Heart failure is not a determinant of central sleep apnea in the pediatric population

Abstract

Background/Objectives:

Adults with heart failure (HF) have high prevalence of central sleep apnea (CSA). While this has been repeatedly investigated in adults, there is a deficiency of similar research in pediatric populations. The goal of this study was to compare prevalence of CSA in children with and without HF and correlate central apneic events with heart function.

Methods:

Retrospective analysis of data from children with and without heart failure was conducted. Eligible children were <18 years old with echocardiogram and polysomnogram within 6 months of each other. Children were separated into groups with and without HF based on left ventricular ejection fraction (LVEF). Defining CSA as Central Apnea-Hypopnea Index (CAHI) >1/hr., the cohort was also classified into children with and without CSA for comparative study.

Results:

120 children (+HF: 19, −HF: 101) were included. The +HF group was younger, with higher prevalence of trisomy 21, muscular dystrophy, oromotor incoordination, and structural heart disease. The +HF group had lower Apnea-Hypopnea Index (median 3/hr. vs. 8.6/hr.) and lower Central Apnea Index (CAI) (median 0.2/hr. vs. 0.55/hr.). Prevalence of CSA was similar in both groups (p=0.195). LogCAHI was inversely correlated to age (Pearson correlation coefficient −0.245, p = 0.022). Children with CSA were younger and had higher prevalence of prematurity (40% vs 5.3%). There was no significant difference in LVEF between children with and without CSA. After excluding children with prematurity, relationship between CAHI and age was no longer sustained.

Conclusions:

In contrast to adults, there is no difference in prevalence of CSA in children with and without heart failure. Unlike in adults, LVEF does not correlate with CAI in children. Overall, it appears that central apneic events may be more a function of age and prematurity rather than of heart function.

INTRODUCTION

In adults with heart failure (HF), central sleep apnea (CSA) is highly prevalent and associated with higher mortality rates^{1; 2}. HF is defined as the insufficiency of the heart to pump adequate blood to organs of the body. The etiology of HF in children often differs from adults³. In adults, coronary artery disease, hypertension, acquired valvular defects, arrhythmias, and myocardial infarction are the leading causes of HF. In children, however, congenital heart defects and the sequelae of time and surgical intervention, and less commonly, cardiomyopathy are the main causes of HF⁴. While the relationship between HF and CSA has been well studied in adults, similar data in children is lacking⁵. Sleep apnea can be obstructive (due to airway obstruction) or central in nature (lack of drive to breathe)⁶. The increased risk of CSA in HF is theorized to be due to a number of hemodynamic and autonomic mechanisms involving the sympathetic nervous system’s upregulated response to periods of hypoxemia and hypercapnia⁷. While both obstructive sleep apnea (OSA) and CSA can present with increased apneic events, the physiology of OSA and CSA are different. In OSA, occlusion of airflow increases cardiac afterload and decreases stroke volume⁸. Relative hypotension triggers the baroreceptor response to increase sympathetic tone⁹. CSA occurs when partial pressure of CO₂ falls below apneic threshold, a level of CO₂ below which the nervous system fails to trigger breathing⁷. Patients with HF have a higher ventilatory response due to altered chemosensitivity to increased levels of CO₂¹⁰. Furthermore, decreased cardiac output delays detection of changes in CO₂ levels which exacerbates apnea/hyperventilation cycles¹¹. Additionally, recurrent arousals at the end of apneic events generate acute rise in arterial pressure evidenced by pulse transit time shortening¹². By treating sleep apnea in children and thereby reducing sympathetic output, these adverse effects can be mitigated. However, the link between CSA and HF in children has not been described. This is important to know, as CSA is a potentially reversible condition with which appropriate treatment may improve morbidity and mortality in children with HF.

CSA in infants with congenital heart disease is associated with a fourfold increase in mortality¹³. However, the study describing these findings did not report prevalence of HF or its association with CSA. In children with HF due to dilated cardiomyopathy, CSA was noted in 19% of the studied population⁵. No correlation was noted between CSA and heart function. In another prospective uncontrolled case series, there was a high prevalence of CSA with significant correlation between left ventricular end diastolic index, left ventricular end systolic index and central apnea index¹⁴. However, this study focused on comparing polysomnogram (PSG) indices with heart function based on the nature of cardiomyopathy but did not compare prevalence of CSA in children with and without HF. While there is a known high prevalence of OSA and CSA in adults with HF, this has not been well studied in children. Moreover, no study has been done to correlate central apnea events with heart function, in children with HF. We hypothesized that the prevalence of CSA will be different in children with HF compared to children without HF.

METHODS

Study Design

We performed a retrospective, cross-sectional analysis on children with and without HF and compared their demographics, medical comorbidities, PSGs and echocardiograms. Institutional Review Board approval was obtained. Data completed between January 1, 2015 and December 31, 2019 from a single tertiary care center was collected and reviewed.

Participants

Children (0 – 18 years of age) with echocardiogram and PSG performed within 6 months of one another were included. Echocardiograms were reviewed and the cohort was divided into groups with (+HF) and without heart failure (−HF). The +HF group was defined by a left ventricular ejection fraction (LVEF) of ≤45% or moderate to severe cardiac dysfunction. The −HF group was defined by a LVEF of ≥58% or normal cardiac function and therefore was deemed appropriate for the control group^{15; 16}. Intermediate group (LVEF 46 – 57%) data was collected post-hoc and compared with the −HF group to ensure that the intermediate group was not significantly different. Review of literature shows that LVEF of 46 – 57% is consistent with mild cardiac dysfunction. While there is considerable evidence that LVEF <45% is associated with CSA, there is no data suggesting that relationship in patients with LVEF >45%. For this reason, we chose a control group with LVEF ≥58%. Exclusion criteria included structural congenital heart disease with right to left shunting, neurological conditions associated with CSA such as Arnold Chiari malformation, congenital central hypoventilation syndrome, patients on non-invasive ventilation or tracheostomy at time of PSG, and patients with inadequate sleep time defined by total sleep time (TST) less than 120 minutes.

Patient charts in the study time period were reviewed for eligibility with Figure 1 showing the screening process. To limit the control cohort to a reasonable size, we limited the −HF group time range from January 1, 2019 through December 31, 2019. Exclusion criteria were then applied with 120 patients included for study. For post-hoc analysis, we collected data on 25 patients included in the intermediate group (LVEF 46 – 57%) from January 1, 2019 through December 31, 2019.

Figure 1.

Flow chart showing the identification of eligible patients and the screening criteria for selection of sample size.

Variables

Demographic data was collected and included: date of birth, age at time of PSG, age at time of echocardiogram, gender, race, BMI z-score (age >2 years) / weight for height z-score (age <2 years) at time of echocardiogram. Data on medical comorbidities was collected based on diagnostic billing codes (ICD-10). Comorbidities studied were prematurity, bronchopulmonary dysplasia, laryngomalacia, craniofacial anomaly, trisomy 21, neuromuscular disorder or Arnold Chiari malformation, hypertension, pulmonary hypertension, feeding dysfunction, and obesity. Surgical history was collected on adenoidectomy, tonsillectomy, or adenotonsillectomy as well as cardiac surgery and type. Heart defects were noted and broadly classified into two groups: structural heart disease or no structural heart disease. Clinical diagnosis of OSA and CSA were collected based on ICD-10 diagnostic codes.

Polysomnography

Children underwent PSG at an accredited sleep laboratory as part of clinically indicated care. PSG tests were performed in accordance with standards proposed by the American Academy of Sleep Medicine (AASM). Electroencephalogram, electro-oculogram, electromyogram (chin and both legs), electrocardiogram, pressure transducer and thermistor airflow, uncalibrated respiratory inductance plethysmography, oximetry, and end-tidal CO₂ (ETCO₂) data with video monitoring of the study for scoring support was collected. All studies were scored by a pediatric pulmonologist board certified in sleep medicine, in accordance with the pediatric scoring rules proposed by the AASM¹⁷. In our institute, post-sigh central apneas are not scored.

PSG data collected included: date of study, age at time of study, percentage of sleep spent in REM, sleep efficiency, Oxygen Desaturation Index (ODI), Arousal Index (AI), Apnea Hypopnea Index (AHI), Central Apnea Index (CAI), Mixed Apnea Index (MAI), Obstructive Apnea Index (OAI), Hypopnea Index (HI), Central Apnea Hypopnea Index (CAHI), average pulse rate, Total Periodic Limb Movement Index, average oxygen saturations in wake versus sleep, nadir saturations in sleep, total duration of desaturations, total sleep time spent between 45 and 49 mmHg ETCO₂, total sleep time spent greater than 50 mmHg ETCO₂, average ETCO₂ during sleep, and peak ETCO₂ during sleep.

In our institute, hypopneas are not scored as central or obstructive. AB reviewed raw sleep study data and scored all hypopneic events as obstructive or central, according to AASM definition¹⁷. We calculated obstructive apnea hypopnea index (OAHI) = obstructive apnea index + obstructive hypopnea index. We calculated central apnea hypopnea index (CAHI) = central apnea index + central hypopnea index. Based on commonly used definitions, we defined pediatric OSA as OAHI >1/hr. and pediatric CSA as CAHI >1/hr. Furthermore, AB reviewed all the raw data for Cheyne-Stokes breathing per AASM definition.

Echocardiogram

All echocardiograms were performed using a lab standard protocol, on ACUSON ultrasound machines. Images were acquired from parasternal long and short axis views, apical four-chamber and two-chamber views, subcostal, and suprasternal views. Images were transferred to dedicated workstations with all measurements and analyses performed offline.

Echocardiogram data collected included: date of study, age at time of study, weight, height, body surface area (BSA), right ventricular (RV) size at end diastole, thickness of interventricular septum (IVS) at end diastole, left ventricular (LV) size at end diastole, thickness of left ventricular posterior wall (LVPW) at end diastole, thickness of IVS at end systole, LV size at end systole, thickness of LVPW at end systole, and LVEF (Teicholz Formula). To accommodate for difference in ages and thus sizes between the +HF group and the −HF group and subsequently between no CSA group and CSA group, we used the Boston Children’s Hospital z-score system for echocardiographic parameters^{15; 19}. The regressions used included 2D LV End-diastolic Septal Thickness vs BSA, 2D LV End-diastolic Dimension vs BSA, 2D LV End-diastolic Free Wall thickness vs BSA, 2D LV End-systolic Sepal Thickness vs BSA, 2D LV End-systolic Dimension vs BSA, and 2D LV End-systolic Free Wall Thickness vs BSA.

Definitions¹⁷

Pediatric OSA –

In children, defined as OAHI ≥1 obstructive event per hour with the event being obstructive or mixed apnea or obstructive hypopnea in nature ¹⁷.

Pediatric CSA –

Definition of CSA is varied throughout pediatric literature¹⁸. In order to help facilitate comparison of our results with other pediatric studies on CSA, we defined pediatric CSA as CAI + central HI >1/hr. For the purposes of this study, all CSA referenced will use the Pediatric CSA definition unless otherwise stated.

Statistical Analysis

Demographics, clinical parameters, and prevalence of comorbidities between groups were analyzed using Mann Whitney U test for continuous variables and Chi-Square tests, with Fisher’s Exact test used to verify results when cell counts were small, for categorical variables. Comparison of PSG parameters was performed between groups using analysis of covariance (ANCOVA), adjusting for age at time of study and prematurity. We performed Pearson correlation between OAHI as well as CAHI with echocardiogram parameters. Furthermore, we analyzed the relationship between log CAHI/log OAHI and age at time of sleep study using Pearson correlation coefficient. We compared demographic and echocardiographic parameters in children with and without CSA. Finally, we excluded children with history of prematurity and compared polysomnographic and echocardiographic parameters between children with and without CSA. In this subgroup, we performed Pearson correlation between OAHI and CAHI with echocardiogram parameters. Furthermore, we analyzed the relationship between log CAHI/log OAHI and age at time of sleep study using Pearson correlation coefficient. All analytic assumptions were verified, and all analyses were performed using SPSS v26 statistical software (Armonk, NY: IBM Corp). In this pilot study we sought to determine the relationship between CSA and HF, so sample size was limited to the small number of patients meeting inclusion criteria. As post-hoc analysis, the intermediate group was compared to the −HF group using Mann Whitney U test for continuous variables and Chi-Square tests, with Fisher’s Exact test being used to verify results when cell counts were small, for categorical variables.

RESULTS

120 children (median age: 28.5 months, 0 – 220) were included. +HF group had 19 children while −HF group had 101 children.

Table 1 compares demographics and medical diagnoses between groups. The +HF group had a significantly higher prevalence of trisomy 21 (15.7% vs. 0%), muscular dystrophy (10.5% vs. 0%), and oromotor incoordination (57.9% vs. 23.8%). The +HF group also had, as expected, a higher prevalence of structural heart disease than the −HF group (78.9% vs 13.9%). The types of structural heart disease in the +HF group included atrial septal defect (ASD), ventricular septal defect (VSD), coarctation of the aorta (CoA), pulmonary stenosis (PS), cardiac fibroma, double outlet right ventricle (DORV), and hypertrophic cardiomyopathy. Children with −HF included ASD, VSD, PS, CoA, DORV, vascular ring, bicuspid aortic valve, and truncus arteriosus. 46 (38.3%) echocardiograms were performed due to reported cardiac signs and symptoms including but not limited to heart murmur, desaturations, chest pain, tachycardia and abnormal EKG. 74 (61.7%) echocardiograms were performed as part of surveillance due to prevalence of risk factors for cardiac dysfunction, including but not limited to severe OSA, genetic syndrome and neuromuscular disorder. 118 (98.3%) sleep studies were performed to evaluate for snoring or desaturations while 2 (1.7%) sleep studies were performed due to presence of risk factors including but not limited to genetic syndromes. None of the patients had Cheyne-Stokes breathing.

Table 1.

Demographics and medical diagnoses in children with and without HF. Age and BMI z-score are recorded as medians with minimum and maximum in parentheses. Demographics and medical diagnoses are listed as prevalence with corresponding percentages in parentheses.

Variables	+HF Group (N=19)	−HF Group (N=101)	p value
Age (months)	24 (0 – 220)	30 (0 – 217)	0.239
BMI z-score	0.25 (−2.62 – 3.01)	0.23 (−5.3 – 3.18)	0.968
Gender (female)	9 (47.4%)	30 (29.7%)	0.181
Race			0.006
White	13 (68.4%)	69 (68.3%)
Black	3 (15.7%)	22 (21.7%)
Unknown	5 (4.9%)	0 (0%)
Prematurity	1 (5.3%)	21 (20.7%)	0.192
Trisomy 21	3 (15.7%)	0	0.003
Obesity	1 (5.26%)	11 (10.7%)	0.688
Hypertension	2 (10.5%)	9 (8.9%)	0.685
Pulmonary Hypertension	2 (10.5%)	8 (7.8%)	0.658
Muscular Dystrophy	2 (10.5%)	0	0.024
Oromotor Incoordination	11 (57.9%)	24 (23.8%)	0.005
Structural Heart Disease	15 (78.9%)	14 (13.9%)	<0.001

Abbreviations: HF, Heart Failure; BMI, Body Mass Index.

Table 2 depicts the relationship between PSG and echocardiogram findings between groups. The +HF group had lower median AHI (p=0.017) and lower median CAI (p=0.016) when compared to the −HF group. Furthermore, the +HF group had a higher median 2D LV End-diastolic Septal Thickness vs. BSA z-score (p=0.02) and a higher median 2D LV End-systolic Dimension vs. BSA z-score (p<0.001). At this point we adjusted for age and prematurity to see if the observed elevations in central apnea measures were a function of age and prematurity, rather than heart failure condition. An ANCOVA model was used with a dependent variable of AHI, fixed factor of +HF group, and covariate of age and prematurity with Bonferroni correction due to the small sample size. After this adjustment, AHI and CAI were no longer significantly different between groups.

Table 2.

PSG and echocardiogram findings of children with and without HF. Median values are recorded for each parameter with corresponding minimum and maximum in parentheses.

Variables	+HF Group (N=19)	−HF Group (N=101)	p value
% Sleep Efficiency	73.9 (59.2 – 94.8)	76.7 (30.4 – 97.2)	0.714
% REM Sleep	17.3 (0.0 – 45.9)	19.2 (0.0 – 62.7)	0.306
AHI (events/hr.)	3 (0.0 – 105.5)	8.6 (0.0 – 136.3)	0.017
OAHI (events/hr.)	2.5 (0.0 – 104.9)	6.9 (0.0 – 135.4)	0.299
CAHI (events/hr.)	0.2 (0.0 – 2.1)	0.6 (0.0 –37.3)	0.107
CAI (events/hr.)	0.2 (0.0 – 1.3)	0.55 (0.0 – 19.9)	0.016
MAI (events/hr.)	0.0 (0.0 – 0.3)	0.0 (0.0 – 5.1)	0.051
HI (events/hr.)	2.8 (0.0 – 80.4)	5 (0.0 – 82.3)	0.052
OAI (events/hr.)	0.49 (0.0 – 24.6)	0.9 (0.0 – 63.7)	0.279
Average duration of central apneic event (seconds)	3.9 (0.0 – 23.5)	5.9 (0.0 –19.3)	0.257
Maximum duration of central apneic event (seconds)	4.6 (0.0 – 35.9)	8.3 (0.0 – 52.6)	0.183
Average sats in sleep	96.2 (90.1 – 99.1)	97.2 (78.3 – 100)	0.190
Nadir sats in sleep	86.2 (73 – 94)	85 (31 – 98)	0.214
Average ETCO2 mmHg	37.45 (30.1 – 51.6)	41.5 (20 – 55.2)	0.096
Peak ETCO2 mmHg	45.8 (32.3 – 61)	49.6 (32.8 – 70)	0.040
RV diastole (mm)	16.7 (7.3 – 38.5)	14.95 (7.7 – 34)	0.639
2D LV end diastolic septal thickness vs BSA z-score	1.14 (−2.9 – 19.34)	0.05 (−5.26 – 6.17)	0.02
2D LV end diastolic dimension vs BSA z-score	0.74 (−6.79 – 7.92)	−1.19 (−6.78 – 3.28)	0.067
2D LV end diastolic free wall thickness vs BSA z-score	0.63 (−4.4 – 17.47)	−0.24 (−4.5 – 8.95)	0.384
2D LV end systolic septal thickness vs BSA z-score	−0.59 (−3.85 – 12.84)	−1.01 (−5.2 – 6.16)	0.617
2D LV end systolic dimension vs BSA z-score	4.39 (−2.31 – 13.2)	−0.94 (−6.13 – 4.85)	<0.001
2D LV end systolic free wall thickness vs BSA z-score	−0.31 (−3.99 – 3.19)	−7.8 (−8.27 – 4.56)	0.582
LVEF	40.8 (21.4 – 45.6)	67 (58 – 87)	<0.001

Abbreviations: PSG, Polysomnogram; HF, Heart Failure; AHI, Apnea Hypopnea Index; OAHI, Obstructive Apnea Hypopnea Index; CAHI, Central Apnea Hypopnea Index; CAI, Central Apnea Index; MAI, Mixed Apnea Index; HI, Hypopnea Index; OAI, Obstructive Apnea Index; RV, Right Ventricle; LV, Left Ventricle; BSA, Body Surface Area; LVEF, Left Ventricular Ejection Fraction.

Table 3 depicts prevalence of various definitions of sleep apnea in both groups. Prevalence of Pediatric OSA was high in the patients with heart failure (73.7%) and was comparable to controls (82.2%). Prevalence of Pediatric CSA was low in both groups and did not differ significantly (p=0.195). Clinical diagnosis of OSA and CSA did not differ significantly between the 2 groups. 4 patients had a clinical diagnosis of CSA. 3 of these patients had concomitant OSA diagnosis, while 1 patient had a clinical diagnosis of CSA with CAHI of 1.3/hr. CAHI and OAHI did not correlate with LVEF or any other echocardiogram parameters.

Table 3.

Prevalence rates of commonly defined sleep apnea in children with and without HF.

Variables	+HF Group (N=19)	−HF Group (N=101)	p value
Pediatric OSA (OAHI >1/hr.)	14 (73.7%)	83 (82.2%)	0.282
Pediatric CSA (CAHI >1/hr.)	4 (21%)	41 (40.6%)	0.195
Clinical Diagnosis of OSA	3 (15.8%)	38 (37.6%)	0.072
Clinical Diagnosis of CSA	0 (0.0%)	4 (3.9%)	1.000

Abbreviations: HF, Heart Failure; OSA, Obstructive Sleep Apnea; OAHI, Obstructive Apnea-Hypopnea Index; CAI, Central Apnea Index; CAHI; Central Apnea-Hypopnea Index; CSA, Central Sleep Apnea.

CAHI was inversely correlated to age at time of sleep study (Pearson correlation coefficient −0.245, p = 0.022). OAHI correlated to age at the time of sleep study (Pearson correlation coefficient 0.259, p=0.004). Due to the wide range of CAHI and OAHI in the study population, we performed logarithmic transformation and depicted the relationship between LogCAHI / LogOAHI and age in Figure 2.

Figure 2.

Relationship between LogCAHI / Log OAHI and age.

Using our definition of CSA (CAHI >1/hr.), we divided our cohort into two groups: no CSA and CSA. Table 4 compares these two groups’ demographics, medical diagnoses, echocardiographic findings, and PSG findings. Children with CSA were younger (p=0.002) and had higher prevalence of prematurity (p<0.001). Sleep efficiency was lower (p=0.026) and percentage REM sleep was higher in children with CSA (p<0.001). Average duration of central apnea and maximum duration of central apnea was significantly higher in children with CSA. There was no statistically significant difference in the LVEF or other echocardiographic parameters between groups.

Table 4.

Demographics, medical diagnoses, echocardiogram parameter data, and PSG characteristics of children with and without CSA. Median values are recorded for each parameter with corresponding minimum and maximum in parentheses.

Variables	Children with no CSA (N=75)	Children with CSA (N=45)	p value
Age (months)	62 (0 – 220)	4 (0 – 217)	0.002
BMI z-score	0.22 (−3.68 – 3.13)	0.24 (−5.3 – 3.18)	0.702
Gender (female)	27 (36%)	12 (26.7%)	0.321
Race			0.210
White	49 (65.3%)	36 (80%)
Black	18 (24%)	7 (15.6%)
Prematurity	4 (5.3%)	18 (40%)	<0.001
Obesity	7 (9%)	5 (11.1%)	0.762
Hypertension	7 (9.3%)	4 (8.9%)	1.000
Pulmonary Hypertension	5 (6.5%)	5 (11.1%)	0.496
Muscular Dystrophy	1 (1.3%)	1 (2.2%)	1.000
Oromotor incoordination	18 (25.9%)	17 (37.8%)	0.221
Structural heart disease	20 (24%)	9 (20%)	0.511
Heart failure	15 (19.5%)	4 (8.9%)	0.127
% Sleep Efficiency	78.5 (43.7 – 97.2)	74.9 (30.4 – 95.9)	0.026
% REM Sleep	17.6 (0.0 – 43.3)	25.4 (0.0 – 62.7)	<0.001
OAHI (events/hr.)	3.5 (0.0 – 135.4)	8.6 (0.0 – 79.7)	0.810
CAHI (events/hr.)	0.2 (0 – 0.9)	3.2 (1.1 – 37.3)	<0.001
Average duration of central apneic event (seconds)	4.6 (0.0 – 16.1)	6.2 (0.0 – 23.5)	<0.001
Maximum duration of central apneic event (seconds)	4.6 (0.0 – 16.8)	9 (0.0 – 52.6)	<0.001
Average sats in sleep	96.9 (78.3 – 100)	97.3 (91.6 – 99.4)	0.196
Nadir sats in sleep	87 (42 – 98)	82.5 (31 – 90)	0.01
Average CO2 mm Hg	42.05 (20 – 55.2)	38.7 (21.8 – 46.9)	0.117
Peak ETCO2 mm Hg	50.2 (32.3 – 70)	47.4 (33.9 – 63.4)	0.294
2D LV end diastolic septal thickness vs BSA z-score	−1.2 (−5.26 – 19.34)	0.41 (−3.88 – 7.67)	0.982
2D LV end diastolic dimension vs BSA z-score	−1.19 (−6.79 – 7.92)	−1.0 (−6.63 – 3.98)	0.974
2D LV end diastolic free wall thickness vs BSA z-score	0.06 (−4.5 – 17.47)	−0.58 (−4.31 – 6.34)	0.133
2D LV end systolic septal thickness vs BSA z-score	−0.81 (−5.2 – 12.84)	−1.16 (−5.08 – 1.91)	0.199
2D LV end systolic dimensions vs BSA z-score	−0.73 (−6.13 – 13.2)	−0.451 (−4.10 – 8.92)	0.822
2D LV end systolic free wall thickness vs BSA z-score	−0.71 (−8.27 – 4.56)	−0.9 (−4.68 – 3.13)	0.414
LVEF	65 (25.2 – 87)	66 (21.35 – 79)	0.55

Abbreviations: PSG, Polysomnogram; CSA, Central Sleep Apnea; BMI, Body Mass Index; OAHI, Obstructive Apnea-Hypopnea Index; CAHI, Central Apnea-Hypopnea Index; LV, Left Ventricle; BSA, Body Surface Area; LVEF, Left Ventricular Ejection Fraction.

After excluding premature children, the difference in sleep efficiency, %REM sleep, average duration of central apneic event, maximum duration of central apneic event and nadir saturations in sleep was sustained. However, after excluding premature children, there was no significant difference in age between children with CSA (8 months, 0 – 217) and without CSA (67 months, 0 – 220).

As part of post-hoc analysis, we identified the intermediate group with LVEF of 46 – 57%. 25 children were eligible for inclusion. We compared this group with the −HF group to reaffirm that the intermediate group was similar to the −HF in terms of demographics and PSG parameters. This data can be found in the supplemental tables 1–3.

DISCUSSION

Our study is the first to compare prevalence of CSA in children with and without HF presenting with snoring or desaturations. Contrary to adults, using the pediatric definition, prevalence of CSA is similar in children with and without HF. After adjusting for age, prevalence of CSA or OSA did not differ between the two groups. Children with CSA were younger with higher prevalence of prematurity. Unlike adults, in children, CAHI is not significantly associated with LVEF. CAHI correlated inversely to age at the time of sleep study. However, after excluding children with prematurity, this relationship was no longer sustained.

We compared our results with the only other study on sleep disordered breathing in children with HF⁵. This was a prospective observational study, where the prevalence of CSA was 19% in children with HF secondary to dilated cardiomyopathy. This study used the pediatric CSA definition of central apnea/hypopnea per hour of sleep >1 as abnormal. Using similar criteria of pediatric CSA definition, prevalence of pediatric CSA in the +HF group of our study was similar at 21% compared to the den Boer study (19%) (Table 3). Interestingly, none of the children with +HF had a clinical diagnosis of CSA. Children with +HF in our study were younger with median age of 24 months compared to the den Boer study with a median age of 11.1 years. If we only compared the CAI in the two groups with and without HF, it was found to be significantly lower in the +HF group. After adjusting for age, CAHI was no longer different between the two groups. Thus, we conclude that, unlike the adult population, central apneic events were not increased in children with HF. Instead, central apneic events were more commonly seen in younger aged children with prematurity, regardless of cardiac function.

Our results support previous findings, which did not show any correlation between AHI and severity of cardiac dysfunction measured by LVEF^{5; 14}. One study focused on assessing frequency of sleep disordered breathing and its relationship to cardiac function in children with cardiomyopathy found significant correlations between CAI+HI and both LV end diastolic volume index and LV end systolic volume index¹⁴. We did not have data on LV end diastolic volume or end systolic volume index. 2D LV End-diastolic Septal Thickness vs BSA z-score and 2D LV End-systolic Dimension vs BSA z-score were higher in the +HF group. The septal thickness being significantly higher is not an intuitive result as the assumption is that dilated, poorly functioning hearts have a thinner septal thickness. With the presence of a subject with hypertrophic cardiomyopathy in conjunction with a small n of the +HF group showing a wide range up to a z-score of 19.34, this was most likely enough to skew the average to significance. However, these parameters did not correlate with CAHI or OAHI. This finding contrasts the adult literature, where sleep disordered breathing is associated with left ventricular remodeling²⁰. However, this is supported by previous studies in pediatric literature which suggest that pediatric sleep disordered breathing is not associated with significant cardiovascular strain and the majority of cardiovascular parameters in children with sleep disordered breathing are within the normal range at baseline²¹.

The relationship between age and CAHI demonstrated in our study is intriguing and a key finding for future research in this field. In the previous study by den Boer, the median age of the 7 children with pediatric CSA was 2.9 years compared to 30 patients without CSA who had a median age of 12.3 years⁵. One of the biggest challenges of pediatric CSA is to have a consistent definition. While some authors have defined CSA as CAI+HI >1/hr., many have used a cut off of 5/hr. While healthy term infants have an estimated median CAI of 5.5/hr. at 1 month of age²², older children aged 7.3 (4) years of age with Chiari 1 malformation had a median CAI of 2.4 (0.63 – 8.95)²³. In our cross-sectional study, we found that CAHI is inversely correlated to age. This has previously been described in children with trisomy 21²⁴. Infants are particularly vulnerable to sleep apnea due to their upper airway structure²⁵, ventilatory control ²⁶, arousal threshold ²⁷, laryngeal chemoreflex²⁸, REM-predominant sleep state distribution²⁹ and physiologically exaggerated laryngeal chemoreflexes that actively induce protective apneas³⁰. Central respiratory pauses from immaturity of control of breathing are frequent during REM sleep in infants²². Review of literature suggests that CAI can be as high as 45/hr. at 1 month of age and decrease to 10–20/hr. for older infants³¹. While the CAI in our study was lower (0 – 19.9/hr.) at median age of 28.5 months, the inverse relationship between age and CAHI supports previous literature. After excluding children with prematurity, the relationship between age and CAHI is no longer sustained. However, after excluding children with prematurity, the median age of children with CSA increases from 4 to 8 months. Prematurity and younger age both may have a significant role in this relationship between CAHI and age.

While there are no studies exploring the relationship between age and central apnea, there are a few possible physiologic explanations involving central and carotid body chemoreception sensitivities. Thus far, chemoreceptor responsiveness has been evaluated in animal and cellular models^{32; 33}. However, the actual implications of this as a causative reason for the relationship between age and central apnea is still grounds for speculation and requires investigation with future study.

The primary weakness of our study is its retrospective nature. The sleep studies were performed as the patients were symptomatic. This gives rise to a referral bias, which can explain the high prevalence of OSA in both the cohorts. Our −HF control group was significantly younger. We did not have data on medications used in heart failure which could also affect control of breathing. The +HF group had only 19 children with heterogenous diagnoses. This may have led to inability to detect differences with the −HF group (for example some of the PSG respiratory indices where the median value of the −HF group was 2–3 times greater than the +HF group). This limitation is due to low prevalence of HF in children and might be overcome by performing multicenter studies and increasing sample size. Along the same lines, another limitation was including controls from 2019 alone. This was done to limit inequality in sample size between groups. If we matched for age and gender, we were likely to have several possible matches for each case and picking only one or two of those would have increased bias. Instead, we limited controls to a single year and controlled for the differing variables in the results. Transcutaneous CO₂ was not collected in the patients, which may have helped in gaining a better understanding of the gas exchange disturbances in this population. Prospective cohort studies correlating objective measurements of LVEF to PSG parameters are needed to better understand the relationship between HF and CSA. Due to the cross-sectional nature of the study, we were unable to prove causality. Longitudinal studies can provide insight regarding the relationship between CAHI and age as well as prematurity. Due to the strong association between younger age, prematurity, and maturation of control of breathing, it is difficult to determine the contribution of age and prematurity on maturation of control of breathing in a cross-sectional study. Finally, this is the experience of a single center and practices may vary between different centers. This study is a pilot initiative to gain a better understanding of this patient population.

Ultimately, unlike adults, after adjusting for age, there is no difference in prevalence of CSA in children with and without HF. CAHI appears to be a function of age in children, rather than a function of ejection fraction, with the younger patients demonstrating higher central apneic hypopneic events. This relationship is not sustained after excluding infants with prematurity. Future studies exploring the relationship between CAI and age as well as prematurity can help us gain a better understanding of determinants of CSA in children.