Download European Journal of Heart Failure

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Baker Heart and Diabetes Institute wikipedia , lookup

Electrocardiography wikipedia , lookup

Coronary artery disease wikipedia , lookup

Management of acute coronary syndrome wikipedia , lookup

Arrhythmogenic right ventricular dysplasia wikipedia , lookup

Saturated fat and cardiovascular disease wikipedia , lookup

Remote ischemic conditioning wikipedia , lookup

Heart failure wikipedia , lookup

Antihypertensive drug wikipedia , lookup

Cardiac surgery wikipedia , lookup

Cardiovascular disease wikipedia , lookup

Cardiac contractility modulation wikipedia , lookup

Quantium Medical Cardiac Output wikipedia , lookup

Transcript
1
Mechanisms underlying increased mortality risk in patients with heart failure with
reduced ejection fraction randomized to adaptive servo-ventilation in the SERVE-HF
study: results of multistate modelling
Christine Eulenburg, Karl Wegscheider, Holger Woehrle, Christiane Angermann, Marie-Pia
d’Ortho, Erland Erdmann, Patrick Levy, Anita K. Simonds, Virend K. Somers, Faiez Zannad,
Helmut Teschler, Martin R Cowie
Department of Medical Biometry and Epidemiology, University Medical Center Eppendorf,
Hamburg, Germany (Dr C Eulenburg, Prof. K Wegscheider); Department for Epidemiology,
University Medical Center Groningen, Groningen, The Netherlands (Dr C Eulenburg);
ResMed Science Center, ResMed Germany Inc., Martinsried, Germany (H Woehrle); Sleep
and Ventilation Center Blaubeuren, Respiratory Center Ulm, Ulm, Germany (H Woehrle);
Department of Medicine I and Comprehensive Heart Failure Center, University Hospital and
University of Würzburg, Würzburg, Germany (Prof. C Angermann); University Paris Diderot,
Sorbonne Paris Cité, Hôpital Bichat, Explorations Fonctionnelles, DHU FIRE, AP-HP, Paris,
France (Prof. M-P d’Ortho); Heart Center, University of Cologne, Cologne, Germany (Dr E
Erdmann); CHU de Grenoble, Grenoble, France (Prof. P Levy); Royal Brompton Hospital,
London, United Kingdom (Prof. A Simonds); Mayo Clinic and Mayo Foundation, Rochester,
Minnesota, USA (Prof. V K Somers); Inserm, Université de Lorraine, CHU Nancy, France
(Prof. F Zannad); Department of Pneumology, Ruhrlandklinik, West German Lung Center,
University Hospital Essen, University Duisburg-Essen, Essen, Germany (Prof. H Teschler);
Imperial College London, London, United Kingdom (Prof. M Cowie)
Correspondence to:
1
2
Dr Christine Eulenburg, Department for Epidemiology, University Medical Center
Groningen, Groningen, The Netherlands; Tel: +31 50 361 0937; Fax: +31 50 361 4493; Email: [email protected]
2
3
ABSTRACT
Background A large randomized treatment trial (SERVE-HF) showed that treatment of
central sleep apnoea (CSA) with adaptive servo-ventilation (ASV) in patients with heart
failure (HF) and reduced ejection fraction (HFrEF) increases mortality, although the analysis
of the primary endpoint, which was the time to first event of death from any cause, lifesaving
cardiovascular intervention, or unplanned hospitalisation for worsening HF, was neutral. This
multistate model analysis of SERVE-HF investigated associations between ASV and
individual study primary endpoint components to try to better understand mechanisms
underlying the increased mortality.
Methods Individual components of the primary SERVE-HF endpoint were analysed
separately in a multistate model investigating individual endpoints separately, before and after
adjustment for potential confounding factors (ICD at baseline, LVEF, proportion of CheyneStokes’ Respiration (CSR)).
Findings Univariate analysis showed an increased risk of both cardiovascular death without
prior hospitalisation (hazard ratio [HR] 2·59, 95% confidence interval [CI] 1·54–4·37;
p<0·001) and cardiovascular death after a life-saving event (HR 1·57, 95% CI 1·01–2·44;
p=0·045) in the ASV versus control group. On adjusted analysis, there was a significant
interaction between ASV treatment and LVEF for cardiovascular death without prior
hospitalisation and hospitalisation for worsening HF, and between the CSR pattern and
hospitalisation for worsening HF. When LVEF was ≤30%, ASV markedly increased the risk
of cardiovascular death without prior hospitalisation (HR 5·21, 95% CI 2·11–12·89).
Interpretation ASV has been shown to be associated with an increased risk of cardiovascular
death. This multistate model analysis shows that this risk is elevated for cardiovascular death
without prior hospitalisation, presumably sudden death. This risk is higher in patients with
poorer left ventricular function.
3
4
Funding ResMed.
Keywords: heart failure; sleep-disordered breathing; central sleep apnea; Cheyne-Stokes’
respiration; adaptive servo-ventilation; cardiovascular death
Research in context
Evidence before this study
The primary intention-to-treat analysis of the Treatment of Sleep-Disordered Breathing With
Predominant Central Sleep Apnea by Adaptive Servo-Ventilation in Patients With Heart
Failure (SERVE-HF) was neutral for the primary endpoint (time to first event of death from
any cause, lifesaving cardiovascular intervention [cardiac transplantation, implantation of a
ventricular assist device, resuscitation after sudden cardiac arrest, or appropriate lifesaving
shock], or unplanned hospitalisation for worsening heart failure) but showed a significant
increase in all-cause, and cardiovascular, mortality in the adaptive servo ventilation (ASV)
versus control group. However, mechanisms underlying this increase in mortality risk and
outcomes in specific patient subgroups remain unclear.
Added value of this study
This multistate model analysis of SERVE-HF data allows the effects of ASV therapy on
different elements of the study’s composite endpoint to be determined, providing insight into
potential mechanisms underlying the increased risk of cardiovascular death in the ASV group,
and defining a subgroup of heart failure patients with more severely reduced ejection fraction
with predominant central sleep apnoea (CSA) who do particularly badly when ASV is added
to optimal medial therapy.
Implications of all the available evidence
ASV therapy should not be used in systolic heart failure patients with predominant CSA due
to the increased mortality risk. This risk is due to sudden death, presumably due to
4
5
arrhythmia, and is particularly marked in those with the lowest ejection fraction. The
pathophysiological mechanism of this effect remains to be elucidated.
5
6
Introduction
The Treatment of Sleep-Disordered Breathing With Predominant Central Sleep Apnea by
Adaptive Servo Ventilation in Patients With Heart Failure (SERVE-HF) study investigated
the effect of adaptive servo ventilation (ASV) added to optimal medical therapy on outcomes
in patients with heart failure and reduced ejection fraction (HFrEF) and predominant central
sleep apnoea (CSA).1 The main study findings were neutral with respect to the primary
endpoint, which was the time to first event of death from any cause, lifesaving cardiovascular
intervention (cardiac transplantation, implantation of a ventricular assist device, resuscitation
after sudden cardiac arrest, or appropriate lifesaving shock), or unplanned hospitalisation for
worsening heart failure. However, all-cause mortality (and in particular cardiovascular
mortality) was significantly increased in the ASV compared with control group.1 Previous
studies, although small and often uncontrolled, had suggested that ASV should have
beneficial effects on the individual components of the SERVE-HF primary endpoint.2-6
The unexpected finding of increased mortality risk in the ASV group in SERVE-HF could
bring into question the appropriateness of the composite primary endpoint used in this study.
Also, associations between ASV and the individual components of the composite endpoint
may reveal important information about potential mechanisms for the excess mortality seen in
the ASV group.
Multistate modelling is a methodological approach for the statistical analysis of multiple
endpoints and their relationships.7, 8 Multistate models can provide examples of disease
progression that include multiple potential endpoints, each of which may affect the occurrence
probability of another endpoint. Such multistate analyses can accurately describe the
transitions between disease states (defined by the potential endpoints), and the associations
with potential risk factors.
This analysis of SERVE-HF used multistate modelling to investigate associations between the
randomised allocation to ASV and the individual components of the study’s endpoints in
6
7
order to better understand the mechanisms underlying the increased mortality observed with
ASV therapy in this population.
Methods
Study design and participants
SERVE-HF was a multinational, multicentre, randomised, parallel-group, event-driven study.
Full details of the study design have been reported previously.1, 9 The SERVE-HF study
protocol was approved by the ethics committee at each participating centre. The trial was
conducted according to Good Clinical Practice and the Principles of the Declaration of
Helsinki. All participants gave written informed consent.
Randomisation and masking
In the SERVE-HF study, patients were randomized in a 1:1 ratio to the ASV or control group.
Randomization was performed using codes generated by a central computer. The study had an
open-label design because of the practical, scientific and ethical issues, and problems with
investigator blinding associated with delivery of sham positive airway pressure therapy.10
Procedures
SERVE-HF participants were randomised to receive optimal medical therapy for heart failure
alone, or in combination with ASV (Auto Set CS, ResMed). For full details of ASV titration
and settings, please see the primary publication.1
Outcomes
In this multistate analysis, individual components of the primary endpoint of the SERVE-HF
study (the first event of death from any cause, lifesaving cardiovascular intervention [cardiac
7
8
transplantation, implantation of a ventricular assist device, resuscitation after sudden cardiac
arrest, or appropriate lifesaving shock], or unplanned hospitalisation for worsening HF) were
investigated separately as competing events. Thus, outcomes analysed were hospitalisation for
worsening heart failure as a first event, one of the life-saving events as a first event (cardiac
transplantation, implantation of a ventricular assist device, resuscitation after sudden cardiac
arrest, or appropriate lifesaving shock), cardiovascular death without prior hospitalisation for
worsening heart failure or life-saving event, non-sudden cardiovascular death, sudden noncardiovascular death and non-sudden non-cardiovascular death, see Figure 1. Death was
classified as cardiovascular unless an unequivocal non-cardiovascular cause of death was
confirmed by the central adjudication committee. Cardiovascular death included sudden
death; death due to myocardial infarction, heart failure, or stroke; procedure-related death
(death during a cardiovascular investigation/procedure/operation); death due to other specified
cardiovascular causes; and presumed cardiovascular deaths (e.g. those for whom a noncardiovascular cause could not be clearly established).
Statistical analysis
For investigating the effects of therapy on the individual disease course, a multistate model
considering the outcomes described above was developed. Multistate models are useful to
analyse multiple endpoints and their relations simultaneously and also to consider covariables.
7, 8
The analysed endpoints and sub-processes are referred to as ‘states’ and ‘transitions’. All
patients were defined as starting the study in a healthy state. State structure and transitions
analysed in the current analysis are shown in Figure 1. To account for the competing-ness of
distinct transitions (i.e. different components of the composite endpoint, and different causes
of death), Cox proportional cause-specific hazards models were applied, in other words a
Cox’s proportional hazards model was used to model the hazard for each type of event
treating the other events as being censored. In the primary analysis, therapy effects were
8
9
estimated without further adjustments. In the next step, analyses were adjusted for the
presence of an implantable cardioverter defibrillator (ICD) at baseline because the occurrence
of an ICD shock as a life-saving event is dependent on this factor. Interactions between ICD
at baseline and allocation to ASV therapy were considered where they were statistically
significant. Thirdly, interactions between the randomised study intervention and baseline
measures of LVEF (≤30%, 31-36%, and >36%) and CSR (<20%, 20-50% or >50%) were
tested because of the potential influence of these factors on the different study outcomes, as
indicated in a univariate subgroup analyses of the main SERVE-HF results.1 The division of
CSR into three categories was introduced prospectively since it was speculated that outcomes
might be different in different subgroups. LVEF was split according to tertiles. Two-sided
tests were performed keeping a significance level of 5%. Results were visualized with
cumulative incidence curves.8
Trial registration: ClinicalTrials.gov, NCT00733343
Role of the funding source
The SERVE-HF study was supported by ResMed Ltd. The steering committee oversaw the
conduct of the trial and data analysis in collaboration with the sponsor according to a
predefined statistical analysis plan. The trial was reviewed by an independent data and safety
monitoring committee. The first draft of the manuscript was prepared by the first three authors
and the final two authors, who had unrestricted access to the data, with the assistance of an
independent medical writer funded by ResMed. The manuscript was reviewed and edited by
all the authors. All authors made the decision to submit the manuscript for publication and
assume responsibility for the accuracy and completeness of the analyses and for the fidelity of
this report to the trial protocol.
9
10
Results
All 1325 patients enrolled in the SERVE-HF study (between February 2008 and May 2013)
were included in this analysis. A total of 666 (50·3%) patients were randomized to the ASV
treatment group and 659 (49·7%) to the control group. Follow-up was 0–80 (median 31)
months. The number of patients who had an ICD at baseline was 316 (47·5%) in the ASV
group and 314 (47·7%) in the control group. Patient characteristics are summarized in Table 1
and have been described in detail previously.1
A schematic representation of the different health states and possible transitions, including the
number of patients with each type of event, is shown in Figure 1. Hospitalisation for
worsening heart failure was the first primary endpoint event in 428/1325 patients (32·3%),
while 166/1325 patients (12·5%) had a ‘life-saving event’ as their first primary endpoint
event. Overall, 357 cardiovascular deaths were documented, of which 68 (19·0%) were
classified as cardiovascular death without prior hospitalisation for worsening heart failure or
life-saving event, and there were 68 deaths from other causes. The frequency of observed
transitions is outlined in Table 2.
Univariate comparisons showed an increased risk of cardiovascular death without prior
hospitalisation for worsening heart failure or life-saving event in patients allocated to the ASV
group compared with control (hazard ratio [HR] 2·59, 95% confidence interval [CI] 1·54,
4·37; p<0·001) (Table 3). In addition, the frequency of cardiovascular death after a life-saving
event was higher in the ASV group than in the control group (HR 1·57, 95% CI 1·01, 2·44;
p=0·045) (Table 3). No significant associations were identified for any of the other transitions
tested (Table 3). Both of the significant associations persisted after adjustment for the
presence of an ICD at baseline (HR 2·59, 95% CI 1·53, 4·36; p<0.001 with no ICD at
baseline and HR 1·57, 95% CI 1·01, 2·44; p=0·044 with an ICD at baseline). There was a
significant interaction between the presence of an ICD at baseline and non-cardiovascular
death after hospitalisation for worsening heart failure (p=0·033); hazard ratios for this
10
11
endpoint in patients with versus without an ICD were 3·03 (95% CI 0·85, 10·75; p=0·086)
and 0·41 (0·11, 1·56; p=0·192), respectively.
A further significant interaction was found for both CSR proportion at baseline (<20%, 2050% or >50%, as classified by the recruiting centres) and LVEF (≤30%, 31-36% or >36% on
baseline echocardiography). With respect to both hospitalisation for worsening heart failure
and cardiovascular death without prior hospitalisation for worsening heart failure or lifesaving event, there was a significant interaction between allocation to ASV and LVEF
(interaction p-values: 0·039 for hospitalisation for worsening heart failure [Table 4] and 0·026
for cardiovascular death without prior hospitalisation for worsening heart failure or life-saving
event [Table 5]). For both of these endpoints, the association between allocation to ASV and
outcome was strongest in patients with LVEF ≤30% compared to those with LVEF 31%-36%
or >36% (Table 4 and 5). Analyses stratified by the presence of an ICD failed to show
statistical significance. However, in the subgroup without an ICD at baseline and LVEF
≤30%, the HR was estimated as 24·08 (95% CI 3·14, 184·46; p=0·003). In patients with an
ICD at baseline, the HR for the same effect was estimated to be 1·89 (95% CI 0·63, 5·68;
p=0·259). There was also a significant interaction between ASV allocation and CSR
proportion at baseline with respect to hospitalisation for worsening heart failure. This event
was somewhat less likely in those with CSR <20% and significantly more likely in those with
CSR >50% (interaction p-value=0·021) (Table 4).
Discussion
The SERVE-HF primary intention-to-treat analysis1 formed the basis of the current multistate
model analysis. It showed a neutral result with respect to the primary composite endpoint and
identified significantly increased all-cause and cardiovascular mortality in the ASV versus
control group.1 The current multistate model analysis provided additional differentiation of
these results by investigating individual components of the composite endpoint.
11
12
Each component was analysed separately and the disease course after the first event was
studied. The multistate analysis showed that the randomised allocation to ASV significantly
increased the risk of cardiovascular death without prior hospitalisation for worsening heart
failure or life-saving event. This suggests sudden death as the key detrimental mechanism. In
addition, there was a trend to increased cardiovascular mortality after a life-saving
intervention, which is also likely to be due to sudden death. Interactions were detected
between ASV therapy and LVEF with respect to cardiovascular death without prior
hospitalisation for worsening heart failure or life-saving event (and hospitalisation for
worsening heart failure), and between ASV therapy and CSR proportion in terms of
hospitalisation for worsening heart failure. There were no associations between randomised
allocation to ASV and hospitalisation for worsening heart failure, or life-saving
cardiovascular intervention or non-cardiovascular mortality.
The multistate model findings of an interaction between baseline LVEF and worse outcomes
in patients allocated to ASV confirms the results of our previously published subgroup
analysis of the intention-to-treat population showing that LVEF significantly modified the
effects of ASV on the primary study endpoint.1 In addition, the interaction between CSR
proportion at baseline and hospitalisation for worsening heart failure in this multistate
analysis provides further insight into the primary ITT results of SERVE-HF, and suggests that
there is heterogeneity within the CSA patient population depending on the proportion of CSR.
The ITT subgroup analysis indicated that HFrEF patients with CSA but a low proportion of
CSR at baseline (<20%) may have better outcomes when allocated to ASV compared with
control.1 This suggests that if CSR is indeed a compensatory mechanism in severe HF,11 then
alleviating this adaptive breathing pattern would be most detrimental in the sickest HF
patients (i.e. those with the lowest left ventricular ejection fraction).
While the precise mechanism by which alleviation of CSR could increase the risk of
presumed sudden death cannot be determined from our data, it is likely that a change in
12
13
autonomic balance plays a role. It has previously been shown that resolution of CSR during
treatment with nocturnal oxygen or carbon dioxide was associated with significant increases
in plasma noradrenaline levels, indicative of increased sympathetic activation.12 Autonomic
dysfunction is an important marker of adverse outcome in coronary heart disease and chronic
heart failure patients,13, 14 and the potential risks associated with therapies that have
sympathomimetic effects in patients with cardiovascular disease has been recognized.13
In addition, the current findings do not support a contribution of ASV pressures to the
increased cardiovascular mortality observed in the ITT analysis: if excessive pressure played
a key role in the outcomes observed then we would expect to also see more hospitalisations
for worsening HF, but this was not the case. Furthermore one would also expect to see a
worsening in the overall heart failure syndrome (i.e. cardiac structure and function, and
cardiac and renal biomarkers), something that was not detected in the SERVE-HF major
echocardiographic substudy (data not shown).
Composite endpoints are often used in clinical trials, especially in cardiology, as has been
summarised previously.16 Advantages include decreased sample size requirements and the
assessment of treatment effects in the presence of competing risks.17, 18 It has been suggested
that composite endpoints should be used with caution in heart failure trials, 18, 19 and our
findings in this multistate analysis of SERVE-HF also raise questions as to whether use of a
composite endpoint is meaningful or whether individual components should be investigated
(and monitored by a data safety committee) separately. In SERVE-HF, components of the
composite endpoint have different levels of importance and occur at different frequencies.
Associations between ASV treatment and individual components differ. By combining these
individual outcomes into a composite endpoint, effects are averaged, and more frequent
events are automatically given more weight in the analysis. In addition, composite outcomes
usually only focus on the first event to occur (e.g. hospitalisation), and subsequent effects
(e.g. death) may not be included in the analysis. 18 As a result, the strong association between
13
14
ASV and cardiovascular death without prior hospitalisation for worsening heart failure or lifesaving event shown in this multistate analysis was diluted in the analysis of the primary
composite endpoint and also in the analysis of cardiovascular mortality.1
In general, theuse of a composite endpoint diminishes the possibility of detecting an important
treatment effect on individual components of that endpoint. The results become more
challenging to interpret,20 and the applicability of study results to clinical practice and
individual patients becomes less certain.
Although our multistate analysis of SERVE-HF data provides useful insights into the effects
of the randomised allocation to ASV on individual components of the primary composite
endpoint, the results need to be interpreted with caution because this is a post-hoc analysis. In
addition, hypotheses on interactions between allocation to ASV and an ICD, CSR proportion
and LVEF were proposed post-hoc, and the number of events available for analysis decreased
for each subsequent transition of the multistate model. Further, the analyses presented here are
on an ITT basis and did not consider crossovers, meaning the change of allocated treatment of
individual patients. Since there is no sham ASV device, patients and physicians were
unblinded in this trial, which again increased the number of crossovers. On-treatment analyses
on the other hand are not protected against bias by randomisation.
In conclusion, this multistate model analysis of the SERVE-HF data identified that the
increased risk of cardiovascular death is mainly seen when death occurs without a preceding
hospitalisation, pointing towards sudden death as the mechanism of harm.
A significant interaction between allocation to ASV and LVEF with respect to cardiovascular
death without prior hospitalisation for worsening heart failure or life-saving event was found.
There was also a significant interaction between both LVEF and CSR proportion at baseline
for hospitalisation for worsening heart failure.
14
15
Author contributions
The SERVE-HF study protocol was designed by the steering committee with the support of
the scientific advisory board. The steering committee oversaw the conduct of the trial and data
analysis in collaboration with the sponsor according to a predefined statistical analysis plan.
The trial was reviewed by an independent data and safety monitoring committee. For this
multistate model, statistical analysis was performed by CE and KW. The first draft of the
manuscript was prepared by CE, KW, HW and MRC, who had unrestricted access to the data,
with the assistance of an independent medical writer funded by ResMed. The manuscript was
reviewed and edited by all the authors. All authors made the decision to submit the
manuscript for publication and assume responsibility for the accuracy and completeness of the
analyses and for the fidelity of this report to the trial protocol.
Conflict of interest
Prof. Cowie’s and Prof. Simond’s salaries are supported by the National Institute for Health
Research Cardiovascular and Respiratory Biomedical Research Units, respectively, at the
Royal Brompton Hospital, London, UK. VKS was supported by NIH R01HL065176. The
content is solely the responsibility of the authors and does not necessarily represent the
official views of the National Institutes of Health.
Dr Eulenburg and Prof. Wegscheider, receiving grant support from ResMed; Dr. Woehrle,
being an employee of ResMed; Prof. Angermann, receiving fees for serving on advisory
boards from ResMed, Servier, Boehringer Ingelheim, and Vifor Pharma, fees for serving on a
steering committee from ResMed, lecture fees from Servier and Vifor Pharma, grant support
from ResMed, Thermo Fisher Scientific, Boehringer Ingelheim, Lundbeck, and Vifor Pharma,
financial support for statistical analyses from Thermo Fisher Scientific, and study medication
15
16
from Lundbeck; Prof. d’Ortho, receiving fees for serving on advisory boards from ResMed
and IP Santé, lecture fees from ResMed, Philips, IP Santé, and VitalAire, grant support from
Fisher and Paykel Healthcare, ResMed, Philips, ADEP Assistance, and IP Santé, and small
material donations from VitalAire; Prof. Erdmann, receiving fees for serving on advisory
boards and honoraria for lecturing from ResMed; Prof. Levy and Prof. Simonds, no potential
conflicts of interest outside the submitted work; Prof. Somers, receiving consulting fees from
PricewaterhouseCoopers, Sorin, GlaxoSmithKline, Respicardia, uHealth, Ronda Grey, Philips
Respironics and ResMed, working with Mayo Medical Ventures on intellectual property
related to sleep and cardiovascular disease, and having a pending patent (12/680073) related
to biomarkers of sleep apnoea; Prof. Zannad, receiving fees for serving on steering
committees from Janssen Pharmaceutica, Bayer, Pfizer, Novartis, Boston Scientific, ResMed,
and Takeda Pharmaceutical, receiving consulting fees from Servier, Stealth Peptides, Amgen,
and CVRx, and receiving lecture fees from Mitsubishi; Prof. Teschler, receiving consulting
fees, grant support, and hardware and software for the development of devices from ResMed;
and Prof. Cowie reports receiving consulting fees from Servier, Novartis, Pfizer, St. Jude
Medical, Boston Scientific, Respicardia, and Medtronic and grant support through his
institution from Bayer. No other potential conflict of interest relevant to this article was
reported.
Acknowledgements
This work was supported by ResMed Ltd. The authors would like to thank the team from CRI
(the Clinical Research Institute, Munich, Germany) for their expertise in overseeing the
SERVE-HF trial. Medical writing support was provided by Nicola Ryan, independent medical
writer, funded by ResMed.
16
17
References
1.
Cowie MR, Woehrle H, Wegscheider K, Angermann C, d'Ortho MP, Erdmann E, Levy
P, Simonds AK, Somers VK, Zannad F, Teschler H. Adaptive Servo-Ventilation for
Central Sleep Apnea in Systolic Heart Failure. N Engl J Med 2015;373(12):1095-105.
2.
D'Elia E, Vanoli E, La Rovere MT, Fanfulla F, Maggioni A, Casali V, Damiano S,
Specchia G, Mortara A. Adaptive servo ventilation reduces central sleep apnea in
chronic heart failure patients: beneficial effects on autonomic modulation of heart rate. J
Cardiovasc Med (Hagerstown) 2013;14(4):296-300.
3.
Hastings PC, Vazir A, Meadows GE, Dayer M, Poole-Wilson PA, McIntyre HF,
Morrell MJ, Cowie MR, Simonds AK. Adaptive servo-ventilation in heart failure
patients with sleep apnea: a real world study. Int J Cardiol 2010;139(1):17-24.
4.
Kourouklis SP, Vagiakis E, Paraskevaidis IA, Farmakis D, Kostikas K, Parissis JT,
Katsivas A, Kremastinos DT, Anastasiou-Nana M, Filippatos G. Effective sleep apnoea
treatment improves cardiac function in patients with chronic heart failure. Int J Cardiol
2013;168(1):157-62.
5.
Oldenburg O, Schmidt A, Lamp B, Bitter T, Muntean BG, Langer C, Horstkotte D.
Adaptive servoventilation improves cardiac function in patients with chronic heart
failure and Cheyne-Stokes respiration. Eur J Heart Fail 2008;10(6):581-6.
6.
Takama N, Kurabayashi M. Effect of adaptive servo-ventilation on 1-year prognosis in
heart failure patients. Circ J 2012;76(3):661-7.
7.
Andersen PK, Keiding N. Multi-state models for event history analysis. Stat Methods
Med Res 2002;11(2):91-115.
8.
Putter H, Fiocco M, Geskus RB. Tutorial in biostatistics: competing risks and multistate models. Stat Med 2007;26(11):2389-430.
17
18
9.
Cowie MR, Woehrle H, Wegscheider K, Angermann C, d'Ortho MP, Erdmann E, Levy
P, Simonds A, Somers VK, Zannad F, Teschler H. Rationale and design of the SERVEHF study: treatment of sleep-disordered breathing with predominant central sleep
apnoea with adaptive servo-ventilation in patients with chronic heart failure. Eur J
Heart Fail 2013;15(8):937-43.
10.
Djavadkhani Y, Marshall NS, D'Rozario AL, Crawford MR, Yee BJ, Grunstein RR,
Phillips CL. Ethics, consent and blinding: lessons from a placebo/sham controlled
CPAP crossover trial. Thorax 2015;70(3):265-9.
11.
Naughton MT. Cheyne-Stokes respiration: friend or foe? Thorax 2012;67(4):357-60.
12.
Andreas S, Weidel K, Hagenah G, Heindl S. Treatment of Cheyne-Stokes respiration
with nasal oxygen and carbon dioxide. Eur Respir J 1998;12(2):414-9.
13.
Curtis BM, O'Keefe JH, Jr. Autonomic tone as a cardiovascular risk factor: the dangers
of chronic fight or flight. Mayo Clin Proc 2002;77(1):45-54.
14.
Nolan J, Batin PD, Andrews R, Lindsay SJ, Brooksby P, Mullen M, Baig W, Flapan
AD, Cowley A, Prescott RJ, Neilson JM, Fox KA. Prospective study of heart rate
variability and mortality in chronic heart failure: results of the United Kingdom heart
failure evaluation and assessment of risk trial (UK-heart). Circulation
1998;98(15):1510-6.
15.
Bitter T, Fox H, Dimitriadis Z, Niedermeyer J, Prib N, Prinz C, Horstkotte D,
Oldenburg O. Circadian variation of defibrillator shocks in patients with chronic heart
failure: the impact of Cheyne-Stokes respiration and obstructive sleep apnea. Int J
Cardiol 2014;176(3):1033-5.
16.
Gomez G, Gomez-Mateu M, Dafni U. Informed choice of composite end points in
cardiovascular trials. Circ Cardiovasc Qual Outcomes 2014;7(1):170-8.
18
19
17.
Ferreira-Gonzalez I, Alonso-Coello P, Sola I, Pacheco-Huergo V, Domingo-Salvany A,
Alonso J, Montori V, Permanyer-Miralda G. [Composite endpoints in clinical trials].
Rev Esp Cardiol 2008;61(3):283-90.
18.
Neaton JD, Gray G, Zuckerman BD, Konstam MA. Key issues in end point selection for
heart failure trials: composite end points. J Card Fail 2005;11(8):567-75.
19.
Kip KE, Hollabaugh K, Marroquin OC, Williams DO. The problem with composite end
points in cardiovascular studies: the story of major adverse cardiac events and
percutaneous coronary intervention. J Am Coll Cardiol 2008;51(7):701-7.
20.
Freemantle N, Calvert M, Wood J, Eastaugh J, Griffin C. Composite outcomes in
randomized trials: greater precision but with greater uncertainty? JAMA
2003;289(19):2554-9.
21.
Ferreira-Gonzalez I, Busse JW, Heels-Ansdell D, Montori VM, Akl EA, Bryant DM,
Alonso-Coello P, Alonso J, Worster A, Upadhye S, Jaeschke R, Schunemann HJ,
Permanyer-Miralda G, Pacheco-Huergo V, Domingo-Salvany A, Wu P, Mills EJ,
Guyatt GH. Problems with use of composite end points in cardiovascular trials:
systematic review of randomised controlled trials. BMJ 2007;334(7597):786.
19
20
Figure Legends
Figure 1. Multistate model, where boxes represent the different possible health states and
arrows show transitions between states (CV, cardiovascular; HF, heart failure)
Figure 2. Cumulative incidence curves comparing incidences of hospitalisation for worsening
heart failure (A), cardiovascular death without prior hospitalisation for worsening heart failure
or life-saving event (B), and cardiovascular mortality after life-saving intervention (C)
between the randomised arms. HR, hazard ratio.
Figure 3. Cumulative incidence curves for different endpoints, stratified by randomization
group and significant interaction variables: (A) cardiovascular death without prior
hospitalisation for worsening heart failure or life-saving event by baseline LVEF; (B)
hospitalisation for worsening heart failure by baseline left ventricular ejection fraction
(LVEF); (B) hospitalisation for worsening heart failure by Cheyne-Stokes respiration (CSR)
proportion at baseline;
20
21
Text Tables
Table 1: Baseline patient demographic data and clinical characteristics1
Characteristic
Control
ASV
Total
69·3±10·4
69·6±9·5
69·4±10·0
599/659 (90·0)
599/666 (89·9)
28·6±5·1
28·4±4·7
28·5±4·9
II
194/654 (29·7)
195/662 (29·5)
389/1316 (29·6)
III or IV
460/654 (69·4)
467/662 (70·5)
927/1316 (70·4)
≤30
237/533 (44·5)
249/536 (46·5)
486/1069 (45·5)
31 – 36
126/533 (23.6)
117/536 (21.8)
243/1069 (22.7)
>36
170/533 (31·9)
170/536 (31·7)
340/1069 (31·8)
252/653 (38·6)
254/660 (38·5)
506/1313 (38·5)
Ischaemic
366/641 (57·1)
390/651 (59·9)
756/1292 (58·5)
Non-ischaemic
275/641 (42·9)
261/651 (40·1)
536/1292 (41·5)
SBP (mmHg)
122·1±19·6
122·3±19·0
122·2±19·3
DBP (mmHg)
73·3±11·5
73·7±11·3
73·5±11·4
cAHI/AHI
81·8±15·7
80·8±15·5
81·3±15·6
Left-bundle branch block¶
135/456 (29·6)
166/467 (35·5)
301/923 (32·6)
Sinus rhythm
395/646 (61·1)
372/650 (57·2)
767/1296 (59·2)
Atrial fibrillation
147/646 (22·8)
178/650 (27·4)
325/1296 (25·1)
CSR <20%
116/585 (19·8)
121/581 (20·8)
237/1166 (20·3)
CSR 20-50%
218/585 (37·3)
221/581 (38.0)
439/1166 (37.7)
CSR >50%
251/585 (42·9)
239/581 (41·1)
490/1166 (42·0)
Age (years)
Male sex – no. (%)
BMI (kg/m2)‡
1198/1325 (90·4)
NYHA class – no. (%)
LVEF
Diabetes mellitus – no. (%)
Heart failure aetiology – no. (%)
ECG findings – no. (%)
21
22
Implanted device – no. (%)
Non-CRT pacemaker
29/364 (8·0)
32/362 (8·8)
61/726 (8·4)
161/364 (44·2)
163/362 (45·0)
324/726 (44·6)
CRT-P
21/364 (5·8)
14/362 (3·9)
35/726 (4·8)
CRT-D
153/364 (42·0)
153/362 (42·3)
306/726 (42·1)
Hemoglobin (g/dL)‡
13·9±1·5
13·9±1·6
13·9±1·5
Creatinine (mg/dL)‡
1·4±0·6
1·4±0.6
1·4±0·6
Males (n=1148)
59·3 ±20·8
58·2±21·1
58·8±21·0
Females (n=121)
59·5±20·9
54·4±20·5
56·8±20·8
340·0±125·0
336·6±123·4
338·3±124·1
ACEI/ARB
603/659 (91·5)
612/666 (91·9)
1215/1325 (91·7)
ß-blockers
611/659 (92·7)
611/666 (91·7)
1222/1325 (92·2)
Aldosterone antagonists
325/659 (49·3)
315/666 (47·3)
640/1325 (48·3)
Diuretics
561/659 (85·1)
560/666 (84·1)
1121/1325 (84·6)
Cardiac glycosides
124/659 (18·8)
148/666 (22·2)
272/1325 (20·5)
Antiarrhythmics
89/659 (13·5)
128/666 (19·2)†
217/1325 (16·4)
ICD
eGFR (mL/min/1.73m2)
6MWD (m)‡§
Concomitant cardiac medication
– no. (%)
Data are mean (standard deviation) or number of patients (%), unless otherwise stated. There were no
significant differences between the control and ASV groups, apart from the rate of antiarrhythmic use,
which was higher in the ASV group († p=0·005).
‡
Data available in a subset of patients: BMI (n=1308); LVEF (n=1067; measurement of LVEF was
added to the study protocol 32 months after the first patient was randomized); BNP (n=544); diabetes
(n=1313); SBP (n=1299); DBP (n=1298); haemoglobin (n=1276); creatinine (n=1269); 6MWD
(n=1240); ESS (n=1303); cAI (n=1324); cAHI (n=1324); central apnoea proportion (n=1324);
desaturation index (n=1315); SaO2 (n=1322); minimum SaO2 (n=1321); time with SaO2 <90%
(n=1311).
¶
In patients without a pacemaker; p=0·054 for difference between groups.
§
6MWD is reported for patients who walked >0 metres.
6MWD, 6-minute walk distance; ACEI, angiotensin converting enzyme inhibitor; AHI, apneahypopnea index; ARB, angiotensin receptor blocker; ASV, adaptive servo-ventilation; BMI, body
mass index; cAHI, central apnea-hypopnea index; cAI, central apnea index; CSR, Cheyne-Stokes
respiration; DBP, diastolic BP; ECG, electrocardiogram; eGFR, estimated glomerular filtration rate;
ESS, Epworth Sleepiness Scale; ICD, implantable cardioverter defibrillator; CRT-D, cardiac
22
23
resynchronization device with defibrillator function; CRT-P, cardiac resynchronization device with
pacemaker function; LVEF; left ventricular ejection fraction; NYHA, New York Heart Association;
SaO2, oxygen saturation; SBP, systolic blood pressure.
23
24
Table 2: Frequency of observed transitions
Number of events
Transition
Overall
Control group
ASV group
Life-saving event
166
86
80
Hospitalisation for
worsening HF
428
211
217
Direct CV death
68
20
48
Direct non-CV death
33
18
15
CV death after lifesaving intervention
81
35
46
Non-CV death after
life-saving intervention
9
6
3
CV death after
hospitalisation
208
103
105
Non-CV death after
hospitalisation
26
11
15
Censored
630
ASV, adaptive servo-ventilation; CV, cardiovascular; HF, heart failure.
24
25
Table 3: Univariate associations between adaptive servo-ventilation therapy and individual
transitions
Hazard ratio
95% Confidence
interval
p-value
Life-saving event
0·97
0·72, 1·32
0·844
Hospitalisation for
worsening HF
1·08
0·90, 1·31
0·407
Direct CV death
2·59
1·54, 4·37
<0·001
Direct non-CV death
0·91
0·72; 1·24
0·780
CV death after lifesaving intervention
1·57
1·01, 2·44
0·045
Non-CV death after
life-saving intervention
0·58
0·14, 2·30
0·433
CV death after
hospitalisation
0·94
0·72, 1·24
0·660
Non-CV death after
hospitalisation
1·26
0·58, 2·75
0·557
Censored
630
Transition
CV, cardiovascular; HF, heart failure.
25
26
Table 4: Associations between adaptive servo-ventilation therapy and hospitalisation for
worsening heart failure, adjusted for implantable cardioverter defibrillator (ICD), CheyneStokes respiration (CSR) proportion at baseline and baseline left ventricular ejection fraction
(LVEF). Significant interactions are reported.
Subgroup
N (%)
Hazard ratio
95% CI
CSR <20%
237 (17·9)
0·64
0·40, 1·02
CSR 20-50%
439 (33.1)
1·31
0·92, 1·86
CSR >50%
490 (37·0)
1·36
1·01, 1·83
LVEF >36%
486 (36·7)
0·85
0·56, 1·30
LVEF 31–36%
243 (18·3)
0·84
0·54, 1·32
LVEF ≤30%
340 (25·7)
1·38
1·02, 1·86
p-value for
interaction
0·021
0·039
CI, confidence interval.
159 and 256 patients (12% and 19.3%) had missing data with respect to CSR and LVEF,
respectively. In these categories, the HR values for hospitalisation for worsening heart failure
were 0·85 (95% CI 0·51, 1·42) for missing CSR and HR 1·21 (95% CI 0·80, 1·83) for
missing LVEF.
26
27
Table 5: Associations between adaptive servo-ventilation therapy and cardiovascular death
without prior hospitalisation for worsening heart failure or life-saving event, adjusted for
implantable cardioverter defibrillator (ICD), Cheyne-Stokes respiration (CSR) proportion at
baseline and baseline left ventricular ejection fraction (LVEF). Significant interactions are
reported.
Subgroup
N (%)
Hazard ratio
95% CI
LVEF >36%
486 (36·7)
1·21
0·48, 3·08
LVEF 31–36%
243 (18·3)
2·33
0·60, 9·03
LVEF ≤30%
340 (25·7)
5·21
2·11, 12·89
p-value for
interaction
0·026
CI, confidence interval.
256 patients (19·3%) had missing data with respect to LVEF. In this category, the HR value
was 3·04 (95% CI 0·84, 11·56).
27