High nocturnal periodic breathing reported by PAP adherence data predicts decompensation of heart failure

Abstract

Study Objectives:

Obstructive sleep apnea (OSA) often coexists with heart failure (HF) and is commonly treated with positive airway pressure (PAP) therapy. Periodic breathing (PB) may be present in HF and is an indicator of poor prognosis, but there is no easy way to detect PB in an outpatient setting. However, it can be detected by analyzing PAP usage data. The study aimed to assess if high PB% detected by PAP machine could predict impending HF exacerbation and if better PAP adherence is associated with reduced hospitalization and mortality.

Methods:

We retrospectively reviewed medical records of 115 patients with OSA from the sleep clinic of our VA Medical Center. The cross-sectional data on demographics, labs, PAP adherence, PB% in the previous 30 days, echocardiogram in the previous 6 months, and hospitalizations and mortality in the subsequent 180 days were extracted. Based on left ventricular ejection fraction (LVEF), patients were classified into (1) HF with normal–midrange LVEF (LVEF ≥40%, n = 74) and (2) HF with reduced LVEF (LVEF < 40%, n = 41). Pairwise correlation and linear regressions were done to assess predictors of PB%. Binomial and logistic regressions assessed the relationship of PB% and PAP adherence with hospitalization from HF and all-cause mortality.

Results:

In the HF with reduced LVEF group, the mean PB% was 2.6 times higher (P < .001) and PAP adherence was 29% lower (P < .001). PB% positively correlated with brain natriuretic peptide level (r = .447, P < .01) and number of hospitalizations (r = .331, P < .01). Higher PB% negatively correlated with LVEF (r = –.423, P < .01) and estimated glomerular filtration rate (r = –.246, P < .01). Every 10% increase in PAP adherence decreased odds of hospitalization by 0.78 times (P < .001) and odds of death by 0.86 (P = .043).

Conclusions:

High PB% detected by PAP machine data is a predictor of impending HF exacerbation and hospitalization. Improved PAP adherence and optimization of medical therapy may reduce hospitalization and all-cause mortality.

Citation:

Ullah MI, Tamanna S, Bhagat R. High nocturnal periodic breathing reported by PAP adherence data predicts decompensation of heart failure. J Clin Sleep Med. 2023;19(3):431–441.

BRIEF SUMMARY

Current Knowledge/Study Rationale: High nocturnal periodic breathing percentage (PB%) is known to be associated with decompensated heart failure, which often results in hospitalization. However, there is no simple way to detect change in PB in the outpatient setting.

Study Impact: Positive airway pressure (PAP) machine adherence data and PB% can be reliably estimated in patients with obstructive sleep apnea and co-existent heart failure during routine clinic visits. This information may be crucial to focus on optimization of medical therapy and improvement in PAP adherence, which could help reduce hospitalization and mortality.

INTRODUCTION

Heart failure (HF) is a major public health problem in the United States,¹ and is responsible for approximately 12 to 15 million office visits and 6.5 million hospital days each year.² Patients hospitalized for HF have a 15% mortality and 30% readmission rate within 30 to 60 days after discharge.³ The prevalence of both obstructive sleep apnea (OSA) and central sleep apnea (CSA) are higher in patients with HF (47–81%) than in the general population (3–28%).^4,5 CSA and periodic breathing (PB) often occur in HF.⁶ The mortality rate is higher in patients with HF who develop PB than those without it.⁷ The recurrent hospitalizations from HF are significant drivers of health care costs in the United States.⁸ Advanced age, ischemic heart disease, chronic kidney disease, untreated OSA, and Black race are some of the important risk factors for hospitalization from HF.^9,10 Blacks are more likely to have a higher prevalence of OSA and HF and lower continuous positive airway pressure (CPAP) adherence,^11,12 which may lead to a higher risk of morbidity and mortality from acute HF.

PB is characterized by recurrent central apneas, alternating with hyperventilation and hypoventilation in a crescendo-decrescendo pattern. A progressively increasing percentage of PB (PB%) in sleep is a marker of decompensating HF with a poor prognosis.¹³ This may be observed clinically by nurses or doctors when a patient is hospitalized or during a routine in-laboratory polysomnography. However, there is no easy way to measure the percentage of PB while the patient is at home, and who could be on the verge of needing hospitalization from HF exacerbation. OSA and CSA associated with HF are commonly treated with CPAP or bilevel positive airway pressure (BPAP) therapy at home. The new-generation positive airway pressure (PAP) machines can identify if patients are still having residual obstructive or central apnea. The time spent having PB during sleep is recorded by the machine as a percentage of total PAP usage per night. Therefore, examining the PAP machine adherence data may reveal increases in PB% that could be a marker of worsening HF.

During routine clinic follow-up visits at the sleep clinic of Veterans Affairs Medical Center (VAMC) Jackson, we observed that some patients with a high PB index, as shown from PAP data, also complained of worsening symptoms of HF needing urgent referral to their primary care provider for evaluation and intervention for acute exacerbation of HF. Others with severely decompensated HF were referred to the emergency room. These observations prompted us to initiate this study. Previous studies have shown improvement in left ventricular ejection fraction (LVEF) among patients with HF and OSA who were treated with PAP.¹⁴ However, adherence to PAP therapy is a major challenge in many patients, and it is not clear if an improved PAP adherence could be an effective way to prevent worsening of HF. We hypothesized that a high PB% detected by PAP machine could predict impending HF exacerbation and improved PAP adherence could reduce hospitalization and mortality.

METHODS

Patient selection

The research was approved by the institutional review board of the G.V. (Sonny) Montgomery Veterans Affairs (VA) Medical Center at Jackson, Mississippi. Veterans diagnosed with OSA by in-laboratory polysomnography and initiated on CPAP/BPAP (Respironics DreamStation, Phillips, USA) equipped with a modem provided by the VA Sleep Laboratory were included in the study. The modem made the adherence data accessible to monitor the PAP usage during their routine follow-up visits in the sleep clinic. Sleep physicians download these data during each visit into the electronic medical record.

Inclusion criteria

Veterans with OSA and history of HF who were seen for a routine follow-up visit in the sleep clinic (index visit) from January 1, 2016, through November 8, 2018, and had the following information available on their electronic medical record were included in the study:

1. Demographic data: age, sex, race, body mass index (BMI)

2. Modem data: PAP adherence (percentage of nights with > 4 hours of PAP use) and percentage of PB/night of PAP use in the last 30 days

3. Clinical data: Echocardiogram documenting LVEF within the last 180 days, serum creatinine (with calculated estimated glomerular filtration rate [eGFR]), and brain natriuretic peptide (BNP) level obtained within 30 days before the index visit. If any participant had > 1 laboratory value within this time frame, only the latest one was recorded.

Exclusion criteria

Patients receiving any narcotic pain medication or dialysis were excluded from the study.

Data extraction

A total of 115 patients were identified who met the above inclusion and exclusion criteria. If a patient had multiple sleep clinic visits during this time frame (January 1, 2016, through November 8, 2018), only the earliest visit was used as the “index visit” for data collection. We looked at each patient’s index visit progress note retrospectively in the electronic medical record and recorded his/her personal data, including age, sex, race, BMI, and presence or absence of chronic atrial fibrillation (from the problem list). We also recorded the PAP adherence data and average PB%, serum creatinine, and BNP in the previous 30 days of index visit. The LVEF from the echocardiogram done in the prior 6 months was recorded from the electronic medical record. Finally, we also documented the number of hospitalizations and deaths (if any) within the 180 days after the index visit.

Definitions of OSA and PB

OSA was defined using the American Academy of Sleep Medicine standard criteria of an apnea-hypopnea index ≥ 15 events/h or apnea-hypopnea index ≥ 5 events/h in patients who reported any of the following: fatigue, excessive daytime sleepiness, unrefreshing sleep, insomnia, gasping or choking, or the bed partner describing snoring or breath interruption during sleep, or history of hypertension, mood disorder, cognitive dysfunction, coronary artery disease, stroke, HF, atrial fibrillation, stroke, or type 2 diabetes.¹⁵

PBs are clusters of breaths separated by intervals of apnea or hypopnea (Figure 1). Cheyne-Stokes breathing (CSB) is a form of PB characterized by a crescendo-decrescendo pattern of respiration with waxing and waning amplitude of tidal volume separated by periods of central apnea or central hypopnea. This term (CSB) has historically been used to describe PB with apneas in the context of HF. The European Respiratory Society task force on CSA recommended replacing the historical term “CSB” with “periodic breathing” in the context of HF.^16,17 We have used the term “periodic breathing” (PB) in our manuscript to denote CSA with CSB. The new-generation CPAP and BPAP machines report the PB% (PB as a percentage of total usage of machine per night; eg, for a patient using the machine for 500 minutes on a given night with PB occurrence for a total of 50 minutes, the PB% was 10% on that night).

Figure 1. A sample of PAP machine adherence data showing periodic breathing from one of the patients in the study.

PAP = positive airway pressure.

Determination of HF status

The patients with OSA with a history of HF in their medical record were included in the study. Then, they were divided into 2 groups based on their LVEF.¹⁸ Those who had LVEF ≥ 40% (n = 74) were classified as HF with normal to midrange ejection fraction (HFnmEF). Those with LVEF < 40% (n = 41) were classified as HF with reduced ejection fraction (HFrEF).

Statistical analysis

The characteristics of the 2 HF groups (HFrEF and HFnmEF) and 2 races (Black and White) were compared using chi-square tests. Correlations among different variables were assessed using pairwise correlation. A multivariate linear regression was done to see what factors predicted PB%. Negative binomial regressions were done to estimate the likelihood of hospital admissions from HF exacerbation within 180 days of the index clinic visit. Logistic regression analyses were also performed to estimate the odds of death within 180 days. PB% was the primary predictor in each univariate analysis. In bivariate models, race, PAP adherence, type of HF, age, and atrial fibrillation were added one at a time to adjust for these potential confounders. A P value < .05 was considered to be statistically significant. Model fit was determined by the Hosmer-Lemeshow goodness-of-fit test. Statistical analyses were performed using STATA software, version 17 (StataCorp, College Station, TX).

RESULTS

Of 115 patients, 41 had HFrEF and 74 had HFnmEF (Table 1). There was no significant difference in mean age between the 2 HF groups. Thirty-seven patients (32.17%) had atrial fibrillation in our study sample. The prevalence of atrial fibrillation was slightly higher in the HFrEF group (34.15%, n = 14) than in the HFnmHF group (31.08%, n = 23) (P = .736).

Table 1.

Characteristics of patients by heart failure status.

Variable	HFrEF (n = 41)		HFnmEF (n = 74)		P
	Mean	SD	Mean	SD
Age, y	67.95	9.33	68.23	10.39	.887
BMI, kg/m2	32.54	8.05	35.87	7.15	.024*
BNP, pg/mL	1008.1	878.55	208.7	284.06	<.001*
Periodic breathing, %/night	24.16	22.47	8.64	12.41	<.001*
eGFR, mL/min/1.73 m2	54.71	27.5	64.69	29.25	.076
PAP adherence, %	57.07	35.7	87.57	23.87	<.001*
LVEF, %	24.63	8.44	52.59	5.37	<.001*

^*P < .05. BMI = body mass index, BNP = brain natriuretic peptide, eGFR = estimated glomerular filtration rate, HFnmEF = HF with normal or midrange ejection fraction, HFrEF = heart failure with reduced ejection fraction, LVEF = left ventricular ejection fraction, PAP = positive airway pressure, SD = standard deviation.

BMI was slightly higher in patients with HFnmEF (35.87 kg/m²) than in the HFrEF group (32.54 kg/m²) (P = .024). In the HFnmEF group, PAP adherence (P < .001) was 29% higher, whereas in the HFrEF group, the mean BNP was 4.2 times higher (P < .001) and the mean PB% was 2.6 times higher (P < .001). The eGFR (CKD-EPI formula)¹⁹ was lower in the HFrEF group, but this was not statistically significant.

We had 63 Black and 52 White patients in our study. Demographic data were analyzed to compare age, BMI, HF type, and other laboratory variables between the 2 races (Table 2). The Black patients were 4.01 years younger than White patients, on average (P = .032). The mean BNP was 2.98 times higher in Black than in White patients (P = .001). PAP adherence and LVEF were significantly lower in Black patients. BMI, eGFR, and PB% were similar in these 2 groups.

Table 2.

Characteristics of patients by race.

Variable	White (n = 52)		Black (n = 63)		P
	Mean	SD	Mean	SD
Age, y	70.33	9.13	66.32	10.36	.032*
BMI, kg/m2	34.69	7.14	34.67	8.05	.988
BNP, pg/mL	267.29	428.57	680.59	797.08	.001*
Periodic breathing, %/night	12.11	16.36	15.87	19.57	.272
eGFR, mL/min/1.73 m2	58.13	22.98	63.61	33.01	.314
PAP adherence, %	84.37	28.61	70.36	32.34	.027*
LVEF, %	46.15	11.73	39.71	16.74	.021*

^*P < .05. BMI = body mass index, BNP = brain natriuretic peptide, eGFR = estimated glomerular filtration rate, LVEF = left ventricular ejection fraction, PAP=positive airway pressure, SD = standard deviation.

Correlation of PB% with variables

Table 3 shows pairwise correlations between PB% and different variables in all patients. PB% was positively correlated with an increase in BNP level (r = .447, P < .01) and the number of hospitalizations from acute HF (r = .331, P < .01). The increase in PB% also correlated with decreased LVEF (r = –.423, P < .01) and lower eGFR (r = –.246, P < .01). The number of hospitalizations was positively correlated with the increase in PB% and BNP, and negatively correlated with PAP adherence and LVEF. Increased PB% predicted higher hospitalization numbers and better PAP adherence had a protective effect for hospitalization. Lower eGFR was associated with increased age (r = –.309, P < .01), higher BNP (r = –.337, P < .01), higher PB% (r = –.246, P < .01), and higher number of hospitalizations (r = –.236, P < .05). Higher PAP adherence was correlated with lower PB% (r = –.064, P = .495), but this was not statistically significant (Figure 2).

Table 3.

Pairwise correlation among variables.

Variable	PB%	BNP	Adherence	LVEF	#Adm	eGFR	Age
PB%	1.000
BNP (pg/mL)	0.447***	1.000
PAP adherence (%)	−0.064	−0.423***	1.000
LVEF (%)	−0.423***	−0.595***	0.476***	1.000
#Adm	0.331***	0.517***	−0.579***	−0.570***	1.000
eGFR (mL/min/1.73 m2)	−0.246***	−0.337***	0.129	0.187**	−0.236**	1.000
Age (years)	0.107	0.111	−0.146	0.017	0.101	−0.309***	1.000

^***P < .01, **P < .05. adherence = PAP adherence in last 30 days, #Adm = number of hospital admissions from acute heart failure, BNP = brain natriuretic peptide, eGFR = estimated glomerular filtration rate, LVEF = left ventricular ejection fraction, PAP = positive airway pressure, PB = periodic breathing.

Figure 2. Correlation of PB% with PAP adherence.

CI = confidence interval, PAP = positive airway pressure, PB = periodic breathing.

Factors predicting PB

A multivariate linear regression was done to see what variables predicted PB% and revealed that LVEF and BNP were strong predictors of PB%. LVEF (coefficient = –0.514, P < .001) was inversely associated with PB%, whereas BNP (coefficient = 0.012, P < .001) was positively associated with PB% (Figure 3). The eGFR was inversely associated with PB% (coefficient = –0.115, P = .035) and there were significant interactions between eGFR and LVEF in predicting PB%. Figure 4 denotes linear predicted values of PB% on the y axis based on eGFR on the x axis in the 2 HF groups. It also shows the statistical interactions between PB% and eGFR in the 2 HF groups by displaying 2 different slops for the 2 groups. The predicted PB% increased more rapidly with lower eGFR in patients with HFrEF compared with those with HFnmEF (Figure 4). Age, race, BMI, presence of atrial fibrillation, and PAP adherence did not significantly predict PB% and were dropped from the model.

Figure 3. Relationship of LVEF and BNP levels with changes in PB%.

BNP = brain natriuretic peptide, LVEF = left ventricular ejection fraction, PB = periodic breathing.

Figure 4. Comparison of predicted PB% by eGFR between HFnmEF and HFrEF groups.

eGFR = estimated glomerular filtration rate, HF = heart failure, HFnmEF = HF with normal or midrange ejection fraction, HFrEF = heart failure with reduced ejection fraction, PB = periodic breathing.

PB predicting hospitalization

A total of 50 participants had no hospitalization in 180 days subsequent to the index visit in the sleep clinic. Of these, 45 belonged to the HFnmEF group (Table 4). The proportion of multiple hospitalizations (> 2) was 8 times higher in the HFrEF group (n = 18, 43.9%) compared with in the HFnmEF group (n = 4, 5.4%). Univariate and bivariate negative binomial regressions (Table 5) were done using PB% as the primary predictor of hospitalization within 180 days from the date of the index office visit. The univariate model showed that the risk of hospitalization increases sharply with higher PB% (Figure 5). The bivariate model (Table 5) revealed that the risk of hospital admission from acute HF exacerbation with higher PB% was most significantly influenced by race, type of HF, and PAP adherence. After adjusting for PB%, Black patients were almost twice as likely to be admitted for acute HF (incidence rate ratio [IRR] = 1.89, P = .001). The presence of HFrEF increased the likelihood of admission by 7.45 times (P < .001). Every 10% increase in PAP adherence decreased the likelihood of admission by 0.78 times (P < .001).

Table 4.

Number of hospitalizations by heart failure status.

Number of Hospitalizations	HF Status
	HFnmEF	HFrEF	Total
0	45	5	50
	60.81%	12.20%	43.48%
1	9	7	16
	12.16%	17.07%	13.91%
2	16	11	27
	21.62%	26.83%	23.48%
3	2	15	17
	2.70%	36.59%	14.78%
4	2	3	5
	2.70%	7.32%	4.35%
Total	74	41	115

The top rows per entry show frequencies, and the bottom rows per entry show column percentages. HF = heart failure, HFnmEF = heart failure with normal to midrange ejection fraction, HFrEF = heart failure with reduced ejection fraction.

Table 5.

Negative binomial regression predicting hospital admissions from increase in PB%.

Variable	Univariate Model			Bivariate Model
	Coefficient	IRR	P	Coefficient	IRR	P
Periodic breathing (%)	.0237415	1.02	.001	—	—	—
Race (Black vs White)	.679948	1.97	.03	.6379596	1.89	.033
Every 10% increase in PAP adherence (%)	−.241369	0.78	<.001	−.256196	0.78	<.001
HFrEF vs HfnmEF	2.110319	8.25	<.001	2.008535	7.45	<.001
Age	.0138551	1.01	.388	.0129279	1.01	.391
Atrial fibrillation	.0526437	1.05	.873	.0321431	1.03	.962

In bivariate models, the variable “periodic breathing” was kept constant and other variables were added one at a time. HFnmEF = heart failure with normal to midrange ejection fraction, HFrEF = heart failure with reduced ejection fraction, IRR = incidence rate ratio, PAP = positive airway pressure, PB = periodic breathing.

Figure 5. Predicted incidence of hospitalization from acute heart failure by PB%.

CI = confidence interval, PB = periodic breathing.

PB predicting the risk of death

In our study, there were 35 deaths (14 White and 21 Black patients) within 180 days of the index clinic visit. Black patients had a higher proportion of participants with HFrEF (46%) than did White patients (23%). Logistic regression predicting the odds of death (Table 6) indicated that neither PB% nor race was a significant predictor of death in 180 days from the index visit. After adjusting for age, PAP adherence and the presence of HFrEF were most significant in predicting death in the multivariate model. The patients with HFrEF were 2.78 times more likely to die and every 10% increase in PAP adherence decreased the odds of death by 0.8 times.

Table 6.

Logistic regressions predicting odds of death in 180 days.

Variable	Univariate Model			Bivariate Model
	Coefficient	OR	P	Coefficient	OR	P
Periodic breathing (%)	.0181805	1.02	.089	—	—	—
Race	.3053816	1.36	.458	.2403505	1.27	.565
Every 10% increase in PAP adherence (%)	−.218283	0.80	.001	−.0146447	0.86	.043
HFrEF vs HFnmEF	1.504077	4.5	.001	1.024191	2.78	.045
Age	.0268781	1.03	.207	.0237905	1.02	.317
Atrial fibrillation	.6817185	1.98	.107	.6170716	1.85	.151

In bivariate models, the variable “periodic breathing” was kept constant and other variables were added one at a time. HFnmEF = heart failure with normal to midrange ejection fraction, HFrEF = heart failure with reduced ejection fraction, OR = odds ratio, PAP = positive airway pressure.

DISCUSSION

PB is commonly observed in patients with HF, and it is associated with a poor prognosis.^13,20 There is no specific device to monitor the pattern of PB for patients at home who may be on the verge of having HF decompensation. Our study shows that an increasing PB index detected by a CPAP/BPAP machine could help identify impending acute HF needing hospitalization (Figure 6). This is consistent with findings by Prigent et al²¹ who reported that a progressive increase in apnea-hypopnea index, particularly the PB/Cheyne-Stokes respiration detected by CPAP telemonitoring, was associated with increased risk developing serious cardiac events, including HF, ventilatory instability, and arrhythmia. Interestingly, the cycle length of PB detected by a CPAP device is reportedly significantly longer in both patients with stable HF and those with acutely decompensated HF compared with controls.²²

Figure 6. Schematic flow diagram of how increased PB% detected by PAP machine in patients with HF and OSA may allow early intervention and help decrease hospitalization and mortality.

CA = central apnea, CPAP = continuous positive airway pressure, GFR = glomerular filtration rate, HF = heart failure, OSA = obstructive sleep apnea, PAP = positive airway pressure, PB = periodic breathing, PCP = primary care physician.

The prevalence of central apnea and PB is high in patients with HFrEF.^5,23 Coordination of breathing is influenced by afferent signals including the carotid and brainstem chemoreflex sensors, which detect fluctuations in PaO₂, PaCO₂, and pH, resulting in modification of ventilatory drive. The overall responses of the ventilatory system determine the persistence of spontaneous rhythmic breathing or the development of respiratory instability and recurrent central apnea.²⁴ In HF, central apnea and PB are believed to result from an unstable ventilatory control system with increased responsiveness to hypercapnia, combined with a prolonged circulation time due to reduced cardiac contractility. This combination causes instability in ventilatory control, leading to alternate periods of over- and under-breathing, which are known as PB or CSB.²⁵ They become even more pronounced as the HF worsens and may occur while the patient is awake as well.²⁶

All patients in our study had primarily OSA and were prescribed CPAP or BPAP therapy. In patients with HF, the upper airway collapsibility may be more common because of pharyngeal edema from rostral fluid accumulation during sleep.²⁷ Both obstructive and central respiratory events are frequently present in same individuals with HF due to the interaction of airway collapsibility, pharyngeal congestion, and instability of the ventilatory system.²⁸ In our study, the mean PB% was 2.6 times higher (P < .001) in patients in the HFrEF group, which may be explained by the above mechanisms.

CSA is associated with increased morbidity and mortality in chronic kidney disease and those with coexisting HF are particularly prone to CSA.²⁹ One study showed that the risk of chronic kidney disease development was 1.5 times greater (95% confidence interval [CI]: 1.29–1.94) among patients with sleep apnea than in the matched non–sleep apnea cohort.³⁰ Concurrent decompensation of cardiac and renal function may cause cardiorenal syndrome where acute or chronic dysfunction in 1 organ may induce acute or chronic dysfunction in the other organ.³¹ It is postulated that the inability of the heart to generate forward flow causes renal hypoperfusion, which activates the renin-angiotensin-aldosterone axis, the sympathetic nervous system, and arginine vasopressin secretion, leading to increased preload, and worsening ventricular pump failure.³² Approximately 25% of patients with acute HF develop cardiorenal syndrome during hospitalization. Several factors have been postulated for this poor outcome. Decreased glomerular filtration leads to accumulation of uremic toxins, causing destabilization of central respiratory control leading to PB. This precipitates volume overload, edema of the upper airway, and collapse of the pharynx, even during the central apnea episodes.³³ An elevated index of CSA/CSB is significantly associated with increased risk of decompensated HF and/or development of clinical HF in older men.³⁴ This is consistent with our study that showed that increasing age, increasing PB% at night, and increased risk of hospitalization from acute HF were associated with declining eGFR. We also found that there were significant interactions between renal function and degree of HF in predicting PB%. The incidence of PB% was much higher with lower eGFR in patients with HFrEF compared with those with HFmrEF (Figure 4).

Atrial fibrillation is one of the important factors in the pathogenesis of central apnea/PB and several studies and case reports have demonstrated such an association.^35,36 Cardioversion to sinus rhythm has been shown to cause a gradual decrease in the PB cycle length and improvement in LVEF in 1 case study.³⁷ In our study, the prevalence of atrial fibrillation was slightly higher in the HFrEF group. Atrial fibrillation was associated with increased mortality and hospitalization in both univariate and bivariate regression analyses, but they were not statistically significant (Table 5 and Table 6), most likely due to the overall total small sample size and small number of participants in the HFrEF group (n = 41). A prospective study in the future with a larger sample size is needed to evaluate atrial fibrillation as a potential predictor of hospitalization and mortality rate related to PB.

Racial disparities with HF incidence and prevalence and hospitalization rates have been well documented in the literature.^38,39 Black patients in our study had lower mean LVEF and higher BNP despite being younger than White patients, along with lower PAP adherence. The prevalence of cardiovascular disease has been historically higher among Black compared with White individuals in the United States.⁴⁰ Nayak et al⁴¹ reported that HF-related deaths are 2.6 times higher in Black men, especially young Black men, than White men. BNP is an important marker of HF status. It may be persistently elevated in chronic HF but may further increase with decompensated HF. Mean BNP was 2.5 times higher in Black patients than in White patients in our study. The National Inpatient Sample used to estimate the rates of HF hospitalization between 2002 and 2013 by sex and race/ethnicity demonstrated that Black men and women had 2.29% and 2.4% higher hospitalization rates, respectively, from acute HF compared with White men and women.¹⁰ We had similar findings in our study sample as well. After adjusting for PB%, Black patients were almost twice as likely to be admitted for acute HF (IRR = 1.89, P = .001).

CPAP is the gold standard of treatment for OSA, and it is also often effective in CSA. Randomized trials showed that alleviation of OSA by PAP therapy in patients with HF lowers sympathetic activity and improves LVEF.^14,42,43 However, PAP adherence is the main hindrance in its effectiveness. Black individuals are generally less compliant than White individuals even after controlling for age, socioeconomic status, BMI, and other comorbidities.⁴⁴ Our study demonstrated a 12.9% lower mean PAP adherence in Black patients than in White patients. Consequently, the mean LVEF was 6.4% lower and mean BNP was 2.5 times higher in Black patients. After adjusting for age, PAP adherence and the presence of HFrEF were most significant in predicting all-cause mortality within 180 days of the index visit. The patients with HFrEF were 2.78 times more likely to die and every 10% increase in PAP adherence decreased the odds of death by 0.8 times. Higher PAP adherence was correlated with lower PB% (r = –.064), but this was not statistically significant in our study (Figure 2). This may have resulted from correlating a single snapshot of PB% and PAP adherence in the previous 30 days from the index visit. PB is a dynamic phenomenon, and it varies nightly in each patient. Our data were cross-sectional and do not reflect correlation of any dynamic fluctuation in PB% with night-to-night variation in duration of PAP machine usage. A prospective study examining variation in nightly PB% with hours of PAP usage/night within each patient as well as across a patient population may better characterize this relationship between PB% and PAP adherence.

In our study, the all-cause mortality rate for all participants was approximately 30%, which seems to be high. This may have resulted from our strict inclusion criteria that required recent laboratory tests and an echocardiogram within a certain time frame from the index visit. Because sicker patients are more likely to have frequent hospital/clinic visits with recent laboratory tests and echocardiogram, this selection bias may have resulted in a sample of participants who might have had more advanced stages of HF and renal failure, resulting in higher mortality rates.

The rates of all-cause mortality and readmission from acute decompensated HF have been an important area of research in recent years. A study using the Danish nationwide registries showed that worsening of HF was associated with a higher rate of all-cause mortality (hazard ratio: 1.22; 95% CI: 1.16–1.28) and HF readmission (hazard ratio: 1.81; 95% CI: 1.69–1.93) compared with new-onset HF.⁴⁵ We found that the risk of hospital admission from acute HF exacerbation was most significantly influenced by Black race, presence of HFrEF, and lower PAP adherence, along with a progressive increase in PB%. However, the directionality of the relationship between HF and CSA/PB is unclear. Javaheri et al³⁴ reported that an elevated CSA/CSB index was associated with increased risk of decompensated HF and/or development of clinical HF in a community-based cohort of older men followed for a mean of 7.3 years. Other reports showed cross-sectional associations between PB and severity of HF.^46,47 Our study demonstrated that HFrEF increased the likelihood of admission by 7.45 times (P < .001). Improved PAP adherence has been previously reported to be associated with decreased hospitalization rate.⁴⁸ In our study, we observed that every 10% increase in PAP adherence decreased the likelihood of admission by 0.78 times (P < .001). Therefore, a progressive increase in PB% can warn the patient and health care provider that an intervention is urgently needed to prevent hospitalization, including medication optimization and improvement in CPAP adherence.

Limitations of the study

This was a retrospective study with limited data access. The sample size was small due to unavailability of several critical laboratory data in many patients, including BNP and echocardiogram, within the period needed for inclusion in this study. In addition, we may not have complete hospitalization and mortality data if some of the veterans were admitted and/or expired in other hospitals outside of the VA hospital system. A future prospective study may be conducted that will allow obtaining these necessary laboratory tests as needed (if they are missing) on the day of the index visit. This may create a longitudinal cohort that may be followed up for a few months to a few years, which will yield more accurate data regarding the association of recurrent admissions and mortality related to the trend of PB. This will also allow the collection of data on the duration of PAP use and PB% for individual nights to longitudinally assess if the night-to-night variation in PAP adherence dynamically affects the occurrence of PB%.

CONCLUSIONS

Our study suggests that an increase in PB% detected by a PAP machine may be used to predict acute decompensation of HF that may require hospitalization. Thus, it may be considered as an indicator for HF status of a single patient over time. Reduced PAP adherence is associated with increased risk of hospital admission and all-cause mortality. Therefore, identification of an increase in PB index by the sleep physicians and alerting the cardiologist or the primary care provider may help optimize CPAP usage and the goal-directed medical therapy for those patients. This may help reduce further complications, including recurrent hospitalization and mortality.

DISCLOSURE STATEMENT

All authors have seen and approved this manuscript. Work for this study was performed at G.V. (Sonny) Montgomery VA Medical Center, Jackson, Mississippi, USA. The authors report no conflicts of interest.

ABBREVIATIONS

BMI

body mass index

BNP

brain natriuretic peptide

BPAP

bilevel positive pressure therapy

CI

confidence interval

CPAP

continuous positive pressure therapy

CSA

central sleep apnea

CSB

Cheyne-Stokes breathing

eGFR

estimated glomerular filtration rate

HF

heart failure

HFnmEF

heart failure with normal to midrange ejection fraction

HFrEF

heart failure with reduced ejection fraction

IRR

incidence rate ratio

LVEF

left ventricular ejection fraction

OSA

obstructive sleep apnea

PAP

positive airway pressure

PB

periodic breathing