Joint Associations of Frailty and Cardiometabolic Diseases With Risk of All-Cause and Cardiac Mortality

Abstract

Background

Cardiometabolic diseases (CMDs) are common for middle-aged and older adults. Whether frailty exacerbates CMD-related risk of mortality outcomes is unclear.

Objectives

The authors sought to investigate the joint associations of frailty and CMDs with mortality risk.

Methods

This prospective cohort study included 467,406 participants from UK biobank. Frailty was assessed using frailty phenotype and frailty index (FI). CMD status was defined as no CMD, single CMD, and cardiometabolic multimorbidity (CMM, coexistence of ≥2 CMDs). Multiplicative and additive interactions between frailty and CMDs two exposures on all-cause and cardiac mortality were examined, and then the joint association of coexisting frailty and CMDs with outcomes were estimated.

Results

During median follow-up of 13.08 years, 33,435 participants (7.2%) died, with 6,709 (1.4%) experiencing cardiac mortality. For frailty phenotype measurement, significant multiplicative and additive interactions existed with CMD status. Coexisting frailty phenotype and CMM were associated with a 4.91 (95% CI: 4.49-5.38) times and 8.33 (95% CI: 7.13-9.72) times higher risk of all-cause and cardiac mortality, with 23% and 36% attributable to the additive interaction, respectively. For FI measurements, a significant multiplicative interaction was observed with CMDs, and similar frailty–CMD joint associations with mortality outcomes were also observed.

Conclusions

Coexisting frailty and CMDs were associated with an accumulatively increased risk of mortality. Comprehensive screening and management of frailty and CMDs is advisable in aged populations.

Central Illustration

With population ageing and extended life expectancy, long-lasting chronic diseases and ageing-related physical function impairment have contributed to an increasing disease burden to both individuals and health care systems.¹ In decades, a cluster of chronic conditions, cardiometabolic diseases (CMDs) (including coronary heart disease [CHD], stroke, and diabetes here), have been a leading cause of mortality for middle-aged and older adults worldwide,² and the coexistence of ≥2 CMDs in an individual, that is, cardiometabolic multimorbidity (CMM), is also highly prevalent.³ Compared with people without CMDs, patients with CMM had an accumulatively increased risk of mortality, dementia, depression, and impaired quality of life.4, 5, 6, 7 And in the context of population ageing, concerns about the coexisting CMDs and frailty has risen.^8,9 Frailty is a geriatric syndrome characterized by an aging-related decline in the physiological functions of comprehensive organs and systems. Featured by decreased physical reserve and increased vulnerability, frail individuals were prone to suffer a substantial decline in physical function after stress, contributing to a series of adverse outcomes.⁸ For community-dwelling older adults, the pooled frailty prevalence was ∼10.7% in high-income countries.¹⁰ In previous literature, frailty was an important predictor for mortality among middle-aged and older adults in a dose-dependent manner. It was also associated with higher risk of hospitalization, falls, loss of independence, and medical expenditure.11, 12, 13

Rising as 2 momentous issues in the field of geriatric medicine, CMDs and frailty have been suggested to be closely related and commonly coexist. Our previous research highlighted the role of frailty in the disease progression trajectory from no CMD to single CMD, CMM, and ultimately to mortality,¹⁴ whereas CMDs were also associated with higher odds of frailty.¹⁵ In a systematic review, 19% of patients with CHD had frailty, whereas 17% of frail patients had CHD.¹⁶ Among older adults with diabetes, the prevalence of frailty and prefrail were also substantially higher (20.1% and 49.1%).¹⁷ Considering their substantial role for predicting poor prognosis and increasing prevalence of coexistence, a comprehensive understanding in the joint associations of coexisting CMDs and frailty may be necessary to improve the long-term prognosis of middle-aged and older adults. However, only a few studies have investigated their joint association with mortality risk. Based on the Taiwan Longitudinal Study on Aging, Chu et al¹⁸ recently found that CMM combined with frailty phenotype was associated with an accumulatively higher risk of mortality (OR: 3.58; 95% CI: 1.96-6.54) relative to CMM alone (OR: 1.24; 95% CI: 1.04-1.49), whereas the research by Nguyen et al¹⁹ did not support the joint associations of frailty and cardiovascular diseases (CVDs) based on the National Health and Aging Trends Study. In addition to the conflict conclusions, a major limitation of the existing literature was that they did not investigate the underlying interaction between frailty and CMD on mortality risk, preventing a comprehensive understanding in the joint role of these 2 intractable issues.

To address this problem and therefore provide solid evidence for enhancing the comprehensive assessment to identify high-risk population, we conducted this study based on 467,406 participants in UK Biobank, described the prevalence of frailty among patients with different CMD status, systematically assessed the multiplicative and additive interactions between frailty and CMD status on all-cause and cardiac mortality, and further estimated the joint associations of coexisting frailty and CMDs with mortality risk.

Methods

Study population and design

The study was based on UK Biobank, a largescale, prospective cohort study with long-term follow-up. From 2006 to 2010, UK Biobank enrolled 0.5 million participants aged 40 to 69 years from 22 centers across the United Kingdom. With written informed consent, participants finished touchscreen questionnaires and verbal interviews about demography, lifestyles, and medical history, underwent physical examinations, and provided biological samples. Health-related information in mortality, inpatient diagnosis, and operations was accessed through linkages to electronic records. UK Biobank has been approved by West Multi-centre Research Ethics Committee (ref: 21/NW/0157).

A total of 502,411 participants in UK Biobank were available for formal analyses. We included 496,885 participants with complete information on baseline frailty status and excluded 29,479 participants with missing data on the date of outcomes or covariates (Supplemental Figure 1). Ultimately, 467,406 participants were included, which fulfilled the necessary sample size required to detect exposure-related difference in mortality events with 80% power (Supplemental Table 1).

Assessment of frailty

We used frailty phenotype and frailty index (FI), 2 widely used and complementary tools to evaluate baseline frailty status. Frailty phenotype was proposed by Fried et al²⁰ based on 5 physical measurements (unintentional weight loss, exhaustion, physical activity, gait speed, and grip strength); we used the adapted version to fit the data availability in UK Biobank (Supplemental Table 2).²¹ Participants with complete information in ≥4 criteria were included and defined with robust (0 fulfilled), prefrail (1-2 fulfilled), and frail (≥3 fulfilled) phenotype. FI was estimated using 49 self-reported items including health conditions, medical history, disabilities, and mental health (Supplemental Table 3).²² Participants with information in ≥40 items were included, and FI was calculated as summed scores divided by the number of items evaluated. Participants were then classified as robust (FI <0.12), mildly frail (FI ≥0.12-≤0.24), or moderately/severely frail (FI >0.24).

Assessment of CMD status

The status and diagnosis date of CMDs were ascertained based on inpatient diagnosis coded by International Classification of Diseases, version 9 and 10, and operation records coded by Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4 (Supplemental Table 4). For participants with certain diseases, the earliest date of diagnosis and that of recruitment were compared. If the date of diagnosis was earlier, the individuals were considered to have the disease at baseline, otherwise have the disease during follow-up. Particularly, according to the number of baseline CMDs, participants were classified as no CMD, single CMD, or CMM (with ≥2 CMDs).

Ascertainment of outcomes

The outcomes were all-cause and cardiac mortality. In UK Biobank, mortality events and dates were acquired via linkages to National Health Service (NHS) Information Centre in England, Wales, and NHS Central Register in Scotland.²³ Primary cause of mortality was encoded by International Classification of Diseases-version 10, and codes I00-I99 were denoted as cardiac mortality. The follow-up started from the date of recruitment to mortality or the latest follow-up (April 2022), whichever came first.

Assessment of covariates

Information in covariates was obtained through touchscreen questionnaires, interview/anthropometric data, and diagnosis records, including demography (age, sex [dichotomous, male or female], race [dichotomous, white or non-white], Townsend Deprivation Index, and body mass index), lifestyle (healthy diet, smoking status [multicategorical, current smoking, previous smoking, or never smoking], alcohol consumption frequency [dichotomous, <3 or ≥3 times per week], physical activity, and sleep duration), clinical factors (hypertension history [dichotomous, yes or no], family history of CMDs [dichotomous, yes or no], and use of antihypertensive agents, lipid-lowering medications, aspirin, and insulin [dichotomous, yes or no]) (Supplemental Table 5). Body mass index was calculated as body weight (kg) divided by the square of height (m²). Healthy diet was measured by healthy diet score according to American Heart Association Guidelines. If ≥2 following items were fulfilled, healthy diet score was considered as “1” (more advisable), or “0” (less advisable) otherwise: 1) total fruit and vegetable intake: >4.5 pieces or servings per day; 2) total fish intake: >2 times per week; 3) processed and red meat intake: ≤2 times of processed meat per week and ≤5 times of red meat per week. Levels of physical activity were assessed using 2017 UK Physical Activity Guidelines (dichotomous, meeting the guidelines of 150 minutes of walking or moderate activity per week/75 minutes of vigorous activity or not).²⁴ In addition, hypertension history was confirmed according to the electronic health records of inpatient diagnoses.

Statistical analyses

First, we described baseline characteristics of participants. Categorical variables were presented as n (%). Continuous variables were measured as mean ± SD, for normal distribution or median (IQR) for abnormal distribution. Histogram and density plots were used to describe the distribution of frailty levels. We also used Kruskal-Wallis and Nemenyi tests to estimate the overall and pairwise difference in frailty levels across CMD status.

In the following survival analyses, follow-up year was used as the timescale. We used Kaplan-Meier survival curves to compare the survival probability related to CMD or frailty status. According to the survival curves, the proportional hazard assumption was fulfilled. Cox proportional hazard models were used to estimate the individual associations of CMD or frailty with outcomes. In the basic model, covariates including sex, race, Townsend Deprivation Index, body mass index, healthy diet, smoking status, alcohol consumption, physical activity, sleep duration, and baseline hypertension were adjusted, and then CMD status and frailty were mutually adjusted to estimate their independent associations with mortality outcomes.

In the analysis for the joint associations between frailty and CMD status with mortality outcomes, we firstly tested their multiplicative and additive interactions on outcomes. Frailty and CMD status were included in Cox proportional hazard models as categorical variables simultaneously, and cross-product terms between the exposures were introduced. Likelihood ratio tests were used to estimate the multiplicative interaction (P_interaction). Additive interaction between frailty and CMD status was tested with 3 indicators including relative excess risk due to interaction, attributable proportion due to interaction, and synergy index.²⁵ Then participants were categorized into 9 groups according to baseline frailty and CMD status, with those without CMD and frailty as reference. Since the proportional hazard assumption was not fulfilled according to Kaplan-Meier curves, we used Royston-Parmar flexible parametric survival models with hazard scale to estimate the joint associations of coexisting CMDs and frailty with mortality outcomes. To estimate the overall HRs, flexible parametric survival models without time-varying effects were used with 2 settings. In model 1, age, sex, and race were adjusted; in model 2, Townsend Deprivation Index, body mass index, healthy diet, smoking status, alcohol consumption, physical activity, sleep duration, and hypertension history were further adjusted. To describe the time-varying effects of the joint association of coexisting frailty and CMDs, flexible parametric survival models with an interaction term between coexisting frailty–CMDs status and follow-up year were further fitted. The details in the settings of flexible parametric survival models are provided in the Supplemental Methods.^26,27 Variance inflation factors for variables included were <5, indicating lack of multicollinearity.

To examine the sex differences in the distribution of frailty levels across CMD status and the joint association of frailty and CMD, stratified analyses were further conducted by sex. To test the reliability and robustness of our research, sensitivity analyses were also conducted (Supplemental Methods).

We used R 4.1.0 (R Foundation for Statistical Computing) and Stata software 16.0 (StataCorp) in the study. Two-sided P values <0.05 indicated a significant threshold.

Results

Descriptive analyses

The baseline characteristics of participants are presented in Table 1. Among 467,406 participants included (mean age 57.01 years), 212,926 (45.6%) were male, and 445,145 (95.2%) were white. A total of 89.5% participants had no CMD, 9.1% had single CMD, and 1.4% had CMM at baseline. Compared with the population free of CMD, participants with CMDs were more likely to be older, male, non-white, have lower socioeconomic status, higher body mass index, history of hypertension, lower alcohol consumption frequency, and were currently smoking. According to frailty phenotype, 15,009 (3.2%) participants were frail. When using FI, 164,752 participants (35.2%) were mildly frail and 27,856 (6.0%) moderately/severely frail. Supplemental Figure 2 shows the overlapping of frailty status with different tools. In addition, baseline characteristics of participants included or excluded in the main analyses are presented in Supplemental Table 6.

Table 1.

Baseline Characteristics Overall and According to CMD Status

	Overall (N = 467,406, 100%)	No CMD (n = 418,292, 89.5%)	Single CMD (n = 42,710, 9.1%)	CMM (n = 6,404, 1.4%)	P Value
Age, y	57.01 ± 8.08	56.51 ± 8.08	61.02 ± 6.80	62.60 ± 5.84	<0.001
Male	212,926 (45.6)	181,553 (43.4)	26,740 (62.6)	4,633 (72.3)	<0.001
Race, White	445,145 (95.2)	399,555 (95.5)	39,760 (93.1)	5,830 (91.0)	<0.001
Townsend Deprivation Index	−2.22 (−3.68 to 0.36)	−2.27 (−3.71 to 0.23)	−1.74 (−3.41 to 1.30)	−0.92 (−3.11 to 2.44)	<0.001
Body mass index, kg/m2	27.34 ± 4.74	27.05 ± 4.56	29.65 ± 5.38	31.12 ± 5.53	<0.001
Healthy diet	263,264 (56.3)	235,483 (56.3)	24,220 (56.7)	3,561 (55.6)	0.133
Current smoking	47,518 (10.2)	41,858 (10.0)	4,832 (11.3)	828 (12.9)	<0.001
Alcohol consumption ≥3 times/week	207,486 (44.4)	189,140 (45.2)	16,346 (38.3)	2,000 (31.2)	<0.001
Physical activity according to guideline	360,674 (77.2)	324,755 (77.6)	31,568 (73.9)	4,351 (67.9)	<0.001
Sleep duration, h/d	7.16 ± 1.09	7.15 ± 1.06	7.20 ± 1.28	7.26 ± 1.56	<0.001
Hypertension history	122,899 (26.3)	92,543 (22.1)	25,277 (59.2)	5,079 (79.3)	<0.001
Frailty phenotype
Fulfilled criteria	0.00 (0.00-1.00)	0.00 (0.00-1.00)	1.00 (0.00-1.00)	1.00 (0.00-2.00)	<0.001
Robust	278,136 (59.5)	258,406 (61.8)	18,034 (42.2)	1,696 (26.5)	<0.001
Prefrail	174,261 (37.3)	149,839 (35.8)	20,988 (49.1)	3,434 (53.6)
Frail	15,009 (3.2)	10,047 (2.4)	3,688 (8.6)	1,274 (19.9)
Frailty by FI
FI	0.10 (0.06-0.16)	0.10 (0.06-0.15)	0.16 (0.12-0.22)	0.22 (0.17-0.29)	<0.001
Robust	274798 (58.8)	262,801 (62.8)	11,420 (26.7)	577 (9.0)	<0.001
Mildly frail	164,752 (35.2)	138,235 (33.0)	23,367 (54.7)	3,150 (49.2)
Moderately/severely frail	27,856 (6.0)	17,256 (4.1)	7,923 (18.6)	2,677 (41.8)

Values are n (%), mean ± SD, or median (Q1-Q3).

CMD = cardiometabolic disease; CMM = cardiometabolic multimorbidity; FI = frailty index.

Figure 1 demonstrates the distribution of frailty levels across CMD status. Irrespective of assessment tools, CMD patients had higher frailty levels (Table 1). Further Kruskal-Wallis tests and Nemenyi tests also supported the overall and pairwise differences in the levels of frailty across CMD status (all P < 0.001). Similar distribution of frailty levels across CMD status was observed in male and female participants, and the prevalence of frailty in female participants was higher (Supplemental Table 7). 26.8% or 52.5% of female participants with CMM were identified as frail phenotype or moderately/severely frail by FI.

Figure 1.

Distribution of Frailty Levels Stratified by CMD Status

(A) Distribution of fulfilled phenotype criteria numbers according to cardiometabolic disease (CMD) status. (B) Density map of FI according to CMD status. CMM = cardiometabolic multimorbidity; FI = frailty index.

Independent association of CMD status and frailty with mortality

During a median follow-up of 13.08 years, 33,435 participants (7.2%) died, with 6,709 (1.4%) experiencing cardiac mortality. Events during follow-up according to baseline CMD and frailty status are presented in Supplemental Table 8.

Kaplan-Meier curves by CMD status and frailty are shown in Supplemental Figure 3, and participants who had CMDs or who were frailer had higher all-cause and cardiac mortality. Supplemental Table 9 shows the independent associations of CMD status and frailty with mortality risk. When CMD and frailty were not mutually adjusted, dose-dependent positive associations existed for CMD and frailty status (measured by either frailty phenotype or FI) with all-cause and cardiac mortality outcomes. Compared with participants without CMD, patients with single CMD or CMM had a 56% (95% CI: 51%-60%) or 152% (95% CI: 140%-165%) higher risk of all-cause mortality. Compared with the robust phenotype, prefrail and frailty phenotype-related HRs for all-cause mortality were 1.40 (95% CI: 1.37-1.43) and 2.63 (95% CI: 2.51-2.75); corresponding HRs for mildly and moderately/severely frail patients by FI were 1.37 (95% CI: 1.341.41) and 2.13 (95% CI: 2.052.21). After mutual adjustment, CMD- and frailty-related associations were partly attenuated but remained significant.

Joint associations of CMD status and frailty with mortality

The coexistence of single CMD/CMM with frailty status increased the risk of mortality outcomes in a dose-dependent manner (Supplemental Figure 4). With frailty phenotype measurements, significant multiplicative interaction existed between CMD status and frailty phenotype for all-cause and cardiac mortality risk (P_interaction < 0.05), and pronounced additive interactions were observed for single CMD with prefrail, CMM with the prefrail/frailty phenotype in relation to all-cause mortality and CMM with the prefrail/frailty phenotype in relation to cardiac mortality (Table 2). Compared with people free of CMD and frailty at baseline, patients with both CMM and frailty phenotype had the highest risk of all-cause and cardiac mortality (HR: 4.91; 95% CI: 4.49-5.38 for all-cause mortality; HR: 8.33l; 95% CI: 7.13-9.72 for cardiac mortality) along the overall follow-up duration, with 23% and 36% of the effect sizes attributed to the additive interaction (Figure 2, Supplemental Table 10). In analysis using FI, significant multiplicative interaction also existed between CMD status and FI-estimated frailty (P_interaction < 0.05), whereas no significant additive interaction was observed (Table 2). And the coexistence of CMM and moderately/severely frail defined by FI also contributed to the highest risk of all-cause (HR: 3.65; 95% CI: 3.4-3.91) and cardiac mortality (HR: 5.93; 95% CI: 5.24-6.71) (Figure 2, Supplemental Table 10).

Table 2.

Additive and Multiplicative Interactions Between Frailty and CMD Status

Exposure Combination	All-Cause Mortality				Cardiac Mortality
	RERI (95% CI)	AP (95% CI)	S (95% CI)	Pinteraction	RERI (95% CI)	AP (95% CI)	S (95% CI)	Pinteraction
Frailty phenotype
Single CMD, prefrail	0.12 (0.02-0.21)	0.06 (0.01-0.10)	1.13 (1.02-1.25)	<0.001	0.08 (−0.19 to 0.34)	0.03 (−0.06 to 0.12)	1.04 (0.90-1.21)	0.024
Single CMD, frail	−0.09 (−0.33 to 0.15)	−0.04 (−0.13 to 0.06)	0.94 (0.81-1.10)		0.23 (−0.77 to 1.23)	0.05 (−0.15 to 0.24)	1.06 (0.82-1.37)
CMM, prefrail	0.76 (0.29-1.24)	0.20 (0.10-0.30)	1.37 (1.16-1.62)		1.01 (0.05-1.96)	0.22 (0.06-0.38)	1.40 (1.08-1.82)
CMM, frail	1.14 (0.64-1.64)	0.23 (0.15-0.32)	1.41 (1.22-1.62)		3.00 (1.59-4.41)	0.36 (0.23-0.49)	1.69 (1.34-2.13)
FI
Single CMD, mild	0.00 (−0.11 to 0.10)	0.00 (−0.06 to 0.06)	1.00 (0.88-1.13)	<0.001	−0.03 (−0.31 to 0.25)	−0.01 (−0.12 to 0.10)	0.98 (0.81-1.18)	0.001
Single CMD, moderate/severe	0.00 (−0.27 to 0.27)	0.00 (−0.09 to 0.09)	1.00 (0.88-1.14)		−0.80 (−1.59 to −0.02)	−0.18 (−0.39 to 0.03)	0.81 (0.64-1.02)
CMM, mild	−0.29 (−0.67 to 0.10)	−0.13 (−0.33 to 0.07)	0.80 (0.57-1.12)		0.18 (−0.80 to 1.15)	0.05 (−0.23 to 0.33)	1.08 (0.71-1.65)
CMM, moderate/severe	−0.10 (−0.62 to 0.42)	−0.03 (−0.17 to 0.12)	0.96 (0.79-1.17)		0.02 (−1.44 to 1.48)	0.00 (−0.25 to 0.25)	1.00 (0.74-1.36)

In the models, follow-up year was included as time scale, and covariates including age, sex, race, Townsend Deprivation Index, body mass index, healthy diet, smoking status, alcohol consumption, physical activity, sleep duration, and hypertension at baseline were adjusted. Additive interaction between CMD and frailty status was tested by RERI, AP, and S. Multiplicative interaction was tested by P_interaction.

AP = attributable proportion due to interaction; RERI = relative excess risk due to interaction; S = synergy index; other abbreviations as in Table 1.

Figure 2.

Joint Association of Frailty and CMD Status on Mortality Outcomes

(A and B) Joint association of frailty phenotype and CMD status on all-cause/cardiac mortality; (C and D) joint association of frailty assessed by FI and CMD status on all-cause/cardiac mortality. In the model, follow-up year is included as the timescale, participants without CMD and frailty status at baseline are considered as the reference group. Overall HR and 95% CI estimates from flexible parametric survival models with hazard scale are presented. Baseline restricted cubic splines with 4 internal knots (df: 5, positions: 20%, 40%, 60%, and 80%) and 3 internal knots (df: 4, default positions: 25%, 50%, 75%) were used for all-cause and cardiac morality as outcomes, respectively. Covariates including age, sex, race, Townsend Deprivation Index, body mass index, healthy diet, smoking status, alcohol consumption, physical activity, sleep duration, and hypertension at baseline were adjusted. Abbreviations as in Figure 1.

By introducing an interaction between coexisting CMD–frailty status and follow-up year, we further described the time-varying effects of the joint role of CMDs and frailty in mortality risk (Supplemental Figures 5 and 6). Results showed that the joint associations of CMDs and frailty status attenuated over time. After 15-year follow-up periods, the HR estimates of CMM combined with frailty phenotype or moderately/severely frail (by FI) remained ∼4 to 5 for mortality risk.

In stratified analyses according to sex (Supplemental Figure 7), consistent results were observed in both sexes, and female participants with CMM and frailty status had a higher risk of all-cause and cardiac mortality.

Sensitivity analysis

We observed results consistent with the main analyses after performing time-lag analysis, additionally adjusting for family history and medication use, modifying FI by excluding 8 items of cardiometabolic conditions, or using age as the time scale (Supplemental Table 11). After excluding participants who developed CMD during follow-up, stronger joint associations of CMD and frailty were observed, especially for cardiac mortality. When using grip strength as a single measurement of frailty, the combination of CMDs and low grip strength was also significantly associated with higher risk of all-cause and cardiac mortality (Supplemental Table 12).

Discussion

According to our results, patients with CMDs had higher frailty levels, especially patients with CMM. Both frailty and CMD status were independently associated with higher risk of all-cause and cardiac mortality. Significant multiplicative interactions were observed between frailty (measured by either frailty phenotype or FI) and CMD status, whereas an additive interaction existed between frailty phenotype and CMD. The coexistence of frailty phenotype and CMM was related to a 391% and 733% higher risk of all-cause and cardiac mortality, respectively, among which 23% and 36% could be attributed to the additive interaction respectively. When evaluated by FI, participants with CMM and moderately/severely frail had a 3.65- and 5.93-times higher risk of all-cause and cardiac mortality (Central Illustration).

Central Illustration.

The Joint Association of Frailty and Cardiometabolic Disease on All-Cause and Cardiac Mortality

In this study, 467,406 participants from UK Biobank were included. Cardiometabolic disease (CMD) status at baseline was categorized into no CMD, single CMD, or cardiometabolic multimorbidity (CMM) (the coexistence of ≥2 CMDs in an individual). Frailty status was evaluated by frailty phenotype and frailty index. The joint association between CMM and frailty on all-cause or cardiac mortality was observed, for their coexistence generated higher risk of mortality compared with individually exposure of CMM or frailty. CHD = coronary heart disease.

Cross-sectionally, we found that frailty levels dose-dependently increased with the progression of CMD, in line with findings in previous studies. According to a systematic review, the prevalence of frailty among older adults with diabetes was 16% (frailty phenotype) and 35% (FI),¹⁷ and among patients with stroke or CHD was 22%²⁸ and 19%, respectively.¹⁶ Here, we extended the scope to different CMD status and found that 19.9% of CMM patients had a frailty phenotype and that the proportion of frailty (mild/moderate/severe) by FI reached up to 91%, far beyond the frailty prevalence in the population with multimorbidity (16%) observed previously.²⁹ This disparity indicated that CMM, an increasingly common comorbidity, was closely related to frailty. Meanwhile, the frailty prevalence among patients with single CMD observed by us was lower than that in previous research, which might be partly explained by the fact that previous studies might not have considered the coexistence of several CMDs. In addition, our data were derived from UK Biobank, the participants of which tended to have better health conditions and lower frailty levels than the general population. In analyses stratified by sex, we found that the prevalence of frailty was higher among females than males, consistent with previous cross-sectional studies.¹⁰ This sex-related difference may be explained by behavioral and social factors.³⁰

Previous studies indicated that frailty and CMDs were risk factors for mortality outcomes. In the general population, frailty phenotype was associated with a 2 times higher risk of mortality across all age groups,³¹ and each 0.1 increment in FI was associated with ∼28% and 31% higher risk of all-cause and CVD-related mortality, respectively.³² For patients with diabetes, CVD, or end-stage renal disease, frailty remained a significant indicator of poor prognosis.33, 34, 35 On the other hand, a cumulative association also existed for coexisting CMDs with all-cause and cardiac mortality.³⁶ By comparing models with or without mutual adjustment for the 2 exposures, we confirmed that both frailty and CMD status were independent risk factors for mortality outcomes, although the estimated effect sizes for both exposures were partly attenuated when included in the statistical models simultaneously.

In light of the common coexistence of frailty and CMDs, and their independent role in predicting poor prognosis, we then explored whether patients with both frailty and CMDs had an accumulatively higher risk of mortality. When analyzed according to the combination of frailty and CMD status, the risk of all-cause and cardiac mortality increased by higher frailty levels and progressed CMD status, and patients with combined frailty and CMM had the highest risk for poor outcomes. Our results were partly consistent with findings in the recent literature. The combination of frailty and CMM generated higher risk of mortality than CMM alone in Taiwan Longitudinal Study on Ageing,¹⁸ and Shi et al³⁷ also showed the combined effect of diabetes and frailty on mortality risk among Chinese older adults. However, Nguyen et al¹⁹ did not observe a joint effect of frailty and CVDs in the National Health and Aging Trends Study, and CVD multimorbidity increased mortality risk only in robust people, rather than frail or prefrail participants. Possible explanations for this controversy included discrepancies in study design, populations, and definitions in CMD patterns and frailty, and the interaction effects between the 2 issues remained unclear. We further strengthened the joint associations by systematically evaluating the interactions between frailty and CMDs in relation to mortality. In addition to the multiplicative interactions between frailty (measured by either frailty phenotype or FI) and CMD status, significant additive interactions also existed for frailty phenotype with CMD status, with 23% and 36% of the risk of all-cause and cardiac mortality among patients with frailty phenotype and CMM attributed to their additive interaction. These findings indicated that the excessive risk of mortality depend majorly on the coexistence of frailty and CMM status.

When frailty and CMDs coexisted, the interplay between alterations in multiple pathological mechanisms including cardiovascular metabolism disorder, chronic inflammation, and dysregulation in maintaining homeostasis could be bidirectional, aggravating the adverse effects of frailty and CMDs, leading to the accumulatively increased mortality risk. Furthermore, frailty status might also aggravate CMD-related excessive risk of mortality by modifying the treatment responses and inducing adverse events. Patients receiving percutaneous coronary intervention after acute coronary syndrome are usually recommended to receive standard dual antiplatelet therapy for 12 months, but those who were frail were more likely to experience major bleeding events within 30 days after the intervention.³⁸ Among patients with acute ischemic stroke, pre-stroke frailty attenuated the symptom improvement after thrombolysis, and also impeded the social and psychological intervention-related improvements in activities of daily living and physical function.^39,40 In this way, it is advisable for health care professionals to take frailty screening and assessment into account when encountering middle-aged and older patients with CMDs, especially those who had CMM. In the management of high-risk population with potentially more severe prognosis, the involvement of frailty conception could also assist therapeutic decision-making to keep the balance between treatment benefits and the risk of adverse events, thereby developing cost-effective strategies for the reduction of disease burden. Despite the performance of frailty phenotype and FI confirmed in the study, their applications in routine frailty assessment were somehow restricted for its complexity. Single frailty metrics such as grip strength and 4/5 meter walk were associated with adverse prognosis,^41,42 and our additional analyses also observed positive associations between CMDs–low grip strength and mortality risk with relatively smaller effect sizes compared with the results generated with frailty phenotype. In this way, it might be advisable to incorporate frailty assessment in CMD management by combining single frailty metrics as routine screening and multidimensional frailty scales (such as frailty phenotype or FI) as further systematic evaluation.

Since there has been no consensus on the standard diagnostic criteria of frailty, we applied 2 commonly used and theoretically complementary tools, frailty phenotype predominantly based on physical function decline and FI based on aging-related deficits accumulation, to assess frailty in this study. In our analyses, frailty status measured by both tools was associated with a higher risk of mortality, with stronger effect sizes observed for frailty phenotype. This difference could be partly attributed to the different conceptions between the 2 tools. Compared with frailty phenotype, FI was more sensitive in distinguishing prefrailty and mild frailty during the progression of frailty,⁴³ hence the prevalence of frailty assessed by FI was higher, with a relatively lower risk of mortality. In analyses of the joint associations, multiplicative interactions were observed between CMD status and frailty measured by both tools, whereas additive interactions existed for frailty phenotype with CMM but not FI-derived frailty status. This phenomenon was probably because we had already included indicators related to cardiac health in the calculation of FI, and the effect related to CMD was somehow masked. However, irrespective of frailty assessment tools used, the dose-dependent associations of frailty and CMD status on mortality existed, highlighting the significance of frailty screening in CMD progression trajectory to reduce disease burden and relevant mortality risk.

This study has several strengths. First, we used data from UK Biobank. With sufficient sample, the “3 × 3” classification of CMD and frailty and further stratification by sex could be realized with adequate statistical power. A follow-up of 13 years also ensured adequate cases of outcome events. Second, we used 2 complementary assessment tools of frailty and observed consistent patterns of the joint associations.

Study limitations

First, participants with missing data in frailty assessment, mortality, or covariates were excluded from the main analyses, which might lead to selection bias since the distribution of baseline characteristics of participants with missing information was different from that of the included people (Supplemental Table 6). Second, 96% of participants included here were white, and “healthy volunteer bias” existed according to previous literature.⁴⁴ Thus, we need to be cautious when extrapolating our findings to other countries and races. Third, limited by the group numbers, we did not include certain CMDs and CMM patterns in analyses, which limited further comparison of effect among specific CMDs. Fourth, CMD and frailty status were evaluated at baseline, and during a long period of follow-up, participants could experience progression of CMD and changes in frailty status. After excluding participants who developed new-onset CMD during follow-up, strengthened effects were observed in sensitivity analysis 3, indicating that potential dilution bias existed here, which might lead to underestimation of the effect sizes. Fifth, self-reported information was used when evaluating frailty and covariates, which might introduce measurement bias. Finally, as an observational study, this research could not eliminate potential bias from reverse causation and residual confounding factors.

Conclusions

Our results indicated that significant interaction existed between frailty and CMD status on the risk of mortality. Coexistence of frailty and single or multiple CMDs was associated with an increased risk of all-cause and cardiac mortality in a dose-dependent manner. Our findings provide epidemiological evidence supporting the comprehensive management of both frailty and CMDs to alleviate disease burden and extend life expectancy in the aged population.

Availability of data and materials

The data that support the findings of this study are available from UK Biobank (approved project 84443), but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are, however, available from the website https://www.ukbiobank.ac.uk/ upon reasonable request and with permission of UK Biobank.

Funding Support and Author Disclosures

This work was supported by Noncommunicable Chronic Diseases-National Science and Technology Major Project grant number 2023ZD0503202, National Natural Science Foundation of China grant numbers 82325006, 82270446, and 82401853, Major Project of Natural Science Foundation of Hunan Province grant number 2021JC0002, National Natural Science Foundation of Hunan Province grant number 2023JJ40930, and China Postdoctoral Science Foundation grant number GZC20242034 and 2024M753687. The sponsors did not participate in the design, methods, subject recruitment, data collections, analysis and preparation of paper. The authors have reported that they have no relationships relevant to the contents of this paper to disclose. The UK Biobank has been approved by the North West Multi-centre Research Ethics Committee as a Research Tissue Bank, and separate ethical clearance is not required for researchers under this approval (updated ref 21/NW/0157, 18 June 2021). All participants of UK Biobank have provided written informed consent.