Longer respiratory events in childhood obstructive sleep apnea syndrome constitute a trait of older children with excessive daytime sleepiness

Abstract

The objective was to evaluate the determinants of desaturation during apnea and apnea–hypopnea duration’s links with the endotypes (pharyngeal compliance, loop gain) of obstructive sleep apnea (OSA) and with heart rate variability (HRV) indices. We retrospectively selected 35 otherwise healthy children (median age: 10.6 years, 16 females) with moderate to severe OSA who had measurements of pharyngeal compliance (acoustic pharyngometry), loop gain (awake tidal ventilation with calculation of controller and plant gains), and respiratory system impedance (impulse oscillometry system). HRV indices were obtained from the whole-night ECG of polysomnography. The slope of desaturation (0.64%.s⁻¹) during apnea was related to baseline end-tidal PO₂ (r² = 0.74; p < 0.001), which is influenced by small airway disease (reactance at 5 Hz and its area). The mean durations of apnea and hypopnea in NREM and REM sleep were combined into the mean apnea–hypopnea duration (MAD₄), which positively correlated with age, sleepiness, and HRV indices. The MAD₄ correlated with two endotypic traits, controller and plant gains. In conclusion, arterial desaturation during apnea is related to alveolar PO₂. Longer respiratory events constitute a trait of older children with excessive daytime sleepiness, higher HRV, and adaptive responses of ventilatory control to maintain its stability.

Introduction

Obstructive sleep apnea (OSA) is a heterogeneous disorder characterized by intermittent pharyngeal obstruction causing recurrent desaturations and arousals. Although the apnea–hypopnea index (AHI) is commonly used to quantify the severity of OSA, this frequency measurement provides an incomplete descriptor of physiological stressors and has variable association with symptom burden, outcomes, and treatment adherence¹. An initial step toward this goal is to better characterize the underlying pathophysiology of each patient and quantify the contributions of anatomic (i.e., upper airway collapsibility) and nonanatomic factors (i.e., arousal threshold and ventilatory control system sensitivity, such as loop gain) to disease severity². Respiratory event duration may also be a useful parameter for improving OSA phenotyping because it can provide an integrated output of the major pathophysiological traits that characterize OSA³. Along this line, Borker and colleagues showed in adult patients with OSA that shorter event durations were associated with lower circulatory delay, lower arousal threshold, reduced pharyngeal collapsibility, and higher loop gain⁴. Most of the reviewed studies support the idea that longer respiratory events have more severe physiological and clinical consequences than shorter events, most probably due to the higher hypoxic burden associated with longer respiratory events. However, a few studies provide clear evidence that short respiratory events have also a deleterious effect on sleep and the physiological and clinical aspects of OSA^5,6.

The consequences of hypoxia versus apnea in childhood OSA have not been studied, probably because these consequences are linked to long-term cardiovascular disease in adults. Nevertheless, neurodevelopmental consequences of OSA in childhood can be observed⁷, which could be due to autonomic nervous system (ANS) dysfunction and associated cerebrovascular disease^8,9. It is often believed that longer respiratory events will be associated with more severe hypoxia and more severe disease, which remains to be demonstrated, since the determinants of arterial desaturation during apnea are multiple, as demonstrated by modeling studies^10,11.

Thus, our retrospective study had three objectives: first, to evaluate the determinants of arterial desaturation during isolated apnea; second, to evaluate the links of apnea–hypopnea duration with pathophysiological endotypes of OSA (pharyngeal compliance, loop gain, and its determinants); and third, to evaluate the relationships between apnea–hypopnea duration and ANS dysfunction using heart rate variability indices.

Materials and methods

Design

From our database, we selected otherwise healthy children (3 to 18 years) with moderate to severe OSA (obstructive AHI determined by night-polysomnography ≥ 5/hour of sleep during a previous hospitalization, as previously described^12,13) who underwent a systematic evaluation on the same day in a multidisciplinary clinic devoted to the multistep therapeutic approach to childhood OSA, as recommended¹⁴. The data collected included demographics, ethnicity, Z-score of body mass index (a nurse measured height, weight, and neck circumference), and the presence or absence of physician-diagnosed asthma and symptoms of sleep-disordered breathing (Brouillette, Spruyt–Gozal, and Epworth questionnaires) or attention-deficit/hyperactivity disorder (Conners questionnaire). The sole non-screening criterion for this study was OSA already treated by non-invasive ventilation. We then selected children who had both SpO₂ and ECG signals available for the whole night of polysomnography, loop gain measurement, and pulmonary function tests (exclusion criteria).

Due to the retrospective nature of the study, the Comité d’Evaluation de l’Ethique des Projets de Recherche de Robert Debré (PHENOSAS: N° 2018 − 416) waived the need of obtaining informed consent. The experimental protocols were those routinely performed in our patients as usual care and were approved by the Comité d’Evaluation de l’Ethique des Projets de Recherche de Robert Debré. The subjects and their parents were informed of the collection of their prospective data for research purposes, and they could request to be exempted from this study in accordance with French law (non-interventional observational research). This study complied with STROBE criteria of observational studies and conformed to the standards set by the latest revision of the Declaration of Helsinki.

Sleep study

In-laboratory (children under 8 years old) or ambulatory (≥ 8 years) polysomnography studies were performed overnight using an Alice 6 LDx or PDX polysomnography system (Philips, Murrysville, PA, USA) as previously described^12,13. The following parameters were recorded: chest and abdominal wall motion using respiratory inductance plethysmography, heart rate by electrocardiogram, arterial oxygen saturation by pulse oximetry, transcutaneous PaCO₂, airflow using a 3-pronged thermistor, nasal pressure by a pressure transducer, electroencephalographic leads (C3/A2, C4/A1, F3A2, F4A1, O1/A2, O2/A1), left and right electrooculograms, submental electromyogram, and tibial electromyogram. Experienced pediatric sleep physicians scored patients using standard pediatric sleep scoring criteria¹⁵.

The kinetics of apnea-induced oxygen desaturation were determined by plotting the duration of apnea events (determined using AASM objective criteria¹⁵) and the corresponding drop in SpO₂; a linear correlation was further evaluated, and the r² value of the Pearson test was used as a crude index of linearity. Only isolated apneas were selected (not preceded by a respiratory event within 3 min) with a normal pre-apneic SpO₂ (97–100%). These kinetics were determined in both NREM and REM sleep.

The mean durations of apnea and hypopnea in both NREM and REM sleep were recorded; these mean durations further allowed us to compute the mean apnea–hypopnea duration (MAD), an index developed by others^16–18. It was recently shown that hypoxic and ventilatory burden (event-specific area under ventilation signal: mean-normalized, area under mean) predicted cardiovascular disease morbidity and that 78% of the variation in hypoxic burden was explained by ventilatory burden¹⁹. Thus, we focused on the durations of respiratory events, and we computed the MAD (mean value of apnea and hypopnea obtained during NREM and REM sleep), which can easily be obtained from polysomnography. This MAD aimed at characterizing the duration of all events independently of the severity of OSA, since we wanted to characterize this specific trait. If MAD is a specific trait of OSA, one may hypothesize that it would be encountered in both NREM and REM sleep and affect both apnea and hypopnea duration. Thus, two MAD values were calculated, one in participants having the four kinds of respiratory events (hypopnea and apnea in both NREM and REM sleep, n = 19: MAD₄) and one in the whole 35 participants (who had at least two kinds of events: MAD_≥2). This latter index was calculated since MAD₄ was obtained in a restricted number of patients.

Heart rate variability analyses

Electrocardiographic recordings obtained during polysomnography were exported as EDF files. The files were then processed using HRVanalysis 1.1 software, downloaded at https://anslabtools.univ-st-etienne.fr and validated by Pichot et al.²⁰ We selected the whole night period for analyses; thus, we studied both the short-term (HRV induced by all events) and long-term (residual ANS dysfunction) effects of events. A potential artifact identification and removal algorithm was applied to all the RR interval data.

Time-domain variables included the mean sinus heart rate (HR), the standard deviation of the RR intervals (SDNN), and the root mean of squared successive differences (RMSSD).

Nonlinear analyses were also performed, since they can be used for non-stationary data from time series²¹and since nonlinear indices have been found to be linked with clinical scores of polysomnography²².

A Poincaré plotwas graphed by plotting every RR interval against the prior interval, creating a scatter plot. We analyzed the Poincaré plot by fitting an ellipse to the plotted points. After fitting the ellipse, we derived two nonlinear measurements, SD1 (short-term variability) and SD2 (long-term variability). SD1 may provide information on the parasympathetic activity, whereas SD2 seems inversely related to sympathetic activity²³.

A fractal dimensionis a statistical index of how details in a pattern change with the scale at which it is measured. From the several algorithms available to calculate the fractal dimension of a time series, we used the algorithm proposed by Katz²⁴. The higher the magnitude of the fractal dimension, the greater is the dynamic complexity.

The Lyapunov exponentis a measure of the system’s dependency on the initial conditions but also quantifies the system’s predictability²⁴. The value of the Lyapunov exponent increases, corresponding to lower predictability, as the degree of chaos becomes higher; a positive Lyapunov exponent is a strong indicator of chaos.

Tidal breathing measurements (loop gain calculation)

Recordings of tidal breathing were performed as previously described²⁵, lasting 20 min, with the first 5 min being discarded. During the recordings, subjects were awake in a calm and non-stimulating atmosphere. Flow rate, end-tidal PO₂ (PETO₂), and end-tidal PCO₂ (PETCO₂) were continuously monitored, and signals were digitized using the MP-100 system (Biopac System Inc., Santa Barbara, CA) at a rate of 50 Hz.

The loop gain model was fitted on the changes from the baseline (mean) levels of the ventilatory parameters ( [minute ventilation], PETO₂, and PETCO₂). We considered three distinct frequency bands: low-frequency oscillations with periods of 16–50 breaths/cycle, medium-frequency oscillations of 5–15 breaths/cycle, and high-frequency oscillations of 2–4 breaths/cycle. Since the medium-frequency band spans the range of cycle durations of periodic breathing observed experimentally, we focused on this frequency range, as previously done²⁵. The means of the absolute values of the corresponding variables are given over the frequency ranges.

As previously described, an unconstrained bivariate ( and PETCO₂) autoregressive model was used²⁵. Since tidal breathing measurements were done while awake, the analyses were specifically related to chemosensitivity.

Ear-nose-throat exam and acoustic exams

Patients underwent acoustic rhinometry and pharyngometry testing using an Eccovision 4.50 acoustic rhinometer-pharyngometer (Sleep Group Solutions), as previously described^12,26. Variables of rhinometry included the volume of the nasopharynx corrected for height to obtain a normalized parameter and calculated nasal resistance given by the apparatus (from nostril to nasopharynx). Nasal airway resistance was determined for each side of the nose, and the total resistance was calculated using the Ohm’s law equation for parallel resistors.

The variable of pharyngometry was the volume of the pharynx (between the oropharyngeal junction and glottis) measured in sitting and supine positions (after they were in this position for 5 min). An estimated pharyngeal compliance (cm³/kPa) was then calculated as described previously^12,13,26.

Fiber endoscopy was performed with a flexible endoscope after local anesthesia using lidocaine hydrochloride 10%. Nasal obstruction was defined as obstruction due to adenoid hypertrophy (Grade 3 or 4 according to Cassano et al.²⁷) or turbinate hypertrophy. Tonsillar hypertrophy was graded according to the Brodsky score²⁸.

Pulmonary function tests

The impedance of the respiratory system was measured using an impulse oscillatory system (IOS: Master Scope Body, CareFusion Technologies, Yorba Linda, California), as previously described²⁹. We used the following IOS variables: resistance (R) and reactance (X) at 5 Hz and 20 Hz and area under the reactance curve (AX). The Z-scores of IOS variables were calculated according to Gochicoa-Rangel et al.³⁰Functional residual capacity (FRC) was measured using the dilution technique (Master Scope Body, CareFusion Technologies, Yorba Linda, California) in the sitting position, and the Z-score was calculated³¹.

Statistical analyses

Results were expressed as median [25th − 75th percentiles]. Comparisons between sleep stages were performed using the Wilcoxon paired test. Correlations were evaluated using Spearman’s coefficient. The other statistical analyses are described in the text. A p-value ≤ 0.05 was deemed significant. No correction for multiple testing was done due to the pathophysiological design of the study³². All statistical analyses were performed with StatView 5.0 software (SAS institute, Cary, North Carolina).

Results

The characteristics of the 35 otherwise healthy children with moderate to severe OSA are described in Tables 1 and 2.

Table 1.

Clinical and polysomnographic characteristics of the 35 children with moderate to severe OSA.

Characteristics, n or median [25th – 75th percentile]	35 children with OSA
Clinical characteristics Age, years Sex, female / male Ethnicity, Caucasian / African / Asian / Mixed Z-score of body mass index Asthma diagnosis, n	10.6 [7.8 ; 12.6] 16 / 19 13 / 17 / 4 / 1 2.28 [1.87 ; 2.62] 3
Questionnaires Brouillette Spruyt-Gozal Epworth Conners	2.56 [−0.99 ; 3.97] 2.72 [2.00 ; 3.41] 8 [3 ; 11] 10 [3 ; 14]
Polysomnography Total Sleep Time / Sleep Period Time, % Obstructive AHI/hour sleep Oxygenation desaturation index/h Sleep time with SpO2 < 90%, % of sleep time Nadir SpO2, %	92 [90 ; 95] 13.1 [7.0 ; 19.9] 9.6 [5.8 ; 16.4] 0.1 [0.0 ; 0.2] 87 [83 ; 90]
Apnea duration NREM sleep, s (n patients) REM sleep, s (n patients)	11.0 [9.6 ; 14.2] (27) 12.5 [10.1 ; 17.1] (28)
Hypopnea duration NREM sleep, s (n patients) REM sleep, s (n patients)	15.9 [14.1 ; 19.0] (35) 16.7 [14.5 ; 18.8] (31)
Nadir SpO2 after apnea NREM sleep, % REM sleep, %	89 [86 ; 92] 86 [82 ; 91]
Nadir SpO2 after hypopnea NREM sleep, % REM sleep, %	90 [87 ; 93] 90 [85 ; 93]
Ear Nose Throat exam Tonsil hypertrophy, n (%) Adenoid hypertrophy, n (%) Nasal obstruction, n (%)	23 (66) 9 (26) 11 (31)

Table 2.

Characteristics of the 35 children with moderate to severe OSA.

Characteristics, n or median [25th – 75th percentile]	35 children with OSA
Acoustic pharyngometry Supine-induced pharyngeal volume reduction, % Calculated pharyngeal compliance, cm3/kPa	31 [19; 40] 12.2 [6.4; 20.4]
Acoustic rhinometry Corrected naso-pharyngeal volume, cm3/m Calculated nasal resistance, kPa.s/L	0.41 [0.24; 0.56] 2.67 [1.81; 5.29]
Pulmonary function tests z-score of FRC z-score of R5Hz z-score of R20Hz z-score of X5Hz z-score of AX	−1.24 [−2.28; −0.03] 1.91 [1.08; 4.69] 1.90 [1.40; 3.66] 0.39 [0.41; 1.00] 1.00 [0.20; 2.63]
Loop gain (LG) indices,n = 33 children* Steady-state Plant Gain (PG), mmHg.min.L−1 Steady-state Controller Gain (CG), L.min−1.mmHg−1 Steady-state LG (log value of │ssPG│* ssCG) τcentral, s τperipheral, s Circulatory delay, s PG-medium frequency, mmHg.min.L−1 CG-medium frequency, L.min−1.mmHg−1 LG-low frequency LG-medium frequency LG-high frequency	−0.74 [−1.19; −0.60] 0.64 [0.24; 1.81] −0.37 [−0.75; 0.10] 20.0 [4.1; 50.5] 4.9 [2.9; 7.4] 6.8 [3.5; 10.3] 0.43 [0.35; 0.61] 0.15 [0.10; 0.25] 0.22 [0.13; 0.38] 0.07 [0.05; 0.11] 0.01 [0.01; 0.03]
HRV analyses from the whole night Mean HR, beat per min RMSSD, ms SDNN, ms SD1, ms SD2, ms SD1/SD2 Fractal dimension (Katz index) Lyapunov exponent	82 [78 ; 88] 54 [39 ; 87] 107 [76 ; 179] 45 [30 ; 54] 131 [95 ; 164] 0.29 [0.23 ; 0.35] 1.71 [1.61 ; 1.89] 0.35 [0.28 ; 0.52]

*: failure of the computing loop gain model in two children.

Determinants of arterial desaturation related to apnea

The AHI correlated with the nadir of SpO₂ (R = −0.61; p < 0.001). When analyzing the different types of respiratory events, we found significant relationships between apnea duration and the nadir of SpO₂ in both NREM and REM sleep: R = −0.45, p = 0.017; and R = −0.41, p = 0.032, respectively. By contrast, we found no significant relationship between hypopnea duration and the nadir of SpO₂ in both NREM and REM sleep: R = 0.07, p = 0.680; and R = 0.06, p = 0.768, respectively.

Kinetics of desaturation during NREM and REM sleep related to apnea duration

These kinetics were determined in 13/35 children (median age: 11.6 years [10.2; 13.3], z-score BMI: 2.26 [1.90; 2.61], OAHI: 18.2 events/hour [8.8; 23.8]) in NREM sleep and 14/35 children (median age: 11.8 years [8.2; 12.9], z-score BMI: 2.26 [1.93; 2.62], OAHI: 16.7 events/hour [7.3; 23.8]) in REM sleep.

The equations of arterial desaturation were 0.64 [0.24; 1.03] × apnea duration − 1.51 [−5.37; +1.61] during NREM sleep (r² = 0.53 [0.40; 0.79]) and 0.37 [0.29; 0.46] × apnea duration + 0.59 [−1.03; +3.21] during REM sleep (r² = 0.46 [0.26; 0.60]). The slope during NREM sleep, but not in REM sleep (R = −0.15; p = 0.632), closely correlated with PETO₂ determined during the measurement of loop gain (during wakefulness) (Fig. 1). These slopes did not correlate with FRC (data not shown).

Fig. 1.

Relationship between arterial desaturation and alveolar PO₂ in NREM sleep. Arterial desaturation was characterized by the slope (%.s⁻¹) of linear decrease in SpO₂ during an isolated apnea in NREM sleep. End-Tidal PO₂ (PETO₂) was measured during measurement of tidal ventilation during wakefulness allowing to compute loop gain.

In the 33 children who had loop gain measurement, there were significant logarithmic relationships between PETO₂ and the Z-scores of both X5Hz (R = −0.43; p = 0.031) and AX (R = −0.44; p = 0.026).

Relationships between apnea and hypopnea durations and characteristics of OSA

NREM and REM sleep apnea durations were correlated (R = 0.70; p = 0.0006), as were NREM and REM sleep hypopnea durations (R = 0.71; p < 0.0001), while apnea and hypopnea durations were not correlated in either NREM or REM sleep.

There was a positive correlation between age and hypopnea duration in NREM sleep (R = 0.39; p = 0.021), apnea duration in REM sleep (R = 0.37; p = 0.05), and hypopnea duration in REM sleep (R = 0.41; p = 0.021), while there was no correlation with apnea duration in NREM sleep (R = 0.04; p = 0.850). There was no difference in apnea and hypopnea durations in either NREM or REM sleep between girls and boys (data not shown). There was no correlation of durations with the Z-score of BMI (data not shown). Apnea duration in REM sleep was significantly longer than in NREM sleep (p = 0.025, paired Wilcoxon test), while only a trend for an increase in hypopnea duration in REM sleep versus NREM sleep was evidenced (p = 0.079, paired Wilcoxon test).

Correlates of the MAD4 and MAD≥ 2

If MAD is an endotypic trait of OSA, MAD₄ would be more strongly associated with other indices than MAD_≥2.

Figure 2 shows that the MAD₄ positively correlated with age (panel A), Epworth score (panel C), and medium-frequency controller gain (panel D), while it negatively correlated with medium-frequency plant gain (panel B). Since age may have affected these relationships due to its correlation with the Epworth score and controller and plant gains, we performed a stepwise regression (forward selection approach) with age and the Epworth score (two variables due to the restricted number of subjects) as independent variables and the MAD₄ as the dependent variable, showing that only the Epworth score remained independently linked to the MAD₄(r² of the model = 0.40; p = 0.005). By contrast, when controller and plant gain were further evaluated as independent variables with age, only age remained independently related to the MAD₄.

Fig. 2.

Relationships between MAD and both clinical or endotypic characteristics. Mean apnea-hypopnea duration (MAD) of both NREM and REM sleep (MAD) was calculated from the whole polysomnography. The upper left panel A describes its relationship with age, the lower left panel C describes its relationship with excessive daytime sleepiness (Epworth score), the right upper panel B describes its relationship with medium-frequency plant gain and the lower right panel D describes its relationship with medium-frequency controller gain.

The MAD₄ also positively correlated with RMSSD (Fig. 3, panel A), SD1 (Fig. 3, panel C), fractal dimension (Fig. 3, panel B), and the Lyapunov exponent (Fig. 3, panel D). The MAD₄ did not correlate with the obstructive AHI or oxygenation desaturation index (data not shown). These HRV indices did not correlate with age.

Fig. 3.

Relationships between MAD and HRV indices. Mean apnea-hypopnea duration (MAD) of both NREM and REM sleep (MAD) was calculated from the whole polysomnography. HRV indices were obtained from the whole night. The upper left panel A describes its relationship with RMSSD (an index of parasympathetic modulation), the lower left panel C describes its relationship with SD1 (an index of parasympathetic modulation), the right upper panel B describes its relationship with the fractal dimension (Katz index, higher values indicate greater dynamical complexity) and the lower right panel D describes its relationship with Lyapunov exponent (higher values indicate lower predictability, or higher chaos).

By contrast, MAD_≥2 only correlated with age (R = 0.39, p = 0.021), medium-frequency plant gain (R= −0.35, p = 0.048), medium-frequency controller gain (R = 0.57, p < 0.001) and RMSSD (R = 0.41, p = 0.020).

Discussion

The main results of our retrospective study are the following: first, the rate of arterial desaturation during isolated apnea is mainly related to the preceding arterial PaO₂, which is influenced by small airway disease; second, the apnea–hypopnea duration in both NREM and REM sleep, combined in an index of severity, the MAD₄, correlated with age, sleepiness, and HRV indices, suggesting its usefulness as a specific trait of childhood OSA.

The kinetics of arterial desaturation during apneas were determined in a minority of children with less severe OSA due to the event selection criteria which were mandatory. Sands and colleagues demonstrated that the main determinant of arterial desaturation in patients with OSA is the PvO₂decrease related to previous respiratory events¹¹; we thus selected isolated events, which is easier in children than in adults due to the former’s less severe apnea–hypopnea index. The median slope of arterial desaturation during NREM (Fig. 1) was almost the same to the theoretical slope calculated by Sands and colleagues¹⁰ (0.5%.s⁻¹ versus 0.64%.s⁻¹), and this slope was closely related to PETO₂, as modeled¹⁰, explaining 74% of its variance, which deserves to be demonstrated. The slope of desaturation during REM sleep did not correlate with PETO₂; REM sleep is associated with an irregular breathing pattern that can lower oxygen saturation and explain the loss of the association with diurnal resting PETO₂ measured during calm tidal breathing. We did not evidence statistically significant relationships with FRC, which could be related to the restricted range of these FRCs in our children. Moreover, FRCs were obtained in a sitting position, and their reduction in supine subjects could be variable.

The close correlation of the rate of desaturation with PETO₂ is important because children with almost normal SpO₂ can have different levels of PETO₂, which explains their rate of desaturation. We further demonstrated a significant relationship between PETO₂and indices of peripheral airway obstruction (X5Hz and AX) in these otherwise healthy children, which could be related to obesity and/or OSA. Along this line, it has been demonstrated that OSA is an independent risk factor for small airway disease defined by indices of the forced oscillation technique³³. Furthermore, the correlation of the rate of desaturation with PETO₂also explains why children with cystic fibrosis or asthma have lower nocturnal oxygen saturation nadir^34,35.

The factors determining respiratory event duration in childhood OSA have not been determined to the best of our knowledge. In adults, sex, age, sleep stage, ventilatory control and obesity have been associated with this duration^4,5. Our study found effects of age and sleep stage on event duration (either apnea or hypopnea).

Both MAD₄ and MAD_≥2 correlated with age (Fig. 2, panel A for MAD₄), which was expected³⁶, and more interestingly MAD₄ correlated with sleepiness (Fig. 2, panel C) independently of the AHI, which is an original finding. We recently showed that alterations in sleep architecture were associated to some extent with increased sleepiness in childhood OSA³⁷. Whether the link between respiratory event duration and sleepiness is related to altered sleep architecture in children warrants further studies. Adult patients with OSA and excessive daytime sleepiness are characterized by worse nocturnal oxygenation than those without excessive daytime sleepiness³⁸. Thus, longer respiratory events could explain this link.

To the best of our knowledge, only one study, a large study with 1546 adult participants, has already assessed the relationship between event duration and OSA endotypes⁴. In our study, we found no relationship between the MAD (both MAD₄ and MAD_≥2) and loop gain, which was related to a positive correlation with controller gain and a negative correlation with plant gain (Fig. 2, panel B and D); thus their product (loop gain) did not correlate with this duration index. Longer MAD₄ and MAD_≥2 were associated with increased controller gain (peripheral chemosensitivity) and decreased plant gain. This association could be interpreted as causal or consecutive; we favor the latter possibility. Long apnea will lead to CO₂ accumulation, and a high controller gain will ensure a higher ventilatory response, favoring CO₂ elimination. A low plant gain will further allow a lesser decrease in PaCO₂ and protect against the occurrence of a central apnea due to hypocapnia, favoring ventilatory control stability (no increase in loop gain). Since the stepwise regressions showed that only age remained independently linked to MAD₄, it suggests that the longer MAD₄associated with increased controller gain and decreased plant gain is related to age-related maturation of ventilatory control, at least in OSA. Along this line, we previously showed that plant gain and age-related minute ventilation negatively correlated in children³⁹, and it has also been shown that the hypercapnic ventilatory response (central CO₂chemosensitivity) decreased with age⁴⁰.

Finally, a longer MAD₄ was also associated with increased parasympathetic modulation (RMSSD and SD1: Fig. 3, panel A and C) and increased dynamic complexity (fractal dimension: Fig. 3, panel B) and chaos (Lyapunov exponent: Fig. 3, panel D) of HRV obtained from the whole night. By contrast, MAD_≥2 only correlated with RMSSD, which could be related to the fact that MAD_≥2 is a weaker index of the endotypic trait than MAD₄. Progressive loss of HRV and complexity has been associated with OSA and its severity⁴¹. Patients with OSA have altered fractal dynamics and decreased unpredictability of heart rate oscillations, a condition consistent with the loss of physiological complexity in patients with OSA²². We did not assess these relationships; however, our results are complementary, since the MAD₄, independently of the number of respiratory events, correlated with several HRV indices. Thus, longer events could be associated with better cardiovascular prognoses in children than shorter events, which is in line with the fact that shorter respiratory events are associated with the development of adverse cardiovascular outcomes in adult OSA⁶and that reduced HRV (associated with shorter respiratory events) has also been linked to the risk of incident hypertension⁴². It would be interesting to evaluate the newly designed pharmacological approaches of OSA⁴³ using noradrenergic and antimuscarinic drugs on both HRV indices and MAD.

Event duration has a strong heritable component⁴⁴and predicts mortality⁶, suggesting that this parameter may help characterize OSA subtypes and risk stratification, constituting per se a specific trait, independently of the number of respiratory events. Along this line, a longer mean apnea–hypopnea duration, but not the AHI, has been associated with worse hypertension in adult OSA¹⁸. Overall, longer respiratory events are a trait of older children with excessive daytime sleepiness and higher HRV. It has previously been suggested that the AHI reflected the frequency of respiratory events, while the MAD represented the severity of respiratory events, and that these two indices are relatively independent but may be complementary in the evaluation of patients with OSA¹⁷.

Our study has as limitations its retrospective design and small sample size of children. Only children with moderate to severe OSA were included, since the endotypes of OSA were determined in these specific children to guide a stepwise treatment approach in a routine design. The kinetics of arterial desaturation has been determined in a limited subgroup of children (~ third of participants) due selection criteria (isolated apneas with a normal pre-apneic SpO₂), which is also a limitation. Two MAD indices were calculated (MAD₄ and MAD_≥2) since we hypothesized that an endotypic trait would affect all the events, whatever the sleep stage. Nevertheless, the effects of age and loop gain indices (plant and controller) were confirmed with the two indices. The HRV indices were obtained from the whole night only because our aim was to take into account both the effects of events (respiratory and arousal events) on HRV and the residual effects of these events (periods free of events) on HRV, as a global effect of OSA on HRV. Nevertheless, awakening from sleep (< 10% of sleep period time, Table 1) may have introduced some bias. On the other hand, this is the first study related to respiratory event duration in children, suggesting the interest of MAD calculation.

In conclusion, longer respiratory events are a specific trait of older children with excessive daytime sleepiness and higher HRV. The between night stability of this trait has to be confirmed and the effect of noradrenergic and antimuscarinic drugs on this trait has to be determined.

Author contributions

substantial contributions to the conception or design of the work (PB, CD2); or the acquisition, analysis, or interpretation of data (all authors); or the creation of new software used in the work (PB, BM); or have drafted the work or substantively revised it (all authors) AND to have approved the submitted version (all authors) AND to have agreed both to be personally accountable for the author’s own contributions and to ensure that questions related to the accuracy or integrity of any part of the work, even ones in which the author was not personally involved, are appropriately investigated, resolved, and the resolution documented in the literature (all authors).

Data availability

data are available on corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Conflict of interest disclosure

the authors have no conflict of interest to disclose.

Ethics approval

approved by our local ethics committee (PHENOSAS: N° 2018 − 416).

Financial disclosure: The experimental part (mice model) of the project “DESSAP” devoted to: apnea causes and consequences in childhood, has been funded by the Fondation Avenir (RM-23-031). This funder has no role in study design, collection, analysis and interpretation of data, writing of the report and decision to submit the article for publication.