Background
The prevalence of dementia in centenarian studies varies widely from 27 % (or 42 % once drop-outs were accounted for [
1]) to 76 % [
2] and even 85 % (albeit in a small sample [
3]). Reasons for this variability include small sample sizes, non-ascertainment of all centenarians within a selected region, the healthy volunteer effect, non-inclusion of residents in skilled nursing facilities, refusal of proxy-consent by ‘protective’ family members, frequent shift in residence owing to care needs, and other potential biases. Longitudinal studies suffer the limitation of selective attrition, particularly due to high mortality in such an elderly sample. Additionally, not all studies demand adequate proof of age, a problem particularly relevant when the claimed age is greater than 110 years old ([
4,
5]).
In the context of dementia diagnosis, accurate cognitive assessments of centenarians can be challenging due to decreased stamina and difficulty with hearing and/or vision, and most studies, at least to date, have not used an adequately sensitive battery of tests to rule out false negative results. Differences in cognitive assessment tools, variability in diagnostic criteria, and difficulty in selecting an appropriate comparison group are further challenges in comparing and/or merging results from different studies. Birth year cohort-specific influences are likely to also explain differences between studies, with accumulating evidence that the age-specific incidence of dementia may be decreasing in high income countries [
6]. Cho et al. [
7] found that a later cohort of centenarians was significantly better off in mental, physical, social and economic domains. A recent UK study found later-born individuals to have a lower risk of dementia [
8]. Steen [
9] found similar cohort effects in five different cohorts of 70 year olds, with later cohorts performing significantly better than their earlier counterparts, although a Swedish study [
10] found no differences in cognitive function between two cohorts. Influences of environment, such as climate [
11] and geography, including rural versus urban [
12], may need to be taken into consideration when examining dementia prevalence and its determinants in different populations. Dietary factors may also play a role, however data remain insufficient to date.
Given the limitations of prevalence data, it can be argued that incidence data on dementia might offer a better metric to examine the cognitive profile of this age group. However, only a handful of studies have provided incidence data on dementia in the oldest-old [
13], most of which have few participants at the oldest ages and sometimes all participants aged over 90 years are combined into a single age category. A review, with sufficient age-specific data, contends that dementia incidence increases in the age range of 95–115 years [
13]. A North American study, with baseline ages of 90–103, also observed an increase in dementia incidence with age [
14]. Other studies that include participants aged 95 or over, observed a
slowing of the age-related increase in incidence of dementia [
15,
16], and a decline in rates for men [
17,
18]. Finally, a study that included the oldest 0.01 percentile (e.g. currently 110+ years old) has demonstrated the progressive compression of both disability and morbidity (in 6 diseases including dementia) with older ages of survival beyond 100 years [
19]. Given these different results further examination of the incidence of dementia for males and females for different age groups among the oldest ages is warranted.
Another approach taken by investigators is to examine specific cognitive functions in this group. Although relatively unexplored in the oldest old, it appears that episodic memory continues to decline through the 10th and 11th decades of life [
20] with processing speed [
21] and attention [
22] being particularly susceptible to ageing. By contrast, many aspects of language [
23] as well as executive functions [
24] may remain intact with increasing age, although the data are limited by small sample sizes. The domains of cognitive function most susceptible to advanced ageing, and their magnitude of decline, have important bearing on the differentiation of the transition from normal cognition to mild cognitive impairment and dementia. In general, the patterns appear similar to cognitive changes observed in younger cohorts [
25], although much more extensive evaluation of this dynamic is needed.
Some research has been conducted on risk and protective factors of dementia onset in centenarians, but little is known about the course and rate of this decline [
26]. The limited data available suggest that some exceptionally long-lived individuals share risk factors for dementia with their younger counterparts [
26], such as African American race [
27], low education [
27], smoking [
28], and poor physical health as evidenced by strength, balance and gait measures [
29]. Conversely, a well-known genetic risk factor for Alzheimer’s disease, the apolipoprotein E ε-4 allele, is rare amongst centenarians [
30,
31]. Similarly, a number of risk factors for cardiovascular disease, although consistent in predicting cognitive impairment in younger cohorts [
32], have different effects on cognition in older populations [
33]. Interestingly, a number of studies recently demonstrated that centenarians have frequencies of disease-associated genetic variants that are similar to the general population [
34,
35]. Yet, as noted above, people achieving ages over 105 years tend to avoid or delay such age-related diseases [
19]. Bergman and colleagues [
36] noted this “buffering effect” of certain genes previously, however limited research exists on the relation of this effect to dementia.
Most existing studies of the oldest old are necessarily small, limiting the power to appropriately examine prevalence and incidence data, cognition and risk factors. Data harmonisation across numerous studies is a cost-effective approach with increased statistical power that offers the potential to explore both existing and novel research questions. Harmonising data across studies to create a single, large database permits evaluation of both study-level and individual-level effects, and the direct comparison of results across studies with opportunity for immediate evaluation of differences, when found, and additional analyses to reconcile such differences [
13,
27]. It is important to note that the provenance and contextual information of each study must be taken into account in any such analysis. Other benefits of collaboration include accelerated accumulation of scientific knowledge and enhanced generalisability associated with using a larger heterogeneous sample [
13].
In the proposed consortium of centenarian and near-centenarian studies, comprising at least fifteen datasets from Asia, Europe, the Americas, and Oceania, The Dementia Harmonisation Project of the International Centenarian Consortium (ICC-Dementia
https://cheba.unsw.edu.au/group/icc-dementia) plans to address the epidemiological challenges confronting the study of this exceptional group of individuals. The Consortium’s objectives are to: (i) determine the sex specific and percentile of survival-standardised prevalence and incidence of dementia and likely dementia type at the extreme tail of survival; (ii) delineate subgroups of cognitive function and their specific patterns of cognitive decline; (iii) identify associated risk and protective factors for dementia and healthy brain ageing that cross or do not cross ethnic and cultural lines; and (iv) examine the influence of contextual factors such as population ageing, survival rates and differential causes of death in different countries on the cognitive health of the oldest old.
Competing interests
The Sydney Centenarian Study is supported by the Centre for Healthy Brain Ageing and Dementia Collaborative Research Centre, University of New South Wales, Sydney, Australia. This study was supported by the National Health & Medical Research Council of Australia (project grant ID 630593 and program grant ID 568969; PS as principal investigator) The Oregon Brain Aging Study is supported in part by grants from the US Department of Veterans Affairs and the US National Institutes of Health (NIH), National Institute on Aging (P30 AG08017). The Polish Centenarian Study is funded by the Medical University of Silesia, Katowice, Poland and by the National Science Centre, Poland (Grant N 404 535439) from the budget for science during the years 2010–2014. The Georgia Centenarian Study is supported by the NIH (Grant PO1 AG17553-01A1; LWP as principal investigator). The Gothenburg 95+ Study was funded by the Swedish Research Council, Swedish Research Council for Health, Working Life and Welfare, and the Alzheimer’s Association. The 100-plus Study is funded by Alzheimer-Nederland, Dioraphte and various private donations. The Monzino 80-plus Study is supported by a research grant from the Italo Monzino Foundation, Milano, Italy. The Tokyo Centenarian Study is supported by a grant from the Japanese Ministry of Health, Labour and Welfare. The Cognitive Function and Ageing study was funded by major awards from the Medical Research Council: Research Grant [G9901400] and the UK Department of Health. The funding bodies played no role in the formulation of the design, methods, subject recruitment, data collection, analysis, or preparation of this manuscript.
Authors’ contributions
PS and HB conceptualized, led the project and critically revised the manuscript. CW, CD and MS collected and managed the data from all the sites. CW drafted the first version of the manuscript. CD drafted the second version. LP was involved in the discussion of the study design. LP, TP and IS critically revised the initial and final versions of the manuscript. JC provided statistical advice. NK analysed and advised on neuropsychological test measures. All other authors outside Australia led studies and were responsible for data collection in their countries. NB, CB, KC, YG, BH, NH, CK, JK, UL, GM, PM, JMR, JS, MT and JV provided feedback on earlier drafts. All authors, including SA, MC, HH, BL and RR, read and approved the final manuscript.
Claudia Woolf and Melissa J. Slavin: Formerly affiliated, now elsewhere.