Modeling the Dementia Epidemic

SUMMARY

The incidence of dementia increases steeply with age in older people, although from the tenth decade the slope may be smoother, perhaps reflecting different pathological processes in the oldest old. The prevalence depends upon interaction of age with other factors (e.g., comorbidities, genetic or environmental factors) that in turn are subject to change. If onset of dementia could be postponed by modulating its risk factors, this could significantly affect its incidence. Analysis of risk and protection factors should take into account the critical period during which these factors play a role. For example, the impact of education and diabetes mellitus occurs in early‐ and midlife, respectively, while maintaining optimal physical and mental activity and controlling vascular factors later in life may slow the rate of cognitive decline. Modifying factors need to be evaluated for different clinical groups, taking into account genetic background, age, and duration at exposure. The aim of the present article is to try to take stock of epidemiological data concerning factors affecting the prevalence of dementia and predict future developments, as well as to look for possible interventions that could affect outcome.

Introduction

Dementia is a syndrome, which rightfully attracts wide attention, since it reaches very high prevalence rates in both western countries and developing communities [1]. The high rates impact individuals, their families, and national economies. It therefore deserves careful study of its causes and pathogenesis in order to try to minimize these effects, organize social services, and predict future developments.

The syndrome of dementia has different causes and manifestations, each of which needs to be discussed separately. In this review we shall focus on the most frequent ones, Alzheimer's disease (AD) and vascular dementia (VaD), which together constitute a large majority of cases with dementia [2].

Dementia and its subtypes are defined using accepted clinical criteria (e.g., DSM‐IV for AD or VaD, Ref. 3, Neary et al. for frontotemporal degeneration [FTD], Ref. 4), and consensus guidelines for dementia with Lewy bodies [5], while biological markers may support the diagnosis of AD or suggest dementia with Lewy bodies (DLB) or FTD [6, 7, 8, 9]. However, the diagnostic accuracy of dementia syndromes is relatively low as there is a large overlap between the clinical and pathological manifestations of AD with those of other diseases [10, 11]. The diagnosis of VaD is even more difficult than for these diseases, partly because VaD itself is heterogeneous, with multiple interacting mechanisms [12] and the lack of specific histological findings [13]. Vascular pathologies can be present in brains of nondemented elderly people [14]. In addition, the epidemiological traits of VaD are affected by those of its risk factors, which are overlapping with those of AD [15, 16, 17]. Furthermore, it has been suggested that in AD, neuronal demise may result from brain hypoperfusion, leading to the concept of “vasocognopathy”[18], while a large majority of cases with vascular brain lesions have coexistent AD pathology [19].

Traditionally, the clinical diagnosis of AD is confirmed according to arbitrary histological criteria [20, 21, 22]. Neurofibrillary tangles (NFT) and neuritic plaques (NP), the hallmarks of AD, are more frequent in hippocampus, neocortex, and entorhinal cortex of elderly people who died with dementia than among those who died without dementia, but the difference between the density of the lesions in demented persons and nondemented ones is stronger below 80 years than among older ones [23, 24]. NFT, more than NP, are correlated with the severity of dementia [25, 26]; however, the relationship between clinical manifestations and the pathological changes is complex. Pathological changes, for example, neocortical NP, common in asymptomatic elderly, are more specifically associated with dementia in relatively younger people [23, 24], while in older people, particularly after the age of 90 years, this relationship is lost [24]. Thus, after the eighth decade, 20–40% of nondemented people may have brain changes sufficient to fulfill the (arbitrary) diagnostic criteria of AD [24, 27, 28, 29], without any or with only subtle cognitive decline [29]. This may suggest that AD changes have less cognitive effects in advanced age. Erten‐Lyons et al. [30] addressed such people as resistant to dementia, but it is possible that the neuropathological lesions associated with AD are not causal but part of a complex process where associated factors interact differently, depending on the time at which they play a role. Terry and Katzman [31] have suggested that “primary senile dementia” could occur, owing to the decrease of neocortical synapses with age, regardless of NFT or SP.

Pathological Phenotype

Clinical phenotypes reflect mainly the anatomical distribution of the underlying pathology and may be related to enhanced vulnerability of brain regions or pathways. However, patients with AD frequently also harbor vascular brain lesions and Lewy bodies [32] and conversely, patients with VaD may have coexistent AD pathology [19, 33]. Lacunar infarcts and cortical microinfarcts have been found in the brains of people fulfilling diagnostic criteria for AD [34, 35]. Such cases are sometimes classified as AD, or as VaD, but probably should be regarded as mixed dementia [36].

Prevalence

Epidemiological studies concluded that the prevalence of dementia increases steeply with age [2, 37, 38]. In the Rotterdam study, the prevalence of dementia increased from 0.4% at age 55–59 years to 40.7% after 89 years, while the Berlin Aging Study of citizens aged 70 years and older demonstrated that from the age 95, dementia prevalence stabilizes at around 45%[39]. A similar effect was reported by Ritchie and Kildea [40] who calculated from a meta‐analysis of prevalence studies of dementia that from the tenth decade on, the prevalence of dementia should level off at around 40%. Because of the difficulties noted above of differentiation the various underlying disorders, it is impossible to compare studies looking at prevalence rates of specific diagnoses, since they used different criteria. However, AD (diagnosed clinically) accounted for 36% of the dementias among 55‐ to 64‐year‐old cases, 65% of the 65‐ to 74‐year‐old cases, 67% of the 75‐ to 84‐year‐old ones, and 78% of those older than 84 years; thus, the relative, as well as the absolute prevalence of AD increased with age, but after the age of 85 years the slope of the curve soothes [2], possibly because of reduced incidence and survival. The prevalence of VaD and the relative proportion of VaD versus AD also increase with age [37]. The prevalence of poststroke dementia is about 30%[41, 42]. Since prevalence reflects the product of incidence by duration of disease, from the age at which life expectancy will be shorter than duration of dementia, after the tenth decade for example, prevalence rates will become similar to incidence figures.

Incidence

The incidence of dementia also increases with age [43, 44, 45], but not uniformly. Its rate increases steeply from age 60, but by the tenth decade, incidence of dementia increases at a slower slope [46, 47]. In a meta‐analysis, Gao et al. [46] showed that although the incidence rate of dementia increases with age, the relative increase over 5 years (incidence rate ratio) is higher for the 65‐ versus 70‐year‐old group than the corresponding figures for the further quinquennial older groups; however, the incidence rate ratio for centenarians versus nonagenarians did not drop below 1. Miech et al. [48] observed even a decrease in incidence rate of AD among men older than 93 years and women older than 97 years, but this was not confirmed by the 90+ Study which showed that dementia incidence continues to increase after the age of 90 years and that the rate of increase doubles every 5 years to reach 40.7% in centenarians [49].

The cumulative incidence of dementia (calculated as uncensored people with dementia who have died up to this age and those with dementia still alive) between the age 65 and 100 years reaches 45% in women and 32.8% in men [50]. But, in 65‐year‐old men, the remaining lifetime risk of dementia is 10.9% (CI: 8–13.8) and that for women 19% (CI: 17.2–22.5), and the remaining risks of AD are 6.3% (CI: 3.9–8.7) and 12% (CI: 9.2–14.8), respectively; these risks take into account potential mortality due to competing causes, and are close to lifetime risks of hip fracture or cancer; the greater the risk of mortality in a given age, the larger the difference between the cumulative incidence and the lifetime risk [50]. The same group, evaluated the remaining risk of AD, from 70 years to 99 years, as 43% (95% CI: 14.4–71.6%) for men and 48.3% (95% CI: 35.6–61.1%), if the calculations had not been corrected for potential competing risks of death but these risks were reduced to 7.1% for men and 16.6% for women after adjustment [51].

In the future, the incidence of AD may be delayed by more aggressive reduction of risk factors, but these treatments may result in higher incidence ratios in older people [52].

Epidemiological figures for different dementia subtypes are probably somewhat inaccurate because cases with VaD could have been classified as AD [2] and vice versa.

In a long‐term follow‐up (median 21.1 months), Desmond et al. [53] observed that patients after stroke are at higher risk of developing dementia: relative risk of 2.7 (95% CI: 1.62–4.8) in octogenarians to 6.3 in nonagenarians (95% CI: 3.5–11.33). The lifetime risk of stroke is equal to or greater than the lifetime risk of AD [54]. The risk of dementia after stroke is greater shortly after stroke occurrence [55]. Pendlebury and Rothwell [56] found that prestroke dementia is associated with medial temporal atrophy and family history of dementia, while dementia developing after stroke is associated with the presence of multiple vascular lesions of the brain. It is possible that stroke precipitated latent AD or had cumulative effect with preexisting subclinical AD. They also showed that prevalence and incidence figures for poststroke dementia are higher in hospital‐based than in population‐based studies and that methodological differences may account for about 90% of the variance of the prevalence and incidence of poststroke dementia figures. They estimated that among stroke survivors with dementia, 10% had dementia before their first stroke, 10% developed dementia within a few months after the first stroke, and about 30–40% after recurrent strokes, while the rate of recurrent strokes is assumed to be 5% per year.

Survival and Mortality

The median survival after diagnosis of AD has been reported to be 5–8.1 years [57, 58]; it has been calculated to be 9.9 years for patients younger than 75 years at diagnosis, 6.9 years for patients 75–84 years at diagnosis, and 4.4 years for those older than 84 years at diagnosis of dementia [59]. After adjustment of disease duration from onset of dementia rather than from onset of cognitive decline (which precedes dementia), Wolfson et al. [60] estimated the median survival of dementia to be reduced from 6.6 years to 3.3 years (CI: 2.66–4.0). The differences observed in the reported survival times of AD patients may be related to the fact that as its onset is insidious, it is difficult to determine. Survival is shorter in VaD and mixed dementia than in AD [58, 61, 62]. As compared to people with normal cognition, their life expectancy is reduced by 65%, 51%, and 35%, respectively [62]. Survival of VaD is shorter than in AD, probably because of coincident cardiovascular complications, and particularly from cerebrovascular disease [62].

Kammoun et al. [63] found that bronchopneumonia and cardiovascular disease were the primary causes of death in demented and nondemented elderly in a geriatric hospital. Autopsy studies showed that patients with dementia die more frequently from bronchopneumonia (38.4–45.5%[64, 65]) than patients without dementia (2.8–28%[64, 65]). Neoplastic diseases were found to be the cause of death in 3.8% of patients with dementia versus 21.3% among patients without dementia [64]. However, cardiac causes of death were more common in VaD than in AD [63]. Cardiovascular disease was cause of death in 46.2% of patients with dementia and 31.3% of nondemented patients [65].

Risk Factors

Women were considered to be at higher risk of dementia, particularly of the Alzheimer type [44, 46], and after age 85 years [48, 66] or 90 years [44]; although the sex difference was found to be nonsignificant in longer or larger incidence studies [45, 50]. It is possible that the female preponderance in prevalence studies is related, at least in part, to their lower schooling [45]. This factor may become less significant in the future as women are more likely to have higher schooling. In addition, women live longer [67] even after development of dementia [68], which adds to the higher proportion of women among patients with dementia.

Theoretically, education may postpone the onset of dementia because it favors “brain reserve”[69], but actually the reported age of onset may be earlier in highly schooled individuals, possibly because of increased awareness [70, 71, 72]; this emphasizes that the definition of onset of dementia is soft and may not reflect the severity of the neuropathological changes, their milder stage lasting at least a few years. However, it is possible that education is a surrogate for other variables, such as lifestyle factors, treatment of vascular comorbidities, lifetime occupation, and cognitive stimulation, all of which could modify the risk of dementia [73]. Education and occupation have additive effects on the incidence of AD [74].

Low education has also been found to be associated with non‐AD dementia, that could be associated with risk factors for stroke for example [72, 75], which suggests the construct of “brain battering” from superimposed vascular lesions that are themselves associated with low education [72]. Cognitive activity delays memory decline, both in late and early life [76] but education has no known neuropathological correlates [71] and neuropathological AD changes are not related to level of intellectual activity [77]. Leisure activities (such as board games, playing a musical instrument, doing crossword puzzle, or reading) are associated with lower risk of vascular cognitive impairment as well as of AD [78].

The apolipoprotein E4 genotype (APOE‐4) is associated with a higher risk of developing AD [48, 79, 80] and is also associated with cerebral amyloid angiopathy [25], which is another neuropathological lesion associated with AD.

APOE‐4 is more frequent among patients with AD (27%) and VaD (21%) than among controls (11%[81]). APOE is involved in maintenance and repair of neurons, and this capacity is allele dependent [82], which could explain the allele effect on age of onset of dementia: APOE‐4 is associated with earlier age of onset of dementia, but has no effect in older populations, possibly because of attrition of individuals who had also other risk factors for dementia or cardiovascular diseases [48].

APOE‐4 also increases the risk of dementia among those with atherosclerosis, peripheral vascular disease, or diabetes mellitus (DM) [83], as well as that of stroke. Thus, the effect of stroke on dementia could be partly explained by APOE genotype [55]. These observations suggest that APOE‐4 is risk modifier since it interacts with other “vascular” risk factors. Although mildly, APOE‐4 isoform is significantly associated with elevated LDL‐cholesterol [84]. Kivipelto et al. [85] showed that APOE‐4, midlife systolic hypertension and midlife hypercholesterolemia are independent risk factors for AD, but have also a cumulative, or perhaps interactive, effect on risk of AD.

From a longitudinal study of elderly men, Feskens et al. [86] calculated that after adjustment for other vascular factors, 22% of the risk of developing impaired cognitive function, of any type, in this population may be attributed to the APOE‐4 allele.

TOMM40 is a gene that encodes Tom40 (involved in mitochondrial function), that is close to the APOE gene, in a region of strong linkage disequilibrium, on chromosome 19 [87], and for which length polymorphism of deoxythymidine polymer (poly‐T) can predict the age of late onset AD: longer poly‐T variants being associated with younger age of onset of AD [87, 88]. In an analysis of sequencing, Roses [87] found that APOE‐4 is virtually connected to long poly‐T variant and APOE‐3 to short poly‐T variants. Therefore, the genetic susceptibility to AD seems to be associated with both the APOE and TOMM40 genes [88]. In case of combination of APOE‐3 allele and long poly‐T, the age of onset is younger, which suggests that the TOMM40 polymorphism prevails over the APOE genotype [88].

Risk factors for stroke are also risk factors for cognitive decline, regardless of the occurrence of clinical stroke [89, 90], which strengthens the association of dementia and vascular pathology and the importance of prevention of vascular diseases particularly in vulnerable persons, for example, APOE‐4 carriers.

Midlife hypertension and late‐life hypotension are associated with late‐life dementia [91, 92]. Midlife hypertension has been found to be associated with the development of NFT and NP at autopsy [93]. In patients whose average age was 72.8 years, blood pressure lowering was not associated with decrease in incidence of dementia [94]. This finding suggests that such intervention should take place at earlier ages, either because there is a critical period of exposure to hypertension for development of dementia, or because in older subjects lowering blood pressure may be hazardous (e.g., by inducing excessive falls leading to further cerebral ischemia).

Transient ischemic attacks (TIA) have also been found to be associated with AD (hazard ratio = 5.1, 95% CI: 1.7–15.5), even among individuals without APOE‐4 [95], as well as for cognitive decline associated with vascular changes [96, 97, 98].

In patients with AF, dementia may occur in 2.7% within 1 year and 10.5% within 5 years [99]. However, among people older than 85 years, AF was associated with stroke but not with dementia [100]. Atrial fibrillation (AF) may increase the risk for poststroke dementia {OR = 2.35 (1.21–4.58), [101]}. Treating AF may decrease the risk of cardioembolic stroke by 60%[102] and hence potentially also the risk of VaD.

Midlife DM is associated with cognitive decline [103]. It increases the risk of dementia for onset <65 years (OR = 2.41 [1.05–5.51]) but not for onset ≥65 years (OR = 0.68 [0.3–1.53]) and is a stronger risk factor for VaD than for AD (OR = 2.17 [1.36–3.47] and OR = 1.69 [1.16–2.36]), respectively [104]. The association between (mild) cognitive decline and DM correlates with the duration and severity of midlife DM [105]. DM has been found to be associated with cerebral infarctions and small vessel disease but not to AD pathology [106]. Knopman et al. [103] in a longitudinal study of more than 10,000 middle‐aged persons found greater decline in cognitive test scores after 6 years among subjects with DM and hypertension. This again suggests that the susceptibility to cognitive decline may be age‐dependent, prior to the age of onset of cognitive decline, and that protection against risk factors should start in the presenium.

Hyperlipidemia has been shown to be associated with increased amyloid deposits in subjects younger than 55 years, without APOE‐4, but not in older ones, possibly reflecting the effect of cardiovascular deaths [107]. Statins use may be associated with lower risk of AD in patients younger than 80 years, but not in older ones [108].

In case‐control studies, smoking has been found to be inversely related to dementia [109] but in incidence studies, the opposite was observed [110, 111]. In the longitudinal Honolulu Heart Program, the risk of AD in smokers increased with pack‐years of smoking [OR = 2.18 (95% CI: 1.07–4.69) and OR = 2.4 (CI: 1.16–5.17), for medium level and heavy smokers, respectively, and neuropathological data indicate that the amount of NP increase with the amount smoked [111]. Very heavy smoking was not associated with the development of AD. The lack of effect of very heavy smoking on the risk of AD may be explained by the assumption that cross‐sectional studies exclude ex‐smokers already dead due to competing risk, as well as due to dementia. The effect of smoking on risk for AD is weaker among subjects without family history [110] and stronger among the APOE‐4 carriers [112].

A late‐life risk index, that included age, APOE‐4genotype, history of cardiovascular disease, internal carotid artery wall thickening on ultrasound, brain MRI findings (infarct‐like lesions, white matter disease, and ventricular enlargement), lifestyle factors, and cognitive performance, has been devised to stratify older people into low, moderate, or high risk of developing dementia depending upon their risk index: in this prospective study, 56% of subjects with high scores (>7) versus 22.8% of subjects with medium score (4–7) and 4.2% of those with low risk score (0–3) developed dementia over 6 years; higher subscores are attributed to older age (>79 years), low cognitive performance at the beginning of follow‐up and poor health behavior (e.g., low physical activity [113]).

Future Prospects

Changes in diagnostic criteria of AD and VaD might modify epidemiological figures since histological changes that define AD are correlated with cognitive status among people who died before the age of 75 years while this correlation is lesser in older people [24]. Also, histological changes of AD can be present in older people without dementia [24]. If the relationship between AD changes and dementia is attenuated with age, this may suggest that additional factors, such as vascular changes or other neurodegenerative changes determine the clinical expression in the oldest old [24].

Figures for VaD will probably be modulated by differential clinical patterns underlined by different pathogenesis: small vessel dementia (SVD), SVD combined with AD, large vessel dementia, or hypoxic‐hypoperfusive dementia [114, 115, 116]. Desmond [116] addressed VaD as a “construct in evolution,” while Merino [117] and Korczyn [12, 118] emphasized the complexity of the possible interaction between the different pathogenetic mechanisms of VaD and AD. If the clinical picture of hypoxic‐hypoperfusion type of VaD was in fact closer to AD than to VaD, epidemiological studies should address this issue, by stratification of cases upon their underlying pathological traits, in their frequency and factors that impact on their development.

Because the prevalence of dementia doubles every 5 years [119], 5‐year delay of onset of dementia would reduce its age‐specific prevalence by 50%, while intervention delaying onset of dementia by 1 or 2 years could reduce its prevalence by about 10% or 20%, respectively [120, 121]. Therefore, in face of the aging of the population, modifying the prevalence of risk factors for dementia could affect that of dementia. The effect of treatment of risk factors may be complex: for example, correction of coronary heart disease (CHD), that has been identified as a risk factor [122], could paradoxically increase the incidence and prevalence of dementia by reducing or delaying cardiac death. Thus, on one hand, the incidence of CHD declined (by 2.2–2.3% from 1996 to 2005) but on the other hand this decline was offset by decline in mortality related to it (4.5–3.4%, [123]), leaving its survivors at risk to develop dementia for longer time.

Secular trends indicate a decline in prevalence, as well as in incidence of VaD, possibly because of successful stroke prevention [124], by reduced smoking, improved control of hypertension and hyperlipidemia, etc.

The course of AD can be modified by the presence or absence of vascular risk factors [103] or by their treatment [125]. Cognitive decline in patients with AD and in whom presumed risk factors for vascular disease are treated is slower than in those without treatment of their vascular risk factors [125]. This, again, emphasizes the overlap between AD and VaD.

Therefore the epidemiological studies of dementia should by now stratify the cases by clinical picture, underlying risk factors, such as genotype, age, education, vascular changes, the age at which exposure to these factors began, and duration of exposure to these factors, rather than by diagnosis.

In Summary

Dementia is an age‐related, probably multifactorial disease, the ultimate diagnosis of which is traditionally based on pathological changes. However, these may not be sufficient to define AD or VaD since nondemented elderly people may have identical brain lesions. The association of these lesions with dementia is nonlinear with age, which suggests that the impact of the factors involved is not uniform, depending on the age at which they act.

Although the incidence of dementia increases with age, after the ninth decade, the increase is perhaps attenuated [46]. This, combined with competing risks of death and control of risk factors [36] may lead to relative stabilization of prevalence rates of dementia. Diagnosis of dementia of the AD type may be modulated by associated vascular lesions, and conversely some VaD pictures may represent AD, with associated subcortical lesions. Therefore, epidemiological predictive studies should take the underlying vascular pathology of the populations studied into account.

Risk factors may not only precipitate the development of dementia but also affect the course of the disease itself and survival of the affected individuals. In addition, their impact may depend upon the age at which they act and their interaction with other factors or concomitant conditions, which underlines the importance of stratified and multivariate analyses by period of exposure.

Conflict of Interest

The authors have no conflict of interest.

Author Contributions

Professor A.D. Korczyn and Dr. T.A. Treves are responsible for the concept, design, and data analysis and Professor A.D. Korczyn for the critical revision of the review.