The Pandora’s Box of Frailty Assessments: Which Is the Best for Clinical Purposes in TAVI Patients? A Critical Review

Frailty assessment is recommended before elective transcatheter aortic valve implantation (TAVI) to determine post-interventional prognosis. Several studies have investigated frailty in TAVI-patients using numerous assessments; however, it remains unclear which is the most appropriate tool for clinical practice. Therefore, we evaluate which frailty assessment is mainly used and meaningful for ≤30-day and ≥1-year prognosis in TAVI patients. Randomized controlled or observational studies (prospective/retrospective) investigating all-cause mortality in older (≥70 years) TAVI patients were identified (PubMed; May 2020). In total, 79 studies investigating frailty with 49 different assessments were included. As single markers of frailty, mostly gait speed (23 studies) and serum albumin (16 studies) were used. Higher risk of 1-year mortality was predicted by slower gait speed (highest Hazard Ratios (HR): 14.71; 95% confidence interval (CI) 6.50–33.30) and lower serum albumin level (highest HR: 3.12; 95% CI 1.80–5.42). Composite indices (five items; seven studies) were associated with 30-day (highest Odds Ratio (OR): 15.30; 95% CI 2.71–86.10) and 1-year mortality (highest OR: 2.75; 95% CI 1.55–4.87). In conclusion, single markers of frailty, in particular gait speed, were widely used to predict 1-year mortality. Composite indices were appropriate, as well as a comprehensive assessment of frailty.


Introduction
The proportion of older people (>65 years) is predicted to constantly increase in Europe and reach around 30% of the total population by 2060 [1]. These significant changes to society's structure must be taken into account by the healthcare system. In fact, it is known that aging is associated with several geriatric syndromes, such as cognition impairments, malnutrition, sarcopenia and frailty [2]. The clinical condition of frailty especially has gained international attention in the last decade [3].
Frailty is a complex clinical condition [4], where physiological deteriorations related to the aging process are accentuated and vulnerability to stressors increases [5,6]. According to the definition of physical frailty, malnutrition, unintentional weight loss, decreased strength and endurance as well as impaired physiological functions characterize the frail population [4,7]. Frail patients have a higher risk of death and compromised independence in activities of daily living [8]. Therefore, in older patients with cardiovascular disease (CVD) undergoing elective transcatheter aortic valve implantation (TAVI), frailty should be evaluated [9].
Older TAVI patients are characterized by decreased physiological functionality (e.g., reduced gait speed), multiple comorbidities and malnutrition, with the prevalence of frailty found to be up to 63% in this population [5,6,10]. Pre-interventional assessment of frailty should be considered alongside common risk scores (e.g., STS score, EuroSCORE) for a better appreciation of the post-interventional prognosis [11]. In the last decade, several investigations have assessed the association of pre-interventional frailty status as an independent predictor of post-interventional mortality. However, due to the lack of consensus on the definition of frailty (e.g., physical frailty, multidimensional) and its measurement (e.g., single markers, composite indices, categorical scales, claims data), the evidence is contradictory [10,[12][13][14] and identification of the most appropriate assessment for clinical practice remains unclear.
Therefore, the aim of our review was to evaluate which frailty assessment is the most commonly used and meaningful for the prediction of short-term, intermediate-term and long-term all-cause mortality in older TAVI patients.

Materials and Methods
The electronic database PubMed (May 2020) was systematically searched. Study inclusion criteria were the following: randomized controlled trials (RCTs) or observational studies (prospective and retrospective, respectively) without time restrictions that quantitatively assessed frailty pre-interventionally in older (≥70 years) patients who underwent elective TAVI. The outcome of interest was the association of frailty with all-cause mortality. Identification, screening, eligibility and inclusion processes were performed according to PRISMA guidelines. The included articles were reviewed by two independent researchers. The overall risk of bias in each study was evaluated using the Quality in Prognosis Studies (QUIPS) [15] and the Newcastle-Ottawa Scale (NOS) [16] tools. According to the QUIPS, we classified studies with four or five low-risk domains as having a low risk of bias overall, studies with two or more high-risk domains as having a high risk for bias overall and the remaining studies as showing a moderate risk of bias overall. For the NOS, the sum of positive adjudications was calculated. Studies were then divided into 3 categories: single markers of frailty, multidimensional frailty scales and composite indices. The single markers category comprised studies that assessed only one component of frailty, like physical function, activities of daily living (ADL), blood parameters, body composition, nutritional status and cognition. The multidimensional frailty scales category comprised studies that rated frailty using categorical scales based on clinician judgment. The third category of composite indices encompassed studies that assessed frailty based on an index resulting from the measurement of several frailty components. Furthermore, studies were stratified into short-term (≤30-day), intermediate (>30-day to <1-year) and long-term (≥1-year) mortality.

Results
In total, n = 79 studies were included for full-text review ( Figure 1) and n = 49 different frailty assessments were identified. All studies were of observational design and most of the findings referred to short-and long-term mortality. The majority of the studies investigated frailty using single markers of frailty, such as physical functionality (e.g., gait speed, handgrip strength), blood parameters (serum albumin), impairments in activities of daily living (ADL), nutritional status/risk (Mini-Nutritional Assessment), cognitive impairments (Mini-Mental State Examination) and body composition (Psoas muscle area index), as shown in Table 1. Several studies used composite indices or multidimensional scales based on clinical judgment (Table 2), and only a few studies defined frailty status based on claimed medical records and comorbidities (Supplementary Material Table S1).   Table S1 displays the characteristics of all included studies and the assessment of bias. This is followed by a summary of the results of the most commonly used assessments (Table 3).
The Mini-Mental State Examination questionnaire [78] was used in n = 3 prospective [9,33,51] studies with relatively small cohorts (89 to 213 participants) to predict shortterm mortality. Only Storteky et al. [9] showed a higher relative risk of mortality with a decrease in cognitive capacity. Long-term mortality was investigated in n = 5 prospective [9,12,22,51,55] (1498 patients) and n = 1 retrospective [40] (1542 patients) studies, with most of the studies reporting no association with MMSE scores. Only n = 2 studies [9,55] with small sample sizes found that pre-interventional lower MMSE results were associated with mortality. Intermediate mortality was investigated in two prospective [14,54] studies (119 and 150 patients) that showed divergent results.
Psoas muscle area index (PMAi) was investigated in only n = 2 prospective studies [56,57] with conflicting evidence for short-term mortality. Kofler et al. [56] found that higher PMAi values were predictive of a lower relative risk of mortality, whereas van Mourik et al. [57] stated no association between PMAi and death rates. Long-term mortality was investigated in n = 5 prospective [25,[56][57][58]60] (2172 patients) and n = 3 retrospective [44,45,59] (2515 patients) studies. The majority of these studies showed that lower PMAi seems to be associated with a negative post-interventional prognosis.
The Timed Up-and-Go test (TUG) was used in n = 7 prospective studies [9,12,14,33,51,54,55] with cohorts ranging from 89 [33] to 344 [12] patients. Storteky et al. [9] found linear (each 5 sec increase) and dichotomized (≥20 versus <20 s) results on the TUG, predictive of short-and long-term mortality in 100 patients. Similarly, Eichler et al. [12] found that a longer time needed to complete the TUG (≥10 to <20 versus <10 s) was predictive of up to a fivefold relative risk of mortality in 344 patients. However, Goudzwaard et al. [51] concluded that only linear values were predictive of 1-year mortality, but not over the short-term, and only when results were dichotomized (>20 versus ≤20 s).
A 5-item index was used in n = 5 prospective [13,[69][70][71][72] (2653 patients) and n = 1 retrospective [21] (498 patients) studies. Rogers et al. [69] found in a cohort of 544 patients that index values ≥ 3 corresponded to a fivefold increase in the relative risk of mortality at 30 days and up to a twofold increase at 1 year post-TAVI. Further, two authors [13,21] stated that higher scores on the index were predictive of long-term mortality.
Four investigations [13,37,73,74] found higher scores of the EFT to be predictive of 1-year mortality, showing up to a threefold higher risk of death in frail patients [13].
The Bern scale was used in n = 5 investigations [9,[12][13][14]55], all of a prospective design (1913 patients). Afilalo et al. [13] found scores ≥ 3 to be predictive of up to a threefold higher relative risk of mortality over the short-term. Similarly, Storteky et al. [9] found a threefold higher risk of death in a cohort of 100 patients per each point increase of the index and for dichotomized results.

Discussion
Besides a distinct heterogeneity of frailty definitions and assessments used across the included studies, the main finding of our systematic review is that single markers of frailty are the most frequently used tools to identify frailty in older TAVI patients, especially parameters of physical functionality (gait speed) and malnutrition (serum albumin level). Patients with a slower walk speed (≥6 s) measured by the 5-mWT showed a higher relative risk of 1-year mortality. Pre-interventional lower levels (≤3.5 g/dL) of serum albumin were associated with a worse long-term prognosis as well. Further, higher scores of composite indices (5-item, EFT, Bern scales) were associated with higher mortality risk over the short and long term after intervention.
A great number of studies (n = 23) investigated gait speed with the 5-mWT and found an association with higher 1-year mortality in older TAVI patients. Slower gait speed is largely used to evaluate impaired health conditions, especially among the elderly [80], and to assess the prevalence of frailty among the general population in a primary care setting [81]. Moreover, it is advocated as a reliable and valid tool to investigate frailty in patients with cardiovascular disease [6]. Gait speed is the result of neuromuscular control, cardiopulmonary condition, physical activity level, patient health status and sensorial pattern interactions, properly expressing the general physical functionality of older patients [82]. According to the physical frailty concept (physical phenotype), slower gait speed is considered a valid clinical indicator of frailty and sarcopenia [7]. Generally, TAVI patients who need ≥6 s (≥0.85 m/s) to perform the 5-mWT are considered frail. However, several investigations advocate that patients' stratification into "slower/slow/normal" might be more appropriate, with the slowest walkers showing the highest risk of long-term mortality [20,39].
Besides physical functionality, malnutrition was mainly assessed as an indicator of frailty considering pre-interventional levels of serum albumin. In the majority of studies, a level of ≤3.5 g/dL was associated with a worse long-term prognosis. Albumin is a protein synthetized in the liver, and a serum concentration under 3.5 g/dL is considered representative of malnutrition [83]. Alteration of serum concentration levels is influenced by vascular injury, renal injury or various cytokine levels, and has been associated with increased risk of mortality in patients with cardiovascular disease [46]. In TAVI patients, serum albumin levels are routinely assessed pre-interventionally [46] and are often used for retrospective analysis. Especially in older and very old patients, malnutrition is a common problem correlated to several factors, such as appetite reduction, physiological changes, altered hormonal responses, mental impairments and chewing or swallowing problems [7]. An impaired nutritional intake contributes to a worsened physical functionality as well as to the process of muscle wasting and sarcopenia, which are closely related to frailty syndrome [84].
Composite indices were developed as measurement tools with the intent to assess several patient components involved in the clinical condition of frailty. Depending on the index, patients with three or more impaired components are considered frail. In our review, several studies found that the assessment of frailty according to composite indices was predictive of 30-day and 1-year mortality in older TAVI patients, in particular the EFT and the 5-item index. Our findings are in line with the recommendation made by the International Conference on Frailty and Sarcopenia Research (ICFSR) [85]. Initially, Fried et al. [86] recommended the evaluation of frailty according to the frail phenotype, an index based on the assessment of slowness, weakness, low physical activity, exhaustion and shrinking. These components were measured according to gait speed, handgrip strength, calculation of kilocalories expended per week, self-reported exhaustion (questionnaire) and >5 kg unintentional weight loss. Several authors revised Fried's index and used alternative measurements tools. Thus, analysis of serum albumin levels as an expression of malnutrition was used to investigate the shrinking component [39]. Further, other indices (EFT, Bern scale) [9,13] were applied to the physical and nutritional component assessments of cognition (MMSE) and disability scales (BADL, IADL).
In this review, we provide a detailed overview of the frailty assessments most commonly utilized in clinical practice, divided by category and their ability to predict mortality risk. However, due to the remarkable heterogeneity of the included studies, meta-analyses of the study results were not indicated. Therefore, this review does not provide pooled effect sizes for the frailty assessments investigated, potentially limiting information for the clinician. Heterogeneity is notable especially in regard to the methodology of frailty measurements, study designs and statistical analyses. For example, seven different cut-off values were identified for gait speed. This variety of cut-off values might have influenced the estimation of patients considered as frail across the studies. For composite indices, different measurement approaches (different tools used to assess the same domain; diverse domain cut-off values) were evident across studies. Additionally, the lack of a general agreement on the definition of frailty [6] and the diversity of TAVI populations included in the studies contributed to the wide range and heterogeneity of the reviewed outcomes.
First of all, there are two fundamentally different approaches to conceptualizing frailty. On the one hand, Fried et al. (2001) describe frailty based on the phenotype, focusing on physical components (e.g., unintentional weight loss, muscle weakness, slow walking speed, low physical activity and exhaustion) [86]. One the other hand, the multidimensional concept developed by Rockwood and Mitnitski (2001) comprises psychological and social components, multi-morbidity, disability in addition to the physical impairments [79,87]. Against this background, several authors purposed a large number of frailty assessments operationalizing these frailty concepts. Recently (2019), the International Conference on Frailty and Sarcopenia Research (ICFSR) recommended the Fried method for the clinical assessment of frailty [85]. However, the majority of studies in this review are conducted and published before 2019.
Although frailty is a multidimensional syndrome among older adults characterized by a marked vulnerability and diminished capability to recover from stressors, a considerable number of single markers of frailty were used in the included studies. From the statistical point of view, in multivariable analyses of mortality, single parameters often show equivalent or stronger predictive effects than composite measures [12,88,89]. Therefore, the use of single frailty markers-also due to their good practicability-in clinical studies seems justified. Given the high complexity of frailty syndrome and frailty definition, for clinical decision-making, composite indices are more appropriate for the characterization of a patient. The use of single parameters, despite being acceptable for clinical studies, is not recommended in this context.
For patients scheduled for TAVI and classified as frail, the effect of prehabilitation programs on morbidity and mortality is currently being investigated [90]. Taking into account the fact that aortic stenosis as well as frailty usually develop over years, the monitoring of malnutrition and physical function seems to be just as important in preventive cardiology as the assessment of aortic stenosis.

Conclusions
Frailty was most commonly assessed using single markers of frailty, especially based on measurements of gait speed (5-mWT) and pre-interventional levels of serum albumin. Slow gait speed (≥6 s) and a level of serum albumin ≤ 3.5 g/dL were predictive of a higher risk of 1-year mortality in older TAVI patients. Composite indices (5 items), considered as a comprehensive assessment of frailty, were associated with short-and long-term mortality. However, considerable heterogeneity was observed among the studies, and the methodology of each frailty assessment should be carefully considered.