Background
The number of people with dementia worldwide is rising within the ageing population, which is a severe issue in all countries. The World Health Organization (WHO) reports that there are almost 50 million people currently living with dementia, and nearly 10 million new cases each year [
1], with the highest growth area being in China and some other developing nations [
2]. According to the China Mental Health Survey, the prevalence of dementia for the population aged 65 and over is 5.6% diagnosed by the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV [
3‐
5]. Moreover, the prevalence is higher if using the 10/66 diagnostic package [
6], which is more sensitive for detecting the early stages of dementia.
Because of the limited availability of therapeutic medications for dementia, it is particularly important to identify its risk factors, which can help policy makers distinguish and intervene in high-risk cases. Therefore, considerable attention to risk factors for dementia has been sustained over the last two decades. Age, smoking, hypertension, high cholesterol, and apolipoprotein E are considered likely risk factors for dementia [
7]. In addition, there is a consensus that the risk factors of incident dementia occurrence likely cluster with lifestyles, chronic diseases, and physical impairment [
8‐
11]. Moreover, in a meta-analysis on dementia, a relationship was found between exposure to increased factor load and subsequent poorer cognitive function or dementia [
12].
Some researchers have suggested integrating risk factors into definition of frailty, which is a comprehensive indicator for evaluating its effects on dementia incidence [
13]. Frailty is an age-related syndrome with the main features of impaired multisystem physiological reserves and decreased resistance to substantial stressors [
14].
It was estimated that frailty increases the likelihood of future cognitive decline and that cognitive impairment elevates the risk of frailty. However, the mechanism of the potential association between frailty and incident dementia is unclear. It has been shown that physical frailty increases the risk of diabetes mellitus and cardiovascular disease [
15,
16]. Moreover, a clinical-pathologic study found physical frailty to be associated with dementia pathology [
17‐
19], and alterations in hippocampal synaptic function, neuronal membrane properties, and axonal trajectories, which might lead to dementia have been reported for people with frailty [
20‐
22].
Preliminary studies on frailty in China have focused on the association between frailty and death. Dupre reported the relationship between frailty, type of death and suffering death, using the frailty index comprising 39 deficits [
23]. The most frequently used frailty assessment methods in China are the Fried phenotype and the frailty index, which are similar to those in other countries. For community-dwelling people aged 65 or over in China, the prevalence of frailty is 5.9%–17.4% [
24], close to that in developed countries. Nonetheless, very few studies on the relationship between frailty and dementia have been carried out in China. In addition, the prevalence of frailty varied in different studies due to the assessment method employed, and previous studies have shown that the variables selected for assessment methods differwith regard to study aim.
Discussion
The study is one part of the series of studies on dementia by the 10/66 International Research Group. To date, very few studies have explored the relationship between frailty and dementia. Based on three different definitions of frailty, we explored the probable influence of frailty on the incidence of dementia in elderly people in a Beijing sample. The results indicated that the participants with multidimensional frailty had a higher risk than did non-frail participants and that the size of the effect varied across different age groups. This finding is consistent with previous studies showing that frailty is a risk factor for dementia [
35]. The uniqueness of this study is the fact that the competing risk of death was considered when exploring the influence of frailty on incident dementia in China [
36], which can result in accurate HRs, despite a limitation to using those experiences directly because of the difference between China and other Western countries. Although the results of the three definitions differ slightly, frailty can generally increase the hazard of incident dementia.
The modified Fried frailty phenotype, which was used in previous frailty studies by 10/66 collaboration group, includes four physical variables, and this definition has been confirmed as a good predictor for dependence and mortality in older adults [
30]. Regardless, the results did not prove the association between the modified Fried frailty phenotype and dementia after adjusting for age, gender and education level. In other words, the modified Fried frailty phenotype could not accurately predict the risk of incident dementia in this Chinese sample. To further evaluate the association between physical frailty and incident dementia, nine additional physical variables, including finger movement disorder, decreased muscular tension, abnormal tendon reflex, eyesight problems, hearing difficulty or deafness, dysdiadochokinesis, waist-hip ratio, sleeping disorders, and slow action, were added to the modified Fried frailty phenotype forming the PFI, but the result was similar to that of the modified Fried frailty phenotype. The association between PFI and incident dementia was only found in the unadjusted model. Previous population-based studies have reported that physical frailty might be associated with vascular dementia, which accounts for 15%~ 20% of all-cause dementia [
37]. As only all-cause dementia was considered in this study, the association between it and physical frailty was not as expected. It is assumed that physical frailty plays different roles in predicting the incidence of different types of dementia. To explore this question more deeply, two age-group analyses were conducted to assess the relationship between frailty and incident dementia.
The mental status and social relations of elderly individuals were added, including impairment of general cognitive ability, delayed recall problems, lack of concentration, lack of interest, delayed recall, depressive emotions, loneliness, and insufficient social support, to the definition of multidimensional frailty. The results showed that multidimensional frailty is an independent predictor for incident dementia and associated with a 47% increased hazard in the competing risk model after adjusting for age, gender, education level, physical multimorbidity, walking distance per week, alcohol use and smoking status. Most previous studies on frailty have only focused on the physical dimension [
16,
17,
38], though there is some evidence showing a relationship between mental status and social relations and incident dementia. In a Japanese dementia study, Shimada interviewed 4570 community-dwelling elderly persons and found that inclusion of a cognitive dimension in the frailty assessment improved the predictive accuracy [
39]. This finding suggests that the identification of high-risk populations of dementia should not be limited to one dimension. A better identification should include multidimensional risk factors for identifying high-risk populations and providing advanced intervention. The report of the Lancet International Commission recommended that dementia prevention include more childhood education, exercise, and management of hearing loss and depression [
40]. In addition, Rogers derived a frailty index comprising 47 health factors and found that frailty should be considered alongside cognitive function when assessing risk factors of dementia [
41]. Furthermore, it has been shown conclusively that cognitive impairment is a potential component of the secondary prevention of dementia [
42]. There are several possible explanations for this result. First, frailty is a condition that is characterized by a reduction in the homeostatic reserves of the individual, which entails an increased vulnerability to endogenous and exogenous stressors [
43]. Hence, frail people face more challenges in adapting to significant threats that may occur in that individual’s life. In other words, frail people’s resilience may decline, which will increase the likelihood of dementia. Second, frailty and dementia might have a similar aetiological pathway because both are related to cognitive decline [
41]. The aetiology of frailty is multidimensional, including nutrition, physical activity, coping, social support and other potential causes [
44,
45]. Inflammation, oxidative stress and cortical atrophy are three factors that play essential roles in the development of both dementia and frailty [
46]. Previous studies reported that in a population of cognitively declining patients, frailty is associated with the presence, degree, and localization of cerebral atrophy. For every 1% increase in FI, the probability of cortical atrophy increased by 2% [
43].
In a long-term follow-up cohort study, participants may become deceased during the follow-up period, especially those who are frail. In this study, the mortality of the participants was almost one-fifth, and these censored data can lead to bias. Therefore, competing risk models were applied to analyse death as a competing event to dementia, which decreased the bias of non-dementia premature death in the association of frailty and incident dementia. Table
3 shows that multidimensional frailty was more highly associated with incident dementia after controlling for age, gender, education level, physical multimorbidity, walking distance per week, smoking status, and alcohol use. The HRs of competing risk hazard models were lower than those of the Cox proportional hazards model. The results of this study suggest that the Cox proportional hazards model might overestimate the risk of incident dementia but that the competing risk hazard model is able to calculate unbiased and accurate outcomes.
According to the data in Table
4, frailty was not associated with incident dementia in the lowest quartile age group of the participants at baseline. In contrast, the other group had a significantly higher risk of incident dementia if they met the definition of multidimensional frailty at baseline. Thus, there was a positive association between age and multidimensional frailty. The gradual decline in physiological reserve with ageing accelerates faster in frail people than in non-frail people [
47]. This observation is also in-line with other studies, showing that the risk of incident dementia increases with decreasing physiological reserve in elderly individuals [
48].
There are two limitations in this study. First, grip strength information was not collected at the baseline, so the models of this study did not analyse grip strength as a variable of frailty. But previous studies of 10/66 group have suggested that the link between grip strength and adverse health conditions weakens when other frailty indicators and confounding factors are accounted for [
49], although grip strength was one of the five indicators of Fried frailty phenotype. Therefore, this study was consistent with former 10/66 dementia study [
30]. Second, this cohort study was not nationally representative, so the external validity of prevalence and cumulative incidence was not as good as the internal validity. However, this study did focus on exploring the association between frailty and dementia, rather than on the prevalence of dementia in the national population.
In the future, additional studies are needed to confirm the associations between subtypes of dementia and frailty to determine potential causes of dementia to prevent elderly people from developing the condition. Other important comorbidities, such as hyperlipidaemia, stroke, or cancers, require further analyses. Moreover, to explain the mechanism between frailty and dementia in dementia studies, biomarkers should be considered vital mediators for predicting the condition. Future cohort studies should take into consideration psycho-social and biological factors to build a more accurate prediction model.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.