Variations in loop gain and arousal threshold during NREM sleep are affected by time of day over a 24-hour period in participants with obstructive sleep apnea

Abstract

We investigated whether time of day affects loop gain (LG) and the arousal threshold (AT) during non-rapid eye movement (NREM) sleep. Eleven men with obstructive sleep apnea (apnea-hypopnea index > 5 events/h) completed a constant-routine protocol that comprised 3-h sleep sessions in the evening [10 PM (1) to 1 AM], morning (6 AM to 9 AM), afternoon (2 PM to 5 PM), and subsequent evening [10 PM (2) to 1 AM]. During each sleep session LG and the AT were measured during NREM sleep with a model-based approach. Our results showed the presence of a rhythmicity in both LG (P < 0.0001) and the AT (P < 0.001) over a 24-h period. In addition, LG and the AT were greater in the morning compared with both evening sessions [6 AM vs. 10 PM (1) vs. 10 PM (2): LG (1 cycle/min): 0.71 ± 0.23 vs. 0.60 ± 0.22 (P = 0.01) vs. 0.56 ± 0.10 (P < 0.001), AT (fraction of eupneic breathing): 1.45 ± 0.47 vs. 1.28 ± 0.36 (P = 0.02) vs. 1.20 ± 0.18 (P = 0.001)]. No difference in LG and the AT existed between the evening sessions (LG: P = 0.27; AT: P = 0.24). LG was correlated to measures of the hypocapnic ventilatory response (i.e., a measure of chemoreflex sensitivity) (r = 0.72 and P = 0.045) and the critical closing pressure (i.e., a measure of airway collapsibility) (r = 0.77 and P = 0.02) that we previously published. We conclude that time of day, independent of hallmarks of sleep apnea, affects LG and the AT during NREM sleep. These modifications may contribute to increases in breathing instability in the morning compared with other periods throughout the day/night cycle in individuals with obstructive sleep apnea. In addition, efficaciousness of treatments for obstructive sleep apnea that target LG and the AT may be modified by a rhythmicity in these variables.

NEW & NOTEWORTHY Loop gain and the arousal threshold during non-rapid eye movement (NREM) sleep are greater in the morning compared with the afternoon and evening. Loop gain measures are correlated to chemoreflex sensitivity and the critical closing pressure measured during NREM sleep in the evening, morning, and afternoon. Breathing (in)stability and efficaciousness of treatments for obstructive sleep apnea may be modulated by a circadian rhythmicity in loop gain and the arousal threshold.

INTRODUCTION

Obstructive sleep apnea is characterized by repeated episodes of apneas and hypopneas, which are associated with declines in oxygen desaturation and increases in respiratory effort that frequently result in arousal from sleep. At least four traits (i.e., mechanisms) are involved in the development of obstructive sleep apnea, with the contribution of each trait varying in a given individual (36, 49). These traits include the degree of collapsibility of the upper airway, the ability of the upper airway to dilate or constrict in response to modifications in the partial pressure of carbon dioxide, the degree of stability of the ventilatory control system (i.e., loop gain), and modifications to the arousal threshold (see Loop gain and arousal threshold analysis for definitions) (36, 49). Investigators have begun to explore novel therapeutic treatments that target these traits in order to mitigate the outcomes associated with the complex pathology of obstructive sleep apnea (18, 49). In line with these efforts, the efficaciousness of these treatments can only be ascertained if the biological factors that modify the traits outlined above are identified.

Several studies have suggested that increases in loop gain are responsible for increasing the severity of apneic events (10, 36, 37). Numerous biological factors modify loop gain, including sex (55), race (35), age (13), and intermittent hypoxia (3). Findings from our laboratory indicate that loop gain might be modulated by time of day (14). However, no studies have been completed to support this speculation. Nevertheless, loop gain is the product of controller gain and plant gain (50). Controller gain reflects chemoreflex sensitivity to blood gas modifications, which we showed is modulated by time of day, independent of hallmarks of sleep apnea (e.g., intermittent hypoxia and sleep fragmentation) (14). On the basis of these relationships, we hypothesized that loop gain is modulated by time of day and is greater in the morning compared with the evening and afternoon. We also hypothesized that the ventilatory response to arousal would be modulated by the time of day, because controller gain might impact the ventilatory response to chemical stimuli (i.e., hypoxia and hypercapnia) at the termination of an apneic event (10, 28, 29).

Previous studies have also indicated that the arousal threshold is greater (i.e., it is more difficult to arouse from sleep in response to respiratory stimuli) in the morning compared with the evening (6, 8, 33, 42). Many studies have suggested that modulation of the arousal threshold from the evening to the morning is a consequence of exposure to repeated apnea. Investigators have proposed that inflammation, edema, and neural damage in response to snoring-related vibrations could progressively dampen excitatory inputs from upper airway sensory receptors from the beginning to the end of the night, leading to alterations in the arousal threshold (6, 8, 33). Likewise, increased exposure to intermittent hypoxia from the beginning to the end of the night could produce a similar result (3). Despite these results, no studies to our knowledge have explored whether the arousal threshold is altered during sleep, over a 24-h period, independent of the hallmarks of sleep apnea.

Thus, the primary aim of the present investigation was to explore the effect of time of day on loop gain and the arousal threshold in individuals with sleep apnea during non-rapid eye movement sleep, using a constant-routine protocol (14, 15). A software program (19, 21, 50) designed to measure loop gain and the arousal threshold was used to avoid more invasive methods that we and others have used previously to measure chemoreflex sensitivity (i.e., nasal positive-pressure mechanical ventilation to induce hyperventilation; 9, 14, 17) and the arousal threshold (i.e., epiglottic or esophageal pressure measurements; 8, 12, 42, 53, 56). Secondarily, we were interested in determining whether the modulation of loop gain or arousal threshold over a 24-h period is correlated with measures of chemoreflex sensitivity and/or upper airway collapsibility (i.e., critical closing pressure). We previously showed that these variables were modulated by time of day (14, 15), in the same individuals who participated in the present investigation. We were interested in exploring these relationships to confirm that loop gain reflects modifications in chemoreflex sensitivity. In addition, we were interested in obtaining additional support for previous findings that showed that other indirect indexes of upper airway collapsibility (i.e., degree of flow limitation, the apnea-hypopnea index, and the ratio of hypopnea to apnea-hypopnea index) were positively correlated to loop gain and the arousal threshold (37).

METHODS

Ethical Approval

Participants were informed about the requirements, benefits, and potential risks of the study before providing their written informed consent to participate. The experimental procedures adhered to the standards set by the Declaration of Helsinki. The Institutional Review Boards of the Wayne State University School of Medicine and the John D. Dingell Department of Veterans Affairs Medical Center approved the experimental protocol and procedures.

Protocol and Procedure

Eleven male participants with untreated obstructive sleep apnea and no additional comorbidities (i.e., hypertension, heart and lung disease, morbid obesity) were enrolled in the study. The participants made six visits to our laboratory. Measures of the arousal threshold and loop gain (see results), which are the focus of this manuscript, were obtained during visit 4 of the protocol. Measures of chemoreflex sensitivity and upper airway collapsibility that were correlated with loop gain (see results) were obtained from visits 5 and 6. The results from visits 5 and 6 have been published previously (14, 15). The reader is referred to these publications for details related to the methods used to determine chemoreflex sensitivity and upper airway collapsibility (14, 15).

On their first visit, the participants provided written informed consent. On the same visit, a physical examination, health and lifestyle questionnaires, and a pulmonary function test were completed. Blood pressure and a 12-lead ECG were also measured. If the inclusion criteria were met, the participants completed a baseline nocturnal polysomnogram on their second visit, to verify the presence of obstructive sleep apnea. Upon confirmation, the participants were given an Actiwatch (Actiwatch Spectrum; Philips Respironics, Murraysville, PA) to wear at home for 2 wk to monitor their sleep-wake schedule before physiological measurements were obtained during subsequent visits. Throughout the 2-wk period, we asked the participants to adhere to a regular sleep-wake schedule, with a sleep onset time between 10 and 11 PM and a wake time between 7 and 8 AM. We also requested that the participants refrain from daytime napping. These measures were implemented to establish a similar circadian rhythm in all the participants. During the 2-wk period, the participants returned to the laboratory for their third visit. During this visit the therapeutic continuous positive airway pressure required to maintain airway patency was determined, so that it could be applied during data collection on visits 5 and 6 (see below). After the 2-wk monitoring period, the participants returned to the laboratory to complete visits 4–6. Each of these visits was separated by 7 days, and the participants were asked not to consume alcohol or any caffeinated beverages at least 1 day before the visits. On each of the last three visits (visits 4–6) the participants completed a constant-routine protocol. Participants arrived at the laboratory at approximately 8 PM. Upon arrival, participants ingested a radiotelemetry pellet (CorTemp Sensor, Palmetto, FL), which measured core body temperature every 10 s throughout each visit. This measure was used to establish the nadir of core body temperature. After instrumentation, the participants completed a constant-routine protocol. The constant-routine protocol completed on visit 4 comprised 3-h sleep sessions in the evening [10 PM (1): 10 PM–1 AM], morning (6 AM: 6–9 AM), afternoon (2 PM: 2–5 PM), and subsequent evening [10 PM (2): 10 PM–1 AM]. The protocol on visits 5 and 6 was similar to the protocol on visit 4, with the exception that the second evening sleep session was not included. This protocol was selected so that the total amount of sleep obtained over 24 h was similar to that obtained during a typical night of sleep. Likewise, the duration of each session was selected to ensure that an adequate amount of non-rapid eye movement sleep could be obtained for each session. Extending the length of each sleep session would likely have resulted in a reduced amount of sleep in the latter sessions, precluding us from adequately exploring time-of-day effects over a 24-h period.

The participants slept in the supine position during all sleep sessions and were placed in a semirecumbent position during wakefulness. During wakefulness the laboratory was dimly lit (i.e., 30 lux) and participants were not exposed to sunlight or any external cues including phones, clocks, radio, or television. During the 5-h wake period between each sleep session participants initially watched a movie for ~120 min, read for ~90 min, and finally sat quietly and did not engage in activity for 90 min before the morning and afternoon sleep sessions. During each wake session, participants received small snacks every 95 min composed of ~15% fat, 75% carbohydrate, and 10% protein. Moreover, participants typically drank 1 L of water over the length of the constant-routine protocol. The data presented in this study were obtained from 11 participants who completed visit 4. We obtained measures of the hypocapnic ventilatory response from 8 of the participants (14) and measures of the critical closing pressure from 11 participants (15).

Polysomnography

During the constant-routine protocol, sleep was monitored with an electroencephalogram (C3/A2, C4/A1, O1/A2, and O2/A1), electrooculograms, submental electromyogram, and an electrocardiogram. Abdominal and thoracic wall movements were measured with inductive plethysmography (Respitrace; Ambulatory Monitoring, Ardsley, NY). Airflow, breathing frequency, and inspiratory and expiratory volumes were obtained with a pneumotachometer (model RSS100-HR; Hans Rudolph, Shawnee, KS) connected to a nasal/facemask. End-tidal oxygen (model 17515; Vacumed, Ventura, CA) and end-tidal carbon dioxide (model 17518; Vacumed) were obtained from expired air through sampling tubes connected to built-in ports on the nasal/facemask. Mask pressure was measured by a port on the nasal mask allowing the connection of a pressure transducer. In addition, epiglottic pressure was measured by using a transducer-tipped catheter (Mikro-Cath 825-0101; Millar, Houston, TX) to determine the nature of events, and the oxygen saturation was measured with a pulse oximeter (Biox 3700; Ohmeda, Boulder, CO). All physiological parameters were converted from analog to digital at a sampling frequency of 100 Hz/channel and then fed into a computer using a commercially available software package (gamma version 4.0; Astro-Med, West Warwick, RI).

Data Analysis

Polysomnograms.

All polysomnography studies were analyzed for sleep stage, arousals, and respiratory-related events according to standard published criteria (5a). An absence or a ≥90% reduction in airflow for a duration of ≥10 s was identified as an apnea. A hypopnea was defined by a ≥30% reduction in airflow that lasted for a minimum of 10 s. To be classified as a respiratory-related event, the reduction in airflow was accompanied by either an arousal or a ≥3% drop in oxygen saturation in the absence of an arousal. Obstructive events were scored based on progressive increase in ventilatory effort that was evident in epiglottic pressure measurements and in abdominal and thoracic wall movement. Arousals were identified by a significant increase in electroencephalography frequency and amplitude for ≥3 s. The start and end of arousals (i.e., arousal duration) were also determined.

Loop gain and arousal threshold analysis.

A MATLAB (MathWorks, Natick, MA) program, which has been previously described (19, 21, 50), was used to quantify loop gain and the arousal threshold from the physiological measures obtained from participants on visit 4. The program identified 7-min windows of non-rapid eye movement or rapid eye movement sleep, and the dominant sleep stage that comprised >50% of the window was determined. Windows without a dominant sleep stage were excluded from data analysis. The window duration of 7 min was selected to allow ~10 obstructive events to occur, which was based on an average interevent interval of ~40 s. This interval was sufficient to distinguish the chemical drive and arousal contribution to the total ventilatory output during and after an event. A categorical breath-by-breath time sequence was created for each window, and the breaths associated with an EEG arousal and/or flow limitation were identified. Thereafter, the program measured loop gain and the arousal threshold.

The program is developed on the premise that obstructive breathing events provide a disturbance to the ventilatory control system leading to increases in carbon dioxide and decreases in oxygen that augment the ventilatory drive (50). The increased ventilatory drive is reflected in the extent of hyperventilation that occurs after the airway is open following termination of an obstructive event. Accordingly, ventilatory drive is modeled as the sum of the ventilatory response to a chemical drive (i.e., changes in the partial pressure of oxygen and carbon dioxide) and a nonchemical drive (i.e., wakefulness drive) that accompanies arousal (50). Furthermore, the time sequence of the ventilatory response to chemical stimuli is modeled with parameters that reflect the circulation time between the lung and chemoreceptors (i.e., time delay), the time course of carbon dioxide buffering in the lung and tissues (i.e., time constant), and the overall gain of the response (50). These temporal parameters coupled with the ventilatory response to arousal are modified until the modeled ventilatory drive closely fits the ventilation measured when the airway is not obstructed. These parameters are then used to compute the magnitude of loop gain at the natural frequency (LG_n) of obstructive events and at a frequency of 1 cycle/min (LG₁; i.e., 60 events/h) for each window (50). A frequency of 1 cycle/min was selected to be consistent with the timing of apneic events and with other published reports. For each window, the ventilatory drive measured immediately before each arousal at the termination of a respiratory event was identified. The mean of the ventilatory drive values is considered to be the arousal threshold.

Measures of loop gain and the arousal threshold were divided into 30-min bins for each session of non-rapid eye movement sleep [i.e., 10 PM (1), 6 AM, 2 PM, 10 PM (2)]. Thereafter, rhythmicity was determined by fitting the loop gain or arousal threshold data with a sine function (SigmaPlot; SPSS, Chicago, IL) {y = y_o + a sin[(2πx/b) + c]}, where y is the loop gain or arousal threshold, y_o is the rhythm-adjusted mean (i.e., mesor), a is the amplitude of the rhythm, x is time (minutes), b is the period of the rhythm, and c is the time corresponding to the peak value (acrophase).

In addition, an average value for loop gain, arousal threshold, time constant, time delay, and ventilatory response to arousal along with indexes of sleep architecture and apnea/hypopnea severity were determined for each participant for each sleep session [i.e., 10 PM (1), 6 AM, 2 PM, 10 PM (2)]. Analysis of loop gain and the arousal threshold was only determined for non-rapid eye movement sleep because participants often did not experience rapid eye movement sleep, particularly during the evening sessions of the protocol. In addition, because differences in loop gain and the arousal threshold across sleep sessions were independent of sleep stage (i.e., N1 and N2, see results) the data for non-rapid eye movement sleep were combined.

Statistical Analysis

Measures of loop gain and arousal threshold that were divided into 30-min bins were fit with a sine function using the method of least squares (SigmaPlot 14.0; Systat Software, Inc., San Jose, CA). Rhythmicity was inferred from a significant analysis of variance corrected for the mean of the observations. In addition, a one-way analysis of variance with repeated measures coupled with a Fisher’s least significant difference post hoc test was used to compare measures of the average loop gain, arousal threshold, and the ventilatory response to arousal during the four sleep sessions [i.e., 10 PM (1), 6 AM, 2 PM, and 10 PM (2)] on visit 4. Pearson’s correlation coefficient was used to correlate loop gain with measures of chemoreflex sensitivity and upper airway collapsibility. Pearson’s correlation coefficient and Spearman rank correlation were used to correlate the arousal threshold with upper airway collapsibility. Individual participant results are presented with a corresponding mean or as a mean ± standard deviation. A value of P ≤ 0.05 was statistically significant.

RESULTS

Table 1 shows the anthropometric variables measured for the group (n = 11 subjects). The participants were young to middle age and not morbidly obese, as indicated by the body mass index. The apnea-hypopnea index determined from the overnight polysomnography on visit 2 ranged from moderate to severe. The level of oxygen desaturation achieved during apnea/hypopnea events was moderate in most of the participants. Systolic and diastolic blood pressure measurements were within normal limits, and the Epworth Sleepiness Scale indicated a history of mild sleepiness (Table 1).

Table 1.

Baseline patient characteristics

Variable	Mean ± SD
Age, yr	29.5 ± 7.0
Height, cm	176.3 ± 8.9
Weight, kg	83.5 ± 10.5
Body mass index, kg/m2	26.9 ± 2.9
Systolic blood pressure, mmHg	120.5 ± 13.2
Diastolic blood pressure, mmHg	72.1 ± 12.0
Epworth Sleepiness Scale	9.0 ± 4.8
Apnea-hypopnea index, events/h	43.4 ± 16.8
Lowest oxygen saturation during apnea, %	86.1 ± 4.2
Race	1 Asian, 1 Hispanic, 3 Caucasian, and 6 African American

n = 11 participants.

Total non-rapid eye movement sleep time was similar across all sleep sessions (P = 0.41) (Table 2). Likewise, the percentage of time spent in N1, N2, or N3 was similar across all sleep sessions (N1: P = 0.12, N2: P = 0.82, N3: P = 0.68) (Table 2). The apnea index was greatest in the morning and afternoon sessions, and these values were significantly greater than the index measured during the 10 PM (2) session (P = 0.002 for the morning, P = 0.01 for the afternoon) (Table 2). No significant difference in the hypopnea index was evident between the sessions (P = 0.23) (Table 2). The average duration of breathing events was longer in the morning compared with the 10 PM (1) session (P = 0.05) (Table 2). The difference between the morning and 10 PM (2) session also approached significance (P = 0.06) (Table 2). Event duration during the afternoon session was longer than the evening sessions [P = 0.05 for 10 PM (1) and P = 0.05 for 10 PM (2)] (Table 2). No difference in event duration was evident between the two evening sessions (P = 0.99) (Table 2). The longest event duration also occurred in the morning compared with the evening sessions [P = 0.01 for 10 PM (1) and P = 0.002 for 10 PM (2)]. The longest event duration was also greater in the afternoon compared with the 10 PM (2) session (P = 0.03). No difference in the longest event duration existed between the two evening sessions (P = 0.46). The average level of oxygen desaturation was similar across all sessions (P = 0.16). Likewise, the lowest level of oxygen desaturation was similar across sessions (P = 0.57).

Table 2.

Sleep measures for each sleep session

Variable	10 PM (1)	6 AM	2 PM	10 PM (2)
% in N1	32.8 ± 22.8	23.8 ± 13.7	28.1 ± 15.6	29.5 ± 16.0
% in N2	58.3 ± 21.1	56.7 ± 19.0	60.0 ± 14.4	58.4 ± 19.9
% in N3	6.9 ± 9.7	7.2 ± 14.4	3.8 ± 6.4	9.2 ± 12.4
% in REM	2.0 ± 3.8	12.3 ± 9.2‡*	8.1 ± 9.5‡*	3.0 ± 5.6
Total NREM sleep time, min	137.6 ± 8.2	147.6 ± 6.6	143.3 ± 6.8	150.4 ± 7.0
Apnea index, events/h	20.6 ± 22.0	28.9 ± 22.2*	24.7 ± 20.8*	9.1 ± 7.5
Hypopnea index, events/h	28.4 ± 18.3	24.8 ± 14.7	24.6 ± 17.3	35.2 ± 22.3
Average event duration, s	20.4 ± 5.6	23.3 ± 6.8‡	23.4 ± 6.5‡*	20.5 ± 6.8
Longest event duration, s	32.6 ± 13.9	44.3 ± 19.0‡*	39.5 ± 16.0*	29.3 ± 10.5
Average oxygen desaturation, %	92.8 ± 2.2	93.3 ± 1.9	93.5 ± 1.8	93.1 ± 1.7
Lowest oxygen desaturation, %	87.9 ± 3.7	87.4 ± 8.4	89.3 ± 3.5	88.3 ± 4.4

Values are means ± SD; n = 11 participants. NREM, non-rapid eye movement sleep; REM, rapid eye movement sleep. A 1-way analysis of variance in conjunction with Fisher’s least significant difference post hoc test was used to make comparisons between sleep sessions.

^‡Significantly different from 10 PM (1);

^*significantly different from 10 PM (2).

Time Constant and Time Delay

The time delay was similar across all sessions [10 PM (1) = 13.08 ± 2.75 s, 6 AM = 12.93 ± 2.04 s, 2 PM = 12.88 ± 1.69 s, 10 PM (2) = 13.72 ± 1.47 s, P = 0.33]. The time constant (τ) during 10 PM (2) (279.0 ± 26.91 s) was statistically greater compared with 10 PM (1) (241.4 ± 45.49 s, P = 0.004) and 6 AM (253.15 ± 39.42 s, P = 0.04), whereas the comparison to 2 PM (254.73 ± 32.94 s) approached significance (P = 0.06).

Loop Gain at Natural Cycling Frequency and at Frequency of 1 cycle/min

Figure 1, A and B, show the presence of rhythmicity (P < 0.0001) in loop gain at the natural cycling frequency (LG_n) and at a frequency of 1 cycle/min (LG₁), with peak values evident in the morning compared with the other sessions. Similarly, average LG_n and LG₁ values were significantly greater in the morning (i.e., 6 AM) compared with measures obtained during both evening sessions [i.e., 10 PM (1) and 10 PM (2)] [LG_n: P < 0.001, LG₁: P = 0.01 for 10 PM (1) and P < 0.001 for 10 PM (2); Fig. 2]. Although LG_n and LG₁ declined in the afternoon compared with the morning (see Fig. 1, A and B), average values for LG_n and LG₁ remained elevated compared with the evening sessions for all but one comparison, which approached statistical significance [LG_n: P = 0.004 for 10 PM (1) and P = 0.01 for 10 PM (2); LG₁: P = 0.06 for 10 PM (1) and P = 0.01 for 10 PM (2); Fig. 2]. In contrast, LG_n and LG₁ were similar during 10 PM (1) and 10 PM (2) (LG_n: P = 0.95; LG₁: P = 0.27; Fig. 2). As stated above (see Loop gain and arousal threshold analysis), the loop gain data for non-rapid eye movement sleep were combined because the findings were independent of sleep stage [e.g., LG₁ N1 10 PM (1) vs. 6 AM: 0.50 ± 0.1 vs. 0.70 ± 0.30, P = 0.04; LG₁ N2 10 PM (1) vs. 6 AM: 0.64 ± 0.24 vs. 0.73 ± 0.26, P = 0.01].

Fig. 1.

Scatterplots showing loop gain at the natural cycling frequency (A), loop gain at 1 cycle/min (B), and the arousal threshold (C) during non-rapid eye movement (NREM) sleep in the evening [10 PM (1)], morning (6 AM), afternoon (2 PM), and subsequent evening [10 PM (2)]. The data shown are an average of loop gain and the arousal threshold calculated at 30-min intervals during each of the sleep sessions. Variability among the participants is shown in Figs. 2 and 4. The sine wave fit to each data set is also shown. Note the presence of a circadian rhythmicity in both loop gain and the arousal threshold. n = 11 participants. The best fit sine wave was determined with the least squares method.

Fig. 2.

Scatterplots showing the natural loop gain (A) and the loop gain at 1 cycle/min (B) during non-rapid eye movement (NREM) sleep in the evening [10 PM (1)], morning (6 AM), afternoon (2 PM), and subsequent evening [10 PM (2)]. Each symbol and corresponding color represent a participant, which is constant across figures. In addition, the horizontal blue line indicates the mean value. Note that both the natural loop gain and the loop gain at 1 cycle/min were greatest in the morning (6 AM) and were significantly greater than values measured during both evening sessions [10 PM (1) and 10 PM (2)]. Likewise, although measures of loop gain declined from the morning to the afternoon, values during the afternoon remained elevated compared with both evening sessions. Finally, loop gain values during both evening sessions were similar. ‡Significantly different from 10 PM (1). *Significantly different from 10 PM (2). n = 11 participants. A 1-way analysis of variance in conjunction with Fisher’s least significant difference post hoc test was used to make comparisons between sleep sessions.

Figure 3, A and B, show that LG_n and LG₁ were significantly correlated to the hypocapnic ventilatory response, a measure of chemoreflex sensitivity (LG_n: r = 0.77, P = 0.03; LG₁: r = 0.72, P = 0.045). Similarly, Fig. 3, C and D, show that LG₁ was significantly correlated to measures of upper airway collapsibility (LG₁ for 3-breath method: r = 0.69, P = 0.02; LG₁ for 5-min method: r = 0.77, P = 0.02).

Fig. 3.

A and B: scatterplots showing the relationship between loop gain at the natural cycling frequency and chemoreflex sensitivity (i.e., the hypocapnic ventilatory response, randomly measured during visit 5 or 6) (A) and loop gain at 1 cycle/min and chemoreflex sensitivity (B) during non-rapid eye movement (NREM) sleep in the evening [10 PM (1)], morning (6 AM), and afternoon (2 PM). Each symbol and corresponding color represent a participant, which is constant across figures. Each data point represents the mean value ± standard deviation obtained from measures in the evening, morning, and afternoon. Note that in all cases loop gain was correlated to measures of chemoreflex sensitivity. C and D: scatterplots showing the relationship between loop gain at 1 cycle/min and a 3-breath measure of the critical closing pressure (i.e., randomly measured during visit 5 or 6) (C) and loop gain at 1 cycle/min and a 5-min measure of the critical closing pressure (D). Each symbol and corresponding color represent a participant. Each data point represents the mean value ± standard deviation obtained from measures in the evening, morning, and afternoon. Note that in all cases loop gain was correlated to measures of the critical closing pressure. n = 11 participants for the 3-breath measure of the critical closing pressure and n = 8 participants for the 5-min measure of the critical closing pressure. A Pearson correlation was used to examine the relationships.

Arousal Threshold and Ventilatory Response to Arousal

Figure 1C shows the presence of a rhythmicity (P < 0.001) in the arousal threshold, with peak values evident in the morning compared with the other sessions. Similarly, the average arousal threshold was significantly higher in the morning compared with both evening sessions [P = 0.02 for 10 PM (1) and P = 0.001 for 10 PM (2); Fig. 4A]. Overall declines in the arousal threshold were evident during the afternoon session, as shown in Fig. 1C. The average values during the afternoon session were not different from values measured during the morning session (P = 0.13; Fig. 4A) or initial evening session [P = 0.41 for 10 PM (1)]. In contrast, the difference in the average values measured in the afternoon compared with the final evening session achieved statistical significance [P = 0.05 for 10 PM (2); Fig. 4A]. The arousal thresholds during both evening sessions were similar (P = 0.24; Fig. 4A). The arousal threshold data for non-rapid eye movement sleep were combined because the findings were independent of sleep stage [e.g., N1 10 PM (1) vs. 6 AM: 1.11 ± 0.11 vs. 1.45 ± 0.63, P = 0.05; N2 10 PM (1) vs. 6 AM: 1.32 ± 0.38 vs. 1.54 ± 0.49, P = 0.03]. The arousal threshold was not correlated to the critical closing pressure with the three-breath method (r = 0.47, P = 0.15) (14). However, the arousal threshold was correlated to the critical closing pressure (r = 0.74, P = 0.03) with the 5-min method (14), as determined by a Spearman rank correlation because of the presence of an obvious single outlier.

Fig. 4.

Scatterplots showing the arousal threshold (A) and the ventilatory response to arousal (B) during non-rapid eye movement (NREM) sleep in the evening [10 PM (1)], morning (6 AM), afternoon (2 PM), and subsequent evening [10 PM (2)]. Each symbol and corresponding color represent a participant, which is constant across figures. In addition, the horizontal blue line indicates the mean value. Note that both the arousal threshold and ventilatory response to arousal were greatest in the morning (6 AM) and were significantly greater than values measured during both evening sessions [10 PM (1) and 10 PM (2)]. In addition, there was a decline in the arousal threshold from the morning to the afternoon, so that no difference existed between the afternoon and initial evening or morning session. In contrast, the ventilatory response to arousal remained elevated in the afternoon compared with both evening sessions. Finally, the arousal threshold and ventilatory response to arousal during both evening sessions were similar. ‡Significantly different from 10 PM (1). *Significantly different from 10 PM (2). n = 11 participants. A 1-way analysis of variance in conjunction with Fisher’s least significant difference post hoc test was used to make comparisons between sleep sessions.

The ventilatory response to arousal was greater in the morning compared with the 10 PM (1) (P = 0.002) and 10 PM (2) (P = 0.002) sessions (Fig. 4B). Likewise, the response in the afternoon remained elevated compared with 10 PM (1) (P < 0.001) and 10 PM (2) (P = 0.003) (Fig. 4B). The ventilatory responses during both evening sessions were similar (P = 0.98; Fig. 4B).

DISCUSSION

Loop Gain

Our results showed that a rhythmicity in loop gain was evident during non-rapid eye movement sleep in individuals with obstructive sleep apnea. Moreover, the peak response (i.e., the acrophase) was evident in the morning compared with the evening and the afternoon. The magnitude of loop gain is in part influenced by controller gain, which mirrors chemoreflex sensitivity. We and others have shown that acute or repeated daily exposure to intermittent hypoxia increases controller gain (i.e., chemoreflex sensitivity) in both healthy humans (22, 24, 34, 46, 52) and humans with obstructive sleep apnea (16, 20, 48, 56) (see 25, 26 for reviews). More recently, we showed directly that exposure to intermittent hypoxia leads to increases in loop gain in participants with obstructive sleep apnea (3). Thus, it is possible that exposure to intermittent hypoxia during the initial sleep session [10 PM (1)] could be responsible for the increase in loop gain that was evident in the morning session. However, this explanation does not fully address our findings, since a steady increase in loop gain would be evident in the subsequent sleep sessions if progressive exposure to intermittent hypoxia was the principal modifying stimulus.

It could also be argued that other factors with a circadian rhythmicity could be responsible for our measures of loop gain. A few studies in awake healthy humans have shown a circadian rhythmicity in controller gain along with a rhythmicity in core body temperature (44, 45) and cortisol (44). Both stimuli can modify chemoreflex sensitivity (5, 38, 44). Nonetheless, it is unlikely that modulation of core body temperature is responsible for circadian modulation in controller gain. Spengler and colleagues (44) reported that the nadir of controller gain occurred 6–8 h before the minimum in core body temperature. Our data are similar to this finding, since the nadir in core body temperature during the constant-routine protocol occurred in the morning compared with the evening (see Fig. 2 in Ref. 14 and Fig. 3 in Ref. 15). In other words, the greatest amplitude in loop gain was evident when temperature reached its lowest point in the morning. In contrast, Stephenson and colleagues (45) stated that the nadir of body temperature and controller gain occurred within a similar time frame. However, these authors discounted the possibility that temperature was directly responsible for the rhythmicity of controller gain, because the circadian peak to valley rise in temperature was much less than 1°C (45), which is required before chemoreflex sensitivity is modified (5, 38). In addition, an increase in temperature of 1°C primarily causes an increase in chemoreflex sensitivity (5, 38), whereas the circadian rhythm observed in controller gain was primarily associated with modifications in chemoreflex threshold and not sensitivity (45). Finally, although the circadian minimum of cortisol has been reported to occur close to the minimum of chemoreflex sensitivity, its direct influence was discounted because no systematic correlation between cortisol and chemoreflex sensitivity was evident within participants (44).

On the basis of the above findings, we believe that the observed modulation in loop gain could be the result of an inherent circadian rhythm. Although the mechanism is unknown, the modulation in loop gain is strongly influenced by variations in controller gain (i.e., a component of loop gain). We previously showed that controller gain, which was reflected in the measurement of the hypocapnic ventilatory response, was greater in the morning compared with the evening and afternoon (14). This measurement was obtained while individuals were treated with continuous positive airway pressure, which mitigated the influence of intermittent hypoxia (14). We also revealed in the present study that loop gain, determined by a model-based approach, was strongly correlated to the previously published measures of the hypocapnic ventilatory response (14). Thus, modulation of controller gain strongly contributed to the observed rhythmicity in loop gain.

An increase in controller gain/loop gain during non-rapid eye movement sleep could lead to an increased ventilatory response to chemical stimuli (i.e., hypoxia and hypercapnia) and arousal at the termination of an apneic event (10, 28, 29). In support of this suggestion, our results showed that elevations in loop gain in the morning and afternoon were coupled to an increased ventilatory response to arousal at the same time points in the constant-routine protocol. It is well established that increases in ventilation that exceed metabolic output at the termination of apneic events will induce reductions in carbon dioxide below the apneic threshold, resulting in a subsequent apnea upon the resumption of sleep (10, 28, 29). Indeed, we have found that loop gain is correlated to indexes of upper airway collapsibility including the degree of flow limitation, the apnea-hypopnea index, and the ratio of hypopnea to apnea-hypopnea index (37). This potential cause-and-effect relationship is supported by results obtained in the present investigation. We showed that loop gain measured at three different time points was correlated to previously published direct measures of upper airway critical closing pressure (i.e., a measure of upper airway collapsibility) (15) made at the same time of day. The correlation between loop gain and the critical closing pressure was positive, in contrast to a previous investigation that indicated that loop gain is negatively correlated to airway collapsibility (11). It is difficult to directly compare our findings to the published investigation. Participants in the published study were treated with continuous positive airway pressure for >3 mo (11), whereas we studied untreated participants. Moreover, participants had a higher body mass and lower apnea-hypopnea index (11) compared with our participants. Furthermore, sleep apnea was defined as an apnea-hypopnea index > 10 events/h (11), whereas sleep apnea was defined as being > 5 events/h in the present investigation. Thus, whether the relationship between loop gain and airway collapsibility is altered by treatment, modified by anthropometric characteristics, or dependent on the characterization of sleep apnea requires further investigation.

The relationship between loop gain and upper airway collapsibility appears to manifest in the measures of apnea frequency and duration in the present investigation. Specifically, the number of apneic events and the duration of breathing events were greater in the morning or afternoon compared with one or both evening sessions. The role that loop gain might have in prolonging apnea duration could be coupled to the degree to which carbon dioxide is reduced below the apneic threshold after arousal from sleep (53). In other words, the greater the reduction the longer the time duration required for carbon dioxide to gradually increase and exceed the apneic threshold.

Arousal Threshold

Arousals from sleep occur when the neuromechanical drive associated with ventilatory effort reaches a threshold, deemed the arousal threshold. With the model-based approach employed in our study, the arousal threshold is determined from the ventilatory drive that immediately precedes arousal from sleep (50). This model-based approach reflects the assessment of the arousal threshold with esophageal pressure measurements, which is considered the gold standard (39). Conceptually, when ventilation decreases during a respiratory event, carbon dioxide and the ventilatory drive increase, often causing respiratory effort-related arousals. Individuals with a higher arousal threshold can tolerate greater increases in ventilatory drive relative to the drive present at eupneic breathing.

Similar to our findings for loop gain, a rhythmicity in the arousal threshold was evident, with peak values occurring in the morning compared with the evening sessions. Our results are similar to published findings of increases in the arousal threshold in the morning compared with the evening (6, 8, 33, 42) in individuals with obstructive sleep apnea. However, the evening to morning increase in the arousal threshold was attributed primarily to a progressive blunting in the arousal response to neural stimuli produced during the respiratory event and a diminished contribution of upper airway sensory receptors to apnea termination toward the end of the night (6, 8, 33, 42). Investigators proposed that inflammation, edema, and neural damage in response to snoring-related vibrations could be responsible for the progressive dampening of excitatory inputs from upper airway sensory receptors from the beginning to the end of the night (6, 8, 33, 42). Likewise, increased exposure to intermittent hypoxia from the beginning to the end of the night could produce a similar result (3). However, as outlined above (see Loop Gain), if these mechanisms were solely responsible for our findings we would anticipate a progressive increase in the arousal threshold from the initial to the final evening session, which was not the case.

It might be debated that an increase in the arousal threshold from the evening to the morning was a consequence of the lack of sleep (i.e., sleep deprivation) between the evening and the morning sessions. More specifically, an increase in sleep pressure could be responsible for the elevated arousal threshold in the morning. We believe this is unlikely for a few reasons. If the arousal threshold is reduced in the morning, as suggested by indirect findings in healthy individuals (40) or the interpretation of non-statistically significant trends in individuals with obstructive sleep apnea (41), then it would be surprising that 3 h of sleep coupled to 5 h of wakefulness would lead to a significant increase in the arousal threshold above evening measures. More importantly, if sleep pressure was the primary driving force for the changes we observed, one would anticipate that the arousal threshold in the 10 PM (2) session should be significantly lower than measurements obtained in the 10 PM (1) session. This suggestion is based on the rationale that 14–15 h of wakefulness preceded the 10 PM (1) session, whereas 10 h of wakefulness coupled to 6 h of sleep preceded the 10 PM (2) session. However, this was not the case, since the arousal threshold was similar in the 10 PM (2) and 10 PM (1) sessions. One might also contend that the arousal threshold would be highest in the 10 PM (1) session compared with the remaining sessions if sleep drive was the primary influence on the arousal threshold.

Rather, we speculate that rhythmicity in the arousal threshold may be influenced in part by circadian variations in monoamine neurotransmitters and orexin released by several sites in the central nervous system (4, 23, 31, 32, 54). For example, our work in tryptophan hydroxylase II-knockout mice has shown that the absence of serotonin leads to an increase in the duration of apneic events and to fewer events terminated by arousals (27). Thus, circadian variations in serotonin, or other neuromodulators involved in the determination of arousal states, might be responsible for the rhythmicity in the arousal threshold that we observed. Circadian modulation in serotonin and its receptor subtypes has been observed in the pineal, hippocampus, hypothalamus, striatum, cerebellum, locus coeruleus, and dorsal raphe nucleus of the brain (1, 30, 47, 51). Jointly, the results indicate that serotonin levels tend to peak in the late afternoon and early evening before reaching a nadir later in the evening. Reductions in serotonin levels are typically associated with increases in arousal threshold (2, 7, 43). Nonetheless, this hypothesis remains to be tested in humans.

Independent of the mechanism, we and others have reported that the arousal threshold is positively correlated to airway collapsibility, as indicated by indirect measures (i.e., degree of flow limitation, ratio of hypopnea to apnea-hypopnea index, apnea-hypopnea index, apnea duration) (12, 37). We have also presented some preliminary evidence in a previous study (37), along with our present findings, that the arousal threshold is positively correlated to direct measures of airway collapsibility (i.e., the critical closing pressure). Our results from the present investigation also support previous findings, since the sessions associated with increases in loop gain and the arousal threshold were coupled with increases in apnea frequency and duration.

Summary and Translational Significance

Our results showed that a circadian rhythmicity in loop gain and the arousal threshold is evident in individuals with obstructive sleep apnea. We have also provided evidence that the rhythmicity in loop gain is linked directly to a circadian rhythmicity in controller gain. Likewise, we provide additional evidence that increases in loop gain are linked to increased upper airway collapsibility, likely via the modulation of partial pressure of carbon dioxide. Finally, we have shown that noninvasive measures of loop gain and the arousal threshold may serve as a surrogate for more invasive measures (14, 53, 56) when examining circadian rhythmicity in respiratory control.

Our findings suggest that the inherent rhythmicity in loop gain and the arousal threshold could account for findings reported of differences in apnea frequency and/or duration at different times of the night for a given stage of sleep. From a translational point of view, attempts to mitigate the severity of sleep apnea by targeting loop gain or the arousal threshold should consider the modulation of these variables when determining the effective dose of a selected treatment option.

GRANTS

This work was supported by awards (I01CX000125 and 15SRCS003; J.H.M.) from the Office of Research and Development, Veterans Health Administration, Department of Veterans Affairs and awards (R56 HL-142757, R01 HL-085537) from the National Heart, Lung, and Blood Institute.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.H.M. conceived and designed research; M.E., D.S., B.I., M.S.B., H.-S.L., and J.H.M. performed experiments; S.P., M.E., D.S., B.I., G.S.P., and J.H.M. analyzed data; S.P., G.S.P., and J.H.M. interpreted results of experiments; S.P. and J.H.M. prepared figures; S.P. and J.H.M. drafted manuscript; S.P., M.E., D.S., B.I., G.S.P., M.S.B., H.-S.L., and J.H.M. edited and revised manuscript; S.P., M.E., D.S., B.I., G.S.P., M.S.B., H.-S.L., and J.H.M. approved final version of manuscript.