Clinical effectiveness and cost-effectiveness of second- and third-generation left ventricular assist devices as either bridge to transplant or alternative to transplant for adults eligible for heart transplantation: systematic review and cost-effectiveness model

Aetiology and pathophysiology of heart failure

Any anatomical or physiological conditions which affect ventricular function can cause HF. In a survey conducted in Hillingdon, West London, which included a population of 151,000 people, researchers found that the most common cause of HF was ischaemic or coronary heart disease (CHD). 10 Similarly, in a UK-based population, a study of coronary artery angiography in new patients aged < 75 years,14 CHD was found to be the commonest cause of HF. Other causes of HF include hypertension, valvular heart disease, myocardial toxins, myocarditis and cardiomyopathy. 4,10,14

The final common pathway for all pathophysiology of HF (CHD, poorly controlled hypertension, cardiomyopathy or valvular heart disease) is ventricular dysfunction. The left ventricle (LV) is most commonly affected with eventual myocardial injury and remodelling leading to a dilated ventricular chamber with a low ejection fraction, activation of non-cardiac factors such as the neurohormonal systems with vasoconstriction and renal sodium retention, and further symptoms such as dyspnoea, fatigue and oedema (Figure 1). 15,16 This can lead to episodes of arrhythmia, increasing pump failure and, finally, premature death. However, aetiology of HF varies by age group, as also do the criteria used to identify its presence. 10,17

FIGURE 1.

The pathophysiology of progression of HF (redrawn and adapted from Cohn16).

Symptoms and signs of heart failure

The most common symptoms of HF are breathlessness, tiredness, loss of appetite, and signs of peripheral oedema, raised jugular venous pressure, tachycardia or tachypnoea. 5,6 The severity of HF is usually assessed using the New York Heart Association (NYHA) functional classification, which is based on the severity of symptoms patients develop in relation to physical activity. Severity of HF can be classified into four grades using the NYHA classification. Patients with NYHA class I are considered to be less severely affected and can perform ordinary physical activity without developing symptoms of HF. Patients with NYHA class IV have advanced HF, are unable to carry out any physical activity and have symptoms at rest (Box 2).

Diagnosis of heart failure

There is no ‘gold standard’ for diagnosis of HF. Initially, it is assessed by patient history and physical examination. 8 In addition, there are no signs and symptoms that are both sensitive and specific for the diagnosis of HF. 19 Investigations such as electrocardiography, measurement of B-type natriuretic peptide (BNP) or both are recommended depending on the condition. If the above tests are abnormal then echocardiography (to measure ventricular performance) and chest radiography (to detect cardiomegaly, pulmonary congestion and pleural fluid accumulation) are undertaken to confirm the diagnosis of HF. BNP and N-terminal pro-B-type natriuretic peptide (NT-pro-BNP) are useful biomarker hormones in the diagnosis of HF. Levels of these biomarkers are raised in patients with HF and the concentrations vary with NYHA class. 5,19,20 Figure 2 gives a schematic representation of recommendations on the diagnosis of HF (adapted from Sutherland). 21

FIGURE 2.

A schematic diagram of HF diagnosis recommendations (redrawn and adapted from Sutherland21).

Quality of life and prognosis of heart failure

People with HF are often heavy users of primary care services. 17,22 The mortality rate of HF is comparable to that of cancer. 17,23 de Giuli et al. 24 studied primary care patients in the UK and found that people with HF have a very poor prognosis, especially the elderly. The Hillingdon Heart Failure study also reported that around 40% of people die within 1 year of a diagnosis of HF. 10 In the Echocardiographic Heart of England Screening Study, QoL was measured by Short Form questionnaire-36 items (SF-36), and impairment in both mental and physical QoL was reported. Impairment was found to be worse in those with more severe HF as measured by NYHA severity assessment class25 and reduction in QoL was particularly evident among elderly people. 21

Management of heart failure

Treatment of patients with HF depends on type and stage of HF.

Descriptions of ventricular assist devices

An LVAD has inflow and outflow cannulae which help to regulate blood flow from the LV or left atrium to the ascending aorta. Similarly, in the RVAD, an inflow cannula regulates blood from the right ventricle (RV) or right atrium to the pulmonary artery.

Left ventricular assist devices help to pump blood from the LV of the heart to the rest of the body in patients with advanced HF. 35 In the UK, LVAD patients with advanced HF wait for a donor heart to become available. 1,36 These devices as are not currently licensed for use as DT in the UK, although they are approved in the USA and in parts of Europe. 1,35 LVADs can be broadly categorised as generation I, generation II and generation III (Figure 3).

FIGURE 3.

A schematic diagram of HF diagnosis recommendations.

Second- and third-generation LVADs are magnetic continuous flow (CF) rotary pumps whereas first-generation LVADs are pulsatile volume displacement devices. 35,37 Compared with first-generation devices, second- and third-generation devices are smaller, quieter and more reliable. Second- and third-generation devices are inserted through a small dissection. They are easier to insert and less traumatic than previous types and are associated with less bleeding and infection. Third-generation LVADs are attached with an impeller which uses magnetic forces or hydrodynamic levitation without mechanical contact. They therefore have greater durability, with no mechanical wear and tear compared with second-generation LVADs. 35,37 In this report the interventions of interest are second- and third-generation devices which have either US Food and Drug Administration (FDA) or Conformité Européenne (CE) approval or both; therefore, this section will describe characteristics only of these devices.

Table 1 shows the VADs which have FDA or CE approval.

Complications of ventricular assist devices

This section summarises papers by Potapov et al. 48 and Barnes,49 which describe complications which can occur for patients with VADs. Complications in patients with VADs can be categorised as acute or late.

Current service provision

With increasing demand for HT, the United Network for Organ Sharing (UNOS) has coded waiting list (WL) patients as Status 1A, 1B or 2 on the basis of medical urgency. Patients who are in a clinically stable condition and patients with LVAD-related complications (infection, thromboembolism or device malfunction) are categorised as Status 1A patients for 30 days. Status 1B is assigned to patients supported by LVAD who do not meet the criteria for Status 1A. Status 2 patients are those receiving long-term LVAD support. 51

In the UK, WL management is based on the urgent and non-urgent WL, developed by the Cardiothoracic Advisory Group of NHS Blood and Transplant (Box 3). 28

Evolving LVAD technology from first generation to second and third generation has led to development of devices which are considerably smaller, more durable and associated with fewer adverse events. As pulsatile and first-generation LVADs have been modified to CF devices, the improvements have been marked with a lower incidence of infection and complications. Use of these devices as a BTT has led to considerable improvements in QoL among patients with advanced HF. 52,53

Patient pathways for management of heart failure

Treatment of patients with HF depends on the type and stage of HF. The following examples of patient pathways indicate how patients are treated at different stages: BTT to VAD (Box 4), BTT to HT (Box 5) and MM (Box 6). The National Protocol for Assessment of Cardiothoracic patients lists below the medical indications for patients eligible for a HT.

• End-stage heart disease with a life expectancy of between 12 and 18 months.

• NYHA class III or IV.

• Refractory to medical therapy, including, if necessary, cardiac resynchronisation therapy. (This assessment should be made by a cardiologist with a special interest in HF.)

• Usually < 60 years of age as there is an increase in comorbidity with the ageing process. However, consider biologically fit older patients.

In the next section of this report we describe the decision problem and research questions.

Objectives

1. To summarise previously published HTA reports by Clegg et al. 4 and Sharples et al. 30 on VADs.

2. To undertake a systematic review and evidence synthesis of the relevant clinical effectiveness and cost-effectiveness literature.

3. To further develop the cost-effectiveness and cost–utility models developed in the 2006 HTA: Evaluation of the ventricular assist device programme in the UK and where possible to compare the use of VADs as BTT firstly to MM and secondly as ATT.

4. To investigate the factors that drive cost-effectiveness estimates.

5. To report on findings and make recommendations for future research.

Identification of literature

Databases searched

Databases searched included MEDLINE; MEDLINE In-Process & Other Non-Indexed Citations; EMBASE; Cochrane Database [including Cochrane Database of Systematic Reviews (CDSR), Database of Abstracts of Reviews of Effects (DARE), NHS Economic Evaluation Database (NHS EED), and HTA databases]; Science Citation Index and Conference Proceedings (Web of Science); UK Clinical Research Network (UKCRN) Portfolio Database; Cumulative Index to Nursing and Allied Health Literature (CINAHL); PsycINFO; and the National Library of Medicine (NLM) Gateway (US Meeting Abstracts and Health Services Research Projects in Progress) were searched. The following trial databases were also searched: Cochrane Central Register of Controlled Trials (CENTRAL); Current Controlled Trials; and ClinicalTrials.gov.

In addition, the reference lists of relevant articles were checked, and the manufacturers' websites screened for relevant publications. Also, the online resources of various regulatory bodies, health services research agencies and professional societies were consulted via the Internet. These included:

1. HTA organisations (including the National Institute for Health Research and the National Research Register Archive)

2. INTERMACS

3. NHS Blood and Transplant (including the Cardiothoracic Transplant Advisory Group)

4. Ventricular Assist Device Forum, National Specialised Commissioning Team

5. The International Society Heart & Lung Transplantation

6. Eurotransplant

7. Scandiatransplant

8. US Transplant

9. The Transplantation Society

10. British Transplantation Society

11. Medicines and Healthcare products Regulatory Agency

12. US FDA.

Citation searches of included studies were undertaken using the Web of Science citation search facility. The reference lists of included studies and relevant review articles were also checked.

Inclusion criteria

Exclusion criteria

The following exclusion criteria were applied:

1. studies in which 20% patients were known to be receiving VADs as DT

2. PVAD

3. TAH

4. first-generation pulsatile volume displacement pumps

5. devices yet to be FDA or CE approved

6. devices for ‘bridge to decision’

7. post-transplant mechanical circulatory support devices for primary graft failure

8. studies involving VADs in conjunction with other interventions where it was not possible to separate out the effects of the different interventions on outcomes

9. animal models and post-mortem studies

10. preclinical and biological studies

11. editorials and opinions

12. reports published as meeting abstracts only, where insufficient methodological details were reported to allow critical appraisal of study quality

13. studies not in English

14. studies before the year 2003

15. case series reports with < 50 cases or where patient recruitment was not consecutive.

Data abstraction strategy

A record of all papers rejected at full-text stage and reasons for exclusion was documented. Titles and abstracts of retrieved studies were examined for inclusion by two reviewers independently. Disagreement was resolved by retrieval of the full publication and consensus agreement.

The full data were extracted independently by one reviewer using a data extraction form informed by the NHS Centre for Reviews and Dissemination (CRD)54 and previous HTA reports4,30 (see Appendix 3 for the complete data extraction forms, this includes publications which did not separate outcomes for BTT patients from DT patients and which are not included in the main text of the report). All studies were checked by a second researcher, and any disagreements were resolved by discussion. Further discrepancies were resolved with involvement of a third reviewer. Data were extracted to allow quality assessment of the included studies.

Critical appraisal strategy

Quality criteria were applied independently by two reviewers and an agreed overall quality assessment was determined for each paper. Any disagreements were resolved by independent assessment by a third reviewer. Included studies were assessed using the following recognised quality assessment scales and/or checklists. Systematic reviews were assessed using criteria developed by NHS CRD. 54 Experimental and non-experimental studies were assessed using an adapted set of criteria developed by Thomas et al. 55 Each study was scored according to (a) selection of participants; (b) study design; (c) confounders; (d) blinding; (e) data collection methods; (f) withdrawal and dropout; and (g) integrity and analysis (see Appendix 4 for further details on quality assessment).

Methods of data synthesis

Data were tabulated and discussed in a narrative review based on indication for treatment, type of VAD, quantity and quality of research evidence, representativeness and outcomes. The remit of this report was to consider BTT but not DT. Some publications presented aggregate results for both groups; such aggregate results are not relevant to BTT, but for completeness we report such results in Appendix 3. Where data specific to BTT patients could be extracted from any publication these are also included in the main text of the report. Outcome results are given for BTT patients with published data selected so as to avoid double counting from overlapping populations.

We analysed patient populations in each included study for overlap between studies, and developed a ‘family tree’ to ascertain which data set included the most recent data on the largest number of unique patient records (as earlier, smaller studies fed into larger, later studies). For each device, we used the largest/latest data set of separately identifiable patients to report baseline characteristics and adverse events.

Baseline characteristics were listed as means for continuous variables and percentages for binary variables. Ninety-five per cent CIs were calculated. Where possible, the reported data for subgroups were combined to obtain a value for the whole study population. Pooling of study baseline characteristic values was undertaken using a random-effects model in MetaAnalyst Version Beta 3.13 (Tufts Medical Centre, Boston, MA, USA) software. Narrative syntheses were used to describe outcomes.

Types of device used

Of the 29 included studies52,53,56–82 involving HMII, 22 studies53,56–59,61–65,67–71,73–75,77–80 presented data on HMII alone. Five studies60,72,76,81,82 compared HMXVE with HMII. One study66 compared HMI with HMII, and a further study52 compared HMII with other devices (type not reported). One study83 reported on HW only, one study84 involved MicroMed DeBakey VAD only, one study86 involved Berlin Heart INCOR only, and two studies42,85 involved DuraHeart only.

A further six studies87–92 reported a mixture of devices but data by device were not reported separately.

Reasons for use of ventricular assist devices in included studies

Studies reported mixed reasons for use of VADs. For example, of the 29 HMII studies,52,53,56–82 12 studies52,57,58,64,65,67,70,71,73–75,82 reported that treatment was for BTT, 12 studies53,56,59,60,62,72,76–81 reported that treatment was for BTT and DT, one study68 reported that treatment was for BTT or DT or BTR and the remaining four studies61,63,66,69 did not report reason for treatment.

Delineating multiple overlapping populations between publications

Many of the identified publications investigated overlapping populations; this was especially true for studies of HMII, most of which were conducted in the USA. Also, in some studies the patient group received different devices and authors did not report results separately for each of the several devices investigated. There were many studies in which different patients were given bridge or destination therapies; however, in most of these outcomes were not reported according to indication.

The US HMII publications can be classified as (a) from single centres (n = 14); (b) deriving from the multicentre FDA approval study and its extension (n = 12); and (c) multicentre registry studies.

Starling et al. 52 and John et al. 65 reported on HMII, while Nativi et al. 89 indicated the number of HMII recipients but reported data for a mix of different VADs. In 12 of the single-centre studies53,56,59,60,62,72,76–81 both DT and BTT patients were included, or indication was not clearly defined; none of these analysed results separately for BTT patients. The other two single-centre publications (Petrucci et al. 74 and John et al. 64) reported results for BTT patients, but these single centres appear to have contributed participants to the FDA approval group of multicentre publications. The FDA approval study publications reflect the gradual accrual of more patients and multiple publications have been produced for overlapping groups of patients. Five of these publications58,59,65,67,82 report on the same 469–486 participants by either dichotomising the population by various criteria,67,82 focusing on a particular outcome,58 combining BTT patients with HMII DT patients,59 or not separating the outcome data according to therapy received. Registry studies, including John et al. 65 and Starling et al. ,52 reported on post-approval HMII BTT patients who were not participants in the FDA extension study. We consider it likely that the 169 patients reported in Starling et al. 52 are participants in the analysis by John et al. 65 The International Society for Heart & Lung Transplantation (ISHLT) registry report by Nativi et al. 89 included 417 patients who received later generation LVADs for BTT. Of these, 291 were implanted with the HMII device and some were likely to also be participants in the FDA approval study or its extension although outcome results for HMII were not reported.

FIGURE 5.

Summary of relationships between the included US HMII publications. All US multicentre BTT HMII FDA protocol publications shared patients with each other; the John 201063 publication reported on 250 patients who received a HT out of 486 BTT participants. The multicentre study by Brewer 201259 combined the 486 FDA study patients with 410 DT patients but did not report results separately according to therapy, some of the DT patients may have come from the single-centre studies. Among the US HMII single-centre publications, those of Petrucci 200974 and John 201164 contributed patients to the multicentre BTT HMII FDA protocol study. All other US HMII single-centre studies (n = 12) included both BTT and DT patients in single analyses (i.e. results not separated according to therapy) or did not state if patients received BTT or DT (Demirozu 2011,61 Martin 201069 and Kato 201266). The two studies of Topilsky 201179,80 investigated overlapping populations as also did the two studies by Schaffer 2009,77 2011. 76 The Nativi 201189 registry study may have included HMII patients from the FDA protocol study extension and may have included patients common to John 201165 which in turn included the patients from Starling 2011. 52 Solid arrows indicate publications that almost certainly shared participants and dashed arrows represent publications that probably shared patients. Numbers of patients are shown in brackets. a, CF VAD numbers only, data for HMII not reported separately; pulsatile VADs (n = 1980) excluded.

Figure 5 attempts to summarise the ‘family tree’ of the large number of US HMII VAD publications. As publications lacked sufficient detail, these relationships between publications cannot be stated with total certainty and it would be valuable to confirm this diagram with the authors. A similar situation of overlapping patient populations applies to the other included publications. These reported on a single device other than HMII, or reported results from studies conducted at European centres employing a mix of LVADs or the HMII device. There were two publications about HMII use with European patients;68,78 these included both BTT and DT patients. In Lahpor et al. 68 outcomes were not stratified by therapy and in Strueber et al. 78 the only outcome reported according to therapy was survival. The relationship between these is summarised in Figure 6. The European multicentre study of patients implanted with the HMII device for BTT, DT or BTR, by Lahpor et al. 68 (n = 184), included the patients (n = 101) reported separately by Strueber et al. ,78 and possibly some of the patients in the mixed VAD studies by Drews et al. 87 and Oswald et al. 90 The two BTT publications by Sandner et al. 91,92 examined the same 86 patient population (who received an amalgam of several devices which did not include HMII and results were not stratified according to device). The source of patients for the multiple device studies by Klotz et al. 88 and Drews et al. 87 were single German centres. Two publications of the DuraHeart (Morshuis et al. 42,85) investigated almost identical patient populations differing slightly in size (n = 6885 and n = 8242). A single-centre publication describing 79 HMII patients (BTT n = 64, BTT and DT n = 15) did not identify the centre and it was uncertain if this was a French or US study. 81

FIGURE 6.

Summary publication relationships: non-HMII single-device publications and European centre studies. One single-device publication was included for each of the INCOR, HW (centres in Europe and Australia) and MicroMed DeBakey devices. The INCOR study included both BTT and DT patients but results were not reported by therapy. The two DuraHeart publications had overlapping populations. Solid arrows indicate publications that almost certainly shared participants and dashed arrows represent publications that probably shared patients. a, The Drews 201087 population was ‘relatively contraindicated for HT’; the reports of Oswald 201090 and Klotz 200688 were also unclear on proportions BTT or DT patients.

The overlapping inter-relationship of populations described above, especially notable for HMII studies, renders any summary of baseline characteristics or of outcome results problematic if double counting is to be avoided. Therefore, where duplication of patients was judged to occur we have included the largest and/or most recent study of the cohort, conditional on availability of data. However, because of the multicentre nature of many of the HMII studies and authors' contention that experience with LVADs over time has influenced study results for some outcomes, we have occasionally also discussed earlier and smaller studies. We have organised baseline characteristics and outcome results according to device; in the main text we have not considered results for DT patients, or where results combine DT and BTT patients, or where this distinction was unclear. For full data on all baseline characteristics, and on outcome results irrespective of therapy, please consult Appendix 3.

Overall quality assessment

The 40 primary included publications42,52,53,56–92 were each quality assessed using an adapted set of criteria developed by Thomas et al. 55 For completeness, see Appendix 4 for copies of the quality assessment sheets for each publication.

Baseline characteristics

An apparent 19,161 participants were described in the 40 included publications; however, please see Delineating multiple overlapping populations between publications, which explains the issue of multiple reporting of patients by publications. The majority of the studies took place in the USA (n = 27); others were listed as taking place in Germany (n = 3), Europe specifically (n = 2), Austria (n = 2), unclear (n = 1) and multiple counties (n = 5).

All included studies reported some baseline characteristic values for the population investigated, one study reported five84 baseline characteristics while two others reported 43. 79,80 Method of reporting varied; for example, age was reported as a mean (standard deviation; SD), a median with range, a proportion or percentage within each of several defined age bands or as a combination of these methods. Some authors reported data for subgroups only but, where possible, subgroups have been combined to provide a value for the whole study population. Authors frequently used baseline characteristics in regression analyses attempting to identify factors that influence outcomes of particular interest in their study (e.g. aortic insufficiency,60,72 renal function62 and stroke58,81).

The baseline characteristics of BTT patients are presented in Figures 7–13. To avoid double counting caused by overlapping populations, the largest or most recent publication from each known cohort, conditional on the availability of data, has been included. Pooled estimates are provided. If the two large registry studies by Nativi et al. 89 and John et al. 65 were to be included in pooling then pooled estimates would merely reflect their input; therefore, these have been omitted from pooling but where possible have been compared with the pooled estimate. A further difficulty concerns whether or not mixed-device studies should be included. We have included data from Sandner et al. 91,92 when reported, as overlap with other studies is likely to be minimal. Pooled estimates should be treated with extreme caution as they:

1. may miss studies that should be included

2. may not be representative either of all included studies or of all patients within a particular VAD study (most studies defined sampling frames and patient selection methods poorly)

3. may include clinical heterogeneity and missing information (not all studies provided analysable information and we excluded studies with fewer than 50 patients).

Nevertheless, pooled estimates provide a picture of baseline characteristics of relevant populations receiving VADs.

Age

Where mean age was reported it varied between 45 years (Klotz et al. 88) and 65 years (Adamson et al. 56). Two large studies involving 115782 and 855789 participants reported comparable mean ages of 51 years. The distribution of mean age in the included non-overlapping studies of BTT patients is summarised in Figure 7. The pooled estimate was 50.8 years, and this is similar to the large registry study by Nativi et al. 89 of 8557 BTT LVAD patients, which found mean ages of 50.1, 50.2, 50.8, 51.4 and 51.8 years, respectively, in patients who received pulsatile first-generation LVADs, pulsatile second-generation LVADs, continuous second-generation LVADs and for second-generation patients (no LVADs) on inotropes and second-generation patients (no LVADs) not on inotropes.

FIGURE 7.

Mean age at baseline for individual studies. (Note: studies with minimal population overlap with other studies; only BTT patients included.) There was statistical heterogeneity between studies (I² = 64%). lCI, lower CI; uCI, upper CI.

Gender

The percentage of males in all these studies ranged from 68.5% (Lahpor et al. 68) to 91.5% (Morshuis et al. 85) (summarised in Figure 8).

FIGURE 8.

Baseline number (%) of patients reported to be male. (Note: studies with minimal population overlap with other studies; only BTT patients included.) lCI, lower CI; uCI, upper CI.

The pooled estimate was 84% with moderate heterogeneity (I² = 40%). Again, the combined result is similar to the large registry study of 8557 LVAD BTT patients reported by Nativi et al. 89 which found percentages of males to be 85.5%, 86.1%, 82.3%, 75.0% and 74.9% for pulsatile first-generation LVADs, pulsatile second-generation LVADs, continuous second-generation LVADs, second-generation patients on inotropes and second-generation patients not on inotropes respectively.

Body mass index

Fourteen studies59,62–65,71,72,77,81,83,86,88,89,91 reported baseline BMI (kg/m²) of patients. All but one of the studies reported BMIs suggestive that patients were overweight. 59,62–65,71,72,77,81,83,86,89,91 The results reported in non-overlapping studies of BTT patients are shown in Figure 9.

FIGURE 9.

Baseline BMI (kg/m²). (Note: studies with minimal population overlap with other studies; only BTT patients included.) lCI, lower CI; uCI, upper CI.

The pooled estimate of 26.5 kg/m² is similar to the value in the large registry study of 8557 LVAD patients reported by Nativi et al. ,89 which found values of 26.7, 27.4, 26.8, 26.2 and 26.3 kg/m², respectively, for BTT patients who received pulsatile first-generation LVADs, pulsatile second-generation LVADs, continuous second-generation LVADs, and for second-generation patients on inotropes and second-generation patients not on inotropes. A somewhat larger value of 28.8 kg/m² was reported by John et al. 65 for 1496 registry patients. It should be noted that in HF BMI may be misleading, owing to the underlying fluid retention of HF.

New York Heart Association functional classification of the extent of heart failure

A minority of studies (n = 16) reported baseline information on the NYHA functional classification of patients and where this was reported, the majority of patients were reported as having NYHA class IV. Overall, there was limited reporting of NHYA classification across all devices. Four-fifths (83.5%) of BTT patients had a rating of NYHA class IV assessed from non-overlapping studies (Figure 10).

FIGURE 10.

Baseline number (%) of patients with NYHA IV classification. (Note: studies with minimal population overlap with other studies; only BTT patients included.) lCI, lower CI; uCI, upper CI.

There was some heterogeneity among studies (I² = 38%). Neither large registry study65,89 included usable NYHA class information.

Diabetes mellitus

A total of 12 studies reported the number of patients with diabetes mellitus at baseline (HMII, n = 8; HW, n = 1; mixed devices, n = 3). The percentage of patients with diabetes mellitus at baseline in these studies ranged from 14% (Strueber et al. 83) to 38.5% (Pak et al. 72). In some studies, subgroups with different rates of diabetes mellitus were reported at baseline [e.g. Sandner et al. 91 reported 17.9% compared with 53.3% (group 1 aged < 60 years vs. group 2 aged > 60 years respectively)]. Figure 11 summarises results reported in non-overlapping studies of BTT patients.

FIGURE 11.

Baseline number (%) of patients with diabetes mellitus. (Note: studies with minimal population overlap with other studies; only BTT patients included.) lCI, lower CI; uCI, upper CI.

The pooled value of 25.2% is similar to reported values, ranging between 20.5% and 28.3%, for patients who received pulsatile first-generation LVADs, pulsatile second-generation LVADs or continuous second-generation LVADs, and for second-generation patients taking inotropes and second-generation patients not taking inotropes, in the large registry study of 8557 LVAD patients (Nativi et al. 89).

Inotropes

The baseline numbers (and percentages) of patients receiving inotropes in non-overlapping studies of BTT patients are shown in Figure 12.

FIGURE 12.

Baseline number (%) of patients using inotropes in non-overlapping studies of BTT patients. lCI, lower CI; uCI, upper CI.

There was heterogeneity between studies (I² = 49%) and percentage values ranged widely, from 40% to 100% of patients receiving inotropes; the pooled estimate of 81% is similar to the value of 80.4% reported for 1496 HMII patients in the large registry study by John et al. 65

Systolic blood pressure

Baseline systolic blood pressure (BP) was reported in 15 studies. 42,52,53,56,59,63,65,70,71,74,75,79,80,83,85 Figure 13 summarises the results reported in non-overlapping studies of BTT patients. Systolic BP can be seen to be low compared with normal physiological levels, reflecting the severity of HF in these patients, with a pooled estimate of 97.3 mmHg.

FIGURE 13.

Baseline systolic BP (mmHg). (Note: Studies with minimal population overlap with other studies; only BTT patients included.) lCI, lower CI; uCI, upper CI.

Heterogeneity (I² = 83%) between studies arose mainly from the study by Morshuis et al. 85 The pooled estimate is similar to the value of 100.9 mmHg reported for 1496 HMII patients in the large registry study by John et al. 65 Nativi et al. 89 did not report this characteristic for registry patients.

Summary of baseline characteristics

In so far as it was possible to separately identify non-overlapping groups of patients, we identified the following baseline characteristics. The majority of patients were white (78–94%), male [84.2% (95% CI 79.4% to 88.0%)] and middle aged [mean age 50.8 years (95% CI 49.3 to 52.4 years)]. Mean BMI was in the overweight range [mean BMI 26.5 kg/m² (95% CI 25.7 to 27.3 kg/m²)] and about one-quarter of patients [25.2% (95% CI 17.4% to 35.1%)] had diabetes mellitus. Study patients had very severe HF, with 83.5% (95% CI 78.0% to 87.9%) overall rated as NYHA class IV. This was supported by the proportion receiving inotrope medication [80.8% (95% CI 50.9% to 94.5%)] and the low mean systolic BP of 97.3 mmHg (95% CI 92.8 to 101.7 mmHg).

Outcomes by device

In this section we describe outcomes including adverse events, survival, causes of death and QoL for each device.

Outcomes for HeartMate II

A total of 29 studies52,53,56–82 met the current inclusion criteria concerning HMII (see Table 3).

Adverse events

Twenty-three studies52,56–61,64–73,75,76,78,79,82 reported adverse events or complications during the follow-up of HMII implantation. Given the problems of reporting on overlapping populations within studies (see Delineating multiple overlapping populations between publications), adverse events for HMII are best described by John et al. 65 (Table 9).

TABLE 9 - Adverse events. Adapted from John et al.65

Adverse event	John et al.65 trial group (n = 486), 511.1 patient-years
	Incidence (% of patients)	Event rate/patient-year
Bleeding requiring re-exploration	21	0.23
Infection
Percutaneous lead infection	20	0.33
Pump pocket infection	3	0.03
Right-side HF requiring RVAD	7	0.06
Stroke
Ischaemic	5	0.05
Haemorrhagic	5	0.05
Other	0	0.00
Device replacement	5	0.06

Adverse events with HMII affect high proportions of patients. Twenty-one per cent of patients had bleeding requiring re-exploration; 20% had percutaneous lead infection and 3% pump pocket infection. Stroke is a very serious adverse event affecting approximately 10% of patients (it is assumed the ischaemic and haemorrhagic stroke did not occur in the same patient). Event rates per year should be treated with caution as follow-up rates are variable and events are highest in the first year after surgery.

Survival

Table 10 summarises the K–M survival results as reported in the HMII studies (see Table 3).

TABLE 10 - Summary of K–M survival results reported in HMII studies

NR, not reported; SE, standard error.

These studies dichotomised the patients from the FDA approval study extension.

Starling 201152 and Miller 200770 did not report an error value.

The first 169 HMII patients post Miller 2007,70 FDA approval study; almost certainly these patients were among the 1496 registry patients in John 2011. 65

Study	Population	n	% (SE) alive
			Month 1	Month 6	Month 12	Month 18
aBogaev 201157	Women	104	96 ± 2	87 ± 4	76 ± 3	73 ± 3
aBogaev 201157	Men	361	93 ± 1	83 ± 2	74 ± 5	73 ± 5
John 201165	HMII FDA approval trial extension	486	NR	83.8 ± 1.8	75.6 ± 2.4	NR
John 201165	INTERMACS post-FDA approval	1496	NR	89.4 ± 0.9	84.9 ± 1.1	NR
aKormos 201067	With early RV failure	65	89 ± 4	66 ± 6	59 ± 7	NR
aKormos 201067	Without early RV failure	386	94 ± 1	87 ± 2	78 ± 3	NR
bMiller 200770	HMII FDA approval trial	133	90	75	68.5	NR
Pagani 200971	HMII FDA approval trial extension	281	92 ± 2	83 ± 3	73 ± 3	72 ± 3
aPal 200973	Concurrent cardiac procedures	111	89 ± 4	77 ± 5	66 ± 7	NR
aPal 200973	No concurrent cardiac procedures	271	94 ± 2	84 ± 3	77 ± 4	NR
bStarling 201152	INTERMACS post-approval study for FDAc (first 169 patients post Miller 200770)	169	96	90	85	NR

The most recent and largest study of survival in the HMII BTT programme for HF appears to be that of John et al. ,65 who included 486 patients from the extension of the HMII FDA approval study and 1496 post-approval patients in the INTERMACS registry. This analysis indicated superior survival for the latter patients (p < 0.0001 for log-rank test comparison). These results led the authors to propose that increasing experience with the HMII device has led to a gradual improvement in survival (Figure 14).

FIGURE 14.

Kaplan–Meier survival results reported by John et al. 65 for patients who received an HMII LVAD.

Earlier publications in the HMII FDA approval series included those of Miller et al. ,70 with 133 patients, and Pagani et al. ,71 with 281 patients; they represent samples from John et al. ,65 a growing cohort of the HMII approval study patients which accumulated 486 patients. The survival results reported are shown in Figure 15 together with those for the first 169 post-approval patients analysed in Starling et al. 52 The results tend to support the proposition that a so-called learning curve leads to improving survival as the cohort grows. Survival at 1 year was 68.5% in Miller et al. ,70 73% in Pagani et al. ,71 75.6% in John et al. 65 and 85% in Starling et al. 52 as well as for the 1496 registry post-approval patients in John et al. 65

FIGURE 15.

Kaplan–Meier survival results for the HMII FDA approval trial and registry patients.

Greater experience with the device offers one explanation for the apparently improving survival; however, the similarity of these populations at baseline is difficult to gauge.

FIGURE 16.

Kaplan–Meier results reported for subgroups of patients receiving HMII LVAD.

Three publications reported survival for subgroups of patients in the extension of the HMII FDA approval study (Bogaev et al. ,57 Kormos et al. 67 and Pal et al. 73). In each of these studies participants were dichotomised according to a single variable. The results are summarised in Figure 16. No significant difference was observed between genders, although early RV failure was associated with poorer survival (p = 0.026), as were concurrent cardiac procedures undertaken (p = 0.048).

In summary, publications suggest that survival at 1 year is approximately 75% but may improve with gain in surgical experience. These K–M analyses censor patients when they receive a HT. The problem here is that if the chance of receiving a donor heart depends on a patient's prognosis, for example if more seriously ill patients are selectively removed from follow-up and receive priority for transplantation, then survival estimates are susceptible to informative censoring and may thereby represent overestimates.

Causes of death

Twelve of the included HMII papers reported causes of death. 56,57,64,68,70–73,78–80,82 Of these, the most recent and largest publication reporting this outcome in the HMII BTT programme for HF appears to be that of Bogaev et al. 57 This publication dichotomised participants according to gender. The reported leading causes of death in men were sepsis (3.9%), right HF (2.8%) and multisystem organ failure (2.2%). The leading causes of death in women were multisystem organ failure (3.8%), haemorrhagic stroke (2.9%), ischaemic stroke (1.9%), right HF (1.9%) and external component device malfunction (1.9%; percutaneous lead trauma in one patient and pump disconnection in another). 57 Table 11 provides a summary of the causes of death reported in the 12 HMII studies.

TABLE 11 - Causes of death reported in included HMII papers

First author	Date	Country	Causes of death
Adamson56	2011	USA	Anoxic brain injury, cardiomyopathy, sepsis, respiratory failure, multiorgan failure, ischaemic stroke, haemorrhagic stroke, device thrombosis, patient disconnected power, cancer, withdrawal of support, unknown causes
Bogaev57	2011	USA	Sepsis, right HF, multisystem organ failure, ischaemic stroke, haemorrhagic stroke (thrombi, pump disconnection, twisted inflow graft, pump pocket infection, loss of power, percutaneous lead trauma), respiratory failure, cardiac failure, bleeding, cancer, elective withdrawal of support, death during transplantation, unknown causes
John64	2011a	USA	Multisystem organ failure, subclavian vein haemorrhage, ventricular fibrillation, respiratory failure, RV failure, intracranial bleed
Lahpor68	2010	Europe	Multiorgan failure mainly occurring as a result of septic complications or right-HF, CVAs
Miller70	2007	USA	Sepsis, ischaemic stroke, multisystem organ failure, haemorrhagic stroke, anoxic brain injury, right HF, miscellaneous other causes, device-related death caused by an inflow graft that was accidentally twisted during implantation
Pagani71	2009	USA	Sepsis, stroke (ischaemic, haemorrhagic), right HF, device related, multiorgan failure, anoxic brain injury, bleeding, cancer, respiratory failure, hyperthermia, air embolism
Pak72	2010	USA	Aortic insufficiency onset of multiorgan system failure
Pal73	2009	USA	Bleeding, sepsis, driveline infection, ventricular arrhythmias, perioperative stroke, renal failure
Strueber78	2008	Multiple	Multiorgan failure, right HF, CVAs, respiratory failure, driveline disconnection, bleeding after ventricular rupture, suffocation after epistaxis (nose bleed)
Topilsky79	2011a	USA	Multiorgan failure, intractable right HF, hyperperfusion brain injury, sepsis, uncontrollable bleeding
Topilsky80	2011b	USA	Uncontrolled right HF, multiorgan failure, intracerebellar bleeding, traumatic head trauma injury, haemorrhagic stroke, unexplained sudden death, RV failure, embolic stroke, complication of myocardial biopsy after transplant, patient withdrawal of support owing to persistent RV failure and need for dialysis
Ventura82	2011	USA	Graft failure, infection, cardiovascular, cerebrovascular, multiorgan failure, haemorrhage, malignancy, unknown causes

Common causes of death included (a) multiorgan failure (n = 12 studies), (b) right heart (ventricular) failure (n = 8 studies), (c) bleeding (n = 7 studies) and (d) stroke and cerebrovascular accident (CVAs) (n = 9 studies).

Quality of life

The HMII studies used several instruments for monitoring QoL and functional status of HF patients. Table 12 provides a summary of HMII studies indicating which QoL measures were used.

TABLE 12 - Quality of life and functional status outcome measures reported for HMII

KCCQ, Kansas City Cardiomyopathy Questionnaire; METs, metabolic equivalent task score; VAS, visual analogue scale.

First author	Date	Country	MLWHF	KCCQ	METs	VAS	NYHA functional class	6-minute walk test
Rogers53	2010	USA	✓	✓	✓		✓	✓
Topilsky79	2011	USA					✓
Topilsky80	2011	USA					✓
Adamson56	2011	USA	✓	✓	✓		✓	✓
Bogaev57	2011	USA	✓	✓	✓		✓	✓
John65	2011	USA		✓		✓		✓
Miller70	2007	USA	✓	✓			✓	✓
Pagani71	2009	USA	✓	✓			✓	✓
Starling52	2011	USA				✓

For this variety of QoL measures, data were presented as group means at various time points, change in group mean from baseline, or as change in mean or median for paired measures for individual patients. The number of patients investigated during individual studies gradually diminished as a result of death, transplantation with a donor heart and loss to follow-up or withdrawal. Baseline values were not always complete because some patients may not have been sufficiently well to participate. Full QoL results can be found in Appendix 3.

Five of the nine studies (Rogers et al. ,53 Bogaev et al. ,57 John et al. ,65 Miller et al. 70 and Pagani et al. 71) contribute to the multicentre HMII BTT FDA approval programme with overlapping populations. By far the fullest QoL information was provided in Rogers et al. 53 (n = 281). Although the largest BTT patient groups were investigated in Bogaev et al. 57 (n = 465) and John et al. 65 (n = 486), relatively limited results were presented and neither report paired measures; therefore, here we focus on the data presented in Rogers et al. 53 (for this study information in Pagani et al. 71 can be used to gauge the completeness of the reported data). Rogers et al. 53 presented results separately for 281 BTT and 374 DT patients; the DT results were similar to those for BTT patients. (They are not considered further here but can be found in Appendix 3.) In the following section, the QoL results presented by Rogers et al. 53 are summarised and considered separately according to the investigatory instrument. Comments on the results from Bogaev et al. 57 and John et al. 65 should be viewed in the knowledge that the populations in these studies included participants from Rogers et al. 53

Minnesota Living With Heart Failure Questionnaire

Scores on the MLWHF questionnaire ranged from 0 to 105, with lower values signifying improved QoL. Scores reported by Rogers et al. 53 decreased over time relative to baseline scores (–10 points at 1 month and –29 points at 6 months; median per cent improvement at 6 months of 38%), indicating an improvement in QoL (p < 0.05). These results for MLWHF scores at 1, 3 and 6 months for BTT patients are summarised in Figure 17.

Reading from the graph in Pagani et al. 71 the number of patients supported on HMII at 6 months was about 132, so that the return of 115 questionnaires at 6 months represents a data set approximately 87% complete.

Rogers et al. 53 also reported paired change (i.e. mean change from baseline for patients with measures at both time points) (Table 13).

TABLE 13 - Paired changes in MLWHF scores reported in Rogers et al.53

p < 0.001 for improvement in median score.

BTT
Month	n	Mean ± SD	Median [25th, 75th percentiles]	% improvement of median
1	167	−12 ± 27	−10 [−28, 4]	13
3	126	−24 ± 31	−30 [−47, −4]	39
6	87	−28 ± 28	−29 [−50, −9]	38

FIGURE 17.

Changes in MLWHF for the BTT group over time (Rogers et al. 53). Bars indicate 25th, 50th and 75th percentiles, and the whiskers indicate fifth and 95th percentiles.

Borgaev et al. ,57 who subdivided the population by gender, reported similar results: a significant improvement (both genders group) between baseline and 6 months (female: 73 ± 22 to 35 ± 22; male: 71 ± 22 to 40 ± 23). There was no significant difference between the sexes (p = 0.661).

The Kansas City Cardiomyopathy Questionnaire (KCCQ) overall summary score (OSS) and clinical summary score (CSS) improved during HMII support. The median KCCQ OSS showed improvements at 1, 3 and 6 months compared with baseline and (p < 0.05 at each time point). KCCQ group OSS also showed similar improvements (p < 0.05 at each time point) (Figure 18).

FIGURE 18.

Kansas City Cardiomyopathy Questionnaire scores (Rogers et al. 53). Bars indicate 25th, 50th and 75th percentiles, and the whiskers indicate fifth and 95th percentiles. (a) OSS; and (b) CSS.

Rogers et al. 53 reported paired KCCQ score changes (i.e. mean of change from baseline for patients with measures at both time points). These results are summarised in Table 14.

TABLE 14 - Paired changes in KCCQ OSS and CSS

OSS					CSS
Month	n	Mean ± SD	Median [25th, 75th percentiles]	Improvement of median	Month	n	Mean ± SD	Median [25th, 75th percentiles]	Improvement of median
1	172	13 ± 25	14 [−3, 29]	0.54	1	170	12 ± 27	11 [−6, 31]	0.3
3	132	22 ± 26	20 [9, 42]	0.77	3	132	21 ± 28	21 [4, 42]	0.57
6	90	27 ± 28	28 [7, 45]	1.08	6	90	25 ± 31	24 [8, 43]	0.65

Again, paired measures support the evidence for an improvement in QoL for patients surviving after implant of the HMII VAD (p-value and statistical significance were not reported).

Borgaev et al. ,57 who subdivided the population by gender, reported significant improvements for both men and women in mean values on the KCCQ OSS and CSS between baseline and 6 months (there was no significant difference between the sexes).

Metabolic equivalent task score

The metabolic equivalent task score (METs) measures patient-reported exercise ability. Rogers et al. 53 presented serial assessments of METs following HMII (Figure 19). At baseline, > 90% of patients described their level of function as low or very low. At 6 months, about two-thirds of patients described their level of function as moderate to very high (p < 0.001 vs. baseline).

FIGURE 19.

Summary of METs. BTT FDA study: data from Rogers et al. 53

Borgaev et al. 57 found no significant difference between males and females in METs improvements (p = 0.348).

Quality of life visual analogue scale

Rogers et al. 53 did not report on this outcome. For purposes of completeness, results from other studies are reported here. The data presented by Starling et al. 52 for the first 169 post-FDA-approval HMII patients are summarised in Figure 20. Visual analogue scale (VAS) scores (scale 0–100; best QoL = 100) improved at 3, 6 and 12 months relative to baseline. Changes were large, but p-value and statistical significance were not reported. Results were based on 253 tests (50%) completed in 508 potential test sessions.

FIGURE 20.

Visual analogue scale for HMII and comparator VADs at different time points (Starling et al. 52). Note: the comparator group consisted of patients who received other LVADs than HMII approved for BTT.

John et al. 65 presented METs QoL data for a sample from the INTERMACS registry BTT post-approval HMII patients (Table 15). Improvements of 32 points at 3 months were sustained at 12 months. The authors provided limited discussion of these findings. The overlapping underlying populations preclude development of a summary estimate combining results with those of Starling et al. 52

TABLE 15 - Visual analogue scale results for the post-trial cohort. Adapted from John et al.65

Item	Pre implant	3 months	6 months	12 months
n (at risk)	1498	1142	822	393
n (completing test)	777	617	432	192
Per cent (completing test)	52	54	53	49
VAS score	42	74	75	76

Functional status

New York Heart Association

Rogers et al. 53 reported that at baseline patients were classified as NYHA class IV, by 1 month 59% had improved to NYHA class I or II, and at 6 months 82% were NYHA classified as class I or II (Figure 21). Relative to baseline scores, highly significant improvements in NYHA functional class were observed at all study intervals (p < 0.001).

FIGURE 21.

Summary of NYHA classification at each time point reported (Rogers et al. 53). Ratings were determined by an independent clinician.

Bogaev et al. 57 reported significant improvements from baseline in the proportion of patients classified as NYHA functional class I/II for both women [0–49 (83%)] and men [0–147 (85%)] (p < 0.001). No significant differences were observed between men and women (p = 0.55).

6-minute walk test

At baseline, Rogers et al. 53 reported that of 281 BTT patients 38 (14%) were able to perform the 6-minute walk test. Baseline distance walked for was 214 ± 125 metres. At 6 months the distance walked was 372 ± 199 metres although only 97 patients completed the test. There was a statistically significant improvement over time.

Bogaev et al. 57 reported that before LVAD implantation, many patients were unable to walk and could not provide baseline values. Group means exhibited significant change for both women and men at 1, 3, and 6 months. Distance walked at all times was further for men (p = 0.037). Registry data for the 6-minute walk test results reported by John et al. 65 similarly indicated improvement from baseline. No statistical analysis was reported.

Summary of quality-of-life results

The results presented by Rogers et al. 53 provide a persuasive indication that those patients who survive implantation of the HMII device as BTT experience an improvement in QoL by 3 months sustained at 6 months. Some of the changes are substantial and statistically significant (e.g. improvements in MLWHF, KCCQ, NYHA and METs in Rogers et al. 53), but data sets were not complete and this may have skewed results.

These results were supported in other publications with somewhat larger populations and where measures are extended to 12 months.

Outcomes for HeartWare

One study reporting on 50 patients implanted with a HW VAD as a BTT fulfilled our inclusion criteria (Strueber et al. 83).

Adverse events

Adverse events were reported in detail and are shown in Table 16. Overall, 22 infections occurred among 50 patients; 20% of patients required repeat surgery for bleeding. Six of the 50 patients suffered from stroke and seven device replacements were required. Other important adverse events suffered by smaller numbers of patients included renal and hepatic dysfunction, haemolysis and right HF. Some patients may have experienced multiple events; therefore, summing percentages within categories may be misleading.

TABLE 16 - Adverse events for patients with a HW device. Adapted from Strueber et al.83

TIA, transient ischaemic attack.

Adverse events	Patients with events (%)	Number of events, overall	Number of events, 0–30 days	Number of events, > 30 days
Infection
Localised non-device related	7 (14)	7	2	5
Sepsis	5 (10)	5	1	4
Driveline exit site	9 (18)	10	0	10
Bleeding
Requiring surgery	10 (20)	11	8	3
Requiring transfusion ≥ 2 units	2 (4)	2	1	1
Requiring hospital stay	3 (6)	3	1	2
Ventricular arrhythmias	2 (4)	2	1	1
Neurological dysfunction
Ischaemic stroke	2 (4)	2	2	0
Haemorrhagic stroke	4 (8)	4	0	4
TIA	2 (4)	3	0	3
Pulmonary dysfunction	8 (16)	9	8	1
Device replacement
Manufacturing defect	2 (4)	2	2	0
Left heart embolus	4 (8)	4	1	3
Inflow occlusion	1 (2)	1	1	0
Pleural effusion	6 (12)	7	5	2
Right HF
RVAD	3 (6)	3	2	1
Intravenous inotropes	3 (6)	3	1	2
Other serious adverse events
Renal dysfunction	5 (10)	5	5	0
Hepatic dysfunction	3 (6)	3	1	2
Haemolysis	1 (2)	1	1	0
HF	3 (6)	3	1	2
Chest pain	1 (2)	1	0	1
Femoral embolism	2 (4)	2	1	1

Survival

Table 17 provides a summary of the K–M survival results for patients who received the HW VAD in the study by Strueber et al. 83

TABLE 17 - Kaplan–Meier survival results for patients who received the HW VAD

NR, not reported.

Study population	Group	n	Month 6	Month 12	Month 18	Month 24
Strueber et al.83 BTT (patients from Europe and Australia)	HW	50	90%	85%	NR	79%
	Virtual control	50	73%	58%	48%	40%

Strueber et al. 83 reported K–M survival results for HW BTT patients (n = 50; Figure 22). Relative to HMII studies (shown in Figures 14 and 15), survival with HW appears to be at least comparable, with 85% of patients alive at 1 year after implant. When compared with the earlier HMII publications, survival appears superior for HW. Strueber et al. 83 also provided a survival curve for a ‘virtual control’ group; this was based on the application of the Seattle Heart Failure Model (SHFM) to the baseline characteristics of the intervention group. This virtual control data fits well when a SHFM score of 2.416 is applied. These results are summarised in Figure 22.

FIGURE 22.

Survival of BTT patients implanted with the HW. Reported by Strueber et al. 83

Causes of death

Reports on nine deaths from the 50 eligible patients in the HW study78 suggested that three were caused by sepsis, three by multiorgan failure and three were thought to be caused by haemorrhagic stroke.

Quality of life

Strueber et al. 83 reported KCCQ data for stated sample sizes of 38, 37, 36 and 21 presurgery and 1, 3 and 6 months post surgery for the 50 patients in their study. Results are summarised in Figure 23. It was unclear if these were paired data.

FIGURE 23.

Group mean scores for the KCCQ (Strueber et al. 83). Bars indicate means with 95% CIs and asterisks indicate 90th percentiles. *Statistically significant improvement reported in KCCQ score at p < 0.05.

Health-related quality of life (HRQoL) improved significantly by 1 month for all subscales of the KCCQ. The authors found statistically significant (p = 0.05) improvement in physical limitations, QoL, symptom burden, and overall functional status across all time points. Greater improvements were found during the first 30 days following the HW implant.

Summary of clinical effectiveness findings

We have reported outcomes for the 40 included publications (for full details see Appendix 3). The lack of prospective comparative study design, modest study quality, diversity of reporting of outcomes, and overlap between populations investigated, render it difficult to draw firm overall conclusions. The only comparisons between devices reported were between early generation VADs and second-/third-generation devices. (See also Appendix 3. )

For all the devices there was a profile of substantial serious adverse events caused by infection, thrombosis, bleeding and stroke, and of mortality from various causes in this already frail population. By 12 months patients had suffered a variety of serious complications. Studies reported the following ranges for adverse events: 4–27% bleeding requiring transfusion; 1.5–40% stroke; 3.3–48% infection (sepsis); 1–14% device failure; 3–30% HF; 11–32% reoperation; and 3–53% renal failure. Table 25 gives a summary of the range of rates of the main adverse events by device type per patient-year demonstrating these high rates of adverse events.

The wide range of rates for stroke, with a very high upper end, represents data from a variety of studies. Higher rates of stroke emanate from shorter studies, as stroke is more likely to occur in the first 3 months after an implant (Figure 24).

FIGURE 24.

Hazard analysis depicting varying incidence over time of four major adverse events: stroke, reoperation for bleeding, pump thrombus and haemolysis (redrawn from Goldstein 200384).

Early follow-up data (e.g. at 1–3 months) cannot reliably be extrapolated to longer time periods (e.g. over 6 months) owing to the changing adverse event profile over time. Also, as few of the papers reported outcomes beyond 12 months, numbers and percentages in Table 25 represent the best estimate of adverse events likely in the first year after the VAD intervention, but cannot reliably be extrapolated to later years after the intervention.

Kaplan–Meier estimates of survival post implant of the HMII device suggest improvement with growing experience. The best 1-year survival estimate for this device was 85%. A similar estimate of 85% survival at 1 year was reported for 50 HW patients (investigated at centres in Europe and Australia). While preparing this report, Aaronson et al. 94 published a larger study of 140 US HW patients and estimated survival at 1 year to be 86%. Estimates of survival at 1 year for other devices as reported in the included publications were less impressive (INCOR 61% and DuraHeart 76%). It should be borne in mind that in estimating survival of BTT patients during VAD support using K–M analyses, those patients who receive a HT are censored at the time of transplant; if these patients are unrepresentative of the overall population studied (e.g. have a poorer prognosis than uncensored patients) then survival may be overestimated (and vice versa). Furthermore, any comparison across device types for any outcome may be confounded by differences in underlying populations (e.g. geography, time period, eligibility criteria, case mix, etc.)

Quality of life and functional status, where these were measured for patients who were still alive, showed a trajectory of improvement in the first year after implant for all groups of patients especially over the first 3 months. Improvements at 6 months were statistically significant in studies of HW and HMII.

In the next chapter we describe the individual patient data (IPD) set provided by the NHS Blood and Transplant National Registry (BTNR) from the UK Blood and Transplant Database (BTDB) maintained on behalf of the UK transplant community and explain derivation of parameters for the Warwick Evidence cost-effectiveness model.

Waiting list patients

The WL data set reported analysable registration data for 33 BTT patients and 1496 MM patients collected during the period January 2002 to December 2011. Figure 27 illustrates the outcomes for the MM subgroup as reported in the data set.

Of the 1496 MM patients, 1005 (67.18%) patients received a HT (Transplanted) and 154 (10.3%) patients died while on the WL (Died on WL) before receiving a HT. The data set included 207 (13.84%) patients who were completely removed from the WL because of substantial improvement in their condition, deterioration in their condition or through patient choice. These patients were referred to as Removed from WL in the data set. Of the remaining 130 (8.69%) patients, some were temporarily removed (Suspended) from the WL due to serious illness such that a HT was inadvisable or because further medical testing was required, while others (Active) were still waiting for a transplant.

FIGURE 27.

Medical management patients.

The next section gives outcomes for the various subgroups identified in the MM arm.

FIGURE 28.

Non-inotrope subgroup.

Baseline characteristics for the non-inotrope group (Figure 28) are reported in Table 26. The median reported age of the patients in this non-inotrope group was 48 years. Almost three-quarters of the group were male. These patients had poor health as evidenced by their NYHA class. More than 95% of the patient groups were categorised either NYHA class III or class IV. Approximately one-third of the patients had had an automatic implantable cardioverter defibrillator (AICD) before registration for a HT. Hypertension was present in a fifth of the patients. One-third of the patients in this subgroup had already had open heart surgery. Just over 3% of patients had diabetes mellitus prior to HT registration.

Patients in the inotrope group were older than those in non-inotrope group; the median age in the inotrope group was 58 years. Almost three-quarters of patients were male. The patients in this subgroup had much poorer health than the non-inotrope group. The NYHA class confirm this, as > 99% of the patients were either NYHA class III or class IV. Almost a fifth of the patients had AICD before registration. A small number (22%) of patients had already undergone open heart surgery and a very small number had already undergone a previous HT. Table 26 reports the baseline characteristics of the ‘other’ group illustrated in Figure 29.

FIGURE 29.

The ‘other’ group.

Patients in the ‘other’ group have a median age of 50 years. Three-quarters were male. Approximately three-quarters of the patients (∼ 73%) in this group belonged to NYHA class III with fewer (∼ 24%) belonging to NYHA class IV. There were some patients for whom NYHA class was not reported. One-quarter of these patients had already had an open heart surgery and three patients had already undergone HT surgery.

Summary table

Baseline characteristics for each of the three groups of MM patients are given in Table 26.

Baseline characteristics for inotrope patients (Figure 30) are reported in Table 26.

FIGURE 30.

Inotrope subgroup.

Bridge to transplant patients

To gain a better understanding of the BTT patients the data were cleaned and merged. Individual identifiers in separate data sheets were matched and then combined so that individual patient characteristics of patients who received a VAD implant were amenable for further analysis.

Figure 31 illustrates the case mix of patients who had received a VAD. The cleaned data set included 235 of the total 389 VAD patients. All these patients were further categorised into three main categories depending on the current status of their VAD implant. If patients still had a VAD and were awaiting HT they were categorised as VAD patients; just over three-quarters of the patients (n = 181; 77%) belonged in this category. Thirty-three patients (14%) had received a HT and are referred to as Transplanted. The remaining patients had had their VAD removed and are referred to as Explanted.

FIGURE 31.

Ventricular assist device patient set.

The number of VAD implants has grown with an overall eightfold increase over a period of 10 years and with significant growth over the last 3 years. The year-over-year graph (Figure 32) illustrates this trend.

FIGURE 32.

Trends of rising VAD implants over a period of 10 years.

Current practice in the UK is to use a second- or higher-generation VAD implant in patients. Table 27 lists the second or higher generation FDA-/CE-approved VADs used for the UK patient cohort. The two most commonly used VADs are HW and HMII.

Baseline characteristics of VAD patients are reported in Table 28. Information mainly includes patient demographics, past medical history and some clinical measurements. The median age for VAD patients was 47 years and > 80% were male. Almost three-quarters of VAD patients had received intravenous inotrope therapy. ARB use was reported for 10% of patients. Beta-blockers were used in more than one-third of the patients, as were ACE inhibitors.

Previous medical history also included previous VAD implant, use of IABP or extracorporeal membrane oxygenation (ECMO). Just over 25% of VADs patients already had IABP before implant. A very small number (∼ 3%) of the patients were on ECMO support during the pre-operative phase.

Within the first 7 days a number of adverse events were reported in BTT patients. These are shown in Table 29. Right HF as well as renal failure were the most common problems faced by these patients. Infection and neurological dysfunction were other prominent adverse events which occurred in the VAD-supported patients. A few patients (n = 3) also suffered from device malfunction.

Adverse events reported during the follow-up period post 7 days VAD implant are reported in Table 30. Infection was the most common, ∼ 15% patients suffered with infection. Other patients had neurological dysfunction and right HF. During the follow-up period two patients were rehospitalised each due to infection and right HF.

Some of the VAD explanted patients had adverse events. Infection (n = 3) and neurological disorder (n = 3) were the two main adverse events among patients after the VAD explant surgery.

During follow-up period of the VADs patients received a HT while some remained on VAD implants. Some patients (n = 68) died. The main reasons reported for death are given in Table 31.

Heart transplant patients

Figure 33 explains patient mix of the HT patients. There were 1101 patients, including BTT patients as well as MM patients, on the WL for HT. Of these, 33 of the patients were in the BTT arm and had received a VAD implant later followed by a HT. The majority of patients (91.2%), however, were from the MM subgroup. Almost half were in the non-inotrope patient subgroup. The remaining half was shared between the inotrope and the ‘other’ subgroup. There were 64 unregistered patients who spent no time on the WL but went straight to a HT.

FIGURE 33.

Heart transplant patients.

Baseline characteristics for the HT patients are given in Table 28. The reported median age for a HT patient was 48 years. Almost three-quarters of the patients were male. Almost all the patients belonged to either NYHA class III or class IV, although there were a few in NYHA class I and class II. Of note are the differences in NYHA classification between transplant patients and VAD patients in Table 28, with very different inotrope usage (∼ 75% for the VAD patients compared with ∼ 25% for the HT patients).

Intra-aortic balloon pump use was reported in almost 10% of transplant patients and ECMO support in 1% of transplant patients during the pre-transplant phase.

Strengths

1. The database was comprehensive in that it contained information on all patients listed for HT within the six designated UK NHS centres between May 2002 and December 2011.

2. It was possible to identify which VAD had been implanted into BTT patients.

3. It was possible to distinguish between BTT and MM patients. The patients who fulfilled the eligibility criteria (see Appendix 6) for HT were clearly registered for HT.

4. The data set reported information about the various important events occurring at different time intervals during the BTT and post-transplant phases.

5. Mortality and time of transplant were recorded in an efficient manner in that patients were followed up routinely and at each time point, current transplant and mortality status were reported. It has therefore been possible to make reasonable estimates of transition probabilities between health states necessary for economic modelling.

Limitations

1. The data set had no information about patients during the pre-VAD phase. Assumptions about clinical treatment and events during that phase had to be made in order to carry out economic modelling.

2. The data set had missing values for some of the important covariates for some of the patients. It had been hoped to use the full set of SHFM covariates in order to make use of the SHFM95 to model survival for VADs patients in the database, but because of incompleteness of data it was necessary to undertake modelling with those few covariate values that were available in the database.

3. A few of the patients (n = 15) in the data set had an apparent negative waiting time for receipt of a donor heart. This happened because such patients are taken off the WL on receiving a VAD and then put on the WL for receiving a HT, and only recent registration dates are available.

4. NYHA stage was not reported for all BTT patients. A full data set for this variable would have allowed a more rigorous assessment of QoL and utility values for use in the economic model.

5. In the systematic review of clinical effectiveness the adverse events were considered according to the type of device. However, we have described here adverse events experienced post-VAD implantation for all devices as the BTDB data did not clarify these in an interpretable format after 7 days of implantation.

6. There were no cost data in the database.

7. It was hoped to use the database to estimate resource-use incidence and intensity for each health state so that an estimate of current costs associated with the NHS transplant programme could be made in conjunction with NHS reference costs and based on the BTDB. However, as data sets were not complete in terms of resource use – especially for hospital stay and duration, and frequency of medication use in the different health states – alternative modelling procedures were necessary.

Quality assessment

Quality assessment of the BTDB was carried out96 (Table 32). It should be noted that the four data sets were updated over a period of 10 years.

Quality assessment was carried out with the help of the statisticians at the BTNR. The completeness of recruitment is very high for all data sets. Three data sets reported 97% catchment of the target population, and one data set captured most of them (90–97%). However, there is some concern about completeness of data in all the data sets. As already mentioned, several important covariates (n = 16) of the SHFM95 model were not reported and only 5 of the total of 21 covariates used for SHFM analysis were reported. The data sets used explicit definitions for covariates which were well explained for each of the four data sets. Three data sets reported thorough range, consistency and validity checks.

Comparison of individual patient data with literature estimates

Table 33 gives a comparison of the BTDB individual patient baseline characteristics for the 235 UK VAD patients with values for similar patients reported in the international literature (see Chapter 3). The UK population implanted with VADs and awaiting a donor heart appear to be younger in mean age, with a less severe NYHA class rating and are more likely to be white.

The proportion of patients using inotropes is of particular relevance to the economic model used in this report. More than three-quarters of BTDB VAD recipients used inotropes, whereas the pooled estimate for published BTT studies with non-overlapping populations and the estimate reported in the registry study by John et al. 65 was slightly higher, at 80%. Just over 20% (307) of the 1496 BTDB MM patients were categorised as using ‘inotrope’ treatment; this lower use of inotropes relative to the BTDB BTT patients tends to support the use of the ‘inotrope’ category group patients for the base-case comparator group in the economic model.

Aim of analysis

The authors aimed to estimate the cost-effectiveness of the HMII using the most robust and recently published evidence about its comparative performance compared with conventional therapy for patients with advanced HF.

Model structure

The model was a semi-Markov discrete-time multistate model with monthly cycles; the model design was the same as used by Sharples et al. 30 The health states were:

1. support on VAD to HT

2. support on conventional care to HT

3. support in the HT state

4. dead.

Model inputs

Transition probabilities between health states 1 and 4 (VAD to dead), 2 and 4 (conventional care to dead) and 3 to 4 (transplanted to dead) were based on data in published studies describing survival rates: these are summarised in Table 34.

For the base-case analysis, all patients alive in each arm at 6 months received a HT [92% and 76%, respectively, for BTT with VAD and conventional therapy patients; transition probability (TP) at 6 months = 1]. For sensitivity analyses receipt of a transplant took place at 12 months or at 18 months. In this way the model did not require extrapolation of survival for patients with BTT or conventional care beyond the observed 18 months of data. Survival after HT was fitted with exponential distributions at 3 months and 7 years using data from Russo et al. ,100 extrapolated to a lifetime horizon.

The utilities attached to the health states used by Moreno et al. 98 were taken from Sharples et al. 30 and are summarised in Table 35. 30

Cost inputs (Table 36) were based on those of Sharples et al. 30 inflated to 2012 prices with the cost of the HMII device set at £94,200, the remaining cost items were given gamma distributions for the probabilistic sensitivity analysis.

Discounting of costs and benefits used a rate of 3.5% and results were expressed as ICERs calculated as cost per life-year gained (LYG) and cost per QALY.

Results

The base-case results (HT received at 6 months) are summarised in Table 37.

Probabilistic sensitivity analysis of the base case indicated a 50% probability of cost-effectiveness at a willingness to pay of £247,000 per QALY. Other sensitivity analyses used 12- and 18-month intervals before receipt of a donor heart. The deterministic results for these are summarised in Table 38.

Base-case results were also reported for a 10-year time horizon; the ICER was £411,227 per QALY. A further analysis was undertaken in which the HMII was provided free of charge; this reduced the lifetime horizon ICER from £133,860 per QALY to £24,063 per QALY when the waiting time for transplant was set at 18 months, indicating that a major driver in these analyses is the cost of the device.

Authors conclusions

The authors concluded that HMII implantation does not offer better value for money than conventional MM and that it is unlikely to be cost-effective with current cost of the device. For a 50% probability of being cost-effective with 6 month transplant delay and lifetime horizon, a payer would need to be willing to pay about £247,000 per QALY.

Quality assessment and comment

This was a good-quality study using model inputs taken wholly from the published literature (refer to authors for full quality assessment according to the Drummond et al. assessment tool97). The clinical effectiveness studies used for input to the model did not derive from a systematic review of the literature; however, in comparing the studies discussed in the study with those which we identified in our own review of the literature it appears that appropriate and relevant studies were identified.

A weakness of this study that is common to all studies of VADs for BTT for advanced HF is the lack of a randomised comparative study in which an appropriate population of patients is randomised to each treatment strategy. A critical element in these economic studies therefore concerns the choice of the comparator population and its associated prognosis (i.e. survival). In the Moreno et al. 98 analysis 76%, 69% and 63% of conventional care patients survived to 6, 12 and 18 months respectively. This survival is somewhat inferior to that of the total MM population in the BTDB, but considerably superior to that of BTDB ‘inotrope’ patients when their survival is modelled on the robust part of the observed K–M survival plot.

A further noteworthy element of the Moreno et al. 98 study is the allocation of a donor heart with equal probability to both groups of live patients; this contrasts with the analysis of Sharples et al. ,30 in which MM patients received a donor heart with much greater probability than BTT patients. 30 It appears sensible for the purposes of a fair comparison between treatment options that each should have an equal opportunity of receiving benefits of a transplant; however, in medical practice, MM patients do in fact receive transplants much earlier than BTT patients, mainly because it makes little sense to remove a VAD from a patient who is doing relatively well to give them a donor heart which is much more urgently required by other patients. Once the premise that equal opportunity of transplant should prevail for both BTT and MM patients is accepted, then the issue becomes one of ‘At what rate should this be set?’ In the Moreno et al. 98 base case all live patients received a transplant at 6 months; in the BTDB the proportion of live BTT patients who had received a donor heart by 6 months was < 20%. It appears possible that the 18-month delay to transplant used by Moreno et al. 98 in sensitivity analysis is more likely to represent a real-world likelihood of receiving a transplant and may have represented a better choice for the base-case analysis (this reduces the estimated ICER to nearly half).

In the next chapter we describe design of the Warwick Evidence model.

Disease state/pathways

Two pathways were modelled for this economic evaluation of VADs. In the base case, patients with more severe HF (based on inotrope medication) either followed the VAD pathway or were allocated to the MM pathway. In both pathways patients received a HT after a certain period of time (which was varied according to different sensitivity analyses). Some of the patients died before receiving a HT.

Strategies/comparators

For the two research questions we compared:

• use of VADs as a BTT with MM using the inotrope subgroup of patients as the comparator group

• use of VADs as an ATT with use of VADs as a BTT. For an ATT, transition probabilities were kept the same as for patients in the BTT base-case arm, except that the probability of receiving a donor heart was set to zero.

For the sensitivity analyses we included comparisons of:

• use of the HW only, as a BTT with MM using the inotrope subgroup of patients as the comparator group as in the base case

• use of VADs as a BTT with all MM patients (both inotrope and non-inotrope)

• use of VADs as a BTT with an artificially constructed MM group using the VAD patients as their own controls. (Based on predicted survival of the VADs group – had they been treated with MM not VADs. Predictions were made using the SHFM; see Chapter 7, Selection of comparator group and sensitivity analyses.)

Selection of comparator group and sensitivity analyses

As mentioned above, the absence of a RCT in which patients have been randomised to VAD or to MM makes it difficult to select an appropriate MM group to act as a comparator for VAD implantation. Of the 1496 BTDB HT-listed patients who received MM, the poorest prognosis (survival) was associated with those patients who were categorised as ‘inotrope’ (see Figure 45). Without randomised evidence it is uncertain if these are equivalent to the 235 BTDB patients who were selected to receive a VAD as BTT; however, in view of comments in the literature (Slaughter and Rogers116) and from our clinical advisors, it is clear that patients who are selected for BTT are perceived as being less well than the generality of patients who are supported with MM.

For this reason the ‘inotrope’ subgroup was selected for the base case in the present economic analysis. This follows previous analysis by Sharples et al. 30 Furthermore, among the MM BTDB patients only 20% were categorised as ‘inotrope’, whereas among BTDB BTT patients 77% were classified as ‘inotrope’ at baseline; this implies that the inotrope subgroup of MM patients may represent a reasonable comparator group for VAD recipients. Sharples et al. 30 found that ‘inotrope-dependent’ patients in their study were associated with greater costs than their ‘no inotrope’ subgroup (mainly because of hospital stays). Therefore, with minor adjustment, we have used Sharples et al. 30 costs for this subgroup, assessed in the light of information provided by Dr Mark Petrie of GJNH, Glasgow. Clinical experts advised that medications used will have remained similar as the previous analysis (Dr Jayan Parameshwar, Papworth Hospital NHS Foundation Trust, 2012, personal communication).

Several options of choice of MM comparator group were selected for sensitivity analysis:

1. All MM: using the K–M function for all MM patients with a constant hazard fitted to 3 months and from 3 months to the time when 10% of patients remained at risk. The monthly transition probabilities are shown in Table 46 and the resulting extrapolation in Figure 46.

2. Exponential fit to 2 months: using exponential distribution fit to 2 months K–M survival for the ‘inotrope’ patients (see Table 47).

3. Predicted survival using the SHFM: survival based on the SHFM (Levy et al. 95).

In this analysis we constructed an artificial ‘MM’ comparator group using the characteristics of the BTDB VAD patients so that the VADs patients could, in effect, act as their own controls. The assumption underlying this use of SHFM scores calculated from baseline characteristics for patients who go on to receive a VAD is that patients with this score would represent the most appropriate comparator group.

The SHFM score is based on multiple baseline covariates and survival may be predicted according to the following equation (Levy et al. 95):

where t is in years and SC is the mean SHFM for a group of patients. This may be modified to predict monthly survival:

Given an SHFM score it is possible to predict survival in order to be able to compare survival with and without a VAD. An SHFM score was obtained for representative populations of VAD patients in the following three ways:

1. Schaffer: Shaffer et al. 77 reported the frequency of SHFM scores at baseline among all VAD patients at a single centre for the period June 2000 to May 2009. The results reported were used to calculate a mean score of 3.036. The resulting survival curves are shown in Figure 49.

2. Strueber: Strueber et al. 83 presented survival to 24 months for a ‘virtual’ comparator group for patients who were implanted with the HW VAD; the survival estimate was based on the SHFM score. These data fitted the equation above when the score was set at 2.416 and generated the curve shown in Figure 49.

3. BTDB SHFM: some covariates required for the calculation of an SHFM score were available at baseline for VAD patients in the BTDB. These were used to estimate a mean score using a Cox's model by using the probability value of ≤ 0.05. The SHFM score was evaluated from the outcomes of the multivariate model, by using the products of the variables, and their β coefficients (natural log of the hazard ratio) were summed. The resulting score was 3.372. Baseline survival was estimated using the equation above and the resulting curve is shown in Figure 49.

The monthly transition probabilities for each of these sensitivity analyses are provided in Table 49.

FIGURE 49.

Modelled survival using SHFM and REMATCH control group.

Informative censoring

Clinical experts advised that among MM patients, those most likely to receive a donor heart are those with the poorest prognosis. This means that censoring these patients in the analysis of survival under MM may represent informative censoring (i.e. those not transplanted would have better prognosis leading to an over estimate of survival).

The distribution of times between listing for transplant and receiving a transplant for the inotrope MM patients is shown in Figure 50. About one-quarter were transplanted within 2–3 days of being listed, half were transplanted within a week of listing and 86% within 4 weeks.

FIGURE 50.

Distribution of times from listing to transplant for BTDB MM patients receiving inotropes.

This is a short delay to transplant and it seems unlikely to reflect variations in likely survival. However, we further investigated the presence of informative censoring and whether or not it is likely to cause a problem in our analysis by considering the sensitivity analysis approach described by Collet117 in which patients censored for transplant were (a) assumed instead to die 1 day after their date of transplant; and (b) assumed instead to die at the last event time for uncensored patients. The resulting K–M plots exhibited median survival times of 0.39 months and 28 months, respectively, clearly indicating a large effect on survival estimates if censoring times are equated to predicted survival (Figure 51).

The median survival in graph Figure 51a mostly reflects the short time between being listed for transplant and receipt of a transplant for those patients transplanted. Although this could be indicative of informative censoring, this could, however, also be partly due to delayed listing of eligible patients so that recorded times between listing and transplantation were too close. Thus, the deviation between the two K–M curves seen is unlikely to be caused by poor prognosis of transplanted patients only, but may also be caused by delayed entry onto the transplant list; this might be interpreted as lead-time bias.

FIGURE 51.

Analysis of inotrope MM patients' survival using modified censoring according to Collet. (a) HT recipients assumed instead to die 1-day post-HT date; and (b) HT recipients assumed instead to die at time of last observed event.

In view of a possible overestimate of survival for the MM group we followed advice from clinical experts to explore the use of the ‘optimum MM’ group from the REMATCH RCT (Rose et al. 45) as our control group. In the REMATCH trial the median survival for optimum MM patients (for whom HT was contraindicated) was 150 days (4.93 months), the data were mature (54 deaths among 61 patients), most were receiving inotropes at baseline, and there was little censoring. Using this median survival value we calculated the exponent (lambda) for a constant hazard fit: λ = (ln 0.5)/4.93 = 0.14065; the fit to observed data is shown in Figure 52 and the corresponding monthly TP of 0.131207633 was used in our economic model for sensitivity analysis. Table 49 and Figure 52 show the survival curve and TP together with other sensitivity analyses.

FIGURE 52.

Constant hazard fit to REMATCH control group based on median survival of 150 days.

Stevenson et al. 46 performed a post-hoc analysis of survival data from the REMATCH trial. Patients were stratified by baseline treatment with inotropes. Among MM patients, survival was poorer for those receiving inotropes; median survival was 120 days (3.94 months). When fitted with a constant hazard model this provided a monthly TP of 0.161217 which was also used for sensitivity analysis in our economic model.

Resource use and associated costs are required for the following health states

1. Support on BTT with VAD until HT.

2. Support on MM until HT.

3. Support on HT.

States 1 and 3 have two phases:

1. a short-term phase associated with preparation for surgery and the immediate aftermath of surgery

2. a prolonged phase of maintenance.

Base case

Our main source of information on cost and resource use was from a previous HTA report undertaken by Sharples et al. 30 In the Sharples et al. 30 study, patient-specific resource-use data were collected for VAD implant, HT, and patients on MM while awaiting transplant. For all three patient groups, costs were collected as monthly costs from the date of VAD implantation, or the date the patients were accepted on to the transplant WL, until the study cut–off date. Post-transplantation and VAD monthly cost were assumed to be constant from month 7 onwards in all groups. Importantly, the study recorded actual costs incurred by patients and, hence, provides a more accurate representation of costs. For our model, these costs were inflated to current levels by applying the projected health services costs. 120 VAD costs were obtained from five UK centres operating under the auspices of the National Specialist Commissioning Team (NSCT), which commissions the VAD programme. We did not include the cost of the VAD supplied by GJNH which was based on a single VAD used at this centre; however, this was used in scenario analysis. The base-case cost inputs for the three health states are summarised in Table 55. It is worth mentioning that Sharples et al. 30 estimated resource use from the perspective of the NHS and centre. In such settings, the cost does not include setting-up costs for a new centre and we would emphasise that setting up a new service would incur additional set-up costs.

State 1: Support on bridge to transplant with a ventricular assist device until heart transplant

This health state consists of an early short-term phase associated with the VAD implant procedure cost which includes the cost of the device, theatre cost and cost of immediate post-operative hospital stay, and a second long-term follow-up phase that includes the costs of outpatient visits, adverse events and rehospitalisation.

We obtained costs of VADs from five centres (listed below) providing long-term VAD support:

• NUT

• Papworth Hospital NHS Foundation Trust

• RB

• UNB

• UHSM.

The base-case cost per VAD (£80,569) was a weighted value according to the number of devices used by BTDB BTT patients and costs of different devices. Details of unit cost of VADs used in this study are given in Table 56.

State 2: Support on medical management until heart transplant

The input required is an estimate of the average monthly cost while patients are medically managed. This includes medication, such as inotropes, and follow-up assessment as inpatient or out-patient visits.

An inotrope-dependent patient subgroup of BTDB MM patients was selected for the base-case analysis. We consulted our clinical advisors and they advised that medications and inotropes used will have remained similar as the previous analysis. 30 The intravenous inotropes used were enoximone 5 μg/kg/minute and dopamine 5 μg/kg/minute. We inflated the cost of inotrope-dependent patients' medications to 2011 prices by applying the projected Health Services Cost Index. 120 The resulting monthly costs with distribution parameters where appropriate are shown in Table 55.

State 3: Support on heart transplant

The cost inputs include average presurgery preparatory cost, procedural cost and short-term post-surgery cost. The transplantation procedure cost was considered to be different between groups to address the increase in theatre time for VAD explant. Post-transplant monthly costs were assumed to be the same for both groups from month 2 onwards. Post-HT support costs include follow-up outpatient visits, investigation, blood test and drugs (see Table 55).

Clinical experts advised that the costs of the transplant donor procedure were trivial and we therefore did not include this cost in our model.

Sensitivity analyses around cost inputs

A sensitivity analysis (I A1] in Table 54) was conducted around the TP for MM to death which assumed the MM group was constituted of all BTDB MM patients (both 307 inotrope patients and 1189 non-inotrope patients). The monthly cost for these is shown in Table 58.

The monthly cost of all MM was a weighted value according to the number of both inotrope and non-inotrope patients from the BTDB. We used previously reported costs30 for inotrope- and non-inotrope-supported MM patients, inflated to 2010/11 prices for the sensitivity analysis Table 59.

Sensitivity analyses II A and II B around cost inputs

In univariate sensitivity analyses, the cost of the VAD was varied from that in the base case as shown in Table 60. This analysis is designated II A in the results section.

In a further sensitivity analysis (designated II B) the cost of patient maintenance on a VAD was decreased from base case by 30%.

This was undertaken because clinical experts advised that patients on second- and third-generation VADs experience relatively fewer adverse events than those supported with earlier VAD designs. To address the potential cost savings of reduced adverse events, we lowered the monthly post-VAD implantation cost by 30% in sensitivity analysis.

Sensitivity analyses (II C and II D) around costs for both arm using Golden Jubilee National Hospital and national schedule of reference costs data

Further sensitivity analyses used a detailed list of resource use and associated costs which were supplied from Glasgow the Glasgow centre, the GJNH, one of the UK centres operating with the NSCT (analysis II C). The GJNH cost data and definitions used are all presented exactly as provided by GJNH. This provided information on all three health states. In addition, we also sourced the mean cost for three health states from the national schedule of reference costs (NSRC) 2010/11 (analysis II D). 122 These alternative sources are described in further detail below.

Support on medical management until heart transplant

Resource-use data from the Glasgow GJNH finance department were collected on hospital stay, drugs, investigations and outpatient visits for patients with advanced HF. The GJNH finance department reported a total cost of £10,111.66 per hospital stay per patient with advanced HF for the year 2009/10, based on a total of 92 patients costing £842.63 (£10,111.66/12) per patient in month 1. Further detailed information on the resource use and costs supplied by GJNH involved in managing patients in hospital with advanced HF is shown in Table 61.

From the cost provided by the GJNH finance department, it was not possible to identify the drugs and inotropes used in hospital at the time of admission. We assumed that patients admitted to hospital for advanced HF were inotrope-dependent and patients at home on the WL were non-inotrope-dependent (Dr Mark Petrie, GJNH, 2012, personal communication).

Patients at home and on the WL were seen in the HF clinic every third month. The cost per initial visit was £437.79, and after 3 months patients are reassessed by a multidisciplinary team consisting of a cardiologist, a cardiac surgeon and a specialist nurse. The cost for every other consecutive follow-up was £160.11 per visit.

We also sourced the mean cost of MM (from the NSRC) 2010/11 and these were explored in the sensitivity analysis (see Table 66).

Support on bridge to transplant with a ventricular assist device until heart transplant

A detailed description of resource-use data and unit cost estimates were collected for VAD implant from the GJNH finance department. Resource-use data were collected on VAD assessment cost, implant procedure cost, cost associated with ward and ICU stay, follow-up outpatient visits, investigation, blood test and drugs. The cost was based on one long-term VAD (HMII) patient for 2009/10. We recommend caution in interpreting this result partly because of this and partly as we believe that the cost from the GJNH might be overestimated (as throughput for this intervention is at present insufficient for economies of scale to be in evidence). However, the GJNH data give us details of the cost component for immediate post-operative hospital stay following VAD implant, and are shown in Table 62.