Comprehensive geriatric assessment in patients undergoing transcatheter aortic valve implantation – results from the CGA-TAVI multicentre registry

Background In older patients with aortic stenosis (AS) undergoing TAVI, the potential role of prior CGA is not well established. To explore the value of comprehensive geriatric assessment (CGA) for predicting mortality and/or hospitalisation within the first 3 months after transcatheter aortic valve implantation (TAVI). Methods An international, multi-centre, prospective registry (CGA-TAVI) was established to gather data on CGA results and medium-term outcomes in geriatric patients undergoing TAVI. Logistic regression was used to evaluate the predictive value of a multidimensional prognostic index (MPI); a short physical performance battery (SPPB); and the Silver Code, which was based on administrative data, for predicting death and/or hospitalisation in the first 3 months after TAVI (primary endpoint). Results A total of 71 TAVI patients (mean age 85.4 years; mean log EuroSCORE I 22.5%) were enrolled. Device success according to VARC criteria was 100%. After adjustment for selected baseline characteristics, a higher (poorer) MPI score (OR: 3.34; 95% CI: 1.39–8.02; p = 0.0068) and a lower (poorer) SPPB score (OR: 1.15; 95% CI: 1.01–1.54; p = 0.0380) were found to be associated with an increased likelihood of the primary endpoint. The Silver Code did not show any predictive ability in this population. Conclusions Several aspects of the CGA have shown promise for being of use to physicians when predicting TAVI outcomes. While the MPI may be useful in clinical practice, the SPPB may be of particular value, being simple and quick to perform. Validation of these findings in a larger sample is warranted. Trial registration The trial was registered in ClinicalTrials.gov on November 7, 2013 (NCT01991444).


Background
Severe symptomatic stenosis of the aortic valve (aortic stenosis; AS) is associated with mortality of up to 50% at 1 year if left untreated [1,2]. The outcome of surgical aortic valve replacement (SAVR) is generally predicted with the aid of quantitative scales such as the EuroSCORE or the Society of Thoracic Surgeons (STS) risk score. However, the accuracy of such algorithms for assessing older-age, multi-morbid patients undergoing transcatheter aortic valve implantation (TAVI) is low [3][4][5]. This is mainly due to the absence of variables related to ageing, such as frailty, mental status, social support, and overall health. There is evidence that additionally evaluating these factors can help provide a more precise estimation of an older person's response to treatment [3,6,7]. Indeed, a recent report from the American College of Cardiology (ACC) advocates assessment of frailty and cognitive function prior to determining a patient's suitability for TAVI [8]. The inclusion of a geriatrician in the Heart Team responsible for assessing patients prior to TAVI may therefore be warranted [9].
A comprehensive geriatric assessment (CGA) is a multidimensional diagnostic process for evaluating an individual's clinical, psychosocial, and functional characteristics [10,11]. It usually consists of functional tests complementing usual clinical evaluation. For example, CGA may include the calculation of a multidimensional prognostic index (MPI) based on mental and nutritional status, number of co-morbid conditions and medications, living arrangements, and the ability to cope with activities of daily living [12,13]. Pilotto et al. showed that the MPI had high predictive power for assessing mortality after hospitalisation of older patients [13]. For TAVI specifically, few studies have evaluated CGA and its components for outcome prediction [3,[14][15][16]. For example, Stortecky et al. evaluated a geriatric assessment that contained many of the same components as the MPI, and found that many of the included items were predictive of mortality and the occurrence of a major adverse cardiac or cerebrovascular event (MACCE) at 30 days and 1 year after the procedure [14]. All previous studies originated from single-centre experiences.
We aimed to determine the power of CGA for predicting the combined endpoint of mortality and stroke within the 3 months subsequent to TAVI based on data from a multi-centre, prospective cohort. We further characterised changes in CGA over time, and provide additional evidence for the utility of TAVI in a geriatric, comorbid population.

Patients and registry design
The CGA-TAVI registry is a prospective, international, multi-centre, observational registry [17].
Patients were enrolled at three centres in Italy (Careggi Hospital, Florence), the Netherlands (Academic Medical Center, Amsterdam) and Canada (Centre Hospitalier de l'Université de Montréal) between August 2013 and December 2015. Individuals were eligible if they were aged >80 years, had symptomatic severe calcific aortic valve (AV) stenosis, and were assigned to undergo transaortic, transapical or transfemoral TAVI. Patients were excluded if TAVI was being performed as an emergency procedure or if patients were unable to participate in the followup. Overall, 603 patients underwent TAVI at one of the three study centres during the study period (Italy: 68, Canada: 95; the Netherlands: 440. Of these, 71 patients were enrolled in the CGA-TAVI registry: 41, 15 and 15 from the sites in Italy, Canada and the Netherlands, respectively.

Baseline assessment
A detailed description of the information documented has been previously published [17]. Briefly, data regarding demographics, comorbidities, and prior cardiovascular interventions were collected at hospital admission. A CGA was performed by a geriatrician for each patient. This included calculation of the MPI [12,13], which consisted of the following components: Activities of Daily Living (ADL; 6 items) [18,19]; Instrumental Activities of Daily Living (IADL; 8 items) [20]; Short Portable Mental Status Questionnaire (SPMSQ; 10 items) [21]; Cumulative Illness Rating Scale (CIRS; 14 items) [22,23]; Mini Nutritional Assessment (MNA; 18 items) [24]; Exton-Smith Scale (ESS; 5 items) [25]; number of drugs used (1 item); and cohabitation status (1 item). In each case, a tripartite hierarchy was used for scoring (no problems: 0 points; minor problems: 0.5 points; severe problems: 1 point). The boundaries for these scores were based on the cut-off points derived from the associated literature [12]. A Silver Code value was also calculated from administrative data for further prognostic stratification [26] and a Short Physical Performance Battery (SPPB) was performed, which involved repeated chair stands, balance testing, and an 8-ft (2.44 m) walk [27]. Detailed breakdowns of the three assessment scores can be found in Appendices 1, 2 (MPI), 3 (Silver Code) and 4 (SPPB). Procedural characteristics were also documented.

Follow-up assessment and outcomes
Patients were scheduled for follow-up at discharge, 30 days and 3 months post-procedure. These visits were conducted at the patient's enrolling centre and involved repetition of the CGA performed at the baseline assessment. Death, all-cause hospitalisation, TAVI-related hospitalisation, stroke, transient ischaemic attack (TIA), myocardial infarction (MI), lifethreatening bleeding, acute kidney injury, coronary artery obstruction requiring intervention, major vascular complications, valve dysfunction requiring repeat procedure, or worsening congestive heart failure (CHF) were recorded, as defined in the Valve Academic Research Consortium-2 (VARC-2) consensus document [28].

Endpoints
The primary endpoint was death and/or hospitalisation within the first 3 months after TAVI. The secondary endpoint was death and/or non-fatal stroke within the same period. Changes in the scores of the components of the CGA from baseline to 3 months were also evaluated.

Data management and statistics
Data were entered into an online database via the completion of an electronic case report form (eCRF; s4trials, Berlin, Germany). Details were automatically checked for plausibility and completeness.
Data were analysed using descriptive statistics, with categorical variables presented as absolute values and percentages and continuous variables as means with standard deviations (SD). A logistic regression was used to evaluate the predictive value of CGA components (MPI, SPBB and Silver Code) for the primary/ secondary endpoints. Age, gender, New York Heart Association (NYHA) class and surgical risk (Euro-SCORE/STS) were used as co-variables. Logistic regression results are presented as odds ratios (OR) with 95% Wald confidence limits (95% CI) and pvalues. CGA changes from baseline to 3-month follow-up were tested for significance using a t-test. P-values of <0.05 were considered significant.
Data were analysed using IBM SPSS statistics version 24 (IBM corporation, Amonk, New York, USA).

Procedural characteristics and periprocedural outcomes
Details about procedural aspects are provided in the supplementary data (Appendix 5). Devices were placed successfully (as defined by VARC-2 [28]) in all patients (Appendix 6). The rate of intraoperative complications was 17.1%, which were vascular complications requiring treatment in 10.0% of patients and access-related in 5.6%. No conversion to open surgery was necessary. Paravalvular regurgitation was moderate in 2.9%, with no severe regurgitation. Postprocedural AV peak and mean gradients were 20.0 ± 12.4 and 11.3 ± 7.2 mmHg, respectively.

Discussion
Of the multiple components of the CGA that were evaluated, the MPI and the SPPB both had value for predicting the likelihood of death and/or hospitalisation in the first 3 months following TAVI. In terms of time-efficiency, the SPPB appears to be the favourable approach, with the MPI perhaps not adding sufficient additional predictive value to warrant such a time-consuming assessment.

Outcomes
In our patients, the rate of all-cause mortality at 30 days (2.8%) was within the range (1.1%-5.9%) reported by large-scale studies in patients with a similar level of surgical risk and mean ages above 80 years, such as the PARTNER II trial and SOURCE 3, WIN-TAVI, Swiss TAVI, and PRAG-MATIC registries [29][30][31][32][33]. Though a less commonly reported outcome, the same was true of the rate of rehospitalisation (7.2% vs. 4.6% in the PARTNER II SAPIEN 3 trial and 6.5% in the overall PARTNER II trial) [29,34]. Variations in rates between studies are likely due to differing patient characteristics, access routes, and the valves/delivery devices available during study periods. Our findings regarding threemonth outcomes could not be easily compared to previous studies, as this time-point is not largely reported upon in the literature.

Surgical risk scores
EuroSCORE and STS algorithms are conventional tools for assessing cardiac operative risk. It has been suggested that the latter is slightly more accurate in TAVI patients [4,5], though neither are ideal. Patients in the present study had expected mortality rates of 22.5% (EuroSCORE I) and 5.8% (STS), and though the latter was closer to the observed rate, both were excessively elevated. Even the more up-todate EuroSCORE II has been shown to have suboptimal discriminatory power, suggesting the need for alternative or additional assessment tools [35,36].

Multidimensional prognostic index
Use of a CGA during clinical assessment of operative risk in AS patients has been suggested as a way to address the shortcomings of the EuroSCORE and STS score, and better predict outcomes [17,36]. In the present study, both MPI and SPPB were found to have predictive value for determining the likelihood of short-term mortality/hospitalisation or stroke after TAVI. Interestingly, while high MPI was a "negative predictor" in the univariate analysis, it became a "positive predictor" after adjusting for age, gender, NYHA class and surgical risk. This change of direction can be explained by the clinical setting; younger patients with a low MPI are typically treated with SAVR, while elderly patients with a high MPI are unlikely to be treated at all. Consequently, our TAVI population was likely composed of patients with a lower age and high MPI or a higher age and low MPI, resulting in a switch of the direction of the odds ratio at multivariate analysis. This reflection of the clinical context is supportive of the "real" association between MPI score and outcomes. Though no other studies appear to have specifically reported on the predictive value of MPI in TAVI, several have shown higher MPI scores to be significantly associated with greater rates of mortality in older patients with a variety of acute illnesses, including heart failure and TIA [12,13,[37][38][39][40]. Other studies have evaluated other multi-component models for predicting mortality and morbidity after cardiac surgery [41]. For TAVI specifically, Green et al. found that patients with a high frailty score, as determined by gait speed, grip strength, serum albumin and ADL, were at greater risk of one-year mortality [15]. The five-component frailty score proposed by Kamga et al. was found to predict one-year mortality after transfemoral TAVI [16]. Stortecky et al. identified numerous parameters in their Multidimensional Geriatric Assessment that were predictive of 30-day and one-year mortality and MACCE after TAVI [14]. Data from the PARTNER trial were used to construct models for predicting a poor outcome, defined as death or a low/significantly decreased quality of life, after TAVI [42]. These models were subsequently validated in a large multi-centre cohort of TAVI patients, with an incremental increase in discriminative ability identified on the addition of markers of frailty and disability [43]. In agreement with the data from our CGA registry, these studies demonstrate the potential value of such multicomponent analyses for predicting outcome after TAVI.

Short physical performance battery
A significant drawback of these multidimensional evaluation tools, however, is that they are extremely time-consuming. In the present work, we found that use of the SPPB alone was equally as effective as the MPI for predicting death and/or hospitalisation, and death and/or non-fatal stroke, in the first 3 months after TAVI. This short series of tests is  [17], although it appears that there is little published evidence in support of using it for assessing TAVI candidates specifically. The concept of tests of physical ability to predict outcome after cardiac surgery has been evaluated in other studies. Afilalo et al. demonstrated that a slow 5-m gait speed was associated with a greater risk of operative mortality and in-hospital mortality and major morbidity in older patients undergoing cardiac surgery [7,44]. They further determined that use of this parameter alone was superior to a variety of other frailty scales [6]. In patients undergoing TAVI specifically, 5-m gait speed has been shown to be independently associated with 30-day mortality after adjustment for STS score and other relevant baseline characteristics [45]. Stortecky et al. reported that the "timed get-up and go" (TUG) test had the greatest predictive ability of all of the individual geriatric assessment tools that they investigated [14]. In combination with either the STS or EuroSCORE, the TUG was superior to the other components evaluated for predicting all-cause mortality and MACCE during the first year after TAVI. A recent report by the ACC recommends that a 5-m gait speed test and a 6 min walk test be used to assess frailty and physical functioning, respectively, when determining a patient's suitability for TAVI [8].
The simplicity of physical tests such as the SPPB is not their only advantage, with the lack of subjectivity on the part of both physician and patient providing a level of accuracy that cannot be obtained using questionnaire-based assessment. This is particularly relevant for the advanced-age TAVI population, where cognitive impairment is a potentially significant confounding factor when evaluating self-reported parameters [6].

Silver code
According to our data, the Silver Code had no value for predicting death/hospitalisation or death/stroke during the first 3 months after TAVI. The calculation of this parameter prior to deciding on the suitability of a patient for TAVI is another recommendation of the EUGMS [17]. Previous studies have demonstrated a relationship between Silver Code and one-year mortality, although this was in the setting of the Emergency Department [26,46]. As the Silver Code is determined from administrative data, it is particularly suited to planned procedures such as TAVI. Therefore, although it was not found to be an independent predictor of outcome in the present analysis, it should perhaps not be discounted. Further evaluation in a larger population may clarify its utility as part of a CGA prior to TAVI.

Limitations
Firstly, as an observational study, inherent limitations such as a higher potential for missing data are present. However, the observational aspect carried several advantages, such as an evaluation of TAVI patients in a real-world setting, avoiding the confounding issue of the strict inclusion and exclusion criteria used for clinical trials. This is particularly important, as geriatric patients are often those excluded due to high levels of comorbidity. Secondly, our findings are currently only applicable to geriatric patients at higher surgical risk. Considering that there is a current shift in clinical practice towards TAVI in lower-risk patients who are normally eligible for surgical heart valve replacement, re-evaluation of CGA in different populations may become necessary. In addition, we were only able to obtain data for a modest number of patients from three participating sites, limiting statistical power and generalisability. Indeed, the relatively low incidence of mortality and stroke at 3 months in a fairly small sample may have resulted in suboptimal power to detect baseline characteristics that are predictive for this outcome. Future studies in larger samples would be useful for clarification.

Conclusion
Several aspects of the CGA have shown promise for being of use to physicians when predicting the likelihood of death, rehospitalisation and non-fatal stroke following TAVI. The strong association between MPI and such outcomes indicates its potential utility in clinical practice. In addition, the SPPB may have significant value, being simple and quick to perform; however, the modest sample size included herein limit the formation of firm conclusions, and validation of these findings in a larger sample of TAVI patients with a greater range of surgical risk is warranted.

Clinical perspectives
The accuracy of conventional surgical risk scores is known to be suboptimal for predicting the outcomes of TAVI in elderly aortic stenosis patients. This study shows a potential benefit of adding items from a comprehensive geriatric assessment (CGA) to preintervention assessments. This now requires further validation in larger cohorts.
Provides an overview about the Multidimensional Prognostic Index (MPI).

Appendix 2
Provides a breakdown of the Multidimensional Prognostic Index (MPI).  Receives no assistance (gets in and out of the tub by self if tub is usual means of bathing) 1 Receives no assistance in bathing only one part of the body (such as back or leg) 1 Receives assistance in bathing more than one part of the body (or not bathed) 0 Dressing (gets clothes from closets and drawers, including underclothes/outer garments and using fasteners/ braces, if worn) Gets clothes and gets completely dressed without assistance 1 Gets clothes and gets dressed without assistance except for assistance in tying shoes 1 Receives assistance in getting clothes or in getting dressed, or stays partly or completely undressed 0 Toileting Goes to "toilet room" cleans self, and arranges clothes without assistance (may use object for support such as cane, walker, or wheelchair and may manage night bedpan or commode, emptying same in morning) 1 Receives assistance in going to "toilet room" or in cleaning self or in arranging clothes after elimination or in use of night bedpan or commode 0 Doesn't go to room termed "toilet" for the elimination process 0 Transfer Moves in and out of bed as well as in and out of chair without assistance (may be using object for support such as cane or walker) 1 Moves in and out of bed or chair with assistance 0 Doesn't get out of bed 0 Heats, serves and prepares meals or prepares meals but does not maintain adequate diet 0

Continence
Needs to have meals prepared and served 0

Housekeeping
Maintains house alone or with occasional assistance (e.g. "heavy work domestic help") 1 Performs light daily tasks such as dishwashing, bed-making 1 Performs light daily tasks but cannot maintain acceptable level of cleanliness 1 Needs help with all home maintenance tasks 0 Does not participate in any housekeeping tasks 0 Laundry Does personal laundry completely 1 Launders small items; rinses stockings, etc. 1 All laundry must be done by others 0 Mode of transportation Travels independently on public transportation or drives own car 1 Arranges own travel via taxi, but does not otherwise use public transportation 1 Travels on public transportation when accompanied by another 1 Travel limited to taxi or automobile with assistance of another 0 Does not travel at all 0

Responsibility for own medications
Is responsible for taking medication in correct dosages at correct time 1 Takes responsibility if medication is prepared in advance in separate dosage 0 Is not capable of dispensing own medication 0   Illustrates how the Silver Code is calculated. In comparison with other people of same age, how they consider their health status?
Not as good 0 Does not know 0.5 As good 1 Better 2 Max MNA score (best-nourished) 30 Legend: The numbering on the left-hand side of score titles corresponds to the numbering of factors in Appendix 1 Total score: 0 pts. = best possible performance, 36 pts. = worst possible performance-Corresponds to gradient risk for mortality 1

. Repeated Chair Stands
• Ask patient if they think it is safe for them to try and stand up from a chair five times without using their arms. • If yes, instruct patient to stand up straight and then sit back down again as quickly as they can five times, without stopping in between, keeping their arms folded over their chest. • Instruct patient to stand with the side of the heel of one foot touching the big toe of the other foot for 10 s (left/right feet as preferred by patient). • Demonstrate • Stand next to patient to help them into a semitandem position, allowing them to hold onto your arms to establish balance. • Begin timing when patient has the feet in position.
• If unable to hold the semitandem position for 10 s → side-by-side.
• If able to hold the semitandem position for 10 s → tandem. → Side-by-side: As for semitandem but with feet together. Patients may use their arms, bend their knees, or move their body to maintain balance, but may not move their feet. → Tandem: As for semitandem but with the heel of one foot in-front-of and touching the toes of the other foot.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.

Authors' contributions
The authors take responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation. The specific contributions of the respective authors are as follows: AU, PB, JK, MT, SM and AWS were involved in the conception and design of the study; GM, NvV, MT, JB, MPT, JBM, GS, MvM, and SJ collected the data, which were analyzed by AU, GM, CD, and PB and interpreted by the author group; PB drafted the manuscript and all other authors revised the article for important intellectual content. All authors have given final approval for the version to be submitted.