate statistical tests. For the purpose of describing our sample, we categorized participants into frailty groups using the median (FI=0∙41) as the cut-point; this cut-point is very close to that typically used to distinguish between moderately and severely frail people.¹⁹

We used logistic regression to examine the relationship between neuropathological markers (exposure, continuous variable), FI (exposure, continuous variable), and dementia status (outcome, binary variable) . We then assessed whether the FI improved model fit when added to an adjusted model including AD pathology, using Aikike Information Criterion (AIC), Bayesian Information Criterion (BIC), and deviance (−2 log likelihood).

We built moderated logistic regression models (Process syntax²⁰) using 5000 bootstrapped samples and bias-corrected confidence intervals. We used this to report the relationship between AD pathology and Alzheimer’s dementia at three different levels of frailty: the mean FI value (“intermediate frailty”), one standard deviation above (“high frailty”), and one standard deviation below (“low frailty”).

All analyses were adjusted for age, sex, and education. Sensitivity analyses examining measures including a modified frailty phenotype, Braak stage, amyloid burden (calculated as percent volume), depression (as measured by the Centre for Epidemiologic Studies 10-item Depression scale), and polypharmacy (as measured by the sum of prescription medications) were conducted. Details on these measures can be found elsewhere.¹¹ All analyses were conducted using SPSS 24.

This project was indirectly funded through various sources supporting individual authors. Funding sources played no role in the design, analysis, or interpretation of the findings.

RESULTS

Since its inception 21 years ago, 3869 persons expressed interest, of those 2112 (54.9%) enrolled in MAP and 193 (9.1%) participants have discontinued study participation, reflecting an average loss to follow-up of ~1%. The dataset frozen January 2017 included 1843 of enrolled participants, of whom 831 had died, and 675 had complete autopsy data. Of those with autopsy records, 456 had either no dementia or Alzheimer’s dementia, and had enough information to create a frailty index and were therefore included in this analysis (Appendix A). Average time from last assessment to death was 0∙89 years (median 0∙61), average age at death was 89∙7 years (standard deviation [SD] 6∙1), and most were women (69∙3%). Over half had a diagnosis of possible or probable Alzheimer’s dementia at their last clinical assessment (53∙1%; n=242). The mean FI for the whole sample was 0∙42 (SD 0∙18), with a median of 0∙41 and range of 0∙04–0∙91. The 95^th and 99^th percentile was 0∙71 and 0∙81, respectively. The FI had a characteristic skewed distribution with a long right tail. People who had high FI scores (FI>0∙41; n=223) were older, had lower MMSE scores, more likely to have a dementia diagnosis, and higher Braak stage (Table 1).

Table 1.

Descriptive characteristics of the sample

	All (n=456)	Frailty index <0∙41 (n=233)	Frailty index ≥ 0∙41 (n=223)
Age at baseline (mean±SD)	83∙1±5∙9	82∙1±5∙8	84∙2±5∙8*
Age at death (mean±SD)	89∙7±6∙1	88∙3±6∙2	91∙2±5∙6*
Sex (% female)	69∙3	64∙4	74∙4
Years of education (mean±SD)	14∙4±2∙9	14∙5±3∙0	14∙3±2∙9
Frailty index (last assessment before death; mean±SD)	0∙42±0∙18	0∙28±0∙09	0∙58±0∙10*
AD-type dementia diagnosed before death (%)	53∙1	33∙9	73∙1*
MMSE (last assessment before death; mean±SD)	19∙8±9∙8	23∙4±8∙2	16∙1±10∙0*
Braak stage (mean±SD)	3∙7±1∙2	3∙4±1∙2	3∙9±1∙1*
APOE genotype (≥1 e4 alleles; %)	23∙2	20∙6	26∙0
10-item CES-D (mean±SD)	2∙0±2∙2	1∙4±1∙8	2∙6±2∙5**
Number of medications (mean±SD)	1∙9±1∙1	1∙7±1∙1	2∙0±1∙2

^*p<0∙05,

^**p<0∙001

SD=Standard deviation; AD=Alzheimer’s Disease; MMSE=Mini Mental State Examination; CES-D=Centre for Epidemiologic Studies Depression Scale

Importantly, 35 people (7%) demonstrated a high burden of AD pathology without having been diagnosed with dementia and 50 people (10%) had Alzheimer’s dementia, but demonstrated a low burden of AD pathology (Table 2). Therefore, for about one person in six the relationship between AD pathology and dementia was weak. The mean FI was significantly higher for people with Alzheimer’s dementia compared to those without, however the mean FI was highest among people with Alzheimer’s dementia with a low burden of AD pathology (Table 2). Among people with a low burden of AD pathology, the prevalence of Alzheimer’s dementia was much higher in those who had high frailty than those with low frailty (69∙0% vs. 5∙3%; Table 3).

Table 2.

Mean frailty index values (mean±standard deviation) by neuropathological burden and dementia status.

	Low burden of pathology	Intermediate burden of pathology	High burden of pathology
No AD	0∙33±0∙14 (n=102)	0∙36±0∙15 (n=77)	0∙35±0∙15 (n=35)	F=1∙01, p=0∙37
AD	0∙55±0∙15 (n=50)	0∙50±0∙18 (n=75)	0∙46±0∙18 (n=117)	F=4∙54, p=0∙012
	F=82∙8, p<0∙001	F=27∙17, p<0∙001	F=10∙88, p<0∙001

Table 3.

Sample size and percent of sample with Alzheimer’s disease by frailty index values and neuropathology burden.

		Frailty index
		Low	Med	High	Total
Neuropathology burden	Low	n=56; 5∙3%	n=54; 33∙3%	n=42; 69∙0%	n=152; 32∙9%
	Med	n=54; 29∙6%	n=48; 45∙8%	n=50; 74∙0%	n=152; 49∙3%
	High	n=45; 66∙6%	n=53;73∙6%	n=54; 88∙8%	n=152; 77∙0%
	Total	n=155; 31∙6%	n=155; 51∙0%	n=146; 78∙1%	n=456; 53∙1%

FI scores and AD pathology were independently associated with dementia status (Odds Ratio [OR] =1∙76, 95% CI 1∙54–2∙02, p<0∙0001; and OR=4∙81, 95% CI 3∙31–7∙01, p<0∙0001, respectively), after adjusting for age, sex, and education. When the FI was added to a model with AD pathology, the model fit improved according to the reduction in AIC (533∙48 to 449∙41), BIC (554∙10 to 474∙15), and deviance (523∙48 to 437∙41), which demonstrated a significant improvement (chi²(1)=86, p<0∙001). Further, there was a significant interaction between FI and AD pathology (OR=0∙73, 95% CI 0.57–0.94, p=0∙015).

Moderation analyses demonstrated that the relationship between AD pathology and dementia status changed over levels of frailty, as increasing frailty weakened the relationship (Figure 1).

Figure 1.

Conditional effect of AD pathology on Alzheimer’s dementia status at values of the moderator (frailty), adjusted for age, sex, and education; n=456.

Sensitivity analyses were undertaken to determine whether the relationship was being driven by the type of AD pathology. Only neuritic plaques demonstrated a significant interaction with the FI in predicting dementia status (OR=0∙80, 95% CI 0.66–0.96, p=0∙019). Moderation analyses suggested that with increasing frailty, the relationship between neuritic plaques and dementia status weakened (consistent with our original results). We also explored the effects of amyloid calculated as percent volume occupied (rather than plaque counts) and found that the relationship was consistent (i.e. significant interaction between amyloid and FI: OR=0∙96, 95% CI 0.93–0.996, p=0∙027). Braak staging also significantly interacted with the FI (OR=0∙80, 95% CI 0∙68–0∙93, p=0∙004) in relation to dementia status, and moderation analyses suggested that with increasing frailty the relationship between Braak stage and dementia status weakened (consistnet with original results). We investigated whether including people with MCI in the reference group (i.e. no dementia) would alter the results and found that while the frailty index and AD pathology remained significant independent predictors of Alzheimer’s dementia, their interaction was non-significant. To be sure that the results were not being driven by activities of daily living variables, we recreated the FI excluding these variables (those indicated with a star in Appendix B; n=14), and found similar results. We also controlled for the effects of possible risk factors, including self-reported history of stroke, hypertension, diabetes, congestive heart failure (CHF), and depression (as measured by the CES-D) and found that while stroke and CHR were significantly associated with dementia status, the frailty*pathology interaction remained significant. When the frailty phenotype (z-score; as detailed elsewhere⁹) was used in place of the FI, it was found to significantly predict the outcome, but with no significant interaction with AD pathology.

DISCUSSION

Three main conclusions can be drawn from our results. First, people with Alzheimer’s dementia but a low burden of AD pathology have the highest frailty levels, suggesting that frailty is implicated in dementia expression in this group. Its effect could arise by reducing the threshold of AD pathology needed to cause clinical disease, or as a marker of impaired repair processes that might otherwise allow for AD pathology to be better tolerated ². Further, frailty levels were no higher than average in people with no Alzheimer’s dementia but a high burden of AD pathology. Together, these results suggest that the frailer an individual is, the less likely that person is able to tolerate a given burden of AD pathology. Second, frailty improves the fit of a model relating AD pathology with dementia status. This is likely because frailty is able to account for several diverse causal pathways in older adults, in whom dementia is multiply determined;^2,21–23 Third, frailty is a significant moderator in the relationship between AD pathology and dementia status; increasing frailty weakens the relationship between AD pathology and dementia.

Together, these findings support the idea that frailty influences the clinical expression of dementia. While frailty may reduce the threshold for AD pathology to cause cognitive decline, frailty is likely also contributing to other mechanisms in the body that give rise to dementia, weakening the direct link between AD pathology and dementia. This suggests that frailty should be considered in clinical care and management. It supports the view of late-life dementia as a multiply determined state, with many factors contributing to its development.

Our results are consistent with existing literature. Dementia is highly linked with aging,²⁴ yet the passage of time alone cannot cause mechanistic failures or account for heterogeneity in health status among people of the same age. Failing mechanisms manifested by age-related signs, symptoms, and diseases can be informative and the FI is able to package this information into a single value that represents health status, reducing dimensionality). Studies are beginning to suggest that ‘age-related’ disease are actually more frailty-related.²⁵

Frailty has also been linked with both cognition and dementia status, cross-sectionally and longitudinally, whether frailty is measured by FI or phenotype.^7,22,26 The construct of cognitive frailty remains debated; some researchers view it as a clinical entity with cognitive impairment related to physical causes and potentially a target for intervention in early or preclinical dementia²⁷. A previous analysis of the dataset used here revealed that higher baseline frailty (as measured by a modified frailty phenotype) and change in frailty increased risk for incident Alzheimer’s dementia at 3-years. Higher baseline frailty was also associated with lower baseline cognition, and faster cognitive decline.¹² Change in frailty has been associated with change in cognition.^13,28

We build on work relating frailty to AD pathology in two ways. First, we show that this relationship still holds with a broadly construed approach to frailty captured by deficit accumulation. Second, we show that the degree of frailty helps explain the circumstances under which AD pathologic markers and dementia are less well correlated; people with a low degree of frailty were better able to tolerate AD pathology, whereas those with higher degrees of frailty were more likely both to have more AD pathology and for it to be expressed as dementia. This finding was robust; it held when controlled for vascular factors and when functional deficits were excluded from the FI.

In a recent scoping review, we synthesized evidence from existing studies that measured both AD biomarkers (such as in-vivo protein abnormalities, MRI abnormalities, or post-mortem plaque and tangle pathology) and frailty²³. Ten studies were identified, of which eight reported direct relationships between biomarkers and frailty. All of these eight studies demonstrated a positive relationship between biomarkers and frailty (regardless of measurement) suggesting that frailty and AD pathology are somehow intrinsically related. Here, we confirm this relationship and suggest some mechanisms of shared etiology, and how these two factors interact to produce dementia.

The dementia literature raises several issues for which no pathophysiologic mechanism has yet been able to account, including (1) the relatively weak relationship between AD pathology and Alzheimer’s dementia (i.e. AD pathology does not appear to be necessary or sufficient), (2) the high prevalence of mixed dementia, and (3) many, diverse risk factors. Our current model responds to these challenges by explaining some of the heterogeneity between AD pathology and dementia, showing AD pathology may be one factor in a continuum of events that can eventually lead to dementia, and opens doors for novel interventions that go beyond halting AD pathology progression.

Our data must be interpreted with caution. Secondary analyses are limited in that not all relevant items necessarily have been collected. Here, however, the relevant measures (cognition and dementia status, AD pathology, and items to construct an FI) were generally available. Further, this was a cross-sectional analysis in which AD pathology was measured post-mortem. To overcome this, we used frailty measurement and dementia status from the last clinical interview before death (mean 0.89 years pre-mortem, median 0.61 years). As AD pathology begins to accumulate decades before clinical manifestations of the disease,²⁹ and pathology at death is related to the rate of cognitive decline over many years prior,³⁰ we can be relatively confident that this limitation would not significantly bias our results. Nevertheless, our results only confirm associations between frailty, Alzheimer’s dementia, and AD pathology. Future studies should examine longitudinal relationships between frailty, cognition, and biomarkers of Alzheimer’s dementia in order to establish causation, and include more diverse pathological lesions to better understand the interplay between pathology and the dementia syndrome.

As frailty measurements were taken close to death, the recorded frailty state might be reflective of terminal decline. If this were the case, the relationship between AD pathology and dementia status among people with high frailty may be overrepresented, though this would not explain cases where frailty was low and AD pathology was high in people with dementia.

Competing risks is also a consideration; if participants died of causes other than those related to dementia without the chance to develop dementia, the results may be confounded. We did not have access to cause of death and could not control for this.

Analyses were limited to Bielschowsky silver stain which may be less robust than immunohistochemistry with specific antibodies³¹. It will be important for future studies to repeat these analyses using more molecularly specific markers of β-amyloid and PHFtau tangles. We attempted to account for vascular risk factors by controlling for them in regression sensitivity analyses, though this cannot fully account for the effects of microinfarcts, hypertensive arteriopathology, and other micro pathologies that have been associated with dementia. It is common for people with Alzheimer’s dementia to exhibit mixed pathologies. We did not control for the contribution of non-AD pathologies, as we aimed to establish links between hallmark diagnostic pathological measures and clinical presentation. Future work will need to address age-related pathologies that contribute to dementia.

Our sampling was largely from retirement homes in Illinois. Though this may introduce bias, the utility of the data to represent varying cognitive and neuropathological profiles was overriding. Further, MAP enjoys extraordinarily high follow-up and autopsy rates (reducing bias and increasing internal validity), and clinical evaluations in MAP are identical to those in the population-based Chicago Health and Aging Project. All risk factor associations that have been examined in both cohorts have cross replicated. Nevertheless, future research should address this hypothesis using a population-based sample.

Our results suggest that dementia expression is multiply determined, and a single mechanism is unlikely to explain the diverse expressions we see in the people who most often develop dementia: those who are older and have multiple comorbidities. Individuals with even a low level of AD pathology may be at risk for dementia if they have high levels of frailty. This contributes to an emerging conceptualization of dementia (and particularly Alzheimer’s) as a complex disease of aging, rather than a single disease entity marked by genetic risk or particular protein abnormality.³² This has the potential to improve our understanding of disease expression mechanisms, explain failures in pharmacological treatment, and aid in development of more appropriate therapeutic targets, approaches and measurements of success. These results may therefore contribute to more effective prevention and management of Alzheimer’s dementia. This work is a novel contribution to the existing literature as it bridges epidemiological methodologies and clinical neuropathology in relation to deficit accumulation in Alzheimer’s dementia and proposes a unique model of Alzheimer’s dementia development.

Research in Context.

Evidence before this study

Using the terms ‘neuropathology’, ‘frailty’, ‘dementia’, ‘Alzheimer’s disease’ and their synonyms we searched GoogleScholar and PubMed for relevant articles between Jan 2017 and Jul 2018 in English or French. The dementia literature raises several issues for which no pathophysiologic mechanism has yet been able to account. 1) The weak relationship between AD pathology and dementia (i.e. AD pathology does not appear to be necessary or sufficient); 2) the high prevalence of mixed dementia; 3) the many, diverse risk factors; 4) the failure of clinical trials.

Added value of this study

The moderation model presented here aims to respond to these challenges by explaining variance in the relationship between AD pathology and dementia, showing how AD pathology may be an interacting risk factor in a continuum of events that can eventually lead to the dementia syndrome, and opens the door for novel interventions that go beyond halting AD pathology progression.

Implications of all the available evidence

Dementia expression is proving to be multiply determined, and a single mechanism is unlikely to explain the diverse expressions we see in the people who most often develop dementia: those who are older and have multiple comorbidities. This characterizes to an emerging conceptualization of dementia (and particularly AD) as a complex disease of aging, rather than a single disease entity marked by genetic risk or particular protein abnormality. In a field with so many competing claims about individual risk factors, understanding how they work together to give rise to clinical dementia may offer a new way to understand dementia risk and target treatment.