The Prevalence of Central Sleep Apnea in New Referrals to a Sleep Surgery Clinic

Abstract

Objective

Surgical treatment of non‐obstructive sleep apnea (OSA) pathology poses the risk of inappropriate surgical indications. Herein, we sought to determine the prevalence of non‐OSA respiratory disorders, specifically central sleep apnea (CSA), in new referrals to a Sleep Surgery Clinic.

Study Design

Prospective observational review.

Setting

Tertiary care academic medical center.

Methods

In a sleep surgery clinic cohort, the presence of clinically significant CSA was defined as having >25% of the total apnea‐hypopnea index (AHI) being central and/or mixed events. Demographics, comorbid disorders, patient‐reported outcome measurements, and sleep study results were compared among patients using linear or logistic regression analysis, unadjusted and adjusted for age, sex, and body mass index (BMI).

Results

On average, the cohort (n = 295) was male (74%), middle‐aged (mean [±SD] 54.2 ± 13.9 years), and overweight (BMI 30.3 ± 5.4), with severe sleep apnea (AHI 30.6 ± 22.6 events/h). Twenty‐nine patients (9.8%) were found to have clinically significant CSA yet only 10% of these cases carried a diagnosis of CSA upon presentation. The remainder were identified by reviewing the pre‐visit sleep study tables (35%), raw data (17%), and tables and raw data of a repeat post‐visit study (38%). Patients with CSA were older and had evidence of more cardiac comorbidities.

Conclusion

The prevalence of CSA in new referrals to a Sleep Surgery Clinic was nearly 1 in 10 despite only 1% (3 of 295) with a known diagnosis upon presentation. Sleep surgeons must remain vigilant for patients with occult CSA, especially in older patients with a history of significant cardiovascular disease.

Obstructive sleep apnea (OSA) is a complex sleep disorder characterized by upper airway collapse despite respiratory effort, leading to drops in oxygen levels or arousals. Untreated OSA has been associated with an increased risk of both cardiovascular and neurocognitive comorbidities and mortality. ¹ , ² The gold standard treatment option for OSA has historically been positive airway pressure (PAP) therapy; however, many patients struggle to tolerate or adhere to using PAP adequately and turn to alternative therapies. Patients often will be referred to a “PAP Alternative Clinic” or a “Sleep Surgery Clinic” to learn about and pursue other treatment options, including mandibular advancement devices or oral appliance therapy, positional therapy, and upper airway surgery (including upper airway stimulation, pharyngeal surgery, and skeletal surgery). Interest in sleep surgery has increased over the years due to enhanced awareness of alternatives through marketing and research, newer techniques and additional treatment options, notably hypoglossal nerve stimulation therapy. ³

Although OSA is the most common type of sleep apnea, other types, including central sleep apnea (CSA), exist. CSA has a considerably lower prevalence than OSA, accounting for around 5% to 10% of sleep medicine clinic patients ⁴ and 1% of diagnoses on polysomnography. ⁵ CSA, however, is far more prevalent in patients with heart failure, occurring in 30% to 40% of this population. ⁵ , ⁶ CSA patient presentation can be similar to that of an OSA patient, with complaints of gasping, intermittent snoring, sleep fragmentation, sleepiness, and cardiovascular history. ⁴ , ⁷ There are many types of CSA including Cheyne‐Stokes, altitude‐induced, opioid‐induced, primary/idiopathic, treatment‐emergent, and other variants (neurologic related, transitional, postarousal events, congenital central hypoventilation syndrome, and more). ⁴ , ⁵ , ⁸ , ⁹ It has also been shown to be more common in patients who are male, >65 years old, with a lower body mass index (BMI) and with lower Epworth Sleepiness Scale (ESS) scores. ⁵ , ¹⁰ It is important to differentiate between central and obstructive sleep apnea, as treatment options are significantly different. Patients with CSA do not respond to OSA surgical interventions but instead can be treated with certain PAP therapies like an Adaptive‐Servo‐Ventilation (machine, oxygen, medications like acetazolamide or theophylline, or the Remedē device (ZOLL Respicardia Inc). ⁶ , ¹¹

As home sleep study testing and referrals to Sleep Surgery Clinics have increased, particularly in older patients since the advent of hypoglossal nerve stimulation, all providers should be aware of the possibility of patients with CSA (ie, not isolated OSA) being referred. The percentage of such “incorrect” referrals has not been previously studied nor reported. As such, the goal of the present manuscript was to comprehensively evaluate a prospective sample referred to a Sleep Surgery Clinic for the presence of clinically significant CSA, defined as having >25% central or mixed events, and to assess specific characteristics of these patients.

Methods

Study Design

This study was a prospective, observational review of all new patients who presented to our ENT Sleep Surgery Clinic. The study met the criteria for Institutional Review Board exemption (IRB #855469) at our institution.

Subjects

All subjects were new patients who consecutively presented to the Sleep Surgery Clinic of the senior author (R.C.D.) from July 2022 to August 2023. Patients were excluded if they were <18 years old, did not have sleep apnea (apnea‐hypopnea index [AHI] <5 events/h) and/or were not seeking PAP alternatives.

Procedures and Data Collection

Electronic Medical Record

Demographics and anthropometrics data including BMI, race, cardiac history, neurological history, and history of narcotic use were recorded for all patients. Patient Reported Outcome Measurements (PROMs) such as Sleep Rating Visual Analog Scale (VAS), Snoring Rating VAS, Epworth Sleepiness Scale (ESS), Nasal Obstruction Symptom Evaluation (NOSE), and Insomnia Severity Index (ISI) were also collected.

Sleep Study Data Review

Sleep study data including type of study, AHI, obstructive apnea index (oAI), central apnea index (cAI), mixed apnea index (mAI), supine AHI, non‐supine AHI, rapid eye movement (REM) and non‐REM (NREM) AHI, oxygen nadir, and percent of time with SpO₂ <90% were recorded. These metrics were reviewed and reported on the initial or most recent pre‐visit sleep study prior to the Sleep Surgery Clinic appointment. This was also performed on a repeat sleep study if ordered during the clinic appointment, either due to suspicion for CSA (older age, male sex, cardiac history, lack of reported snoring, and/or previsit study with elevated cAI/mAI but not meeting criteria) or due to need for an updated study (previsit study >2 years old). If the pre‐visit or repeat post‐visit study was performed at our institution, raw data was reviewed and assessed for central events including central hypopneas. Scoring was performed according to American Academy of Sleep Medicine (AASM) criteria (Rule 1B, 4% hypopnea) (AASM Scoring Manual, Version 3) and central hypopneas specifically were defined by criteria published by Javaheri et al. ⁷ Major criteria from AASM to differentiate central from obstructive hypopneas include lack of snoring during the event, lack of increased inspiratory flattening of the nasal pressure signal compared to baseline, and lack of associated thoracoabdominal paradox during the event. Central hypopneas were able to be scored but not counted or differentiated on our software's sleep study report. If the majority of hypopneas were central in etiology, the interpreting physician attributed the hypopnea index as central in origin. Post‐arousal central events were not scored if the sleep study was at our institution.

Clinically significant CSA (having >25% of the total AHI being central and/or mixed events) was encountered and distinguished at 2 possible timepoints: (1) at the initial new patient visit or (2) after a repeat sleep study (see Figure 2). During the first timepoint or initial visit, clinically significant CSA could be detected in 3 different ways using their pre‐visit sleep study: (1) a previous diagnosis of CSA as defined on the sleep study report with a central and mixed apnea index (cAI + mAI) > 50% of total AHI, (2) cAI + mAI > 25% of total AHI as seen within the sleep study tables, or (3) central hypopnea index and central and mixed apnea indices (cHI + cAI + mAI) > 25% of total AHI after review of the raw data. At the second timepoint, CSA could be diagnosed similarly with a (4) central hypopnea index and central and mixed apnea indices (cHI + cAI + mAI) > 25% of total AHI on repeat or post‐visit sleep study. Thus clinically significant CSA was detected by using either the pre‐ or post‐visit sleep study's report, tables, or raw data.

Figure 2.

Pathway to Diagnosis in CSA Cohort. AHI, apnea‐hypopnea index; cAI, central apnea index; cHI, central hypopnea index; CSA, central sleep apnea; mAI, mixed apnea index.

Statistical Analysis

Continuous characteristics are summarized using means and standard deviations and categorical traits are summarized using frequencies and percentages. To compare continuous and categorical variables between patients with CSA versus OSA, we utilized Student's t tests and χ ² tests, respectively, in unadjusted analyses and linear regression or logistic regression, respectively, and adjusted for age, sex, and BMI. Where presented, a P < .05 was considered statistically significant. Analyses were performed using Stata/SE v14.3 (StataCorp).

Results

Participant Characteristics

From July 2022 to August 2023, there were 345 new patient visits to our ENT Sleep Surgery Clinic. Fifty patients (14.5%) were excluded as they either presented for nasal obstruction or snoring intervention without a sleep study (n = 22) or did not have sleep apnea on their sleep study (n = 28). The study flow is presented as a Consolidated Standards of Reporting Trials (CONSORT) diagram (Figure 1). Of the remaining patients (n = 295), this total cohort (Table 1) was comprised of mostly males (74%) and was, on average, middle‐aged (54.2 ± 13.9 years) and obese (BMI 30.3 ± 5.4 kg/m²). Overall, the average AHI was 30.6 ± 22.6 events/h with a mean O₂ nadir of 79.5 ± 8.3%. Of the total cohort, the mean cAI and mAI were 0.9 ± 2.9 and 0.9 ± 2.9 events/h, respectively.

Figure 1.

CONSORT diagram. AHI, apnea‐hypopnea index; CONSORT, Consolidated Standards of Reporting Trials; CSA, central sleep apnea; OSA, obstructive sleep apnea.

Table 1.

Demographics, Anthropometrics, and Medical History

Characteristic	Total cohort (n = 295)	OSA cohort (n = 266)	CSA cohort (n = 29)	P value	P value adjusteda
Percentage	100	90.2	9.8
Age, y	54.2 ± 13.9	53.2 ± 13.7	63.1 ± 13.6	.0007
Male sex % (#)	74.2 ± 219	73.3 (195)	82.8 (24)	.269
BMI, kg/m2	30.3 ± 5.4	30.3 ± 5.4	30.4 ± 5.4	.908
Race/ethnicity, %				.929
White	69.5	69.6	69.0
Black	14.2	13.9	17.2
Asian	4.4	4.5	3.5
Hispanic	2.0	1.9	3.5
Other/unknown	9.8	10.2	6.9
Cardiac history, %	17.0	14.3	41.4	.0002	.056
Neurologic history, %	8.2	7.6	13.8	.244	.531
On narcotic medication, %	2.0	1.9	3.5	.570	.900

Bold formatting signifies statistical significance.

Abbreviations: BMI, body mass index; CSA, central sleep apnea; OSA, obstructive sleep apnea.

^a P value adjusted for age, sex, and BMI.

Prevalence of CSA

In the primary analysis, 29 of the 295 patients (9.8%) were found to have clinically significant CSA. As clinically significant CSA was defined in 4 ways (see Methods), we divided the CSA cohort into 4 groups based on the method of diagnosis (see Figure 2). Of these 29 patients, 3 patients (10%) had either a known diagnosis of CSA or >50% of total events were central/mixed on their pre‐visit sleep study (Group 1). Ten patients (35%) were found to have a cAI + mAI > 25% of total AHI on review of the report of the initial/pre‐visit sleep study (Group 2). Five patients (17%) were noted to have a cAI + cHI + mAI > 25% of the total AHI on review of their raw data (from the pre‐visit sleep study) (Group 3); these 5 patients were not noted to have CSA on the report of their sleep study as central hypopneas are not typically scored or differentiated. As they were performed at our institution, the raw data of the sleep studies were able to be reviewed and found to have predominantly central and not obstructive hypopneas. Lastly, 11 patients (38%) were diagnosed with CSA based on a repeat or post‐visit study that had cAI + cHI + mAI > 25% of the total AHI (Group 4). All post‐visit repeat sleep studies were completed at our institution; 8 were in lab polysomnograms (PSGs) and 3 were home sleep tests (HSTs), with flow and effort belt signals. Of these 11 patients, 6 had predominantly central hypopneas account for the diagnosis of clinically significant CSA (cHI + cAI + mAI > 25% AHI) while the remaining 5 had higher cAI and mAI (cAI + mAI > 25% AHI).

Differences in Characteristics of CSA and OSA Patients

We performed exploratory analyses examining the characteristics of patients with CSA versus those with OSA (see Table 1). There were no statistically significant differences in sex, BMI or race between the OSA and CSA cohorts. However, CSA patients were ∼10 years older than OSA patients (63.1 ± 13.6 vs 53.2 ± 13.7 years, P < .01). There were no differences between OSA and CSA patients regarding neurologic history or narcotic usage, but patients with CSA were more likely to have a history of cardiovascular disease (including heart failure, heart arrhythmia, myocardial infarction, stroke, cardiomyopathy) than OSA patients (41.4% vs 14.3%, respectively, P = .0002, P _adj = .0563). This was only significant when age, sex, and BMI were not adjusted.

In addition, there was no significant difference in sleep‐related PROM's (sleep rating VAS, snoring rate VAS, NOSE, ISI) except for ESS when adjusted for by age, sex, and BMI (see Table 2). Patients in the CSA cohort were found to have more severe OSA with an average AHI of 38.3 ± 19.8 compared to OSA patients with an average AHI of 29.9 ± 22.7 but only when not adjusted for age, sex, and BMI (P = .046, P _adj = .1456). CSA patients also had higher cAI (5.4 ± 7.2 vs 0.4 ± 1.1, P = .0017, P _adj = <.0001) and mAI (3.8 ± 6.9 vs 0.5 ± 1.8, P = .023, P _adj = <.0001) than OSA patients. Among sleep study measurements, supine AHI was higher in CSA patients (59.7 ± 24.2 vs 41.8 ± 27.7, P = .0025, P _adj = .0018).

Table 2.

Subjective Sleep Metrics and Sleep Study Results

Characteristic	Total cohort (n = 295)	OSA cohort (n = 266)	CSA Cohort (n = 29)	P value	P value adjusteda
Sleep rating VAS	63.1 ± 24.9	63.3 ± 24.5	61.6 ± 28.3	.760	.848
Snoring rating VAS	62.9 ± 35.2	63.7 ± 34.9	55.4 ± 38.0	.270	.412
ESS	8.35 ± 5.38	8.18 ± 5.36	9.90 ± 5.39	.112	.027
NOSE	35.9 ± 23.3	36.2 ± 23.8	33.6 ± 18.4	.495	.835
ISI	13.4 ± 6.1	13.3 ± 6.1	14.3 ± 6.0	.406	.140
AHI/REI	30.6 ± 22.6	29.9 ± 22.7	38.3 ± 19.8	.046	.146
oAI	14.0 ± 17.0	13.9 ± 17.3	14.6 ± 14.1	.838	.814
cAI	0.928 ± 2.930	0.417 ± 1.140	5.388 ± 7.191	.002	<.0001
mAI	0.863 ± 2.924	0.520 ± 1.753	3.788 ± 6.880	.023	<.0001
Supine AHI	43.5 ± 27.9	41.8 ± 27.7	59.7 ± 24.2	.003	.002
Non‐supine AHI	21.1 ± 23.9	20.6 ± 24.1	24.9 ± 22.1	.395	.404
REM AHI	33.5 ± 26.9	34.3 ± 26.4	27.5 ± 30.8	.441	.263
NREM AHI	31.1 ± 28.5	30.6 ± 29.4	34.3 ± 22.4	.590	.699
O2 nadir, %	79.5 ± 8.3	79.5 ± 8.4	80.3 ± 6.7	.552	.271
T90, %	10.9 ± 15.3	10.8 ± 15.4	12.1 ± 14.0	.673	.910
Sleep study type, %				.108	.646
PSG	40.0	39.5	44.8
HST	58.0	59.0	48.3
Other	2.0	1.5	6.9

Bold formatting signifies statistical significance.

Abbreviations: AHI, apnea‐hypopnea index; BMI, body mass index; cAI, central apnea index; ESS, Epworth Sleepiness Scale; HST, home sleep test; ISI, insomnia severity index; mAI, mixed apnea index; NOSE, Nasal Obstruction Symptom Evaluation; NREM, non‐REM; oAI, obstructive apnea index; OSA, obstructive sleep apnea; PSG, polysomnogram; REM, rapid eye movement; REI, respiratory event index; T90, percent of time SpO₂ < 90%; VAS, Visual Analog Scale.

^a P value adjusted for age, sex, and BMI.

Sleep Study Characteristics

In regards to the type of pre‐visit sleep study, 40% of patients in this cohort had PSGs, 58% had HSTs, and 2% had a type that did not report or specify on type of apnea (ie, WatchPAT). There was no significant difference in the type of pre‐visit sleep study performed between OSA and CSA cohorts (see Table 2).

Of the total 295 patients, 91 (30.8%) had new sleep studies ordered during the initial patient visit, either due to the need for an updated study or suspicion of CSA. Fourteen of these studies (15.4%) were positive for clinically significant CSA: 11 new diagnoses (Group 4) and 3 confirmed diagnoses (Group 2). Not all ordered studies were completed.

Of the 29 patients found to have clinically significant CSA, 17 patients had a new sleep study ordered at the Sleep Surgery consultation. Only 2 studies were ordered due to needing an updated sleep study (prior previsit study >2 years old). The remainder of the repeat sleep studies were ordered due to suspicion by the sleep surgery provider team. Eight patients had a new study ordered because of their prior study having some element concerning CSA (cAI + mAI > 5 or almost 25% of events) or to confirm findings with a PSG. The remaining 7 studies were ordered because of suspicion related to the patient's reported cardiac history, older age, male sex, and/or lack of reported snoring.

Discussion

This is the first study to describe the prevalence of central sleep apnea among patients presenting as OSA referrals in a Sleep Surgery Clinic. Overall, 10% of new patient referrals had clinically significant CSA that would preclude patients from having most OSA surgical interventions. Only 10% (3) of the central apnea cohort presented with a diagnosis of CSA and thus, 90% (26) were incorrectly labeled on presentation to their new patient visit. To avoid inappropriate surgical recommendations, providers need to be vigilant in assessing the sleep study report, tables and raw data (if available). It is also prudent for providers to be aware of the possibility of central hypopneas as these are not often differentiated or reported. ¹² In our study, 11 of 29 patients in the CSA cohort (37.9%) had central hypopneas account for the majority of central/mixed events.

CSA in our study appeared as frequently as previously reported in general sleep medicine clinics. ⁴ This is surprising given that patients who present to Sleep Surgery Clinics are typically referred for only CPAP alternatives for treating OSA and not general sleep complaints. Similar to previous studies, we found the CSA cohort to be older in age and with a higher likelihood of having cardiovascular comorbidities. Although not significant when adjusted for age, sex, and BMI, the adjusted Pvalue for cardiac comorbidities between groups was close at 0.0563. This difference could be due to the fact that as patients age, they have a higher likelihood of developing cardiovascular disease. In contrast, we did not find a significant difference in sex between cohorts (CSA patients included ∼10% less females), although this may reflect the fact that male gender is a risk factor for both CSA and OSA. ¹ Since the development of upper airway stimulation, there has been a shift toward older patients undergoing surgical treatment for OSA. Sleep surgery providers should be particularly mindful of possible CSA in these older patients, especially with a cardiac history, as upper airway surgery does not treat CSA (see Table 3 for risk factors).

Table 3.

Risk Factors for Central Sleep Apnea

Patient characteristics	Sleep study report variables
Older age (>60 y old)	Frequent central and/or mixed apneas
Male gender	Positional sleep apnea
Cardiovascular comorbidities	NREM‐predominant sleep apnea
Reported lack of snoring

Abbreviation: NREM, non‐rapid eye movement.

In regards to sleep study metrics, there are certain clues within the sleep study tables that can be associated with CSA and/or OSA. In our study, the supine AHI was higher in patients with clinically significant CSA. Although not frequently discussed, CSA, like OSA, is often positional and worse in the supine position. ⁴ , ¹³ , ¹⁴ It should not be assumed that positional sleep apnea is obstructive. Prior studies have shown that CSA is often worse during NREM sleep, ⁷ , ¹² where normal breathing is controlled by the level of partial pressure of carbon dioxide, while REM sleep can override metabolic control of ventilation. ⁷ While AHI was higher in the CSA cohort during NREM and lower during REM sleep, these differences were not statistically significant compared to those with OSA. Still, elevated NREM AHI may be a marker of underlying CSA.

As discussed previously, repeat sleep studies were ordered at the end of the new patient visit for certain patients due to suspicions of CSA. These suspicious traits included: older age (>65 years old), cardiac history, higher cAI/mAI, and reported lack of snoring. As stated above, older age and cardiac history were 2 common risk factors in our CSA cohort. While we did not find a difference in reporting snoring severity, we surmise that the hyperpneic breaths between central events which produce a “snore‐like” sound could have dampened the difference in reported snoring between groups. ⁷

There are important limitations of this study. First, central hypopneas are rarely reported or differentiated within sleep reports and sleep study scoring and reporting at our institution only report the total hypopneas or hypopnea index even if central hypopneas were scored. As we were only able to review the raw data of studies performed at our institution, we could be underestimating the prevalence of central hypopneas and, thus, CSA in our sample. We could not verify the scoring of sleep studies performed at outside sleep centers and confirm that central events were not post‐obstructive arousal events. Central hypopneas can also be scored differently whether using AASM criteria alone or that set forth by Jahaveri et al. ⁷ In addition, as insurance companies do not always approve PSGs, HSTs were commonly performed in the total cohort (58%). Since in‐laboratory PSGs are better at differentiating between obstructive and central events, we could be missing patients with central sleep apnea among those who had HSTs. Finally, we only ordered repeat sleep studies in those who were “suspicious” of CSA or who needed an updated study, which could result in selection bias and underdiagnosis of true central apnea prevalence.

Our study also has several strengths worth stating, including the large sample size accrued within a 13‐month period. Greater than half (51%) of patients had sleep studies at our institution, allowing access to the raw data. In these patients, we were able to distinguish between obstructive and central hypopneas. Additionally, the sleep studies were reviewed by board‐certified sleep medicine specialists (J.G.R. and R.C.D.). Finally, we defined clinically significant CSA as >25% central or mixed events, consistent with surgical indications for Food and Drug Administration‐approved surgical therapies for OSA such as Inspire® (Maple Grove). Therefore, the patients in our CSA cohort would be ineligible to receive Inspire implantation. For these patients, we first try to identify and treat the possible cause for their central events (ie, refer to cardiology), and often retrial and optimize them on different PAP therapies. Alternatively, providers could consider other PAP alternatives like Remedē (if they qualify) and infrequently could consider different upper airway surgeries if the majority of their other events (>50%) were obstructive.

Conclusions

Central sleep apnea is relatively common (10%) in “OSA” patient referrals to a Sleep Surgery Clinic, despite only 1% of our sample (3 of 295) labeled as having known CSA on initial presentation. Those with CSA were generally older, with a cardiac history, and had more severe sleep apnea (especially in the supine position). Our results highlight that providers should also be aware of central hypopneas, as these are generally not characterized on sleep study reports. Thus, if patients present with CSA risk factors but no known diagnosis, sleep surgeons should consider obtaining an updated PSG with special emphasis on ruling out central diathesis. Differentiation and correct identification of clinically significant CSA is critical to avoid inappropriate surgical treatment.

Author Contributions

Julianna G. Rodin, design, conduct, analysis, presentation, manuscript writing; Tice Harkins, conduct, analysis; Erica Kent, conduct, analysis; Chau Phung, conduct, analysis; Rafa Khan, conduct, analysis; Everett Seay, conduct, analysis; Brendan T. Keenan, analysis, manuscript writing; Raj C. Dedhia, design, conduct, manuscript writing.

Disclosures

Competing interests

Julianna G. Rodin—consultant Inspire; Everett Seay—consultant Inspire; Raj C. Dedhia—grant research funding by NIH, Cryosa, Inspire Medical, Nyxoah Medical; clinical advisory board for Lunair Medical.

Funding source

This work was supported by the National Institute of Health (1R01HL144859‐04).