Management of end-stage heart failure patients with or without ventricular assist device: an observational comparison of clinical and economic outcomes^†

Abstract

OBJECTIVES

Heart transplantation (HT) and ventricular assist devices (VAD) for the management of end-stage heart failure have not been directly compared. We compare the outcomes and use of resources with these 2 strategies in 2 European countries with different allocation systems.

METHODS

We studied 83 patients managed by VAD as the first option in Bad Oeynhausen, Germany (Group I) and 141 managed with either HT or medical therapy, as the first option, in Paris, France (Group II). The primary end-point was 2-year survival. Kaplan–Meier analyses were performed after the application of propensity score weights to mitigate the effects of non-random group assignment. The secondary end-points were resource utilization and costs. Subgroup analyses were performed for patients undergoing HT and patients treated with inotropes at the enrolment time.

RESULTS

The Group I patients were more severely ill and haemodynamically compromised, and 28% subsequently underwent HT vs 55% primary HT in Group II, P < 0.001. The adjusted probability of survival was 44% in Group I vs 70% in Group II, P <0.0001. The mean cumulated 2-year costs were €281 361 ± 156 223 in Group I and €47 638 ± 35 061 in Group II, P < 0.0001. Among patients who underwent HT, the adjusted probability of survival in Group I (n = 23) versus Group II (n = 78) was 76% versus 68%, respectively (0.09), though it differed in the inotrope-treated subgroups (77% in Group I vs 67% in Group II, P = 0.04).

CONCLUSIONS

HT should remain the first option for end-stage heart failure patients, associated with improved outcomes and better cost–effectiveness profile. VAD devices represent an option when transplant is not possible or when patient presentation is not optimal.

INTRODUCTION

Heart failure (HF) is a global pandemic disease affecting an estimated 26 million people worldwide, prompting over 10 × 10⁶ hospitalizations annually in the USA and Europe [1]. Slightly less than 10% of patients suffer from advanced HF, ranging from New York Heart Association functional Class IV to cardiogenic shock, despite optimal management [2]. End-stage (ES) HF is associated with major concomitant disorders, loss of autonomy, high mortality and high societal costs [1, 2]. For these patients, heart transplantation (HT) or ventricular assist devices (VAD) are the most successful and most often recommended treatments [2, 3].

Although HT is a highly effective therapy for advanced, refractory HF, it is limited to <10% of candidates due to a severe shortage of donor organs and a variety of contraindications [3, 4]. This paucity of effective therapy promoted the development of VAD, which may be used as (i) bridges to HT or (ii) a long-term alternative to HT, also known as destination therapy [4–8]. HT and VAD have rarely been compared [9, 10], and the cost–effectiveness of VAD therapy remains to be confirmed [11], particularly when considering the considerable technical progress, lower morbidity and higher survival rates in the last decade [3–10].

Germany and France allocate donor hearts differently. In the large cardiac surgical Heart and Diabetes Centre of Bad Oeynhausen (BO), Germany, the average waiting period for a highly emergent HT is 3.0 months, and very few patients undergo transplantation without prior listing as an emergency. Patients presenting with end-stage heart failure (ESHF) are preferentially managed by VAD therapy, particularly if unstable and inotrope dependent [4, 10, 12]. In contrast, in major French surgical heart centres, the average waiting period for highly emergent HT is 0.3 months, and HT is the first-line treatment of ESHF [13].

These differences between the German and French management strategies for ESHF have not been studied in a real-world setting and thoroughly compared, particularly with respect to the respective benefits they confer to patients and clinical outcomes. Therefore, the main objective of this study was to compare the outcomes of these 2 strategies in the management of ESHF and to analyse consumption of resources for each strategy.

METHODS

Study organization and design

The 3 specialized university hospitals participating in this observational study include the Heart and Diabetes Centre in BO, Germany, and the Groupe Hospitalier Pitié-Salpêtrière and Hôpital Européen Georges Pompidou in Paris, France. This study design allowed a comparison of the results of management strategies currently in use in Germany versus France in the care of patients with ESHF. We were able to consecutively enrol all eligible patients, with a high acceptance of the protocol by the physicians and patients. We used propensity score-weighting techniques to compensate for the non-random assignment of the treatment strategies.

This observational study was in compliance with the principles outlined in the Declaration of Helsinki, and its protocol was reviewed and approved by the institutional review board of each participating centre. According to the French regulation, the project was approved by the national data protection authority (Commission National Informatique et Liberté, agreement number 1165361).

Study populations

All patients presenting with ESHF between November 2010 and October 2011 at the Heart and Diabetes Centre of BO and who underwent VAD implantations were included in Group I. All patients presenting with ESHF to the Hôpital Européen Georges Pompidou or the Groupe Hospitalier Pitié-Salpêtrière during that same period were included in Group II. ESHF was defined as severe depressed left ventricular (LV) function (LV ejection fraction <25%) and New York Heart Association functional Class III or IV or cardiogenic shock, refractory to optimal drugs and other therapies [2]. Patients presenting with severe, irreversible pulmonary, hepatic dysfunction or active infections were excluded from the study.

Patient management

Medical therapy

In both the groups, the patients were treated with inotropes for clinical or haemodynamic manifestations of circulatory failure [2].

Device therapy

In candidates for HT, short-term, veno-arterial extracorporeal membrane oxygenation (ECMO) was delivered via a biventricular assist device implanted mainly for cardiogenic shock with severe end-organ dysfunction, despite high doses of inotropes [3]. The VAD used for long-term treatment in BO were the HeartMate^® II (Thoratec, Pleasanton, CA, USA), the HeartWare^® (HeartWare International Inc., Framingham, MA, USA), the Paracorporeal Biventricular Assist Device™ (Thoratec) or the CardioWest™ (Syncardia, Tucson, AZ, USA) total artificial heart [3, 5, 8].

Heart transplantation

Indications and contraindications for HT were similar in both countries and adhered to the current guidelines [2]. We can summarize the indications for high urgency (HU) listing in both countries by 2 situations: (i) the first situation as ‘patient presenting clinical and biological signs of cardiogenic shock while on inotrope or under ECMO’ and (ii) the second situation as ‘complications on long-term VAD’ [12, 13]. The basic criteria for HU listing are reported in Supplementary Material, Appendix 1 [10, 12, 13].

Procedures according to countries

In Germany, the median waiting time for all HUHT during the inclusion period (2010–11) was 3.0 months (0.07–31.8). One patient underwent HT without HU listing. Unstable inotrope-dependent ESHF patients were preferentially managed with long-term VAD as bridge to transplant. ECMO was used in patients having refractory cardiogenic shock, as bridge to decision or bridge to recovery or when the patients were very gravely ill and haemodynamically compromised. Indications of long-term VAD are reported in Supplementary Material, Appendix 2.

In France, during the same time period, HT was the first-line therapy for the management of ESHF, and the median waiting time was 0.3 months (0.1–2.4) for HUHT versus 6.7 months (2.1–19.6) in the absence of urgent listing [13]. For HT candidates in refractory cardiogenic shock, veno-arterial ECMO was used as the first-line therapy, as a bridge to transplant. HU listing was reserved for patients who could not be weaned from inotropes or from veno-arterial ECMO. ECMO was replaced by a long-term VAD in 4 patients (2.7%) while waiting for a suitable donor heart or for the resolution of temporary contraindications to HT.

Study end-points

The primary study end-point was survival at 2 years. We used a Cox proportional hazards regression to estimate the relative risks and 95% confidence intervals (CIs) and to adjust for differences in baseline predictors of outcome. Kaplan–Meier analyses were used to estimate 2-year survival in the 2 groups, before and after application of propensity score weights and compared using log-rank statistics.

The secondary study end-points were (i) resource consumption, (ii) costs and (iii) cost versus survival of the treatment strategies up to 2 years of follow-up. The costs were estimated from a payer’s perspective. Only hospital costs were included. Patient-related information was retrieved from the hospitals’ information systems, and admission records were linked over the duration of the follow-up period. In both countries, itemized patient billing was available. The number of days spent in the intensive care units, the devices used and the use of life-sustaining interventions were recorded. In Germany, every hospital resource was recorded and billed. In France, the billing was based on the patients’ Diagnosis-Related Group approximation of hospital costs, including outpatient days and manufacturer’s billing for devices. We added post-transplant outpatient costs. We calculated the cost per additional month of life (cost–effectiveness ratio) for the patients who underwent HT. Bootstrap methods were used to examine the distribution of the incremental cost and incremental survival across the cost–effectiveness plane. All costs are expressed in 2013 Euros. Additional secondary end-points included adverse events (Supplementary Material, Appendix 3) and unplanned hospitalizations at 2 years [14].

Statistical analysis

We used a propensity score weighting to minimize treatment selection biases and simulate a randomization scheme [15, 16]. Propensity scores were estimated, using a stepwise logistic regression model based on variables with high prognostic value, including age, gender, ischaemic heart disease, haemoglobin, serum creatinine and new simplified acute physiology score (SAPS II). Propensity scores represented the likelihood of undergoing VAD treatment based on the cofactors observed a priori. This probability was used to assign each patient a weight depending on their characteristics, to create 2 study groups as similar as possible at the time of enrolment (Supplementary Material, Appendix 4). Subgroup analyses were performed for patients undergoing HT and patients treated with inotropes at the time of study enrolment. Demographic, clinical, biological and imaging data were recorded at the time of study entry in each study group. Details of the scores used are reported in the Supplementary Material, Appendix 5. Categorical variables are expressed as counts and percentages, and continuous variables as means ± standard deviation or medians (interquartile ranges), as appropriate. Between-group differences were examined by the χ² test or the Fisher’s exact test for categorical variables, and by the Student’s t-test or the Wilcoxon test for continuous variables, as appropriate. Censored data were compared by log-rank tests. Variables emerging at a P-value <0.2 by a single-variable analysis were entered in a multiple variable analysis in search of independent predictors of clinical outcomes.

The statistical analyses were performed using the SAS software (SAS Institute Inc., Cary, NC, USA). All P-values are 2-sided and considered to indicate statistical significance when <0.05.

RESULTS

Baseline characteristics

Between November 2010 and October 2011, 83 patients were enrolled in Group I (57 left VAD, 13 paracorporeal BiVAD and 13 total artificial hearts) and 141 patients in Group II (Supplementary Material, Appendix 6). The baseline characteristics of the patients are listed in Table 1. The main baseline characteristics of the 2 treatment groups were similar, including age, gender, cardiovascular risk factors and long-term, non-pharmacologic treatments. However, at inclusion, more patients in Group I (i) presented with a <3 INTERMACS profile, (ii) were hypotensive and hyponatraemic and (iii) required inotrope therapy or ultrafiltration (Table 1).

Baseline characteristics after the weighting procedure are reported in Supplementary Material, Appendix 7.

Clinical outcomes

At 2 years of follow-up, 23 patients in Group I (28%) underwent HT vs 78 patients (55%) in Group II (P < 0.001). All patients in Group I (100%) underwent HT on a high emergency basis (HU listing), compared with 42 patients (53.8%) in Group II (P < 0.001). The median waiting time before HT was 3.0 months (2.3–5.7) in Group I vs 1.6 months (0.2–5.3) in Group II (P < 0.001). Among all candidates, 32 patients died before HT in Group I (45%), vs 21 patients (15%) in Group II (P < 0.05).

Primary end-point

The adjusted probability of 2-year survival was 44% in Group I (95% CI 36–53) vs 70% (95% CI 66–74) in Group II (P < 0.0001). Figure 1 shows 2-year survival in both study groups, before and after adjustment by propensity score weights.

Secondary end-points

Resource utilization and costs

Resource consumption and costs at 1 and 2 years are listed in Table 2. The mean cumulated 2-year costs were €281 361 ± 156 223 in Group I and €47 638 ± 35 061 in Group II (P < 0.0001). The main expenditures incurred in the first year were the (i) VAD implanted in 100% of Group I, (ii) 100 ± 58 vs 30 ± 37 days hospitalization and (iii) €1880 vs €808 per diem costs for the intensive care unit, in Group I versus Group II, respectively (Table 2).

Adverse events

The overall rates of (i) adverse events and (ii) rehospitalizations were significantly lower (P < 0.001 for both comparisons) in Group I than in Group II (Supplementary Material, Appendix 8).

Post hoc analyses

Heart transplant recipients

The baseline characteristics of the 101 HT recipients are shown in Supplementary Material, Appendix 9. The adjusted probability of survival at 2 years was 76% in Group I (95% CI 58–87) vs 68% in Group II (95% CI 62–72; P = 0.09, Fig. 2).

Among the transplanted patients, the higher 2-year costs in Group I were associated with a longer survival (Table 2), with an estimated cost–effectiveness ratio of €156 527 per added month of survival (Fig. 3).

Recipients of inotropic drugs

Among the 149 recipients of inotropes at the time of study inclusion, the 2-year adjusted probability of survival was 44% in Group I (95% CI 40–58) vs 62% in Group II (95% CI 56–67; P = 0.11, Fig. 4).

Recipients of inotropic drugs and heart transplantation

Among the 69 patients treated with inotropes at the time of study inclusion who underwent HT, the 2-year adjusted probability of survival was 77% in Group I (95% CI 72–90) vs 67% in Group II (95% CI 58–69; P = 0.04, Supplementary Material, Appendix 10).

DISCUSSION

Our study compared 2 strategies applied in the management of ESHF in a real-world setting in 2 Western European countries and how medical systems actually adapt to differences in heart donor availability and allocation policies. Compared with Group II, the Group I patients were more severely ill and haemodynamically compromised, had lower survival rates, less often underwent HT and waited longer before HT. In a subgroup analysis, HT after VAD therapy yielded results similar to HT non-preceded by VAD and better results in patients treated with inotropes at the time of study inclusion. The strategy adopted in Group I was more expensive, especially during the first year.

HT was associated with a higher 2-year survival than primary VAD therapy. The greater than 70% 2-year survival in Group II is similar to that reported by the French Agence Nationale de Biomedicine and by the Organ Procurement and Transplantation Network [13, 18]. The 2-year survival in Group I is lower than the approximately 80% survival of VAD-treated patients in recent studies [5–7, 10]. This comparatively low survival rate must be put in perspective granting the frequency of patients receiving biventricular assist devices or total artificial hearts in Group I (nearly one-third of the population). These devices are associated with a >50% 6-month mortality, distinctly higher than what can be observed with LVAD [19, 20]. Furthermore, some patients were severely haemodynamically compromised, with deteriorating end-organ function (INTERMACS 1 and 2) at the time of enrolment in the study. Earlier studies of the outcomes of VAD recipients presenting with INTERMAC levels 1 and 2 have reported survival rates of approximately 40%, lower than that observed in our study [9, 21, 22].

In the subgroup analysis of patients treated with inotropes at the time of enrolment in the study, 2-year survival was similar in the 2 groups. If they subsequently underwent HT (i.e. has received a VAD in first intention), the 2-year survival was higher in Group I. These observations are concordant with previous studies, which suggested that transition from inotropic therapy to VAD implantation might increase post-transplant survival [9, 17, 21–24]. In a systematic review of 31 studies including over 19 500 patients, Alba et al. [24] found that VAD did not worsen post-transplant outcomes but did not document improved survival either.

The median waiting time for HT was longer in Group I than in Group II, and more Group II patients underwent HT. The management strategies for ESHF studied here resulted from different allocation systems in each country and, consequently, different waiting times to HT, both under usual circumstances and in highly emergent situations [4,12, 13]. In reality, the possibility of allocation not only is dependent on the allocation system but, more importantly, also depends on the number of patients on the waiting list and the number of transplantations performed. In 2010 and 2011, in France, 1566 patients were on the HT list and 754 were actually transplanted [13]. During the same period in Germany, 1921 patients were on the HT list and 710 were transplanted [12]. One hundred thirty-eight patients died while on the HT list in France, compared with 387 patients in the same situation in Germany [12, 13]. Of note, heart donation rates are higher in France (6.5 vs 3.8 per million population in Germany) [12, 13].

A waiting time of more than 3.0 months in advanced HF patients being treated with intropes is a clinically uncomfortable, and often unrealistic, situation. Some patients in Germany are transplanted in high urgency status without prior VAD, when they are thought able to tolerate such a long waiting period [10]. However, when end-organ function deteriorates on continued medical management, the risk becomes truly huge if other dispositions are not taken.

Some studies reported that patients receiving intravenous inotropic support while awaiting transplantation for more than 21 days have high mortality before HT and a >50% post-transplant mortality [21, 22]. We think that in these inotrope-dependent ESHF patients, LVAD is a reliable option. It allows recovery of end-organ perfusion and function and a better nutritional status, although specific complications can occur. The German allocation policy does not permit urgent HT for patients on VAD unless life-threatening device-related complications occur. Thus, some VAD patients can die before HT due to complications related to their assist devices [18, 25]. This has obvious ethical implications, in the light of waiting times of months and more, with clinicians in centres such as BO facing a serious circular dilemma in their decision making, which should be addressed by regulators [10]: either implant VAD and impede the chance of HT or wait for HT and decrease the chance of survival. In addition, there have been major advances in this field of VAD, with a significant improvement in device outcomes [6, 7, 18, 24]. In the USA, VAD patients constitute more than one-third of all listed adult HT candidates [26].

What are the overall lessons for France and other countries with a low use of VAD?

Our findings provide evidence that in ESHF patients dependent on inotropes, LVAD is a reliable option. The ideal situation would be to transplant the VAD patients after recovery (end-organ perfusion and nutritional status) and before complications develop. We also think that implanting LVAD to ESHF patients can allow to keep donor hearts for others patients with biventricular heart failure for which BiVAD or total artificial heart decrease survival after HT [19, 20, 26]. Such a management strategy would allow better distribution of the limited resources in terms of donor hearts.

Finally, VAD therapy appears to be expensive and hardly cost–effective according to published thresholds [27]. This is not surprising in the light of the device cost in BO patients, a significantly higher hospitalization rate, as well as higher unit costs in Germany. This is further emphasized by the fact that per capita health expenditure in Germany is higher than in France (€4881 vs €4288) [17], which appears to be associated with higher overall Diagnosis-Related Group tariff levels [28]. At the same time, French costs for patients needing cardiac therapy/care have been found to be vastly lower than for US patients [29], whereas the costs of VADs patients managed in BO are comparable with those managed in the USA [11, 30]. Furthermore, BO-transplanted patients presented fewer adverse events and rehospitalizations than APHP patients. Finally, future estimates of the cost–effectiveness ratio should take into account the potential quality-of-life benefits.

Limitations and strengths

Our emulation of randomization, using propensity scores, was based on a set of observable variables. However, other factors may have contributed to the measured outcome that were not accounted for, including selection effects by patients and caregivers in the highly specialized treatment centre of BO. Second, our comparisons of costs ignore the differences in care pathways and provision of care between countries. In this perspective, the differences in rates of repeat hospitalizations due to different referral patterns in France versus Germany need to be underscored. Despite these limitations inherent to a non-randomized, between-country study design, we believe that our analysis allowed assessments, which can only be made using the ‘natural experimental’ design provided by distinct regulations and traditions of care in the countries involved.

CONCLUSION

HT continues to be the first option for ESHF, associated with better outcomes and better cost–effectiveness profile. VAD devices nevertheless constitute possible options when transplant is not possible or when the patients are not in optimal clinical conditions.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

ACKNOWLEDGEMENTS

We thank all the study coordinators in the Heart and Diabetes Centre, in the Groupe Hospitalier Pitié-Salpêtrière and in Hôpital Européen Georges Pompidou, for their assistance in the data collection. We also thank Christian Piek, Annette Fihlon, Catherine Amrein, Stefan Wlost, Christelle Cantrelle and Richard Dorent for their assistance in the data collection and Rodolphe Ruffy for his review of our article for style and language.

Conflict of interest: none declared.

REFERENCES

Author notes