[ Commentary ]
SERVE-HF: More Questions Than Answers
Shahrokh Javaheri, MD, FCCP; Lee K. Brown, MD, FCCP; Winfried Randerath, MD; and Rami Khayat, MD, FCCP
The recent online publication of the SERVE-HF trial that evaluated the effect of treating central
sleep apnea (CSA) with an adaptive servoventilation (ASV) device in patients with heart failure
and reduced ejection fraction (HFrEF) has raised serious concerns about the safety of ASV in
these patients. Not only was ASV ineffective but post hoc analysis found excess cardiovascular
mortality in treated patients. The authors cited as one explanation an unfounded notion that
CSA is a compensatory mechanism with a protective effect in HFrEF patients. We believe that
there are several possible considerations that are more likely to explain the results of SERVE-
HF. In this commentary, we consider methodological issues including the use of a previous-
generation ASV device that constrained therapeutic settings to choices that are no longer in
wide clinical use. Patient selection, data collection, and treatment adherence as well as group
crossovers were not discussed in the trial as potential confounding factors. We have developed
alternative reasons that could potentially explain the results and that can be explored by
post hoc analysis of the SERVE-HF data. We believe that our analysis is of critical value to the
field and of particular importance to clinicians treating these patients.
CHEST 2016; 149(4):900-904
KEY WORDS: adaptive servoventilation; cardiac arrhythmia; central sleep apnea; congestive heart
failure; Hunter-Cheyne-Stokes breathing; noninvasive ventilation; sudden cardiac death
The recent publication of SERVE-HF, a with protective effects in HFrEF patients
trial that evaluated the effect of treating and (2) excess positive intrathoracic pressure
Hunter-Cheyne-Stokes breathing with caused by ASV might have had adverse
central sleep apnea (CSA) in patients with cardiovascular consequences.
heart failure and reduced ejection fraction
(HFrEF), has raised serious concerns about In our opinion, the former hypothesis has
the safety of adaptive servoventilation little scientific basis and lacks experimental
(ASV) in this population.1 In addition to support. In contrast, much experimental
having no effect on the primary composite evidence exists that hypoxia, arousals, and
end point, post hoc analysis revealed an increased sympathetic activity consequent
unexpected association between randomization to CSA have adverse cardiovascular effects
to ASV and excess cardiovascular mortality. that are reversed with positive-airway
The investigators speculated that (1) CSA pressure therapy.2-4 In addition, the post hoc
might represent a compensatory mechanism analysis of the Canadian continuous
ABBREVIATIONS: AHI = apne-ahypopnea index; ASV = adaptive Medicine (Dr Randerath), Bethanien Hospital, University of Cologne,
servoventilation; CSA = central sleep apnea; EPAP = expiratory posi- Solingen, Germany; and the Sleep Heart Program (Dr Khayat), the
tive airway pressure; HF = heart failure; HFpEF = heart failure with Ohio State University, Columbus, OH.
preserved ejection fraction; HFrEF = heart failure and reduced ejection CORRESPONDENCE TO: Shahrokh Javaheri, MD, FCCP, Sleep Labo-
fraction; IPS = inspiratory pressure support; LVEF = left ventricular ratory, Bethesda North Hospital, 10535 Montgomery Rd, Cincinnati,
ejection fraction; SDB = sleep-disordered breathing OH 45242; e-mail:
AFFILIATIONS: From the Sleep Laboratory (Dr Javaheri), Bethesda Copyright Ó 2016 American College of Chest Physicians. Published by
North Hospital, Cincinnati, OH; the Departments of Internal Medicine Elsevier Inc. All rights reserved.
and Electrical and Computer Engineering (Dr Brown), University of DOI: http://dx.doi.org/10.1016/j.chest.2015.12.021
New Mexico, Albuquerque, NM; the Center of Sleep and Ventilatory
900 Commentary [ 149#4 CHEST APRIL 2016 ]
, positive-airway pressure therapy randomized controlled the published trial and the safety notice, that ASV
trial (CANPAP) suggested that when CSA is effectively remains safe in patients with HFpEF.
treated, survival improves.5
We are also concerned about the possibility of additional
Our careful reading of the paper,1 its online supplement, protocol violations involving LVEF because values for
and the study protocol6 revealed important design, this crucial parameter were missing for 126 and 130
data collection, and analysis concerns. We address subjects in the control and ASV cohorts, respectively,
some of these issues below. as disclosed in the legend of Table 1. Again, we have
been told (oral communication to S. J. and R. K.)
Population Characterization and Protocol that ven though the actual values of LVEF were not
Adherence recorded, all had LVEF less than 45%. Written
confirmation of this by the investigators would be
A critical inclusion criterion was the requirement that
helpful in this regard.
enrolled subjects have a documented left ventricular
ejection fraction (LVEF) # 45%. However, a range of We also note in Figure 1 of the study1 that 87 patients
LVEF values from 9.0% to 71.0% in the control arm in the control arm withdrew at some point in the
and 10.0% to 54.0% in the ASV arm are noted in Table 1 study (most of whom subsequently started treatment
in the article.1 Therefore, a number of patients did with ASV); 168 subjects in the ASV cohort discontinued
not meet the inclusion criteria and had heart failure this treatment at some point before study completion;
with preserved ejection fraction (HFpEF). We have 21 patients assigned to ASV never received it; and
been informed, however, that only a small number 85 patients withdrew from the ASV arm. The intention-
of patients had HFpEF (oral communication to S. J. to-treat analysis considers all of these patients in
and R. K.), in which case elimination of such patients calculating mortality, possibly skewing the results
should not change the results. Nonetheless, it is (see subsequent sections regarding this and other
important to know whether these patients were among analysis issues). Sensitivity testing that includes
those with increased cardiovascular mortality because adherence and time of exposure to the device is
treating physicians are under the impression, based on critical to understanding the effect of ASV on mortality.
Intervention Methodology and Adaptive previous-generation ASV device failed to adequately control SDB in
some of the subjects studied, and it is therefore surprising that the
Servoventilation Issues authors asserted that SDB was “well controlled.”1 It is conceivable
The particular ASV device used in SERVE-HF was a first-generation that patients in whom ASV failed to suppress CSA or obstructive
model no longer manufactured by the sponsor. This technology may sleep apnea contributed to the negative primary outcome or even
have applied pressures that were too low for some patients and the excess mortality. This scenario would be similar to that
excessive for others, with adverse cardiovascular consequences. As which was predicted10 and in fact materialized in a previous
detailed in a recent publication,7 later generations of ASV devices unsuccessful trial of positive-pressure therapy.4 We also note that after
have incorporated important advances in technology that might have the first 12 months, clinic visits with device downloads occurred only
affected the ultimate results had they been used in this trial. annually.
First, the ASV device used in SERVE-HF allows for only fixed These long intervals without assessment raise the possibility that
expiratory positive airway pressure (EPAP). Data from Table 2 in mortality in some patients may have been attributable to SDB even if
the study1 indicate that at the baseline, on average, 20% of the the last download showed adequate control of sleep apnea. This
apnea-hypopnea index (AHI) for both control and ASV subjects was possibility is further strengthened by the fact that the ASV device
composed of obstructive events. It is known that the phenotype of used in SERVE-HF used fixed EPAP settings, although the phenotype
sleep-disordered breathing (SDB) may change over time from of SDB demonstrably changed with time. Initially, central AHI
predominantly central to predominantly obstructive events.8 In a accounted for 80% of all disordered breathing events; with time,
fluid-overload state such as heart failure (HF), the latter may largely however, this percentage decreased to about 40%, meaning that on
be related to shifts in fluid status.9 Given the variability of the sleep average 60% of the events were considered obstructive in nature.
apnea phenotype, the fixed EPAP might at times prove inadequate, Given the wide range of residual AHIs reported in the Supplementary
with consequent residual SDB events. Evidence that this might have Appendix,1 there inevitably would have been patients who developed
occurred in SERVE-HF is found in the large range of AHIs significant obstructive apneas not adequately suppressed by the fixed
downloaded from the ASV devices across the months of follow-up expiratory pressure of the device. When and if that occurred, the
(Table S4 in the Supplementary Appendix). Recorded AHI values at ASV device used was equipped with only one strategy for suppressing
3, 12, 24, 36, and 48 months ranged as high as 72, 51, 46, 61, and these events: progressively increasing inspiratory pressure support in
38 events/hour and mean values were all more than 5/hour. an attempt to open the closed airway. Once the airway opened, the
Times below an oxyhemoglobin saturation of 90% were as high as prevailing high pressures may have resulted in an excessive rise in
344, 269, 285, 291, and 278 minutes at the same respective time intrathoracic pressure with consequent adverse hemodynamic effects.
points, and means were all 18 minutes or higher. Clearly, this The current generation of ASV devices can be set to increase EPAP
journal.publications.chestnet.org 901
SERVE-HF: More Questions Than Answers
Shahrokh Javaheri, MD, FCCP; Lee K. Brown, MD, FCCP; Winfried Randerath, MD; and Rami Khayat, MD, FCCP
The recent online publication of the SERVE-HF trial that evaluated the effect of treating central
sleep apnea (CSA) with an adaptive servoventilation (ASV) device in patients with heart failure
and reduced ejection fraction (HFrEF) has raised serious concerns about the safety of ASV in
these patients. Not only was ASV ineffective but post hoc analysis found excess cardiovascular
mortality in treated patients. The authors cited as one explanation an unfounded notion that
CSA is a compensatory mechanism with a protective effect in HFrEF patients. We believe that
there are several possible considerations that are more likely to explain the results of SERVE-
HF. In this commentary, we consider methodological issues including the use of a previous-
generation ASV device that constrained therapeutic settings to choices that are no longer in
wide clinical use. Patient selection, data collection, and treatment adherence as well as group
crossovers were not discussed in the trial as potential confounding factors. We have developed
alternative reasons that could potentially explain the results and that can be explored by
post hoc analysis of the SERVE-HF data. We believe that our analysis is of critical value to the
field and of particular importance to clinicians treating these patients.
CHEST 2016; 149(4):900-904
KEY WORDS: adaptive servoventilation; cardiac arrhythmia; central sleep apnea; congestive heart
failure; Hunter-Cheyne-Stokes breathing; noninvasive ventilation; sudden cardiac death
The recent publication of SERVE-HF, a with protective effects in HFrEF patients
trial that evaluated the effect of treating and (2) excess positive intrathoracic pressure
Hunter-Cheyne-Stokes breathing with caused by ASV might have had adverse
central sleep apnea (CSA) in patients with cardiovascular consequences.
heart failure and reduced ejection fraction
(HFrEF), has raised serious concerns about In our opinion, the former hypothesis has
the safety of adaptive servoventilation little scientific basis and lacks experimental
(ASV) in this population.1 In addition to support. In contrast, much experimental
having no effect on the primary composite evidence exists that hypoxia, arousals, and
end point, post hoc analysis revealed an increased sympathetic activity consequent
unexpected association between randomization to CSA have adverse cardiovascular effects
to ASV and excess cardiovascular mortality. that are reversed with positive-airway
The investigators speculated that (1) CSA pressure therapy.2-4 In addition, the post hoc
might represent a compensatory mechanism analysis of the Canadian continuous
ABBREVIATIONS: AHI = apne-ahypopnea index; ASV = adaptive Medicine (Dr Randerath), Bethanien Hospital, University of Cologne,
servoventilation; CSA = central sleep apnea; EPAP = expiratory posi- Solingen, Germany; and the Sleep Heart Program (Dr Khayat), the
tive airway pressure; HF = heart failure; HFpEF = heart failure with Ohio State University, Columbus, OH.
preserved ejection fraction; HFrEF = heart failure and reduced ejection CORRESPONDENCE TO: Shahrokh Javaheri, MD, FCCP, Sleep Labo-
fraction; IPS = inspiratory pressure support; LVEF = left ventricular ratory, Bethesda North Hospital, 10535 Montgomery Rd, Cincinnati,
ejection fraction; SDB = sleep-disordered breathing OH 45242; e-mail:
AFFILIATIONS: From the Sleep Laboratory (Dr Javaheri), Bethesda Copyright Ó 2016 American College of Chest Physicians. Published by
North Hospital, Cincinnati, OH; the Departments of Internal Medicine Elsevier Inc. All rights reserved.
and Electrical and Computer Engineering (Dr Brown), University of DOI: http://dx.doi.org/10.1016/j.chest.2015.12.021
New Mexico, Albuquerque, NM; the Center of Sleep and Ventilatory
900 Commentary [ 149#4 CHEST APRIL 2016 ]
, positive-airway pressure therapy randomized controlled the published trial and the safety notice, that ASV
trial (CANPAP) suggested that when CSA is effectively remains safe in patients with HFpEF.
treated, survival improves.5
We are also concerned about the possibility of additional
Our careful reading of the paper,1 its online supplement, protocol violations involving LVEF because values for
and the study protocol6 revealed important design, this crucial parameter were missing for 126 and 130
data collection, and analysis concerns. We address subjects in the control and ASV cohorts, respectively,
some of these issues below. as disclosed in the legend of Table 1. Again, we have
been told (oral communication to S. J. and R. K.)
Population Characterization and Protocol that ven though the actual values of LVEF were not
Adherence recorded, all had LVEF less than 45%. Written
confirmation of this by the investigators would be
A critical inclusion criterion was the requirement that
helpful in this regard.
enrolled subjects have a documented left ventricular
ejection fraction (LVEF) # 45%. However, a range of We also note in Figure 1 of the study1 that 87 patients
LVEF values from 9.0% to 71.0% in the control arm in the control arm withdrew at some point in the
and 10.0% to 54.0% in the ASV arm are noted in Table 1 study (most of whom subsequently started treatment
in the article.1 Therefore, a number of patients did with ASV); 168 subjects in the ASV cohort discontinued
not meet the inclusion criteria and had heart failure this treatment at some point before study completion;
with preserved ejection fraction (HFpEF). We have 21 patients assigned to ASV never received it; and
been informed, however, that only a small number 85 patients withdrew from the ASV arm. The intention-
of patients had HFpEF (oral communication to S. J. to-treat analysis considers all of these patients in
and R. K.), in which case elimination of such patients calculating mortality, possibly skewing the results
should not change the results. Nonetheless, it is (see subsequent sections regarding this and other
important to know whether these patients were among analysis issues). Sensitivity testing that includes
those with increased cardiovascular mortality because adherence and time of exposure to the device is
treating physicians are under the impression, based on critical to understanding the effect of ASV on mortality.
Intervention Methodology and Adaptive previous-generation ASV device failed to adequately control SDB in
some of the subjects studied, and it is therefore surprising that the
Servoventilation Issues authors asserted that SDB was “well controlled.”1 It is conceivable
The particular ASV device used in SERVE-HF was a first-generation that patients in whom ASV failed to suppress CSA or obstructive
model no longer manufactured by the sponsor. This technology may sleep apnea contributed to the negative primary outcome or even
have applied pressures that were too low for some patients and the excess mortality. This scenario would be similar to that
excessive for others, with adverse cardiovascular consequences. As which was predicted10 and in fact materialized in a previous
detailed in a recent publication,7 later generations of ASV devices unsuccessful trial of positive-pressure therapy.4 We also note that after
have incorporated important advances in technology that might have the first 12 months, clinic visits with device downloads occurred only
affected the ultimate results had they been used in this trial. annually.
First, the ASV device used in SERVE-HF allows for only fixed These long intervals without assessment raise the possibility that
expiratory positive airway pressure (EPAP). Data from Table 2 in mortality in some patients may have been attributable to SDB even if
the study1 indicate that at the baseline, on average, 20% of the the last download showed adequate control of sleep apnea. This
apnea-hypopnea index (AHI) for both control and ASV subjects was possibility is further strengthened by the fact that the ASV device
composed of obstructive events. It is known that the phenotype of used in SERVE-HF used fixed EPAP settings, although the phenotype
sleep-disordered breathing (SDB) may change over time from of SDB demonstrably changed with time. Initially, central AHI
predominantly central to predominantly obstructive events.8 In a accounted for 80% of all disordered breathing events; with time,
fluid-overload state such as heart failure (HF), the latter may largely however, this percentage decreased to about 40%, meaning that on
be related to shifts in fluid status.9 Given the variability of the sleep average 60% of the events were considered obstructive in nature.
apnea phenotype, the fixed EPAP might at times prove inadequate, Given the wide range of residual AHIs reported in the Supplementary
with consequent residual SDB events. Evidence that this might have Appendix,1 there inevitably would have been patients who developed
occurred in SERVE-HF is found in the large range of AHIs significant obstructive apneas not adequately suppressed by the fixed
downloaded from the ASV devices across the months of follow-up expiratory pressure of the device. When and if that occurred, the
(Table S4 in the Supplementary Appendix). Recorded AHI values at ASV device used was equipped with only one strategy for suppressing
3, 12, 24, 36, and 48 months ranged as high as 72, 51, 46, 61, and these events: progressively increasing inspiratory pressure support in
38 events/hour and mean values were all more than 5/hour. an attempt to open the closed airway. Once the airway opened, the
Times below an oxyhemoglobin saturation of 90% were as high as prevailing high pressures may have resulted in an excessive rise in
344, 269, 285, 291, and 278 minutes at the same respective time intrathoracic pressure with consequent adverse hemodynamic effects.
points, and means were all 18 minutes or higher. Clearly, this The current generation of ASV devices can be set to increase EPAP
journal.publications.chestnet.org 901
