Impact of Frailty on Mortality and Hospitalization in Chronic Heart Failure: A Systematic Review and Meta‐Analysis

Introduction

Frailty is a complex systemic syndrome that has been associated with poor outcomes, including increased rates of mortality and hospitalization in frail patients with heart failure (HF) compared with those without frailty.^{1, 2, 3, 4, 5, 6, 7} More commonly observed in association with advanced age,¹ frailty can also affect young patients with HF and can be reversible after heart transplantation⁸ or heart function replacement.⁹ Characterizing frailty status in patients with HF may provide clinicians an indicator for gauging disease severity, prognosis, and disease progression or reversal.

Although most studies have shown increased risks for hospitalization and mortality in chronic HF,^{2, 3, 4, 5, 6, 7} offset by 2 relatively small studies with borderline neutral results,^{10, 11} the precise average effect of frailty has not previously been summarized with certainty. Recently, a systematic review included a total of 8 articles from 2004 to 2014 and suggested that patients with HF and frailty had increased risks for adverse outcomes.¹² However, 2 of 8 articles^{5, 13, 14, 15} in that study contained data from overlapping studies (ie, double counting), and no meta‐analysis was performed. Since 2014, 6 additional independent studies^{2, 3, 4, 11, 16, 17} with potential to be included in a meta‐analysis have been published. Recently, a systematic review and meta‐analysis has summarized data on the prevalence of HF‐associated frailty from 26 studies at 44.5% after removal of 14 overlapping studies.¹⁸ However, that study did not assess the impact or effect size of frailty on adverse outcomes, including death and hospitalization, and included patients with acute HF and those receiving cardiac resynchronization therapy, left ventricular assist device (LVAD), or heart transplant.¹⁸ In this study, we have carefully summarized the existing evidence, explored for heterogeneity, and completed the first meta‐analysis of frailty on the hazards of mortality and incident hospitalization in patients with chronic HF.

Methods

Data Sources and Search Strategy

The data that support the findings of this study are available from the corresponding author on reasonable request. From October 2016 to April 2018, we searched MEDLINE, EMBASE, and the Cochrane databases for articles published between January 1966 and March 2018. Study design (Figure 1 and Figure S1) and data reporting were compliant with MOOSE, as recommended in the EQUATOR Network guidelines.^{19, 20} Institutional Review Board approval was not required for this systematic review and meta‐analysis.

• Download figure
• Download PowerPoint

Eligibility Criteria

The following Medical Subject Headings or MeSH search terms were used: “frailty AND heart failure,” “frail AND heart failure,” “fragility AND heart failure,” “gait speed AND heart failure,” “grip strength AND heart failure,” “weight loss AND heart failure,” and “cognitive frailty AND heart failure.” Fragility is a term used by some European investigators with whom we have individually confirmed about its use in their publications as being synonymous with frailty for the purpose of this study. Literature search was independently performed by 2 reviewers (E.F. and X.Y.), according to a prespecified workflow. Conflicting findings were resolved by a third researcher (J.W.). The inclusion criteria were as follows: (1) frailty assessed using validated assessment instruments; (2) confirmed diagnosis of HF on the basis of international criteria and guideline definitions^{21, 22}; (3) human individuals; and (4) articles in English. The exclusion criteria were as follows: (1) unpublished data, abstracts, conference proceedings, comments, letters, correspondences, editorials, or duplicates; (2) studies without any relevance to the clinical outcomes of hospitalization and mortality; (3) studies that investigated the effects of medical or surgical intervention, including cardiac resynchronization therapy, LVAD, and cardiac transplantation; (4) data on acute HF; and (5) non‐English articles.

Data Extraction

The main clinical outcomes in this meta‐analysis were all‐cause mortality and hospitalization. Where data on all‐cause hospitalization were not available,³ cardiac hospitalization was used. A structured data form was used to organize information^{2, 3, 4, 5, 6, 7, 10, 11, 13, 14, 15, 16, 17, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37} (Table1). Two reviewers (E.F. and X.Y.) extracted the raw data and independently evaluated study quality (described later).

Data Synthesis and Statistical Analysis

The effect sizes of risks were reanalyzed and recalculated as hazards, pooled, and represented by adjusted or unadjusted hazard ratios (HRs) for the clinical outcomes of all‐cause mortality or hospitalization in frail and nonfrail patients with chronic HF. When available, all‐cause rather than cardiac hospitalization was used. This is justified by evidence in the literature that hospitalizations secondary to HF with preserved left ventricular ejection fraction (EF) are often attributable to complex comorbidities.^{21, 22} Of 6 studies included in the meta‐analysis of frailty on incident hospitalization, 5 used all‐cause hospitalization and 1 used cardiac hospitalization³ as end points. Where possible, adjusted, rather than unadjusted, HRs were used (Table S1).

To standardize classification of frailty/nonfrailty status, we reclassified prefrailty, a nonexistent category in some frailty assessment scales,^{38, 39} as nonfrailty. To minimize data redundancy and double counting,⁴⁰ we contacted authors of the original studies for clarification and data reanalysis, where necessary. Among a total of 10 groups authoring 20 studies, 8 groups responded to our query with clarification and/or assisted in reanalysis (response rate, 80%). Where data redundancy from overlapping studies was suspected or in cases in which the original authors could not be reached, we proceeded to remove those articles from further analysis. The research group of Bayés‐Genís, Lupón, and Gastelurrutia consolidated their cohort data from 3 publications and reanalyzed risk estimates.^{3, 15, 34} Originally presented as odds ratios in the study of Vidán and colleagues,⁴ the risk for rehospitalization was recalculated as HRs using data from telephone interviews at 1, 3, 6, and 12 months after hospital discharge. For 66 patients in that study who did not have the exact dates of rehospitalization, the time to hospitalization was imputed using the calculated mean time to hospitalization from the remaining 340 patients. The data on mortality and hospitalization from the original article by Ferguson and colleagues were newly calculated and presented as HRs.¹⁶

Results from eligible studies were pooled and meta‐analyzed using a random‐effects model with inverse‐variance weighting. Heterogeneity of studies was assessed using Cochrane's Q statistic and I². Prespecified I² threshold values of 25%, 50%, and 75% were used to indicate low, moderate, and high levels of heterogeneity, respectively. A 2‐tailed P<0.05 was considered statistically significant. Analyses were performed using Review Manager 5.3 (The Cochrane Collaboration) and R 3.3.3 (R Foundation).

Assessment of Publication Bias and Study Quality

Publication bias was assessed using funnel plots and Duval‐Tweedie's trim‐and‐fill test (Figure S2).

Results

Search and Study Selection

Our query to EMBASE, MEDLINE, and Cochrane databases returned 6886 records (Figure 1 and Figure S1). After further screening and evaluation, 6641 irrelevant records and 217 ineligible full‐text articles were excluded. Among 28 eligible articles,^{2, 3, 4, 5, 6, 7, 10, 11, 13, 14, 15, 16, 17, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37} 8 and 6 unique studies were finally included in the meta‐analysis for all‐cause mortality and incident hospitalization, respectively (Figure 1 and Table1).

Characteristics of Patients With Chronic HF

Among studies that reported on the prevalence rates of atrial fibrillation, ischemic heart disease (coronary artery disease), hypertension, and/or diabetes mellitus, the respective median prevalence rates were 53% (quartile 1–quartile3, 46.5%–62.2%), 44.5% (quartile 1–quartile 3, 26.9%–54.5%), 80% (quartile 1–quartile 3, 64.0%–87.9%), and 38.4% (quartile 1–quartile 3, 34.8%–47.0%). The median prevalence rate of frailty from 17 of 19 nonoverlapping studies was 49.0% (quartile 1–quartile 3, 21.7–64.9); data were unavailable from 2 studies.

Frailty Is Significantly Associated With Increased Mortality in Chronic HF

On the basis of data from 8 unique studies, the presence of frailty is significantly associated with an increased hazard for mortality in chronic HF (HR, 1.54; 95% confidence interval [CI], 1.34–1.75; P<0.001) (Figure 2A and Table S2). The median duration of follow‐up for 2645 patients in the 8 studies^{2, 3, 4, 5, 6, 10, 11, 16} was 1.82 years (quartile 1–quartile 3, 1.0–3.9 years). In studies that used only the Fried phenotype for frailty assessment (n=5 studies; Figure 2B), the effect size estimate was increased by 16.9% (HR, 1.80; 95% CI, 1.41–2.28; P<0.001; n=1127 patients).^{2, 4, 5, 10, 11} Irrespective of the frailty assessment instrument used, the level of study heterogeneity was low (I²=0%) (Figure 2A and 2B).

• Download figure
• Download PowerPoint

Frailty Is Significantly Associated With an Increased Rate of Hospitalization in Chronic HF

During a median follow‐up of 1.12 years (quartile 1–quartile 3, 1–2 years), the presence of frailty was significantly associated with an increased hazard for hospitalization by 56% (adjusted HR, 1.56; 95% CI, 1.36–1.78; P<0.001; n=2541 patients in 6 studies), even after adjusting for factors in the respective studies^{2, 3, 4, 7, 11, 16} (Figure 2C and Table S1). In 4 studies that used Fried assessment,^{2, 4, 7, 11} the estimate was, however, similar (adjusted HR, 1.57; 95% CI, 1.30–1.89; P<0.001; n=1142 patients) to the overall pooled estimate (Figure 2D). The level of heterogeneity between studies was also low (I²=0%) (Figure 2C and 2D).

Publication Bias and Study Quality

Although the effect size estimates for mortality fell within the pseudo 95% confidence limits of the funnel plot (Figure S2A), we proceeded to using the trim‐and‐fill test for ascertainment and found no significant difference with or without adjustment (adjusted HR, 1.48 [95% CI, 1.25–1.75] [P<0.001] 1.54 [95% CI, 1.34–1.75] [P<0.001]) (Figure S2C and Figure 2A), thereby ruling out large publication bias effects. The high quality of the included studies^{2, 3, 4, 5, 6, 7, 10, 11, 16} was indicated by a composite score of ≥7 on the Newcastle‐Ottawa scale⁴¹ (Table S3).

Discussion

Frailty is increasingly recognized as an important target for monitoring and intervention in contemporary cardiovascular care and management.^{42, 43} Clinical pathways using frailty assessment in management decision making and for determination of procedural eligibility have previously been shown to affect patient outcomes and provide prognostic indication of well‐being and survival associated with procedures, ranging from transcatheter and surgical aortic valve replacement to heart transplantation.^{43, 44, 45} However, the precise negative effects of frailty on chronic HF^{2, 3, 4, 5, 6, 7, 10, 11} have not been previously established with certainty.

This systematic review and meta‐analysis is the first to summarize the adverse impact of hospitalization and mortality associated with frailty in chronic HF. Mortality and incident hospitalization were both significantly increased by ≈1.5‐fold in >1300 patients with HF with frailty, compared with >1200 patients with HF without frailty (Figure 2). This study has reanalyzed and recalculated risk estimates as HRs, consolidated data from 3 independent sources (study cohorts from Barcelona, Spain; Madrid, Spain; and Sydney, Australia) to minimize or eliminate double counting, and focused on patients with chronic HF. Current available studies in the literature have not included any meta‐analyzed data on mortality and hospitalization associated with frailty in HF,^{12, 46} and frailty prevalence was the sole focus of another recently published meta‐analysis that included patients with acute decompensated HF, patients with chronic HF, and patients receiving cardiac resynchronization therapy, LVAD, and heart transplant.¹⁸

One of the major findings in this study was the higher estimate of hazard for mortality, but not incident hospitalization, by the Fried phenotype compared with the overall pooled (combined Fried and non‐Fried) estimates (Figure 2). The reason for this observation is unclear, but it can be possibly explained by the overlapping characteristics (and pathophysiological features) between functional components of the Fried phenotype (eg, progressive unintentional weight loss and weakness) and cardiac cachexia of advanced chronic HF, which carries a poor prognosis. However, the concept of cardiac cachexia cannot be simply explained by reduced body mass index alone because this relationship is complex,⁴³ depending on the population or patient subset with HF, the cause, the pathogenesis, and the chronicity of the pathophysiological features. For instance, patients waiting for heart transplant tend to be younger, may have developed HF with reduced EF over a relatively shorter period (eg, post–viral dilated cardiomyopathy), and may have different physiological reserve levels and body composition compared with elderly patients with long‐standing, chronic HF. Older adults may have more comorbidities,⁴⁷ latent chronic HF⁴⁸ (particularly, HF with preserved EF), and sarcopenic obesity,⁴⁹ a disorder characterized by low lean skeletal muscle mass relative to abundant intermuscular adiposity. This meta‐analysis was focused on patients with chronic HF and frailty outside the setting of heart function replacement or acute HF, acknowledging the variability in phenotypic expression of (or individuals’ variable resistance to) frailty, and its manifestation and reversibility across the HF spectrum.

The clinical phenotype proposed by Fried and colleagues³⁸ in 2001 is based on the concept of aging‐related failure of homeostasis in physiologic systems (or domains) represented by 5 specific items: low physical activity, fatigue, shrinkage (or unintentional weight loss over a defined period), weakness, and slowness. The 5 items of the Fried phenotype are scored out a total of 5 points, with 1 point awarded for each positive item; scores of ≥3, 1 to 2, and 0 indicate subjects in frail, prefrail, and robust states, respectively. The prefrail state may be a clinically relevant indicator of underlying cardiometabolic disorder, reduced physiological reserve, and a window of opportunity for workup and intervention before development of systemic decompensation. However, the absence of prefrailty in some frailty assessment scales precluded comparison between studies for this condition. Another major frailty concept is centered on the frailty index developed by Rockwood et al.⁵⁰ The frailty index is based on a cumulative multiple deficit approach, using clinical and laboratory variables with an emphasis on the number rather than type of derangements. As noted, these and other commonly used frailty assessment tools, including the modified Fried‐based FRAIL scale that obviates physical testing (eg, hand grip strength and walking tests⁵¹) with improved clinical operationality, have been previously compared in specific settings and found to perform similarly.^{51, 52} Other surrogate markers of frailty include gait speed,^{53, 54} which has been reported to offer prognostic information in patients with HF.⁵⁵ Our search strategy in this study has included individual components of frailty assessment scales (see Methods) to maximize identification of articles in the literature.

A major challenge lies in the harmonization of a wide range of frailty assessment scales used in studies. Woo and coworkers have previously reported that the major frailty scales (frailty index by Rockwood and Mitnitski, Fried phenotype from the Cardiovascular Health Study, FRAIL scale, and the Hubbard modified frailty score) perform similarly in predicting death and physical limitation in a geriatric study population.⁵¹ Chong and colleagues found that several frailty assessment instruments (frailty index, FRAIL scale, Tilburg Frailty Index, and Clinical Frailty Scale) tested did not significantly differ in their ability to diagnose frailty.⁵² Those findings suggest that pooling of studies for meta‐analysis can be feasible and informative for the specific purpose of this study, as supported by available evidence in the literature.

HF and aging‐related frailty share common signs, symptoms, and manifestations, including sarcopenia or skeletal muscle weakness, impaired cardiorespiratory and physical fitness, increased inflammatory burden,⁵⁶ and central nervous system dysfunction.^{57, 58} Although definitions and assessment scales do vary, there has been impetus toward a universal definition of frailty in the geriatric professional community.^{1, 59, 60} Currently, frailty is not routinely assessed for or systematically categorized in patients with HF.^{1, 21, 22, 42, 61} However, findings from this and other studies encourage the use of frailty assessment for risk stratification of patients with HF to inform prognosis and management decisions. Recent data from the LVAD and heart transplant literature suggest that frailty in patients with advanced HF can be reversed by intervention or organ replacement^{8, 43} and that frailty is not necessarily age or functional class related,¹⁸ suggesting that inclusion of frailty assessment in patients with HF can inform outcomes. Indeed, Jha and colleagues have shown that pre‐LVAD or pretransplant frailty status has a significant impact on survival after LVAD implantation or heart transplantation.⁴³

There are several limitations in this study. First, the competing risks between hospitalization and mortality across time could not be assessed given the limitations inherent in the original studies.⁶² Second, hospitalization as an outcome measure is complex and may be HF associated, cardiac related, or unrelated to cardiovascular events.^{63, 64} Details on hospitalization and contemporary metrics (eg, 30‐day readmission and length of stay) for characterizing hospitalization were unavailable from the included studies. Future studies using standardized metrics may improve accuracy of risk estimates. Third, there were insufficient data in the available articles to perform a meta‐analysis on HF subtypes (eg, HF with reduced EF and HF with preserved EF).^{21, 22} Although some studies have shown that the adjusted survival and hospitalization rates are similar between HF subtypes,^{47, 65} there is a need for future studies to clarify this.

In conclusion, this meta‐analysis provides the first summary on the effects of frailty on mortality and hospitalization and confirms the significant negative impact of frailty on chronic HF. Stratification of patients with HF by frailty status provides prognostic information and may inform priorities for HF interventions and management.

Sources of Funding

Bayés‐Genís and Gastelurrutia are supported by grants from the Ministerio de Educación y Ciencia (SAF2014‐59892) and Generalitat de Catalunya (Pla estratègic de recerca i innovació en salut Programme, Departament de Salut, SLT002_16_00209), Fundació La MARATÓ de TV3 (201502‐30, 201516‐10), Red de Terapia Celular–TerCel (RD16/0011/0006), and Centro de Investigación en Red en Enfermedades Cardiovasculares (CB16/11/00403) projects as part of the Plan Nacional de I+D+I, and jointly funded by Instituto de Salud Carlos III–Sudirección General de Evaluación y el Fondo Europeo de Desarrollo Regional. Centro Nacional de Investigaciones Cardiovasculares (CNIC) is supported by the Ministerio de Economía, Industria y Competitividad (MINECO), and the Pro‐CNIC Foundation, and is a Severo Ochoa Center of Excellence (MINECO award SEV‐2015‐0505). Ferguson was supported by a University of Technology Sydney, Chancellor's Postdoctoral Research Fellowship (2015001232). Fung is recipient of the Direct Grant award from the Faculty of Medicine, The Chinese University of Hong Kong, and of an investigator‐initiated research grant from the Food and Health Bureau of the Hong Kong Special Administrative Region. Woo is supported by research grants from the Hong Kong Jockey Club and the Research Grants Council.

Disclosures

None.

Footnotes