Association between obstructive sleep apnea and sex hormones in U.S. adult males: a population-based study Authors: Yuan Ji, Pengbo Hua, Haoyue Gao, Qiming Liang, Jingqing Wu, Liwei Peng, Jingqi Wang Published: 2026 ABSTRACT AbstractBackgroundThe relationship between obstructive sleep apnea (OSA) and comprehensive sex hormone profiles, including free testosterone, in men remains unclear.AimThis study aimed to explore the independent association between OSA risk and a panel of sex hormone indices after adjusting for obesity.MethodsWe conducted a cross-sectional analysis of 2218 adult males from the National Health and Nutrition Examination Survey 2015–2016. OSA was assessed using an adjusted multivariable apnea prediction index and categorized into tertiles. Sex hormones were measured via isotope dilution liquid chromatography–tandem mass spectrometry and immunoassays. Associations were examined using weighted multivariable linear and logistic regression, adjusted for demographics, and clinical factors. Restricted cubic spline and mediation analyses were performed to assess nonlinearity and the mediating role of sex hormone-binding globulin.OutcomesPrimary outcomes were total testosterone, sex hormone-binding globulin, and free androgen index; secondary outcomes included estradiol, total testosterone/estradiol ratio, and testosterone deficiency prevalence.ResultsIn fully adjusted models, higher OSA risk was significantly associated with lower ln(total testosterone) (β = -0.23, 95% CI, -0.40 to -0.06) and ln(sex hormone-binding globulin) (β = -0.49, 95% CI, -0.66 to -0.31). No significant association was found with free testosterone or estradiol. Testosterone deficiency prevalence was markedly higher in the high-risk OSA group (43.83%) versus the low-risk group (12.70%), with a borderline significant association for the continuous OSA score (OR = 3.85, P = .04). Sex hormone-binding globulin mediated a substantial proportion of the OSA-total testosterone relationship (P < .001).Clinical ImplicationsRoutine screening for sex hormone-binding globulin in obese OSA patients may help prevent misdiagnosis of hypogonadism and guide targeted intervention.Strengths and LimitationsStrengths include a nationally representative sample and comprehensive hormone profiling; limitations include the cross-sectional design and subjective OSA assessment.ConclusionOSA risk is associated with a hormonal profile of low total testosterone and sex hormone-binding globulin but preserved free testosterone, a relationship strongly moderated by body mass index. Introduction Obstructive sleep apnea (OSA) is a pervasive sleep-disordered breathing condition characterized by recurrent upper airway collapse, leading to chronic intermittent hypoxia (CIH) and profound sleep fragmentation.1 Recent epidemiological evidence indicates a rising prevalence, affecting approximately 27% of adult males in the United States, thereby posing a substantial burden on public health and economic stability.2 Beyond its hallmark symptoms of daytime sleepiness and apnea, OSA frequently entails systemic damage across multiple organ systems.3 OSA is closely associated with a constellation of severe comorbidities, including hypertension, cardiovascular disease, sudden cardiac death, diabetes, and chronic kidney disease.4 While its cardiovascular and metabolic sequelae are well-documented, a significant proportion of male patients also exhibit markedly reduced serum testosterone levels.5 Currently, OSA is increasingly recognized as a critical driver of male hypogonadism.6 This androgen deficiency not only impairs sexual health and quality of life but also synergistically exacerbates concomitant metabolic disturbances, fostering a self-perpetuating vicious cycle of endocrine and respiratory decline.5 The interplay between OSA and the male endocrine system is mediated through a complex bidirectional feedback loop. On one hand, the hallmark features of OSA—CIH and sleep fragmentation—disrupt the hypothalamic–pituitary–gonadal (HPG) and hypothalamic–pituitary–adrenal (HPA) axis, impairing Leydig cell steroidogenesis and attenuating the pulsatile secretion rhythm of gonadotropin-releasing hormone.7,8 This state is further compounded by systemic inflammation and oxidative stress, which upregulate peripheral aromatase activity in adipose tissue, shunting testosterone toward estradiol conversion and deepening hormonal imbalances.9 On the other hand, testosterone is vital for maintaining respiratory homeostasis; it preserves pharyngeal dilator muscle tone and enhances medullary chemoreceptor sensitivity to hypoxic stimuli, thereby stabilizing ventilatory drive.10,11 Consequently, diminished testosterone levels increase the risk of upper airway collapsibility, potentially inducing or exacerbating OSA.12 Despite this compelling biological rationale, epidemiological evidence remains fragmented and contradictory. While large-scale meta-analyses support an association between OSA and testosterone deficiency (TD), certain investigations suggest that obesity and aging are the primary drivers of androgen suppression, fueling intense academic debate.5,13-15 Recent National Health and Nutrition Examination Survey (NHANES)-based studies have attempted to clarify this link but suffer from notable methodological limitations. For instance, some studies utilized mixed-sex cohorts, potentially obscuring the sexual dimorphism inherent in HPG regulation,16 while others focused exclusively on binary outcomes of TD, neglecting the broader hormonal spectrum.17 Different from the above two studies, our study specifically targets adult males, employs continuous hormone profiling, and systematically investigates sex hormone-binding globulin (SHBG) mediation and body mass index (BMI) stratification. Most critically, the profound confounding effect of obesity frequently calls the independent role of OSA into question. Furthermore, the role of SHBG is often overlooked.14 Its hepatic synthesis is highly sensitive to the systemic inflammatory milieu and metabolic stress characteristic of OSA.18 Thus, prior research exhibits a significant "blind spot" by underestimating the biological weight of SHBG and failing to identify BMI as a potential effect modifier that may reveal divergent pathophysiological pathways across different populations. Previous studies have yielded inconsistent findings regarding the relationship between OSA and male sex hormones, largely due to inadequate adjustment for obesity, neglect of SHBG’s regulatory role, and failure to explore potential mediators.19 To address these gaps, the present study utilizes data from the NHANES 2015–2016 cycle, employing isotope dilution liquid chromatography–tandem mass spectrometry (ID-LC–MS/MS) for high-accuracy sex hormone quantification. Therefore, we aim to move beyond simple association to elucidate the specific mechanistic pathways—with SHBG at the core—that link sleep-disordered breathing to male hypogonadism. Specifically, we seek to answer the following refutable questions: (1) After rigorous adjustment for obesity and clinical comorbidities (including cardiovascular disease (CVD), diabetes, and chronic kidney disease), is OSA risk independently associated with serum total testosterone (TT) and SHBG levels in U.S. adult males? (2) Does SHBG act as a key mediator in the association between OSA risk and testosterone suppression? (3) Does BMI significantly modify these associations, indicating divergent mechanisms across weight categories? Methods Study design and population This cross-sectional study utilized data from the 2015–2016 cycle of the NHANES, administered by the National Center for Health Statistics (NCHS) under the Centers for Disease Control and Prevention (CDC). NHANES employs a complex, multistage, stratified probability sampling design to obtain a sample representative of the non-institutionalized civilian U.S. population to biennially assess health and nutritional status in a nationally representative U.S. population. Detailed protocols regarding the survey design, weighting methodology, and data collection are officially documented by the NCHS.20 Data collection normally included dietary intake, health conditions, lifestyle factors, and biochemical measurements from blood/urine samples, metabolite, etc., which are all obtained through household interviews, and mobile examination centers. The analysis was restricted to data from the NHANES 2015-2016 survey cycle. Participant inclusion criteria: Male participants aged ≥18 years with an availability of complete sex hormone testing; Participant exclusion criteria: Incomplete sex hormone data or missing of OSA risk data; Current taking of sex hormone-related medications (testosterone, progesterone, estrogen, or “other sex hormones” as specified in NHANES questionnaires) to ensure the integrity of the endocrine profiles. A total of 2218 participants were included in the final analysis. A flowchart of participant selection is provided in Figure 1. Detailed NHANES protocols, officially documented in the CDC/NCHS Analytic Guidelines, are available at the CDC website. The original survey protocol was approved by the NCHS Ethics Review Board, and written informed consent was obtained from all participants. The dataset utilized in this study is publicly available and de-identified; thus, no additional ethical clearance was required for this secondary analysis. The study adhered to the principles of the Declaration of Helsinki. Flow chart of the selection of the study participants from the NHANES 2015-2016 cycle. The diagram illustrates the stepwise exclusion criteria applied to derive the final analytical sample of 2218 adult male participants. NHANES, National Health and Nutrition Examination Survey Exposure and outcome definitions The “risk of OSA” (0-1.0) was evaluated with an adapted multivariable apnea prediction (MAP) index incorporating NHANES variables, originally developed and validated by Maislin et al.21 The MAP index quantifies the frequency of three key symptoms on a 4-point Likert scale: loud snoring (NHANES variable SLQ030: “How often do you snore?”), witnessed apnea,s, and gasping/choking (SLQ040: “How often do you snort, gasp, or stop breathing?”) [“never” = 0; “almost always” = 3 (Five times a week or more)]. Following the original MAP algorithm, an apnea index was first calculated as the average of nonmissing responses for the three symptom items. These symptom scores formed the core component of the MAP index, ranging from 0 (lowest risk) to 1 (highest risk). The adjusted MAP index, used in this study, integrated self-reported sleep symptoms with physiological variables (age, sex, BMI) to enhance the accuracy and universality.21 The MAP index has been widely used in epidemiological studies utilizing NHANES data to assess OSA risk.22-24 MAP scores were categorized into tertiles (T1: <0.33; T2: 0.33-0.66; T3: >0.66) based on population distribution. TT and estradiol (E2) were measured via the CDC-developed ID-LC–MS/MS at a single timepoint (morning, afternoon, or evening). This methodology is recognized as the gold standard for sex steroid analysis.25 Following NCHS analytical guidelines, hormone values below the lower limit of detection (LLOD) were imputed as LLOD/√2. The concentration of SHBG was quantified by measuring the reaction product of SHBG with immune antibodies and chemiluminescent immunoassays through a photomultiplier tube. Free androgen index (FAI) was calculated as [TT (nmol/L)/SHBG (nmol/L)] × 100, and FT was estimated using the Vermeulen equation. The TT/E2 ratio reflected the balance between the androgen (testosterone, TT) and the estrogen (estradiol, E2), also an indirect reflection for the amount of FT in circulation and the aromatase activity. A secondary outcome was TD, defined as TT < 300 ng/dL per American Urological Association guidelines.26 This methodology has been widely used in previous NHANES-based studies examining sex hormone profiles.27,28 Study covariates Based on previous studies,29 potential confounding variables that may influence the association between OSA risk and sex hormones were included in multivariable models. The covariates included age, BMI, race/ethnicity, CVD, diabetes mellitus (DM), chronic kidney disease (CKD), and hypertension, alcohol consumption status, smoking status, education level, waist circumference, marital status, family income level, and venipuncture time. BMI was calculated based on examination measurements (self-reported weight and height measurements) and categorized into three groups: < 25, 25–29.9, ≥ 30 kg/m2. Race/ethnicity was classified as Non-Hispanic Black, Non-Hispanic White, Mexican American, or Other Race. Education level was categorized into less than high school, high school graduate, some college or associate degree, and college graduate or higher. Drinking status was divided into heavy, moderate, mild, never, and former. Age was dichotomized as <45, ≥ 45 years based on physiological testosterone levels. Smoking status was classified as current, never, or former. Marital status was grouped into widowed/divorced, unmarried, or married/partnered. PIR was categorized into three levels using cutoff points of 1.5 and 3.0 times the poverty-to-income ratio. Venipuncture time was classified as morning, afternoon, or evening to account for potential diurnal variations in sex hormone levels. Comorbidities including CVD, DM, CKD, and hypertension were defined based on standardized self-reported medical history, prescription medication use, and relevant clinical and laboratory measurements (eg, HbA1c, fasting glucose, estimated glomerular filtration rate (eGFR), urine albumin-to-creatinine ratio (UACR), and blood pressure) in accordance with recognized guidelines. Statistical analysis All statistical analyses were conducted in strict accordance with the NCHS analytic and reporting guidelines to account for the complex survey architecture. To produce nationally representative estimates and adjust for non-response bias, we applied the 2-year mobile examination center (MEC) weights (WTMEC2YR) to all models.30 Variance estimation was performed using the Taylor series linearization method, incorporating primary sampling units (SDMVPSU) and strata (SDMVSTRA) to accommodate the cluster-sampling design. Continuous variables are presented as weighted means ± standard error (SE), and categorical variables as weighted proportions. Differences across risk of OSA tertiles were assessed using weighted chi-square tests (categorical variables) or survey-weighted linear models (continuous variables). Due to right-skewed distributions of sex hormones, natural logarithmic (ln) transformation was applied to improve normality for linear regression analyses. Sex hormones and related indices, as continuous variables, were analyzed with multivariable linear regression models. Male serum testosterone levels (< 300 ng/dL) were analyzed as a binary variable via logistic regression. Multivariable models included an unadjusted model (Model I) and a fully adjusted model (Model III), which adjusted for age, BMI, race/ethnicity, CVD, DM, CKD, hypertension, alcohol consumption status, smoking status, education level, waist circumference, marital status, family income, and phlebotomy time. Given the debated relationship between BMI and OSA and sex hormones, BMI was treated as a potential effect modifier. Stratified analyses were performed across BMI categories (<25, 25–29.9, ≥30 kg/m2). RCS models explored nonlinear relationships between risk of OSA and sex hormones. Given the unclear associations between OSA and sex hormones across subgroups, subgroup and interaction analyses were conducted by stratifying age, race/ethnicity, smoking status, alcohol consumption status, marital status, income level, phlebotomy time, BMI, and education level, and these analyses evaluated the interaction effects (TT, E2, FAI, SHBG, TT/E2 ratio, TD, FT) across different sex hormone indices in interaction terms between all covariates and OSA risk across these hormonal markers. Mediation analysis evaluated SHBG’s role in the OSA-TT association using non-parametric bootstrap resampling (2000 iterations) to generate confidence intervals and P-values, controlling for the aforementioned confounding factors. Mediation models were fitted using the “mediation” package in R (version 4.3.2), with the same set of covariates included in both the mediator (SHBG) and outcome (TT) models, All paths were adjusted for age, race, education, marital status, BMI, waist circumference, CKD, DM, CVD, hypertension, drinking status, smoking status, time of blood sampling, and PIR. Sensitivity analysis reversed independent (risk of OSA) and dependent (sex hormones) variables to assess robustness of associations. The amount of missing data for all analyzed variables was minimal (< 2%) and was handled by imputation with medians for continuous variables or modes for categorical variables. All statistical analyses were performed using R statistical software (version 4.3.2; R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was defined as a two-sided P-value <.05. Results Population characteristics The study included 2218 participants, with baseline characteristics categorized by risk of OSA tertiles (presented in Table 1). The mean age was 45.95 ± 0.59 years, progressively increasing with higher OSA risk (T3 group: 55.43 ± 0.66 years). The overall mean risk of OSA score was 0.51 ± 0.01. Mean BMI was 28.93 ± 0.25 kg/m2, progressively increasing across risk of OSA tertiles. TT levels averaged 425.39 ± 7.01 ng/dL, with TD prevalence of 26.42%. TT levels (T3: 350.32 ± 7.19 ng/dL), free androgen index (T3: 32.58 ± 1.29), and TT/E2 ratio (T3: 14.34 ± 0.39) decreased significantly with higher risk of OSA. Smoking status showed a strong association with risk of OSA (P < .01): former smokers were overrepresented in the high-risk group (T3), whereas current smokers predominated in low/moderate-risk groups (T1/T2). Alcohol consumption exhibited no clear dose–response relationship, but former drinkers, similar to former smokers, had higher representation in T3. Among obese males (BMI ≥ 30 kg/m2), 64.95% were classified into the high-risk tertile (T3). Education level and phlebotomy time showed no significant association with risk of OSA stratification. SHBG levels did not differ significantly across groups (P = .60), while E2 levels were slightly higher in the high-risk OSA group (P = .02). Compared to the low-risk group, men in the high-risk tertile (T3) were more likely to have BMI ≥30 kg/m2, identify as non-Hispanic White, report former smoking/drinking habits, and belong to higher-income households. Table 1Baseline characteristics of study participants stratified by risk of OSA tertiles. Variable Total T1 T2 T3 P No. pts2218740741737Mean ± SE age (yrs)45.95 ± 0.5934.01 ± 0.6348.54 ± 0.5355.43 ± 0.66< .01Mean ± SE BMI (kg/m2)28.93 ± 0.2524.37 ± 0.1528.59 ± 0.2133.87 ± 0.33< .01Mean ± SE TT (ng/dL)425.39 ± 7.01514.42 ± 10.39410.60 ± 6.60350.32 ± 7.19< .01Mean ± SE E2 (pg/mL)24.72 ± 0.3723.97 ± 0.6024.28 ± 0.4525.91 ± 0.44.02Mean ± SE SHBG (nmol/L)43.87 ± 1.0544.80 ± 1.4043.52 ± 1.0743.28 ± 1.78.60Mean ± SE FAI39.27 ± 0.8245.90 ± 1.1339.26 ± 1.0032.58 ± 1.29< .01Mean ± SE TT/E218.38 ± 0.4222.92 ± 0.6617.84 ± 0.4414.34 ± 0.39< .01% TD26.4212.7022.8043.83Mean ± SE risk of OSA0.51 ± 0.010.22 ± 0.010.54 ± 0.000.78 ± 0.00< .01Mean ± SE energy (kcal)2464.47 ± 33.262555.35 ± 54.402465.70 ± 44.892372.40 ± 42.07.02Race, n (%):< 0.01 Mexican394 (9.93)105 (32.16)136 (34.94)153 (32.91) Non-Hispanic Black425 (9.80)150 (39.82)142 (32.24)133 (27.94) Non-Hispanic White740 (63.64)216 (31.09)242 (32.57)282 (36.34) Other659 (16.62)269 (40.26)221 (34.66)169 (25.08)Smoke, n (%):< 0.01 Former617 (28.43)119 (21.05)211 (32.34)287 (46.60) Now493 (20.48)176 (33.67)185 (39.16)132 (27.17) Never1108 (51.09)445 (40.51)345 (31.13)318 (28.35)Drinking, n (%):0.02 Mild226 (11.76)77 (37.71)74 (31.89)75 (30.40) Moderate921 (45.20)306 (33.10)298 (31.20)317 (35.71) Heavy491 (22.66)173 (33.96)170 (35.22)148 (30.82) Former327 (11.51)68 (21.29)118 (36.46)141 (42.25) Never253 (8.88)116 (45.52)81 (34.85)56 (19.63)BMI, n (%):< 0.01 < 25635 (26.33)445 (72.48)148 (22.25)42 (5.28) 25-30809 (36.68)265 (35.93)356 (41.74)188 (22.33) ≥30774 (37.00)30 (3.57)237 (32.32)507 (64.12)PIR, n (%):< 0.01 < 1.5699 (20.17)259 (41.93)239 (33.96)201 (24.11) 1.5-3802 (32.31)283 (38.16)254 (31.49)265 (30.35)≥3717 (47.52)198 (26.92)248 (33.87)271 (39.21) P-values were calculated using weighted t-tests for continuous variables and weighted chi-square tests for categorical variables. OSA risk tertile ranges: T1, 0 to 0.33; T2, 0.33 to 0.66; T3, 0.66 to 1.Abbreviations: BMI, body mass index; E2, estradiol; FAI, Free Androgen Index; OSA, obstructive sleep apnea; PIR, poverty income ratio; SHBG, sex hormone-binding globulin; TD, testosterone deficiency; TT, total testosterone. Baseline characteristics of study participants stratified by risk of OSA tertiles. P-values were calculated using weighted t-tests for continuous variables and weighted chi-square tests for categorical variables. OSA risk tertile ranges: T1, 0 to 0.33; T2, 0.33 to 0.66; T3, 0.66 to 1. Abbreviations: BMI, body mass index; E2, estradiol; FAI, Free Androgen Index; OSA, obstructive sleep apnea; PIR, poverty income ratio; SHBG, sex hormone-binding globulin; TD, testosterone deficiency; TT, total testosterone. Association between risk of OSA and sex hormones Table 2 displays the associations between risk of OSA and sex hormones in linear regression models. In crude models, all variables except ln(SHBG) showed significant linear relationships with OSA risk (P for trend <0.01). TT, FAI, TT/E2 and FT were inversely associated, whereas E2 and TD were positively associated. After full adjustment, OSA risk remained inversely associated with ln(SHBG), ln(TT), and ln(TT/E2), and positively associated with ln(FAI). For TD, the continuous OSA risk score showed a borderline association (P = 0.04). No significant associations were observed with ln(FT) or ln(E2). When analyzing risk of OSA as tertiles, significant decreasing trends emerged for ln(SHBG) (P for trend <0.01), ln(TT) (P for trend = 0.04) and ln(TT/E2) (P for trend = 0.01). Table 2Linear relationship between risk of OSA and sex hormones. Risk of OSA TT β (95% CI) E 2 β (95% CI) FAI β (95% CI) SHBG β (95% CI) TT/E 2 β (95% CI) TD OR (95% CI) FT β (95% CI) Crude model Continuous -0.74 (-0.81, -0.66) 0.15 (0.05, 0.24) -0.66 (-0.78, -0.53)-0.08(-0.21, 0.05) -0.88 (-0.97, -0.79) 23.15 (10.31, 52.01) -0.78 (-0.88, -0.69)Tertile 1ReferenceReferenceReferenceReferenceReferenceReferenceReferenceTertile 2 -0.22 (-0.26, -0.17)0.03(-0.02, 0.08) -0.17 (-0.23, -0.11)-0.04(-0.12, 0.04) -0.25 (-0.29, -0.20) 1.82 (1.10, 3.00) -0.23 (-0.27, -0.18)Tertile 3 -0.40 (-0.46, -0.35) 0.08 (0.02, 0.15) -0.36 (-0.45, -0.27)-0.04(-0.13, 0.04) -0.49 (-0.55, -0.43) 4.72 (2.91, 7.65) -0.43 (-0.50, -0.35) P for trend < 0.01 0.01 < 0.01 0.29 < 0.01 < 0.01 < 0.01 Adjusted model Continuous -0.23 (-0.40, -0.06)-0.06(-0.25, 0.12) 0.26 (0.02, 0.50) -0.49 (-0.66, -0.31) -0.17 (-0.31, -0.02) 3.85 (1.10, 13.50)0.07(-0.15, 0.28)Tertile 1ReferenceReferenceReferenceReferenceReferenceReferenceReferenceTertile 2 -0.07 (-0.14, 0.00)0(-0.06, 0.06) 0.12 (0.04, 0.19) -0.19 (-0.26, -0.11) -0.07 (-0.12, -0.02)1.08(0.63, 1.83)0.03(-0.05, 0.11)Tertile 3 -0.11 (-0.21, 0.00)-0.01(-0.12, 0.09)0.08(-0.04, 0.21) -0.19 (-0.31, -0.08) -0.1 (-0.17, -0.03)1.60(0.85, 3.02)0.01(-0.11, 0.12) P for trend 0.04 0.800.24 < 0.01 0.01 0.090.97Multivariable linear regression analyses of the association between OSA risk and sex hormone levels. Results are presented as β coefficients with 95% CIs from both unadjusted and fully adjusted models. The fully adjusted model controlled for age, race, education, marital status, BMI, waist circumference, CKD, DM, CVD, hypertension, drinking status, smoking status, time of blood sampling, and PIR.Abbreviations: BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; E2, estradiol; FAI, Free Androgen Index; FT, free testosterone; OSA, obstructive sleep apnea; PIR, poverty income ratio; SHBG, sex hormone-binding globulin; TD, testosterone deficiency; TT, total testosterone. The bold values indicate P < 0.05. Linear relationship between risk of OSA and sex hormones. Multivariable linear regression analyses of the association between OSA risk and sex hormone levels. Results are presented as β coefficients with 95% CIs from both unadjusted and fully adjusted models. The fully adjusted model controlled for age, race, education, marital status, BMI, waist circumference, CKD, DM, CVD, hypertension, drinking status, smoking status, time of blood sampling, and PIR. Abbreviations: BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; E2, estradiol; FAI, Free Androgen Index; FT, free testosterone; OSA, obstructive sleep apnea; PIR, poverty income ratio; SHBG, sex hormone-binding globulin; TD, testosterone deficiency; TT, total testosterone. The bold values indicate P < 0.05. Restricted cubic spline Restricted cubic spline (RCS) models explored nonlinear relationships between OSA risk and sex hormones (Figure 2). Significant nonlinear associations were detected for ln(SHBG) (P-nonlinear <0.0001) and ln(FAI) (P-nonlinear <0.0001), with FAI exhibiting a U-shaped relationship and SHBG showing a negative trend. No significant nonlinearity was observed for ln(TT) (P-nonlinear = 0.8817), ln(E2) (P-nonlinear = 0.9399), TD (P-nonlinear = 0.2637), or ln(TT/E2) (P-nonlinear = 0.9258). Restricted cubic spline models depicting the nonlinear associations between risk of OSA and sex hormone profiles. The figure contains six panels showing relationships with TT, E2, TD, SHBG, FAI and TT/E2 ratio. Solid lines represent estimated effect sizes, and shaded areas indicate 95% CIs. All models were adjusted for age, race, education, marital status, Abbreviations: BMI, waist circumference, CKD, DM, CVD, hypertension, drinking status, smoking status, time of blood sampling, and PIR. BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; E2, estradiol; FAI, Free Androgen Index; OSA, obstructive sleep apnea; PIR, poverty income ratio; SHBG, sex hormone-binding globulin; TD, testosterone deficiency; TT, total testosterone. Subgroup analyses After adjusting for covariates, the associations between sex hormones and risk of OSA differed across BMI groups, though most showed no significant differences. In the normal-weight group (BMI <25 kg/m2), FT exhibited a significant association with risk of OSA. In the overweight group, SHBG showed a significant inverse association (continuous β = -0.47, 95% CI, -0.79 to -0.15), and TD was also significantly associated with OSA risk (continuous OR = 15.68, 95% CI, 1.13–217.97). In the obese group, TT, FT, and TT/E2 showed significant inverse associations with OSA risk, whereas SHBG associations did not reach statistical significance. These results suggest BMI may act as an effect modifier in the relationship between sex hormones and risk of OSA. However, the estimated OR for TD in the normal-weight group (0.07, 95% CI, 0.00–1.00) should be interpreted with caution due to the limited number of TD cases (n = 8) in the highest OSA risk tertile (Table 3). Table 3Association between risk of OSA and sex hormones stratified by BMI. Risk of OSA TT β(95%CI) E 2 β(95%CI) FAI β(95%CI) SHBG β(95%CI) TT/E2 β(95%CI) TD OR(95%CI) FT β(95%CI) < 25 kg/m 2 (n = 635)Continuous0.1(-0.20, 0.40)0.15(-0.09, 0.38) 0.30 (0.01, 0.58)-0.20(-0.66, 0.26)-0.05(-0.39, 0.29)0.07(0.00, 1.00) 0.27 (0.06, 0.49)Tertile 1ReferenceReferenceReferenceReferenceReferenceReferenceReferenceTertile 2-0.04(-0.17, 0.09)0.01(-0.09, 0.10)0.08(-0.02, 0.18)-0.12(-0.25, 0.01)-0.05(-0.17, 0.08) 0.35 (0.14, 0.86)0.05(-0.05, 0.15)Tertile 30.07(-0.14, 0.28)0.11(-0.09, 0.30)0.07(-0.12, 0.27)0.00(-0.24, 0.25)-0.03(-0.23, 0.16)0.27(0.05, 1.42)0.09(-0.10, 0.28)p for trend0.850.350.230.460.56 0.03 0.23 25 - 29.9 kg/m 2 (n = 809)Continuous-0.26(-0.59, 0.06)-0.14(-0.40, 0.12)0.21(-0.16, 0.58) -0.47 (-0.79,-0.15)-0.12(-0.37, 0.13) 15.68 (1.13217.97)0.03(-0.31, 0.37)Tertile 1ReferenceReferenceReferenceReferenceReferenceReferenceReferenceTertile 2-0.05(-0.17, 0.08)0.03(-0.05, 0.11)0.04(-0.08, 0.15)-0.08(-0.21,0.05)-0.07(-0.17, 0.02)1.15(0.49, 2.69)0(-0.11, 0.11)Tertile 3-0.13(-0.29, 0.04)-0.04(-0.16, 0.09)0.07(-0.13, 0.27) -0.20 (-0.39,0.00) -0.09 (-0.21, 0.03)1.98(0.58, 6.76)-0.01(-0.18, 0.16)p for trend0.120.510.48 < 0.05 0.140.230.89 ≥ 30 kg/m 2 (n = 774)Continuous -0.57 (-0.85,-0.29)0.18(-0.11, 0.46)-0.28(-0.63, 0.08)-0.30(-0.66,0.07) -0.75 (-1.06,-0.44)10.04(1.19,84.84) -0.42 (-0.69, -0.15)Tertile 1ReferenceReferenceReferenceReferenceReferenceReferenceReferenceTertile 2-0.13(-0.28, 0.02)0.02(-0.11, 0.14)0.12(-0.03, 0.28) -0.25 (-0.45,-0.06)-0.15(-0.31, 0.02)2.13(0.51, 8.76)0(-0.14, 0.13)Tertile 3 -0.20 (-0.39, -0.01)0.05(-0.10, 0.18)0.07(-0.13, 0.27) -0.27 (-0.55, 0.01) -0.24 (-0.42,-0.06)3.43(0.75,15.80)-0.06(-0.23, 0.10)p for trend 0.04 0.440.720.28 < 0.01 0.03 0.13Stratified analyses of the association between OSA risk and sex hormone profiles by BMI categories. Results are presented as β coefficients with 95% confidence intervals for continuous outcomes and odds ratios with 95% confidence intervals for TD. Models were adjusted for age, race, education, marital status, waist circumference, CKD, DM, CVD, hypertension, drinking status, smoking status, time of blood sampling, and PIR. The bold values indicate P < 0.05.Abbreviations: BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; OSA, obstructive sleep apnea; PIR, poverty income ratio; TD, testosterone deficiency. Association between risk of OSA and sex hormones stratified by BMI. Stratified analyses of the association between OSA risk and sex hormone profiles by BMI categories. Results are presented as β coefficients with 95% confidence intervals for continuous outcomes and odds ratios with 95% confidence intervals for TD. Models were adjusted for age, race, education, marital status, waist circumference, CKD, DM, CVD, hypertension, drinking status, smoking status, time of blood sampling, and PIR. The bold values indicate P < 0.05. Abbreviations: BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; OSA, obstructive sleep apnea; PIR, poverty income ratio; TD, testosterone deficiency. Given the unclear associations between OSA and sex hormones across subgroups, subgroup, and interaction analyses were conducted by stratifying age, race/ethnicity, smoking status, alcohol consumption status, marital status, income, phlebotomy time, BMI, and education level, and applying different sex hormone indices (TT, E2, FAI, SHBG, TT/E2, TD, FT). Results with P for interaction <0.05 are all listed in Figure 3. Significant interactions were observed between race and risk of OSA across multiple sex hormone indices, with non-Hispanic Whites showing distinct differences compared to other ethnic groups. OSA-smoking interactions influenced TT; OSA-alcohol interactions affected E2; Both OSA-BMI and OSA-race interactions modulated SHBG and FAI; OSA-marital status/income/smoking status interactions impacted FT; and OSA-phlebotomy time interactions modified TD associations. Forest plots of subgroup analyses examining the association between OSA risk and sex hormone indicators. Results are presented as β coefficients with 95% CIs for continuous outcomes (TT, E2, FAI, SHBG, TT/E2, FT) and odds ratios with 95% CIs for TD. Interaction analyses were stratified by age, race, smoking status, alcohol consumption, marital status, income, phlebotomy time, Abbreviations: BMI, and education level. BMI, body mass index; E2, estradiol; FAI, Free Androgen Index; FT, free testosterone; OSA, obstructive sleep apnea; SHBG, sex hormone-binding globulin. The bold values indicate P < 0.05. Mediation analysis results Mediation analysis demonstrated a significant mediating role of SHBG in the association between OSA and TT. Specifically, OSA exerted a significant indirect effect on TT through SHBG (β = −160.82, 95% CI, −162.05 to −72.66; P < 0.01). The direct effect of OSA on TT was not statistically significant (β = −2.73, 95% CI, −55.35 to 68.22; P = 0.95). The total effect of OSA on TT was statistically significant (β = −163.55, 95% CI. −188.25 to −32.78; P < 0.01). The proportion of the total effect mediated by SHBG was 0.98 (95% CI, 0.66 to 2.71; P < 0.01). Sensitivity analysis To evaluate the robustness of OSA-sex hormone associations, sensitivity analyses were performed by swapping the independent (X) and dependent (Y) variables. In primary analyses, risk of OSA served as the independent variable and sex hormone levels as the dependent variables. In sensitivity models, sex hormone levels were the independent variables and risk of OSA was the dependent variable. After adjusting for the same covariates, the reversed models confirmed significant associations between sex hormone profiles and risk of OSA (Table 4). Table 4Sensitivity analysis of the linear association between the risk of OSA and sex hormones. Risk of OSA TT β (95% CI) E2 β (95% CI) FAI β (95% CI) SHBG β (95% CI) TT/E2 β (95% CI) TD OR (95% CI) FT β (95%CI) Crude model Continuous -0.21 (-0.24, -0.17) 0.06 (0.02, 0.11) -0.15 (-0.20, -0.11)-0.02(-0.05, 0.01) -0.26 (-0.30, -0.23) 0.16 (0.12, 0.20) -0.23 (-0.28,- 0.17) P < 0.01 0.01 < 0.01 0.18 < 0.01 < 0.01 < 0.01 Adjusted model Continuous -0.02 (-0.03, 0.00)-0.01(-0.03, 0.01) 0.02 (0.00, 0.04) -0.03 (-0.05, -0.02) -0.02 (-0.03, 0.00)0.01(0.00, 0.03)0.01(-0.01, 0.03) P 0.02 0.48 0.03 < 0.01 0.03 0.080.48 < 25 kg/m 2 Crude model Continuous-0.07(-0.14, 0.00)0.00(-0.07, 0.06) -0.16 (-0.20, -0.12) 0.10 (0.05, 0.16) -0.10 (-0.17, -0.02)0.01(-0.10, 0.11) -0.16 (-0.21, -0.12) P 0.060.90 <0.01 < 0.01 0.01 0.90 < 0.01 Adjusted model Continuous0.01(-0.03, 0.05)0.02(-0.01, 0.05) 0.04 (0.00, 0.08)-0.02(-0.07, 0.03)-0.01(-0.06, 0.04)-0.04(-0.09, 0.01) 0.04 (0.01, 0.08) P 0.470.16 0.04 0.370.760.14 0.02 25 - 29.9 kg/m 2 Crude model Continuous -0.09 (-0.12, -0.05)-0.02(-0.08, 0.04) -0.14 (-0.19, -0.09) 0.1 (0.07, 0.13) -0.09 (-0.14, -0.04) 0.08 (0.03, 0.13) -0.15 (-0.20, -0.09) P < 0.01 0.54 < 0.01 < 0.01 < 0.01 < 0.01 < 0.01 Adjusted model Continuous-0.02(-0.06, 0.01)-0.02(-0.05, 0.01)0.02(-0.01, 0.05) -0.05 (-0.08, -0.02)-0.01(-0.05, 0.02) 0.08 (0.03, 0.13)0.00(-0.03, 0.03) P 0.160.270.24 < 0.01 0.35 0.04 0.86 ≥ 30 kg/m 2 Crude model Continuous -0.10 (-0.13, -0.07)0.04(-0.01, 0.08) -0.14 (-0.17, -0.11) 0.06 (0.03, 0.09) -0.13 (-0.16, -0.09) 0.07 (0.04, 0.10) -0.17 (-0.21, -0.13) P < 0.01 0.08 < 0.01 < 0.01 < 0.01 <0.01 < 0.01 Adjusted model Continuous -0.03 (-0.05, -0.01)0.02(-0.01, 0.04)-0.02(-0.04, 0.01)-0.01(-0.03, 0.01) -0.05 (-0.07, -0.03)0.02(0.00, 0.04) -0.03 (-0.06, 0.01) P 0.01 0.210.190.15 < 0.01 0.04 0.01 Sensitivity analysis examining the robustness of associations by reversing independent and dependent variables. Results are presented as β coefficients with 95% CIs for continuous outcomes and odds ratios with 95% CIs for TD. Both unadjusted and fully adjusted models are shown, with the latter controlling for age, race, education, marital status, BMI, waist circumference, CKD, DM, CVD, hypertension, drinking status, smoking status, time of blood sampling, and poverty-to-income ratio. Analyses are further stratified by BMI categories.Abbreviations: BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; OSA, obstructive sleep apnea; TD, testosterone deficiency. Sensitivity analysis of the linear association between the risk of OSA and sex hormones. Sensitivity analysis examining the robustness of associations by reversing independent and dependent variables. Results are presented as β coefficients with 95% CIs for continuous outcomes and odds ratios with 95% CIs for TD. Both unadjusted and fully adjusted models are shown, with the latter controlling for age, race, education, marital status, BMI, waist circumference, CKD, DM, CVD, hypertension, drinking status, smoking status, time of blood sampling, and poverty-to-income ratio. Analyses are further stratified by BMI categories. Abbreviations: BMI, body mass index; CKD, chronic kidney disease; CVD, cardiovascular disease; DM, diabetes mellitus; OSA, obstructive sleep apnea; TD, testosterone deficiency. Discussion Based on the NHANES database, this study represents the first systematic investigation into the association between OSA risk and sex hormone profiles among U.S. adult males within a BMI-stratified framework, while uncovering the pivotal mediating role of SHBG. The analysis has demonstrated significant negative correlations between OSA risk and TT, SHBG, and TT/E2 ratio, alongside a positive correlation with FAI–relationships that persist independently of obesity and age covariates. The association with TD was also observed in the continuous model, though it appeared less robust than that of SHBG or TT after adjustment. Critically, SHBG emerged as a key mediator in the pathway linking OSA to suppressed testosterone levels, indicating that OSA-induced testosterone suppression primarily occurs through inhibition of SHBG synthesis and subsequent impairment of testosterone transport. Furthermore, BMI stratification revealed heterogeneous mechanistic patterns underlying OSA-sex hormone associations across body composition categories. All of this corrects conventional hypotheses that attribute OSA-related hypogonadism solely to HPG axis disruption, providing novel insights into the pathophysiology of endocrine dysfunction in OSA. Specifically, the negative correlation between OSA and SHBG is independent of traditional confounders such as obesity and age, which implies: First, OSA itself can activate systemic inflammatory responses (eg, elevated IL-6 and TNF-α) through chronic hypoxia and inflammation. Inflammatory factors have been confirmed to downregulate hepatocyte nuclear factor 4α, thereby inhibiting SHBG gene transcription.31 In this study, the OSA–SHBG association remained robust after BMI adjustment (P < 0.0001), supporting the direct regulatory effect of hypoxia/inflammation on SHBG. Second, although obesity is also a common risk factor for both OSA and SHBG decline, our research has isolated the influence of obesity on SHBG, demonstrating the independence of OSA's impact on testosterone through the SHBG pathway. The relationship between SHBG and OSA risk is nonlinear, with SHBG levels declining more precipitously at higher OSA risk levels. This accelerated decline may be caused by intensified suppression of SHBG synthesis through inflammatory pathways. This novel mechanism of OSA-related testosterone suppression provides important clues for understanding the metabolic consequences of OSA. BMI stratification revealed distinct pathways: In normal-weight individuals, Continuous hormone outcomes (eg, FT) showed significant associations in this subgroup, suggesting possible HPG axis involvement. In obese individuals, the inverse association between SHBG and OSA risk was not statistically significant; instead, TT, FT, and TT/E2 showed more consistent inverse relationships. This demonstrates that obesity, as a comorbid factor of OSA, may synergistically amplify the metabolic disruption of SHBG homeostasis and potentially affect SHBG levels through complex metabolic pathways. This underscores the need for BMI-specific management: weight loss may mitigate SHBG suppression in obese OSA patients, while non-obese patients may benefit more from OSA-directed therapies to preserve HPG function. Given that obesity-related metabolic disturbances may exacerbate the impact of OSA on SHBG and testosterone levels, future research should prioritize investigating how OSA may further aggravate SHBG decline by influencing obesity-associated metabolic dysregulation. Our findings provide a critical endocrine mechanism to the well-documented association between OSA and male sexual dysfunction, particularly erectile dysfunction (ED).32,33 ED in OSA is not merely a psychogenic byproduct but a multifaceted condition driven by a "multi-hit" pathophysiology.34,35 The primary hit is vascular: CIH and sleep fragmentation induce severe oxidative stress and systemic inflammation, which directly compromise endothelial nitric oxide synthase activity.36,37 This resulting endothelial dysfunction impairs penile hemodynamics, acting as a profound organic driver of ED.38 Concurrently, the second hit is endocrine, characterizing a state of secondary hypogonadism.39,40 Our mediation analysis underscores that CIH-induced metabolic stress strongly suppresses SHBG, exacerbating the decline in TT. This androgen deficiency not only diminishes libido but also further destabilizes the corpus cavernosum's endothelial and smooth muscle integrity, compounding the vascular deficits.41 Furthermore, recent advanced models indicate that these core organic insults are frequently complicated by psychoneurological and systemic factors—such as impaired vigilance, comorbid insomnia, and sleep apnea, and gut dysbiosis—which synergistically worsen erectile function.34,42,43 From a clinical management perspective, this intertwined vascular-endocrine pathology explains why treating OSA with CPAP alone often yields suboptimal ED remission.44 A multimodal clinical pathway is therefore essential. For obese men presenting with OSA and ED, clinicians must perform early screening for SHBG and TT to differentiate between isolated endothelial dysfunction and concurrent HPG axis suppression. Patients with pronounced androgen deficiency and endothelial impairment may require combined interventions—such as CPAP combined with phosphodiesterase type 5 inhibitors for endothelial rescue,45 or carefully monitored testosterone replacement therapy (TRT). However, it is imperative to note that unmonitored TRT may paradoxically exacerbate untreated OSA, highlighting the need for a highly individualized, multidisciplinary approach to male sexual health.6,46 Several limitations should be noted. This cross-sectional design precludes causal inference, and risk of OSA assessment relied on subjective symptom reporting rather than polysomnography. Longitudinal studies tracking SHBG and testosterone pre/post OSA treatment are needed to confirm temporal relationships. The absence of luteinizing hormone/follicle stimulating hormone measurements limits HPG axis evaluation, and racial homogeneity (predominantly non-Hispanic Whites) restricts generalizability. Socioeconomic and genetic modifiers of OSA-endocrine interactions warrant exploration.47 In BMI-stratified analyses, the small number of TD cases within subgroups (eg, only 8 cases in the normal-weight high-risk tertile) led to unstable logistic regression estimates, as reflected by the extremely wide confidence intervals. Therefore, primary conclusions are based on continuous hormone outcomes. For symptomatic OSA patients with hypogonadism (eg, fatigue, reduced libido), routine SHBG monitoring may enable early detection of hormone imbalance. SHBG normalization post-CPAP (if observed) could validate its mediating role. By establishing SHBG as the linchpin connecting OSA and TD, this study redefines OSA-related endocrine dysfunction and informs targeted management strategies. Future research should prioritize mechanistic studies of SHBG regulation and intervention trials assessing OSA treatment’s hormonal benefits. Conclusions Risk of OSA is closely associated with sex hormone levels, with correlations varying substantially by BMI stratification. SHBG primarily mediates the OSA-testosterone relationship, though longitudinal studies are needed to confirm mechanistic pathways. Screening for OSA should prioritize SHBG and testosterone monitoring, particularly in obese populations.