Preinterventional frailty assessment in patients scheduled for cardiac surgery or transcatheter aortic valve implantation: a consensus statement of the European Association for Cardio-Thoracic Surgery (EACTS) and the European Association of Preventive Cardiology (EAPC) of the European Society of Cardiology (ESC)

Table of Contents

1 INTRODUCTION 4

 1.1 General considerations and rationale 4

 1.2 Methodology and results 4

2 FRAILTY AND THE PREDICTION OF OUTCOME AFTER CARDIAC SURGICAL PROCEDURES 5

 2.1 Frailty as a predictor of short-term, intermediate-term and long-term mortality after cardiac surgery 5

  2.1.1 Gait speed and walking distance 6

  2.1.2 Katz Index of activities of daily living 6

  2.1.3 Psoas muscle measurements and sarcopenia 7

  2.1.4 Fried frailty phenotype: 5 items assessment 8

  2.1.5 Clinical Frailty Scale 9

  2.1.6 Short Physical Performance Battery 9

  2.1.7 Composite indexes 10

 2.2 Frailty as predictor of neurological complications/delirium and prolonged ventilation/hospitalization after cardiac surgery 10

  2.2.1 Fried frailty phenotype—5-item assessment, modified Fried criteria and single components 11

  2.2.2 Psoas muscle measures and sarcopenia 11

  2.2.3 Nutritional state measurements 12

  2.2.4 Short Physical Performance Battery 12

  2.2.5 Edmonton Frail Scale 13

  2.2.6 Cognitive measurements: Mini-Mental State Examination and Montreal Cognitive Assessment 13

 2.3 Frailty as predictor of quality of life, discharge location and readmission after cardiac surgery 13

  2.3.1 Clinical frailty scale and multiscale assessment of frailty to predict quality of life and time to discharge to home after cardiac surgery 13

  2.3.2 Fried frailty phenotype—5 items assessment and modified scales 14

  2.3.3 Psoas muscle measures and sarcopenia 14

 2.4 Summary of consensus statements for frailty assessment in cardiac surgery 15

  2.4.1 Consensus statements: prediction of mortality after cardiac surgery 15

  2.4.2 Consensus statements: prediction of neurological complications/delirium and prolonged hospitalization/ventilation after cardiac surgery 15

  2.4.3 Consensus statements: prediction of quality of life, discharge location and probability of readmission after cardiac surgery 15

 2.5 Expert task force condensed advice on frailty assessment to predict outcome of cardiac surgical procedures by frailty assessment 15

  2.5.1 The European Frailty Score for Cardiovascular Interventions (cardiac surgery) [EuroFORECAST(cs)] 16

3 FRAILTY AND THE PREDICTION OF OUTCOME OF TRANSCATHETER AORTIC VALVE IMPLANT PROCEDURES 16

 3.1 Frailty as a predictor of short-, intermediate- and long-term mortality 16

  3.1.1 Gait speed and walking distance 16

  3.1.2 Serum albumin 17

  3.1.3 Activities of daily living 18

  3.1.4 Handgrip strength 18

  3.1.5 Clinical Frailty scale 18

  3.1.6 Psoas muscle measurements 19

  3.1.7 Mini-Mental State Examination 19

  3.1.8 Composite indexes 20

 3.2 Frailty as predictor of neurological complications/delirium and prolonged ventilation/hospitalization after transcatheter aortic valve implantation 21

  3.2.1 Nutritional state measurements 21

  3.2.2 Gait speed 22

  3.2.3 Mini-Mental State Examination 22

  3.2.4 Psoas muscle measures and sarcopenia 22

  3.2.5 Composite frailty indexes 23

 3.3 Frailty as a predictor of quality of life, discharge location, readmission and functional impairment after transcatheter aortic valve implantation 23

  3.3.1 Assessment of frailty to predict quality of life or functional improvement 23

  3.3.2 Assessment of frailty to predict discharge location and probability of readmission 24

   3.3.2.1 Gait speed as single parameter or in a composite score predicting discharge location 24

   3.3.2.2 Fried criteria and albumin as an isolated marker to predict discharge location 24

  3.3.3 Prediction of probability of readmission by different frailty tools 25

 3.4 Summary of consensus statements on frailty assessment in transcatheter aortic valve implantation 25

  3.4.1 Consensus statements: prediction of mortality after transcatheter aortic valve implantation 25

  3.4.2 Consensus statements: prediction of neurological complications/delirium and prolonged hospitalization/ventilation after TAVI 26

  3.4.3 Consensus statements: prediction of quality of life, discharge location and probability of readmission after transcatheter aortic valve implantation 26

 3.5 Expert task force condensed consensus statements on frailty assessment to predict outcome of transcatheter aortic valve implantation procedures by frailty assessment 26

  3.5.1 The European Frailty Score for Cardiovascular Interventions (EuroFORECAST) 26

4 MANAGEMENT OF FRAIL PATIENTS AND INTEGRATION OF FRAILTY ASSESSMENT IN CLINICAL ROUTINE 26

 4.1 Use of serum markers for frailty assessment 26

 4.2 Use of administrative data for frailty assessment 26

 4.3 Prehabilitation 27

5 PRACTICAL SUMMARY AND KEY MESSAGES 28

 5.1 General considerations 28

 5.2 Frailty assessment 28

  5.2.1 Frailty assessment in patients scheduled for cardiac surgery 28

  5.2.2 Frailty assessment in patients scheduled for cardiovascular interventions (surgery and transcatheter aortic valve implantation) 29

 5.3 Prehabilitation 29

6 GAPS IN EVIDENCE 29

7 SEARCH STRATEGY 29

8 REFERENCES 30

List of Boxes

BOX 1: 5-meter gait speed assessment  6

BOX 2: 6-minute walk test (6MWT) assessment  6

BOX 3: Katz Index of activities of daily living assessment  7

BOX 4: Psoas muscle area measures and sarcopenia assessment  7

BOX 5: Fried Criteria ( Cardiovascular Health Study) assessment  8

BOX 6: Clinical Frailty Scale assessment  9

BOX 7: Short Physical Performance Battery (SPPB) assessment  10

BOX 8: Edmonton Frail Scale (EFS) assessment  13

BOX 9: Composite indexes assessment  20

List of Figures

Figure 1: Paper identification process  5

Figure 2: Illustration of the condensed EACTS/EAPC consensus statement on frailty assessment to predict outcome of cardiac surgical procedures (EuroFORECAST(cs)) based on the actually available evidence  16

Figure 3: Illustration of the condensed EACTS/EAPC consensus statement on frailty assessment to predict outcome of cardiovascular interventions (cardiac surgery and transcatheter aortic valve implantation, EuroFORECAST) based on the actually available evidence  27

Figure 4: Illustration of the concept of cardiac prehabilitation before cardiac surgery or transcatheter aortic valve implantation procedures (adapted from [273])  28

Consensus statements

Consensus statements: Gait speed as a prediction tool for early mortality after cardiac Surgery  6

Consensus statements: Assessment of ADLs as a prediction tool early mortality after cardiac surgery  7

Consensus statements: Psoas muscle area measurement as a prediction tool for early mortality after cardiac surgery  8

Consensus statements: Fried criteria as a prediction tool for early mortality after cardiac after cardiac surgery  9

Consensus statements: Clinical Frailty Scale as a prediction tool for early mortality after cardiac surgery  9

Consensus statements: SPPB as a prediction tool for early mortality after cardiac Surgery  10

Consensus statements: Composite indexes as prediction tools for early mortality after cardiac surgery  10

Consensus statements: The Fried criteria as prediction tool for postoperative delirium and prolonged hospitalization  11

Consensus statements: Psoas muscle measures as prediction tool for postoperative delirium and prolonged hospitalization  11

Consensus statements: Measurement of nutritional status as prediction tool for postoperative delirium and prolonged hospitalization  12

Consensus statements: SPPB as prediction tool for postoperative delirium and prolonged hospitalization  12

Consensus statements: Edmonton Frail Scale as prediction tool for postoperative delirium and prolonged hospitalization  13

Consensus statements: Cognitive measurements as prediction tools for postoperative delirium and prolonged hospitalization  13

Consensus statements: Fried criteria as prediction tool for quality of life after cardiac surgery  14

Consensus statements: Fried criteria as prediction tool for readmission and discharge to an intermediate-care facility  14

Consensus statements: Psoas muscle size measurement as prediction tool for readmission and discharge to an intermediate-care facility  15

Summary of consensus statements on frailty assessment in cardiac surgery Consensus statements: prediction of mortality after cardiac surgery  15

Consensus statements: Prediction of neurological complications/delirium and prolonged hospitalization/ventilation after cardiac surgery  15

Consensus statements: prediction of quality of life, discharge location and probability of readmission after cardiac surgery  16

Consensus statements: Gait speed as prediction tool for mortality after TAVI  17

Consensus statements: Serum albumin as prediction tool for mortality after TAVI  17

Consensus statements: Measurement of ADLs as prediction tool for mortality after TAVI  18

Consensus statements: Handgrip strength as prediction tool for mortality after TAVI  18

Consensus statements: The Clinical Frailty Scale as prediction tool for mortality after TAVI.  19

Consensus statements: Psoas muscle measurements as prediction tool for mortality after TAVI  19

Consensus statements: MMSE as prediction tool for mortality after TAVI  20

Consensus statements: Composite indices as prediction tools for mortality after TAVI  20

Consensus statements: Nutritional state as prediction tool for neurological complications/delirium and prolonged ventilation/hospitalization after TAVI  22

Consensus statements: 5m-gait speed as prediction tool for neurological complications/delirium and prolonged ventilation/hospitalization after TAVI  22

Consensus statements: MMSE as prediction tool for neurological complications/delirium and prolonged ventilation/hospitalization after TAVI  22

Consensus statements: Psoas muscle measures as prediction tool for neurological complications/delirium and prolonged ventilation/hospitalization after TAVI  23

Consensus statements: Frailty assessment as prediction tool for quality of life after TAVI  24

Consensus statements: 5-meter gait speed as prediction tool to estimate the probability of discharge to another location than home after TAVI  24

Consensus statements: Fried criteria and serum albumin as prediction tool to estimate the probability of discharge to another location than home after TAVI  25

Consensus statements: Different frailty assessments as prediction tool to estimate the probability of early or late readmission after TAVI  25

Summary of consensus statements on frailty assessment in TAVI Consensus statements: Prediction of mortality after TAVI  25

Consensus statements: prediction of neurological complications/delirium and prolonged hospitalization/ventilation after TAVI  26

Consensus statements: Prediction of quality of life, discharge location and probability of readmission after TAVI  26

Abbreviations and acronyms

 
• 5-mWT

  5-meter walk test

 
• 6MWT

  6-minute walk test

 
• ADL

  Instrumental activity of daily living

 
• AUC

  area under the curve

 
• BMI

  body mass index

 
• CABG

  coronary artery bypass grafting

 
• CFS

  Clinical Frailty Scale

 
• CI

  confidence interval

 
• CS

  cardiac surgery

 
• CT

  computed tomography

 
• DXA

  dual-energy X-ray absorptiometry

 
• EACTS

  European Association for Cardio-Thoracic Surgery

 
• EAPC

  European Association of Preventive Cardiology

 
• EFS

  Edmonton Frail Scale

 
• EFT

  Essential Frailty Toolset

 
• EWGSOP

  European Working Group on Sarcopenia in Older People

 
• GNRI

  Geriatric Nutritional Risk Index

 
• HR

  hazard ratio

 
• IADL

  instrumental activities of daily living

 
• ICU

  intensive care unit

 
• KCCQ

  Kansas City Cardiomyopathy Questionnaire

 
• KI

  Katz Index

 
• MCI

  mild cognitive impairment

 
• MMSE

  Mini-Mental State Examination

 
• PAI

  psoas muscle area index

 
• PMA

  psoas muscle area

 
• OR

  odds ratio

 
• QoL

  quality of life

 
• SAVR

  surgical aortic valve replacement

 
• SF-36

  36-item short-form

 
• SPPB

  Short Physical Performance Battery

 
• STS

  Society of Thoracic Surgeons

 
• TAVI

  transcatheter aortic valve implantation

1 INTRODUCTION

1.1 General considerations and rationale

In an ageing population of patients with heart disease, preinterventional frailty assessment for risk prediction is gaining relevance both in cardiac surgery and transcatheter valve interventions like transcatheter aortic valve implantation (TAVI). However, there is to date no consensus or widely accepted recommendations on how to determine frailty for prediction of outcomes of cardiovascular operations or interventions despite an exponentially growing number of studies in this field. Adding frailty assessment to conventional risk scores for cardiovascular interventions began in the context of heart-team discussions when TAVI for patients with aortic valve stenosis became an alternative for surgical aortic valve replacement (SAVR) [1]. Validated algorithms such as the EuroSCORE II [2] and the Society of Thoracic Surgeons (STS) Risk Score [3, 4] are usually used to estimate the perioperative risk of surgical procedures. They predominantly assess the risk of short-term mortality. For the decision between treatment modalities (surgery vs transcatheter procedure), this is not always sufficient because the typical patient population comprises elderly patients with a complex panel of comorbidities, physiological particularities and age-related risk factors. Therefore, risk assessment must take into account more and different factors than those considered in younger patients. For example, the above-mentioned scoring systems just include the chronological ages of the patients. Consequently, they tend to overestimate the perioperative risk of otherwise resilient elderly patients and, on the other hand, tend to underestimate the risk of patients who have lower resistance against stressors like a cardiac intervention or an operation. Additionally, frailty is accompanied by other patient-related factors like a decline in quality of life (QoL). Today, SAVR and TAVI are complementary techniques for the treatment of severe aortic stenosis with comparable short and mid-term outcomes according to several randomized controlled trials. Due to the remaining uncertainties concerning long-term (>5 years) TAVI valve durability, SAVR remains the preferred therapy in younger patients (<75 years of age) who are low risk for surgery (STS PROM or EuroScore II ≤4%) according to current European Society of Cardiology/European Association for Cardio-Thoracic Surgery guidelines [5]. Usually, the end points that are mainly taken into consideration are short-term mortality and neurological complications like stroke. Softer end points like QoL are handled as a lower priority. This approach is in contrast to patients’ growing demands for prediction of their postoperative or postinterventional QoL and independent living situation. However, conventional risk scores are not validated for these end points. Frailty assessment is an additional factor for estimating the outcome of surgery or interventions in elderly patients who are low risk based on conventional risk scoring and can thus sharpen the estimation of the individual patient’s risk for an unfavourable outcome. A growing body of literature is available investigating frailty assessment as a predictor for the above-mentioned end points. Unfortunately, the literature is very heterogeneous. Maybe because of this lack of consistency, assessment tools are not routinely used in daily clinical routine. As a result, the European Association for Cardio-Thoracic Surgery (EACTS) and the European Association of Preventive Cardiology (EAPC) decided to make an effort to review the available literature and to condense major findings in order to develop EACTS/EAPC consensus statements. Because the available literature is extensive and inconsistent, the members of the working group decided to limit the recommendations to surgical procedures and TAVI in order to maintain some clarity. Of course, there is also literature on frailty assessment related to other transcatheter interventions such as transcatheter edge-to-edge mitral valve repair. We propose to apply the recommendations in this consensus statement to these patient groups as well, because they are similar cohorts within the same risk group as TAVI patients.

2.2 Methodology and results

Members of this task force were selected by the EACTS and EAPC to represent professionals involved with the medical care of these patients. Besides surgical and cardiological experts, a geriatrician and an anaesthesiologist became task force members. A biostatistician, who gave advice on the methodology, literature review and development of the consensus statements, participated in the developmental process. All experts on the writing and the reviewing panel signed conflict of interest forms, declaring potential sources of conflicting interests. The task force received its entire financial support from the EACTS and the EAPC. There were no external funds or support from the industry. EACTS, supported by EAPC, supervised and coordinated the development of this consensus statement. Both entities agreed to submit the document in parallel to the European Journal of Cardiothoracic Surgery and the European Journal of Preventive Cardiology.

A systematic review was performed based on EACTS and EAPC policies for the development of scientific documents. Search terms were used that included frailty assessment, transcatheter aortic valve interventions and cardiac surgical procedures. The complete list of search terms is printed in the Supplement. The search was restricted to Medline. Every identified abstract was reviewed by 2 task force members. In case of disagreement, a third task force member also reviewed the relevant paper and gave a judgement in regards to inclusion or exclusion. The detailed review process of full-text papers was the same as that mentioned previously. One of the main criteria for inclusion of a paper was that the study examined the predictive ability of a particular frailty assessment tool for one of the outcomes of interest.

As a result, a total of 1181 papers published between the earliest available point of time in the database and January 2022 were identified, 357 of which met inclusion criteria and were read in full. Of these, 103 publications had to be excluded based on predefined exclusion criteria, so that 254 papers were included in the final analysis (Fig. 1).

Statistical methods to combine the individual studies in regard to their ability to predict the outcomes and their validity could not be applied due to the large variety of tools investigated, the different methods applied in the studies to calculate associations and the variable end points. Therefore, consensus statements are based on how often certain tools were described and investigated as successful predictors of certain outcomes. Additionally, the value of a study to be considered for inclusion in our consensus statements was based on the expertise of the task force members. There was a continuous discussion about the consensus statements during the development of the manuscript.

Three main areas of outcomes were identified that have mainly been investigated and are seen as most relevant from the point of view of the patients and the surgeons/interventionalists: (i) mortality, (ii) neurological complications including delirium and consecutive prolonged hospitalization and (iii) quality of life including discharge location and readmission rate. The consensus statements are structured according to these 3 aspects.

2 FRAILTY AND THE PREDICTION OF OUTCOME AFTER CARDIAC SURGICAL PROCEDURES

2.1 Frailty as a predictor of short-term, intermediate-term and long-term mortality after cardiac surgery

In cardiac surgery, the most frequently used risk scores to predict short- (up to 30 days) or intermediate- (up to 1 year) term mortality are the EuroSCORE II [2] and the STS score [3, 4]. However, these risk scores have been shown to over- or underestimate the perioperative risk. One possible reason might be that they do not include frailty assessment. In this systematic review, the literature was searched meticulously (see above): A total of 58 papers was found reporting frailty assessment tools as predictors of short-, intermediate- and long-term mortality in patients undergoing cardiac surgery. In these 58 papers, almost all authors used a different frailty assessment tool. The most commonly described tools were gait speed, activities of daily living (ADL), the 5-item Canadian Health Study Assessment, psoas muscle measures and the Clinical Frailty Scale (CFS).

2.1.1 Gait speed and walking distance

Gait speed is most often described as the time needed to walk a short distance, i.e. 4 [6] or 5 meters [7–16] at a steady walking speed, but also as an endurance test with regard to the distance covered in the 6-minute walk test (6MWT) (Box 1 and Box 2) [17,18,19]. Gait speed was applied in 13 studies and was the most commonly applied tool to assess 1 component of frailty in the context of the prediction of mortality. Of these 13 identified papers, 10 found gait speed to be predictive of mortality [6–11, 13, 14, 18, 20].

Most of the studies assessed perioperative and short-term mortality (≤ 30 days) [6, 8–10, 13]. Two studies with the largest number of patients deserve particular mention: Afilalo and colleagues included 15,171 patients above the age of 60 from 109 centres who underwent cardiac surgery [coronary artery bypass grafting (CABG), aortic or mitral valve surgery, or combinations of these]. Approximately one-third of the patients were identified as slow walkers, defined as a gait speed of less than 0.83 meters per second and had an increased risk of operative mortality [odds ratio (OR) 3.16; confidence interval (CI) 2.31–4.33]. There was also a correlation between gait speed and perioperative mortality. It has been shown that when measured as a continuous variable, gait speed added relevant information about the risk of mortality independently from other risk scores (OR 1.11 per 0.1 m/s decrease gait speed; CI 1.07–1.16, P < 0.001).

The second study by Shih and colleagues [9] included 19,743 patients from the STS database of whom 2,061 had recorded gait speed measures (5-m gait speed). Of these, around 20% were classified as slow walkers, also defined as a gait speed of less than 0.83 meters per second, which was predictive for perioperative mortality (OR 2.15; CI 1.27–3.63, P = 0.004). In other studies, gait speed was not used as an isolated frailty component, but as part of a multicomponent frailty assessment tool, for example as part of the comprehensive assessment of frailty (CAF) test [21]. However, even as an isolated component, Bäck and colleagues found gait speed to be an independent predictor of 30-day mortality [OR 5.0 (CI 1.6–16); P = 0.006] [6].

Intermediate- (up to 1 year) [11, 14] and long-term (more than 1 year) [11, 18] mortality has been investigated as well, e.g. in a large study by Afilalo and colleagues [11], which included 8,287 patients scheduled for cardiac surgery (CABG, aortic or mitral valve surgery or combinations of these) and for whom assessment of gait speed was recorded in the STS database. The investigators performed a landmark analysis for different points of time up to 36 months. The patients’ median age was 74 years. Frailty was assessed by the 5-meter gait speed. They found that slow gait speed, defined as a gait speed of less than 0.83 meters per second, was a strong predictor of mortality both between 30 days and 1 year (adjusted HR 2.28; CI 1.57–3.32) and also beyond 1 year (adjusted HR 1.41; CI 1.00–1.99).

Gait speed as a continuous variable was also predictive of 1-year mortality (HR 2.16 per 0.1m/s decrease in gait speed; 95% CI, 1.59–2.93). This result remained significant when gait speed was measured as a continuous variable, with the highest hazard ratio (HR) found at 1 year (adjusted HR 2.16 per 0.1 m/s decrease gait speed; 95% CI 1.59–2.93).

Arenaza and colleagues analysed the correlation of frailty defined by endurance measured by the 6MWT and intermediate term survival [18]. Frailty was assessed by the 6MWT. They included 208 patients undergoing AVR and found 49% to be fast walkers (>300 meters in 6 min). Similarly, Clavel and colleagues reported that a walking distance of below 320 meters in the 6MWT was a predictor of increased mortality in patients undergoing SAVR (HR 2.55; CI 1.6–4.35) [19]. Similar to the publication by Afilalo and colleagues [11], where slow walkers had higher mortality, fast walkers had a lower HR (HR 0.28; CI 0.09–0.85, P = 0.25) for mortality at 1 year. Also, in the STICH trial [17], endurance gait speed was derived from 6MWTs in 1,212 patients to predict long- term mortality after cardiac surgery, but no significant correlation of endurance gait speed and long- term mortality was found.

Gait speed as part of a multicomponent frailty assessment (Green Score [22]) was also described as an incremental and independent predictor of 1-year mortality in a publication by Bo and colleagues [13], who found a significant correlation of gait speed and 1-year mortality in patients undergoing TAVI or SAVR (HR 0.14; CI 0.02–0.85, P = 0.033).

Conclusions for gait speed as prediction tool for mortality after cardiac surgery:

• 5-Meter gait speed is an easily accessible assessment tool.

• 5-Meter gait speed is a predictor of short-, intermediate- and long-term survival or mortality, respectively.

• For the 6MWT, only limited data in regard to predicting mortality are available in the target population of cardiac surgical patients.

2.1.2 Katz Index of activities of daily living

The second most common test, the Katz Index (KI) of ADL [24] assessment, was reported as a frailty parameter to predict mortality in 6 publications [10, 13, 25–28]. The KI of ADL is a measurement of deficiency of independence in the daily activities as described in Box 3. Lee et al. reported a study that included 3,826 patients with a median age of 66 (15–94) years [28]. Patients with any deficiency in the KI were deemed frail (4.1%) and had a significantly increased risk of in-hospital mortality (OR 1.8; CI 1.1–3.0, P = 0.03). In contrast, in a study by Nagai and colleagues, frailty was not a predictor of in-hospital mortality [29] but of reduced midterm survival (HR 1.5; CI 1.1–2.2, P = 0.01). A more recent study by Hiraoka and colleagues included 113 patients undergoing elective total arch surgery. The median age of the cohort was 72.6 years. The KI of ADLs was one of several indices used to predict midterm mortality and was identified as a predictive categorical index with an HR of 7.6 (CI 1.16–29.53, P = 0.038). Only these 2 studies identified the KI of ADLs as a predictor of mortality. In contrast, 3 studies [10, 13, 25] found ADLs not to be predictors of mortality. Two of them used the KI [10, 12]; 1 used other definitions of decline in ADLs [10, 13, 25].

Conclusions for ADLs as a prediction tool for mortality after cardiac surgery:

• The Katz Index of ADLs has been described as an independent predictor of mortality in just 2 out of 6 papers.

• Overall ADLs have not been shown to be predictors of mortality in cardiac surgical patients.

2.1.3 Psoas muscle measurements and sarcopenia

Lately, measurements of the dimensions of the psoas muscle as parameters for sarcopenia and frailty have been suggested. The psoas muscle can reproducibly be measured from computed tomography (CT) scans. The main body of the psoas muscle can be found in the CT scans at the height of L3/L4. Different measures like psoas muscle area, volume or diameters were described. More detailed information can be found in Box 4. In 6 studies, psoas muscle measurements (PMM) [27, 30–34] were used as correlates for sarcopenia. Sarcopenia was defined either by PMM below the lower percentiles of the population, control groups or the study population. Unfortunately, there is no standard definition of low psoas muscle size. In the recommendations of the European Working Group on Sarcopenia in Older People (EWGSOP), the reference values for the appendicular skeletal muscle mass (AMMI) measured by dual-energy X-ray absorptiometry (DXA) or bioelectrical impedance analysis are provided (<7 kg/m² for men and <5.5 kg/m² for women). Due to the less broad availability of CT equipment, the cross-sectional areas of specific muscle groups assessed by CT are recommended to be used in research rather than in clinical routines [35]. The majority of studies identified in this systematic review used a psoas muscle area index (PAI) below the gender- and/or age-specific 25th percentile of the study cohort. In publications from non-Asian groups, also including the studies that investigated TAVI cohorts [30, 36, 37], low PAI for men was defined as <8.5 cm²/m² and for women <6.5 cm²/m². All but 1 [27] study showed the prediction of mortality by PMMs. Ikeno et al. included 266 patients undergoing elective total aortic arch replacement with a mean age of 76.2 years [31]. They found that 30.5% of the patients were frail (PAI > 2 standard deviations below the PAI of the general population) with a 2.5-fold increased risk of 5-year mortality (HR 2.59; CI 1.27–5.29, P = 0.011) but no differences in in-hospital mortality. Similar results were found in a study by Hawkins and colleagues [30]: 1-year survival in non-sarcopenic (defined by PAI) patients increased (OR 0.84, P = 0.02), with a C-statistic value of 0.65; no differences were found for perioperative mortality. Three more studies using PAI as a frailty correlate confirmed these results: Okamura and colleagues [33] had 25% frail patients in a cohort of 304 patients with increased late (HR 4.25; CI 2.18–8.28, P < 0.001) but no differences in short-term mortality. Kurumisawa and colleagues [32] included 138 haemodialysis patients undergoing cardiac surgery of which 25% were frail. They found low PAI as an independent risk factor for 5-year mortality (HR 1.94; CI 1.19–3.17, P = 0.01). Yamashita and colleagues [34] included 773 patients undergoing all types of cardiac surgical procedures and found low muscle attenuation to be a predictor of increased long-term mortality (HR 2.23; CI 1.17–4.23, P = 0.022). In 2 other studies, PAI was used as part of a multicomponent frailty assessment that was predictive for late mortality, excluding hospital death [38, 39]. Unfortunately, in both studies, the predictive value of PAI alone was not described. Therefore, it is not clear which of the assessed parameters had the highest impact on the total score. The AMMI measured by DXA to measure sarcopenia as defined by the EWGSOP was assessed in a publication by Joshi and colleagues [40]. They found AMMI to be a predictor of all-cause mortality [median 4.3 years, HR 1.84 (95% CI 1.18–2.86)] in a cohort of 141 patients. Skeletal muscle mass measured in CT was assessed in 874 patients undergoing SAVR in a study by Lee and colleagues [41]. In addition to 30-day and 1-year mortality, they found SMM also to be predictive of long-term mortality (median 2.8 years, HR per unit decrease in the z-score, 1.52; 95% CI, 1.33–1.75; P < 0.001).

Conclusions for psoas and skeletal muscle measures as prediction tools for mortality after cardiac surgery:

• PMMs obtained from CT images are independent of the patient's ability to move and therefore can, for example, also be used in patients in the ICU.

• The PAI has been shown in several studies to be a predictor of intermediate- to long-term survival.

• The PAI does not predict perioperative mortality.

• The PAI standard values for specific patient populations have to be defined.

• Alternatively, the skeletal muscle mass measured by CT or by DXA (as defined by the EWGSOP) can be assessed as a prognostic variable for short-, intermediate- and long-term mortality.

2.1.4 Fried frailty phenotype: 5 items assessment

The 5-item frailty assessment based on data from the Cardiovascular Health Study was the first standardized frailty assessment tool and is considered the gold standard in geriatric medicine for measuring physical frailty. In the landmark publication by Linda Fried and colleagues published in 2001 [42], frailty has been defined as present when 3 or more of the following apply (See Box 5):

The 5,317 participants were 65 years or older men and women from a community-dwelling cohort. The study showed that frailty was a significant predictor of mortality for up to 7 years after baseline evaluation.

The Fried criteria have also been used as a frailty assessment tool to predict mortality after cardiac surgery as described in 7 publications [7, 10, 45–49]. Three of these showed that frailty measured by the Fried criteria predicted mortality [46–48]. The study published by Piñón and colleagues [46] included 183 patients; 31% were deemed frail with a greater than 3-fold increased risk of 1-year mortality (HR 3.41; 95% CI 1.02–11.45, P = 0.05). This result could not be confirmed in the analysis of the 3-year results.

Rodriguez-Pascual and colleagues showed that frailty measured as suggested by Fried [42] can add value by predicting outcomes of reduced survival up to 98 weeks after interventions related to the aortic valve (HR 3.2; CI 2.37–4.32) [47]. The cohort comprised 606 patients. TAVR and SAVR were performed. Because the patients were not split into groups, it was not possible to determine for which procedure frailty assessment might be a predictor. Afilalo and colleagues reported on 1,020 patients, of whom 374 underwent SAVR and found that frailty as defined by the Fried criteria were predictive of 1-year mortality (OR 1.63) [CI 1.12–2.37, total cohort; OR 1.35 (CI 1.36–3.09) [48]].

In all other papers mentioned at the beginning of this chapter, there was no significant predictive value.

Conclusions for the Fried criteria as a prediction tool for mortality after cardiac surgery:

• The Fried criteria are considered the gold standard tool to measure physical frailty in geriatric medicine.

• Fried criteria have been extensively validated in many studies with a strong prediction of in- hospital complications and mortality.

• Fried criteria can be assessed with intermediate effort.

• For cardiac surgery, limited studies have used the Fried criteria to assess short-, intermediate- and long-term mortality. This may be explained by the greater effort required to apply the 5-item multicomponent tool.

2.1.5 Clinical Frailty Scale

The CFS is an easily accessible tool because it does not require any instruments like e.g. dynamometers and does not rely on individual measures like gait speed. But neither is it an eyeball test, because a considerable amount of time has to be invested in the evaluation of the patient to assess the relevant domains like disability for basic and instrumental activities of daily living, mobility, activity, energy and disease-related symptoms. As a result, it has the potential to provide a subjective and comprehensive estimation of a patient's frailty status. Rockwood and colleagues first described the CFS [50]. It is a semi-quantitative tool providing a score from 1 (very fit) to 7 (severely frail, terminally ill) that is given by a trained medical professional who has direct contact with the patient, for example at an interview for admission or a similar situation (see Box 6). The CFS was used as a frailty assessment tool in 8 studies [14, 27, 48, 51–55]. Reichart and colleagues included 6,156 patients undergoing isolated CABG [51]. Preoperative frailty status was assessed by applying the CFS. They reported the highest area under the curve (AUC) mentioned in the literature for a frailty assessment tool in patients undergoing cardiac surgery (AUC 0.823; CI 0.783–0.863) and an almost 6 times higher risk of 30-day mortality in patients with a CFS > 5 when adjusted for the EuroSCORE II [OR 5.90 (2.67–13.05); P < 0,001] and a 3-fold increased risk of 1-year mortality when adjusted for the EuroSCORE II (HR 3.05; CI 1.83–5.06, P < 0,001). An increased risk of 1-year mortality was shown in 2 studies: Hiraoka and colleagues [27] (HR = 4.05; CI 1.22–14.06, P = 0.023) and Lytwyn and colleagues [53] [(adjusted for EuroSCORE II) (OR = 2.08; CI 1.05–4.12)] published similar findings. A significant predictive value for the CFS was also found by Kovasc and colleagues with an AUC of 0.778 (0.649–0.878) for postoperative mortality [52] and by Rodrigues et al. (OR 2.763; CI 1.206–6.331, P = 0.0001). A study by Afilalo and colleagues [48] included 1,020 patients; these authors found the CFS to be predictive of 1-year mortality (OR = 2.40; CI 1.63–3.52). However, this study included patients who underwent TAVR or SAVR. The predictive ability of the CFS for the 374 patients undergoing SAVR was found to be not significant (OR 2.52; CI 0.95–6.68). Naganuma and colleagues found the CFS to be predictive for midterm mortality in a study including 2019 patients undergoing SAVR (OR 1.18; 95% CI 0.91–1.53, P = 0.204) [55].

Conclusions for the CFS as a prediction tool for mortality after cardiac surgery:

• The CFS is easily accessible (no tools necessary).

• The CFS is a standardized subjective estimation of a patient's frailty status, but it is not an eyeball test and requires training.

• The CFS has been shown to add predictive value for the estimation of perioperative and intermediate-term mortality.

2.1.6 Short Physical Performance Battery

The Short Physical Performance Battery (SPPB) comprises 3 batteries of tests: balance (standing with feet side by side, semi-tandem and tandem), gait speed and the chair stand test (see Box 7). The SPPB was assessed and analysed for prediction of mortality in patients undergoing cardiac surgery in 4 articles [48, 53, 56, 57]. Three of the papers found SPPB to be a significant predictor of 1-year mortality [48, 53, 56]. Lytwyn and colleagues included 188 patients with a median age of 71 years: 52.6% were deemed frail. The predictive value for 1-year mortality was significant (OR 3.47; 95% CI 1.69–7.12) [53]. Afilalo and colleagues included 1,020 patients and found an increased risk of 1-year mortality (OR 3.62; 95% CI 1.33–9.81) in the 374 patients undergoing SAVR [48]. Goldfarb and colleagues found in their cohort of 1158 patients from the FRAILTY-AVR Study that SPBB is a predictor of 1-year mortality with an OR of 1.14 (CI 1.07–1.20) for every 1-point decrease in the SPBB measures [56].

Conclusions for SPPB as a prediction tool for mortality after cardiac surgery:

• SPPB is an easily accessible functional test with 3 components (balance, gait speed, chair stand test)

• SPPB measured frailty values and showed good predictive abilities for intermediate-term mortality.

2.1.7 Composite indexes

The CAF score was the first established and validated frailty assessment specifically for cardiac surgical patients; it was published by Sündermann and colleagues in 2011 [21]. It was applied in a cohort of 400 patients and consists of Fried criteria, physical performance tests (get-up-and-sit-down, balance tests, pick up a pen from the floor and put on and remove a jacket), laboratory and physiological measures (albumin, creatinine, forced expiratory volume in 1 s) and a CFS assessment by 2 physicians. A defined point value is given for all parameters, and the points are combined to create a comprehensive score. To simplify the test, the FORECAST (Frailty predicts death One year after Elective Cardiac Surgery Test) was condensed from the CAF using the parameters with the highest AUC [58, 59].

The CAF and FORECAST scores were found to be predictive for short- and intermediate- term mortality in all studies [6, 21, 58–60]. Sündermann and colleagues also included patients undergoing TAVI, whereas the study by Bäck and colleagues included 604 patients only undergoing cardiac surgery. Frail patients had an increased 30-day mortality and CAF was an independent predictor (OR 1.1; CI 1.0–1.2, P = 0.044). The CAF score was associated with good predictive ability with an AUC of 0.70 (CI 0.56–0.84) for 30 days and an AUC of 0.73 (CI 0.64–0.82) for 1-year mortality [6, 60]. Bäck and colleagues found the CAF and the FORECAST also to be predictive for long-term mortality for up to 3 years (2.31, 95% CI 1.74–3.07, P < 0.001 and 2.68, 95% CI 1.88–3.81, P < 0.001 respectively) [61].

The Essential Frailty Toolset (EFT) is mainly addressed in studies including patients having TAVI. Nevertheless, it should be mentioned here because it gained relevant importance in the heart team. In patients undergoing cardiac surgery, 2 studies of interest are available. The first is the initial publication of the EFT by Afilalo and colleagues [48]. In the subgroup of patients undergoing SAVR, they found frailty to be predictive of 1-year mortality when added to the STS-PROM score [adjusted OR (aOR) 4.38; 95% CI 1.81–10.61]. Solomon and colleagues used the EFT to assess frailty in a cohort of 500 patients undergoing CABG [62]. They found a clear correlation of frailty and all-cause mortality. Each incremental EFT point was associated with an HR of 1.28 (95% CI, 1.05–1.56). Additionally, frail patients had a threefold increase in all-cause mortality. Lately, population-based cohort studies using linked health administrative data have become available. For example, McIsaac and colleagues investigated a cohort of 61,389 patients 65 years or older undergoing cardiac surgery in Canada [63].

Conclusions for composite indexes as prediction tools for mortality after cardiac surgery:

• CAF combines several dimensions of frailty assessment in 1 test.

• CAF in its original form is time consuming (approx. 20 min to complete).

• The EFT should be validated in more studies with surgical patients.

2.2 Frailty as predictor of neurological complications/delirium and prolonged ventilation/hospitalization after cardiac surgery

In previous decades, mortality after cardiac surgery continued to decrease; therefore, QoL after cardiac surgery became an important measure of success. Treatment of complications is usually expensive and lengthy; thus, prevention of complications would be ideal. In addition, a complicated perioperative period can undermine recovery and can have long-term adverse consequences. Postoperative complications such as neurological events and delirium and length of stay in the intensive care unit (ICU) and the hospital are frequently used markers for the measurement of postoperative morbidity and higher resource utilization. These parameters are rarely investigated in studies. Therefore, we summarized these end points under one “umbrella” of end points that lead to higher resource utilization and finally to higher mortality. Frailty assessment tools seem to help with the development of a more precise prediction of morbidity during the hospital stay. In the systematic review, 27 studies were identified that investigated postoperative complications like delirium and length of ICU and/or hospital stay as primary or secondary outcomes. Several comprehensive, functional, nutritional, mental and cognitive frailty measures were investigated in these studies: the Fried test [64–72], the SPPB battery [66, 73, 74], sarcopenia [30, 75–78] and nutritional status [67, 73, 75, 79, 80] were mainly applied, either alone or concurrently with other frailty tools and medical variables. In addition to reduced ADL assessed from health-care records [81], nationally translated and modified tests, like the Japanese version of the Cardiovascular Health Questionnaire [82] or university-developed protocols and score sheets, like the Duke Activity Status Index [45, 74] or the Johns Hopkins Adjusted Clinical Groups frailty indicator [83, 84], were also used as well as handgrip strength, which has been demonstrated to be predictive for postoperative length of stay, but only in 1 study [85]. Comprehensive geriatric assessments were used in 2 studies [86, 87].

2.2.1 Fried frailty phenotype—5-item assessment, modified Fried criteria and single components

Assessment of frailty with the Fried criteria (see Box 4) has been used as a primary frailty tool in 7 studies [65, 66, 68–72] or in combination with other batteries in 2 studies [64, 67]. Gait speed as a component of the Fried criteria was investigated in 1 study [88]. Four papers investigated the relationship between delirium and the Fried criteria [64, 65, 68, 69]. Jung prospectively investigated the occurrence of delirium defined by the Confusion Assessment Method for the Intensive Care Unit test [89] in 133 patients [64]. The investigators found that frail patients defined by the modified Fried criteria had an increased risk for postoperative delirium [adjusted OR 5.05 (CI 1.58–16.13)] after adjustment for the EuroSCORE II. Additionally, frail patients had a higher risk for a prolonged hospital stay (aOR 2.31; CI 1.15–4.65, P = 0.02). In a preliminary observational study, Brown and colleagues reported that frailty measured by the 5-domain Fried criteria was independently associated with increased risk for delirium after adjustment for age, prior stroke, depression and the Charlson comorbidity score (absolute risk reduction: 18.3; CI 2.1–161.8, P = 0.01) [65]. These results were confirmed after a propensity score adjustment. Nomura and colleagues divided their patients into frail, prefrail and non-frail groups using the Fried criteria [68]. They found that, among frail patients, prefrail (adjusted odds ratio aOR 6.43; CI 1.31–31.64, P = 0.02) and frail patients (aOR 6.31; CI 1.18–33.74, P = 0.03) had a higher risk for delirium. Frail patients also had a higher adjusted z-score of cognitive decline at 1 month but not at 1 year after surgery. Li and colleagues found frailty assessed by the Fried criteria in a cohort of 298 patients to be predictive for postoperative delirium (OR 4.9; P < 0.001) [69]. These results were confirmed by Cheng and colleagues, who also found frailty measured by the Fried criteria to be predictive for postoperative delirium in a cohort of 152 patients [72]. Goldfarb investigated the relationship between frailty and hospital costs as a surrogate marker of perioperative adverse events in patients above 60 years of age undergoing CABG [66]. In the multivariable model, hospital costs were independently associated with frailty in patients undergoing CABG combined with valve surgery. Frailty was defined as Fried criteria >3 out of 5 or SSPB battery score <5 out of 12. The Fried test was also applied in correlation with new frailty markers, such as phase angle measured by bioimpedance [67] (see also the sections Mortality and Sarcopenia). The modified Fried criteria have been used in a study including 95 patients undergoing cardiac surgery by Delaney and colleagues [70]. A prediction of length of ICU stay could be shown (OR 0.9; CI 0.1–0.7, P = 0.02). Similar results were described by Nakano and colleagues, who showed an increased duration of hospitalization in frail patients defined by the Fried criteria (OR 1.35; CI 1.19–1.52, P < 0.001) [71].

Conclusions for the Fried criteria as a prediction tool for neurological complications, postoperative delirium and prolonged hospitalization:

• Fried criteria show predictive values for postoperative delirium, higher resource utilization and longer ICU and hospital lengths of stay.

• Fried criteria can be assessed with intermediate effort.

2.2.2 Psoas muscle measures and sarcopenia

Sarcopenia, assessed by the measurement of the cross-sectional area of the psoas muscle and their derivates (see Box 3), has been found to be independently associated with prolonged length of stay, increased resource utilization and higher hospital costs [30, 75–78]. Hawkins et al. found that sarcopenia defined as the PAI was independently associated with postoperative length of stay (–0.46, P < 0.0001) and hospital cost (−0.03, P = 0.0010 [30]). In the study by Teng et al., sarcopenia defined with PAI plus low physical performance or low gait speed was associated with longer hospitalization after adjustment for other clinical variables [77]. Kiriya defined sarcopenia by total psoas muscle index and intramuscular adipose tissue content [75]. In their multivariable model, both the low total psoas muscle index (OR: 2.581; CI 1.152–5.783, P = 0.036) and the high intramuscular fat index (aOR 3.973; CI 1.737–9.088, P < 0.001) were independently associated with prolonged hospitalization after cardiac surgery. In the study of Taniguchi and colleagues, trunk muscle (erector spinae muscles) cross-sectional area but not psoas muscle was found to be independently associated with length of hospitalization after SAVR [76]. Zuckermann et al. compared the discriminative value of muscle area at different levels [67]; thoracic, lumbar and psoas cross-sectional areas were correlated with the length of hospital stay. After adjustment for age, sex and body surface area in a multivariable linear regression model, the measurement of psoas muscle remained associated with length of postoperative hospital stay (beta: –2.35, CI –4.48 to –0.22).

Conclusions for psoas muscle measures as a prediction tool for neurological complications, postoperative delirium and prolonged hospitalization

• Psoas muscle area has been shown to be a predictor of length of hospital stay after cardiac surgery.

2.2.3 Nutritional state measurements

Malnutrition is a known risk factor of poor physical reserve in addition to age and comorbidities. Malnutrition has been recognized as an important part of geriatric and frailty syndromes, and measurement of a patient's nutritional state has become an essential part of comprehensive frailty tools. The Geriatric Nutritional Risk Index (GNRI) is calculated by including body weight, height and albumin levels and has been widely validated in cardiac and other hospitalized populations. GNRI is an objective measurement and predicts nutritional state more accurately than albumin or body mass index (BMI) alone [90]. We found 5 papers investigating the correlation of nutrition and adverse outcomes in cardiac surgery [67, 73, 79, 91, 92]. In the study by Ogawa and colleagues [73], a total of 131 patients awaiting cardiac surgery were enrolled. Complex physical performance tests (e.g. SPPB, 6MWT, grip strength) were performed and the GNRI were measured. In a multivariable regression analysis, low GNRI values were independently associated with the length of hospital stay (beta -0.36, P = 0.037) and prolonged time to walk without assistance (beta: −0.33, P = 0.02) after adjustment for other medical variables. Unosawa and colleagues also found that malnutrition defined as GNRI < 91 was independently associated with the length of hospital stay for more than 1 month (OR: 3.428; CI 1.687–6.964, P < 0.001) and with a prolonged bedridden stay (OR 7.377; CI: 1.874–29.041, P = 0.004) in 287 patients undergoing elective cardiac surgery [79]. Mullie and colleagues used phase angle measured by bioimpedance to objectively determine the body composition and hydration state [67]. Phase angle was strongly associated with physical functioning, nutritional state and sarcopenia. In this study, the Mini-Nutritional Assessment Short Form score was used. Phase angle was associated with the length of hospitalization after adjustment for STS risk factors (adjusted by 4.8 days per 1° decrease in PA; 95% CI 1.3–8.2 days) in a linear regression model. Taniguchi and colleagues showed that a low albumin level was an independent predictor of prolonged hospitalization (per 0.1 g/dl; OR 0.83; CI = 0.70–0.99, P < 0.05) and that it exhibited a strong relationship with erector spinae muscles surface area [76]. In a large database analysis (7,446 patients), Chassé and colleagues found that waist circumference was independently associated with several postoperative adverse outcomes, such as prolonged hospitalization and ICU stay after adjustment for medical variables [92]. Their results indicate that variations in the distribution of the adipose tissue are also an important marker of postoperative complications.

Conclusions for nutritional state assessment as a prediction tool for neurological complications, postoperative delirium and prolonged hospitalization:

• The estimation of the nutritional state by the simple GNRI may help capture the frailty state in older cardiac patients.

• Assessment of nutritional state and body composition is useful for the prediction of prolonged hospitalization.

2.2.4 Short Physical Performance Battery

The SPPB is described above (see Box 6). In 4 publications, SPPB either alone or simultaneously with other frailty tools was used to predict length of in-hospital stay [66, 73, 93, 94]. In the study by Goldfarb and colleagues, a composite frailty measure of Fried criteria and SPPB was used in 235 patients scheduled for bypass or valve surgery who had a median age of 73 years [66]. In the multivariate analysis, Goldfarb et al. found that frail (defined as an SPPB equal to or below 5) patients had a higher risk for prolonged hospitalization of more than 14 days (aOR 5.67; CI 2.73–11.77, P < 0.001). Ogawa and colleagues investigated the role of nutritional status regarding physical condition and postsurgical rehabilitation in patients undergoing cardiac surgery with the use of cardiopulmonary bypass machines and who were older than 65 years. Besides the nutritional state, SPPB was assessed preoperatively. In the multivariable regression model, SPPB (adjusted beta for 1 unit: −0.31, P = 0.018) and nutritional status were independently associated with prolonged hospitalization [73]. In the publication by Rao and colleagues, SPPB was used as a predictor of delirium [93]. Han and colleagues found SPPB and its single components to be predictive for a prolonged hospitalization after CABG (OR 0.55; 95% CI 0.48–0.64) [94].

Conclusions for assessment of SPPB as a prediction tool for neurological complications, postoperative delirium and prolonged hospitalization:

• SPPB assessment has been shown to be predictive for prolonged in-hospital stay and postoperative delirium but only in a small number of publications.

2.2.5 Edmonton Frail Scale

The Edmonton Frail Scale (EFS) was developed at the University of Alberta in 1999 by Darryl Rolfson [95]. EFS is a multidimensional frailty index extending from physical frailty alone to 9 domains including cognitive performance, general health status, mood, functional dependence and performance, nutritional state, medication, continence and social support (see Box 8). The EFS score ranges from 0 to 17 points and cut-offs are defined for the severity of frailty. In the context of this consensus paper, 3 publications are relevant [96–98]. Kiss and colleagues found that frail and prefrail patients had longer ICU and hospital stays compared to the non-frail patients; they also found significant differences in their fibrinogen and C-reactive protein levels as well as white blood cell counts [96]. In the study of Amabili and colleagues, patients were dichotomized by the EFS as frail or non-frail. The authors found that, after adjustment for their EuroSCORE, frail patients had a lower probability of being discharged from the ICU [adjusted HR (aHR): 0.67; CI: 0.48–0.94, P = 0.02), and a higher probability of being discharged to a health-care facility [97]. Abdullahi and colleagues assessed the Edmonton Frail Scale in a cohort of 80 patients undergoing different cardiac surgical procedures [98]. Additionally, the patients wore wrist accelerometers 2 weeks before and 1 month after the operation. The EFS and the physical activity measured by the accelerometers predicted the length of the hospital stay (OR 0.134; CI 0.106–0.162, P < 0.001 and OR 1.76; CI 0.003–3.524, P = 0.05, respectively) and the grade of postoperative activity levels. A prolonged length of stay was correlated with the occurrence of postoperative complications (stroke, renal failure, reoperation, pacemaker implant).

2.2.6 Cognitive measurements: Mini-Mental State Examination and Montreal Cognitive Assessment

The Mini-Mental State Examination (MMSE) has been used widely in clinical research and care to assess cognitive function [99]. Thus, the MMSE may be considered a measure of cognitive frailty. It comprises tests of orientation, attention, memory, language and visual–spatial skills. The highest score is 30 points; the cut-offs are tailored by the education of the patient. Below 24 points, the MMSE score is pathological. In the Comprehensive Geriatric Assessment-Based Frailty Index, the MMSE score is ranked as 27–30 (best cognitive function); 24–26; 21–23 and below 21 points. Increasingly in research and clinical care, the Montreal Cognitive Assessment (MoCA), a more challenging and complex cognitive test, is used. Ziad Nasreddine developed it in 1996 in Canada [100]. The MoCA assesses several cognitive domains, like visual–spatial skills, memory, attention and language. The highest score is 30 points. Similar to the MMSE, the test is also used to detect mild cognitive impairment (MCI). Consistent with the fact that older adults with cognitive impairment have an increased risk of delirium, in the study of Eide and colleagues, patients with surgically treated aortic valve disease with a higher MMSE score were found to have a lower occurrence of delirium (OR 0.85; CI 073–0.98, P = 0.03) [101]. Itagaki and colleagues enrolled 89 patients and investigated how MCI and physical frailty influence the occurrence of postoperative delirium [82]. They found that both MCI and physical frailty were associated with postoperative delirium in the multivariate model. The coexistence of MCI and frailty had the highest risk for postoperative delirium (aOR 7.494; CI 1.539–36.494; P = 0.013).

Conclusions for cognitive measurements as a prediction tool for neurological complications, postoperative delirium and prolonged hospitalization:

• Assessment of cognition is important in the prediction of the development of delirium after cardiac surgery.

• The MMSE and MoCA have successfully been used in the prediction of postoperative delirium.

2.3 Frailty as predictor of quality of life, discharge location and readmission after cardiac surgery

In addition to postoperative morbidity and mortality, older patients are primarily interested in their postoperative QoL, functional independence and the course of their re-convalescence. Ideally, patients are discharged to their own homes after surgery, do not need to be hospitalized again and have no new functional limitations. Elderly patients have much more interest in the prediction of these softer end points than in the prediction of mortality. As already discussed in the introduction part in section 3.2, these end points are less often investigated and therefore are summarized under the “umbrella” of end points related to the postoperative QoL and living situation without readmission to the hospital. Twenty-two papers were identified in the systematic review that investigated end points describing QoL, readmission, discharge location and impaired function after surgery [7, 11, 28, 30, 38, 45, 64, 68, 73, 83, 85, 97, 102–111]. Tools to measure QoL were the 36-item short-form (SF-36) [105] or the EuroQoL 5-dimensions questionnaire [106]. Another end point that was chosen to be relevant for the consensus statements was the discharge location. Several studies investigated if frailty is a factor enabling prediction of patient discharge location, for example, to an intermediate health-care hospital instead of being discharged to their home [7, 28, 30, 38, 64, 83, 97, 104, 107]. Readmission to hospitals was investigated in 5 publications [11, 30, 45, 83, 112].

2.3.1 Clinical frailty scale and multiscale assessment of frailty to predict quality of life and time to discharge to home after cardiac surgery

In the identified publications that aimed to show that frailty is a predictor of QoL after cardiac surgical procedures, the SF-12 [45], SF-36 [105], the Kansas City Cardiomyopathy Questionnaire (KCCQ) [113] and the EuroQol 5-dimension questionnaire [106] as baseline and outcome measurements were used.

In the publication of Marshall and colleagues [105], 123 patients were assessed for frailty before open cardiac surgery. Eleven different frailty scales were used and condensed into 1 score, grading the patient as “robust”, “borderline” or “frail”, suggesting that there is a correlation between frailty and the postoperative QoL. In particular, the authors could show that “borderline” patients had the most relevant improvement in the postoperative QoL for most of the categories compared to “robust” and “frail” patients, but these results were only by trend. Nevertheless, it shows that the assessment of frailty adds relevant information, especially when it comes to categorizing patients with regard to their perioperative risk. Miguelena and colleagues [106] included 137 patients in their study. Frailty was measured using the Fried criteria (see Box 3), CFS (see Box 4) and the FRAIL (fatigue, resistance, ambulation, illnesses and loss of weight) scale. QoL was assessed before and 6 months after the operation using the EuroQoL 5-dimension 5-level questionnaire [114]. The study had findings similar to those of Marshall et al. showing that the borderline and frail patients had greater improvement in regard to QoL after open cardiac surgery. Kotajarvi and colleagues also used the Fried criteria to assess frailty in patients undergoing SAVR [45]. They found that physical well-being in particular improved more in the group of frail patients. None of the studies investigated whether frailty assessment is a tool to predict QoL after cardiac surgery but showed that there might be a correlation. Only 1 study in patients undergoing heart valve surgery by Borregaard and colleagues showed that frailty at discharge, assessed by the Fried criteria, is predictive of reduced QoL as measured by the EQ-5D-5L Index (OR 3.38; CI 1.51–7.52), the visual analogue scale (OR 2.41; CI 1.13–5.14) and the KCCQ (OR 2.84; CI 1.35–5.97) [113]. Hill and colleagues did a longitudinal follow-up of patients who underwent cardiac surgery (49 patients) [115]. They found that frail patients according to the CFS needed longer time to be discharged to home (HR 0.54; CI 0.34–0.86) and that frailty reduced the number of patients discharged to home.

Conclusions for the Fried criteria as a prediction tool for QoL, discharge location and readmission after cardiac surgery:

• There are insufficient studies with regard to frailty assessment and the relevance for QoL after cardiac surgery. This can be seen as a gap in evidence.

• The Fried criteria have mainly been used to assess frailty in the context of QoL.

• Only 1 study showed that frailty is a predictor of QoL after cardiac surgery.

• QoL assessment might be performed before and after surgery as a quality control and as a baseline for further studies.

2.3.2 Fried frailty phenotype—5 items assessment and modified scales

The Fried frailty phenotype (see Box 3) has been used in 3 publications [7, 45, 68, 74] for the prediction of discharge location and readmission as well as for the prediction of postoperative functional/neurological impairment. Ad and colleagues found in a study of 166 patients who underwent frailty assessment with the Fried criteria (see Box 5) that frailty remained significantly related to greater odds for discharge to an intermediate-care facility (OR 3.13; CI 1.24–7.91, P = 0.016) in multivariate analysis [7]. None of the other end points such as morbidity and mortality and readmission were predicted by the frailty status of the patients in this study. The odds for readmission were not significant either. However, gait speed alone as one component of frailty was a predictor for discharge to an intermediate-care facility (OR 1.95; CI 1.24–3.06, P = 0.004). Kotajarvi and colleagues also found that the assessment of the frailty status by using the Fried frailty phenotype criteria can add valuable information to predict rehospitalization after discharge (OR 2.61; CI 1.08–6.75) [45]. The Fried frailty phenotype was assessed in 103 study participants with aortic stenosis who underwent surgical procedures or TAVI. In the same paper, it could also be shown that non-frail participants had a greater improvement in their QoL. Unfortunately, prediction of postoperative outcomes by frailty parameters was not investigated separately between surgical and transcatheter procedures. Therefore, it is not possible to draw any conclusions from this cohort with regard to these comparisons. Nakano also found that frail patients, defined by the Fried criteria, have a higher probability of not being discharged to home (OR 3.25; CI 1.37–7.69, P = 0.007) as well to have a functional decline (OR 2.41; CI 1.03–5.63, P = 0.04) [71].

Conclusions for the Fried frailty phenotype assessment as a prediction tool for QoL, discharge location and readmission after cardiac surgery:

• The Fried frailty phenotype assessment is associated with discharge location and readmission after cardiac surgery.

• Gait speed alone is a predictor for discharge to an intermediate-care facility.

2.3.3 Psoas muscle measures and sarcopenia

The PAI [30] and the total psoas volume as part of a multiscale frailty assessment [38] were used as predictors of discharge to an intermediate-care facility [30] or discharge to a place other than home [38]. Hawkins and colleagues assessed frailty by measuring the PAI (see Box 2) in a cohort of 245 patients who were scheduled for SAVR with or without coronary artery bypass surgery [30]. They found that PAI remained predictive after risk adjustment for postoperative discharge to a facility instead of discharge to the previous home (OR 0.88; CI 0.8–0.96, P = 0.005, C-statistic 0.612). Ganapathi and colleagues used a multiscale frailty assessment including psoas muscle volume measurement (see Box 2) in a cohort of patients who underwent aortic and proximal arch surgery [38]. They found frailty to be predictive for discharge to a destination other than home (OR 3.7; CI 1.8–7.7, P < 0.01).

Conclusions for PMM as a prediction tool for QoL, discharge location and readmission after cardiac surgery:

• PMM can be performed easily as part of routine CT scans even on unconscious patients.

• Based on limited data from 2 studies, the PAI may predict discharge to a facility or to a place other than home.

• Psoas muscle volume measurement has been shown in 1 publication to be not predictive for readmission within the first 30 days after surgery.

2.4 Summary of consensus statements for frailty assessment in cardiac surgery

2.4.1 Consensus statements: prediction of mortality after cardiac surgery

 

2.4.2 Consensus statements: prediction of neurological complications/delirium and prolonged hospitalization/ventilation after cardiac surgery

2.4.3 Consensus statements: prediction of quality of life, discharge location and probability of readmission after cardiac surgery

2.5 Expert task force condensed advice on frailty assessment to predict outcome of cardiac surgical procedures by frailty assessment

2.5.1 The European Frailty Score for Cardiovascular Interventions (cardiac surgery) [EuroFORECAST(cs)]

Based on the available evidence, it is advised to follow the frailty assessment pathway illustrated in Figure 2. Work-up should be started with the easy-to-use tools 5-Metre Gait Speed Test and Clinical Frailty Scale to assess if the probability of frailty is high or low. If it is high, more comprehensive frailty assessment is indicated. The Fried frailty too has been widely validated also for outcomes other than mortality and for patients in other fields of medicine. If a CT-scan is not part of routine patient examination, the assessment of the psoas muscle should not be performed as it adds radiation exposure to the patient. We suggest grading a patient as not frail if none of the initial assessments is positive, as frail if 2 initial plus 1 additional or 1 initial plus 2 additional assessments are positive and as severely frail if 2 initial plus 2 additional assessments are positive (illustrated in Figure 2).

The Comprehensive Assessment of Frailty tool [21] can be used, too. It incorporates all of the advised assessments and has been validated in large patient cohorts [21, 60, 61]

3 FRAILTY AND THE PREDICTION OF OUTCOME OF TRANSCATHETER AORTIC VALVE IMPLANT PROCEDURES

3.1 Frailty as a predictor of short-, intermediate- and long-term mortality

More than 100 studies [23, 24, 36, 48, 50, 91, 99, 116–216] were included, and more than 50 different assessments were used to define the frailty status of the patients. The most used assessments were gait speed [5-meter walk test (5-mWT) or the 6MWT], serum albumin level, activities of daily living (ADL), handgrip strength, the CFS, instrumental activities of daily living (IADL), psoas muscle area (PMA by CT scan), MMSE and composite indexes (Box 6). There was pronounced heterogeneity in the methodologies of the assessment methods used, study designs and statistical analyses. Overall, most of the findings were related to short (up to 30 days)- and long-term (more than 1 year) mortality.

3.1.1 Gait speed and walking distance

A total of 29 studies assessed gait speed with the 5-mWT (Box 1) or the 6MWT (Box 8) [22, 37, 116, 118, 120–122, 128, 134, 135, 139, 148, 150, 154, 156, 157, 165, 180, 189, 197, 200, 208, 217, 218]. A cut-off value of ≥ 6 s (gait speed < 0.8 m/s) and the patient’s stratification as a “unable/slow/fast walker” were the most used methods to identify frail patients. Short-term mortality was assessed in 13 investigations [118, 120–122, 139, 144, 150, 156, 169, 170, 180, 197, 200], and conflicting information was found. For example, Kiani et al. [156] found in their retrospective analysis (36,242 participants) that lower results of the 5-mWT (cut-off > 6 s) or “being unable to perform the test” was predictive of 30-day death (HR 1.21; CI 1.00–1.47). Similarly, Afilalo et al. found an OR of 3.17 (CI 1.17–8.59) in 131 participants for reduced walking speed (cut-off > 6 s) [120]; Alfredsson et al. [118] showed an OR of 1.16 (CI 1.06–1.28) per every 0.2 m/s decrease in 8,039 patients. However, Arnold et al. [121] reported that only dichotomized results (unable to walk vs able to walk/speed first percentile) were associated with short-term mortality (OR 1.27; CI 1.02–1.58) but not gait speed reduction (every 0.2 m/s decrease OR 0.95; CI 0.89–1.02) in 21,661 patients investigated retrospectively. In contrast, 7 authors [121, 122, 150, 165, 180, 197, 200] stated that reduced gait speed was not associated with mortality. In fact, Steinvil et al. [200] found that lower results of the 5-mWT were not associated with a 30-day risk of mortality (OR 1.74; CI 0.36–8.50) in 498 patients analysed retrospectively. This result was also seen in a study by Kure and colleagues in which a 15-foot walk test was not predictive for cumulative mortality in a cohort of 280 patients [165]. Long-term mortality was assessed in 16 investigations (7 prospective [22, 116, 135, 157, 189, 208, 218], 2,360 patients/9 retrospective [128, 134, 139, 150, 154, 156, 200, 217], 44,222 patients). Eleven studies [116, 128, 134, 135, 139, 154, 156, 157, 208, 217, 218] found reduced gait speed to be predictive of mortality. Kano et al. [154] showed in a cohort of 1,256 patients who were slower walkers (< 0.5 m/s, HR 1.83; CI 1.03–3.26) and patients unable to walk had higher long-term mortality (unable, HR: 4.28; CI 2.22–8.72). Further, Steinvil et al. [200] found an OR of 2.34 (CI 1.03–5.32) in 498 patients (retrospective), and Abdul-Jawad et al. [116] showed in his prospective study that preinterventional slow gait speed (6MWT) was associated with an increased risk of long-term mortality (HR 2.30; CI 1.35–3.93). Despite this finding, Sathananthan et al. [189] found that 5-mWT results were not predictive of long-term mortality (OR: 0.78; CI 0.35–1.72) in 755 participants. Also, Hermiller et al. [150] reported that analysis of walking speed was predictive only at the univariate level (HR: 1.42; CI 1.06–1.91) in a sample size of 3,687 subjects. In contrast, the same group found in a study with 2,037 patients that the 6MWT is a strong predictor of 2-year mortality (HR 0.87 per 50 m; CI 0.83–0.92, P < 0.001) [190].

Four studies [37, 143, 144, 184] were available for intermediate-term mortality. One found after subanalyses that gait speed alone (cut-off value 0.5 m/s) was not associated with mortality (P = 0.174); however, the mortality rate was higher in patients with reduced walking speed [37]. Goel and colleagues found that a slow gait speed 1 year after the procedure, independently from the initial gait speed, is a strong predictor of death/hospitalization between 1 and 2 years [143]. For those patients in whom slow gait speed normalized after the procedure, no reduced risk was observed. Similar, Roca and colleagues identified low walking speed to be a predictor for 1-year mortality [184].

Conclusions for gait speed as a prediction tool for mortality after TAVI:

• 5-mWT is a predictor of long-term mortality.

• 5-mWT is a feasible test for patients having TAVI and might be easily incorporated into clinical practice.

• Different cut-off values and test methodologies were used in the literature; therefore, results have to be interpreted with caution.

• For the 6MWT, limited data are available referring to long-term mortality.

3.1.2 Serum albumin

Preoperative serum albumin levels were measured in 17 studies [22, 123, 124, 128, 139, 148–150, 156, 159, 162, 174, 180, 191, 198, 200, 208]. Serum albumin is considered a marker of malnutrition, is easy to measure and has high reliability [159]. Mostly, a cut-off value of < 3.5 g/dl was used to classify frail patients. Notably, however, the utility of serum albumin and the cut-off of < 3.5 g/dl as a measure of nutritional status has been questioned in older patients with low physical function and even among those without inflammation. Nonetheless, it has been used broadly in cardiac patients and is therefore addressed in these consensus statements.

Short-term mortality was investigated in 7 retrospective studies [139, 150, 156, 180, 191, 198, 200] with conflicting data. For instance, Kiani et al. [156] reported an HR of 1.29 (CI 1.12–1.48) in 36,242 patients for serum albumin level < 3.5 g/dl. Hermiller et al. [150] found an HR of 1.60 (CI 1.04–2.47) in 3,687 subjects, and Shimura et al. [197] noted an HR of 2.36 (CI 1.64–3.40) in 1,542 participants when results were considered as a categorical variable.

Long-term mortality was investigated in 14 studies (4 prospective [22, 123, 159, 208], 10 retrospective [124, 128, 139, 148–150, 156, 162, 174, 200]). Most of the investigations found that a preinterventional lower level of serum albumin was associated with a higher hazard of 1-year mortality with HR values ranging from 1.40 (CI 1.04–1.91) [150] to 3.12 (CI 1.80–5.42) [128]. Accordingly, Steinvil et al. [200] reported an OR of 2.21 (CI 1.12–4.37) in 498 patients (≤ 3.5 vs > 3.5 g/dl) and Shimura, an OR of 0.39 (CI 0.28–0.52, P < 0.001) [197].

Conclusions for serum albumin as a prediction tool for mortality after TAVI:

• A lower (≤ 3.5 g/dL) preoperative serum albumin level is a predictor of long-term mortality.

• A lower (≤ 3.5 g/dL) preoperative serum albumin level is not consistently associated with increased short-term mortality.

• The serum albumin level is an easily accessible parameter but may not be reliable in older patients with impaired function or an inflammatory state.

3.1.3 Activities of daily living

Fifteen studies [22, 126, 128, 129, 135, 139, 146, 148, 157, 172, 173, 182, 200, 203, 218] assessed ADL with the Katz Index [24], where generally the presence of at least 1 impaired activity was used to define frailty (cut-off values < 4 or < 6/6). Short-term mortality was investigated in 7 studies (3 prospective [129, 146, 182], 5 retrospective [139, 172, 173, 200, 219]). Two [172, 173] of them reported increased odds of mortality for a Katz score < 6/6 (OR 1.27; CI 1.11–1.44 and 2.10; CI: 1.39–3.15). However, this was not confirmed in 2 of the prospective studies [129, 146] (OR 1.07; CI 0.64–1.77 in 312 patients [129]) and in 2 retrospective [200, 219] analyses (OR 2.43; CI 0.58–10.20) accounting for 498 participants. Long-term mortality was assessed in 15 studies (9 prospective [22, 126, 129, 135, 146, 157, 182, 203, 218], 6 retrospective [128, 139, 148, 172, 173, 200]) and for this outcome, too, controversial information was found. For example, Puls et al. [182] and Chauhan et al. [128] showed similar adjusted HR of 2.50 (CI 1.60–3.90) and 2.45 (CI 1.42–4.22) in 330 and 342 patients, respectively, for Katz scores < 6. However, in 2 [173, 200] retrospective analyses of larger cohorts (498 and 2624 subjects), an OR of 1.43 (CI 0.59–3.45) [200] for Katz scores ≤ 4/6 and an adjusted HR of 1.23 (CI 0.86–1.75) [173] for scores < 6 were reported. For intermediate-term mortality, only 2 studies [126, 173] reported a score < 6/6 predictive of mortality (HR 1.74; CI 1.19–2.55) [173].

Instrumental Activities of Daily Living (IADL) were assessed in 9 prospective studies [122, 125, 126, 136, 146, 192, 193, 201, 208], mostly with the Lawton Index [167]. Most of these studies found that any impairment in the IADL scale (cut-off < 8/8) was not predictive of short- [122, 146, 201], intermediate- [125, 193] and long-term [136, 192, 201, 208] mortality. Only 2 investigations [126, 146] reported IADL results as predictive for intermediate- and long-term mortality (HR 1.20; CI 1.07–1.33) [146].

Conclusions for ADL assessment as a prediction tool for mortality after TAVI:

• A Katz Index of ADL showed controversial information on the prediction of short- and long-term mortality after TAVI.

• For the Katz Index of ADL, information about the prediction of intermediate-term mortality after TAVI is limited.

• The Lawton Index of IADL is not a predictor of mortality after TAVI.

3.1.4 Handgrip strength

In 11 studies [22, 128, 135, 139, 146, 148, 150, 157, 197, 200, 218], reduced handgrip strength values were considered a marker of frailty in TAVI patients. In general, strength values were obtained using a hand dynamometer and then normalized to the patient’s BMI and stratified by sex. Different reference values were used between studies resulting in heterogeneity of cut-off scores. Further, some investigations rated handgrip strength based on the clinician's judgement of the patient's hand grasping (weak/mild/strong). Short-term mortality was investigated in 5 studies (2 prospective [146, 197], 3 retrospective [139, 150, 200]): All authors found lower handgrip values not predictive of mortality rate (e.g. Steinvil et al. [200], OR 2.24; CI 0.28–17.80).

Long-term mortality was assessed in 9 investigations (4 prospective [22, 135, 157, 218], 5 retrospective [128, 139, 148, 150, 200]) showing conflicting information. For instance, Chauhan et al. [128] showed that a weak handgrip was associated with an increased mortality hazard at 1 year (HR 3.31; CI 1.01–10.85). Similarly, Forcillo et al. [139] stated that the odds of 1-year mortality decreased as the values of handgrip strength increased. Conversely, Steinvil et al. [200] found that reduced handgrip values were not predictive of 1-year mortality in 498 patients (OR 1.63; CI 0.66–4.06) and Hermiller et al. [150], in 3,687 participants. No data were available for intermediate-term mortality.

Conclusions for handgrip strength as a prediction tool for mortality after TAVI:

• Handgrip strength seems not to be predictive of short-term mortality.

• Handgrip strength showed controversial predictive abilities for long-term mortality.

3.1.5 Clinical Frailty scale

Twelve studies [37, 48, 151, 175, 189, 194, 197, 198, 202, 213, 214, 216] rated frailty using the CFS [50]. Based on clinical judgement, patients were classified on a categorical scale where lower scores indicated independence and absence of frailty (Box 3). Short-term mortality was considered in 3 studies (2 prospective [48, 197], 1 retrospective [198]). Two investigations [197, 198] showed that per each increase in the score, a higher mortality hazard was found (HR 1.42; CI: 1.04–1.95 in 1,215 patients [197] and HR 1.17; CI: 1.01–1.34 in 1,542 patients [198]). However, Afilalo et al. [48] showed no association between higher CFS scores (≥ 5/9) and odds of mortality (OR 1.87; CI 0.99–3.53) in 1, 020 subjects.

Long-term mortality was investigated in 9 studies (4 prospective [48, 189, 194, 214], 4 retrospective [151, 175, 187, 198]) presenting divergent information. Afilalo et al. [48] found that higher CFS results were associated with 1-year mortality (OR 2.40; CI 1.63–3. 52). Seiffert et al. [194] reported an HR of 1.31 (CI 1.13–1.52) in 845 participants; Yamamoto and colleagues found an OR of 1.30 (CI 1.15–1.47, P < 0.001 per 1 scale). In contrast, Yokoyama et al. [214] and Saji et al. [187] found that the results of the CFS were not associated with a death hazard (scores ≥4/9 vs <4/9 HR 1.03; CI 0.60–1.8677 and per 1 score increase HR 1.21; CI 0.35–1.5557) in 767 and 455 patients, respectively. Mortality 2 years after the intervention in nonagenarians was investigated by Noguchi and colleagues [176]. A CFS ≥4 was associated with increased 2-year mortality (HR 1.82; CI 1.02– 3.42, P = 0.04). Zisiopoulou and colleagues also found the CFS to be a predictor for midterm mortality for up to 2 years [HR 0.25 (95% CI =1.09–1.44, P = 0.002)] [216].

Conclusions for the CFS as a prediction tool for mortality after TAVI:

• For the CFS, limited information is available with regard to short-term mortality.

• CFS showed controversial evidence for long-term mortality.

• CFS is an easily accessible assessment tool based on clinical judgement.

3.1.6 Psoas muscle measurements

These values were used primarily to classify patients into 3 groups, additionally stratified by sex. Short-term mortality was assessed in 2 prospective studies [36, 207] (1,659 patients) in which Kofler et al. [36] found mortality associated with PMA measurements (OR L3: 0.082; CI 0.011–0.589 and OR L4: 0.049; CI 0.005–0.536). Long-term mortality was investigated in 11 studies (5 prospective [36, 142, 171, 207, 218], 6 retrospective [138, 140, 141, 162, 174, 210]), mostly showing an association with PMA measurements. In fact, Krishnan et al. [162] showed an association between lower PMA values and higher mortality risk (HR 2.50; CI 1.10–4.60) in 381 patients, and Kofler et al. [36] found that PMA values were inversely associated with mortality odds (OR L3: 0.200; CI 0.083–0.482 and OR L4: 0.083; CI 0.029–0.235) in 1,076 patients. However, Mamane et al. [171] found PMA tertile values predictive of mortality only for female patients (per 1 cm² increase in PMA HR: 0.88; CI 0.78–0.99). Michel et al. [174] found that risk of death was not associated with results of psoas muscle assessments when organized in quartiles and stratified by sex in 1,731 subjects analysed retrospectively. Waduud and colleagues found that PMMs were not associated with mortality [211].

Two recent studies investigated psoas muscle [210] attenuation and psoas muscle density [140] in the context of mortality prediction after TAVI. Both found these parameters to be relevant predictors of long-term mortality [5-year mortality: OR 2.871; 95% CI 0.880–9.371 (psoas muscle attenuation); 3- year mortality: OR 4.55; 95% CI 2.41–40.00, P = 0.03 (psoas muscle density)].

Intermediate-term mortality was assessed by only 1 retrospective analysis (Saji et al. [37]); the authors reported that PMA tertiles were an independent predictor for mortality at 6 months (HR: 1.53; CI 1.06–2.21) in 232 patients.

Conclusions for psoas muscle assessment as a prediction tool for mortality after TAVI:

• PMA measurement can be performed in TAVI patients without additional burden, because a routine CT scan is performed for procedure planning.

• PMA is a predictor of long-term mortality.

• Data on the association of PMA and short- or intermediate-term mortality are scarce and inconclusive.

• PMA is easily accessible from routine preprocedural CT scans.

3.1.7 Mini-Mental State Examination

Nine studies [122, 125, 136, 146, 192, 193, 198, 201, 208] used the MMSE questionnaire [99], where a cut-off value of < 27/30 points was representative of cognitive frailty. The MMSE assesses cognitive function (orientation, attention, memory, language and visual-spatial skills) and is widely used among older populations. Short-term mortality was assessed in only 3 prospective [122, 146, 201] studies with small cohorts (402 patients). Two [122, 146] of them showed no association between lower MMSE scores and mortality; however, Storteky et al. [201] reported the opposite (OR 7.62; CI 1.44–40.19) result. Long-term mortality was investigated in 6 studies (5 prospective [136, 146, 192, 201, 208], 1 retrospective [198]), most of which reported no association with MMSE results. For instance, Goudzwaard et al. [146] stated that reduced MMSE scores were not predictive of 1-year mortality (HR 1.60; CI 0.76–3.22). Eichler et al. [136] and van der Wulp et al. [208] also found no association with mortality in sample sizes of 344 and 511 patients investigated prospectively. Only 2 studies [192, 201] found lower MMSE results (categorical variable) associated with higher odds of long-term mortality (OR 2.98; CI 1.07–8.31 [201] and OR 2.35; CI 1.33–4.14 [192]). Intermediate-term mortality was referred to in 2 small prospective studies ([125, 193], 349 patients total) that showed divergent information.

Conclusions for the Mini-Mental State Examination as a prediction tool for mortality after transcatheter aortic valve implantation:

• MMSE seems not to be predictive of long-term mortality.

• Information about MMSE that refers to short- and intermediate-term mortality is minimal and conflicting.

3.1.8 Composite indexes

Overall, 4- or 5-item indexes, the EFT [48] and the Bern scale [201] (Box 9) were the most frequently used tools. Assessments addressing mobility, strength, physical activity, exhaustion, malnutrition, cognition and level of independence are used, according to the concept of frailty as a multifactorial syndrome [132]. Test results were then converted into a categorical index; higher scores usually indicated frailty of greater severity. Fifteen studies used prospective and 5 studies used retrospective observations, with most of them showing an association between index scores and postinterventional deaths. A 4-item index was used in 7 investigations (4 prospective [22, 48, 148, 177], 3 retrospective [128, 152, 178]). Afilalo et al. [48] found that a score of ≥ 3/5 (dichotomized) was associated with higher odds of mortality (OR: short-term 2.65; CI 1.28–5.49 and OR: long-term 3.04; CI 1.98–4.66) in 1,020 patients. Similarly, Chauhan et al. [128] showed for higher scores an increased mortality hazard (score ≥ 3/4 HR 3.05; CI 1.24–7.46) and an AUC of 0.7133 in a cohort of 342 participants. A 5-item index was used in 6 studies (5 prospective [48, 119, 137, 185, 195], 1 retrospective [200]). Rogers et al. [185] found that, in a cohort of 544 patients, those who had higher scores (≥ 3/5) were associated with higher odds of short- (OR 5.06; CI 1.36–18.80) and long-term mortality (OR 2.75; CI 1.55–4.87). Similarly, Afilalo et al. [48] and Steinvil et al. [200] reported an OR of 1.63 (CI 1.12–2.37) and 2.23 (CI 1.14–4.34) in 1,020 and 498 participants, respectively, for dichotomized results of the index. The Bern scale was used in 5 prospective investigations [48, 136, 192, 193, 201]. Afilalo et al. [48] found higher scores on the scale (score ≥ 3/7) associated with short- (OR 3.29; CI 1.53–7.07) and long-term mortality (OR 2.57; CI 1.69–3.91); similar results were found by Stortecky et al. [201] per each point increase of the score (OR 2.18; CI 1.32–3.61). The EFT was used in 6 studies [48, 133, 181, 188, 189, 199]. Four investigations [48, 133, 181, 189] found higher scores on the EFT to be predictive of a 1-year death rate (e.g. OR: 3.72; CI 2.54–5.45) [48] and 3 of them, of short-term [48, 133, 181] mortality (e.g. OR: 3.50; CI 1.74–7.07) [133] in cohorts ranging from 723 to 1,020 patients. Additionally, Pighi et al. [181] showed an AUC of 0.80 (CI 0.74–0.86) in female and of 0.77 (CI 0.71–0.83) in male participants. Saji and colleagues showed good abilities for up to 2 years using the EFT to predict mortality up to 2 years (AUC 0.74; CI 0.62 –0.84) [188]. Two more composite indexes are found in single publications: Multidimensional assessments based on the comprehensive geriatric assessment [147] and the Hospital Frailty Risk Score, which is calculated on the basis of diagnoses and conditions reported in the International Classification of Diseases (ICD) [117]. Both found frailty to be predictive for intermediate and long-term mortality. A new score has been introduced in Japan: The Kihon Checklist. It is a questionnaire with questions about instrumental and social activities of daily living, physical functions, nutritional status, oral function, cognitive function and depressive mood. It is easily accessible but is used in only 1 study [165]. The results of the latter 3 studies have to be confirmed in further studies to be advisable for frailty assessment in the cohort of cardiovascular patient.

Conclusions for composite indexes as prediction tools for mortality after TAVI:

• The 5-item index is a predictor of short- and long-term mortality.

• The EFT is a predictor of short- and long-term mortality.

• The 4-item index and the Bern scale seem to be predictive of short- and long-term mortality.

3.2 Frailty as predictor of neurological complications/delirium and prolonged ventilation/hospitalization after transcatheter aortic valve implantation

This section includes studies that investigated the use of frailty tools to predict postoperative neurological complications, length of ICU and/or hospital stay after TAVI as primary or secondary outcomes. Several comprehensive, functional, nutritional, mental and cognitive frailty measures have been used in that context: nutritional status [139, 145, 146, 160, 180, 181, 196, 197, 212, 220–228], gait speed [118, 139, 180, 196, 197, 220, 228–233], Canadian Study of Health and Ageing [173, 196, 197, 228, 231], MMSE [122, 192, 196, 220, 231], sarcopenia [142, 171, 225, 234, 235], Fried criteria [137, 181, 223] and SPPB battery [181, 236] were the main measures used. The EFT was used in 1 study [237]. These measures were applied, either alone or together with other frailty tools, along with medical variables.

3.2.1 Nutritional state measurements

Nutritional state measurements have been widely used in patients undergoing TAVI procedures as patient-selection criteria, as preoperative risk stratification or as an important factor of early safety. In this regard, the preprocedural albumin level [139, 180, 196, 197, 220, 221, 223–226], BMI [160, 181, 197, 212, 227, 228], GNRI score [145, 196] and Malnutrition Universal Screening Tool [146, 222] were used either alone or as a component of the frailty assessment. Gassa et al. examined 457 patients retrospectively [221]. The preprocedural serum albumin level was dichotomized as low level (<3.5 g/dl) versus normal level (>3.5 g/dl). In multivariate logistic regression models, occurrence of acute kidney failure (OR: 3.09; CI1.5–6.5, P = 0.003), infection (aOR: 2.17; CI 1.1–4.2, P = 0.024), need for blood transfusions (OR: 2.85; CI 1.4–5.9, P = 0.004) and prolonged postinterventional ventilation (OR: 3.24; CI 1.7–6.2, P < 0.001) were independently associated with albumin levels lower than 3.5 g/l. From the OCEAN-TAVI (Optimized transCathEter vAlvular iNtervention) Japanese multicentre registry (OCEAN-TAVI registry), Yamamoto et al. reported a higher prevalence of postprocedural complications and prolonged length of ICU and hospital stay in patients with hypoalbuminemia compared to the group of patients without hypoalbuminemia, after propensity matching [days; 2.0 (1.0 to 3.0) vs 1.0 (1.0 to 3.0), P < 0.001; and 12.0 (8.0 to 20.0) vs 10.0 (7.0 to 15.0); P < 0.001, respectively] [226]. Forcillo found that each 1 g/dl increase in the preprocedural albumin level decreased the odds of the 30-day composite outcome (1 g/dl albumin; OR 0.32, P < 0.0001) [139]. Osnabrugge reported that albumin levels below 3.3 g/l showed a tendency toward a poor outcome (OR 1,8; CI 0.9–3.5, P = 0.073) [224]. Patel combined low albumin levels with a 5-mWT and divided the patients into 2 groups: (i) frail and (ii) non-frail [180]. The adjusted mean difference in hospital stay was 4.3 (2.5–6.2) days, P < 0.001. Shimura et al. reported a strong correlation between the CFS grade and albumin levels (ρ=0.22, P < 0.001) [197]. Shibata et al. investigated the role of GNRI in 1,613 patients from the OCEAN-TAVI registry [196]. They found that low GNRI had a higher occurrence of AKI, major vascular complications, major bleeding, surgical conversion and prolonged length of hospital and ICU stay. Furthermore, correlation existed between the GNRI value and gait speed (ρ = −0.15, P < 0.001), grip strength (ρ  =  0.25, P < 0.001), MMSE (ρ  =  0085, P = 0.004) and the CFS (ρ = −0.24, P < 0.001). In addition, significant correlation was found between the GNRI and STS scores (ρ = −0.29, P < 0.001). In the study of Tzeng et al., hypoalbuminemia was not an independent predictor of length of hospital stay, although the incidence of hypoalbuminemia was significantly higher (68.3% vs 41.0%, P < 0.001) in patients who spent more than 14 days or less than 14 days, respectively [225]. Frei et al. also reported a univariate association between albumin and length of hospital stay (r -0.19; P = 0.004) [220]. Goudzwaard et al. found that the ‘Malnutrition Universal Screening Tool’ (‘MUST’) with a score of 2 or more was independently associated with the occurrence of postoperative delirium (OR 2.9; CI 1.06–7.81, P = 0.04) [146]. Malnutrition was predictive for prolonged in-hospital length of stay in a study by Bobet and colleagues [229]. BMI is evaluated differently in clinical trials: (i) several investigators did not consider it as a frailty marker; (ii) in those studies using BMI as a frailty marker, BMI lower than 20 was associated with worse outcome, for example higher vascular complications. Shimura et al. found a strong correlation between CFS grade and BMI (Spearman’s ρ = –0.077, P = 0.007) [197], although the authors did not define BMI as a frailty variable. Tokarek et al. reported lower rates of bleeding complications and transfusion rates in obese and overweight patients compared to the normal weight group [228]. In contrast, Yamamoto et al. did not find a correlation between early safety outcomes and BMI < 20 kg/m [212]. Similar results were found by Arsalan et al., who evaluated the data of 917 patients [227]: Neither BMI nor body surface area was associated with post-procedural complications.

Conclusions for nutritional state as a prediction tool of neurological complications/delirium and prolonged ventilation/hospitalization after TAVI:

• The nutritional state has a strong correlation with postinterventional complications and postponed convalescence.

• The nutritional state can be assessed objectively by measuring the serum albumin level.

• Results are inconsistent for BMI as a prognostic marker,

3.2.2 Gait speed

Gait speed and postprocedural outcomes have been assessed in 8 papers as measures of frailty, either alone or in a composite score. The ability to predict postoperative neurological complications and length of stay were investigated [118, 180, 196, 220, 228, 230, 232, 238]. Alfredsson et al. evaluated 5-m gait speed in the Society of Thoracic Surgeons/American College of Cardiology Transcatheter Valve Therapy Registry [118]. The data of 8,039 patients were divided into 3 groups: (i) slowest; (ii) slow; (iii) normal walkers. They found significant differences in the length of hospital stay among the groups (P < 0.001). Frei et al. investigated the relationship between length of hospital stay and gait speed in a multiple regression model [220]. They found an independent association between length of hospital stay and gait speed (beta: 0.39 ± 0.16, P = 0.018) in a multiple linear regression analysis. Patel et al. used the 5-mWT in combination with the measurement of albumin levels to assess frailty [180]. They found an adjusted mean difference in the in-hospital length of stay of 4.3 (CI 2.5–6.2, P < 0.001) days. Assmann et al. found that gait speed was protectively associated with the occurrence of delirium (m/s; OR 0.05; CI 0.01–0.50; P = 0.01) [122]. Koh and colleagues also found 5-meter gait speed to be an independent predictor of index hospitalization duration (P < 0.05) [238]. In the remaining studies, a correlation between gait speed and neurological complications/delirium and/or length of stay could not be shown.

Conclusions for 5-m gait speed as a prediction tool for neurological complications/delirium and prolonged ventilation/hospitalization after TAVI:

• Slow gait speed predicts neurological complications/delirium and prolonged length of hospital stay in most studies.

3.2.3 Mini-Mental State Examination

The MMSE has been used as a frailty tool in 6 papers [122, 146, 192, 196, 220, 231]. Assman investigated the incidence of delirium and different frailty measurements [122]. In the multivariate regression analysis, MMSE scores were independently associated with delirium (aOR 0.73; CI 0.53–0.99, P = 0.04). Lauck reported lower MMSE points among those who were discharged later, compared to the early discharge group (MMSE < 24: late: 10.4% vs early 4.0%; P = 0.04) [231]. Shibata found a correlation between GNRI scores and MMSE points (ρ  = 0.0085, P = 0.004) [196]. In a study by Goudzwaard et al., MMSE scores lower than 27 points were associated with postprocedural delirium (OR 2.7; CI 1.27–5.58, P = 0.009), adjusted for age, sex, stroke and dyslipidaemia [146].

Conclusions for MMSE as a prediction tool for neurological complications/delirium and prolonged ventilation/hospitalization after TAVI:

• Most studies showed that low MMSE values predict neurological complications/delirium and prolonged length of hospital stay.

3.2.4 Psoas muscle measures and sarcopenia

Multislice CT imaging of the thoracic and abdominal regions is routinely performed for device selection for TAVI planning. Therefore, measurement of the psoas muscle can be easily carried out for the quantitative measurement of sarcopenia and is not associated with an additional burden for the patient. In a prospective cohort study by Dayha et al., the skeletal muscle index was independently associated with length of hospital stay (beta = −0.07, SE = 0.03) [230] after adjustment for other risk factors. Tzeng et al. found that psoas muscle density was independently associated with hospital stay longer than 14 days (aOR 0.934; CI 0.878–0.994, P = 0.030), and neither the psoas muscle area nor the psoas muscle index remained in the final model [225]. Garg et al. reported that sarcopenia (defined as the PAI) was less than the median value and independently predicted the early poor outcome (aHR 3.18; CI 1.29–7.83, P = 0.012) and the high resource utilization (aHR 2.648; CI 1.322–5.307, P = 0.006) [142]. Mamane found that the PMA in males was related with increased risk for bleeding (aOR 0.78; CI 0.62–0.97) [171]. Nemec measured the skeletal muscle index at different levels. Length of hospital stay was associated with the muscle index at the third lumbar vertebra and at the twelfth thoracic vertebra [235]. Similarly, Damluji found a clear correlation between higher skeletal muscle mass and reduced length of in-hospital stay (10-unit increase in skeletal muscle index: reduction in LOS (β = −0.42, CI −0.75 to −0.09, P = 0.012) [239].

Conclusions for PMM as a prediction tool for neurological complications/delirium and prolonged ventilation/hospitalization after TAVI:

• PMM is a predictor of prolonged hospital length of stay.

• PMM is easily accessible from routine preprocedural CT scans.

3.2.5 Composite frailty indexes

Chaucan reported that frailty (using a 0/4 to 4/4 categorization) was associated with prolonged length of hospital stay [128]. In the multivariate regression, frailty status 4/4 had a coefficient of 3.51 (0.72–6.31), P = 0.014 for long hospitalization. Goudzwaard found that frailty (defined by the Erasmus Frailty Set) was associated with the occurrence of delirium (OR 3.3; CI 1.55 –7.10, P = 0.002) and (OR 2.50; CI 1.30–4.82; P = 0.006) in 213 and 543 patients, respectively [146, 222]. Khan et al. also reported longer hospital stays and higher occurrence of delirium among frail patients [155]. The SMARTI tool set assessed depression. In the study of van Mourik et al., the Multidimensional Prognostic Index was independently associated with death and non-fatal stroke after 30 days (aOR 10; CI 1.48–68.75; P = 0.018) [236]. Ungar et al. also reported that Multidimensional Prognostic Index had a strong predictive value for the estimation of the 3-month mortality and non-fatal stroke (aOR 4.75; CI 1.40–16.08; P = 0.0123) [205]. Tzeng reported that EFT higher than 4 was associated with hospital stay longer than 14 days (OR 0.285 (0.089–0.912); P = 0.034) [225].

Conclusion for composite frailty indexes as a prediction tool for neurological complications/delirium and prolonged ventilation/hospitalization after TAVI:

• The literature about composite frailty indexes is very diverse and does not allow for any specific recommendations. However, all reports show an association between frailty and longer hospital stay and/or neurological complications.

3.3 Frailty as a predictor of quality of life, discharge location, readmission and functional impairment after transcatheter aortic valve implantation

Besides postoperative mortality and morbidity, older patients are highly interested in their postoperative QoL and the course of their convalescence. Even if the procedure is not as demanding for the patient as a surgical procedure, it is essential that patients are discharged to their previous home, are not readmitted to the hospital and do not have any new functional impairment after the procedure.

In the current literature, 15 papers were identified that investigated association or prediction of QoL, readmission, discharge location and impaired function after TAVI using different frailty assessment tools [118, 128, 152, 158, 162, 163, 180, 240–247].

3.3.1 Assessment of frailty to predict quality of life or functional improvement

In the identified publications, only a few studies investigated postinterventional QoL. The overall KCCQ (KCCQ-OS-Score) was used as a disease- specific health status assessment in a cohort of 34,466 patients from the U.S. CoreValve Extreme and High-risk trials and the Society of Thoracic Surgeons/American College of Cardiology Transcatheter Valve Therapy Registry. In the cohort from the U.S. CoreValve trials [240], 1,181 patients had a poor outcome, defined by death, poor QoL or decreased QoL. Frailty was defined by the Fried criteria; disability was defined as more than 1 dependency in ADL. The frailty syndrome was an independent predictor of poor outcome at 1 year (OR 1.42; CI 1.18–1.69, P < 0.001). Single parameters with significant impact were exhaustion and unintentional weight loss (OR 1.35; CI 1.12–1.63, P = 0.002 and OR 1.61; CI 1.21–2.14, P = 0.001 respectively).

One additional dependency in an activity of daily living led to an approximately 20% increased odds of poor outcome (OR 1.19; CI 1.09–1.3, P < 0.01). In the second study, data for 5MGS was available for 6,151 patients from the total cohort [241]. Slow and slowest gait speeds were associated with worse 1-year KCCQ-OS scores (-3.5 points; CI -5–-2.1 and -5.7; CI -7.6–-3.7 respectively). Limitation of mobility (5MGS test) was also predictive for deterioration of QoL (EQ-5D index) after 1 year in the study by Goudzwaard and colleagues (OR 2.29; CI 1.35–6.17, P = 0.006) [248]. Paleri et al. used the Rockwood CFS (see Box 5) to assess frailty in 142 patients and investigated the ability to predict functional improvement, defined as improvement in the New York Heart Association class [242]. They could show that frailty was a predictor of no functional improvement (OR 0.174; CI 0.035–0.869, P = 0.033). The CFS was also assessed in a study by Yoshijima and colleagues [249] and was found to be predictive for poor symptomatic improvement up to 1 year after the procedure, when a CFS of 4 or more was measured. In contrast, Abdelaziz and colleagues [250] did not find the CFS to be associated with QoL scores. Mok identified a decreased distance in the 6MWT as predictor of the futility of TAVI procedures in 319 patients (death or no functional improvement, OR 1.16; IQR 1.02–1.19, P = 0.011) [243]. A walking distance of less than 170 m in the 6MWT was identified as a cut-off point for the prediction of futility [area under the ROC curve (AUC): 0.67, CI 0.55–0.78; P = 0.002]. Kobe et al. used the FORECAST to measure frailty in a TAVI cohort of 130 patients [158]. Based on that, they built 3 groups of patients with different stages of frailty and found that frail patients had significantly lower SF-36 scores for general health, social functioning and role emotional after the procedure and that severely frail patients had worse postoperative SF-36 scores than moderately and not-frail patients. Damluji and colleagues investigated skeletal muscle index as a frailty parameter for prediction of health-related QoL after TAVI. In a cohort of 300 patients, they found that an increase of 10 units of skeletal muscle index leads to an eightfold increase of QoL after the procedure measured by the KCCQ (β1 = 0.81; CI 0.14–1.5, P = 0.017). One self-combined frailty index was used in a study by Bertschi and colleagues to investigate its predictive ability for functional decline 1 year after TAVI [251]. It is based on geriatric assessment and includes mainly ADL, mental disorder and mobility. The combined score was shown to be predictive for functional decline 1 year after TAVI (OR 3.26, 95% CI 1.72–6.16, P < 0.001).

Conclusions for frailty assessment as a prediction tool for QoL and functional improvement after TAVI:

• There is little evidence regarding frailty assessment and the relevance for QoL or functional improvement after TAVI. This can be seen as a gap in evidence.

• Gait speed (5-mWT) and the Fried criteria have been shown to be predictors for QoL after TAVI.

• The Rockwood CFS and 6MWT have been shown to be predictors for functional improvement (improvement in the NYHA functional class).

• QoL assessment should be performed before and after TAVI as a quality control and as a baseline for further studies.

3.3.2 Assessment of frailty to predict discharge location and probability of readmission

3.3.2.1 Gait speed as single parameter or in a composite score predicting discharge location

The largest study (8,039 patients) by Alfredsson et al. investigated the association of 5-meter gait speed with the probability of discharge other than to home [118]. They found that the slowest walkers were significantly more often discharged to rehabilitation, acute care or a nursing home than faster walkers. Unfortunately, the predictive ability of gait speed was not investigated in this study. Chauhan et al. assessed gait speed (15 ft/5m) in 342 patients within a composite test additionally including the Katz Index, serum albumin levels and handgrip strength [128]. They found that the composite score was predictive for discharge to rehabilitation or acute care facilities instead of home (OR 13.10; CI 2.9–59.15, P = 0.001). Horne et al. investigated 285 patients who underwent a TAVI procedure and had a 5-meter gait speed assessment before the intervention [244]. They found a gait speed of more than 7 s for 5 meters to be predictive for discharge to a skilled nursing home facility (RR 2.0; CI 1.4–3.0, P = 0.0002). Similar findings were reported in a study by Huded et al. with 191 patients; these authors found that slowness as part of the Fried criteria assessment to be predictive for discharge to a rehabilitation facility instead of discharge to home (OR = 3.64; CI 1.59–.35, P = 0.002) [125].

Besides prediction of discharge location, Patel et al. found reduced gait speed as 1 parameter in a composite score together with serum albumin (< 3.5 g/dl) to be predictive for readmission to the hospital (OR 2.28; CI 1.16–4.51, P = 0.038) [180]. Additionally, they also found it to be predictive for discharge to acute care after the intervention (OR 3.40; CI 1.87–6.18, P < 0.001). In 2 studies, sarcopenia was defined as loss of skeletal muscle mass [252] or as the sarcopenia index (serum creatinine–cystatin C ratio) [253]. The sarcopenia index was predictive for all-cause mortality or readmission at 1 year. In the study by Brousseau and colleagues [252], only preinterventional gait speed was predictive for rehospitalization 1 year after the intervention (HR 0.32; 95% CI 0.1–0.97).

Conclusions for gait speed as a prediction tool for QoL and functional improvement after TAVI:

• Discharge to a location other than home after TAVI was predicted by gait speed in several studies.

3.3.2.2 Fried criteria and albumin as an isolated marker to predict discharge location

Krishnan et al. performed a retrospective study in 381 patients who underwent a TAVI procedure and had a CT scan to assess psoas and paravertebral muscle sizes [162]. Additionally, they assessed the modified Frailty Index) [254] and serum albumin as frailty markers. Only the latter predicted a decreased risk of being discharged to a rehabilitation center (OR 0.46; CI 0.3–0.8, P < 0.01). No frailty markers were associated with 30-day readmission. Okoh et al. investigated the discharge disposition of 851 patients who were admitted from home for a TAVI procedure [245]. Frailty was assessed with the Fried criteria (see Box 4). Frail patients had a higher likelihood of non-home discharge (OR 1.27; CI 1.03–1.56, P = 0.0263). In the study of Huded et al. [152], frailty measured by the Fried criteria was predictive for discharge to a rehabilitation facility (OR 4.8; CI 1,66–13,85, P = 0.004).

Conclusions for Fried criteria and albumin as prediction tools for discharge location after TAVI:

• Frailty assessment using the Fried criteria was predictive for non-home discharge.

• Normal serum albumin levels were predictive for a decreased risk of being discharged to a rehabilitation centre.

3.3.3 Prediction of probability of readmission by different frailty tools

In 4 publications, readmission or rehospitalization was described as the end point, and the researchers investigated if frailty is predictive for these end points [163, 180, 246, 247]. Patel et al. included 401 patients who had a TAVI procedure and a preinterventional complete frailty assessment. The main end point was the different cost of the procedure between the frail and non-frail patients. A combined assessment including gait speed assessment and serum albumin levels was used as a frailty marker. The mean costs for frail patients were significantly higher, driven by longer in-hospital stays. Every 1-s increase in the 5-meter walk test was associated with decreased costs of almost $2,000, and for every 1 g/dl increase of serum albumin, a more than $9,000 decrease of costs was found. Also, frail patients were more likely to be readmitted within 30 days after the index procedure (OR 2.28; CI 1.16–4.51, P = 0.038). Arai investigated data from 1,215 patients in the OCEAN-TAVI multicentre registry [246]. They found a clinical frailty score > 5 and serum albumin < 3.5 g/dl to be predictive for late readmission in the univariate analysis (OR 1.37; CI 1.03–1.81, P = 0.02 and OR 1.77; CI 1.33–2.36, P = 0.01). Albumin remained predictive in the multivariate analysis (OR 1.37; CI 1.01–1.86, P = 0.04). Kundi et al. retrospectively analysed data from 28,531 patients who were identified from the Centres for Medicare and Medicaid Services Medicare Provider and Review database with a procedural code for TAVI. Frail patients were identified by using the ICD-10 claims-based frailty score [255]. This frailty assessment is based on a cohort of older people (n = 22,139) that was characterized by higher resource use and diagnosis of frailty. ICD-10 codes were collected that characterized this patient cohort, and the Hospital Frailty Risk Score was developed based on that. Rehospitalization was predicted by the Hospital Frailty Risk Score (OR 1.061; CI 1.058–1.064, P < 0.001) and also by the risk categories “intermediate risk” (Hospital Frailty Risk Score 5–15) (OR 1.931; CI 1.762–2.117, P < 0.001) and “high risk” (Hospital Frailty Risk Score > 15, 3.644; CI 3.214–4.132, P < 0.001). Saji et al. performed a study including 155 patients undergoing a TAVI procedure with readmission as the end point [247]. Frailty was assessed by SPPB (see Box 6), the PARTNER frailty scale, the Frailty Index (Japan), the CFS (see Box 5) and modified Fried criteria (see Box 4). All frailty assessments except the modified Fried criteria were identified as independent predictors of unplanned readmission.

Conclusions for prediction of probability of readmission after TAVI:

• There is only limited evidence about prediction of probability of readmission after TAVI by frailty tests.

• Gait speed and albumin levels were shown to be predictive in 2 of the 4 studies.

3.4 Summary of consensus statements on frailty assessment in transcatheter aortic valve implantation

3.4.1 Consensus statements: prediction of mortality after transcatheter aortic valve implantation

3.4.2 Consensus statements: prediction of neurological complications/delirium and prolonged hospitalization/ventilation after TAVI

3.4.3 Consensus statements: prediction of quality of life, discharge location and probability of readmission after transcatheter aortic valve implantation

3.5 Expert task force condensed consensus statements on frailty assessment to predict outcome of transcatheter aortic valve implantation procedures by frailty assessment

3.5.1 The European Frailty Score for Cardiovascular Interventions (EuroFORECAST)

Based on the available evidence, it is advised to follow the frailty assessment pathway illustrated in Figure 3. Work-up should be started with the easy-to-use tools 5-Meter Gait Speed Test and Serum Albumin to assess if the probability of frailty is high or low. If it is high, more comprehensive frailty assessment is indicated. The Fried frailty too has been widely validated also for outcomes other than mortality and for patients in other fields of medicine. The advised assessment pathway includes those tools that are validated for outcome prediction in cardiac surgical patients and patients having TAVI. Therefore, it is ideal to be used in the heart-team decision pathway and is advised especially for this application. We suggest grading a patient as not frail if none of the initial assessments is positive, as frail if 2 initial plus 1 additional or 1 initial plus 2 additional assessments are positive and as severely frail if 2 initial plus 2 additional assessments are positive.

4 MANAGEMENT OF FRAIL PATIENTS AND INTEGRATION OF FRAILTY ASSESSMENT IN CLINICAL ROUTINE

It has been widely demonstrated that frailty is a clinical condition that has a relevant impact on the outcome of patients undergoing cardiac surgery or TAVI. Not only mortality but also functional outcomes, QoL and resource utilization are affected by the presence of frailty. With these consensus statements, our goal was to streamline frailty assessment in clinical routine by condensing the available literature to suggest an easily performable toolset that has, based on the evidence in the literature, the potential to predict all aforementioned end points. This document should help to standardize frailty assessment and to create a base for further creation of evidence, which is highly warranted.

4.1 Use of serum markers for frailty assessment

Besides clinical tests, few studies investigated specific laboratory values as parameters to detect frailty [256, 257]. In fact, only albumin was investigated in several papers, because it may serve as a serum marker for the patient's nutritional state. However, as discussed in the respective sections, it is only an indirect marker for frailty.

The evidence about specific frailty biomarkers is extremely sparse, but it might be a field of interest for further investigations, because a biomarker to detect a clinical condition is always an elegant, reproducible and easy step in the diagnostic pathway.

4.2 Use of administrative data for frailty assessment

Systematic integration of frailty assessment in clinical routine can help to optimize therapeutic decisions and resource utilization in health-care systems. Defining frailty by clustering frailty-defining diagnoses like malnutrition, dementia, decubitus, urinary incontinence, weight loss and so forth has been proposed. Based on the presence of diagnoses from these clusters, frailty can be diagnosed [258, 259]. This approach has also been used for patients undergoing cardiac interventions [83, 84, 164, 260, 261].

Tram and colleagues showed in a population-based, retrospective cohort study in Canada in more than 40,000 patients that frailty was a predictor for long-term mortality (aHR 1.20; CI 1.12–1.28) [84]. Kundi and colleagues showed an improved mortality prediction above and beyond traditional risk factors (integrated discrimination improvement: 0.019; P < 0.001) [164].

The advantage of such scoring systems is that they include factors that are documented in the system anyway, so no further assessment is necessary. Unfortunately, extensive validation of such models is missing.

4.3 Prehabilitation

As the relevance of frailty for outcomes of cardiac interventions has been proven, development of methods to treat or decrease the level of frailty is an important next step.

One approach is a so-called prehabilitation program before interventions. Several domains and systems can be addressed:

• Muscle strength with physical exercise training

• Respiratory system with training of the respiratory muscles

• Body composition by nutritional interventions or diet

• Anxiety and depression with patient education and mental health support

Physical training before interventions has been suggested [262, 263] and investigated in a few studies. Sawatzki and colleagues for example showed that patients in the prehabilitation group increased their gait speed in contrast to patients who did not undergo prehabilitation. Additionally, enrollment in cardiac rehabilitation programs was higher in the group of patients who underwent the prehabilitation program. Steinmetz and colleagues showed that patients scheduled for CABG who underwent an exercised-based prehabilitation program experienced a better improvement in functional capacity and in QoL [264]. Nonetheless, the Enhanced Recovery After Surgery group gave positive recommendations for general surgery [265] but did not clearly recommend it for cardiac surgery due to the small number of patients enrolled in relevant studies. Although not performed explicitly in the frail, Herdy and colleagues revealed that in-hospital prehabilitation in patients awaiting CABG was superior to standard care and led to reduced rates of postoperative complications and shorter hospital stays [266].

Aquino and colleagues described a positive effect of respiratory muscle strength training before CABG compared to regular preoperative care [267]. The interval before the procedure and the duration of the interventions varied among the studies, starting from a few days up to several weeks. In most of the studies, an interval from 4 to 8 weeks was considered useful. In fact a minimum of 4 weeks seems necessary to induce clinically meaningful changes, especially if physical training is the main intervention.

Nutritional interventions are applied either to reduce (diet) or to gain weight in patients with malnutrition. The latter is assumed to be extremely important in the perioperative care not only of cardiac surgical patients but in patients anticipating all different types of operations. Unfortunately, relevant studies are missing. Nevertheless, there is a recommendation from an international multidisciplinary expert group on nutrition in cardiac surgery for optimization of the nutritional state prior to cardiac surgery [268]. They suggest a high caloric supplement starting from at least 2 to 7 days before the scheduled surgery.

Although exercise training, nutritional interventions and other interventions play a pivotal role in prehabilitation, Vigorito and colleagues rightly point out that frailty assessment should always include cognitive and social aspects as well [269]. Depression was found in almost one-third of the patients who underwent cardiac surgery [270]. Dao and colleagues investigated the effects of preoperative cognitive and behavioural therapy on postoperative recovery after coronary artery bypass graft surgery in a randomized clinical trial [271]. Patients who underwent the respective therapies had a shorter length of hospital stay, fewer symptoms of depression and anxiety and better QoL 4 weeks after discharge. All these preprocedural interventions had a positive effect based on the available literature, although the described effects are sometimes small. Thus, further studies are needed to help identify the most effective interventions or combinations thereof.

Therefore, we strongly recommend putting maximum effort into investigating, optimizing and further developing such therapeutic approaches to optimize the outcomes of the increasing number of frail patients undergoing cardiovascular interventions.

5 PRACTICAL SUMMARY AND KEY MESSAGES

5.1 General considerations

Frailty assessment has become an important tool in the daily clinical routine to estimate the individual perioperative or peri-interventional risk. A large body of evidence is available. This task force thoroughly searched and condensed the literature in order to draft this consensus statement on how to use frailty assessment tools for the prediction of important outcomes like hospital morbidity, length of stay, readmission, mortality, neurological sequelae and postoperative living situation as well as QoL.

5.2 Frailty assessment

5.2.1 Frailty assessment in patients scheduled for cardiac surgery

Five tools to assess frailty have been identified that were broadly used and showed good performance in predicting the 3 outcomes that represent hospital morbidity, length of stay and resource utilization, i.e. mortality, neurological sequelae, length of hospital stay and end points relevant to the postoperative course, i.e. postoperative QoL/discharge situation/re-admission (see Fig. 2):

• 5-Meter gait speed with a cut-off for frailty of <0.8 m/s

• Psoas muscle area index with a cut-off for frailty of <8.5 cm²/m² for men and <6.5 cm²/m² for women

• Fried criteria with a cut-off for frailty of ≥ 3 criteria

• Short Physical Performance Battery with a cut-off for frailty with a score < 10

• The Clinical Frailty Scale with a cut-off for frailty with a score of ≥ 4

5.2.2 Frailty assessment in patients scheduled for cardiovascular interventions (surgery and transcatheter aortic valve implantation)

Five tools to assess frailty have been identified that were broadly used and showed a good performance in predicting the 3 outcomes that represent hospital morbidity, length of stay and resource utilization, i.e. mortality, neurological sequelae, length of hospital stay and end points relevant to the postoperative course, i.e. postoperative QoL/discharge situation/re-admission (see Fig. 3):

• 5-meter gait speed with a cut-off for frailty of <0.8 m/s

• Serum albumin with a cut-off for frailty of ≤ 3.5 g/l

• Psoas muscle area index with a cut-off for frailty of <8.5 cm²/m² for men and <6.5 cm²/m² for women

• Fried criteria with a cut-off for frailty of ≥ 3 criteria

• The Essential Frailty Toolset with a score of ≥ 3

5.3 Prehabilitation

1. Prehabilitation (physical, nutritional and mental conditioning before the intervention) seems to be beneficial for the postoperative outcome. Evidence in the field of structural heart interventions remains sparse.

2. Studies in the field of prehabilitation are strongly recommended.

6 GAPS IN EVIDENCE

1. Frailty assessment to predict QoL after cardiac surgery

2. Frailty assessment to predict QoL and functional improvement after TAVI

3. Sarcopenia as a frailty parameter

4. Serum albumin level as biomarker for frailty

5. New biomarkers for frailty

6. Correlation of mental frailty (i.e. assessed by Mini-Mental State Examinations/Montreal Cognitive Assessment Test and resource-relevant end points (i.e. length of stay)

7. Benefit of prehabilitation programs before cardiac surgery or TAVI

8. The role of frailty assessment from the patient’s perspective

7 SEARCH STRATEGY

Search Terms

(frailty [Title/Abstract]) OR frail [Title/Abstract]) OR frailty assessment [Title/Abstract]) OR walk test [Title/Abstract]) OR walking test [Title/Abstract]) OR gait speed [Title/Abstract]) OR frailty measure* [Title/Abstract]) OR cognitive de* assessment [Title/Abstract]) OR physical de* assessment [Title/Abstract]) OR mental de* assessment [Title/Abstract]) OR geriatric assessment [Title/Abstract]) OR sarcopenia [Title/Abstract]) OR psoas muscle [Title/Abstract]) OR grip strength* [Title/Abstract]) AND (cardiac surgery [Title/Abstract] OR cardiothoracic surgery [Title/Abstract] OR cardiac procedures [Title/Abstract] OR cardiac procedure [Title/Abstract] OR cardiac operations [Title/Abstract] OR cardiac operation [Title/Abstract] OR cardiovascular care [Title/Abstract] OR thoracic surgery [Title/Abstract] OR TAVI [Title/Abstract] OR Transcatheter Aortic Valve Impl* [Title/Abstract] OR Surgical Aortic Valve Replacement [Title/Abstract] OR valve re* [Title/Abstract] OR SAVR [Title/Abstract] OR mitral valve [Title/Abstract] OR mitral valve surgery [Title/Abstract] OR surgical mitral valve replacement [Title/Abstract] OR surgical mitral valve repair [Title/Abstract] OR surgical mitral valve reconstruction [Title/Abstract] OR coronary artery bypass [Title/Abstract] OR coronary artery bypasses [Title/Abstract] OR aortic surgery [Title/Abstract] OR arch replacement [Title/Abstract] OR coronary artery bypass surgery [Title/Abstract]) AND (survival [Title/Abstract] OR mortality [Title/Abstract] OR death [Title/Abstract] OR risk [Title/Abstract] OR risk score [Title/Abstract] OR prediction model [Title/Abstract] OR predictive model [Title/abstract] OR predictive power [Title/Abstract] OR stroke [Title/Abstract] OR dementia [Title/Abstract] OR length of stay [Title/Abstract] OR hospital length of stay [Title/Abstract] OR intensive care unit* [Title/Abstract] OR quality of life [Title/Abstract] OR morbidity [Title/Abstract] OR hospitalization [Title/Abstract] OR rehospitalization [Title/Abstract] OR rehospitalization [Title/Abstract] OR complication* [Title/Abstract])

Funding

This work was supported by the European Association for Cardio-Thoracic Surgery and the European Association of Preventive Cardiology.

Conflict of interest: None of the authors declares any conflict of interest.

8 REFERENCES

Author notes