Association between ln(FPG/HDL-C) × BMI and self-reported obstructive sleep apnea

Abstract

Obstructive sleep apnea (OSA) is linked to fasting plasma glucose, high-density lipoprotein cholesterol, and body mass index. However, the association between OSA and the ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index (FHB) index: calculated as ln(FPG/HDL-C) × BMI: has not been extensively investigated in US adults. This research aims to investigate whether the FHB index is associated with the risk of self-reported OSA symptoms among US adults. The analysis utilized National Health and Nutrition Examination Survey datasets collected during the 2005 to 2008 and 2015 to 2018 survey cycles. Weighted multivariable regression and curve fitting were applied to explore the FHB index-OSA correlation, while subgroup and sensitivity analyses were conducted to evaluate the consistency of this relationship. Among 7414 study subjects, 3694 reported symptoms of OSA. Curve fitting showed a nonlinear positive relationship of the FHB index for OSA risk (P for nonlinearity <.001), with an identified threshold of 25.84. Below this value, each 1-unit increase in the FHB index was linked to a 5.2% increase in OSA risk (odds ratio [OR] = 1.052; 95% confidence interval [CI]: 1.009–1.096). Above this value, each 1-unit increase was linked to a 2.6% increase (OR = 1.026; 95% CI: 1.022–1.031). When the FHB index was analyzed as a categorical variable, individuals in the third tertile of the FHB index had odds of OSA that were 211.9% higher than those in the first tertile (OR = 3.119; 95% CI: 2.675–3.635). Subgroup and sensitivity analyses confirmed the robustness of this association. This cross-sectional study revealed a nonlinear positive relationship between the FHB index and self-reported OSA among US adults. The FHB index may serve as a novel and convenient candidate marker for identifying individuals at high-risk of OSA, meriting further investigation. However, before considering clinical application, these findings require validation through longitudinal studies.

1. Introduction

Obstructive sleep apnea (OSA), characterized by recurrent episodes of partial or complete upper airway obstruction during sleep, represents a significant global public health challenge.^[1,2] Epidemiological studies estimate a global OSA burden of around 1 billion individuals, and its prevalence is known to increase with age and be higher in men.^[3,4] OSA is strongly associated with excessive daytime sleepiness, diminished quality of life, and an elevated risk of cardiovascular morbidity and mortality, including hypertension, coronary heart disease, stroke and arrhythmias.^[5–7] Furthermore, OSA imposes substantial economic burdens on healthcare systems due to its chronic nature and associated complications.^[8,9] Screening for OSA in high-risk individuals remains crucial for timely intervention and mitigation of its adverse consequences.

Evidence from the literature suggests that OSA can exacerbate insulin resistance and impair glucose metabolism. Consistent with this pathophysiological link, OSA is also more prevalent among individuals with diabetes compared to nondiabetic populations.^[10] Furthermore, elevated fasting plasma glucose (FPG) is associated with an increased risk of OSA.^[11] In addition, dyslipidemia is associated with OSA, with serum high-density lipoprotein cholesterol (HDL-C) levels being inversely associated with OSA risk.^[11,12] Obesity is strongly linked to an elevated risk of OSA. It is estimated that at least 30% of obese individuals are affected by OSA, and approximately 60% of OSA patients present with obesity.^[12] Body mass index (BMI) serves as a key diagnostic criterion for obesity.^[13] Previous studies have indicated that an increase of 6 units in BMI is linked to a fourfold elevation in the risk of developing OSA.^[14] Most previous research has focused on the association between individual biomarkers: such as FPG, HDL-C, and BMI: and OSA. In contrast, composite indices that integrate both metabolic and anthropometric measures may offer a more reliable and practical tool for assessing OSA risk. FPG–HDL-C–BMI ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index (FHB) index, defined as ln [FPG (mmol/L)/ HDL-C (mmol/L)] × BMI (kg/m²), incorporates FPG, HDL-C, and BMI into a single metric. The natural logarithm transformation was applied to this ratio to normalize its distribution. The multiplicative combination (×) of the metabolic term (ln[FPG/HDL-C]) and BMI was intended to model their potential synergistic contribution to OSA risk, rather than a simple additive effect. By simultaneously reflecting glucose metabolism, lipid profile, and obesity, the FHB index overcomes the limitations associated with relying on individual parameters. However, no research has explored the FHB index-OSA association in the general population.

To bridge this gap, we analyzed the National Health and Nutrition Examination Survey (NHANES) data to assess the FHB index-OSA association in US adults. We hypothesized a positive association between a higher FHB index and OSA risk. Elucidating this relationship may facilitate OSA risk stratification and support the early implementation of targeted interventions.

2. Materials and methods

2.1. Study design and participants

This cross-sectional research utilized 4 NHANES cycles (2005–2006, 2007–2008, 2015–2016, and 2017–2018). This survey, conducted by the National Center for Health Statistics (NCHS), is a national health assessment program for the noninstitutionalized US civilian population.

The initial pooled sample consisted of 39,722 participants. We excluded 17,520 individuals aged under 20 years. After further excluding 8 participants with missing data on OSA, 12,754 participants with missing FHB index data, and 2026 participants with missing data on key covariates or sampling weights, a final analytical sample of 7414 participants was included in the present study (Fig. 1). Data collection was conducted from September 2025 to October 2025. The original NHANES received ethical approval from the NCHS Research Ethics Review Board, and informed consent was secured from all participants. Therefore, this study, which utilized de-identified NHANES data, was exempt from additional review.

Figure 1.

Flowchart of study participant selection. NHANES = National Health and Nutrition Examination survey, OSA = obstructive sleep apnea; FHB = ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index.

2.2. Definition of FHB index

The following equation was used to derive the FHB index:

$${\displaystyle \mathrm{F}\mathrm{H}\mathrm{B}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}=\ln [\mathrm{F}\mathrm{P}\mathrm{G}(\mathrm{m}\mathrm{m}\mathrm{o}\mathrm{l}/\mathrm{L})/\mathrm{H}\mathrm{D}\mathrm{L}\text{-}\mathrm{C}(\mathrm{m}\mathrm{m}\mathrm{o}\mathrm{l}/\mathrm{L})]\times \mathrm{B}\mathrm{M}\mathrm{I}(\mathrm{k}\mathrm{g}/{\mathrm{m}}^2)}$$

2.3. Definition of OSA

According to earlier research,^[7] participants were classified as having symptoms indicative of high-risk for OSA based on 3 specific questions from the NHANES questionnaire (questionnaire item codes: SLQ030, SLQ040, SLQ120/SLD010H/ SLD012). The first question assessed whether snoring occurred at least 3 times weekly over the past year. The second question evaluated if episodes of gasping, snorting, or stopping breathing happened at least 3 times weekly during the same period. The third question evaluated whether, although obtaining ≥7 hours of sleep each night, an individual had 16 to 30 episodes of excessive daytime sleepiness within a month. Any positive response among the questions was interpreted as indicative of OSA. It should be noted that this is not equivalent to a polysomnography-based diagnosis.

2.4. Covariates

According to established literature^[15–18] and clinical expertise, covariates incorporated in the analysis comprised age, sex, race, level of education, marital status, the poverty income ratio, alcohol use, history of hypertension, coronary heart disease and diabetes; physical activity; and total sleep time. Smoking status was determined according to the number of cigarettes smoked (lifetime) and current use. Participants reporting <100 cigarettes were never smokers; those with ≥100 cigarettes were further categorized as former or current smokers.^[19] Alcohol drinkers were defined as those consuming ≥12 standard alcoholic beverages per year; all others were classified as nondrinkers.^[20] Hypertension was identified based on self-report of a prior diagnosis by a healthcare provider. Participants with ≥2 clinical diagnoses of hypertension were classified as having hypertension; all others were considered non-hypertensive.^[21] Diabetes and coronary heart disease status were ascertained through self-reported history of previous clinical diagnosis.^[22,23] Physical activity levels were classified as sedentary, moderate, or vigorous based on self-reported bouts of at least 10 minutes during the preceding 30 days. Moderate-intensity activity referred to exertion resulting in light sweating or modest cardiorespiratory elevation. Vigorous activity required intensity sufficient to cause heavy sweating or a substantial increase above the moderate benchmark.^[21] Sleep duration was classified into 3 groups: short (<7 hours/night), recommended (7 to 9 hours/night), and long (>9 hours/nigh).^[24]

2.5. Statistical analysis

In accordance with the NHANES complex sampling design, all analyses applied the fasting subsample 2-year MEC weight (WTSAF2YR), with sampling weights calculated as 1/4 × WTSAF2YR. Descriptive statistics are reported as unweighted counts (weighted percentages) for categorical variables and mean ± standard error for continuous variables. Differences across groups were assessed using ANOVA/Kruskal–Wallis tests (continuous variables, per normality) and χ² tests (categorical variables). Multicollinearity was ruled out (all variance inflation factor <5), and complete-case analysis addressed missing data. The fixed nature of the NHANES dataset precluded prospective power calculation.

The association between the FHB index and OSA was evaluated using a series of multivariable logistic regression models with progressive adjustment for covariates. An unadjusted model was initially constructed, followed by 3 adjusted models: Model 1 included age, sex, and race; Model 2 further adjusted for level of education, marital status, poverty income ratio, smoking status, alcohol use, physical activity, and sleep duration; Model 3 additionally incorporated hypertension, coronary heart disease, and diabetes. Data are reported as odds ratio (OR) with 95% confidence interval (CI).

To examine potential nonlinearity, restricted cubic splines (RCS) with 3 knots (placed at the 10th, 50th, and 90th percentiles) were incorporated into the logistic regression models. The presence of a nonlinear association was formally tested using a likelihood ratio test, comparing the model with the RCS terms to a model with only a linear term. If the nonlinear test was significant (P < .05), the inflection point was identified as the value of the FHB index corresponding to the point of maximum absolute slope on the fitted spline curve. Subgroup analyses were performed by age (<65 vs ≥65 years), sex, race/ethnicity, smoking status, alcohol consumption, hypertension, and coronary heart disease. Interaction effects were examined using likelihood ratio tests. To assess robustness, we conducted sensitivity analyses, including the exclusion of participants aged >70 years. Additionally, multiple imputation was applied to address missing covariate data as a further robustness check. All analyses were carried out with R version 4.2.2 and Free Statistics version 2.1.1. Statistical significance was set at a 2-sided P value < .05.

3. Results

3.1. Characteristics of participant

This study ultimately included 7414 participants aged 20 years or older, with an OSA prevalence of 49.23%. The high prevalence is consistent with a broad, symptom-based case definition designed for sensitivity in a population survey. Table 1 presents the demographic characteristics stratified by tertiles of the FHB index. Compared with those in the lowest tertile, participants in the highest FHB index tertile exhibited a higher prevalence of hypertension, coronary heart disease, diabetes, and OSA, along with higher BMI and FPG levels. Conversely, those in the highest tertile showed lower poverty income ratio and HDL-C levels. The prevalence of OSA progressively increased from tertile 1 to tertile 3 (T1, 34.62%; T2, 49.53%; T3, 65.57%).

Table 1.

Baseline characteristics of the participants by FHB index tertiles.

Variables	Tertiles of FHB index				P value
	Overall	T1 (−4.53–34.45)	T2 (34.45–49.25)	T3 (49.25–210.76)
Participants (n)	7414	2471	2471	2472
Age (yr), mean (SE)	47.03 (0.37)	45.32 (0.51)	47.23 (0.47)	48.78 (0.47)	<.001
Sex, male, n (%)	3642 (48.63)	893 (34.43)	1344 (54.87)	1405 (58.21)	<.001
Race/Ethnicity, n (%)					<.001
Non-Hispanic white	3244 (68.86)	1161 (71.83)	1057 (67.82)	1026 (66.57)
Non-Hispanic black	1498 (10.49)	519 (10.51)	468 (10.08)	511 (10.89)
Mexican American	1243 (8.22)	297 (5.52)	459 (9.17)	487 (10.30)
Other Race/Ethnicity	1429 (12.44)	494 (12.14)	487 (12.94)	448 (12.25)
Education level, n (%)					<.001
<high school	728 (4.88)	163 (3.27)	278 (5.77)	287 (5.76)
High school or equivalent	2795 (34.75)	874 (31.01)	910 (34.31)	1011 (39.48)
College or above	3891 (60.37)	1434 (65.71)	1283 (59.91)	1174 (54.75)
Marital status, n (%)					.006
Never married	1207 (17.02)	511 (19.94)	368 (16.51)	328 (14.24)
Married or living with a partner	4576 (64.58)	1419 (61.32)	1579 (65.21)	1578 (67.63)
Widowed or divorced or separated	1631 (18.40)	541 (18.74)	524 (18.28)	566 (18.13)
Poverty income ratio, mean (SE)	3.07 (0.04)	3.20 (0.05)	3.09 (0.05)	2.90 (0.06)	<.001
Body mass index (kg/m²), mean (SE)	29.12 (0.14)	23.72 (0.09)	28.64 (0.09)	35.80 (0.18)	<.001
Smoking status, n (%)					<.001
Never	3998 (53.19)	1421 (56.41)	1332 (52.22)	1245 (50.53)
Former	1939 (26.31)	532 (22.17)	660 (27.29)	747 (30.01)
Current	1477 (20.50)	518 (21.42)	479 (20.49)	480 (19.46)
Alcohol consumption, n (%)	5570 (80.04)	1856 (81.01)	1858 (79.68)	1856 (79.31)	.480
Hypertension, n (%)	2218 (25.55)	490 (16.39)	699 (23.82)	1029 (37.83)	<.001
Coronary heart disease, n (%)	319 (3.57)	56 (2.16)	106 (3.39)	157 (5.38)	<.001
Diabetes, n (%)	1004 (9.36)	74 (1.51)	220 (5.45)	710 (22.45)	<.001
Physical activity, n (%)					<.001
Sedentary	3648 (41.95)	1047 (33.65)	1193 (41.94)	1408 (51.42)
Moderate	1940 (28.45)	659 (28.38)	647 (28.58)	634 (28.39)
Vigorous	1826 (29.60)	765 (37.97)	631 (29.48)	430 (20.19)
Sleep duration, n (%)					.005
<7 h	2339 (29.08)	699 (25.40)	788 (30.50)	852 (31.78)
(7, 9)	4572 (64.99)	1589 (68.29)	1537 (64.16)	1446 (62.11)
>9 h	503 (5.93)	183 (6.31)	146 (5.35)	174 (6.11)
Survey cycle, n (%)					<.001
2005–2006	1757 (24.53)	682 (27.31)	590 (24.93)	485 (20.93)
2007–2008	1996 (24.15)	639 (23.85)	722 (26.91)	635 (21.56)
2015–2016	1842 (25.48)	616 (25.31)	559 (22.84)	667 (28.47)
2017–2018	1819 (25.84)	534 (23.52)	600 (25.32)	685 (29.05)
Fasting plasma glucose (mmol/L), mean (SE)	5.95 (0.03)	5.30 (0.02)	5.68 (0.02)	6.98 (0.06)	<.001
High-density lipoprotein cholesterol, (mmol/L), mean (SE)	1.42 (0.01)	1.78 (0.01)	1.32 (0.01)	1.10 (0.01)	<.001
Obstructive sleep apnea, n (%)	3694 (49.23)	871 (34.62)	1249 (49.53)	1574 (65.57)	<.001

Categorical variables were reported as unweighted counts (weighted %), and continuous variables as mean ± SE.

FHB, ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index; T = tertile, SE = standard error.

3.2. Association between FHB index and OSA

To investigate the FHB index-OSA association, we developed 4 multivariate regression models (Table 2). A significant positive association was observed between the FHB index and OSA in the unadjusted model (OR: 1.031, 95% CI: 1.026–1.035). In the fully adjusted Model 3, each 1-unit increment in the FHB index remained significantly associated with a 2.8% higher risk of OSA (OR: 1.028, 95% CI: 1.024–1.033).

Table 2.

Association of FHB index and obstructive sleep apnea.

Variables	n	Crude OR (95%CI)	P value	Model 1 OR (95%CI)	P value	Model 2 OR (95%CI)	P value	Model 3 OR (95%CI)	P value
FHB index	7414	1.031 (1.026–1.035)	<.001	1.028 (1.023–1.032)	<.001	1.027 (1.023–1.031)	<.001	1.028 (1.024–1.033)	<.001
FHB index tertiles
T1 (−4.53–34.45)	2471	1 (Reference)		1 (Reference)		1 (Reference)		1 (Reference)
T2 (34.45–49.25)	2471	1.853 (1.562–2.198)	<.001	1.683 (1.418–1.997)	<.001	1.658 (1.408–1.953)	<.001	1.650 (1.402–1.943)	<.001
T3 (49.25–210.76)	2472	3.596 (3.085–4.192)	<.001	3.200 (2.745–3.731)	<.001	3.102 (2.666–3.610)	<.001	3.119 (2.675–3.635)	<.001
P for trend			<.001		<.001		<.001		<.001

Crude: unadjusted model.

Model 1: adjusted for age, sex, race.

Model 2: adjusted for Model 1 and education level, marital status, poverty income ratio, smoking status, alcohol consumption, physical activity and sleep duration.

Model 3: adjusted for Model 2 and hypertension, coronary heart disease and diabetes.

FHB = ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index; T = tertile.

When the FHB index was categorized into tertiles, the fully adjusted model (Model 3) showed that compared to the lowest tertile, subjects in the second tertile had 65.0% higher odds of OSA (OR: 1.650, 95% CI: 1.402–1.943), while those in the highest tertile had 211.9% higher odds (OR: 3.119, 95% CI: 2.675–3.635); both associations were statistically significant.

To further investigate the relationship, RCS analysis was performed using the covariates from Model 3, revealing a marked nonlinear association between the FHB index and OSA (P for nonlinearity < .001; Fig. 2). Consistent with the nonlinear relationship identified by the RCS analysis (inflection point at FHB index = 25.84), we performed a piecewise linear regression. The association was positive on both sides of the threshold but differed in magnitude. Below the inflection point, each 1-unit increase in the FHB index was associated with an OR of 1.052 (95% CI: 1.009–1.096). Above the inflection point, the corresponding OR was 1.026 (95% CI: 1.022–1.031) per 1-unit increment (Table 3).

Figure 2.

Dose–response relationship of the FHB index with obstructive sleep apnea. The odds ratio is shown by the solid orange line, with the surrounding pink band representing the 95% confidence interval. Covariate adjustment matched Model 3 specifications. Data are presented for 99% of the participants. OSA = obstructive sleep apnea, FHB = ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index.

Table 3.

Threshold effect analysis of FHB index with obstructive sleep apnea.

FHB index	Adjusted model
	OR (95%CI)	P value
<25.84	1.052 (1.009–1.096)	.019
≥25.84	1.026 (1.022–1.031)	<.001

Covariate adjustment matched Model 3 specifications. Data are presented for 99% of the participants.

FHB, ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index.

3.3. Subgroup analyses

Subgroup analyses assessed the FHB index-OSA association in various demographic and clinical characteristics (Fig. 3). Notable interactions were identified for sex (P for interaction = .045) and alcohol consumption (P for interaction = .035). The link was slightly stronger in females (OR: 1.030, 95% CI: 1.023–1.037) than in males (OR: 1.024, 95% CI: 1.018–1.031), and more pronounced among alcohol consumers (OR: 1.030, 95% CI: 1.026–1.035) than nonconsumers (OR: 1.019, 95% CI: 1.010–1.028). No notable interactions were found for other variables, and the positive relationship was robust within all subgroups (OR range: 1.016–1.030). The forest plot visually demonstrates that all effect estimates and their 95% CI exceed 1.0, further supporting the robust positive relationship between the FHB index and OSA risk.

Figure 3.

Subgroup analyses of FHB index and obstructive sleep apnea. Odds ratio and their 95% confidence interval are depicted by green dots and blue horizontal lines, respectively. Diamonds illustrate summary estimates (center line: odds ratio; width: 95% confidence interval). Covariate adjustment matched Model 3 specifications, while avoiding adjustment for the stratification variable itself. CI = confidence interval, FHB = ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index, OR = odds ratio

3.4. Sensitivity analysis

The robustness of the primary finding was assessed through sensitivity analyses. After multiple imputation of missing covariates (Table 4), the FHB index-OSA association persisted (OR: 1.028, 95% CI: 1.024–1.032). This result was corroborated by an analysis excluding individuals over 70 years of age (OR: 1.029, 95% CI: 1.024–1.034, Table 5). After adjusting for all covariates including the survey cycle, the weighted multivariate logistic regression analysis was performed, and the results are presented in Table 6. The OR for OSA risk was 1.029 (95% CI: 1.024–1.033). The remarkable consistency of these effect sizes with the primary result (OR: 1.028; 95% CI: 1.024–1.033) reinforces the reliability of our main conclusion.

Table 4.

Sensitivity analysis after multiple imputation of missing covariate data.

Variables	n	Crude OR (95%CI)	P value	Model 4 OR (95%CI)	P value
FHB index	8702	1.030 (1.026–1.034)	<.001	1.028 (1.024–1.032)	<.001
FHB index tertiles
T1 (-4.53–34.51)	2901	1 (Reference)		1 (Reference)
T2 (34.52–49.18)	2900	1.818 (1.552–2.129)	<.001	1.627 (1.395–1.898)	<.001
T3 (49.19–210.76)	2901	3.441 (2.959–4.001)	<.001	2.998 (2.581–3.483)	<.001
P for trend			<.001		<.001

FHB, ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index; T, tertile.

Crude: unadjusted model.

Model 4: Covariate adjustment was identical to that of Model 3.

Table 5.

Sensitivity analysis among participants aged ≤70 years.

Variables	n	Crude OR (95%CI)	P value	Model 5 OR (95%CI)	P value
FHB index	6291	1.032 (1.027–1.037)	<.001	1.029 (1.024–1.034)	<.001
FHB index tertiles
T1 (-4.53–34.45)	2097	1 (Reference)		1 (Reference)
T2 (34.46–49.49)	2097	1.910 (1.609–2.267)	<.001	1.664 (1.406–1.969)	<.001
T3 (49.49–210.76)	2097	3.840 (3.256–4.528)	<.001	3.239 (2.727–3.847)	<.001
P for trend			<.001		<.001

Crude: unadjusted model.

Model 5: Covariate adjustment was identical to that of Model 3.

FHB, ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index, T = tertile.

Table 6.

Sensitivity analysis with survey cycle as an additional covariate.

Variables	n	Crude OR (95%CI)	P value	Model 6 OR (95%CI)	P value
FHB index	7414	1.031 (1.026–1.035)	<.001	1.029 (1.024–1.033)	<.001
FHB index tertiles
T1 (−4.53–34.45)	2471	1 (Reference)		1 (Reference)
T2 (34.45–49.25)	2471	1.853 (1.562–2.198)	<.001	1.654 (1.405–1.947)	<.001
T3 (49.25–210.76)	2472	3.596 (3.085–4.192)	<.001	3.168 (2.718–3.693)	<.001
P for trend			<.001		<.001

Crude: unadjusted model.

Model 6: adjusted for Model 3 and survey cycle.

FHB, ln[(fasting plasma glucose)/(high-density lipoprotein cholesterol)] × body mass index, T = tertile.

4. Discussion

This cross-sectional study of US adults identified a nonlinear positive association between the FHB index and OSA risk, with an inflection point at 25.84. Both subgroup and sensitivity analyses confirmed the robustness of this relationship.

Our findings are consistent with and extend the existing literature on the association between the individual components of the FHB index and OSA. Previous studies have indicated that elevated FPG is associated with an increased risk and severity of OSA.^[25,26] While low HDL-C levels are a common feature in patients with OSA and have been linked to its disease severity,^[12,27,28] and elevated BMI is a well-established major risk factor for the condition.^[29] In line with these observations, our cross-sectional analysis also demonstrated significant positive associations between the FHB index and high-risk OSA status. The novelty of our study lies in the integration of 3 critical metabolic and anthropometric parameters into a single composite index: the FHB index. While previous studies have explored other combined metrics, such as the triglyceride-glucose (TyG) index, in relation to OSA,^[30] the FHB index is unique in its inclusion of HDL-C, a protective lipid marker known to be inversely associated with metabolic dysfunction and cardiovascular risk.^[31] Even after adjusting for established comorbidities including diabetes, the FHB index maintained a strong and independent association with high-risk OSA status. This indicates that the FHB index captures dimensions of metabolic dysregulation relevant to OSA pathogenesis that are not fully reflected by a diabetes diagnosis alone. The choice of HDL-C in constructing the FHB index may offer practical clinical value. HDL-C exhibits high biological stability with low intraindividual variability, enhancing the index’s reliability. Moreover, low HDL-C is a central feature of metabolic syndrome and a robust marker of insulin resistance and cardiovascular risk, all of which share overlapping pathophysiology with OSA. By incorporating HDL-C, the FHB index capitalizes on its dual role in metabolic and vascular health, potentially offering a more integrated risk assessment. While triglycerides (as captured in indices like TyG) are significant, the focus on HDL-C provides a complementary emphasis on protective lipid deficiency, which may be especially pertinent in metabolically dysregulated OSA populations. Moreover, the use of a large, well-characterized, population-based dataset from NHANES strengthens the generalizability and robustness of our findings. However, it is important to emphasize that due to the cross-sectional design, the associations observed in this study do not imply causality, and we cannot determine the temporal sequence between the FHB index and OSA status.

The components of the FHB index are each linked to OSA through distinct yet interrelated biological pathways, as proposed in the existing literature. Glucose metabolic dysfunction predisposes may predispose the pharyngeal muscles to impaired function and collapse, potentially contributing to airway obstruction and resultant hypoventilation.^[32,33] Similarly, low HDL-C, characterized by its impaired atheroprotective and anti-inflammatory functions, is linked to endothelial dysfunction and elevated oxidative stress. This pathophysiological milieu may adversely affect the neural and muscular control of the upper airway, which in turn may elevate the risk for OSA.^[12] Furthermore, an elevated BMI, acting as a surrogate for increased adiposity, is thought to promote OSA through both mechanical and metabolic pathways. Mechanically, excess peri-pharyngeal fat deposition can directly narrow and increase the collapsibility of the upper airway.^[34] Metabolically, the associated adipose tissue dysfunction may exacerbate systemic inflammation and insulin resistance, which might further impair neuromuscular control of airway patency.^[35] These pathways collectively contribute to the development of OSA. It is important to note that these are hypothesized pathways derived from the established literature, which provide a plausible biological framework for our epidemiological association. Furthermore, the relationship is likely bidirectional. The intermittent hypoxia characteristic of OSA can also directly trigger disturbances in glucose and lipid metabolism,^[36] potentially creating a vicious cycle that further links metabolic dysregulation with OSA progression.

In subgroup analyses, we observed potential effect modification by sex and alcohol consumption (P for interaction = .045 and .035, respectively). Although a significant positive association between the FHB index and OSA risk was consistent across all subgroups, point estimates suggested that the strength of association per 1-unit increase in the FHB index was slightly higher in females (OR: 1.030, 95% CI: 1.023–1.037) than in males (OR: 1.024, 95% CI: 1.018–1.031), and also higher among alcohol drinkers (OR: 1.030, 95% CI: 1.026–1.035) than among nondrinkers (OR: 1.019, 95% CI: 1.010–1.028). These findings require validation in future prospective studies. The nonlinear association between the FHB index and OSA risk, characterized by an inflection point at 25.84, offers a clinically interpretable framework for risk stratification. Below this threshold, each 1-unit increment in the FHB index was associated with an OR of 1.052 (95% CI: 1.009–1.096), reflecting a relatively steep increase in OSA risk: suggesting that this range may represent a window of opportunity for early metabolic intervention. Above the threshold, the slope attenuated to an OR of 1.026 (95% CI: 1.022–1.031) per unit increase, indicating that although the risk remains elevated, its rate of increase slows, which may prompt comprehensive clinical evaluation for OSA. Nevertheless, external validation in independent and diverse populations is essential before this threshold can be translated into routine clinical practice.

Our study has several limitations. First, the cross-sectional design of NHANES precludes causal inference and temporality determination. Reverse causation cannot be ruled out: while metabolic abnormalities may increase OSA susceptibility, OSA-related intermittent hypoxia and sleep fragmentation may likewise promote metabolic dysfunction, including insulin resistance, dyslipidemia, and weight gain. Thus, the observed association may reflect a bidirectional relationship. Longitudinal studies with repeated metabolic measurements and objective OSA assessment are needed to establish temporality and causality. Second, OSA diagnosis in NHANES relied on self-reported predetermined questionnaire and related symptoms rather than the gold standard polysomnography. Although this approach is practical for large-scale surveys and captures clinically significant cases, it may lead to overestimation of the true prevalence of OSA, particularly mild forms, and making our effect estimates conservative. Future validation of our findings in cohorts with polysomnographic data is therefore warranted. Third, despite extensive adjustment for known confounders, residual confounding from unmeasured variables (such as diet, medication use, and fat distribution) cannot be excluded. Fourth, the present study did not include a comparison with other relevant metabolic indices, such as the triglyceride-to-high-density lipoprotein cholesterol (TG/HDL-C) ratio or TyG index.^[37] Future research in multi-center cohorts or longitudinal studies is warranted to systematically evaluate and compare the clinical utility of various metabolic indices for OSA risk assessment. Fifth, numerous studies have confirmed the existence of seasonal variations in lipid profiles,^[38–40] including HDL-C. These fluctuations may originate from alterations in circadian light exposure patterns.^[41,42] Although we adjusted for the survey cycle in our analysis, NHANES does not provide data on the specific month of examination, daily light duration, or individual light exposure. Therefore, we were unable to fully control for potential residual confounding arising from differences in latitude and light exposure. Sixth, evidence suggests that light exposure influences glucose levels^[43,44] and BMI^[45,46] through temperature-modulated pathways. However, because NHANES does not provide data on environmental variables such as season or geographic latitude, potential residual confounding effects arising from these factors cannot be fully ruled out. Seventh, the exclusion of a large number of participants due to missing data could introduce bias. To minimize this potential bias and retain the full analytical sample, we handled missing covariate data using multiple imputation, followed by sensitivity analyses. This method operates under the missing at random assumption. Despite its robustness, bias could remain if this assumption is untenable. Finally, although NHANES provides strong population representativeness, residual selection bias cannot be fully excluded; therefore, our findings should be generalized with caution.

5. Conclusions

This cross-sectional study revealed a nonlinear positive relationship between the FHB index and self-reported OSA among US adults. The FHB index may serve as a novel and convenient candidate marker for identifying individuals at high risk of OSA, meriting further investigation. However, before considering clinical application, these findings require validation through longitudinal studies.

Acknowledgments

We thank the NCHS and all NHANES participants for making this research possible.

Author contributions

Conceptualization: Yunzhou Zheng, Shanshan Zhang.

Data curation: Yunzhou Zheng, Shanshan Zhang.

Formal analysis: Yunzhou Zheng, Shanshan Zhang.

Funding acquisition: Yunzhou Zheng.

Methodology: Yunzhou Zheng, Shanshan Zhang.

Supervision: Yunzhou Zheng.

Validation: Yunzhou Zheng, Shanshan Zhang.

Investigation: Shanshan Zhang.

Project administration: Shanshan Zhang.

Resources: Shanshan Zhang.

Software: Shanshan Zhang.

Writing – original draft: Yunzhou Zheng, Shanshan Zhang.

Writing – review & editing: Yunzhou Zheng, Shanshan Zhang.