Circadian blood pressure dysregulation in children with obstructive sleep apnea

Abstract

Study Objectives

Obstructive sleep apnea (OSA) adversely affects normal blood pressure (BP) and may disrupt circadian BP patterns. We sought to examine 24-hour circadian BP rhythms in children with OSA and healthy controls.

Methods

Children 5–14 years with OSA and healthy controls underwent 24-hour BP monitoring and actigraphy to quantify sleep. Shape invariant statistical models compared circadian BP patterns (e.g. times of BP peaks, time arrived at peak BP velocity [TAPV]) in the OSA and control groups.

Results

The analytic sample included 219 children (mild OSA: n = 52; moderate-to-severe OSA (MS-OSA): n = 50; controls: n = 117). In the morning, the MS-OSA group had earlier TAPV for DBP than controls (51 minutes, p < 0.001). TAPV in the evening was earlier for the MS-OSA group than controls (SBP: 95 minutes, p < 0.001; DBP: 28 minutes, p = 0.028). At mid-day, SBP and DBP velocity nadirs were earlier for the MS-OSA group than controls (SBP: 57 minutes, p < 0.001; DBP: 38 minutes, p < 0.01). The MS-OSA group reached most BP values significantly earlier than controls; the largest differences were 118 minutes (SBP) and 43 minutes (DBP) (p < 0.001). SBP and DBP were elevated in the MS-OSA group (hours 18–21 and 7–-12, respectively, p < 0.01) compared to controls. The MS-OSA group was prone to “non-dipping” compared to controls (SBP: odds ratio [OR] = 2.16, 95% CI: 1.09, 4.29; DBP: OR = 3.45, 95% CI: 1.21, 10.23).

Conclusions

Children with MS-OSA had changes in circadian BP patterns, namely earlier TAPV and BP peaks and nadirs than controls. Circadian disturbances in BP rhythms may be key to mapping the natural history of BP dysregulation in children with OSA.

Graphical Abstract

Graphical Abstract.

Statement of Significance.

Obstructive sleep apnea (OSA) can disrupt normal blood pressure (BP) control and is associated with cardiovascular sequelae. In this study, we evaluated 24-hour circadian BP rhythms in children with OSA. This is the first study to identify differences in the timing of peak systolic and diastolic BP velocities in children with OSA compared to healthy controls. The natural history of circadian BP rhythm disruptions in children with OSA is not yet known; the changes in BP peaks and troughs, and how these shifts relate to the onset of hypertension and other cardiovascular sequelae of OSA, must be investigated further. Timing disturbances in BP velocity may be important to understanding the cardiovascular effects of OSA in children.

Introduction

Ambulatory blood pressure (BP) monitoring is frequently used in clinical care and is integral to hypertension research [1, 2]. BP patterns are inherently linked with circadian rhythms, as well as other physiological rhythms, and thus variation in BP predictably occurs within a 24-hour period. Typically, there is a morning BP surge that occurs upon waking, a small decline in the early afternoon followed by an increase, and then a decline leading to a nadir during sleep [3–5]. Disruption of circadian BP patterns has been strongly associated with adverse cardiovascular outcomes [6] and other chronic diseases [7]. Genetic variations in circadian rhythm genes appear to increase susceptibility to myocardial infarction [8]. Preclinical studies indicate that disrupting circadian rhythm by alternating light–dark cycles aggravates atherosclerosis development in APOE*3-Leiden.CETP mice [9]. Excessive morning BP surges [10, 11] and loss of, or reduction in, the physiologic nadir of BP during sleep [12, 13], are examples of problematic alterations in physiologic BP patterns that are associated with cerebral small blood vessel disease [14], left ventricular hypertrophy [15], and cardiovascular diseases. In experiments of forced Desynchrony, circadian misalignment increased wake-time BP, as well as 24-hour average systolic and diastolic BP, and dampened the healthy sleep-associated systolic BP dipping effect linked to increased cardiovascular risk and mortality [16].

BP alterations in obstructive sleep apnea (OSA) are complex and multifactorial as sleep stage, hypoxia, and arousal, among other factors [17], may contribute to BP dysregulation and variability. Those with OSA often have increased morning BP surges, reduced BP dipping, and increased BP variability [17]. Children with severe OSA, relative to healthy controls, have more notable elevations of morning BP, significant increases in BP ceiling/peak values (i.e. BP load), and elevated mean arterial diastolic BPs [18]. Reduced nocturnal dipping, as well as increased variability in BP during wake and sleep states, have also been documented [19]. Cardiovascular sequelae from OSA, such as left ventricular remodeling, have been identified in children with OSA as well [18]. Although patients with OSA are prone to disruptions in circadian systems [20], the nuances of circadian BP patterns among patients with OSA, especially children, have not been fully elucidated.

It is vital to understand the mechanisms underlying BP dysregulation in pediatric OSA; specifically, determining how disruptions in circadian BP rhythmicity contribute to hypertension and other cardiovascular sequelae is essential. Previous studies examining BP patterns in children with OSA have highlighted changes in mean SBP and DBP [21, 22]. However, investigating other BP parameters, including BP trajectory over 24-hour periods and other indices associated with diurnal BP variations, may be valuable. For example, the rate of BP change (velocity), and the timing of velocity changes, in addition to other unique measures, may be informative; these parameters have not yet been studied in the context of pediatric OSA. Quantifying BP with these parameters may provide a more in-depth, temporal understanding of the pathophysiology of OSA and the natural history of cardiovascular changes in children.

To further investigate BP patterns in children with OSA, we aimed to analyze 24-hour SBP and DBP measurements in children using shape-invariant models. We anticipated that a model successfully used to analyze children’s growth patterns, the Super Imposition by Translation And Rotation (SITAR) model [23, 24], could be applied to 24-hour BP data and may more fully identify OSA-related disruptions in BP patterns, including changes in BP rhythmicity. We also investigated the relationship between OSA and dipping and hypothesized that children with OSA may be more prone to non-dipping when compared to healthy controls.

Methods

Study design and population

This study combined data from two prospective longitudinal studies with children ages 5–14 years to compare children with OSA to healthy controls, and has been previously described [25]. Briefly, the OSA group included children with palatine tonsil hypertrophy who were scheduled for treatment with adenotonsillectomy (T&A). These children were recruited from otolaryngology and pulmonary clinics at our tertiary care center and through community advertisements. The inclusion criteria for the OSA group included (1) absence of chronic medical conditions and genetic syndromes and (2) overnight polysomnography (PSG) used to diagnose OSA. OSA was defined by the obstructive apnea–hypopnea index (oAHI) and was categorized as mild if oAHI was ≥ 1 and < 5, and moderate-to-severe if oAHI ≥ 5.

The control group included healthy children age- and gender-matched to those in the OSA groups. Children in the control group were recruited using hospital and community-based advertisements. Additional inclusion criteria for controls were (1) absence of habitual snoring and (2) no OSA on PSG (defined as OI < 1). Children were excluded from the control group if they were on chronic medication and were unable to discontinue the medication temporarily. The control and OSA groups were then compared with regard to multiple BP indices to evaluate for differences in circadian BP patterns. All data collection for the first prospective study occurred between 2003 and 2006, while it was between 2007 and 2013 for the second prospective study. The institutional review board at Cincinnati Children’s Hospital Medical Center approved this study, and parents and patients provided written informed consent and/or assent, respectively.

Twenty-four-hour ambulatory BP monitoring data

All children had at least one 24-hour period of ambulatory BP monitoring. Random days were assigned for both OSA and control groups for BP monitoring. BP measurements were taken within two weeks of the PSG date. During this period, no treatment was introduced. Spacelabs Medical Models Ambulatory BP monitors were used to collect BP measurements. The monitors were carried in pouches that were strapped and/or belted to the side of the patient. The measurements were obtained using a BP cuff attached to the patient’s arm. Monitor-recorded BP data were then downloaded for analysis. SBP and DBP measurements were recorded for all children every 15 minutes within a 24-hour period of ambulatory monitoring and were averaged hourly for each subject. Data were synchronized so that time zero was designated as time of sleep onset; actigraphy readings confirmed time of sleep onset, as well as sleep duration, wake duration, and wake time.

Demographic and clinical characteristics

Demographic data and clinically relevant characteristics were recorded, including sex, age, BMI, BMI Z-score, baseline SBP and DBP, sleep duration, wake duration, wake time, and sleep onset time. Hypertension status was defined as follows: normal: <90th percentile; elevated: ≥90th percentile to < 95th percentile or 120/80 mmHg to < 95th percentile (whichever is lower); stage 1: ≥95th percentile to < 95th percentile + 12 mmHg, or 130/80 to 139/89 mmHg (whichever is lower); stage 2: ≥95th percentile + 12 mmHg, or ≥ 140/90 mmHg (whichever is lower) [26]., Other variables recorded included apnea–hypopnea index (AHI), obstructive AHI (oAHI), sleep efficiency, sleep latency, arousal index, max end-tidal CO₂ (Et CO₂), average Et CO₂, hypoventilation (%), % of N1, N2, and N3, REM, SpO2 average REM, SpO2 average NREM, and SpO2 nadir. Chi-square tests and Kruskal–Wallis rank-sum tests were used to evaluate differences in characteristics by OSA status for categorical and continuous variables, respectively.

Standard BP indices

BP indices for both SBP and DBP, including average BP during the 24-hour period, sleep-only, and wake-only periods, BP load, BP surge, the first hour mean BP, and the last hour mean BP by OSA status were calculated. BP load was defined as the percentage of BP measurements falling at or above the 95th percentile, while the BP surge was defined as the difference between the first hour of wake reading and the last hour of sleep reading. For average BP, we derived the least square means and standard errors from the mixed-effects models; the corresponding P-values were obtained using mixed-effects analysis of variance (ANOVA). We generated means and standard deviations (SDs) for BP load, and P-values were obtained using the Kruskal–Wallis rank-sum test. For BP surge, the first and the last hour BP means, the least square means, standard errors, and the corresponding P-values were obtained using ANOVA.

SITAR model parameters and BP curve derivation

SBP and DBP data were analyzed individually using the SITAR model which accounts for the number of study participants, the time when measurements were taken during a 24-hour period, and OSA status (mild, moderate-to-severe, or control). Fixed-effect parameters in the model included means for each study group of the following: overall 24-hour period mean, time arrived at peak velocity (TAPV or phase), and velocity of BP change. Velocity was defined as rate of BP change per hour.

The fitted mean BP and mean velocity curves over the 24-hour period, for both OSA and control groups, were generated using the overall mean values in coordination with other parameters, separately, for SBP and DBP. These fitted mean curves were used to derive individual fitted curves based on estimated random effects of three parameters (mean BP, velocity, and TAPV). Subject-specific random effects parameters represented individual variation from the corresponding mean SBP and DBP, velocity, and TAPV, respectively. When predicting the SBP and DBP values (group mean and individual), the 24-hour time interval was divided into 1000 time points to ensure that the BP-predicted curves appeared smooth. The SBP and DBP velocity values (group mean and individual) were evaluated by the first-order differential of the predicted SBP and DBP values against time, respectively. The time corresponding to the peak velocity was defined as TAPV and was determined following the grid search algorithm. TAPV could vary from the mean by shifting left on the x-axis (a negative change indicating earlier time arrived at peak velocity compared to the reference category (here, “control”)), or right on the x-axis (similarly, a positive change indicating later time arrived at peak velocity) [23]. Group-specific peak and nadir data were used to estimate control and OSA group-specific amplitudes.

The SITAR models captured the shape and variation in BP measurements over the study period using a natural cubic spline function. The optimal number of knots needed to ensure the best model fit was determined using model selection criteria, namely Akaike information criterion and Bayesian information criterion. An R package sitar [27], developed using a frequentist approach, was used to estimate all model parameters, and all other computations were carried out in the R system [28]. The SITAR model may also be implemented in Bayesian framework [29, 30].

BP dipping

Separate subject-level circadian index (CI) data for SBP and DBP were generated by calculating the percent change in BP during the sleep state compared to the awake state [31]. These CIs for SBP and DBP were later dichotomized to “non-dipping” and “dipping” using the threshold value of 10%; if the percent change was < 10%, this was classified as “non-dipping.” A logistic regression model was used to find the effect of OSA on odds of non-dipping (“dipping” was the reference category).

Results

Study population by OSA status

There were 373 children that initially met inclusion criteria. After including children with at least 15 hours of complete data for hourly measurements (including a minimum of 6 hours of sleep and 6 hours of wake data) over a 24-hour period, the final study population included 219 children (117 healthy controls, 52 children with mild OSA, and 50 children with MS-OSA). Overall, there were 4716 hourly BP observations. Control and OSA groups did not differ significantly regarding average age, sex, BMI, baseline SBP and DBP, sleep duration, wake duration, wake time, and hypertension status. However, the BMI Z-score was significantly different only between MS-OSA and healthy controls (Table 1).

Table 1.

Demographic, Sleep–Wake, and Polysomnographic Characteristics of the Study Population by OSA Status

Characteristics	OSA status			P-value
	Control (n = 117)	Mild (n = 52)	Moderate-to-severe (n = 50)
Demographic and BMI
Male*	48.36%	53.15%	46.95%	0.72
Age (year)†	9.76 ± 2.41	9.26 ± 2.64	9.05 ± 2.52	0.19
BMI†	19.89 ± 4.75	20.48 ± 5.11	21.94 ± 6.05	0.25
Sleep–wake
Sleep duration†	8.72 ± 1.14	8.76 ± 1.41	8.60 ± 1.32	0.781
Wake duration†	12.55 ± 1.71	12.35 ± 1.95	12.52 ± 1.96	0.783
Wake up time (am)†	9.74 ± 1.21	9.86 ± 1.42	9.76 ± 1.32	0.725
BMI Z-score	0.79 ± 0.86	0.95 ± 1.21	1.24 ± 0.97‡	0.015
Baseline SBP	105.66 ± 11.39	107.21 ± 11.71	107.08 ± 12.09	0.717
Baseline DBP	59.27 ± 7.50	59.50 ± 6.16	61.74 ± 7.71	0.236
Hypertension status
Normal	78.63%	65.38%	66.00%	0.14
Stage 1	5.98%	19.23%	12.00%
Stage 2	0.85%	1.92%	4.00%
Elevated	14.53%	13.46%	18.00%
Polysomnographic
AHI events/hour†	1.07 ± 1.18	3.22 ± 1.47‡	14.81 ± 9.58‡	<0.001
oAHI events/hour†	0.31 ± 0.28	2.35 ± 1.13‡	14.19 ± 9.52‡	<0.001
Sleep efficiency†	0.81 ± 0.10	0.80 ± 0.10	0.79 ± 0.11	0.471
Sleep latency	46.12 ± 35.45	52.44 ± 38.34	54.47 ± 40.81	0.408
Arousal index events/hour†	8.93 ± 2.95	9.82 ± 2.23‡	17.02 ± 9.98‡	<0.001
Max EtCO2 mmHg†	50.42 ± 4.76	49.58 ± 4.32	51.99 ± 5.39	0.103
Average EtCO2 mmHg†	43.94 ± 2.60	43.16 ± 3.88	41.41 ± 7.26	0.197
%Hypoventilation*	0.85%	7.49%	10.00%‡	0.016
% of N1	3.28 ± 3.70	2.98 ± 1.53	3.65 ± 1.93	0.132
% of N2	47.42 ± 8.41	46.86 ± 7.71	46.76 ± 9.14	0.769
% of N3	2.46 ± 3.10	2.26 ± 1.49	2.43 ± 1.13	0.131
REM	19.96 ± 4.78	20.37 ± 5.56	19.40 ± 5.91	0.706
SpO2 average REM	97.53 ± 1.57	95.72 ± 11.78	96.93 ± 1.55‡	0.020
SpO2 average NREM	96.85 ± 3.79	95.29 ± 11.67	96.92 ± 1.33	0.392
SpO2 nadir	96.71 ± 3.92	91.23 ± 3.81‡	86.37 ± 8.79‡	<0.001

^* P-value is obtained by chi-squared test.

^†Tests have been performed by using the Kruskal–Wallis rank-sum test, a non-parametric approach, as the populations do not satisfy the normality assumption. To test the normality of residuals obtained from the analysis of variance, the Shapiro–Wilk test is used.

^‡Significantly different from control.

AHI, oAHI, arousal index, hypoventilation (%), SpO2 average during REM, and SpO2 nadir differed significantly according to OSA (Table 1). The pairwise comparison showed that the percent of participants with hypoventilation and SpO2 average during REM were significantly different only between MS-OSA and healthy controls. Other characteristics, including sleep efficiency, sleep latency, max end-tidal CO₂ (Et CO₂), average Et CO₂, % of N1, N2, and N3, REM, and SpO2 average during NREM did not differ based on OSA severity (Table 1).

Standard BP indices by OSA status

There was a statistically significant difference in the average DBP during sleep only (p = 0.006). During sleep, children with MS-OSA showed higher mean DBP than healthy controls (58.98 mmHg vs. 55.93 mmHg). During the first hour, the DBP means were also significantly different (p = 0.048) by OSA status. Otherwise, the standard BP indices, including average BP during 24-hour period, sleep and wake time, BP load, BP surge, and 24-hour mean BP did not differ significantly by OSA status (Supplementary Table S1).

Circadian SBP and DBP rhythms

We present results generated from the SITAR model comparing healthy controls and children with MS-OSA. The SITAR model takes into consideration the nonlinear trajectory of 24-hour BP and shows significant differences in the TAPV and velocity parameters when comparing healthy controls with MS-OSA; no significant differences were identified between controls and children with mild OSA for any of the BP indices investigated (Supplementary Table S2). Individual SBP and DBP predicted curves for healthy controls and MS-OSA children are presented in Supplementary Materials, Figure S1.

Time arrived at peak velocity.

Predicted SBP and DBP velocities by OSA status are displayed in Figure 1, and individual-level predicted velocities are displayed in Supplementary Materials, Figure S2. The time arrived at each SBP and DBP peak velocity, and the times of mid-day velocity nadirs, are displayed in Table 2. In the morning, TAPV for SBP differed between the MS-OSA group and controls, but this difference was not statistically significant (19 minutes, 10.55 (1.62) vs. 10.87 (1.51), p = 0.24). However, at mid-day, the SBP velocity nadir was significantly earlier in the MS-OSA group compared to controls (57 minutes, 14.48 (1.48) vs. 15.43 (1.73), p < 0.001). TAPV in the evening was also significantly earlier by 95 minutes for the MS-OSA group than controls (18.68 (2.24) vs. 20.27 (2.01), p < 0.001).

Figure 1.

The estimated 24-hour SBP and DBP velocity of control (solid line) and moderate-to-severe OSA (dotted line) groups are exhibited in panels A and B, respectively. The solid and dotted vertical lines indicate the times arrived at peak (or, nadir) velocity for control and moderate-to-severe OSA groups, respectively.

Table 2.

Time Arrived at SBP and DBP Velocities (Peaks and Nadirs) by Group

Feature	SBP				DBP
	Control	Moderate-to-severe OSA	Difference (mins)	p	Control	Moderate-to-severe OSA	Difference (mins)	p
	Mean (SE)	Mean (SE)			Mean (SE)	Mean (SE)
Time arrived at first peak velocity	10.87 (1.51)	10.55 (1.62)	19	0.24	10.38 (1.68)	9.53 (1.43)	51	<0.001
Time arrived at mid-day nadir	15.43 (1.73)	14.48 (1.48)	57	<0.001	15.33 (1.25)	14.70 (1.37)	38	<0.001
Time arrived at second peak velocity	20.27 (2.01)	18.68 (2.24)	95	<0.001	19.45 (0.66)	18.98 (1.29)	28	0.028

In the morning, TAPV for DBP was significantly earlier for the MS-OSA group than in healthy controls (51 minutes, 9.53 (1.43) vs. 10.38 (1.68), p < 0.001) (Table 2). At mid-day, the timing difference of DBP velocity nadir between the MS-OSA group and controls was 38 minutes, with the MS-OSA group reaching the velocity nadir earlier (14.70 (1.37) vs. 15.33 (1.25), p < 0.01). TAPV in the evening was significantly earlier as well for the MS-OSA group compared to controls (28 minutes, 18.98 (1.29) vs. 19.45 (0.66), p = 0.028).

Time differences in reaching BP peaks, nadirs, and various SBP and DBP values by group

The predicted times when the control group and the MS-OSA group reached SBP and DBP peaks and nadirs are displayed in Table 3. The MS-OSA group reached each peak and nadir significantly earlier than controls (Table 3). The greatest time difference to reaching an SBP peak was 118 minutes (late evening peak), and for DBP the greatest time difference was 43 minutes (early afternoon BP peak, Table 3). Additional detailed comparisons for time arrived at various SBP and DBP values are displayed in Supplementary Materials, Table S3. The MS-OSA group reached most SBP and DBP values significantly earlier than the controls.

Table 3.

Timing of SBP and DBP Peaks and Nadirs by Group

Feature	SBP				DBP
	Control	Moderate-to-severe OSA	Difference (mins)	p	Control	Moderate-to-severe OSA	Difference (mins)	p
	Mean (SE)	Mean (SE)			Mean (SE)	Mean (SE)
First BP peak	13.42 (0.52)	12.77 (0.72)	39	0.032	13.48 (0.50)	12.77 (0.59)	43	<0.001
Mid-day nadir	17.48 (0.51)	16.28 (0.68)	72	<0.001	17.35 (0.58)	16.80 (0.70)	33	0.003
Second BP peak	11.08 (0.45)	9.12 (0.47)	118	<0.001	9.55 (0.49)	9.17 (0.69)	23	0.003

Twenty-four-hour BP curves: healthy controls versus moderate-to-severe OSA.

The predicted BP rhythms for the MS-OSA and control groups are depicted in Figure 2, Panels A and B. Mean hourly time point comparisons for SBP and DBP are illustrated in Figure 3. At each hour from 18–21, the MS-OSA group had significantly higher SBP than controls (p < 0.01 for each hour). At the 24–hour, the MS-OSA group had lower SBP compared to controls (p < 0.001). DBP was significantly higher in those with MS-OSA compared to healthy controls at every hour from 1 to 2, 7 to 12, and at 19; only at hour 24 was DBP for the MS-OSA group significantly lower than controls (p < 0.05 for all comparisons).

Figure 2.

The estimated 24-hour SBP and DBP trajectories by OSA groups (control: solid line; and moderate-to-severe OSA: dotted line) are exhibited in panels A and B, respectively.

Figure 3.

Error bar plots with mean ± SE of predicted SBP (A) and DBP (B) by control and moderate-to-severe OSA and children at every hour. At the top of each error bar, “*” indicates a significant P-value from testing the hourly difference between the control and moderate-to-severe OSA groups. Two-tailed t-tests were used to detect significant differences in the predicted hourly mean BPs.

Odds of non-dipping: healthy controls versus moderate-to-severe OSA

Estimates from logistic regression models indicated that the MS-OSA group was significantly prone to “non-dipping” compared to controls for SBP and DBP (SBP: odds ratio [OR] = 2.16, 95% CI: 1.09, 4.29; DBP: OR = 3.45, 95% CI: 1.21, 10.23; Table 4). Along with OSA, when BMI Z-score was added as a covariate, it did not show any significant effect on non-dipping status, and the OR estimates of OSA remained very similar.

Table 4.

Odds Ratio Estimates of OSA Group for the Event of Non-dipping from the Logistic Regression Model^*

OSA Group	SBP		DBP
	OR (95% CI)	p	OR (95% CI)	p
Healthy control	1.00		1.00
Mild OSA	1.46 (0.73, 2.90)	0.2749	3.74 (1.35, 10.92)	0.0120
Moderate-to-severe OSA	2.16 (1.09, 4.29)	0.0265	3.44 (1.21, 10.23)	0.0209

^*In each logistic regression model, the covariates: age, sex, and BMI were not included since their effects were not statistically significant.

Discussion

This study is the first to identify differences in TAPV, as well as the timing of systolic and diastolic BP peaks and nadirs, in children with MS-OSA compared to healthy controls. The differences consisted of a shift in velocity and BP curves to the left, meaning an earlier acceleration of BP velocity and rise in BP; compared to healthy children, TAPV and the timing of BP peaks and troughs were disrupted in children with MS-OSA. Together with the increased odds of non-dipping, our findings suggest a deviation from normal circadian rhythm in children with MS-OSA. Our model identified these previously uncharacterized BP rhythm disturbances by analyzing thousands of BP data points. Hourly BP comparisons over 24 hours demonstrated distinct time periods when SBP and DBP were elevated or decreased in children with MS-OSA compared to controls; the patterns identified for SBP and DBP were distinct. Specifically, SBP was elevated in the evening (hours: 18–21) in children with MS-OSA compared to controls, and DBP was elevated in the morning and early afternoon (hours: 1–2 and 7–12). The SBP and DBP nadirs were significantly lower for those with MS-OSA than controls around midnight as well, demonstrating substantial swings in BP between peaks and troughs. These subtle disturbances in BP rhythms provide a more in-depth, temporal understanding of the natural history of cardiovascular changes in children with OSA.

OSA is believed to disrupt circadian rhythms through frequent arousals and intermittent hypoxia [20, 32]. Circadian rhythms are regulated by retinal ganglion cells that sense light and communicate with the suprachiasmatic nuclei in the hypothalamus [33]. However, peripheral circadian clocks exist throughout the body and regulate gene transcription for a host of physiologic processes [20]. Intermittent hypoxia, a main feature of OSA, interferes with circadian gene expression patterns [32] that regulate cardiovascular health and other physiologic processes [20]. Fragmented sleep and repeated hypoxic episodes also stress the cardiovascular system and interfere with relative rest periods when cellular growth, repair, and metabolic changes occur [34, 35]. Certain molecules are known to be diurnally regulated and appear necessary to counter-regulate inflammatory processes [36, 37]. For example, one such molecule, RvDn-3, which has a role in vascular protection, is reduced 3-fold in the early morning hours; this reduction has been linked to the onset of cardiovascular disease [36]. When components of the cardiovascular system are altered (such as peak BP), changes in inflammatory responses, or the times at which anti-inflammatory mechanisms are up- and down-regulated, may occur.

Among adults, OSA has been associated with impaired nocturnal BP dipping, referred to as “non-dipping,” and more pronounced variability in BP [17, 35]. “Non-dipping” has been associated with an increased risk of cardiovascular sequelae in many studies, though not universally [38–42], and dichotomous categorizations such as dipping and non-dipping are likely inadequate to fully characterize cardiac risk. Recent evidence suggests that subgroups, such as reverse or extreme dippers, may be important to consider when evaluating risk for cardiac outcomes [41] and that the relationship between the extent and type of BP dipping may be affected by age [43] among adults.

Existing studies indicate that children with OSA are prone to increased sympathetic tone, elevated BP, impaired nocturnal dipping, and increased variability in diurnal and nocturnal BP [19, 44–48]. Our results indicate that children with MS-OSA are prone to “non-dipping” when compared with healthy controls, a finding previously identified [19]. The extent of “non-dipping,” and the timing of onset, are important to consider in the context of other evolving cardiovascular changes in children, circadian, and non-circadian. Increased severity of OSA has been associated with elevated BP [49] and was found to be positively associated with SBP after adjusting for body habitus and other covariates [47]. Other aspects of BP, such as TAPV, have not yet been explored in children. We suspect that subtle changes in TAPV may be important as they may predate overt hypertension and other subclinical cardiac changes that coincide with underlying inflammatory pathways; however, the clinical relevance of changes in TAPV has not been definitively established. Changes in BP, including disruptions in circadian rhythmicity, and other cardiovascular changes from OSA, likely occur on a spectrum and develop gradually [48]. The exact chronology of these changes is not yet known but is essential to determine [48]; evidence indicates that adverse effects on BP from OSA persist from childhood to adulthood, even when OSA resolves over time [50].

The changes we identified in DBP in children with OSA are noteworthy and distinct from changes in SBP. We identified significant DBP elevation in the morning hours among children with MS-OSA between hours 1 and 2, as well as from 7 to 12, and more prominent differences in SBP between MS-OSA and control groups in the evening. This is consistent with our previous observation of a significant association between OSA severity and diastolic BP in children [18]. The DBP pattern noted in children with OSA should also be considered in the context of relevant adult literature. In adults, MS-OSA has been specifically associated with DBP and diastolic dysfunction [51, 52]. One study of 1171 adult patients with complaints suggestive of OSA found that patients with severe OSA had higher morning DBP compared to healthy participants or those with mild to moderate OSA [52]. In contrast, there was no association with changes in SBP. Although increased DBP and myocardial diastolic dysfunction have been identified in children with OSA [25], the natural history of circadian rhythm disruptions in BP in children with OSA is not yet known. The changes in BP rhythm peaks and troughs, and how these shifts relate to the onset of hypertension and other cardiovascular sequelae of OSA, must be investigated further.

Despite understanding the endogenous and independent nature of BP patterns, and how and why overt hypertension is associated with an increased risk of cardiovascular events, little information exists regarding subclinical changes that may occur in untreated OSA prior to the development of overt end-organ damage [53]. Epidemiologic data demonstrate a time-of-day dependent pattern in adverse cardiovascular events, usually between 6:00 am and noon [54–58]. However, patients with OSA experience a higher incidence of cardiovascular events at night; these events are thought to be related to hemodynamic and autonomic responses to hypoxic events _ENREF_57 [59]. These data suggest that rhythmicity of cardiovascular responses exists and is altered in the setting of disease.

Many factors may contribute to disturbances in BP rhythmicity in children and adolescents (e.g. age, sex, hormonal fluctuations, and multiple possible environmental factors). However, disruptions in the circadian clock must not be discounted as a potentially powerful contributing factor to abnormal BP patterns. It is possible that changes in the timing of BP peaks and troughs could impact how the body compensates for cardiovascular stress. Such mechanisms have not been identified to date, but could be a focus of future studies. Alternatively, these disruptions in the circadian rhythm of BP may be an early indicator of more serious physiologic perturbations to come. Regardless, quantifying circadian BP patterns may be useful to assess progression of cardiovascular sequelae of OSA, and response to treatment, if circadian BP patterns are correlated with cardiovascular outcomes.

This study has several limitations. Our sample included children recruited at a single institution, which may limit the generalizability of our results. The completeness of BP data (89.7%) is also important to consider as some children did not have BP measurements at every hour. The estimates generated may be less reliable for the selected time points where there was a greater impact of missing data. This issue is mitigated in part, but not completely, by our statistical model; gaps in data that were addressed by our model may have introduced errors in estimates. Sleep and wake states were not consistent in all children and variations in timing of sleep and wake states may have affected our results. Although changes to the 24-hour BP rhythms are influenced by dysregulation of the circadian clock, orchestration of these complex cardiovascular processes are multifactorial, likely the result of control from the circadian clock, the autonomic nervous system, and renal function [60]. Future studies should be conducted to further delineate the nature of these findings in children with this complex disorder. Finally, some additional factors should be considered when interpreting results. BMI Z-score and SpO2 nadir were not included in the SITAR models to avoid multicollinearity and model overfitting; oAHI was; however, significantly correlated with BMI Z-score (though not BMI) and SpO2 nadir. Adding these variables to the SBP model attenuated the effect of OSA on SBP, BMI Z-score significantly predicted mean SBP, and TAPV and velocity parameters were significantly associated with SpO2 nadir. The DBP model showed BMI Z-score was significantly associated with mean BP, which could be due to multicollinearity.

Conclusions

Children with MS-OSA have circadian disruptions in BP patterns, namely disrupted TAPV and earlier arrival at SBP and DBP peaks and troughs than controls; these changes are likely to go unrecognized clinically with routine BP monitoring. Circadian disruptions in BP that occur in children with OSA need to be quantified in larger samples, over time, and in relation to the onset of hypertension and other cardiovascular sequelae, such as changes in cardiac structure and function. Defining the natural history of circadian BP disruptions in OSA will likely inform diagnostic and therapeutic approaches if interruptions of normal circadian patterns are early signs of moderate-to-severe cardiovascular disease.

Funding

The study was supported by grants: National Institutes of Health (NIH) R01HL080670 and R01HL070907.

Data Sharing

The data is available upon request to other investigators.

Conflict of Interest

This project was totally federally funded. None of the investigators have relationships that may influence the findings and the conclusion described in the manuscript.

Disclosure Statement

Financial disclosure: None. Nonfinancial disclosure: None.

Clinical Trial Registration: Clinical trials were registered with ClinicalTrials.gov. The unique identifier numbers, trial names, and URL were: NCT00059111, Mechanisms Mediating Cardiovascular Disease in Children With Obstructive Sleep Apnea, URL: https://www.clinicaltrials.gov/ct2/show/NCT00059111

NCT01837459, A Research Study for Children About Heart Changes and Obstructive Sleep Apnea (OSA), URL: https://clinicaltrials.gov/ct2/show/NCT01837459