Humans and other animals are mortal, and the risk of mortality in all organisms increases exponentially with age (Ledberg, Reference Ledberg2020). At least in humans, impairments of both the body and mind – studied either individually or as a collective – are linked to increases in mortality with aging (Soysal et al., Reference Soysal2017). As a collection of impairments, frailty – thought of either as a clinical syndrome (Fried et al., Reference Fried2001) or as a state (Mitnitski et al., Reference Mitnitski, Mogilner and Rockwood2001) – is associated with complex age-related outcomes, such as dementia (Ward et al., Reference Ward, Wallace and Rockwood2020) and death (Theou et al., Reference Theou, Brothers, Mitnitski and Rockwood2013). Similarly, depression in later life increases the risk of mortality, especially if persistent (Li et al., Reference Li2018) and quite possibly in a dose-related manner (White et al., Reference White2016). Previous longitudinal evidence has demonstrated frailty and depression to be closely related (Collard et al., Reference Collard2015; Soysal et al., Reference Soysal2017). At the very least, frailty assessed as deficit accumulation and depressive symptoms appear to increase the risk of mortality in an additive manner (Chang et al., Reference Chang2019). However, it is not yet clear whether the frailty phenotype interacts with depressive symptoms in a similar manner. It is in this context that Ruiz-Grao and colleagues (Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020) set out to examine how frailty in older adults may be related to depression, and to investigate whether these quality-of-life reducing conditions interact to affect a person’s risk of death.
Although the authors did not state a specific hypothesis, it is reasonable to consider that frailty may act as a vulnerability factor to increase the additional risk of mortality conferred by depression, or vice versa. The authors conducted their statistical analyses using data from the Frailty and Dependence in Albacete Study (FRADEA), which is an impressive population-based cohort study set in the city of Albacete, Spain (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). As per the study protocol, information on a range of sociodemographic, health, and functional domains was collected at baseline, with follow-up assessments conducted 2 years later by telephone and 4 and 10 years later in person by trained nurses. The population from which the sample was drawn were 18,137 individuals who lived in Albacete aged 70 years or older and who held a health card between 2007 and 2009. Full details of the FRADEA cohort, design, and methodology are published elsewhere (Abizanda Soler et al., Reference Abizanda Soler2011). Briefly, frailty was assessed using the frailty phenotype proposed by Fried and colleagues (Fried et al., Reference Fried2001), depression was assessed using the 15-item Geriatric Depression Scale (GDS; Yesavage et al., Reference Yesavage1982), and data on mortality over the subsequent 10-year period were available. Risk of depression was considered present at baseline if participants scored 5 or more on the GDS, a tool which has been shown to be accurate in classifying the severity of depressive symptoms in community-dwelling older adults (Shin et al., Reference Shin2019).
At baseline, the 800 analyzed participants were mostly classified as prefrail (56.3%), with fewer classified as non-frail (24.5%) and fewer still classified as frail (19.3%). Importantly, and supporting previous work (e.g. Soysal et al., Reference Soysal2017), depression was highly comorbid with frailty: a 57.1% prevalence of depression was observed in frail participants, with a 33.3% and 15.3% prevalence in prefrail and non-frail participants, respectively. Almost half of the analytical samples died during the observation period (45%). For their primary analysis, the authors presented covariate-adjusted survival curves for combined frailty/depression groups yielded from a Cox proportional hazards model. Broadly, higher frailty was observed to substantially increase the risk of mortality, and the presence of depressive symptoms appeared to further add to this risk, at least in those classified as non-frail or prefrail.
The authors concluded that the presence of depressive symptoms increased mortality risk in a manner that depended upon the person’s level of frailty, whereby depressive symptoms increased the risk of death only in those classified as prefrail, but not in those classified as non-frail or frail (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). This trend is indeed observable at the level of the effect visualizations (i.e. survival curves), but the statistical evidence that was provided to support this conclusion was borderline. For instance, when the GDS was recoded as depressive risk being present (GDS score greater than or equal to 5) or absent (GDS score less than 5), and subsequently combined with frailty groups (non-frail, prefrail, frail), mortality risk did appear to increase in conjunction with depressive risk in those who were prefrail, but confidence intervals overlapped after models were adjusted for all covariates. Similarly, in a logistic regression model that examined the odds of death as a function of depressive symptoms (continuous score) in each frailty group individually, an increase in these symptoms was not associated with the odds of mortality at a statistically significant level in any frailty group, although a possible increase in the odds associated with depressive symptoms was observed in the prefrail group (i.e. OR = 1.05 [95% CI = 0.98 – 1.13]). We note that solid evidence for the authors’ conclusions may well emerge in a study with a larger sample size. What this investigation does contribute clearly, however, is further evidence of the dramatic increase in the risk of death due to advancing frailty, with this risk probably increasing further, in many individuals, if depressive symptoms are also present.
It is long known that depression is associated with a higher risk of death in older individuals (Penninx et al., Reference Penninx, Geerlings, Deeg, van Eijk, van Tilburg and Beekman1999). It is then worth pointing out that depressive symptoms and mortality, although clearly related in descriptive statistics and in Kaplan–Meier curves, were not associated at a statistically significant level after adjusting for possible confounders, even in those classified as non-frail (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). For context, the covariates included in the statistical models included age, sex, institutionalization, polypharmacy, and high comorbidity, which included information on diagnoses such as dementia and stroke. It is reasonable that depression may not be associated with a higher mortality risk in those individuals who are classified as frail, given that mortality risk may already be “saturated” in these people. Moreover, baseline depression may not be a good predictor of mortality (Li et al., Reference Li2018). Therefore, the authors are right to pay close attention to outcomes associated with depressive symptoms in non-frail and prefrail individuals, in which intervention may be most likely to alter an adverse trajectory.
This cohort was compiled using broad, population-based sampling (Abizanda Soler et al., Reference Abizanda Soler2011). Individuals with many comorbidities, and even those related to neurodegeneration, were included in the analytical sample. For example, in this study (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020), 71 participants had dementia (8.9%) and 67 had ischemic cerebrovascular disease (8.4%), with 118 participants living in institutions (14.8%; Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). On one hand, this is a strength, as such individuals are often excluded from research studies. On the other hand, the inclusion of individuals who have conditions that alone can cause both sufficient physical impairment to meet frailty criteria and depression (such as Parkinson’s disease or ischemic stroke) could potentially confound the results. Indeed, such participants were excluded for this reason from the investigation by Fried and colleagues (Reference Fried2001) that first described the frailty phenotype.
Instead of excluding ill participants, Ruiz-Grao and colleagues adjusted their analyses using the Charlson Comorbidity Index (Charlson et al., Reference Charlson, Pompei, Ales and MacKenzie1987), which is a composite score representing 17 categories of disease weighted by their relationship with mortality. The authors further categorized the resulting continuous score into a bivariate factor prior to its use as a covariate. Not only is it unlikely that this statistical adjustment would sufficiently correct for the presence of those confounding conditions described above, but many of the conditions in the index would be considered by others to indicate a higher degree of frailty if the deficit–accumulation approach were instead used (Rockwood and Howlett, Reference Rockwood and Howlett2019). Inclusion of this information could weaken the relationships between frailty and mortality, and so it is useful that the authors presented some of their key findings at different levels of covariate adjustment.
A minor consideration that is also worth touching upon is the authors’ method of handling missing data. For instance, this analysis used a complete cases dataset and opted to exclude participants with data missing who otherwise could possibly be retained if imputation techniques were used. The inclusion of these participants could have strengthened the findings of the study, as missing data is most often observed in individuals who are older and less healthy (Tsiampalis and Panagiotakos, Reference Tsiampalis and Panagiotakos2020). Consequently, the retention of participants with data missing may have increased the number of participants in the frail subgroup, which was the least populous of the frailty groups.
This paper identifies some appealing avenues for future research, namely, which factors modify the impact that frailty exerts on the risk of adverse health outcomes, such as mortality (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). Conversely, it is equally important to also understand the contexts in which frailty acts as a vulnerability factor, increasing the risk of adverse health outcomes from other stand-alone factors. This has, for example, been demonstrated with Alzheimer’s neuropathological burden and its expression as Alzheimer’s dementia (Wallace et al., Reference Wallace, Theou, Godin, Andrew, Bennett and Rockwood2019). It is perhaps too early to state with absolute confidence that frailty and depression interact to affect the risk of death, rather than both contributing to the risk in an additive manner. This may also be related to how frailty is conceptualized given that depressive symptoms are integral to the frailty phenotype, but can be studied independently with the deficit–accumulation model. This study was an admirable effort to address this important topic and we can expect significant work to build off this investigation in the coming time.
You have
Access
Humans and other animals are mortal, and the risk of mortality in all organisms increases exponentially with age (Ledberg, Reference Ledberg2020). At least in humans, impairments of both the body and mind – studied either individually or as a collective – are linked to increases in mortality with aging (Soysal et al., Reference Soysal2017). As a collection of impairments, frailty – thought of either as a clinical syndrome (Fried et al., Reference Fried2001) or as a state (Mitnitski et al., Reference Mitnitski, Mogilner and Rockwood2001) – is associated with complex age-related outcomes, such as dementia (Ward et al., Reference Ward, Wallace and Rockwood2020) and death (Theou et al., Reference Theou, Brothers, Mitnitski and Rockwood2013). Similarly, depression in later life increases the risk of mortality, especially if persistent (Li et al., Reference Li2018) and quite possibly in a dose-related manner (White et al., Reference White2016). Previous longitudinal evidence has demonstrated frailty and depression to be closely related (Collard et al., Reference Collard2015; Soysal et al., Reference Soysal2017). At the very least, frailty assessed as deficit accumulation and depressive symptoms appear to increase the risk of mortality in an additive manner (Chang et al., Reference Chang2019). However, it is not yet clear whether the frailty phenotype interacts with depressive symptoms in a similar manner. It is in this context that Ruiz-Grao and colleagues (Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020) set out to examine how frailty in older adults may be related to depression, and to investigate whether these quality-of-life reducing conditions interact to affect a person’s risk of death.
Although the authors did not state a specific hypothesis, it is reasonable to consider that frailty may act as a vulnerability factor to increase the additional risk of mortality conferred by depression, or vice versa. The authors conducted their statistical analyses using data from the Frailty and Dependence in Albacete Study (FRADEA), which is an impressive population-based cohort study set in the city of Albacete, Spain (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). As per the study protocol, information on a range of sociodemographic, health, and functional domains was collected at baseline, with follow-up assessments conducted 2 years later by telephone and 4 and 10 years later in person by trained nurses. The population from which the sample was drawn were 18,137 individuals who lived in Albacete aged 70 years or older and who held a health card between 2007 and 2009. Full details of the FRADEA cohort, design, and methodology are published elsewhere (Abizanda Soler et al., Reference Abizanda Soler2011). Briefly, frailty was assessed using the frailty phenotype proposed by Fried and colleagues (Fried et al., Reference Fried2001), depression was assessed using the 15-item Geriatric Depression Scale (GDS; Yesavage et al., Reference Yesavage1982), and data on mortality over the subsequent 10-year period were available. Risk of depression was considered present at baseline if participants scored 5 or more on the GDS, a tool which has been shown to be accurate in classifying the severity of depressive symptoms in community-dwelling older adults (Shin et al., Reference Shin2019).
At baseline, the 800 analyzed participants were mostly classified as prefrail (56.3%), with fewer classified as non-frail (24.5%) and fewer still classified as frail (19.3%). Importantly, and supporting previous work (e.g. Soysal et al., Reference Soysal2017), depression was highly comorbid with frailty: a 57.1% prevalence of depression was observed in frail participants, with a 33.3% and 15.3% prevalence in prefrail and non-frail participants, respectively. Almost half of the analytical samples died during the observation period (45%). For their primary analysis, the authors presented covariate-adjusted survival curves for combined frailty/depression groups yielded from a Cox proportional hazards model. Broadly, higher frailty was observed to substantially increase the risk of mortality, and the presence of depressive symptoms appeared to further add to this risk, at least in those classified as non-frail or prefrail.
The authors concluded that the presence of depressive symptoms increased mortality risk in a manner that depended upon the person’s level of frailty, whereby depressive symptoms increased the risk of death only in those classified as prefrail, but not in those classified as non-frail or frail (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). This trend is indeed observable at the level of the effect visualizations (i.e. survival curves), but the statistical evidence that was provided to support this conclusion was borderline. For instance, when the GDS was recoded as depressive risk being present (GDS score greater than or equal to 5) or absent (GDS score less than 5), and subsequently combined with frailty groups (non-frail, prefrail, frail), mortality risk did appear to increase in conjunction with depressive risk in those who were prefrail, but confidence intervals overlapped after models were adjusted for all covariates. Similarly, in a logistic regression model that examined the odds of death as a function of depressive symptoms (continuous score) in each frailty group individually, an increase in these symptoms was not associated with the odds of mortality at a statistically significant level in any frailty group, although a possible increase in the odds associated with depressive symptoms was observed in the prefrail group (i.e. OR = 1.05 [95% CI = 0.98 – 1.13]). We note that solid evidence for the authors’ conclusions may well emerge in a study with a larger sample size. What this investigation does contribute clearly, however, is further evidence of the dramatic increase in the risk of death due to advancing frailty, with this risk probably increasing further, in many individuals, if depressive symptoms are also present.
It is long known that depression is associated with a higher risk of death in older individuals (Penninx et al., Reference Penninx, Geerlings, Deeg, van Eijk, van Tilburg and Beekman1999). It is then worth pointing out that depressive symptoms and mortality, although clearly related in descriptive statistics and in Kaplan–Meier curves, were not associated at a statistically significant level after adjusting for possible confounders, even in those classified as non-frail (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). For context, the covariates included in the statistical models included age, sex, institutionalization, polypharmacy, and high comorbidity, which included information on diagnoses such as dementia and stroke. It is reasonable that depression may not be associated with a higher mortality risk in those individuals who are classified as frail, given that mortality risk may already be “saturated” in these people. Moreover, baseline depression may not be a good predictor of mortality (Li et al., Reference Li2018). Therefore, the authors are right to pay close attention to outcomes associated with depressive symptoms in non-frail and prefrail individuals, in which intervention may be most likely to alter an adverse trajectory.
This cohort was compiled using broad, population-based sampling (Abizanda Soler et al., Reference Abizanda Soler2011). Individuals with many comorbidities, and even those related to neurodegeneration, were included in the analytical sample. For example, in this study (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020), 71 participants had dementia (8.9%) and 67 had ischemic cerebrovascular disease (8.4%), with 118 participants living in institutions (14.8%; Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). On one hand, this is a strength, as such individuals are often excluded from research studies. On the other hand, the inclusion of individuals who have conditions that alone can cause both sufficient physical impairment to meet frailty criteria and depression (such as Parkinson’s disease or ischemic stroke) could potentially confound the results. Indeed, such participants were excluded for this reason from the investigation by Fried and colleagues (Reference Fried2001) that first described the frailty phenotype.
Instead of excluding ill participants, Ruiz-Grao and colleagues adjusted their analyses using the Charlson Comorbidity Index (Charlson et al., Reference Charlson, Pompei, Ales and MacKenzie1987), which is a composite score representing 17 categories of disease weighted by their relationship with mortality. The authors further categorized the resulting continuous score into a bivariate factor prior to its use as a covariate. Not only is it unlikely that this statistical adjustment would sufficiently correct for the presence of those confounding conditions described above, but many of the conditions in the index would be considered by others to indicate a higher degree of frailty if the deficit–accumulation approach were instead used (Rockwood and Howlett, Reference Rockwood and Howlett2019). Inclusion of this information could weaken the relationships between frailty and mortality, and so it is useful that the authors presented some of their key findings at different levels of covariate adjustment.
A minor consideration that is also worth touching upon is the authors’ method of handling missing data. For instance, this analysis used a complete cases dataset and opted to exclude participants with data missing who otherwise could possibly be retained if imputation techniques were used. The inclusion of these participants could have strengthened the findings of the study, as missing data is most often observed in individuals who are older and less healthy (Tsiampalis and Panagiotakos, Reference Tsiampalis and Panagiotakos2020). Consequently, the retention of participants with data missing may have increased the number of participants in the frail subgroup, which was the least populous of the frailty groups.
This paper identifies some appealing avenues for future research, namely, which factors modify the impact that frailty exerts on the risk of adverse health outcomes, such as mortality (Ruiz-Grao et al., Reference Ruiz-Grao, Sanchez-Jurado, Molina-Alarcon, Hernandez-Martinez, Avendano-Cespedes and Abizanda Soler2020). Conversely, it is equally important to also understand the contexts in which frailty acts as a vulnerability factor, increasing the risk of adverse health outcomes from other stand-alone factors. This has, for example, been demonstrated with Alzheimer’s neuropathological burden and its expression as Alzheimer’s dementia (Wallace et al., Reference Wallace, Theou, Godin, Andrew, Bennett and Rockwood2019). It is perhaps too early to state with absolute confidence that frailty and depression interact to affect the risk of death, rather than both contributing to the risk in an additive manner. This may also be related to how frailty is conceptualized given that depressive symptoms are integral to the frailty phenotype, but can be studied independently with the deficit–accumulation model. This study was an admirable effort to address this important topic and we can expect significant work to build off this investigation in the coming time.