And Maria Rosa Costanzo, MD

Compared with obstructive sleep apnea (OSA), the prevalence of central sleep apnea (CSA) is low in the general population. However, in adults, CSA may be highly prevalent in certain conditions, most commonly among those with left ventricular systolic dysfunction, left ventricular diastolic dysfunction, atrial fibrillation, stroke, and opioid users (Javaheri S, et al. J Am Coll Cardiol. 2017; 69:841). CSA may also be found in patients with carotid artery stenosis, cervical neck injury, and renal dysfunction. CSA can occur when OSA is treated (treatment-emergent central sleep apnea, or TECA), notably, and most frequently, with continuous positive airway pressure (CPAP) devices. Though in many individuals, this frequently resolves with continued use of the device.

In addition, unlike OSA, adequate treatment of CSA has proven difficult. Specifically, the response to CPAP, oxygen, theophylline, acetazolamide, and adaptive-servo ventilation (ASV) is highly variable, with individuals who respond well, and individuals in whom therapy fails to fully suppress the disorder.

Our interest in phrenic nerve stimulation increased after it was shown that CPAP therapy failed to improve morbidity and mortality of CSA in patients with heart failure and reduced ejection fraction (HFrEF) (CANPAP trial, Bradley et al. N Engl J Med. 2005;353(19):2025). In fact, in this trial, treatment with CPAP was associated with significantly increased mortality during the first few months of therapy. We reason that a potential mechanism was positive airway pressure that had adverse cardiovascular effects (Javaheri S. J Clin Sleep Med. 2006;2:399). This is because positive airway pressure therapy decreases venous return to the right side of the heart and increases lung volume. This could increase pulmonary vascular resistance (right ventricular afterload), which is lung volume-dependent. Therefore, the subgroup of individuals with heart failure whose right ventricular function is preload-dependent and has pulmonary hypertension is at risk for premature mortality with any PAP device.

Interestingly, investigators of the SERVE-HF trial (Cowie MR, et al. N Engl J Med. 2015;373:1095) also hypothesized that one reason for excess mortality associated with ASV use might have been due to an ASV-associated excessive rise in intrathoracic pressure, similar to the hypothesis we proposed earlier for CPAP. We expanded on this hypothesis and reasoned that based on the algorithm of the device, in some patients, it could have generated excessive minute ventilation and pressure contributing to excess mortality, either at night or daytime (Javaheri S, et al. Chest. 2016;149:900). Other deficiencies of the algorithm of the ASV device could have contributed to excess mortality as well (Javaheri S, et al. Chest. 2014;146:514). These deficiencies of the ASV device used in the SERVE-HF trial have been significantly improved in the new generation of ASV devices.

Undoubtedly, therefore, mask therapy with positive airway pressures increases intrathoracic pressure and will adversely affect cardiovascular function in some patients with heart failure. Another issue for mask therapy is adherence to the device remains poor, as demonstrated both in the CANPAP and SERVE-HF trials, confirming the need for new approaches utilizing non-mask therapies both for CSA and OSA.

Given the limitations of mask-based therapies, over the last several years, we have performed studies exploring the use of oxygen, acetazolamide, theophylline, and, most recently, phrenic nerve stimulation (PNS). In general, these therapies are devoid of increasing intrathoracic pressure and are expected to be less reliant on patients’ adherence than PAP therapy. Long-term randomized clinical trials are needed, and, most recently, the NIH approved a phase 3 trial for a randomized placebo-controlled low flow oxygen therapy for treatment of CSA in HFrEF. This is a modified trial proposed by one of us more than 20 years ago!

Regarding PNS, CSA is characterized by intermittent phrenic nerve (and intercostal nerves) deactivation. It, therefore, makes sense to have an implanted stimulator for the phrenic nerve to prevent development of central apneas during sleep. This is not a new idea. In 1948, Sarnoff and colleagues demonstrated for the first time that artificial respiration could be effectively administered to the cat, dog, monkey, and rabbit in the absence of spontaneous respiration by electrical stimulation of one (or both) phrenic nerves (Sarnoff SJ, et al. Science. 1948;108:482). In later experiments, these investigators showed that unilateral phrenic nerve stimulation is also equally effective in man as that shown in animal models.

The phrenic nerves comes in contact with veins on both the right (brachiocephalic) and the left (pericardiophrenic vein) side of the mediastinum. Like a cardiac pacemaker, an electrophysiologist places the stimulator within the vein at the point of encounter with the phrenic nerve. Only unilateral stimulation is needed for the therapy. The device is typically placed on the right side of the chest as many patients may already have a cardiac implanted electronic device such as a pacemaker. Like the hypoglossal nerve stimulation, the FDA approved this device for the treatment of OSA. The system can be programmed using an external programmer in the office.

Phrenic nerve stimulation system is initially activated 1 month after the device is placed. It is programmed to be automatically activated at night when the patient is at rest. First, a time is set on the device for when the patient typically goes to bed and awakens. This allows the therapy to activate. The device contains a position sensor and accelerometer, which determine position and activity level. Once appropriate time, position, and activity are confirmed, the device activates automatically. Therapy comes on and can increase in level over several minutes. The device senses transthoracic impedance and can use this measurement to make changes in the therapy output and activity. If the patient gets up at night, the device automatically stops and restarts when the patient is back in a sleeping position. How quickly the therapy restarts and at what energy is programmable. The device may allow from 1 to 15 minutes for the patient to get back to sleep before beginning therapy. These programming changes allow for patient acceptance and comfort with the therapy even in very sensitive patients. Importantly, no patient activation is needed, so that therapy delivery is independent of patient’s adherence over time.

In the prospective, randomized pivotal trial (Costanzo et al. Lancet. 2016;388:974), 151 eligible patients with moderate-severe central sleep apnea were implanted and randomly assigned to the treatment (n=73) or control (n=78) groups. Participants in the active arm received PNS for 6 months. All polysomnograms were centrally and blindly scored. There were significant decreases in AHI (50 to 26/per hour of sleep), CAI (32 to 6), arousal index (46 to 25), and ODI (44 to 25). Two points should be emphasized: first, changes in AHI with PNS are similar to those in CANPAP trial, and there remained a significant number of hypopneas (some of these hypopneas are at least in part related to the speed of the titration when the subject sits up and the device automatically is deactivated, only to resume therapy in supine position); second, in contrast to the CANPAP trial, there was a significant reduction in arousals. Probably for this reason, subjective daytime sleepiness, as measured by the ESS, improved. In addition, PNS improved quality of life, in contrast to lack of effect of CPAP or ASV in this domain. Regarding side effects, 138 (91%) of 151 patients had no serious-related adverse events at 12 months. Seven (9%) cases of related-serious adverse events occurred in the control group and six (8%) cases were reported in the treatment group.—3.4% needed lead repositioning, a rate which is like that of cardiac implantable devices. Seven patients died (unrelated to implant, system, or therapy), four deaths (two in treatment group and two in control group) during the 6-month randomization period when neurostimulation was delivered to only the treatment and was off in the control group, and three deaths between 6 months and 12 months of follow-up when all patients received neurostimulation. Of 73 patients in the treatment group, 27 (37%) reported nonserious therapy-related discomfort that was resolved with simple system reprogramming in 26 (36%) patients but was unresolved in one (1%) patient.

Long-term studies have shown sustained effects of PNS on CSA with improvement in both sleep metrics and QOL, as measured by the Minnesota Living with Heart Failure Questionnaire (MLWHF) and patient global assessment (PGA). Furthermore, in the subgroup of patients with concomitant heart failure with LVEF ≤ 45%, PNS was associated with both improvements in LVEF and a trend toward lower hospitalization rates (Costanzo et al. Eur J Heart Fail. 2018; doi:10.1002/ejhf.1312).

Several issues must be emphasized. One advantage of PNS is complete adherence resulting in a major reduction in apnea burden across the whole night. Second, the mechanism of action prevents any potential adverse consequences related to increased intrathoracic pressure. However, the cost of this therapy is high, similar to that of hypoglossal nerve stimulation. Large scale, long-term studies related to mortality are not yet available, and continued research should help identify those patients most likely to benefit from this therapeutic approach.

Compared with obstructive sleep apnea (OSA), the prevalence of central sleep apnea (CSA) is low in the general population. However, in adults, CSA may be highly prevalent in certain conditions, most commonly among those with left ventricular systolic dysfunction, left ventricular diastolic dysfunction, atrial fibrillation, stroke, and opioid users (Javaheri S, et al. J Am Coll Cardiol. 2017; 69:841). CSA may also be found in patients with carotid artery stenosis, cervical neck injury, and renal dysfunction. CSA can occur when OSA is treated (treatment-emergent central sleep apnea, or TECA), notably, and most frequently, with continuous positive airway pressure (CPAP) devices. Though in many individuals, this frequently resolves with continued use of the device.

In addition, unlike OSA, adequate treatment of CSA has proven difficult. Specifically, the response to CPAP, oxygen, theophylline, acetazolamide, and adaptive-servo ventilation (ASV) is highly variable, with individuals who respond well, and individuals in whom therapy fails to fully suppress the disorder.

Our interest in phrenic nerve stimulation increased after it was shown that CPAP therapy failed to improve morbidity and mortality of CSA in patients with heart failure and reduced ejection fraction (HFrEF) (CANPAP trial, Bradley et al. N Engl J Med. 2005;353(19):2025). In fact, in this trial, treatment with CPAP was associated with significantly increased mortality during the first few months of therapy. We reason that a potential mechanism was positive airway pressure that had adverse cardiovascular effects (Javaheri S. J Clin Sleep Med. 2006;2:399). This is because positive airway pressure therapy decreases venous return to the right side of the heart and increases lung volume. This could increase pulmonary vascular resistance (right ventricular afterload), which is lung volume-dependent. Therefore, the subgroup of individuals with heart failure whose right ventricular function is preload-dependent and has pulmonary hypertension is at risk for premature mortality with any PAP device.

Interestingly, investigators of the SERVE-HF trial (Cowie MR, et al. N Engl J Med. 2015;373:1095) also hypothesized that one reason for excess mortality associated with ASV use might have been due to an ASV-associated excessive rise in intrathoracic pressure, similar to the hypothesis we proposed earlier for CPAP. We expanded on this hypothesis and reasoned that based on the algorithm of the device, in some patients, it could have generated excessive minute ventilation and pressure contributing to excess mortality, either at night or daytime (Javaheri S, et al. Chest. 2016;149:900). Other deficiencies of the algorithm of the ASV device could have contributed to excess mortality as well (Javaheri S, et al. Chest. 2014;146:514). These deficiencies of the ASV device used in the SERVE-HF trial have been significantly improved in the new generation of ASV devices.

Undoubtedly, therefore, mask therapy with positive airway pressures increases intrathoracic pressure and will adversely affect cardiovascular function in some patients with heart failure. Another issue for mask therapy is adherence to the device remains poor, as demonstrated both in the CANPAP and SERVE-HF trials, confirming the need for new approaches utilizing non-mask therapies both for CSA and OSA.

Given the limitations of mask-based therapies, over the last several years, we have performed studies exploring the use of oxygen, acetazolamide, theophylline, and, most recently, phrenic nerve stimulation (PNS). In general, these therapies are devoid of increasing intrathoracic pressure and are expected to be less reliant on patients’ adherence than PAP therapy. Long-term randomized clinical trials are needed, and, most recently, the NIH approved a phase 3 trial for a randomized placebo-controlled low flow oxygen therapy for treatment of CSA in HFrEF. This is a modified trial proposed by one of us more than 20 years ago!

Regarding PNS, CSA is characterized by intermittent phrenic nerve (and intercostal nerves) deactivation. It, therefore, makes sense to have an implanted stimulator for the phrenic nerve to prevent development of central apneas during sleep. This is not a new idea. In 1948, Sarnoff and colleagues demonstrated for the first time that artificial respiration could be effectively administered to the cat, dog, monkey, and rabbit in the absence of spontaneous respiration by electrical stimulation of one (or both) phrenic nerves (Sarnoff SJ, et al. Science. 1948;108:482). In later experiments, these investigators showed that unilateral phrenic nerve stimulation is also equally effective in man as that shown in animal models.

The phrenic nerves comes in contact with veins on both the right (brachiocephalic) and the left (pericardiophrenic vein) side of the mediastinum. Like a cardiac pacemaker, an electrophysiologist places the stimulator within the vein at the point of encounter with the phrenic nerve. Only unilateral stimulation is needed for the therapy. The device is typically placed on the right side of the chest as many patients may already have a cardiac implanted electronic device such as a pacemaker. Like the hypoglossal nerve stimulation, the FDA approved this device for the treatment of OSA. The system can be programmed using an external programmer in the office.

Phrenic nerve stimulation system is initially activated 1 month after the device is placed. It is programmed to be automatically activated at night when the patient is at rest. First, a time is set on the device for when the patient typically goes to bed and awakens. This allows the therapy to activate. The device contains a position sensor and accelerometer, which determine position and activity level. Once appropriate time, position, and activity are confirmed, the device activates automatically. Therapy comes on and can increase in level over several minutes. The device senses transthoracic impedance and can use this measurement to make changes in the therapy output and activity. If the patient gets up at night, the device automatically stops and restarts when the patient is back in a sleeping position. How quickly the therapy restarts and at what energy is programmable. The device may allow from 1 to 15 minutes for the patient to get back to sleep before beginning therapy. These programming changes allow for patient acceptance and comfort with the therapy even in very sensitive patients. Importantly, no patient activation is needed, so that therapy delivery is independent of patient’s adherence over time.

In the prospective, randomized pivotal trial (Costanzo et al. Lancet. 2016;388:974), 151 eligible patients with moderate-severe central sleep apnea were implanted and randomly assigned to the treatment (n=73) or control (n=78) groups. Participants in the active arm received PNS for 6 months. All polysomnograms were centrally and blindly scored. There were significant decreases in AHI (50 to 26/per hour of sleep), CAI (32 to 6), arousal index (46 to 25), and ODI (44 to 25). Two points should be emphasized: first, changes in AHI with PNS are similar to those in CANPAP trial, and there remained a significant number of hypopneas (some of these hypopneas are at least in part related to the speed of the titration when the subject sits up and the device automatically is deactivated, only to resume therapy in supine position); second, in contrast to the CANPAP trial, there was a significant reduction in arousals. Probably for this reason, subjective daytime sleepiness, as measured by the ESS, improved. In addition, PNS improved quality of life, in contrast to lack of effect of CPAP or ASV in this domain. Regarding side effects, 138 (91%) of 151 patients had no serious-related adverse events at 12 months. Seven (9%) cases of related-serious adverse events occurred in the control group and six (8%) cases were reported in the treatment group.—3.4% needed lead repositioning, a rate which is like that of cardiac implantable devices. Seven patients died (unrelated to implant, system, or therapy), four deaths (two in treatment group and two in control group) during the 6-month randomization period when neurostimulation was delivered to only the treatment and was off in the control group, and three deaths between 6 months and 12 months of follow-up when all patients received neurostimulation. Of 73 patients in the treatment group, 27 (37%) reported nonserious therapy-related discomfort that was resolved with simple system reprogramming in 26 (36%) patients but was unresolved in one (1%) patient.

Long-term studies have shown sustained effects of PNS on CSA with improvement in both sleep metrics and QOL, as measured by the Minnesota Living with Heart Failure Questionnaire (MLWHF) and patient global assessment (PGA). Furthermore, in the subgroup of patients with concomitant heart failure with LVEF ≤ 45%, PNS was associated with both improvements in LVEF and a trend toward lower hospitalization rates (Costanzo et al. Eur J Heart Fail. 2018; doi:10.1002/ejhf.1312).

Several issues must be emphasized. One advantage of PNS is complete adherence resulting in a major reduction in apnea burden across the whole night. Second, the mechanism of action prevents any potential adverse consequences related to increased intrathoracic pressure. However, the cost of this therapy is high, similar to that of hypoglossal nerve stimulation. Large scale, long-term studies related to mortality are not yet available, and continued research should help identify those patients most likely to benefit from this therapeutic approach.

Compared with obstructive sleep apnea (OSA), the prevalence of central sleep apnea (CSA) is low in the general population. However, in adults, CSA may be highly prevalent in certain conditions, most commonly among those with left ventricular systolic dysfunction, left ventricular diastolic dysfunction, atrial fibrillation, stroke, and opioid users (Javaheri S, et al. J Am Coll Cardiol. 2017; 69:841). CSA may also be found in patients with carotid artery stenosis, cervical neck injury, and renal dysfunction. CSA can occur when OSA is treated (treatment-emergent central sleep apnea, or TECA), notably, and most frequently, with continuous positive airway pressure (CPAP) devices. Though in many individuals, this frequently resolves with continued use of the device.

In addition, unlike OSA, adequate treatment of CSA has proven difficult. Specifically, the response to CPAP, oxygen, theophylline, acetazolamide, and adaptive-servo ventilation (ASV) is highly variable, with individuals who respond well, and individuals in whom therapy fails to fully suppress the disorder.

Our interest in phrenic nerve stimulation increased after it was shown that CPAP therapy failed to improve morbidity and mortality of CSA in patients with heart failure and reduced ejection fraction (HFrEF) (CANPAP trial, Bradley et al. N Engl J Med. 2005;353(19):2025). In fact, in this trial, treatment with CPAP was associated with significantly increased mortality during the first few months of therapy. We reason that a potential mechanism was positive airway pressure that had adverse cardiovascular effects (Javaheri S. J Clin Sleep Med. 2006;2:399). This is because positive airway pressure therapy decreases venous return to the right side of the heart and increases lung volume. This could increase pulmonary vascular resistance (right ventricular afterload), which is lung volume-dependent. Therefore, the subgroup of individuals with heart failure whose right ventricular function is preload-dependent and has pulmonary hypertension is at risk for premature mortality with any PAP device.

Interestingly, investigators of the SERVE-HF trial (Cowie MR, et al. N Engl J Med. 2015;373:1095) also hypothesized that one reason for excess mortality associated with ASV use might have been due to an ASV-associated excessive rise in intrathoracic pressure, similar to the hypothesis we proposed earlier for CPAP. We expanded on this hypothesis and reasoned that based on the algorithm of the device, in some patients, it could have generated excessive minute ventilation and pressure contributing to excess mortality, either at night or daytime (Javaheri S, et al. Chest. 2016;149:900). Other deficiencies of the algorithm of the ASV device could have contributed to excess mortality as well (Javaheri S, et al. Chest. 2014;146:514). These deficiencies of the ASV device used in the SERVE-HF trial have been significantly improved in the new generation of ASV devices.

Undoubtedly, therefore, mask therapy with positive airway pressures increases intrathoracic pressure and will adversely affect cardiovascular function in some patients with heart failure. Another issue for mask therapy is adherence to the device remains poor, as demonstrated both in the CANPAP and SERVE-HF trials, confirming the need for new approaches utilizing non-mask therapies both for CSA and OSA.

Given the limitations of mask-based therapies, over the last several years, we have performed studies exploring the use of oxygen, acetazolamide, theophylline, and, most recently, phrenic nerve stimulation (PNS). In general, these therapies are devoid of increasing intrathoracic pressure and are expected to be less reliant on patients’ adherence than PAP therapy. Long-term randomized clinical trials are needed, and, most recently, the NIH approved a phase 3 trial for a randomized placebo-controlled low flow oxygen therapy for treatment of CSA in HFrEF. This is a modified trial proposed by one of us more than 20 years ago!

Regarding PNS, CSA is characterized by intermittent phrenic nerve (and intercostal nerves) deactivation. It, therefore, makes sense to have an implanted stimulator for the phrenic nerve to prevent development of central apneas during sleep. This is not a new idea. In 1948, Sarnoff and colleagues demonstrated for the first time that artificial respiration could be effectively administered to the cat, dog, monkey, and rabbit in the absence of spontaneous respiration by electrical stimulation of one (or both) phrenic nerves (Sarnoff SJ, et al. Science. 1948;108:482). In later experiments, these investigators showed that unilateral phrenic nerve stimulation is also equally effective in man as that shown in animal models.

The phrenic nerves comes in contact with veins on both the right (brachiocephalic) and the left (pericardiophrenic vein) side of the mediastinum. Like a cardiac pacemaker, an electrophysiologist places the stimulator within the vein at the point of encounter with the phrenic nerve. Only unilateral stimulation is needed for the therapy. The device is typically placed on the right side of the chest as many patients may already have a cardiac implanted electronic device such as a pacemaker. Like the hypoglossal nerve stimulation, the FDA approved this device for the treatment of OSA. The system can be programmed using an external programmer in the office.

Phrenic nerve stimulation system is initially activated 1 month after the device is placed. It is programmed to be automatically activated at night when the patient is at rest. First, a time is set on the device for when the patient typically goes to bed and awakens. This allows the therapy to activate. The device contains a position sensor and accelerometer, which determine position and activity level. Once appropriate time, position, and activity are confirmed, the device activates automatically. Therapy comes on and can increase in level over several minutes. The device senses transthoracic impedance and can use this measurement to make changes in the therapy output and activity. If the patient gets up at night, the device automatically stops and restarts when the patient is back in a sleeping position. How quickly the therapy restarts and at what energy is programmable. The device may allow from 1 to 15 minutes for the patient to get back to sleep before beginning therapy. These programming changes allow for patient acceptance and comfort with the therapy even in very sensitive patients. Importantly, no patient activation is needed, so that therapy delivery is independent of patient’s adherence over time.

In the prospective, randomized pivotal trial (Costanzo et al. Lancet. 2016;388:974), 151 eligible patients with moderate-severe central sleep apnea were implanted and randomly assigned to the treatment (n=73) or control (n=78) groups. Participants in the active arm received PNS for 6 months. All polysomnograms were centrally and blindly scored. There were significant decreases in AHI (50 to 26/per hour of sleep), CAI (32 to 6), arousal index (46 to 25), and ODI (44 to 25). Two points should be emphasized: first, changes in AHI with PNS are similar to those in CANPAP trial, and there remained a significant number of hypopneas (some of these hypopneas are at least in part related to the speed of the titration when the subject sits up and the device automatically is deactivated, only to resume therapy in supine position); second, in contrast to the CANPAP trial, there was a significant reduction in arousals. Probably for this reason, subjective daytime sleepiness, as measured by the ESS, improved. In addition, PNS improved quality of life, in contrast to lack of effect of CPAP or ASV in this domain. Regarding side effects, 138 (91%) of 151 patients had no serious-related adverse events at 12 months. Seven (9%) cases of related-serious adverse events occurred in the control group and six (8%) cases were reported in the treatment group.—3.4% needed lead repositioning, a rate which is like that of cardiac implantable devices. Seven patients died (unrelated to implant, system, or therapy), four deaths (two in treatment group and two in control group) during the 6-month randomization period when neurostimulation was delivered to only the treatment and was off in the control group, and three deaths between 6 months and 12 months of follow-up when all patients received neurostimulation. Of 73 patients in the treatment group, 27 (37%) reported nonserious therapy-related discomfort that was resolved with simple system reprogramming in 26 (36%) patients but was unresolved in one (1%) patient.

Long-term studies have shown sustained effects of PNS on CSA with improvement in both sleep metrics and QOL, as measured by the Minnesota Living with Heart Failure Questionnaire (MLWHF) and patient global assessment (PGA). Furthermore, in the subgroup of patients with concomitant heart failure with LVEF ≤ 45%, PNS was associated with both improvements in LVEF and a trend toward lower hospitalization rates (Costanzo et al. Eur J Heart Fail. 2018; doi:10.1002/ejhf.1312).

Several issues must be emphasized. One advantage of PNS is complete adherence resulting in a major reduction in apnea burden across the whole night. Second, the mechanism of action prevents any potential adverse consequences related to increased intrathoracic pressure. However, the cost of this therapy is high, similar to that of hypoglossal nerve stimulation. Large scale, long-term studies related to mortality are not yet available, and continued research should help identify those patients most likely to benefit from this therapeutic approach.