Background
A systematic analysis for the Global Burden of Diseases, Injuries, and Risk Factors Study 2016 reported that the number of people worldwide living with dementia has more than doubled from 1990 to 2016 [
1]. According to Swedish register data, there are an estimated 150,000 dementia patients and approximately 24,000 individuals develop dementia each year [
2]. Lower education, hypertension, hearing impairment, smoking, obesity, depression, physical inactivity, diabetes, low social contact, excessive alcohol consumption, traumatic brain injury, and air pollution were recently identified in the 2020 report of the Lancet Commission [
3] as 12 common risk factors for incident dementia. However, because current medical treatment cannot cure or reverse dementia, but only alleviate the symptoms, identifying new potential modifiable risk factors is crucial in preventing or delaying the onset of dementia.
Meta-analyses and systematic reviews have reported that frailty predicts incident dementia [
4,
5]. The studies have, however, almost exclusively focused on individuals aged 65 and older, leaving younger and middle-aged adults understudied. The association appears robust regardless of the scale used to measure frailty; both the frailty phenotype (FP) [
6] and frailty index (FI) [
7] are predictive of dementia. However, the mechanisms underlying the association are unclear. Frailty and dementia share several risk factors and clinical manifestations [
8] and may even share a common pathological basis [
9]. It is also possible that some genetic variants increase the risk of both conditions or modify the effect of environmental and lifestyle-related factors, such as education, smoking, and physical activity, on frailty and dementia. Twin studies have shown that Alzheimer’s disease, the most common form of dementia, is highly heritable [
10], with the
apolipoprotein E (
APOE) ɛ4 allele being the strongest genetic risk factor [
11]. Although the genetics of frailty are less well understood, twin studies have estimated that the heritability of the FI ranges from 30 to 52% [
12,
13], and a genome-wide association study has indicated that variants in brain pathways underlie the risk of frailty [
14].
In this study, we sought to address the gaps in current evidence by analyzing whether frailty predicts dementia in a large cohort of twins including younger and older adults and discerning whether the association is explained by genetic and/or shared environmental factors. As monozygotic (MZ) twins can be considered genetically identical, whereas dizygotic (DZ) twins share on average 50% of their segregating genes, associations within twin pairs discordant for dementia can inform about the involvement of genetic and shared environmental factors [
15]. Should there be no involvement (“confounding”) of genetic and shared environmental factors, the effect size of the exposure (frailty) observed in the population-level analysis would have to persist in the within-pair analysis in both DZ and MZ twins. Such a scenario would be consistent of at least partially causal relationship between frailty and dementia, making frailty a potential target for the prevention of dementia. A similar attenuation of the effect in both DZ and MZ twins would indicate the involvement of shared environmental factors, including but not limited to early life exposures and lifestyle-related factors. The familial environmental factors are thus all types of “anonymous” influences that make the twins similar to each other—even in later life. Further attenuation of the effect in MZ twins relative to DZ twins would indicate that shared genetic influences explain the association between frailty and dementia as genetic effects are fully accounted for in the within-MZ pair analysis. In such a scenario, intervening the exposure (decreasing frailty) is unlikely to lead in the effective improvement of the outcome (preventing dementia).
Discussion
This study assessed the FI-dementia association in middle-aged and older individuals during a 19-year follow-up in a large cohort of twins and analyzed whether familial effects (shared environment and genetics) affect the association. In the full sample, a 10% increase in the FI was associated with a 17% increase in the risk of dementia, after adjusting for age, sex, education, and tobacco use. In the cognitive sample, a 10% increase in FI associated with a 14% increase in the risk of dementia, after adjusting for age, sex, and baseline cognitive level. The FI-dementia association remained significant after further adjustment for APOE ɛ4 carrier status and when considering the competing risk of death. After controlling for familial factors in within-pair analyses and taking into account an age-varying (age at FI measurement) risk of frailty on dementia, we could neither find evidence that the effect sizes in DZ and MZ twin pairs significantly differed from the population-level estimates, nor that the MZ estimate significantly differed from the DZ estimate. These findings would suggest that familial factors, including genetics, do not account for the association between frailty and dementia. The risk carried by increased frailty on dementia was seemingly constant after age 50 until very old age. A sensitivity analysis using the FI-NTRF that was stripped from traditional dementia risk factors showed the association between frailty and dementia remained significant.
Our study adds to the current understanding of the relationship between frailty and dementia by showing that it is independent of familial factors, i.e., everything that makes the family members similar to each other, even in old age. This means that in identical twins, the twin with a higher level of frailty is more likely to develop dementia regardless of their similar genetic propensities and shared environmental risks, suggesting that the association between frailty and dementia is consistent with a causal hypothesis. Such study designs have not been used before to assess the relationship between frailty and dementia. Another novel contribution is the finding that the risk conferred by a higher level of frailty is similar in magnitude in midlife as in old age, suggesting that screening for frailty before or at midlife might provide benefit in identifying at-risk individuals. In the light of our previous finding that the rate of increase in frailty more than doubles after age 65 [
32], extending clinical frailty assessments to younger age groups might indeed be beneficial in tackling frailty before it escalates and makes interventions less likely to succeed.
Meta-analyses and several observational studies [
5,
33‐
35] have showed a significant association of frailty with the risk of dementia, independent of other risk factors and regardless of the scale used to measure frailty. Our study is in line with these findings as we observed that higher FI is associated with an increased risk of all-cause dementia after adjusting for age, sex, education, tobacco use, and cognition. Similar to Rogers et al. [
34] who reported an association between higher FI and incident dementia in a competing risk model in the English Longitudinal Study of Aging, we found that the association remained significant after accounting for the competing risk of death. Regarding the role of the
APOE genotype, our study is the second to find that the FI is associated with incident dementia independent of the
APOE ε4 carrier status, consistent with Ward et al. [
36] who recently found that the risk effect of the FI on incident dementia was similar in both carriers and non-carriers of the
APOE ε4 allele. Lastly, similar to the findings by Rogers et al. [
34], we found that the risk of frailty increased in a dose-response manner across the categorized FI (non-frail, pre-frail, and frail).
Although the association between frailty and dementia appears robust and seemingly independent of other risk factors, the role of familial environmental factors and genetic predisposition, other than APOE, in the association has not been studied. Using a unique twin design, our study assessed the impact of such environmental and genetic factors. Comparing the population-level estimates to the within-pair estimates in DZ twins in the full and cognitive samples, we did not find evidence of attenuation of the effect, indicating that shared environmental factors do not explain the frailty-dementia association. When further analyzing the difference between the DZ and MZ twin estimates in the within-pair analysis—that for the MZ twins fully adjusts for genetic factors—we found that although the DZ estimate appeared higher in the full sample and in the age-varying analysis, the difference was not statistically significant. Finally, when comparing the within-MZ twin estimate with the population estimate, the estimates were very similar and not statistically significantly different. This finding would lend support to the hypothesis that genetic factors do not explain the FI-dementia association either; note, though, that the within-pair analyses yielded estimates with wide confidence intervals, which indicates that inferences should be made with caution. Should further studies corroborate our findings that frailty predicts dementia independent of familial factors, frailty would be an actionable target for preventing or delaying dementia.
To the best of our knowledge, there are currently no trials aimed at preventing or delaying dementia through reversing or slowing down the progression of frailty. However, better management of frailty may contribute favorably to the prevention of dementia in several ways. Individuals living with frailty or at the risk of becoming frail can be subjected to medication reviews and fall prevention measures, decreasing the odds of escalating frailty and its negative sequelae that pave the way for dementia. The multidimensional nature of frailty nevertheless creates necessities for more comprehensive approaches, such as multicomponent interventions. A 6-month intervention based on group exercise, nutritional supplementation, depression management, deprescribing medications, and home hazard reduction showed sustained beneficial effects on frailty up to 1 year [
37]. Moreover, as frailty can present both in the absence and presence of multimorbidity, disease management may be more beneficial to those living with comorbidities, while slowing down cellular aging might benefit those whose frailty is driven by accelerated biological aging rather than age-related diseases [
38]. A geroscience hypothesis posits that targeting fundamental aging processes at the cellular level might delay the onset or severity of multiple chronic diseases and frailty [
38,
39]. While still in its infancy, senolytic treatments to clear senescent cells and manipulation of signal transduction pathways linked to metabolism and nutrient sensing might be such geroscience approaches that have the potential to mitigate frailty [
38,
39]. Another emerging approach for frailty therapeutics is drug repurposing in which existing drugs are used for new therapeutic targets. Our group has recently shown that lipid-lowering therapeutics might decrease the odds of frailty in midlife and older age [
40]. Nevertheless, even if the above approaches showed beneficial effects on frailty, they are likely non-specific such that they affect the risk of dementia, too. As frailty and cognitive decline develop over a long period of time and share a common pathologic basis [
9], finding approaches that specifically target frailty might be challenging. However, as we found that the relationship between the FI and the risk of dementia is linear, with even lower levels increasing the risk, targeting frailty long before dementia is manifest might provide a specific time window for future therapeutic approaches.
Our study also sheds light into the age-varying and sex-specific risk of frailty on dementia. In the within-pair analysis—that adjusts for covariates and familial factors—we found that the risk was seemingly constant from age 50 to 90. This finding indicates that even though the risk of dementia increases with age, the risk conferred by increased frailty is similar in magnitude from midlife into old age. In the sex-stratified analysis, the risks appeared similar in men and women, although in MZ men the effect size of FI decreased and attenuated to null—a finding that can be attributed to a small sample size or true null finding.
This study has several strengths. Firstly, we used a large, genetically informative sample of twins with a wide age range and a long follow-up to dementia (up to 19 years). Secondly, the role of familial factors in the frailty-dementia association has not been studied before, and our study provides a unique opportunity to do so. Thirdly, the BW model used in the within-pair analysis is a robust approach that typically produces a more powerful test of the within-pair effect than a stratified conditional Cox regression [
22]. One of the limitations in our study is that the FI was measured based on self-reported items of which a large proportion were medical conditions, leading to potential misclassification and a FI that is skewed towards comorbidities. Some of the comorbidities are also known risk factors for dementia, potentially driving the association beyond the construct of frailty itself. Nevertheless, the estimates for FI-NTRF and FI-TRF were similar. In addition, pertinent to all aging studies, a relatively large number of individuals were censored due to death, leading to a limited number of informative MZ twin pairs, especially in the sex-stratified analysis. Furthermore, within-pair estimates, especially for MZ twins, were imprecise with wide confidence intervals, which warrants caution in conclusions based on the study. For example, the within-MZ pair estimate does not only cover the population estimate, it also covers the null (i.e., a HR of 1), and the results are thus compatible with the FI-dementia association being explained by the factors shared within MZ pairs (e.g., genetics). Nevertheless, the within-MZ pair estimate does not significantly differ from the population-level estimate. Lastly, for those dementia diagnoses that were obtained from the NPR, there is a possibility of uncertainty in the timing of the onset of dementia as the diagnoses in the NPR are recorded approximately 5 years after the age of onset [
41].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.