Abstract
Objective
Obstructive sleep apnea (OSA) is generally considered to lower serum testosterone concentration in men, although data supporting this as a direct effect are limited. The aim of this study was to determine the relationship between the presence and severity of OSA and testosterone in a community-based cohort of men aged over 40 years.
Design and methods
Anthropometry, polysomnography and biomedical information were collected from enrolled, consenting men from the prospective, longitudinal MAILES study cohort. Fasting morning blood samples (n = 1869) were drawn between 2010 and 2012 for measurement of testosterone using liquid chromatography mass spectrometry. Home polysomnography was completed in 861 men between 2010 and 2012. The final analysis sample consisted of 623 men aged 41–86 years. The effect of OSA on testosterone were analyzed using linear regression models controlling for potential confounders (age, BMI and sex hormone binding globulin (SHBG)).
Results
The mean (s.d.) cohort characteristics were: age 59.0 (10.2) years, testosterone 16.8 (5.3) nmol/L, SHBG 32.9 (13.1) nmol/L, BMI 28.6 (4.2) kg/m2 and apnoea hypopnoea index (AHI) 14.9 (13.7). OSA was present in 51.5%. There was an inverse relationship between AHI and testosterone (P = 0.01), which was lost after covariate adjustment.
Conclusions
These data suggest that obesity, rather than OSA per se, determine testosterone concentration. This accords with the graded effect of weight loss, but limited effect of continuous positive airway pressure to increase testosterone, and highlights the importance of managing obesity in men with low testosterone concentration, particularly in the context of OSA.
Introduction
Obstructive sleep apnea (OSA) is characterised by the recurrent interruption of breathing during sleep due to a collapse of the upper airways. Recent data from different countries reveal it to be a common disorder, with a reported prevalence around 50% in middle-aged and older men (1, 2). Although not an exclusively obesity-related condition, the prevalence of OSA rises with increasing obesity. In patients being evaluated for bariatric surgery, 71% of individuals with a BMI between 35–40 kg/m2 were diagnosed with OSA on polysomnography, with the figure rising to 95% among those with class 3 obesity and a BMI >60 kg/m2 (3). These findings were supported by a recent systematic review that found rising BMI to be a factor associated with an increase in OSA prevalence (4).
An inverse relationship between the presence of OSA and serum testosterone concentration in men has been reported in some (5, 6, 7, 8) but not all studies (9, 10, 11). Interpretation of many of these studies has been hampered by observational study design, small sample sizes and limited or absent covariates in the data analyses (5, 6, 7, 8). The largest study cohort studying this association detected a relationship between low testosterone concentration and OSA; however, this association became non-significant after adjustment for either BMI or waist circumference (9).
Given the discrepant data and still widely held clinical view that OSA causes low testosterone levels, we examined the relationship between the presence and severity of OSA and serum testosterone concentration in a comprehensively characterized community dwelling cohort of middle-aged and elderly men, approximately half of whom were randomly invited to undergo a home sleep study (12).
Subjects and methods
Study participants and protocol
The Men Androgen Inflammation Lifestyle Environment and Stress (MAILES) Study, as has previously been described (12), includes randomly selected men from the community, aged 35 years or more at enrolment. The study was approved by the Human Research Ethics Committees of the Royal Adelaide Hospital (approval number: 020305) and North West Adelaide Health Service (approval number: 2010054). All participants provided written informed consent after full explanation of the purpose and nature of the study. Data for analysis in the current study were collected from follow-up biomedical assessments, including fasting blood samples completed between 2007 and 2010 (time point codes M2, M3) and sleep assessments (M4) completed between September 2010 and February 2012.
A computer-assisted telephone interview was completed by MAILES study participants (n = 1,629) to identify men with a prior diagnosis of OSA. Subjects reporting ‘No’ when asked ‘Have you ever been diagnosed with OSA with a sleep study?’ were invited to undergo a sleep study. Overall, 1087 of 1445 without a previous OSA diagnosis (75.2%) initially agreed to participate (12). From this group, a randomly selected sample (n = 861) underwent polysomnography. A total of 238 men were subsequently excluded from the data analysis due to known hypothalamic-pituitary-gonadal (HPG) disease (n = 43), pre-existing use of medications known to affect testosterone concentration, SHBG or sleep (n = 76), presence of active malignancy (n = 20) or missing primary measurement values (BMI, AHI or testosterone) (n = 99). The final analysis group consisted of 623 men, aged 41 to 86 years. The methods of population selection for the analysis are outlined in Fig. 1.
Participant flow chart for study analysis cohort. MAILES, men androgen inflammation lifestyle environment and stress study; CATI, computer assisted telephone interview; OSA, obstructive sleep apnoea; PSG, polysomnography; AHI, apnea hypopnea index; BMI, body mass index; SHBG, sex hormone binding globulin; HPG, hypothalamic pituitary gonadal.
Citation: European Journal of Endocrinology 183, 1; 10.1530/EJE-19-0978
Participant flow chart for study analysis cohort. MAILES, men androgen inflammation lifestyle environment and stress study; CATI, computer assisted telephone interview; OSA, obstructive sleep apnoea; PSG, polysomnography; AHI, apnea hypopnea index; BMI, body mass index; SHBG, sex hormone binding globulin; HPG, hypothalamic pituitary gonadal.
Citation: European Journal of Endocrinology 183, 1; 10.1530/EJE-19-0978
Participant flow chart for study analysis cohort. MAILES, men androgen inflammation lifestyle environment and stress study; CATI, computer assisted telephone interview; OSA, obstructive sleep apnoea; PSG, polysomnography; AHI, apnea hypopnea index; BMI, body mass index; SHBG, sex hormone binding globulin; HPG, hypothalamic pituitary gonadal.
Citation: European Journal of Endocrinology 183, 1; 10.1530/EJE-19-0978
Anthropometry
Anthropometric measures and current medications were recorded at the time of sleep study completion. Weight, height and waist circumference were recorded to the nearest 0.1 kg and 0.1 cm, respectively, according to methods as previously described (13). BMI in kilograms per metre squared was calculated according to routine methodology. Overweight and obesity were defined by the standard BMI thresholds 25 and 30 kg/m2, respectively.
Polysomnography
Participants were visited in their homes by trained staff to complete 8-channel in-home unattended polysomnography (Embletta X100, Embla Systems, Thornton, CO, USA) for measurement of EEG, electrooculogram, electromyogram, nasal pressure, thoracic and abdominal effort, oximetry and body position.
All home polysomnography reports were scored manually by a single experienced sleep technician according to the 2007 American Academy of Sleep Medicine criteria (14). Hypopneas were defined as a greater than 50% reduction in nasal pressure or in both thoracic and abdominal excursions, in conjunction with a ≥3% oxygen desaturation or arousal on EEG. An apnea hypopnea index (AHI) ≥10 per h was used to define the presence of OSA, with severity based on the International Classification of Sleep Disorders (ICSD-2) and classified as follows: mild: AHI 10–19; moderate: 20–29 and severe: ≥30 per h. The rationale for selecting these thresholds in our cohort has been previously described (15) and was based on an AHI cut off of five events/h, as used to define sleep-disordered breathing in the Wisconsin Sleep Cohort Study (16), having been determined to be approximately equivalent to an AHI of ten events/h using the alternate American Academy criteria (17). The total sleep time with oxygen desaturation below 90% (TST90) and oxygen desaturation index (ODI), describing the average number of oxygen desaturation events ≥3% per h, were extracted from oximetry data.
Assays
Blood plasma samples were collected by venepuncture in the morning following an overnight fast and stored at −80°C. Total testosterone was assayed by API-5000 triple-quadruple mass spectrometer (Applied Biosystems/MDS SCIEX) (18). The inter-assay coefficients of variation (CV) for the testosterone assay were 10.2% at 0.43 nmol/L, 11.1% at 1.66 nmol/L and 4% at 8.17 nmol/L. Sex hormone binding globulin (SHBG) and glycated haemoglobin (HbA1c) were measured using two-site chemiluminescent immunoassay and HPLC, respectively, in a NATA certified laboratory (SA Pathology, Adelaide, Australia).
Statistical analysis
Associations between baseline covariates and measured testosterone concentrations were assessed using Jonckheere’s trend tests. Baseline covariates included general demographics (age, marital status and years of shift work), average consumption habits (alcohol, caffeine and smoking), measures of adiposity (BMI and waist circumference), metabolic morbidities (diabetes and triglycerides), endocrine (SHGB) and polysomnography measures (AHI, arousal index, TST90 and ODI). For these analyses subjects were divided by testosterone concentration quartiles (Q1: ≤13.1, Q2: 13.1–16.3, Q3: 16.3–20.2 and Q4: >20.2).
To evaluate the effect of OSA on testosterone concentration, linear regression analyses were constructed for each of the sleep parameters AHI, ODI, TST90 and arousal index. For each linear regression model, covariate adjustment included log SHBG (model 1), secondly for a combination of log SHBG, BMI and study arm (model 2) and finally all three of these factors plus age (model 3). In each of these models, the measures of OSA included were as linear continuous predictors, with total testosterone being log transformed. Examination of residual distributions ensured that assumptions concerning error distributions were not unduly violated. For each model coefficent described, we report the standardized estimate, 95% CI and P-value.
Each of the multiply adjusted models (model 3) were extended using restricted cubic splines with four degrees of freedom. Log-likelihood ratio tests of the nested linear continuous models and the non-linear models were used to assess the presence of non-linear associations between OSA measures and total testosterone. To present non-linear estimates of the final models (model 3), the OSA continuous measures were replaced with these measures divided into quartiles. The exception in this model being AHI, for which standard thresholds based on OSA severity categories (no OSA: AHI ≤ 10; mild OSA: 10 < AHI ≤ 20; Moderate: 20 < AHI ≤ 30 and Severe AHI: > 30) are presented.
As data collection was sequential rather than from a single time point, post-hoc stability analyses were undertaken in order to evaluate the assumption of stability of both total testosterone concentration and OSA within individual subjects over the study period. Correlations in testosterone concentrations between time points M1 and M2 and BMI measurements from M2 and M4 were evaluated. A subsequent subgroup sensitivity analysis was completed to determine the effect that variation within these parameters may have had on the primary outcome in men with minimal variation in testosterone. In all analyses significance is set at 0.05 (two-sided). Statistical analyses were conducted using R version 3.5.0.
Results
The mean (s.d.) age of the 623 men analyzed was 59 (10) years, with an average (s.d.) BMI of 28.2 (4.2) kg/m2 (Range: 17.2–46.7 kg/m2). Testosterone concentration ranged from 4.5 to 38.5 nmol/L, with a mean (s.d.) of 16.8 (5.3) nmol/L. The proportion of individuals with testosterone concentration measured below the thresholds of 6.0 mmol/L, 8.0 mmol/L and 10.0 mmol/L were 0.3%, 2.4% and 8.0%, respectively. The prevalence of OSA of any severity was 51%. The breakdown by OSA severity category was as follows: mild: 26%, moderate: 14% and severe: 11%.
The distribution of demographic, lifestyle, comorbidities, hormonal, anthropomorphic and sleep parameters according to measured total testosterone quartiles are provided in Table 1. There were strong associations between testosterone levels and both measures of adiposity, the presence of either diabetes or hypertriglyceridemia, SHBG, TST90 (all P < 0.0001) and ODI (P = 0.0002). A weak negative association was apparent between total testosterone levels and AHI (P = 0.04).
Patient characteristics by total testosterone quartiles.
| All | Q1 | Q2 | Q3 | Q4 | P-value | |
|---|---|---|---|---|---|---|
| n = 623 | n = 157 | n = 157 | n = 153 | n = 156 | ||
| Cohort | 0.008 | |||||
| NWAHS | 262 (42%) | 60 (38%) | 55 (35%) | 68 (44%) | 79 (51%) | |
| FAMAS | 361 (58%) | 97 (62%) | 102 (65%) | 85 (56%) | 77 (49%) | |
| Demographics | ||||||
| Age (M4) | 59 (10.2) | 61.3 (9.8) | 57.2 (9.8) | 59.2 (9.7) | 58.5 (11.3) | 0.08 |
| Marital Status (M4) | 0.24 | |||||
| Partner | 502 (81%) | 131 (83%) | 126 (80%) | 125 (82%) | 120 (77%) | |
| Single | 120 (19%) | 26 (17%) | 31 (20%) | 28 (18%) | 35 (23%) | |
| Shift work, years (M3) | 0.08 | |||||
| None | 302 (49%) | 67 (44%) | 74 (47%) | 81 (54%) | 80 (51%) | |
| less than 3 years | 105 (17%) | 22 (14%) | 35 (22%) | 23 (15%) | 25 (16%) | |
| 3 or more years | 211 (34%) | 65 (42%) | 48 (31%) | 47 (31%) | 51 (33%) | |
| Lifestyle (M2) | ||||||
| Alcohol, StD/day | 1.7 (3.7) | 1.7 (5.5) | 1.9 (3.1) | 1.5 (2.4) | 1.6 (3.1) | 0.32 |
| Coffee, cups/day | 1.8 (1.6) | 1.7 (1.5) | 1.8 (1.5) | 1.8 (1.8) | 1.9 (1.6) | 0.66 |
| Smoking history | 0.37 | |||||
| Never | 248 (40%) | 64 (41%) | 58 (37%) | 64 (42%) | 62 (40%) | |
| Past | 270 (44%) | 80 (51%) | 65 (42%) | 61 (40%) | 64 (42%) | |
| Current | 101 (16%) | 13 (8%) | 33 (21%) | 27 (18%) | 28 (18%) | |
| Adiposity (M4) | ||||||
| BMI | 28.6 (4.2) | 30.5 (4.3) | 29 (4.2) | 28 (3.6) | 27 (3.9) | <0.0001 |
| Waist circumference (cm) | 102 (11.7) | 106.6 (11) | 102.8 (11.8) | 101.2 (10.8) | 97.5 (11.6) | <0.0001 |
| Metabolic Morbidities | ||||||
| Diabetes (M2) | <0.0001 | |||||
| No diabetes | 471 (77%) | 93 (61%) | 120 (77%) | 124 (83%) | 134 (87%) | |
| Diabetes | 141 (23%) | 60 (39%) | 36 (23%) | 25 (17%) | 20 (13%) | |
| High triglycerides (M2) | <0.0001 | |||||
| No, TG ≤1.7 mmol/L | 397 (64%) | 77 (49%) | 106 (68%) | 94 (61%) | 120 (77%) | |
| Yes, TG >1.7 mmol/L | 226 (36%) | 80 (51%) | 51 (32%) | 59 (39%) | 36 (23%) | |
| Endocrine | ||||||
| SHBG (M2) | 32.9 (13.1) | 23.6 (8.4) | 30 (9.6) | 34.4 (10.8) | 43.9 (13.8) | <0.0001 |
| Polysomnography (M4) | ||||||
| AHI (1/h) | 14.9 (13.7) | 17.2 (15.5) | 14.6 (11.6) | 14.1 (14.5) | 13.8 (12.6) | 0.04 |
| Arousal Index (1/h) | 18.1 (7.8) | 19.2 (8.8) | 17.5 (7.5) | 17.9 (7.8) | 17.7 (7.1) | 0.39 |
| TST90 (min) | 16.2 (40.6) | 21.4 (45.9) | 18.2 (50.9) | 16 (34.9) | 9.3 (25.1) | <0.0001 |
| ODI (1/h) | 12.2 (12.5) | 14.8 (14.7) | 11.7 (9.6) | 11.6 (13.6) | 10.5 (11.2) | 0.0002 |
Testosterone quartile divisions were as follows: Q1: ≤13.1, Q2: 13.1–16.3, Q3: 16.3–20.2, Q4: >20.2 nmol/L. Time points for data collection are annotated by M2, M3 and M4.
AHI, apnea hypopnea index; FAMAS, Florey Adelaide Male Aging Study cohort; NWAHS, North West Adelaide Health Study cohort; ODI, oxygen desaturation index; SHBG, sex hormone binding globulin; StD, standard drinks; TG, triglycerides; TST90, total sleep time with oxygen saturation below 90%.
In the linear regression models of log testosterone, after adjustment for SHBG there were negative associations with AHI, ODI (both P = 0.01) and TST90 (P = 0.0008), but not arousal index (P = 0.08). After additional adjustment for BMI and study (Florey Adelaide Male Aging Study (FAMAS) vs North West Adelaide Health Study (NWAHS)), there were no detectable linear associations between testosterone and either AHI or ODI, though the association between testosterone and TST90 remained weakly significant. No detectable linear associations between testosterone and any of the four OSA measures remained after further adjustment for age (Fig. 2 and Table 2). In these analyses, BMI and age were both strongly negatively associated with log testosterone, while SHBG was strongly positively associated (all P < 0.0001, Supplementary Table 1, see section on supplementary materials given at the end of this article).
Effect of OSA parameters on testosterone concentration, after adjustment for covariates. The linear regression model estimates and 95% CIs for the effect of different OSA parameters on (log) testosterone concentration, after adjustment for covariates are represented. Model 1: adjustment for log SHBG. Model 2: adjustment for log SHBG, BMI and study arm. Model 3: adjustment for log SHBG, BMI, study arm and age. AHI, apnea hypopnea index; ODI, oxygen desaturation index; TST90, total sleep time with oxygen saturation below 90%.
Citation: European Journal of Endocrinology 183, 1; 10.1530/EJE-19-0978
Effect of OSA parameters on testosterone concentration, after adjustment for covariates. The linear regression model estimates and 95% CIs for the effect of different OSA parameters on (log) testosterone concentration, after adjustment for covariates are represented. Model 1: adjustment for log SHBG. Model 2: adjustment for log SHBG, BMI and study arm. Model 3: adjustment for log SHBG, BMI, study arm and age. AHI, apnea hypopnea index; ODI, oxygen desaturation index; TST90, total sleep time with oxygen saturation below 90%.
Citation: European Journal of Endocrinology 183, 1; 10.1530/EJE-19-0978
Effect of OSA parameters on testosterone concentration, after adjustment for covariates. The linear regression model estimates and 95% CIs for the effect of different OSA parameters on (log) testosterone concentration, after adjustment for covariates are represented. Model 1: adjustment for log SHBG. Model 2: adjustment for log SHBG, BMI and study arm. Model 3: adjustment for log SHBG, BMI, study arm and age. AHI, apnea hypopnea index; ODI, oxygen desaturation index; TST90, total sleep time with oxygen saturation below 90%.
Citation: European Journal of Endocrinology 183, 1; 10.1530/EJE-19-0978
Linear regressions of log total testosterone on AHI, ODI, TST90 and arousal index.
| Model 1 | Model 2 | Model 3 | ||||
|---|---|---|---|---|---|---|
| Coef (95%CI) | P-value | Coef (95%CI) | P-value | Coef (95%CI) | P-value | |
| AHI | −0.026 (−0.047, −0.0055) | 0.01 | −0.0019 (−0.022, 0.018) | 0.85 | 0.014 (−0.0047, 0.033) | 0.14 |
| ODI | −0.027 (−0.048, −0.0063) | 0.01 | 0.00083 (−0.02, 0.021) | 0.94 | 0.014 (−0.0049, 0.033) | 0.15 |
| TST90 | −0.035 (−0.056, −0.015) | 0.0008 | −0.022 (−0.041, −0.0026) | 0.03 | −0.0045 (−0.023, 0.014) | 0.63 |
| Arousals | −0.019 (−0.039, 0.0019) | 0.08 | −0.0085 (−0.028, 0.011) | 0.39 | 0.0047 (−0.013, 0.023) | 0.61 |
For each measure of obstructive sleep apnea, three models are constructed adjusting for (model 1) log SHBG, (model 2) log SHBG, BMI and study (FAMAS vs NWAHS) and (model 3) log SHBG, age, BMI and study. Standardized coefficients are reported.
AHI, apnea hypopnea index; ODI, oxygen desaturation index; TST90, total sleep time with oxygen saturation below 90%.
Extending these multivariable models with restricted cubic splines and comparing linear vs non-linear models using log-likelihood ratio tests indicated a weak non-linear association with AHI (P = 0.04). There was no evidence for non-linear associations with either ODI (P = 0.17), TST90 (P = 0.99) or arousal index (P = 0.92). In the model with non-linear AHI, 54.6% of the variation of log testosterone was explained, as compared to 53.8% that is explained by the four covariates alone (age, BMI, log SHBG and study). Replacing continuous AHI with the four-level discretised version (none, mild, moderate vs severe) indicated an inverted U-shape with log-transformed total testosterone higher in men with mild (P = 0.04) or moderate (P = 0.03) OSA as compared to men with negligble OSA (Supplementary Table 2).
The post-hoc stability analyses demonstrated that BMI remained very stable overtime (correlation, 0.95), though there was some variability in total testosterone (correlation 0.72) (Supplementary Fig. 1). However, restriction of the cohort to include only men in whom changes of less than 25% in total testosterone were observed (n = 466), the associations between total testosterone and OSA measures were almost unchanged from the analyses in the full cohort, though widening of CIs is noted, in keeping with the smaller sample size (Supplementary Table 3).
Discussion
OSA is a common disorder, with an overall prevalence of more than 50% found in our study population of community dwelling men with no known history of OSA. This rate is similar to those reported in other populations of middle-aged and older men in recent years (1, 2). It appears that OSA remains significantly under-diagnosed in the community, evident by only 11% of our population screened for sleep study participation reporting a history of diagnosed OSA.
The main findings from our results suggest that OSA is not an independent determinant of serum testosterone concentration in community dwelling middle-aged and older men. Though there was a linear association between lower measured total testosterone concentration and more severe sleep disordered breathing as measured by both AHI and ODI, the association was lost with adjustment for BMI. Supplementary Table 1 reports the parameter estimates for the AHI M3 model (Table 2) in full and confirms there is no association between testosterone and AHI after adjustment for confounding by age and BMI. This model assumes linearity, and under this assumption there is no evidence of differences in testosterone levels with mild to moderate OSA. Increased total duration of oxygen desaturation, as measured by the TST90, was also associated with lower testosterone concentration; and though the relationship remained borderline significant after adjustment for BMI, the association was again lost after adjustment for age.
Several past observational studies have reported outcomes contrary to our results, finding a negative relationship between testosterone concentration and the presence of OSA that is independent of BMI. However, all of these studies have been limited by small study populations, with cohorts of less than 100 subjects (5, 6, 7, 8, 10, 11). Other limitations include failure to clearly account for other factors that may influence testosterone (5), failure to adjust for the effect of BMI on testosterone when evaluating the relationship with OSA (6) and a lack of objective evaluation of OSA in all patients studied (8). It is also worth noting that though the two small studies by Luboshitzky (10, 11) found persistence of the relationship between lower testosterone and the categorical presence of OSA, the authors concluded that the primary cause for altered gonadal function in this population was due to the effects of obesity and aging, though sleep fragmentation and hypoxia may also contribute. We have reported results from a large sample of community dwelling middle-aged and older men, with outcomes similar to the findings by Barrett-Connor’s study of more than 1300 elderly community dwelling men (9); There is an inverse relationship between testosterone and polysomnography measures of OSA including AHI, though these associations are attenuated after adjustment for BMI. In comparison to Barrett-Connor’s cohort, our patient population was on average 15 years younger (59 years compared with 73 years) and had proportionally fewer men with a low testosterone concentration below 8.7 nmol/L (250 ng/dL) (4% compared with 9%). Notably, the mean BMI across each of our testosterone quartiles was higher than in Barrett-Connor’s cohort, though a similar trend of rising BMI with lower testosterone concentration was still observed.
Our data are also supported by evidence showing that testosterone concentration is not improved by treatment of OSA with continuous positive airway pressure (CPAP), irrespective of treatment duration (19, 20, 21, 22, 23, 24, 25). Only a single study has reported an increase in total testosterone after 3 months of nasal CPAP treatment, used in 43 men with severe OSA (26). However, in this report there was no change in measured free testosterone due to a rise in SHBG in conjunction with testosterone rise during the intervention. The outcome in this study was also not adjusted for the change in body mass over the treatment period, with a mean weight loss of 1.4 kg reported during the intervention which reached borderline significance (P = 0.06). Placebo-controlled evaluation of the effect of CPAP on testosterone concentration has been evaluated in only a single study; though more severe OSA correlated with lower testosterone concentrations at baseline, 1 month of active nasal CPAP treatment as compared to sham CPAP did not result in a significant rise in the measured testosterone concentration (27). A recent meta-analysis of this topic also concluded that, in men with OSA, CPAP has no influence on testosterone concentration (28). The three published studies excluded from the meta-analysis due to insufficient or missing key data, all still reported findings in support of this conclusion (19, 25, 29). By contrast, improvement in testosterone concentration has been consistently demonstrated in response to weight loss, with the magnitude of improvement in T proportional to the degree of weight loss achieved (30). As far as we can determine, there is no evidence to support a converse hypothesis that OSA is more common in hypogonadal men, in the absence of concurrent obesity.
A strength of our study is the inclusion of a large number of subjects with baseline characteristics representative of the broader community. We have evaluated sleep disordered breathing using a validated and reliable diagnostic method in all participants and utilised a variety of different objective polysomnography parameters, including AHI, ODI, TST90 and arousals in the analyses. Though we have selectively employed BMI as the measure of adiposity in our study for statistical evaluation, there was a very high degree of positive correlation between all adiposity measures recorded (results not shown), such that it is unlikely that substituting BMI for an alternative parameter of adiposity would significantly alter the OSA-testosterone associations found. We have also previously reported an analysis of self-selection bias, which did not find any significant differences in relation to obesity or sleep symptoms between those who did and did not participate in the sleep study (31).
The main limitation of our study is its observational design, meaning that associations may be drawn from the data, though this is insufficient to prove causality of BMI in mediating the effect of OSA on total testosterone concentration. A further limitation is the stepwise collection of data at different time points; as the timing of assessments was not identical, the data is not a true cross-sectional analysis from a single time point. For transparency, the timing of data collection has been indicated in the relevant tables by the codes M1, M2, M3 and M4, respectively. As testosterone concentration was not remeasured at the time of the sleep study, it is not possible to definitively exclude a change in measured testosterone between the time of testosterone measurement used in the analysis (M2) and the time of the sleep study (M4), confounding the relationship between these parameters. To address this issue within the limitations of the available data, we performed post-hoc stability analyses of testosterone measurements and BMI. From these analyses, we conclude that the high level of stability of BMI between time point M2 and M4 supports that it is a reasonable assumption that the measures of OSA magnitude would also be stable over this same period. Testosterone levels were relatively stable between M1 and M2 and the time interval between M2 and M4 was significantly shorter (median 20 months, IQR 5 months, compared with 5 years), thus it is also reasonable to assume that testosterone measurements would have similar or greater stability over the period of interest. Furthermore, on restriction of the primary analysis to only those individuals whom showed minimal variability in measured testosterone, the result of the primary analysis remains essentially unchanged. This provides evidence supporting the assumption that the OSA-testosterone associations in the overall cohort are unaffected by the difference in assessment times (M2 for testosterone, M4 for sleep parameters).
We acknowledge that our study design excluded individuals with pre-existing obstructive sleep apnoea from the analysis. It is possible that these individuals had more severe OSA to have come to clinical attention. In such an instance, their exclusion from analysis may have resulted in a bias towards the detection of milder OSA, obscuring the relationship with testosterone concentration. However, a pre-existing diagnosis of OSA was uncommon in the screened population, particularly when compared to the diagnosis rates following sleep study completion, and despite the exclusion those with a pre-existing diagnosis, moderate or severe OSA was still detected in half of all individuals who met the OSA diagnostic criteria following sleep study. In combination, this seems to support the proposition that OSA is under-diagnosed in the community and less likely that our study design resulted in a skew towards the detection of mild OSA only. The inclusion of individuals with pre-diagnosed OSA would also result in the incorporation of a proportion of individuals treated for OSA, though the evidence, as discussed, indicating that the treatment of OSA with CPAP does not improve testosterone concentration independent of change in weight. With respect to the results obtained, it is also important to acknowledge that the CIs of the direct OSA effect, as measured by AHI, on testosterone after adjustment for covariates, remain large, and therefore could be consistent with any of three different outcomes, that there is no effect, a positive effect or a negative effect of OSA on testosterone concentration.
Overall the findings from our study lend further support to the argument that low serum testosterone concentrations, in the context of obesity, are not the consequence of comorbid OSA and rather reflect the effect of obesity. From a clinical standpoint, treatment of OSA with CPAP cannot be expected to lead to an improvement in testosterone concentration though, in contrast, weight loss can increase testosterone (30) and improve objective measures of OSA (32, 33). Taken together, these findings highlight the importance of addressing obesity in men with OSA, particularly if testosterone is decreased, irrespective of whether CPAP is instituted or not.
Supplementary materials
This is linked to the online version of the paper at https://doi.org/10.1530/EJE-19-0978.
Declaration of interest
G A W is the Editor-in-Chief of Obesity Research and Clinical Practice and a recipient of research support from Bayer, Eli Lilly, Weight Watchers, ResMed Foundation and Lawley Pharmaceuticals, as well as speaking fees from Bayer and Besins Health Care. R A is a recipient of funding support from the ResMed Foundation. The other authors have nothing to disclose.
Funding
This work was supported by the National Health and Medical Research Council of Australia (Project grant number 627227). Unrestricted financial support for the conduct of sleep studies was also obtained from the ResMed Foundation, California, USA.
Acknowledgements
The authors would like to acknowledge the staff of the Florey Adelaide Male Ageing Study, North West Adelaide Health Study, Clinic and Sleep Department, in particular Leanne Owen, Janet Grant, Sandy Pickering and Tina Stavropoulos, for their contributions. The authors also acknowledge Siemens Healthcare for providing the RIA kits for all Immulite assays and Embla Systems, Inc. (Colorado, USA) for the use of the polysomnography machines. Thanks are also extended to the participants and their families for their invaluable contributions.
References
- 1↑
Adams R, Appleton S, Taylor A, McEvoy D, Wittert G. Are the ICSD-3 criteria for sleep apnoea syndrome too inclusive? Lancet: Respiratory Medicine 2016 4 e19–e20. (https://doi.org/10.1016/S2213-2600(16)00109-0)
- 2↑
Heinzer R, Marti-Soler H, Haba-Rubio J. Prevalence of sleep apnoea syndrome in the middle to old age general population. Lancet: Respiratory Medicine 2016 4 e5–e6. (https://doi.org/10.1016/S2213-2600(16)00006-0)
- 3↑
Lopez PP, Stefan B, Schulman CI, Byers PM. Prevalence of sleep apnea in morbidly obese patients who presented for weight loss surgery evaluation: more evidence for routine screening for obstructive sleep apnea before weight loss surgery. American Surgeon 2008 74 834–838.
- 4↑
Senaratna CV, Perret JL, Lodge CJ, Lowe AJ, Campbell BE, Matheson MC, Hamilton GS, Dharmage SC. Prevalence of obstructive sleep apnea in the general population: a systematic review. Sleep Medicine Reviews 2017 34 70–81. (https://doi.org/10.1016/j.smrv.2016.07.002)
- 5↑
Hammoud AO, Walker JM, Gibson M, Cloward TV, Hunt SC, Kolotkin RL, Adams TD, Meikle AW. Sleep apnea, reproductive hormones and quality of sexual life in severely obese men. Obesity 2011 19 1118–1123. (https://doi.org/10.1038/oby.2010.344)
- 6↑
Canguven O, Salepci B, Albayrak S, Selimoglu A, Balaban M, Bulbul M. Is there a correlation between testosterone levels and the severity of the disease in male patients with obstructive sleep apnea? Archivio Italiano di Urologia, Andrologia 2010 82 143–147.
- 7↑
Kirbas G, Abakay A, Topcu F, Kaplan A, Unlu M, Peker Y. Obstructive sleep apnoea, cigarette smoking and serum testosterone levels in a male sleep clinic cohort. Journal of International Medical Research 2007 35 38–45. (https://doi.org/10.1177/147323000703500103)
- 8↑
Gambineri A, Pelusi C, Pasquali R. Testosterone levels in obese male patients with obstructive sleep apnea syndrome: relation to oxygen desaturation, body weight, fat distribution and the metabolic parameters. Journal of Endocrinological Investigation 2003 26 493–498. (https://doi.org/10.1007/BF03345209)
- 9↑
Barrett-Connor E, Dam TT, Stone K, Harrison SL, Redline S, Orwoll E & Osteoporotic Fractures in Men Study Group. The association of testosterone levels with overall sleep quality, sleep architecture, and sleep-disordered breathing. Journal of Clinical Endocrinology and Metabolism 2008 93 2602–2609. (https://doi.org/10.1210/jc.2007-2622)
- 10↑
Luboshitzky R, Lavie L, Shen-Orr Z, Herer P. Altered luteinizing hormone and testosterone secretion in middle-aged obese men with obstructive sleep apnea. Obesity Research 2005 13 780–786. (https://doi.org/10.1038/oby.2005.88)
- 11↑
Luboshitzky R, Aviv A, Hefetz A, Herer P, Shen-Orr Z, Lavie L, Lavie P. Decreased pituitary-gonadal secretion in men with obstructive sleep apnea. Journal of Clinical Endocrinology and Metabolism 2002 87 3394–3398. (https://doi.org/10.1210/jcem.87.7.8663)
- 12↑
Grant JF, Martin SA, Taylor AW, Wilson DH, Araujo A, Adams RJ, Jenkins A, Milne RW, Hugo GJ & Atlantis E et al. Cohort profile: the men androgen inflammation lifestyle environment and stress (MAILES) study. International Journal of Epidemiology 2014 43 1040–1053. (https://doi.org/10.1093/ije/dyt064)
- 13↑
Shi Z, Araujo AB, Martin S, O’Loughlin P, Wittert GA. Longitudinal changes in testosterone over five years in community-dwelling men. Journal of Clinical Endocrinology and Metabolism 2013 98 3289–3297. (https://doi.org/10.1210/jc.2012-3842)
- 14↑
Berry RB, Budhiraja R, Gottlieb DJ, Gozal D, Iber C, Kapur VK, Marcus CL, Mehra R, Parthasarathy S, Quan SF, Redline S, et al. Rules for scoring respiratory events in sleep: update of the 2007 AASM Manual for the Scoring of Sleep and Associated Events. Deliberations of the Sleep Apnea Definitions Task Force of the American Academy of Sleep Medicine. Journal of Clinical Sleep Medicine 2012 8 597–619. (https://doi.org/10.5664/jcsm.2172)
- 15↑
Adams RJ, Appleton SL, Vakulin A, Lang C, Martin SA, Taylor AW, McEvoy RD, Antic NA, Catcheside PG, Wittert GA. Association of daytime sleepiness with obstructive sleep apnoea and comorbidities varies by sleepiness definition in a population cohort of men. Respirology 2016 21 1314–1321. (https://doi.org/10.1111/resp.12829)
- 16↑
Peppard PE, Szklo-Coxe M, Hla KM, Young T. Longitudinal association of sleep-related breathing disorder and depression. Archives of Internal Medicine 2006 166 1709–1715. (https://doi.org/10.1001/archinte.166.16.1709)
- 17↑
Ruehland WR, Rochford PD, O’Donoghue FJ, Pierce RJ, Singh P, Thornton AT. The new AASM criteria for scoring hypopneas: impact on the apnea hypopnea index. Sleep 2009 32 150–157. (https://doi.org/10.1093/sleep/32.2.150)
- 18↑
Harwood DT, Handelsman DJ. Development and validation of a sensitive liquid chromatography-tandem mass spectrometry assay to simultaneously measure androgens and estrogens in serum without derivatization. Clinica Chimica Acta: International Journal of Clinical Chemistry 2009 409 78–84. (https://doi.org/10.1016/j.cca.2009.09.003)
- 19↑
Macrea MM, Martin TJ, Zagrean L. Infertility and obstructive sleep apnea: the effect of continuous positive airway pressure therapy on serum prolactin levels. Sleep and Breathing 2010 14 253–257. (https://doi.org/10.1007/s11325-010-0373-0)
- 20↑
Hoekema A, Stel AL, Stegenga B, van der Hoeven JH, Wijkstra PJ, van Driel MF, de Bont LG. Sexual function and obstructive sleep apnea-hypopnea: a randomized clinical trial evaluating the effects of oral-appliance and continuous positive airway pressure therapy. Journal of Sexual Medicine 2007 4 1153–1162. (https://doi.org/10.1111/j.1743-6109.2006.00341.x)
- 21↑
Bratel T, Wennlund A, Carlstrom K. Pituitary reactivity, androgens and catecholamines in obstructive sleep apnoea. Effects of continuous positive airway pressure treatment (CPAP). Respiratory Medicine 1999 93 1–7. (https://doi.org/10.1016/s0954-6111(99)90068-9)
- 22↑
Knapp A, Myhill PC, Davis WA, Peters KE, Hillman D, Hamilton EJ, Lim EM, Davis TM. Effect of continuous positive airway pressure therapy on sexual function and serum testosterone in males with type 2 diabetes and obstructive sleep apnoea. Clinical Endocrinology 2014 81 254–258. (https://doi.org/10.1111/cen.12401)
- 23↑
Celec P, Mucska I, Ostatnikova D, Hodosy J. Testosterone and estradiol are not affected in male and female patients with obstructive sleep apnea treated with continuous positive airway pressure. Journal of Endocrinological Investigation 2014 37 9–12. (https://doi.org/10.1007/s40618-013-0003-3)
- 24↑
Zhang XB, Lin QC, Zeng HQ, Jiang XT, Chen B, Chen X. Erectile dysfunction and sexual hormone levels in men with obstructive sleep apnea: efficacy of continuous positive airway pressure. Archives of Sexual Behavior 2016 45 235–240. (https://doi.org/10.1007/s10508-015-0593-2)
- 25↑
Vlkova B, Mucska I, Hodosy J, Celec P. Short-term effects of continuous positive airway pressure on sex hormones in men and women with sleep apnoea syndrome. Andrologia 2014 46 386–390. (https://doi.org/10.1111/and.12092)
- 26↑
Grunstein RR, Handelsman DJ, Lawrence SJ, Blackwell C, Caterson ID, Sullivan CE. Neuroendocrine dysfunction in sleep apnea: reversal by continuous positive airways pressure therapy. Journal of Clinical Endocrinology and Metabolism 1989 68 352–358. (https://doi.org/10.1210/jcem-68-2-352)
- 27↑
Meston N, Davies RJ, Mullins R, Jenkinson C, Wass JA, Stradling JR. Endocrine effects of nasal continuous positive airway pressure in male patients with obstructive sleep apnoea. Journal of Internal Medicine 2003 254 447–454. (https://doi.org/10.1046/j.1365-2796.2003.01212.x)
- 28↑
Zhang XB, Jiang XT, Du YP, Yuan YT, Chen B. Efficacy of continuous positive airway pressure on testosterone in men with obstructive sleep apnea: a meta-analysis. PLoS ONE 2014 9 e115033. (https://doi.org/10.1371/journal.pone.0115033)
- 29↑
Zhuravlev VN, Frank MA, Gomzhin AI. Sexual functions of men with obstructive sleep apnoea syndrome and hypogonadism may improve upon testosterone administration: a pilot study. Andrologia 2009 41 193–195. (https://doi.org/10.1111/j.1439-0272.2008.00914.x)
- 30↑
Grossmann M. Low testosterone in men with type 2 diabetes: significance and treatment. Journal of Clinical Endocrinology and Metabolism 2011 96 2341–2353. (https://doi.org/10.1210/jc.2011-0118)
- 31↑
Appleton SL, Vakulin A, McEvoy RD, Wittert GA, Martin SA, Grant JF, Taylor AW, Antic NA, Catcheside PG, Adams RJ. Nocturnal hypoxemia and severe obstructive sleep apnea are associated with incident type 2 diabetes in a population cohort of men. Journal of Clinical Sleep Medicine 2015 11 609–614. (https://doi.org/10.5664/jcsm.4768)
- 32↑
Kuna ST, Reboussin DM, Borradaile KE, Sanders MH, Millman RP, Zammit G, Newman AB, Wadden TA, Jakicic JM & Wing RR et al. Long-term effect of weight loss on obstructive sleep apnea severity in obese patients with type 2 diabetes. Sleep 2013 36 641–649A. (https://doi.org/10.5665/sleep.2618)
- 33↑
Mitchell LJ, Davidson ZE, Bonham M, O’Driscoll DM, Hamilton GS, Truby H. Weight loss from lifestyle interventions and severity of sleep apnoea: a systematic review and meta-analysis. Sleep Medicine 2014 15 1173–1183. (https://doi.org/10.1016/j.sleep.2014.05.012)
