Continuous Positive Airway Pressure Increases Pulsatile Growth Hormone Secretion and Circulating Insulin-like Growth Factor-1 in a Time-Dependent Manner in Men With Obstructive Sleep Apnea: A Randomized Sham-Controlled Study

Abstract

Study Objectives:

To assess the time-dependent effect of continuous positive airway pressure (CPAP), on insulin-like growth factor-1 (IGF-1), IGF binding proteins (IGFBPs) and pulsatile growth hormone (GH) secretion.

Design:

A randomized, double-blind, sham-controlled, parallel group study.

Participants:

Sixty-five middle-aged men with moderate to severe obstructive sleep apnea.

Intervention:

Active (n = 34) or sham (n = 31) CPAP for 12 weeks, followed by 12 weeks of active CPAP (n = 65).

Measurements and Results:

Fasting morning IGF-1, IGFBP-3, and IGFBP-1 blood levels at 0, 6, 12, and 24 weeks. Overnight GH secretion was calculated by mathematical deconvolution of serial GH measurements from serum samples collected every 10 min (22:00-06:00) during simultaneous polysomnography in a subset of 18 men (active n = 11, sham n = 7) at week 12. Active, compared with sham, CPAP increased IGF-1 at 12 weeks (P = 0.006), but not at 6 weeks (P = 0.44). Changes in IGFBP-3 and IGFBP-1 were not different between groups at 6 or 12 weeks (all P ≥ 0.15). At week 24, there was a further increase in IGF-1 and a decrease in IGFBP-1 in the pooled group (P = 0.0001 and 0.046, respectively). In the subset, total (P = 0.001) and pulsatile (P = 0.002) GH secretion, mean GH concentration (P = 0.002), mass of GH secreted per pulse (P = 0.01) and pulse frequency (P = 0.04) were all higher after 12 weeks of CPAP compared with sham. Basal secretion, interpulse regularity, and GH regularity were not different between groups (all P > 0.11).

Conclusions:

Twelve weeks, but not 6 weeks, of CPAP increases IGF-1, with a further increase after 24 weeks. Total and pulsatile GH secretion, secretory burst mass and pulse frequency are also increased by 12 weeks. CPAP improves specific elements of the GH/IGF-1 axis in a time-dependent manner.

Clinical Trials Registration:

Australia New Zealand Clinical Trials Network, www.anzctr.org.au, number ACTRN12608000301369.

Citation:

Hoyos CM; Killick R; Keenan DM; Baxter RC; Veldhuis JD; Liu PY. Continuous positive airway pressure increases pulsatile growth hormone secretion and circulating insulin-like growth factor-1 in a time-dependent manner in men with obstructive sleep apnea: a randomized sham-controlled study. SLEEP 2014;37(4):733-741.

INTRODUCTION

Obstructive sleep apnea (OSA) is a common condition characterized by repetitive collapse of the upper airway during sleep, leading to transient episodes of hypoxemia and arousals that disrupt sleep architecture and reduce slow wave sleep (SWS).¹ Because growth hormone (GH) is secreted in a highly organized fashion predominately during SWS,^2–4 reductions in GH and abnormalities in other components of the growth hormone/insulin-like growth factor-1 (GH/IGF-1) axis would be anticipated to accompany OSA, and might even be reversed with continuous positive airway pressure (CPAP) therapy.⁵ These hormonal abnormalities could underpin the metabolic dysfunction that characterizes OSA because GH and circula-tory insulin-like growth factor 1 (IGF-1) have major effects on carbohydrate, fat, and lipid metabolism.^6–8 Furthermore, GH deficiency is associated with many health consequences including altered body composition, worsened cardiovascular health, and poor quality of life; low IGF-1 also predicts cardiovascular morbidity and mortality.⁹ Understanding the effects of OSA, CPAP, and the underlying sleep architectural and/ or hypoxemic mechanisms, on the entire GH/IGF-1 axis is therefore important.

The GH/IGF-1 axis consists of pulsatile GH secretion by the anterior pituitary, which signals the production of circula-tory IGF-1 by the liver; circulatory liver-derived IGF-1 is then sequestered and transported by the two main binding proteins, insulin-like growth factor binding protein-3 (IGFBP-3) and insulin-like growth factor binding protein-1 (IGFBP-1).^6–8,10,11 For a number of reasons, a comprehensive understanding of this entire sequence is required to fully elucidate GH/IGF-1/ IGFBP action. First, the secretory pattern of GH is critical for its action. Larger body growth and greater tissue IGF-1 expression in bone occurs when the same total dose of GH is administered in a pulsatile rather than a continuous fashion.^12,13 Second, although IGF-1 is a surrogate of GH action, some GH effects occur independently of circulatory IGF-1 due to direct local stimulation of IGF-1 in specific tissues.⁸ Furthermore, in muscle, these GH effects occur without even local tissue generation of IGF-1,¹⁴ and this could plausibly be true in other tissues. Third, binding proteins sequester and transport circula-tory IGF-1 thereby altering IGF-1 action, but also have direct effects on growth, independent of IGF-1.^10,11

The effect of CPAP therapy on many of the hormonal deficiencies associated with OSA is also inadequately studied. Such information would be useful to guide the need for additional hormonal supplementation, particularly because hormonal therapy may inadvertently exacerbate sleep disordered breathing when administered supraphysiologically, without individualized dose titration.^15–18 Uncontrolled studies show significant increases in blood IGF-1 with 1 to 8 mo of CPAP.^19–21 However, the only randomized sham-controlled study did not show any between-group differences after 1 mo, despite IGF-1 significantly increasing in both groups.²² Only two small studies have assessed nocturnal GH concentrations.^23,24 Both studies used only a single night of CPAP, and neither study was controlled. The first study (n = 8) showed that CPAP significantly increased GH plasma levels and secretion rates as determined by deconvolution analysis of serial GH concentrations measured every 10 min for 8 h (22:00 to 06:00) overnight.²³ The deconvolution method used did not allow for determination of important pulse characteristics.²⁵ The second study (n = 6) measured GH every hour for 6 h overnight and showed that the GH area under the curve increased with CPAP.²⁴ Hourly assessment of GH does not allow for mathematical deconvolution of GH secretion rates.²⁵

To date, there has not been a randomized sham-controlled study of any duration examining the effect of CPAP therapy on pulsatile GH secretion. A single randomized sham-controlled study has examined the effect of 1 mo of CPAP on IGF-1, without further delineation of time course. We therefore examined the effect of 12 weeks of CPAP on IGF-1, IGFBP-3 and IGFBP-1 in a randomized sham-controlled study. Additionally, we measured overnight GH secretion and pulsatility with frequent overnight blood sampling using gold-standard deconvolution analysis in a subset of participants. We also assessed by correlation whether these changes were driven by changes in SWS, or by hypoxemia, because both typify OSA. To further delineate time effects, all participants received an additional 12 weeks of active CPAP irrespective of initial treatment allocation. The effects of CPAP on visceral abdominal fat, insulin sensitivity, and metabolic syndrome in this study have recently been published.^26,27

METHODS

Subjects

Participants were recruited from tertiary referral sleep clinics at Royal Prince Alfred Hospital and the Woolcock Institute of Medical Research, Sydney, Australia. Eligibility criteria have been previously described in detail.²⁷ In brief, eligible participants were men with moderate-severe OSA, defined as an apnea-hypopnea index (AHI) ≥ 20 events/h and an oxygen desaturation index 3% (ODI) ≥ 15 events/h measured by in-laboratory polysomnography. Exclusion criteria included type II diabetes, severe OSA that required immediate treatment, past CPAP use, uncontrolled medical conditions, and previous clinical trial participation within 30 days.

The study was approved by the Sydney South West Area Health Service Human Research and Ethics Committee (RPAH Zone) and all participants provided written informed consent prior to commencing the study. The study is registered with the Australia New Zealand Clinical Trials Network, www.anzctr.org.au, number ACTRN12608000301369.

Study Design

This was a randomized, double-blind, sham-controlled, parallel group study. Participants were randomized to receive 12 weeks of treatment with either real or sham CPAP. Data collection visits occurred before, midway at 6 weeks and after treatment. At the end of the 12-week blinded period, all participants received an additional 12 weeks of treatment with open-label real CPAP. Assessments were recollected at 24 weeks.

Participants were assigned to real or sham CPAP in a 1:1 ratio using a computer-generated randomization list with a block size of four. At baseline, each participant was assigned a unique number in sequential, ascending, chronological order that corresponded to the treatment allocation. CPAP and sham machines were prepared by a person separate from the study investigators and not involved in participant assessments. The study investigators were blinded to treatment allocation for the duration of the study. The intensive overnight sampling substudy was initiated after specialized facilities became available, and after the start of the main study. All participants from the 22^nd randomized participant were consecutively given the opportunity to also participate in this substudy.

CPAP Machines and Titration

The real and sham CPAP machines (Remstar Auto, Philips Respironics, Andova, MA, USA) were identical in appearance to each other and have been used previously at our center²⁸ and by others.²⁹ The sham device delivered airflow with minimal pressure (0.5 cm H₂0). Prior to randomization, every subject was fitted with a mask and received a standard CPAP education program operating in our clinic. Once randomized, each participant underwent a multiple night home imitation (sham group) or auto-titrating (real group) CPAP pressure determination study. Compliance data were recorded by an internal clock within all real and sham CPAP machines and were downloaded after the home titration and at each visit.

Blood Collection and Polysomnography

Fasting venous blood was collected in the early morning at each visit in all participants for measurements of IGF-1, IGFBP-3, and IGFBP-1.

A subset of patients (n = 18) underwent an 8-h (22:00-06:00) intensive (every 10 min) blood sampling session via intravenous cannula while simultaneously undergoing full polysomnography (PSG, Sandman Elite v9.2, Tyco Healthcare, Denver, CO) using either real or sham CPAP. Each patient slept in a soundproofed room separate to a registered nurse who collected blood through a purpose- built access panel designed to maintain soundproofing. Sleep staging and respiratory events were scored by personnel blinded to treatment allocation using published criteria.³⁰ Apneas were classified as a complete cessation of airflow for at least 10 sec and hypopneas were defined as either a minimum 10-sec period of greater than 50% airflow reduction or a lesser airflow reduction with associated > 3% oxygen desaturation or arousal.³¹ AHI was calculated as the total number of apneas and hypopneas per hour of sleep. ODI was defined as the number of times oxygen desaturated by more than 3% per hour of sleep, as determined by pulse oximetry.

Assays

IGF-1 concentrations were assayed by standard platform assays. IGFBP-1 and IGFBP-3 were measured by radioimmunoassays using polyclonal antibodies by previously validated methods.^32,33 The limits of detection for IGFBP-1 and IGFBP-3 were 2.5 μg/L and 0.2 mg/L, respectively. The within-assay imprecision and between-assay imprecision were 4.5-6.2% and 12-14.5%, respectively, across the analytical range for IGFBP-1. The within-assay imprecision and between-assay imprecision were 6.4-9.8% and 12-14%, respectively, across the analytical range for IGFBP-3.^32,33 Serum GH concentrations were assayed in each sample (49 samples per patient) by commercially available Delfia® assay (Perkin-Elmer Life Sciences, Rowville, Australia). The assay has a detection level of 0.04 mU/L with intra-assay and interassay coefficients of variation of less than 10%. All samples were stored at -80°C for subsequent batched analysis and all samples from an individual patient were run within a single assay.

Deconvolution Analysis

GH concentration times-series (all 8 h) were analyzed by an automated deconvolution method.^34,35 The program first detrends and normalizes GH concentrations to the unit interval (0, 1). Then, a smoothing process (a nonlinear adaptation of the heat-diffusion equation) sequentially produces potential pulse-sets, each including one fewer pulse. Next, secretion and elimination rates are calculated simultaneously for each and every pulse-set interval by a maximum likelihood equation (Matlab Direct-Search Algorithm). The parameters that can be estimated include basal secretion (β₀), two half-lives (α₁, α₂), random effects of burst mass (σ_A), an accumulation process and weak interpulse length dependency for secretory-burst mass (η₀, η₁), procedural and measurement error (σ_ε), a three-parameter secretory-burst waveform (β₁, β₂, β₃), and observed interpulse intervals defined by a two-parameter Weibull process. For the current analyses, elimination of half-lives of 3.5 min, contributing 37% of total decay, and 16 min were selected from previous data.³⁶ Candidate pulse-sets are then compared using the Akaike information criterion, in order to objectively determine the actual pulse-set model that best fits the observed data. Outcome variables (and units) are mean total concentration (concentration per unit), basal and pulsatile secretion rates (concentration per unit of time), total secretion (sum of basal and pulsatile), the mass secreted per burst (micrograms per liter), interpulse regularity (unitless γ of Weibull) and pulse frequency (number of bursts per unit time, λ of Weibull distribution) and pulse number.^35,37 Model precision is established by the analytical standard deviation (SD) of the likelihood function.^35,37 The coefficients of variance of parameter estimates are 2-14%.³⁷

Approximate Entropy

Approximate entropy (ApEn) was calculated for each individual subject's GH concentration time series. ApEn is the measure of regularity, with a greater ApEn indicating a more random pattern of hormone secretion.³⁸ Because ApEn is a family of statistics, ApEn (0.35, 1) specifically, was calculated.³⁹

Power Spectral Analysis

Power spectral analysis was performed on the central leads (C3) of the EEG to determine the non-rapid eye movement (NREM) mean delta power, NREM relative delta power density (% delta power/total power across all frequency bands), and total NREM delta power (mean delta power × number of 30-sec epochs ×2) after manual removal of electroencephalogram (EEG) artefact. EEG was transformed into European Data files and aligned with traditional visual scoring by one dedicated scorer. If noise artefact was present in over 25% of the channel, it was discarded from analysis (n = 1). Fast Fourier transform was performed on 5-sec epochs over the entire frequency bands, with the delta range (0.75-4.5 Hz) the primary focus for analysis.

Statistical Analysis

Data collected at week 0, 6, 12, and 24 were used to calculate the outcomes, which were differences from baseline. Mixed model analysis was used to determine between treatment group differences of these calculated differences from baseline at week 6 and 12 during the blinded period. A single sample Student t-test determined whether changes from baseline at week 24 were nonzero, in the pooled group which included all participants irrespective of initial treatment allocation. Associations between IGF-1 and ODI were explored using Pearson correlation coefficient. Data were log-transformed as required to ensure normality.

Further analyses explored the influence of treatment compliance, age, baseline severity (AHI), and obesity (body mass index [BMI] and visceral abdominal fat). These potential confounders were included as linear covariates. Additionally, covariates with predefined cut points were included in separate mixed models as a dichotomised factor (compliance 4 h/night; AHI 30 events/h; BMI 30 kg/m²; median age). The analyses were determined by mixed model analysis for repeated outcome measures (IGF-1 and IGFBPs) and linear regression for outcomes collected at only one time point (GH secretion). Pooled mean changes and the standard error of the mean changes from baseline at week 24 were determined regardless of initial treatment allocation.

Analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC). Data were considered significantly different at P < 0.05 (two-sided) and are presented as mean differences (95% confidence interval [CI]) and mean (standard deviation) as indicated.

RESULTS

Sixty-five men were randomized to receive either real (n = 34) or sham (n = 31) CPAP treatment. Fifty-two men, 29 in the CPAP group and 23 in the sham group, completed the 12-week treatment period. Baseline characteristics were comparable in the two treatment groups (Table 1). Eighteen men, 11 in the CPAP group and seven in the sham group, underwent intensive overnight blood sampling and simultaneous PSG. There were also no differences in baseline characteristics between the two treatment groups in this subset of participants (Table 2). Baseline characteristics were not statistically different between those included and not included in the substudy (data not shown). None of the participants had acromegaly. Average nightly use (mean ± standard deviation) during the 12-week treatment period was 3.6 ± 1.9 and 2.8 ± 2.1 h in the CPAP and sham treatment groups, respectively, for all participants and 4.3 ± 1.4 h in the CPAP and 3.1 ± 2.0 h in the sham groups in the 18 participants who underwent the overnight sampling. Nightly use was not significantly different between the sham and CPAP groups in either all subjects (P = 0.07) or in those who underwent the overnight sampling (P = 0.14).

Table 1.

Subject characteristics

Table 2.

Subject characteristics in subset (n = 18)

The overall change in IGF-1 from baseline during the first 12 weeks was significantly greater in the CPAP group compared to the sham group (mean difference 2.0 nmol/L, 95% CI 0.19 to 3.82, P = 0.03); Figure 1A. Significant between-group differences were observed at week 12 (P = 0.006), but not week 6 (P = 0.44). At week 24, IGF-1 had increased significantly from baseline in all participants (P = 0.0001; Figure 1A). There were no significant between-group differences in changes from baseline during the first 12 weeks in IGFBP-1 (Figure 1B) and IGFBP-3 (Figure 1C). At week 24, IGFBP-1 (Figure 1B) but not IGFBP-3 (Figure 1C) had decreased in all participants. There were no significant between-group differences in the changes from baseline in testosterone concentration during the first 12 weeks (P > 0.90) nor had testosterone changed from baseline at week 24 in all participants (P = 0.47). Adjustment for CPAP compliance, age, OSA severity, or obesity did not alter the statistical significance of any of these findings (data not shown). At baseline, IGF-1 was correlated with ODI (Figure 2A, r = -0.26, P = 0.038). Examination of other correlation coefficients confirms that this association was not due to age, BMI, or visceral abdominal fat. At week 24, the increase in IGF-1 was negatively correlated with the decrease in oxygen desaturation index (Figure 2B, r = -0.45, P = 0.002) but not SWS (r = -0.004, P = 0.98).

Figure 1.

The left-hand plot (line graph) shows the of mean change from baseline and standard error of the mean change at week 6 and week 12 in the real continuous positive airway pressure (CPAP; filled circles) and sham CPAP (open circles) groups for (A) insulin-like growth factor-1 (IGF-1), (B) insulin-like growth factor binding protein-1 (IGFBP-1), and (C) IGFBP-3. The P values denote the between-group difference at each time point as determined by mixed model. The right-hand plot (vertical bar graph) is the pooled mean change from baseline, and standard error of the change determined by Student t-test, at week 24 in all participants irrespective of initial treatment allocation.

Figure 2.

Scatter plot of (A) oxygen desaturation index (ODI, events/h) and insulin-like growth factor-1 (IGF-1; nmol/L) levels at baseline and (B) the change from baseline in ODI and the change from baseline in IGF-1 levels at the end of the study (week 24). The r and P values were determined by Pearson correlation.

Figure 3 shows illustrative overnight GH concentrations in a participant who had used CPAP (A) and different participant who had used sham (B) for 12 weeks. Twelve weeks of CPAP use, compared to sham use, resulted in significantly higher total and pulsatile GH secretion (Figures 4A and 4C), pulse frequency, GH mass per pulse, and mean GH concentration (Figures 4D-F). Basal GH secretion, pulse number, and GH regularity (ApEn) were not significantly different between the CPAP and sham groups (Figures 4B, 4G, 4H). Interpulse regularity (CPAP: mean (SD) 21.4 (64.9), sham: 11.4 (23.0), P = 0.63) was also not significantly different between groups. Adjustment for CPAP compliance, baseline age, OSA severity, or obesity did not alter these findings (data not shown). GH pulse secretion, total secretion, and mean total GH concentration were all negatively correlated with ODI (Figures 5A-C) and positively correlated with minimum oxygen saturation (all P < 0.02), but none were associated with SWS (% total sleep time; Figure 5E-G) or NREM delta power (Figure 5I-K). Pulse frequency was not correlated with ODI (Figure 5D), but was negatively correlated with SWS (% total sleep time; Figure 5H) and NREM delta power (Figure 5L).

Figure 3.

Plot of 10 minutely growth hormone (GH) blood concentrations (filled circles) taken overnight (22:00-06:00) at week 12 in two patients: (A) a 57-year-old man with a body mass index (BMI) of 33 kg/m² who had used sham continuous positive airway pressure (CPAP) for the 12 weeks prior and (B) a 61-year-old man with a BMI of 37 kg/m² who had used CPAP for the 12 weeks prior. The hypnogram displayed below the GH concentrations on both figures show the stages of sleep over the 8-h period. The sleep stages are defined in the displayed legend. REM, rapid eye movement.

Figure 4.

Vertical bar plot of mean and standard error of the mean at week 12 in the real continuous positive airway pressure (CPAP; shaded) and sham CPAP (unshaded) groups for (A) total growth hormone (GH) secretion, (B) basal GH secretion, (C) pulse GH secretion, (D) mean GH concentration, (E) mass per pulse, (F) pulse frequency, (G) pulse number, and (H) GH regularity. The P value denotes the between group difference as determined by Student t-test. CPAP n = 11, Sham n = 7.

Figure 5.

Plot of correlation at week 12 between (A-D) oxygen desaturation index (ODI; events/h), (E-H) slow wave sleep (% total sleep time) and (I-L) nonrapid eye movement (NREM) delta power (μV²) with total growth hormone (GH) secretion, pulse GH secretion, mean GH concentration, and pulse frequency. r and P values determined by Pearson correlation. Deconvolution variables are log-transformed.

At week 12, the AHI was significantly lower in the CPAP group compared to the sham group in the 18 participants who underwent overnight sampling as expected (Table 3). Measures of hypoxia (ODI, minimum saturation levels, and the percent of time spent below 90%) were also all significantly reduced with CPAP compared to sham (Table 3). Total sleep time or SWS were not different between groups in this subset (Table 3).

Table 3.

Sleep characteristics at 12 weeks and the between-group differences in subset (n = 18)

DISCUSSION

Here we present the first randomized sham-controlled trial showing the effect of CPAP treatment in men with OSA on pulsatile GH secretion, as measured by a recently validated deconvolution method. We have demonstrated significantly higher mean total GH concentrations, pulsatile and total GH secretion, GH secretory burst size, and pulse frequency after 12 weeks of CPAP compared to sham. Additionally, we have shown 12 weeks of CPAP, but not 6 weeks, to significantly increase total IGF-1 levels compared to sham in the longest randomized sham-controlled study to date. This occurred in the absence of between-group changes in the two main binding proteins, IGFBP-3 and IGFBP-1, suggesting that unbound IGF-1 concentrations would have paralleled those of total IGF-1. A further increase in total IGF-1, observed in the presence of a significant decrease in IGFBP-1, was also detected at 24 weeks. Both the further increase in IGF-1 and significant decrease in IGFBP-1 would be expected with persisting GH action.

These findings resolve earlier controversies. Prior uncontrolled trials showed that CPAP increases IGF-1 after 1-8 mo,^19–21 whereas the sole randomized sham-controlled trial did not show any between-group effect of 1 mo of CPAP.²² Of note, the latter study showed that IGF-1 concentrations increased significantly in both the sham and CPAP groups, bringing into doubt the preceding uncontrolled trials, which our current findings now resolve. The discrepancy between the controlled and uncontrolled studies of 1 mo duration^20,22 could not previously be explained by sample size because the randomized study, which did not report any significant between-group differences treated almost twice as many adults with CPAP. In conjunction with our findings, the most parsimonious conclusion is that the effect of CPAP on IGF-1 is time dependent, requiring more than 6 weeks, is unequivocally present at 12 weeks, and persists for at least 24 weeks. Metabolic or other effects that are IGF-1 dependent would therefore be expected to follow a similar time course.

Changes in GH secretion would be anticipated to underpin these time-dependent changes in IGF-1. Two small studies, albeit both uncontrolled, report increases in GH concentration or in GH secretion after a single “first-night” of CPAP usage.^23,24 If future controlled trials confirm that GH secretion is increased, compared with sham, within the first 6 weeks of CPAP therapy, then the lack of increase in circulating IGF-1 during this same period of time would most likely suggest liver resistance to GH (because the liver normally produces IGF-1 in response to GH, unless the liver is resistant to GH action), an ineffectual pattern of GH secretion, or both. Only one of these earlier studies²³ assessed pulsatile GH secretion because the other study sampled GH every hour,²⁴ which is too infrequent to assess the rapid rise and decay observed with pulsatile GH secretion. Although mathematical methods to calculate GH secretion and pulse characteristics have become more accurate,²⁵ this early study showed that CPAP increased GH secretion and the amplitude of the secretory pulses,²³ which is consistent with the findings of the current study. In contrast, pulse frequency did not change after 1 night of CPAP,²³ although we showed the pulse frequency to be significantly higher after 12 weeks of CPAP. Furthermore, the increase in pulsatile, and not basal GH secretion observed in the current study is of particular importance as pulsatile secretion contributes to the majority of daily GH release in healthy men.⁴⁰ Furthermore pulsatile, but not basal, GH secretion has been positively correlated with IGF-1 levels,⁴⁰ which is consistent with our data showing an increase in IGF-1 with CPAP.

The pathophysiological basis for this time dependency is unknown, but our correlational analyses suggest that OSA-related hypoxemia is responsible. We showed that baseline IGF-1 levels were negatively correlated with two measures of baseline hypoxemia. This finding replicates previous studies in humans^19,41 and rats.⁴² Additionally, we showed that the increase in IGF-1 was negatively correlated with the decrease in oxygen desaturation index, but not SWS. These findings suggest, but do not prove, that liver hypoxemia due to OSA is the underlying mechanism for time-dependent liver resistance to GH action. Although reduced GH action in OSA has previously been observed,⁴³ neither the mechanism nor the time course has previously been documented. Our data also showed a correlation between hypoxemia, not SWS, with total and pulsatile GH secretion, and with mean GH concentrations, in those with OSA. Because previous studies have shown a positive correlation between GH secretion and SWS in normal adults without OSA,^2–4 our analyses imply that the pathophysiological process of hypoxemia in OSA overrides normal physiological relationships. A novel finding was that SWS correlated negatively with GH pulse frequency. Previous studies in normal individuals without OSA have not reported the relationship between SWS and GH pulse frequency and hence whether disrupted versus consolidated sleep modifies this relationship is not known. Nevertheless, we conclude that in the presence of OSA, intermittent hypoxemia is largely responsible for most but not all of the defects in GH secretion.

One limitation is that we measured GH secretion only at the end of treatment in a subgroup of the participants. However, the random nature of treatment allocation should ensure that pulsatile GH secretion at baseline should be equivalent between the two groups. Furthermore, IGF-1, IGFBP-3, and IGFBP-1, which all are markers of GH action, were equivalent between groups both in the entire cohort as well as in the subset who underwent high fidelity blood sampling. If anything, IGF-1 was (non-significantly) higher in the seven men who subsequently received sham therapy, compared with the 11 men who received CPAP, which would bias against our findings of CPAP over sham effects.

In conclusion, GH secretion, specifically pulsatile secretion and the amount secreted with each pulse, was higher in those treated with CPAP for 12 weeks compared to sham CPAP. Furthermore CPAP treatment increased IGF-1 levels after 12 weeks but not 6 weeks, with a further increase at 24 weeks. These findings indicate that CPAP improves specific components of the GH/IGF-1 axis by improvement in hypoxemia in middle-aged men with OSA in a time-dependent manner. Future research should incorporate these findings when investigating any time-dependent improvements into longer term measures of cardiometabolic health, in the ever-increasing population of those with OSA. Future studies could relate these changes with metabolic or other specific improvements.

DISCLOSURE STATEMENT

This was not an industry supported study. Financial support by the National Health and Medical Research Council of Australia (NHMRC) through a project grant (512498), a Centre for Clinical Research Excellence in Interdisciplinary Sleep Health (571421) and fellowships to CH, RK, and PYL (512057, 633161 and 1025248, respectively). The project described was also supported by the National Center for Advancing Translational Sciences through UCLA CTSI Grant UL1TR000124. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NHMRC or National Institutes of Health. Sham continuous positive airway pressure devices were provided by Philips Respironics. The authors have indicated no financial conflicts of interest.