Mortality
Globally, dementia (all-cause) is the fifth leading cause of death, with 4.4% of all deaths attributable to dementia in 2016 [
31]. Deaths due to dementia have steadily increased over time, partly due to population growth and population aging, more than doubling from 1990 to 2016 [
31]. AD dementia-related deaths have also risen. An analysis of AD dementia mortality in Europe found that deaths from AD among patients aged ≥ 50 years (as identified by diagnostic codes) more than doubled from 1994 [41,255 deaths] to 2013 [86,822 deaths] [
32]. The age-standardized mortality rate for deaths caused by AD dementia in Europe was 45.2 per 100,000 in 2013 [
32].
Recording of cause of death is variable and AD dementia may be either a contributing cause or an underlying cause [
33,
34]. In the USA in 2017, 46.4% of dementia-related deaths were attributable to AD [
34]. One study of AD dementia in Canada investigated multiple causes of death when AD was listed on death certificates [
33]. In this analysis, 4.3% of all deaths occurring from 2004 to 2011 were AD dementia-related, but 2.6% of deaths had AD dementia as an underlying cause, in contrast to 1.7% of deaths having AD dementia as a contributing cause [
33]. Usually, when AD dementia was listed as an underlying cause, cardiovascular disease was the most common contributing cause (46%); when AD dementia was listed as a contributing cause, cardiovascular disease was the most common underlying cause (41%) [
33]. The crude mortality rate in Canada for deaths from AD dementia as an underlying cause increased over the period 2004 to 2011, from 10.1 to 11.5 per 100,000 for men and from 24.4 to 25.4 per 100,000 for women [
33].
Data on survival from early stages of AD dementia are scarce. However, median survival among persons with AD dementia has been estimated to be approximately 7 years from presentation with cognitive decline (according to an analysis of data from one geographic area in England) [
35].
MCI due to AD
Patients with MCI due to AD are more likely to progress to dementia than those with MCI that is not associated with toxic species of Aβ and/or the development of neurofibrillary tangles of hyperphosphorylated tau protein [
5,
15,
21,
36‐
38]. Data from 13 cohort studies representative of multiple countries has demonstrated a high prevalence of AD at the MCI stage according to the IWG-1, IWG-2, and NIA-AA criteria, which identifies patients at higher risk of dementia based on the presence of AD-related biomarkers (abnormal Aβ and/or tau) and or neuronal injury [
38]. On the basis of the IWG criteria, 53% (IWG-1) and 40% (IWG-2) of subjects with MCI were identified as having AD; 3-year progression rates were 50% (IWG-1 criteria) in persons with prodromal AD versus 21% without prodromal AD, and 61% (IWG-2) with prodromal AD versus 22% without prodromal AD. On the basis of NIA-AA criteria, 46% of subjects with MCI were classified as being in the “high likelihood AD” group [
38]; and the 3-year progression rate was 59% (subjects in lower risk groups had 3-year progression rates of 5–24%).
The incidence of clinically diagnosed MCI has been examined in a systematic review of studies in Europe, the Americas, and Australia [
39]. Incidence rates ranged from 22.5 to 60.1 per 1000 person-years depending on age, but meta-analysis suggests substantial heterogeneity in incidence estimates due to methodological and population sample characteristics in the included studies. Analyses resulted in MCI incidence of 22.5 (95% CI 5.1, 51.4) per 1000 person-years for ages 75–79, 40.9 (95% CI 7.7, 97.5) for ages 80–84, and 60.1 (95% CI 6.7, 159.0) for ages 85 and older [
39]. Other research has examined etiologic diagnoses of AD in MCI (as evidenced by NIA-AA clinical criteria), finding that 75% of subjects with MCI had an etiology of AD, while the remainder were classified with etiologies such as CVD and Lewy body dementia [
40].
Several recent studies in China have documented population-based prevalence of clinically diagnosed MCI [
37,
41,
42]. In a meta-analysis of 41 studies of Chinese community-dwelling populations over 55 years of age, the pooled prevalence of MCI was 12.2% (95% CI 10.6, 14.2) for MCI and 10.9% (95% CI 7.7, 15.4) for amnestic MCI. There were slight differences in prevalence according to diagnostic criteria, with rates of MCI of 13.5% using the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSM-IV) criteria, 12.9% using the Petersen criteria, and 10.3% using NIA-AA [
42]. There was a higher prevalence among women than men (13.8% [95% CI 9.7, 13.6] vs. 11.5% [95% CI 11.7, 16.3]), and people living alone versus in families (18.2% [95% CI 13.6, 24.4] vs. 14.1% [95% CI 11.0, 18.2]); those that had education levels lower than primary school had much higher prevalence of MCI (17.2% [95% CI 12.2, 24.3]) when compared to persons with higher levels of education. Studies in this meta-analysis based on more recent observation periods resulted in higher prevalence estimates (before 2005, 3.7% [95% CI 1.6, 8.7]; in 2005 or later, 14.1 [95% CI 12.4, 16.0]) [
42]. One study of a rural area in Northern China in 2015 reported prevalence of MCI and AD, finding rates of 27.8% (overall MCI), 18.4% (MCI due to AD), and 6.5% (AD) [
41]. Another national cross-sectional study of adults over the age of 60 in China reported age- and sex-adjusted overall MCI prevalence of 15.5% (95% CI 15.2, 15.9) [
37].
Preclinical Stages of AD
AD dementia develops by progression through several stages, from no accumulation of toxic Aβ species to normal cognition with Aβ deposition and then neurodegeneration (as evidenced by elevated CSF tau protein and fluorodeoxyglucose (FDG) positron electron tomography (PET) imaging), and subsequently to MCI. According to a multistate disease model, the lifetime risk of AD for a 60-year-old person with normal cognition and no toxic Aβ accumulation has been estimated to be 20.1% (range 10.6–34.0%) among women and 13.9% (range 6.9–25.1%) among men [
43]. For patients with Aβ deposition and neurodegeneration, estimates of lifetime risk are 41.9% among women and 33.6% among men, and for patients with MCI, lifetime risk is more than twice that of the earlier stage. By contrast, a much higher percentage of patients with MCI due to AD were estimated to progress to dementia (men 92.9%; women 95.6%).
Prevalence estimates studied in a systematic literature review that measured amyloid PET or CSF in subjects with normal pathology showed rates of 22% (CI 18–27%) and 21% (CI 15–29%) for PET and CSF, respectively; however, prevalence rates across the studies including biomarker data ranged widely (7–48% for PET studies, and 4–52% for CSF studies) [
44]. Nevertheless, results suggest that measuring biomarkers is an important component of identifying risk of progression from preclinical AD to MCI and ultimately to AD dementia. In the same systematic literature review, an analysis of the studies that provided data across NIA-AA preclinical stages resulted in prevalence estimates of 13% (CI 9–18%), 16% (CI 9–25%), and 5% (CI 3–9%) for stages 1, 2, and 3, respectively.
Other research using the NIA-AA criteria has estimated the prevalence of preclinical and clinical AD by stage [
36]. Rates were calculated on the basis of a multistate Markov model with data on incidence from longitudinal studies, as well as mortality rates and population projections. Most cases with AD pathology (i.e., Aβ deposition, CSF tau, and neurodegeneration based on FDG-PET imaging) were preclinical, with rates from this model declining from early to later stages of AD owing primarily to attrition from death (from age, other conditions, or dementia). Projected 2060 prevalence estimates increased across all stages from 2015, more than doubling for each of the later stages: MCI due to AD (2.4 million in 2017; 5.7 million in 2060), early-stage AD dementia (2.1 million in 2017; 5.3 million in 2060), and late-stage AD dementia (1.5 million in 2017; 4.0 million in 2060). Prevalence estimates for preclinical AD also increased in patients with Aβ deposition only (22.1 million in 2017; 31.9 million in 2060) and in those with Aβ deposition and neurodegeneration (16.2 million in 2017; 30.2 million in 2060).
This multistate model was also used to assess the impact of hypothetical disease-modifying health interventions on forecasts. In one scenario, an intervention delaying the annual risk of progression to MCI by 50% resulted in a decrease in the prevalence of MCI (from 5.70 million to 5.01 million) as well as a decrease in the prevalence of AD dementia (from 9.3 million to 6.95 million) [
36].
Disease Progression
Progressive deterioration of cognitive function in persons with AD varies according to the affected elements of cognition. In an analysis of data from 1495 adults with clinically diagnosed AD dementia from the Geriatric Education and Research in Aging Sciences (GERAS) study, deterioration on both the Mini Mental State Exam (MMSE) and the Alzheimer’s Disease Assessment Scale-cognitive subscale (ADAS-cog) showed a similar pattern of progressive decline in which word recall and orientation were affected first, with declines in attention and concentration, language, constructional praxis, and executive function occurring next, and immediate memory deteriorating in severe AD dementia [
45]. A systematic literature review of studies examining acceleration points in cognitive decline showed that verbal memory deficits appeared first, followed by deficits in visuospatial ability, executive functions, fluency, and verbal IQ [
46].
Data on the risk of progression from preclinical AD to MCI are sparse, but a systematic literature review and meta-analysis (2011–2018) included studies that provided progression data across the three preclinical NIA-AA stages [
44]. Rates of progression increased across stages: 73% (95% CI 40–92%) at the preclinical AD stage 3 (subtle cognitive decline with AD biomarker positivity), 38% (95% CI 21–59%) at stage 2 (toxic Aβ accumulation and neurodegeneration), and 20% (95% CI 10–34%) at stage 1 (toxic Aβ accumulation but asymptomatic) [
44]. The relative risk of progression increased across stage and was over six times higher at stage 3 (RR = 6.38; 95% CI 3.33–12.24) compared to individuals with normal biomarkers [
44]. Authors suggested that on the basis of these data, NIA-AA stage 3 should be used along with MCI due to AD in treatment decision-making [
44].
Data from studies of progression from clinically diagnosed MCI to AD dementia have shown that the majority of patients with MCI either remain cognitively stable or revert to normal values, versus progressing to AD [
11,
20,
47,
48]. However, comparing results across studies can be challenging because of inconsistent study designs, differing clinical definitions of MCI, and differences in patient populations [
20]. For example, in a meta-analysis of 28 studies (through 2014) including 2365 patients with MCI, 38.7% progressed to AD dementia over a mean follow-up of 31 months; however, progression rates in the individual studies varied from 6% to 39% per year.
One factor that has implications for estimating progression rates is the subtype of MCI. A meta-analysis of 33 studies (1999–2017) of 4907 patients with clinically diagnosed MCI showed that approximately half of patients with sd-aMCI or md-aMCI progressed (sd-aMCI odds of progression to AD, 0.47; 95% CI 0.33, 0.66;
p < 0.001; md-aMCI odds of progression to AD dementia, 0.52; 95% CI 0.36, 0.75;
p < 0.001), but that risk of progression was substantially lower for patients with naMCI (sd-naMCI odds of progression to AD dementia, 0.11; 95% CI 0.07, 0.16;
p < 0.001; md-naMCI odds of progression to AD dementia, 0.18; 95% CI 0.11, 0.27;
p < 0.001) [
20]. This was also illustrated in the prospective Australian Imaging, Biomarkers and Lifestyle (AIBL) study in which patients (
n = 866) were classified by subtype of clinically diagnosed MCI and then also by the severity of their memory impairment [
11]. After 3 years, patients with aMCI were more likely than the healthy controls to develop AD dementia (positive predictive value [PPV] 24.1%; 95% CI 18.4, 30.6), and risk increased with the severity of memory impairment (grade 1 PPV = 10.0%; 95% CI 5.1, 17.2; grade 2 PPV = 43.0%; 95% CI 32.8, 53.7) [
11]. Patients with md-aMCI had a higher risk of progression vs. sd-aMCI (PPV of 47.3%; 95% CI 33.7, 61.2 vs. PPV of 15.5%; 95% CI 10.1, 22.4), and risk also increased with increasing severity of memory loss [
11]. The annual progression rates were 8.0% for aMCI, 5.2% for sd-aMCI, and 15.8% for md-aMCI [
11]. Other studies have reported rates of progression from clinically diagnosed MCI to AD dementia varying from 18.3% (9 of 51 people) to 48.3% (86 of 178 people) within 2 years [
49,
50]. A larger study found that 33.3% (181 of 544 patients) progressed from clinically diagnosed MCI to AD dementia within a median of 46 months [
51].
The progression rates reported in studies can also be influenced by the source of recruitment of patients: the aforementioned meta-analysis of 33 studies also classified patients as community-dwelling or from specialist clinics and showed that rates of progression were higher for patients from specialist clinics vs. community samples (sd-aMCI, 40% vs. 18%; md-aMCI, 35% vs. 21%) [
20]. This concept was also demonstrated in a systematic literature review and meta-analysis of 59 studies (1999–2016) [
52]. The overall progression rate (timeframe not reported) from clinically diagnosed MCI (subtype not stated) to AD dementia (adjusted for prevalence and weighted by follow-up time) was 28% (95% CI 22, 33) and to dementia was 34% (95% CI 27, 40), with 45% (95% CI 34, 55) remaining stable and 15% (95% CI 10, 19) improving their status to normal [
52]. When the data were grouped according to community vs. specialist clinic patients, progression to AD was higher in the clinic sample compared to the community sample (31% vs. 13%), which may be due differences between clinic and community populations [
52].
Another confounding factor is that a sizable percentage of healthy elderly adults score below normal on at least one cognitive test when given a battery of tests [
47]. In order to resolve this issue, some researchers used the concept of “base rate of low scores” (BRLS) to classify patients from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database [
47]. According to this approach, a person was deemed to have MCI if their number of low scores on a panel of cognitive tests was greater than or equal to the number of the 10% of worst-performing patients [
47]. According to these criteria, 34% of patients defined as MCI progressed to AD dementia, compared to 22% defined by the Petersen criteria, and 11% of sd-aMCI and 29% of md-aMCI defined by the Winblad criteria [
47].
Several modeling studies have estimated the progression of dementia severity in patients with AD. An analysis based on patients from the National Alzheimer’s Coordinating Center (NACC) (
n = 3009) estimated that a starting population of patients with mild (69.5%) or moderate (30.5%) dementia would progress to mild (38.6%), moderate (44.1%), or severe (7.3%) dementia or death (10.0%) after 1 year [
53]. Another analysis based on a larger NACC population (
n = 18,103) estimated that patients aged 65 years with mild AD dementia had a 25% chance of progression to moderate (19%) or severe (1%) dementia or death (4%) within 1 year, whereas those with moderate AD dementia would have a 36% chance of progression to severe dementia (27%) or deal (9%) [
54]. Slightly higher progression rates were estimated for patients aged 75 years. Unfortunately, neither of these studies took patient amyloid status into consideration. A more recent analysis limited to Aβ-positive patients estimated higher annual rates of progression. Among Aβ-positive patients with mild dementia, 45.1% of patients with progressed to moderate (31.6%) or severe (4.3%) dementia or death (9.2%) within 1 year. Among those with moderate dementia, 59.8% progressed to severe dementia (28.6%) or death (31.2%) [
55]. These analyses highlight the limitation of studying AD progression without incorporating biomarker estimates including amyloid status.
Risk Factors for Disease Progression
Understanding risk factors associated with disease progression can help individual patients make decisions regarding their future as well as guide clinical development and use of treatments for slowing progression [
56,
57]. However, the factors noted previously that influence data on reported rates of progression also are likely to influence data on risk factors, sometimes resulting in conflicting results.
Some researchers have examined which tests (neurocognitive or biomarkers) might be the most valuable in identifying patients who are likely to progress from clinically diagnosed MCI to AD dementia or from mild to more severe AD dementia. In the meta-analysis of 28 studies (through 2014) that included 2365 patients with MCI, described previously, the objective was to evaluate the predictive value of neurocognitive tests for progression to dementia [
48]. Five tests were shown to have excellent (≥ 0.90) sensitivity and specificity: Addenbrooke’s Cognitive Examination (a global measure), Visual Object and Space Perception Silhouettes (visuospatial), Object Function Recognition (language), and three tests of verbal episodic memory (Face–Name Association Task; Rey Auditory Verbal Learning Test, and Guild Paragraph) [
48]. The meta-analysis of 24 studies (through 2018) that included 2689 patients with clinically diagnosed MCI (primarily aMCI) examined the ability of neuropsychological testing to predict progression to AD dementia [
58]. A mean of 37% of these patients progressed to AD dementia [
58]. Overall, patients who progressed did worse on a variety of measures than did non-progressors, affirming the relevance of neuropsychiatric testing [
58]. In a systematic analysis of 48 studies that examined biomarkers in patients with AD dementia, little information was available from studies with follow-up of more than 2.5 years or with repeated testing [
59]. Although magnetic resonance imaging (MRI) scan of whole brain or hippocampal atrophy was the most common testing measure, the investigators noted it is not specific to AD [
59]. In most studies, evaluation of progression was based on cognitive testing, but there was no consensus on which tests or combinations of tests were most valuable [
59].
Risk factors for progressing from unimpaired cognition to aMCI were evaluated on the basis of data from approximately 1500 individuals in the ADNI database [
60]. Over 4 years, 17% of unimpaired participants converted to aMCI [
60]. Significant risk factors for this conversion were low hippocampal volume, a high CSF tau/Aβ ratio, and a low memory score; neither increasing age nor family history of AD was significant [
60]. The risk of conversion from no impairment to aMCI increased with the number of risk factors; participants with three risk factors had an estimated probability of developing aMCI of 0.35 over 4 years of follow-up [
60].
Multiple risk factors for progressing from MCI to AD dementia have been proposed. A meta-analysis of 53 studies with 12,396 patients with AD dementia and 1934 with MCI examined progression risk [
61]. For patients with MCI, rapid progression to AD dementia was associated with positive
APOE4, greater white matter hyperintensity volume, entorhinal cortex atrophy, high CSF neurogranin levels, and high dependency in instrumental activities of daily living (IADLs) [
61]. More engagement in social activities and a higher body mass index (BMI) were associated with slower progression to AD dementia [
61]. Evaluation of the ADNI database showed that of patients with aMCI at baseline, 35% had progressed to AD dementia by 4 years [
60]. Significant risk factors for progression were
APOE4 positive, high FAQ score, a low memory score, hippocampal atrophy, and a high tau/Aβ ratio [
60]. Risk increased with the number of risk factors; participants with five risk factors had an estimated probability of developing aMCI of 0.90 over 4 years of follow-up [
60]. In a meta-analysis of 60 studies (1966–2015) of patients with MCI, the greatest risk factors for progression to AD dementia were abnormal CSF tau, abnormal CSF tau/Aβ ratio, hippocampal atrophy, medial temporal lobe atrophy, and entorhinal atrophy [
62]. Other risk factors were
APOE4 positive, CSF total tau, white matter hyperintensity volume, depression, diabetes, hypertension, older age, female gender, lower MMSE score, and higher ADAS-cog score [
62]. Protective factors were high BMI and higher auditory verbal learning test delay score. Individuals with a combination of risk factors had an even greater risk for progression [
62]. It has been demonstrated that declines in both cognitive and financial skills in patients with clinically diagnosed MCI are significant predictors of clinical progression [
50].
Many studies have attempted to identify risk factors for progression from mild to more severe AD dementia. Several factors increasing the risk of AD dementia have been identified in this review (Table
3). There are conflicting results for some factors (e.g., age, comorbidities, education level), but many studies agree that atrophy in various brain regions, poor baseline memory and ADAS-cog scores, and abnormal CSF biomarkers are risk factors [
49]. Other potential factors that vary by AD dementia severity include hair cortisol levels, which have been observed to be higher in patients with more severe dementia compared to those with mild dementia, and salivary IgA levels, which exhibit the inverse pattern [
63]. Systemic inflammation, as evidenced by consistently elevated concentrations of inflammatory biomarkers and cytokines, has been observed in patients with AD dementia [
64]. However, inflammatory biomarkers have not always been associated with measures of clinical disease progression in patients with clinically diagnosed MCI or AD dementia [
65]. A recent study in a small number of patients (
n = 39 patients with AD dementia and
n = 21 healthy controls) found that the levels of oligomeric Aβ in nasal discharge were higher in patients with AD dementia and that the presence of specific Aβ oligomers could distinguish between patients with mild and moderate cognitive dysfunction [
66].
Table 3
Factors shown to increase risk of AD
Hippocampal atrophy |
Medial temporal atrophy |
Entorhinal atrophy |
Abnormal CSF tau |
White matter hyperintensities |
Abnormal CSF tau/Aβ ratio |
APOE4 |
Increasing age |
Female sex |
Low memory/MMSE scores |
High ADAS-cog score |
High dependency/IADL score |
High FAQ score |
Neuropsychiatric symptoms |
Hypertension |
Depression |
Some analyses have focused on factors associated with rapid or slow decline. The previously described systematic literature review of 53 studies of patients with MCI and AD dementia reported that risk factors for rapid progression in patients with dementia were
APOE4 positive, early age at onset, higher level of education, early appearance of extrapyramidal signs, and neuropsychiatric conditions [
61]. Factors such as age ≥ 75 years, diabetes, and multidrug therapy lowered the risk of rapid progression [
61]. In the Impact of Cholinergic Use (ICTUS) study, a prospective study in 12 European countries that followed 1005 patients with AD dementia, a worse baseline ADAS-cog score was significantly associated with rapid decline, while increased age at baseline was a protective factor [
57]. The Progression of Alzheimer's Disease and Resource use (PADR) study evaluated 282 Norwegian patients with AD dementia for a mean of 2 years [
67]. Over 40% of patients progressed slowly (47% by change in Clinical Dementia Rating Scale-Sum of Boxes [CDR-SB] and 44% by change in MMSE) [
67]. Baseline factors associated with slower progression on the CDR-SB were younger age at diagnosis, higher education, better IADL function, better scores on cognitive tests, used fewer drugs, and 1-point lower CSR-SB score, while better scores on cognitive tests and better CDR-SB score were associated with slower progression on the MMSE and IADL [
67].
The Canadian Outcomes Study in Dementia (COSID) evaluated overall progression using data from 488 patients with AD dementia [
68]. After adjustment for age, patients who progressed were significantly more likely at baseline to have poorer cognition, greater dependence, and more neuropsychiatric symptoms [
68]. Male sex and having a worse baseline Global Deterioration Scale (GDS) score were significant protective factors [
68]. A systematic literature review of 11 studies (to 2016) evaluating the influence of comorbid conditions on progression of late-onset AD dementia [
56] found associations between comorbidities and declines in cognitive function and functional abilities, as well as increases in neuropsychiatric symptoms [
56]. Other studies have examined specific factors protective or predictive of clinical progression. Blood pressure variability [
69] has been proposed as a risk factor for progression, and cognitive reserve (e.g., IQ prior to disease onset) [
70] has been proposed as a protective factor. A recent analysis of healthy controls (
n = 318) and individuals with MCI (
n = 168) or AD dementia (
n = 269) found that neuroinflammation (as evidenced by translocator protein (TSPO) identified during PET) increases with more severe AD dementia [
71]. Compared with healthy controls, individuals with MCI had increased TSPO levels in the neocortex and those with AD had even greater levels of TSPO present throughout the brain [
71]. There was a significant inverse association between TSPO levels in the parietal region of the brain and MMSE scores (
p = 0.024) [
71].