Mortality Risk after Diagnosis of Early-Onset Alzheimer’s Disease versus Late-Onset Alzheimer’s Disease: A Propensity Score Matching Analysis

Abstract

Background/Objective: We aimed to compare the risk of mortality in patients with early-onset Alzheimer’s disease (EOAD) versus those with late-onset AD (LOAD) using a large number of study subjects. We applied propensity score matching (PSM) to minimize confounding biases in the comparison between EOAD and LOAD.

Methods: We obtained data from elderly Korean subjects with AD (n = 3,611) at baseline from the CREDOS cohort study, which was conducted from November 2005 to July 2013. We conducted PSM to reduce the bias due to confounding variables related to survival in patients with AD. The risks of mortality associated with EOAD and LOAD were evaluated by Cox proportional hazard analyses, controlling for relevant covariates.

Results: After propensity score matching, 312 subjects with EOAD and 624 subjects with LOAD were selected for further analysis. The Cox proportional hazard analysis showed that patients with EOAD are at a greater risk for mortality compared to those with LOAD (Hazard Ratio: 2.01, 95% CI: 1.01–4.00, p-value: 0.04) when controlling for the direct effect of aging on mortality. The results did not change after adjusting for age at diagnosis, general cognitive function, nutritional factor related to body mass index, and physical disability using activities of daily living. The results support the assumption that EOAD takes a more malignant course than LOAD.

Conclusions: Our results provide support for the idea that EOAD takes a clinical course that is distinct from that of LOAD, especially as pertains to the risk of mortality.

Keywords

Early-onset Alzheimer’s disease late-onset Alzheimer’s disease mortality propensity score matching

INTRODUCTION

There has been growing interest in early-onset Alzheimer’s disease (EOAD), a form of AD which strikes before the age of 65 [1 –4]. Although AD that presents at an early age was once thought to be rare, this perception has changed recently. In a recent meta-analysis on the epidemiology of EOAD, the percentage of AD patients suffering from EOAD was found to be 5.5%, not 1–2% as previously thought [5]. More importantly, patients with EOAD suffer from a more malignant disease course and face devastating consequences earlier in life than those with late-onset AD (LOAD) [3 , 6–8].

It is not surprising that patients with AD have reduced survival rates than do older people without dementia [9, 10]. However, it is unclear whether the risk of mortality is related to age at disease onset, although the course of EOAD is thought to be more aggressive than that of LOAD. Some studies found that age of onset is not associated with the risk of death in dementia patients [11, 12]. Other studies found that patients suffering from dementia from a younger age have a lower death risk than do people with late-onset dementia, but still have a higher mortality risk when compared to healthy individuals [13, 14]. Other studies showed that dementia patients who develop disease before the age of 65 are at a much greater mortality risk than are patients with late-onset dementia or healthy individuals [2 , 16]. In one study, the mortality rate of patients with early-onset dementia was forty times greater than that of non-demented controls, while that of patients with late-onset dementia was only three times greater than that of controls [2].

The inconsistency of the results detailed above is caused by methodological differences [17]; definitions of survival time, inclusion of subjects who have AD, vascular dementia, and other forms of dementia, and small sample sizes all contribute to the discrepancy. In particular, some studies which investigated the relationship between EOAD and mortality risk compared EOAD patients to non-demented subjects, not to those with LOAD [2]. Overcoming the intrinsic problems inherent in comparing EOAD and LOAD is necessary before analyzing the effect of age of onset on mortality risk [18].

Hence, this study aimed to examine the mortality risk in patients with EOAD compared to that in those with LOAD in a large population. We also applied propensity score matching (PSM) to minimize confounding biases and the intrinsic problems in comparing EOAD and LOAD. Through this analysis, we identify a distinct clinical course of EOAD compared to LOAD.

METHODS

Participants

The Clinical Research Center for Dementia of South Korea (CREDOS) dataset was used for our analyses. The CREDOS study was a nationwide multicenter study designed to assess the occurrence of and risk factors for cognitive deterioration in the elderly. The CREDOS study is registered on ClinicalTrials.gov (identifier NCT 01198093). The study recruited subjects from 31 university-affiliated hospitals from November 2005 to December 2013. All participants and their caregivers underwent comprehensive medical, neurological, and psychiatric interviews at outpatient clinics. Evaluations (physical, psychological, and neurological exams) were conducted at baseline and during each visit in the follow-up period. The subjects in the CREDOS study consisted subjects of Asian ethnicity living in Korea. Of the 14,875 participants in the CREDOS study, 3, 611 subjects were diagnosed with AD according to the criteria set forth by the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) [19], as well as the Diagnostic and Statistical Manual of Mental disorders, Fourth edition (DSM-IV) [20]. Then we defined AD with onset age before 65 as EOAD and AD with onset age over 65 as LOAD. Among the study subjects, 331 were diagnosed with EOAD and 3,280 were diagnosed with LOAD. Through PSM with a ratio of 1 to 2, a total of 936 patients were selected (312 EOAD patients and 624 LOAD patients) from among the initial 3,611 subjects. Subjects with physical or psychiatric disorders that could interfere with the clinical study, such as hearing/vision loss, aphasia, schizophrenia, and mental retardation were excluded. Data on mortality in this study was obtained from the nationwide mortality database of Statistics Korea. In Korea, all deaths are compiled into a national database and managed by Statistics Korea. Each subject in this study was tracked for mortality from cohort entry to December 31, 2013. Cause of death was categorized according to the International Classification of Disease, 10th revision. The Institutional Review Boards (IRBs) of the participating centers approved this study. After the subjects were given a full explanation of the study, written informed consent was obtained from each one.

Measurements

According to van Dijk et al., dementia-related mortality is attributable to the following factors: mortality as a consequence of declining cognitive function, mortality as a result of specific disease processes (AD, vascular dementia), and mortality resulting from comorbid diseases [21]. Thus, it is necessary to comment on the association between mortality and age and sex, the severity of symptoms, and comorbidity in patients with AD. The data included information on patient demographics (age at cohort entry, age at diagnosis of AD, age at symptom onset, sex, and level of education at cohort entry). The event time was defined as the time of death. Survival time was defined as the interval between the initial diagnosis of AD and the date of death. The severity of AD was measured by the Clinical Dementia Rating (CDR) [22]. A Korean version of the Mini-Mental State Examination (MMSE) was used to evaluate general cognitive function [23]. A Korean study defined the cut-off MMSE score that indicates dementia as 17/18 points; the sensitivity and specificity of the findings were 91% and 86%, respectively [23]. Physical disabilities were evaluated by activity of daily living (ADL) [24]. Medical comorbidities included diabetes mellitus, hypertension, cardiovascular disease, stroke, history of brain surgery, thyroid disease, kidney disease, respiratory disease, arthritis, gastritis, history of cancer, alcohol dependence, seizure, osteoarthritis, and history of fracture. These factors determined a dichotomous variable, “yes = 1 or no = 0”, based on information obtained in the baseline interview with patients and caregivers, medical records, and other sources of prescription information. Apathy and delusion also determined dichotomous variable, “yes = 1 or no = 0”, respectively. We assessed the presence of delusion and apathy using subscale of the Korean version of the Neuropsychiatric Inventory. The extent of obesity was identified using body mass index (BMI). Degree of white matter hyperintensity (WMH) was evaluated according to the modified criteria proposed by Fazekas et al. and Scheltens et al. using T2-axial or fluid-attenuated inversion recovery images [25, 26]. CDR, MMSE, and other variables including ADL, BMI, WMH, apathy, delusion, and medical comorbidity was assessed at time of cohort entry.

Statistical analyses

Categorical variables are reported as frequencies and percentages, while continuous variables are reported as means with standard deviations. Discrete variables were compared using Chi-square tests. Independent Student’s t-tests were used to compare the characteristics of patients who developed EOAD and LOAD. We conducted propensity score matching to reduce the bias due to confounding variables related to survival in patients with AD [27, 28]. The propensity score was estimated using logistic regression modeling of the mortality rate of subjects with EOAD based on covariates related to survival in patients with dementia. The 20 variables chosen for propensity score matching were as follows: gender, disease duration, years of education, severity of dementia (CDR), and comorbid conditions (diabetes mellitus, hypertension, cardiovascular disease, stroke, history of head trauma, history of brain injury, thyroid disease, kidney disease, respiratory disease, arthritis, osteoarthritis, gastritis, cancer, alcohol dependence, seizure, and history of fracture). Both discrimination (c-index = 0.701) and calibration (p-value = 0.750, obtained from a Hosmer and Lemeshow Goodness of Fit test) were adequate (Supplementary Figure 1). A 1:2 matching was used to pair patients with EOAD and LOAD [29, 30]. To assess the balance of covariates before and after matching, we analyzed standardized differences. In general, a standardized difference (d) value of less than 10% is desirable, as a greater value of standardized difference may indicate a serious imbalance between the covariates before and after matching [31]. The standardized difference of all covariates was less than 10% after propensity score matching (Supplementary Table 1). Through propensity score matching with ratio of 1 to 2, a total of 936 patients were selected (312 EOAD patients and 624 LOAD patients) from the initial 3,611 subjects. To examine the differential effects of EOAD and LOAD on survival, a Cox proportional hazards model was used after controlling for covariates. Data are presented as hazard ratios (HRs) and 95% confidence intervals (CIs). To deal with methodological issues frequently encountered in studies related to survival time, we also conducted a secondary analysis.

RESULTS

Characteristics of the study population before and after propensity score matching

Before conducting propensity score matching, the study population contained 3,611 patients with AD (331 EOAD patients and 3,280 LOAD patients). Table 1 shows the demographics of the original cohort. The difference in the mean age at diagnosis and age at symptom onset between EOAD and LOAD patients was 18 and 19 years, respectively. Subjects with EOAD were more likely to have moderate to severe dementia and suffer from longer disease duration than were those with LOAD; they were also more apathetic and more functionally disabled. The prevalence of delusion and apathy was 14.9% and 36.1% in this study. The proportion of EOAD patients who reported a family history of dementia, after excluding missing data, was higher (23.2%) than that of LOAD patients (17.2%). After propensity score matching, several characteristics, including educational level, severity of dementia and duration of disease were not different between EOAD and LOAD patients (Table 2). The differences in apathy and functional disability between EOAD and LOAD patients were diminished after matching. Interestingly, WMH as a sign of cerebrovascular burden was more common in LOAD patients and family history of dementia in patients with EOAD marginally increased after matching. Table 3 also showed a higher proportion of deaths “directly” related to dementia rather than to concomitant diseases in EOAD patients after matching (EOAD = 28.1%, LOAD = 15.5%). A total of 161 patients died during the study period. That figure includes 129 LOAD patients and 32 EOAD patients. The causes of death of LOAD patients in order of decreasing frequency were malignancy (17.1%), dementia (15.5%), senescence (13.2%), stroke (11.6%), cardiovascular disease (11.6%), and pneumonia (10.2%). On the other hand, for EOAD patients, the causes of death in order of decreasing frequency were dementia (28.1%), malignancy (12.5%), cardiovascular disease (12.5%), diabetes mellitus (9.4%), and other (9.4%).

Differential effects of EOAD and LOAD on survival after propensity score matching

Table 4 shows Cox proportional hazard regression analyses adjusting for covariates including age at diagnosis, general cognitive function, Barthel index, BMI as a nutritional factor, degree of WMH as a vascular risk factor and delusions and apathy as neuropsychiatric symptoms. In an analysis using the Kaplan-Meyer method without adjusting for age at diagnosis, the impact of LOAD on mortality appeared to be greater than that of EOAD (Log-rank test: p < 0.001). A similar result was also found in an unadjusted Cox proportional hazard regression analysis when we did not control for the effect of aging (unadjusted HR: 0.48, 95% CI: 0.33–0.71, p < 0.001). However, Cox Models 1 and 2, which adjust for the effect of aging effect, showed that patients with EOAD have a higher risk of mortality compared to those with LOAD (HR: 2.01, 95% CI: 1.01–4.00, p = 0.04 in Model 2) (Table 4). When we analyzed the survival curve based on the age-adjusted Cox analysis, the risk conferred by EOAD was higher than that conferred by LOAD (Fig. 1). These results indicate that EOAD, as a discrete subtype of AD, is more malignant than is LOAD when aging is taken into account.

Secondary analysis: Age at diagnosis versus age at first symptom onset

In order to deal with the methodological differences in definition of survival time that are frequently found between studies, we conducted two separate Cox proportional hazard regression analyses according to two types of definition of survival time (Table 5). There was no difference in the results of Cox regression when defining survival time from either age at diagnosis or age at first symptom onset.

DISCUSSION

The main finding of this study was that the mortality risk among patients with EOAD is two times higher than in patients with LOAD when the effects of age on mortality are taken into account. This result supports the idea that EOAD takes a more malignant course than does LOAD. The result was unchanged after adjusting for age at diagnosis, general cognitive function, BMI, physical disability using ADL, WMH as a vascular risk factor, and neuropsychiatric symptoms including apathy and delusions.

Previous studies suggested that EOAD has a negative impact on mortality in an older population, which is in line with our findings [2 , 16]. Koedam et al. compared the mortality risk of 273 patients with early-onset dementia and 620 patients with late-onset dementia versus non-demented older adults in the same age range [2]. They found that the patients with early-onset dementia had a strongly increased mortality risk compared to non-demented patients, while those with late-onset dementia had a moderately increased mortality risk compared to the older controls (HR: 43.3 versus 3.4). Although they did not perform a direct comparison of mortality between patients with early- and late-onset dementia, their results suggested that the impact of dementia on mortality in patients with early-onset dementia might be more severe than it is in those with late-onset dementia. Barkely et al. also found results that are in line with ours through life table analysis [16]: patients with EOAD had a decreased relative duration of survival compared with patients over 65 at disease onset. In the Cache Country Study, Tschanz JT et al. revealed that AD shortened survival time most dramatically in younger participants, while vascular dementia posed a greater mortality risk to older participants. The relative risk of mortality fell from 7.24 at a younger age to 2.20 in those 85 years or older, suggesting that dementia is a greater cause of reduced lifespan in the ‘young-old’ [15].

There may be a possible interpretation about higher mortality in EOAD patients compared to LOAD based on our findings. First, in descriptive analysis, 1) an overrepresentation of deaths “directly” related to dementia rather than to concomitant diseases in EOAD patients. 2) The paucity of signs of cerebrovascular burden and the marginal increase of familial history of dementia was more prominent in patients with EOAD than LOAD. These bring some support to the conclusion that EOAD and LOAD may represent partly different clinical entities. Second, in regression analysis, EOAD itself increased risk of mortality directly, despite we controlled other factors related to survival in EOAD including gender, age of onset, cognitive function, medical and psychiatric comorbidities [6]. The above two findings support the assumption that a burden of dementia itself in EOAD might be more malignant compared with LOAD. Our findings serve to support the theory that EOAD might be pathologically distinct from LOAD. Most studies, which examined differences in neuropathological burden between EOAD and LOAD patients, have reported a greater burden of pathology in EOAD[32, 33].

At first, in our results of Kaplan-Meyer and Cox proportional hazard analysis without adjusting for age at diagnosis, the impact of LOAD on mortality appeared to be greater than that of EOAD. However, when we controlled for the effect of age on mortality, the opposite result was found; EOAD morenegatively affected mortality than did LOAD. Previous study suggested that the mortality risk of people with LOAD is higher than that of patients with EOAD simply because of the advanced age of patients with LOAD, not because of the added risk owing to LOAD itself [34]. EOAD as a distinct disease compared with LOAD takes a more aggressive course, but overall life expectancy may be increasing because patients with EOAD are younger [35]. Similarly, when we controlled for age at onset and the protective effect of being younger disappeared, the mortality risk of EOAD was more prominent than that of LOAD, indicating a more malignant clinical course of EOAD. An alternative explanation for the phenomenon of decreasing mortality with increasing age could be that there is a survival bias, such that patients who are more susceptible to mortality die earlier, and the ones who make it to an older age are less likely to die in general [2 , 37].

One of the reasons for the inconsistent results of previous studies comparing the mortality rate between EOAD and LOAD is methodological differences [2, 17]. These include different definitions of survival time, heterogeneous sample characteristics, and small sample sizes. To minimize such methodological issues, we recruited 3,611 patients diagnosed with AD by neurologists or psychiatrists according to validated criteria and conducted propensity score matching to avoid overfitting in the Cox model. In particular, as we conducted statistical analyses using two different definitions of survival time (measured from either age at diagnosis or age at first symptom onset) [2 , 35], we showed that our findings did not change according to the definition of survival time. Further study to replicate the results using other databases is warranted to confirm this finding.

Our study has some limitations. First, since the subjects who participated in this study were enrolled after seeking medical attention in hospitals, they may not be representative of the general population. Second, we used a clinical diagnosis of AD, as neuropathological diagnoses were not available, which may have been a problem even though all subjects were carefully screened and diagnosed by psychiatrists and neurologists. Third, since autopsy records were not available in this study, autopsy-based diagnostic confirmation was not performed. Future study using Pittsburgh Compound B position emission tomography might be helpful in overcoming this limitation. Fourth, since we collected data on age at first symptom onset that depended on the subjects’ or their caregivers’ memories, a recall bias is most likely present to some extent. Because of the above limitation, ‘survival time’ was calculated from the date of initial diagnosis of AD while the first symptoms of the disease may have started years before. We thought that the use of the date of diagnosis of AD is preferable as a starting point, as a clearly defined point in time by professionals. Fifth, this study might be limited by the possible occurrence of overcorrected bias related to age after matching. EOAD and LOAD patients were an originally discrete mean age difference according to their definitions. Further study is warranted to explore health survival effect and competing risk of death without AD-related death.

In summary, Korean EOAD patients showed a greater risk of mortality compared to their counterparts with LOAD through PSM. This might indicate that the EOAD and LOAD represent two distinct AD phenotypes. Moreover, the more malignant disease course of EOAD is strictly related to the AD itself rather than the comorbidities. Prompt identification and management of EOAD is important to decrease the risk of mortality in dementia patients.

Footnotes

ACKNOWLEDGMENTS

This study received funding from the Korea Healthcare Technology Research and Development Project, Ministry of Health & Welfare, Republic of Korea (HI10C2020).

This work was also supported by the National Research Foundation of Korea (NRF) Grant funded by the Korean Government (MSIP) (No.2015R1A5A7037630)

Authors’ disclosures available online ().

Appendix

References

Harvey

, Skelton-Robinson

, Rossor

(2003) The prevalence and causes of dementia in people under the age of 65 years. J Neurol Neurosurg Psychiatry 74, 1206–1209.

Koedam

, Pijnenburg

, Deeg

, Baak

, van der Vlies

, Scheltens

, van der Flier

(2008) Early-onset dementia is associated with higher mortality. Dement Geriatr Cogn Disord 26, 147–152.

Kim

, Cho

, Jeong

, Park

, Kang

, Kim

, Lee

, Na

(2005) Glucose metabolism in early onset versus late onset Alzheimer’s disease: An SPM analysis of 120 patients. Brain 128, 1790–1801.

Koss

, Edland

, Fillenbaum

, Mohs

, Clark

, Galasko

, Morris

(1996) Clinical and neuropsychological differences between patients with earlier and later onset of Alzheimer’s disease: A CERAD analysis, Part XII. Neurology 46, 136–141.

Zhu

, Tan

, Wang

, Jiang

, Cao

, Wang

, Tan

, Meng

, Yu

(2015) Rate of early onset Alzheimer’s disease: A systematic review and meta-analysis. Ann Transl Med 3, 38.

Ueki

, Shinjo

, Shimode

, Nakajima

, Morita

(2001) Factors associated with mortality in patients with early-onset Alzheimer’s disease: A five-year longitudinal study. Int J Geriatr Psychiatry 16, 810–815.

Chui

, Teng

, Henderson

, Moy

(1985) Clinical subtypes of dementia of the Alzheimer type. Neurology 35, 1544–1550.

Seltzer

, Sherwin

(1983) A comparison of clinical features in early- and late-onset primary degenerative dementia. One entity or two? Arch Neurol 40, 143–146.

Guehne

, Riedel-Heller

, Angermeyer

(2005) Mortality in dementia. Neuroepidemiology 25, 153–162.

10.

Helmer

, Joly

, Letenneur

, Commenges

, Dartigues

(2001) Mortality with dementia: Results from a French prospective community-based cohort. Am J Epidemiol 154, 642–648.

11.

Newens

, Forster

, Kay

, Kirkup

, Bates

, Edwardson

(1993) Clinically diagnosed presenile dementia of the Alzheimer type in the Northern Health Region: Ascertainment, prevalence, incidence and survival. Psychol Med 23, 631–644.

12.

Kay

, Forster

, Newens

(2000) Long-term survival, place of death, and death certification in clinically diagnosed pre-senile dementia in northern England. Follow-up after 8-12 years. Br J Psychiatry 177, 156–162.

13.

Ganguli

, Dodge

, Shen

, Pandav

, DeKosky

(2005) Alzheimer disease and mortality: A 15-year epidemiological study. Arch Neurol 62, 779–784.

14.

Larson

, Shadlen

, Wang

, McCormick

, Bowen

, Teri

, Kukull

(2004) Survival after initial diagnosis of Alzheimer disease. Ann Intern Med 140, 501–509.

15.

Tschanz

, Corcoran

, Skoog

, Khachaturian

, Herrick

, Hayden

, Welsh-Bohmer

, Calvert

, Norton

, Zandi

, Breitner

, Cache County Study G (2004) Dementia: The leading predictor of death in a defined elderly population: The Cache County Study. Neurology 62, 1156–1162.

16.

Barclay

, Zemcov

, Blass

, McDowell

(1985) Factors associated with duration of survival in Alzheimer’s disease. Biol Psychiatry 20, 86–93.

17.

Todd

, Barr

, Roberts

, Passmore

(2013) Survival in dementia and predictors of mortality: A review. Int J Geriatr Psychiatry 28, 1109–1124.

18.

Park

, Choi

, Park

, Kim

, Lee

, Han

, Kim

, Han

, Moon

, Yang

, Park

, Yoon

, Seo

, Na

, Lee

(2015) Cognitive profiles and neuropsychiatric symptoms in Korean early-onset Alzheimer’s disease patients: A CREDOS study. J Alzheimers Dis 44, 661–673.

19.

McKhann

, Drachman

, Folstein

, Katzman

, Price

, Stadlan

(1984) Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939–944.

20.

American Psychiatric Association (1994) Diagnostic Statistical Manual of Mental Disorders, 4th edition. American Psychiatric Association, Washington, DC.

21.

van Dijk

, Dippel

, Habbema

(1991) Survival of patients with dementia. J Am Geriatr Soc 39, 603–610.

22.

Morris

(1993) The Clinical Dementia Rating (CDR): Current version and scoring rules. Neurology 43, 2412–2414.

23.

Kang

, Na

, Hanhn

(1997) A validity study on the Korean Mini-Mental State Examination (K-MMSE) in demented patients. J Korean Neurol Assoc 15, 300–307.

24.

Sinoff

, Ore

(1997) The Barthel activities of daily living index: Self-reporting versus actual performance in the old-old (>or =75 years). J Am Geriatr Soc 45, 832–836.

25.

Fazekas

, Chawluk

, Alavi

, Hurtig

, Zimmerman

(1987) MR signal abnormalities at 1.5 T in Alzheimer’s dementia and normal aging. AJR Am J Roentgenol 149, 351–356.

26.

Scheltens

, Barkhof

, Leys

, Pruvo

, Nauta

, Vermersch

, Steinling

, Valk

(1993) A semiquantative rating scale for the assessment of signal hyperintensities on magnetic resonance imaging. J Neurol Sci 114, 7–12.

27.

Rubin

, Thomas

(1996) Matching using estimated propensity scores: Relating theory to practice. Biometrics 52, 249–264.

28.

Rubin

(1997) Estimating causal effects from large data sets using propensity scores. Ann Intern Med 127, 757–763.

29.

Austin

(2010) Statistical criteria for selecting the optimal number of untreated subjects matched to each treated subject when using many-to-one matching on the propensity score. Am J Epidemiol 172, 1092–1097.

30.

Rassen

, Shelat

, Myers

, Glynn

, Rothman

, Schneeweiss

(2012) One-to-many propensity score matching in cohort studies. Pharmacoepidemiol Drug Saf 21(Suppl 2), 69–80.

31.

Normand

, Landrum

, Guadagnoli

, Ayanian

, Ryan

, Cleary

, McNeil

(2001) Validating recommendations for coronary angiography following acute myocardial infarction in the elderly: A matched analysis using propensity scores. J Clin Epidemiol 54, 387–398.

32.

Marshall

, Fairbanks

, Tekin

, Vinters

, Cummings

(2007) Early-onset Alzheimer’s disease is associated with greater pathologic burden. J Geriatr Psychiatry Neurol 20, 29–33.

33.

Boekhoorn

, Joels

, Lucassen

(2006) Increased proliferation reflects glial and vascular-associated changes, but not neurogenesis in the presenile Alzheimer hippocampus. Neurobiol Dis 24, 1–14.

34.

van der Flier

, Pijnenburg

, Fox

, Scheltens

(2011) Early-onset versus late-onset Alzheimer’s disease: The case of the missing APOE varepsilon4 allele. Lancet Neurol 10, 280–288.

35.

, Lee

, Seo

, Chin

, Kang

, Moon

, Na

, Cheong

(2013) Survival of Alzheimer’s disease patients in Korea. Dement Geriatr Cogn Disord 35, 219–228.

36.

Rountree

, Chan

, Pavlik

, Darby

, Doody

(2012) Factors that influence survival in a probable Alzheimer disease cohort. Alzheimers Res Ther 4, 16.

37.

Wolfson

, Wolfson

, Asgharian

, M’Lan

, Ostbye

, Rockwood

, Hogan

, Clinical Progression of Dementia Study, Group (2001) A reevaluation of the duration of survival after the onset of dementia. N Engl J Med 344, 1111–1116.