ABSTRACT
Background
Inflammation-related mechanisms may be important in the development of obstructive sleep apnea (OSA), and diet plays a crucial role in modulating inflammation. Current epidemiologic evidence for the associations between dietary patterns and OSA risk is limited to cross-sectional studies.
Objectives
We investigated prospectively the associations of overall diet quality and proinflammatory diet with OSA risk.
Methods
We followed 145,801 participants in the Nurses’ Health Study (NHS) (2002–2012), NHS II (1995–2013), and Health Professionals Follow-up Study (1996–2012). Alternative Healthy Eating Index 2010 (AHEI) and Empirical Dietary Inflammatory Pattern (EDIP) scores were calculated based on validated FFQs administered every 4 y. Cox models were used to estimate HRs and 95% CIs.
Results
We documented 8856 incident OSA cases during follow-up. In pooled analyses adjusted for potential confounders, higher diet quality (higher AHEI scores) was associated with lower OSA risk (HR comparing the highest with the lowest quintile of AHEI: 0.76; 95% CI: 0.71, 0.82; P-trend < 0.001), and higher dietary inflammatory potential (higher EDIP scores) was associated with significantly increased risk (HR comparing the highest with the lowest quintile of EDIP: 1.94; 95% CI: 1.81, 2.08; P-trend < 0.001). Additional adjustment for metabolic factors attenuated both associations. The association with AHEI score was no longer statistically significant (comparable HR: 0.98; 95% CI: 0.91, 1.05; P-trend = 0.54), whereas the association with EDIP score remained statistically significant (comparable HR: 1.31; 95% CI: 1.22, 1.41; P-trend < 0.001).
Conclusions
A healthier diet, particularly one with anti-inflammatory potential, was associated with lower OSA risk.
Keywords: diet quality, proinflammatory diet, obstructive sleep apnea, Alternative Healthy Eating Index, Empirical Dietary Inflammatory Pattern, prospective cohort study
Introduction
Obstructive sleep apnea (OSA) is a common sleep disorder caused by repeated episodes of partial or complete obstruction of the upper airway during sleep, affecting 10%–20% of the US adult population and leading to increased morbidity and mortality (1–6). A growing number of prospective studies have linked poorer diet quality with worse pulmonary functions and increased risk of chronic health conditions comorbid with OSA (e.g., obesity, diabetes, cardiovascular disease, and chronic obstructive pulmonary disease) (7–10). Another line of evidence suggests a close relation of an unhealthy diet with shorter sleep duration and poorer sleep quality (11, 12), which are common among individuals with OSA. Therefore, it is highly plausible that the risk of OSA may also be modulated by diet. However, current epidemiologic evidence for the potential role of diet in OSA development remains limited. Although 4 cross-sectional studies reported consistent associations of lower diet quality with higher OSA prevalence (13–16), these results may reflect bidirectional relations and prospective studies are needed to clarify the directionality of the associations (17).
Diet also modulates inflammation, and higher diet quality has been associated with favorable inflammation profiles (18–21). Further, the dietary inflammatory index, which was developed to specifically assess the inflammatory potential of diet (22), has been robustly associated with increased risk of several inflammation-related diseases including diabetes (23), cardiovascular disease (24), and colorectal cancer (25). Given that elevated inflammation has been implicated in the pathogenesis of OSA (26–28), the inflammatory dietary pattern may be particularly relevant for OSA risk. There are initial clues from cross-sectional studies suggesting adverse associations between the inflammatory potential of diet and several OSA-related parameters [e.g., Apnea-Hypopnea Index (AHI) and daytime sleepiness] (16, 29).
We investigated the associations of overall diet quality and proinflammatory diet with OSA risk in the Nurses’ Health Study (NHS), NHS II, and Health Professionals Follow-up Study (HPFS). We hypothesized that higher overall diet quality and lower dietary inflammatory potential may be associated with lower OSA risk, independently of confounders and major OSA predictors.
Methods
Study population
The NHS, NHS II, and HPFS are ongoing US prospective cohort studies, the details of which have been elaborated elsewhere (30–32). In brief, the NHS was initiated in 1976, enrolling 121,700 female nurses between 30 and 55 y of age (30), and NHS II in 1989, enrolling 116,429 female nurses aged 25–42 y (31). The HPFS began in 1986, when 51,529 male health professionals aged between 40 and 75 y were included (32). Demographics and disease history were collected at cohort enrollment. Through biennial follow-up using self-administered questionnaires, information on anthropometrics, lifestyle characteristics, diet, menstrual and reproductive history (females only), and medical history were assessed, with response rates exceeding 90% of eligible person-time. The study protocols were approved by the Institutional Review Boards of the Brigham and Women's Hospital and Harvard TH Chan School of Public Health. We excluded individuals who had no dietary information at the predefined analytic baseline (NHS: 2002, NHS II: 1995, and HPFS: 1996), reported implausible total energy intake (<500 or >3500 kcal/d for females and <800 or >4200 kcal/d for males), did not answer the question on sleep apnea diagnosis, or reported sleep apnea diagnosis before the analytic baseline, leaving 57,391 participants in the NHS, 69,180 in the NHS II, and 19,536 in the HPFS for analysis (Supplemental Figure 1).
Dietary assessment
A semiquantitative FFQ was first administered in the NHS in 1980, with expanded 131-item FFQs in 1984, 1986, and every 4 y thereafter. Starting from 1991 in the NHS II and 1986 in the HPFS, the same 131-item FFQs were administered quadrennially. Participants were queried about their consumption frequency of each food during the previous year based on a specified standardized portion size or a commonly used unit, and responses included 9 categories ranging from “never or <1 time per month” to “≥6 times per day.” The validity and reproducibility of the FFQs in our cohorts have been comprehensively reported (33–35). We estimated the average daily intake of nutrients by multiplying the consumption frequency of each food by the corresponding nutrient content (based primarily on the USDA Nutrient Database, and supplemented by other published sources). We further calculated energy-adjusted, cumulative average dietary intakes from repeated FFQs to capture long-term exposure and reduce random within-subject variation (36, 37).
The Alternative Healthy Eating Index 2010 (AHEI) was derived from 11 foods and nutrients predictive of chronic disease risk based on existing clinical and epidemiologic investigations (38), including information from the original Alternate Healthy Eating Index (39). The details of variable selection and scoring criteria for the AHEI have been elaborated elsewhere (38). Each component was scored from 0 (worst) to 10 (best), with a total score ranging from 0 (nonadherence to healthy eating) to 110 (perfect adherence to healthy eating). Components that receive a lower score with higher intake include sugar-sweetened beverages (SSBs), red and processed red meat, trans fat, sodium, and alcohol; components that receive a higher score with higher intake include fruits, vegetables, whole grains, nuts and legumes, PUFAs, and omega (ω)-3 fatty acids (38).
The Empirical Dietary Inflammatory Pattern (EDIP) score was derived according to food groups’ ability to predict plasma concentrations of 3 inflammatory markers (IL-6, CRP, and TNF α receptor 2) to assess the overall inflammatory potential of diet (18, 22). Using reduced rank regression (RRR; to identify an RRR dietary pattern based on linear functions of food groups simultaneously maximizing the explained variation in inflammatory markers) and stepwise linear regression (to identify the most important food group components contributing to the RRR dietary pattern), a data-driven index of dietary inflammatory potential was generated based on the weighted sum of 18 food groups, with higher scores indicating higher inflammatory potential (18, 22). Dietary components that contribute to higher EDIP scores with higher intake (proinflammatory) include processed meat, red meat, organ meat, fish other than dark-meat fish, vegetables other than green leafy or dark yellow vegetables, refined grains, high-energy beverages, low-energy beverages, and tomatoes. Components that contribute to lower scores with higher intake (anti-inflammatory) include beer, wine, tea, coffee, dark yellow vegetables, green leafy vegetables, snacks, fruit juice, and pizza. This EDIP score has been widely used in multiple cohorts with robust associations with risk of cardiometabolic disease and cancer (24, 25, 40, 41).
The AHEI and EDIP were moderately correlated (Spearman correlation = −0.34 in the pooled baseline sample) (Supplemental Table 1a). Supplemental Table 1b presents correlations of individual food/nutrient groups with the AHEI and EDIP. The scoring criteria for both indexes have previously been reported (18, 22, 38).
Assessment of OSA and related symptoms
On the 2012 NHS questionnaire, participants reported whether they had sleep apnea diagnosed by a clinician or sleep study (yes or no), and (if yes) the year of first diagnosis (response categories: before 2002, 2002–2005, 2006–2007, 2008–2009, 2010–2011, and after 2012) (42–44). Therefore, the analytic baseline was 2002 in the NHS. In 2013 (NHS II) and 2012 (HPFS), clinical diagnoses of sleep apnea were assessed using a similar approach, with the analytic baseline determined as 1995 and 1996, respectively (42–44). The validity of self-reported sleep apnea in our cohorts has been studied previously and showed excellent reliability (42). Specifically, in 108 NHS/NHS II participants with self-reported sleep apnea, all diagnoses were confirmed by medical record review based on ≥1 objective monitoring method (92% by in-laboratory polysomnography, 98% were classified as obstructive, and 89% were considered as moderate-to-severe with AHI ≥15) (42). Moreover, the projected prevalence of moderate-to-severe OSA in the United States measured by polysomnography (1) was similar to that of self-reported OSA in our cohorts (42). Key OSA symptoms were assessed repeatedly throughout follow-up, including habitual snoring (defined as snoring every night or most nights) and excessive daytime sleepiness (EDS; defined as ≥4 d/wk of disrupted daily activities due to sleepiness) (4). We considered incidence of clinically diagnosed OSA as the primary outcome of interest, and incidence of symptomatic OSA with EDS (potentially more severe OSA) as a secondary outcome.
Statistical analysis
The analytic baseline was 2002 in the NHS, 1995 in the NHS II, and 1996 in the HPFS, with follow-up until 2012 in the NHS, 2013 in the NHS II, and 2012 in the HPFS. We calculated person-years of follow-up from the return date of the baseline questionnaire until the date of OSA diagnosis or the return date of the 2012/2013 questionnaire, whichever arrived earlier.
We first performed analyses in each cohort, and then conducted pooled analyses with assessment of potential heterogeneity by cohort using meta-analysis. Multivariable-adjusted HRs and 95% CIs for OSA risk across quintiles of AHEI and EDIP scores were estimated using Cox proportional hazards models; we found no evidence for violation of the proportional hazards assumption. Dietary exposures and covariates (except race/ethnicity and cohort) were modeled as time-varying in the analysis. Tests for linear trend were conducted by modeling median values in each quintile of AHEI or EDIP score as a continuous variable.
The first multivariable model was stratified by cohort (NHS, NHS II, and HPFS, in the pooled analyses only), age (mo), and questionnaire cycle, and adjusted for race/ethnicity (white, nonwhite), smoking (never, past, current), pack-years of smoking (continuous), menopausal status (premenopausal, postmenopausal) (females only), duration of postmenopausal hormone therapy (HT) by type (never, <5, 5 to <10, ≥10 y for estrogen-only and estrogen + progestin HT separately) (females only), physical exams (yes, no), total energy intake (in quintiles; kcal/d), habitual sleep duration (≤5, 6, 7, 8, ≥9 h/d), and physical activity (<6.0, 6.0–11.9, 12.0–20.9, 21.0–35.9, ≥36.0 metabolic equivalent task hours per week). Details for assessment of these covariates can be found in Supplemental Methods. The second multivariable model further adjusted for metabolic factors including BMI (continuous; in kg/m2), waist circumference (continuous; cm), history of diabetes (yes, no), and history of hypertension (yes, no), which were likely consequences of dietary patterns and considered as potential mediators in the analysis. We estimated the percentage of the association explained by these metabolic factors using the publicly available %Mediate macro (https://www.hsph.harvard.edu/donna-spiegelman/software/mediate/). The proportion was estimated by applying the partial likelihood function to 2 Cox models postulated on the same failure time variable, and the CIs were constructed by using the direct normal approximation to the point estimator (45). Secondarily, we repeated these analyses using OSA with EDS as the outcome.
We further explored the associations of individual foods/nutrients in each dietary score with OSA risk to understand what may drive potential differences in the results between the AHEI and EDIP. We also performed stratified analyses by age, sex, BMI, diabetes status, waist circumference, and physical activity to explore whether these factors modified the associations. Statistical significance of effect modification was assessed by testing multiplicative interactions using the likelihood ratio test. Finally, given the long time span of our exposure assessment, we repeated our primary analyses using AHEI and EDIP scores assessed 2–4 y before OSA diagnosis as the exposure to test the robustness of our findings based on the most recent diet.
To minimize missing data, information from the most recent previous questionnaire cycle was carried forward for imputation. Missing indicators were used if the data were still missing. This approach has been widely applied in prior studies in the NHS and HPFS (25, 46, 47).
Data analyses were conducted using SAS statistical software (version 9.4; SAS Institute Inc.), with 2-sided P values < 0.05 considered statistically significant.
Results
Population characteristics
A total of 8856 incident cases of OSA (NHS: 1743, NHS II: 5022, and HPFS: 2091) were identified during ≤18 y of follow-up, encompassing 2,051,278 person-years. Participants with higher AHEI scores were more likely to be older and white, engaged in more physical activity, had lower BMI and waist circumference, reported lower total energy intake and fewer pack-years of smoking, and were less likely to be current smokers, or report habitual snoring and EDS (Table 1). They were also less likely to have a history of diabetes and hypertension, and were more likely to have regular physical exams or use estrogen + progestin HT. Largely similar differences were observed comparing individuals with lower EDIP with those with higher EDIP (Table 2).
TABLE 1.
Age-adjusted baseline characteristics by AHEI score in the NHS, NHS II, and HPFS1
| Characteristic | AHEI score | |||||
|---|---|---|---|---|---|---|
| NHS (2002) | NHS II (1995) | HPFS (1996) | ||||
| Quintile 1, unhealthiest (n = 11,534) | Quintile 5, healthiest (n = 11,455) | Quintile 1, unhealthiest (n = 13,825) | Quintile 5, healthiest (n = 13,843) | Quintile 1, unhealthiest (n = 4015) | Quintile 5, healthiest (n = 3824) | |
| AHEI score, median (range) | 40.8 (17.3–44.6) | 64.8 (60.4–94.7) | 35.8 (16.0–39.7) | 61.7 (56.9–95.4) | 40.1 (16.5–44.4) | 67.2 (62.3–94.3) |
| Age, y | 65.1 ± 6.4 | 67.8 ± 6.5 | 40.1 ± 4.7 | 41.7 ± 4.4 | 57.3 ± 6.7 | 60.6 ± 7.3 |
| Nonwhite | 4.8 | 7.2 | 4.1 | 5.5 | 3.4 | 3.8 |
| BMI,2 kg/m2 | 27.5 ± 5.6 | 25.8 ± 4.7 | 26.4 ± 6.4 | 24.6 ± 5.0 | 22.8 ± 9.7 | 21.8 ± 9.5 |
| Waist circumference, cm | 88.5 ± 11.8 | 84.1 ± 11.4 | 88.9 ± 14.4 | 83.9 ± 12.9 | 101.7 ± 10.7 | 97.6 ± 10.1 |
| Smoking status | ||||||
| Never smoked | 52.2 | 42.5 | 71.2 | 61.3 | 53.4 | 55.3 |
| Past smoker | 36.8 | 53.8 | 17.1 | 31.4 | 38.1 | 42.2 |
| Current smoker | 11.0 | 3.7 | 11.7 | 7.3 | 8.5 | 2.5 |
| Pack-years of smoking3 | 21.7 ± 18.0 | 15.1 ± 13.8 | 13.6 ± 10.0 | 10.6 ± 7.7 | 23.5 ± 17.2 | 16.4 ± 12.8 |
| Habitual sleep duration, h | 7.2 ± 1.2 | 7.1 ± 1.1 | 7.0 ± 1.1 | 7.0 ± 1.0 | 7.2 ± 1.0 | 7.1 ± 1.0 |
| Habitual snoring, ≥4 nights/wk | 31.1 | 23.6 | 31.6 | 22.8 | 38.4 | 29.3 |
| Excessive daytime sleepiness, ≥4 d/wk | 3.9 | 2.9 | 12.1 | 9.3 | 11.2 | 8.6 |
| Physical activity,4 MET-h/wk | 12.8 ± 13.2 | 25.2 ± 21.2 | 15.0 ± 19.0 | 31.6 ± 33.3 | 22.7 ± 19.3 | 36.3 ± 27.1 |
| Total energy intake, kcal/d | 1848 ± 441 | 1691 ± 434 | 1964 ± 491 | 1703 ± 491 | 2132 ± 531 | 1886 ± 522 |
| History of diabetes | 7.3 | 4.5 | 0.6 | 0.2 | 1.8 | 1.3 |
| History of hypertension | 52.8 | 44.3 | 10.3 | 7.8 | 29.3 | 26.3 |
| Physical exams in the past 2 y | 92.6 | 94.6 | 88.2 | 90.5 | 64.5 | 72.3 |
| Postmenopausal | 99.3 | 99.3 | 7.5 | 7.0 | — | — |
| Ever estrogen-only HT use | 40.1 | 44.8 | 4.8 | 3.5 | — | — |
| Ever estrogen + progestin HT use | 36.6 | 44.5 | 1.6 | 2.2 | — | — |
Values are means ± SDs or percentages. AHEI, Alternative Healthy Eating Index 2010; HPFS, Health Professionals Follow-up Study; HT, hormone therapy; MET, metabolic equivalent task; NHS, Nurses’ Health Study.
Calculated as weight in kilograms divided by height in meters squared.
Among ever smokers.
Weekly energy expenditure in MET-hours per week from recreational and leisure-time physical activity.
TABLE 2.
Age-adjusted baseline characteristics by EDIP score in the NHS, NHS II, and HPFS1
| Characteristic | EDIP score | |||||
|---|---|---|---|---|---|---|
| NHS (2002) | NHS II (1995) | HPFS (1996) | ||||
| Quintile 1, anti-inflammatory (n = 11,743) | Quintile 5, proinflammatory (n = 11,189) | Quintile 1, anti-inflammatory (n = 13,852) | Quintile 5, proinflammatory (n = 13,799) | Quintile 1, anti-inflammatory (n = 3949) | Quintile 5, proinflammatory (n = 3879) | |
| EDIP score, median (range) | −0.4 (−3.5 to −0.2) | 0.3 (0.1–2.2) | −0.4 (−2.7 to 0.3) | 0.3 (0.2–2.4) | −0.4 (−3.2 to −0.3) | 0.3 (0.2–1.9) |
| Age, y | 66.5 ± 6.4 | 65.6 ± 6.4 | 41.9 ± 4.3 | 40.1 ± 4.8 | 58.6 ± 6.7 | 58.3 ± 7.2 |
| Nonwhite | 3.8 | 8.3 | 3.1 | 7.2 | 2.6 | 5.9 |
| BMI,2 kg/m2 | 25.2 ± 4.3 | 28.7 ± 5.9 | 24.5 ± 4.8 | 27.4 ± 6.7 | 22.1 ± 9.5 | 22.7 ± 10.1 |
| Waist circumference, cm | 83.8 ± 11.0 | 89.7 ± 12.4 | 84.3 ± 12.6 | 90.3 ± 15.1 | 99.7 ± 10.2 | 102 ± 11.3 |
| Smoking status | ||||||
| Never smoked | 37.2 | 54.1 | 55.2 | 73.3 | 43.7 | 62.8 |
| Past smoker | 54.0 | 38.6 | 31.4 | 18.3 | 50.0 | 32.3 |
| Current smoker | 8.8 | 7.3 | 13.4 | 8.4 | 6.3 | 4.9 |
| Pack-years of smoking3 | 18.4 ± 16.1 | 19.5 ± 16.8 | 12.0 ± 8.6 | 12.7 ± 9.5 | 20.3 ± 14.6 | 21.0 ± 16.0 |
| Habitual sleep duration, h | 7.2 ± 1.1 | 7.1 ± 1.2 | 7.0 ± 1.0 | 6.9 ± 1.1 | 7.2 ± 1.0 | 7.1 ± 1.0 |
| Habitual snoring, ≥4 nights/wk | 24.5 | 33.0 | 25.6 | 32.4 | 34.9 | 38.2 |
| Excessive daytime sleepiness, ≥4 d/wk | 2.5 | 4.3 | 9.7 | 12.4 | 9.6 | 10.9 |
| Physical activity,4 MET-h/wk | 21.5 ± 19.3 | 15.4 ± 15.1 | 25.4 ± 28.7 | 19.6 ± 24.7 | 30.4 ± 23.2 | 27.0 ± 24.6 |
| Total energy intake, kcal/d | 1746 ± 438 | 1893 ± 472 | 1803 ± 504 | 1966 ± 539 | 2007 ± 520 | 2199 ± 595 |
| History of diabetes | 2.4 | 11.6 | 0.2 | 1.1 | 1.1 | 2.9 |
| History of hypertension | 41.7 | 59.2 | 7.3 | 13.2 | 26.3 | 32.0 |
| Physical exams in the past 2 y | 93.8 | 93.7 | 89.3 | 89.4 | 71.7 | 66.8 |
| Postmenopausal | 99.4 | 99.2 | 6.8 | 8.5 | — | — |
| Ever estrogen-only HT use | 40.3 | 43.8 | 3.6 | 5.4 | — | — |
| Ever estrogen + progestin HT use | 45.1 | 35.3 | 2.2 | 1.9 | — | — |
Values are means ± SDs or percentages. EDIP, Empirical Dietary Inflammatory Pattern; HPFS, Health Professionals Follow-up Study; HT, hormone therapy; MET, metabolic equivalent task; NHS, Nurses’ Health Study.
Calculated as weight in kilograms divided by height in meters squared.
Among ever smokers.
Weekly energy expenditure in MET-hours per week from recreational and leisure-time physical activity.
AHEI and OSA risk
In pooled analyses adjusted for potential confounders, higher AHEI score was associated with lower OSA risk (Table 3). The HR was 0.76 (95% CI: 0.71, 0.82) comparing participants in the highest with those in the lowest AHEI score quintile. However, the inverse trend was only evident in the NHS II, but not in the NHS or HPFS (P-heterogeneity by cohort < 0.001). The HR comparing the extreme quintiles of AHEI score was 0.97 (95% CI: 0.83, 1.14) in the NHS (P-trend = 0.33), 0.65 (95% CI: 0.59, 0.72) in the NHS II (P-trend < 0.001), and 0.90 (95% CI: 0.78, 1.04) in the HPFS (P-trend = 0.27). After in addition controlling for metabolic factors, no association was detected between AHEI score and OSA risk in the pooled sample (P-trend = 0.54) or individual cohorts; metabolic factors fully attenuated the association between AHEI score and OSA risk (proportion mediated: 91.9%; 95% CI: 35.1%, 99.6%).
TABLE 3.
AHEI score and risk of incident obstructive sleep apnea in the NHS, NHS II, and HPFS1
| AHEI score | ||||||
|---|---|---|---|---|---|---|
| Quintile 1, unhealthiest | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5, healthiest | P-trend2 | |
| NHS | ||||||
| Cases, n (total n = 1743) | 417 | 377 | 331 | 304 | 314 | |
| Person-years, n | 110,164 | 110,416 | 110,870 | 110,797 | 110,569 | |
| Multivariable3 | 1.00 (Reference) | 0.96 (0.83, 1.10) | 0.89 (0.77, 1.03) | 0.85 (0.73, 0.99) | 0.97 (0.83, 1.14) | 0.33 |
| Multivariable + metabolic factors4 | 1.00 | 0.98 (0.85, 1.13) | 0.93 (0.80, 1.08) | 0.94 (0.80, 1.09) | 1.16 (0.99, 1.35) | 0.21 |
| NHS II | ||||||
| Cases, n (total n = 5022) | 1266 | 1113 | 951 | 945 | 747 | |
| Person-years, n | 237,574 | 240,482 | 241,411 | 242,441 | 241,842 | |
| Multivariable3 | 1.00 | 0.88 (0.81, 0.95) | 0.77 (0.70, 0.83) | 0.78 (0.71, 0.85) | 0.65 (0.59, 0.72) | <0.001 |
| Multivariable + metabolic factors4 | 1.00 | 0.95 (0.88, 1.04) | 0.88 (0.80, 0.96) | 0.96 (0.88, 1.05) | 0.91 (0.82, 1.00) | 0.06 |
| HPFS | ||||||
| Cases, n (total n = 2091) | 473 | 438 | 404 | 414 | 362 | |
| Person-years, n | 60,144 | 59,005 | 58,858 | 58,642 | 58,064 | |
| Multivariable3 | 1.00 | 0.96 (0.84, 1.10) | 0.92 (0.80, 1.05) | 0.99 (0.87, 1.14) | 0.90 (0.78, 1.04) | 0.27 |
| Multivariable + metabolic factors4 | 1.00 | 0.99 (0.86, 1.13) | 0.95 (0.83, 1.09) | 1.08 (0.95, 1.24) | 1.04 (0.90, 1.20) | 0.34 |
| Pooled | ||||||
| Cases, n (n = 8856) | 2156 | 1928 | 1686 | 1663 | 1423 | |
| Person-years, n | 407,882 | 409,903 | 411,139 | 411,880 | 410,474 | |
| Multivariable3 | 1.00 | 0.91 (0.86, 0.97) | 0.82 (0.77, 0.88) | 0.84 (0.79, 0.90) | 0.76 (0.71, 0.82) | <0.001 |
| Multivariable + metabolic factors4 | 1.00 | 0.96 (0.91, 1.03) | 0.90 (0.85, 0.96) | 0.98 (0.92, 1.05) | 0.98 (0.91, 1.05) | 0.54 |
Values are HRs and 95% CIs unless otherwise indicated. P-heterogeneity by cohort < 0.001. AHEI, Alternative Healthy Eating Index 2010; HPFS, Health Professionals Follow-up Study; HT, hormone therapy; NHS, Nurses’ Health Study.
P-trend was calculated using the midpoint of each category of AHEI score.
HRs and 95% CIs were estimated by Cox proportional hazards analyses. Multivariate analyses were stratified by cohort (NHS, NHS II, and HPFS, in the pooled analyses only), age (mo), and questionnaire cycle; adjusted for race/ethnicity (white, nonwhite), smoking (never, past, current), pack-years of smoking (continuous), menopausal status (premenopausal, postmenopausal) (females only), duration of postmenopausal HT by type (never, <5, 5 to <10, ≥10 y for estrogen-only and estrogen + progestin HT separately) (females only), physical exams (yes, no), total energy intake (in quintiles; kcal/d), habitual sleep duration (≤5, 6, 7, 8, ≥9 h/d), and physical activity (<6.0, 6.0–11.9, 12.0–20.9, 21.0–35.9, ≥36.0 metabolic equivalent task hours per week).
Metabolic factors included BMI (continuous; kg/m2), waist circumference (continuous; cm), history of diabetes (yes, no), and history of hypertension (yes, no).
EDIP and OSA risk
Higher EDIP score was linearly associated with increased risk of OSA in the pooled analysis (Table 4). Compared with the lowest quintile of EDIP score, the HR for OSA was 1.18 (95% CI: 1.10, 1.27) for quintile 2, 1.38 (95% CI: 1.28, 1.48) for quintile 3, 1.54 (95% CI: 1.43, 1.65) for quintile 4, and 1.94 (95% CI: 1.81, 2.08) for the highest quintile (P-trend < 0.001). Significant positive trends were consistently observed across all 3 individual cohorts (P-trend < 0.001 for all), although the association appeared stronger in the NHS and NHS II than in the HPFS (P-heterogeneity by cohort < 0.001). The observed association was partially attenuated after further adjusting for metabolic factors but remained statistically significant (HR comparing the highest and lowest quintiles of EDIP score: 1.31; 95% CI: 1.22, 1.41; P-trend < 0.001). The percentage of the association between EDIP score and OSA risk explained by metabolic factors was 64.8% (95% CI: 58.2%, 70.8%).
TABLE 4.
EDIP score and risk of incident obstructive sleep apnea in the NHS, NHS II, and HPFS1
| EDIP score | ||||||
|---|---|---|---|---|---|---|
| Quintile 1, anti-inflammatory | Quintile 2 | Quintile 3 | Quintile 4 | Quintile 5, proinflammatory | P-trend2 | |
| NHS | ||||||
| Cases, n (total n = 1743) | 233 | 260 | 351 | 406 | 493 | |
| Person-years, n | 113,860 | 112,312 | 110,619 | 109,270 | 106,753 | |
| Multivariable3 | 1.00 (Reference) | 1.13 (0.94, 1.35) | 1.51 (1.28, 1.78) | 1.71 (1.45, 2.01) | 2.01 (1.72, 2.36) | <0.001 |
| Multivariable + metabolic factors4 | 1.00 | 1.00 (0.84, 1.19) | 1.22 (1.03, 1.45) | 1.25 (1.06, 1.47) | 1.23 (1.04, 1.45) | 0.002 |
| NHS II | ||||||
| Cases, n (total n = 5022) | 663 | 839 | 966 | 1097 | 1457 | |
| Person-years, n | 242,210 | 243,059 | 241,969 | 240,417 | 236,096 | |
| Multivariable3 | 1.00 | 1.26 (1.13, 1.39) | 1.44 (1.30, 1.59) | 1.64 (1.49, 1.81) | 2.18 (1.99, 2.40) | <0.001 |
| Multivariable + metabolic factors4 | 1.00 | 1.16 (1.05, 1.29) | 1.21 (1.09, 1.34) | 1.21 (1.10, 1.34) | 1.31 (1.19, 1.44) | <0.001 |
| HPFS | ||||||
| Cases, n (total n = 2091) | 373 | 393 | 409 | 426 | 490 | |
| Person-years, n | 60,293 | 60,088 | 58,286 | 58,302 | 57,745 | |
| Multivariable3 | 1.00 | 1.07 (0.93, 1.23) | 1.15 (1.00, 1.33) | 1.20 (1.04, 1.39) | 1.38 (1.20, 1.59) | <0.001 |
| Multivariable + metabolic factors4 | 1.00 | 1.07 (0.93, 1.24) | 1.16 (1.00, 1.33) | 1.16 (1.01, 1.34) | 1.22 (1.06, 1.40) | 0.003 |
| Pooled | ||||||
| Cases, n (total n = 8856) | 1269 | 1492 | 1726 | 1929 | 2440 | |
| Person-years, n | 416,363 | 415,459 | 410,873 | 407,989 | 400,594 | |
| Multivariable3 | 1.00 | 1.18 (1.10, 1.27) | 1.38 (1.28, 1.48) | 1.54 (1.43, 1.65) | 1.94 (1.81, 2.08) | <0.001 |
| Multivariable + metabolic factors4 | 1.00 | 1.12 (1.04, 1.21) | 1.22 (1.13, 1.31) | 1.24 (1.15, 1.33) | 1.31 (1.22, 1.41) | <0.001 |
Values are HRs (95% CIs) unless otherwise indicated. P-heterogeneity by cohort < 0.001. EDIP, Empirical Dietary Inflammatory Pattern; HPFS, Health Professionals Follow-up Study; HT, hormone therapy; NHS, Nurses’ Health Study.
P-trend was calculated using the midpoint of each category of EDIP score.
HRs and 95% CIs were estimated by Cox proportional hazards analyses. Multivariate analyses were stratified by cohort (NHS, NHS II, and HPFS, in the pooled analyses only), age (mo), and questionnaire cycle; adjusted for race/ethnicity (white, nonwhite), smoking (never, past, current), pack-years of smoking (continuous), menopausal status (premenopausal, postmenopausal) (females only), duration of postmenopausal HT by type (never, <5, 5 to <10, ≥10 y for estrogen-only and estrogen + progestin HT separately) (females only), physical exams (yes, no), total energy intake (in quintiles; kcal/d), habitual sleep duration (≤5, 6, 7, 8, ≥9 h/d), and physical activity (<6.0, 6.0–11.9, 12.0–20.9, 21.0–35.9, ≥36.0 metabolic equivalent task hours per week).
Metabolic factors included BMI (continuous; in kg/m2), waist circumference (continuous; cm), history of diabetes (yes, no), and history of hypertension (yes, no).
Stratified analyses
In subgroup analysis, we detected significant effect modification by sex (P-interaction < 0.001), age (P-interaction < 0.001), and BMI (P-interaction < 0.001) for the association between AHEI score and OSA risk, with stronger inverse associations observed specifically for females, those aged <65 y, or those with BMI ≥25 kg/m2 than for their counterparts. In subgroup analysis of the association between EDIP score and OSA risk, we observed significant heterogeneity by sex (P-interaction = 0.005), age (P-interaction = 0.009), and diabetes status (P-interaction = 0.001). Dietary inflammatory potential was more strongly positively associated with OSA risk in females, those aged <65 y, or individuals without diabetes than in their counterparts (Supplemental Figure 2).
Dietary component analyses
In multivariable analyses examining individual dietary components in the AHEI, the strongest ones included positive associations with intakes of red and processed meat (HR per 1-SD increment: 1.18; 95% CI: 1.16, 1.21), sodium (1.18; 95% CI: 1.15, 1.20), and trans fat (1.10; 95% CI: 1.08, 1.13), and an inverse association with alcohol intake (0.86; 95% CI: 0.84, 0.88) (Supplemental Figure 3). Other AHEI components were either not associated with OSA risk (e.g., whole grains, ω-3 fatty acids) or moderately associated with OSA risk in an unexpected direction (e.g., vegetables, PUFAs, SSBs). Given these unexpected associations, we conducted a post hoc analysis excluding vegetables, PUFAs, SSBs, and nuts/legumes from the AHEI, and the AHEI score showed a stronger association with OSA risk. The HR comparing extreme quintiles was 0.59 (95% CI: 0.55, 0.64) before adjusting for metabolic factors (P-trend < 0.001) and 0.88 (95% CI: 0.81, 0.94) after adjusting for metabolic factors (P-trend < 0.001). By contrast, the significant association observed for EDIP was mainly due to inverse associations with intakes of wine (HR per 1-SD increment: 0.84; 95% CI: 0.82, 0.87), coffee (0.90; 95% CI: 0.88, 0.92), and beer (0.93; 95% CI: 0.91, 0.95), and strong positive associations with intakes of red meat (1.14; 95% CI: 1.12, 1.16), processed meat (1.11; 95% CI: 1.09, 1.13), and low-energy beverages (1.15; 95% CI: 1.13, 1.17) (Supplemental Figure 4).
Sensitivity analyses
Overall, the association patterns for OSA with EDS were similar to those observed in the primary analysis (Supplemental Tables 2 and 3). When using AHEI and EDIP scores assessed 2–4 y before OSA diagnosis as the exposure, our results remained essentially unchanged (Supplemental Tables 4 and 5).
Discussion
In this large prospective cohort study, lower diet quality and higher dietary inflammatory potential were associated with increased OSA risk after rigorously controlling for potential confounders. Additional adjustment for metabolic factors partly explained the association for EDIP score but fully attenuated the association with AHEI score. Our findings underscored the potential benefits of maintaining a healthier diet (particularly one with anti-inflammatory potential) in reducing OSA risk, and suggested the possible mediating role of metabolic factors in the association between diet and OSA risk. To our knowledge, we present the first prospective epidemiologic evidence to have characterized the magnitude of long-term excess risk of OSA in relation to overall diet quality and dietary inflammatory potential.
Comparison with other studies
Prior population-based evidence regarding the association between overall diet quality and OSA risk remains sparse and is limited to 4 cross-sectional studies (13–16). In a subset of MESA (Multi-Ethnic Study of Atherosclerosis) participants who completed an FFQ and 1-night in-home polysomnography, OSA was cross-sectionally associated with lower overall diet quality as measured by the AHEI (13). Two other smaller studies in obese adolescents/adults reported that severe OSA may be cross-sectionally associated with poor diet quality independently of degree of obesity (14, 15). A more recent study linked high-quality and anti-inflammatory diets with lower OSA prevalence (16). Our prospective study, leveraging the repeated, validated, longitudinal dietary assessment and self-reported, validated OSA diagnosis, reported a similar inverse association between overall diet quality and OSA diagnosis; however, no association was observed after accounting for metabolic factors including BMI and waist circumference, which is inconsistent with results from previous cross-sectional investigations (13–15). On the other hand, few epidemiologic studies have focused on the role of proinflammatory diet in OSA development, and we reported a robust positive association between dietary inflammatory potential and OSA risk independent of obesity and a wide spectrum of potential confounders. Collectively, our study added important prospective evidence to the existing cross-sectional investigations, suggesting that a healthier diet, particularly diet with lower inflammatory potential, may be associated with lower risk of OSA.
Of note, in our further analyses exploring individual dietary components, we observed that increased OSA risk associated with a higher intake of red/processed meat and a lower intake of alcohol for both the AHEI and EDIP. Interestingly, 2 prior cross-sectional studies reported similar positive associations between intake of red/processed meat and OSA (13, 48). However, the association between alcohol consumption and OSA risk remains equivocal. Although acute alcohol ingestion before sleep is known to exacerbate OSA severity and alter sleep architecture (49), there is a lack of evidence on the long-term impact of alcohol consumption on OSA risk, particularly the potential benefits of moderate social drinking that have been observed for multiple cardiometabolic outcomes (50–52). Of note, the strong positive association observed for sodium intake in the current study was consistent with prior well-designed studies linking sodium intake with OSA (53, 54). High sodium consumption may induce peripheral fluid retention, which has been implicated in OSA pathogenesis (55–58). For other AHEI components, we observed increased OSA risk associated with higher intakes of PUFAs, vegetables, and nuts/legumes and lower SSB intake, which are considered to be part of a healthy dietary pattern. These counterintuitive findings may to some extent dilute the overall inverse association with the AHEI. In a post hoc analysis excluding these 4 components from the AHEI, we found a stronger inverse association between the new AHEI score (including red and processed red meat, trans fat, sodium, alcohol, fruits, whole grains, and ω-3 fatty acids) and OSA risk, which remained statistically significant even after adjusting for metabolic factors. To our knowledge, the associations between individual dietary components (e.g., nuts/legumes) and OSA risk have been rarely studied and require further investigation.
For EDIP components, higher energy drink consumption and lower coffee consumption were associated with increased OSA risk. Prior cross-sectional studies have reported positive associations of carbonated and noncarbonated beverages with risk of OSA (13, 59), although our study suggests that low-energy drinks appear more strongly associated with the risk than do high-energy drinks. Similar to alcohol consumption, the long-term association between coffee intake and OSA risk is unclear despite evidence that caffeinated beverages may contribute to sleep disturbances such as prolonged sleep latency, reduced sleep duration, and poorer perceived sleep quality. However, in the Jackson Heart Sleep Study, evening caffeine use was not associated with sleep parameters measured by concurrent actigraphy and sleep diary (60). Other coffee components, such as polyphenols and trigonelline, have been shown in experimental studies to reduce inflammation and improve insulin resistance (61, 62), which are potential pathogenic pathways for OSA. Because coffee consumption was not included in the AHEI, the inverse association with coffee intake may drive the overall more significant association observed for the EDIP. However, coffee intake was not correlated with OSA in a previous small case–control study (63). For coffee and alcohol intakes, additional studies are needed to understand their relations with OSA risk in more detail, such as timing of coffee/alcohol consumption and type of coffee/alcohol (e.g., caffeinated compared with decaffeinated coffee).
Biological mechanisms
Increasing biological and epidemiologic evidence has recognized the role of chronic metabolic dysregulation and low-grade inflammation in the etiology of OSA (26, 27, 64). We consistently detected effect attenuation in the associations of both AHEI and EDIP with OSA risk after in addition accounting for metabolic factors in fully adjusted analyses, supporting that a healthier diet may lower OSA risk by reducing obesity, improving metabolic function, and mitigating chronic inflammation. Of note, although the EDIP was more strongly correlated with metabolic factors than was the AHEI in our study (Supplemental Table 1), adjustment for metabolic factors only partly explained the association with the EDIP but fully attenuated the association with the AHEI. Interestingly, the EDIP components associated with OSA risk, namely low-energy beverages, red meat, processed meat, wine, coffee, and beer, are also top food groups most strongly associated with circulating inflammatory biomarkers (18, 22), suggesting inflammatory pathways underlying the observed associations. Specifically, in our mediation analysis, 64.8% of the association between EDIP and OSA risk was explained by metabolic factors. We thus cannot exclude the possibility that a proinflammatory diet may influence OSA risk through other plausible biological mechanisms, such as lipid dysregulation and insulin resistance (40, 65, 66). In contrast, the mediation analysis indicated that the association of AHEI with OSA risk may be predominantly (91.9%) mediated by metabolic-related mechanisms. Given significant age and sex differences in OSA endotypes such as pharyngeal collapsibility and ventilator control (67, 68), future studies are warranted to elucidate whether the observed effect modification by age and sex for the associations between diet and OSA risk may be explained by the dietary impact on these OSA-related physiologic traits.
Strengths and limitations
Our study had several notable strengths, including the prospective study design, large sample size, long-term follow-up, high follow-up rates, and a large number of documented OSA cases with high reliability. Repeated and validated longitudinal assessment of dietary information and a wide spectrum of covariates allowed us to accurately capture long-term dietary exposure and rigorously control for confounding in a time-varying manner. The nature of our study participants (all health care professionals) further ensured data quality and internal validity and minimized socioeconomic confounding.
Our study had several limitations. First, although a previous validation study showed higher accuracy of self-reported OSA diagnoses in our cohorts of health care professionals (42), the impact of clinically undiagnosed OSA and differential OSA diagnosis on our results cannot be completely ruled out; such misclassification may potentially lead to underestimation of the associations. Second, information on OSA severity and treatment was unavailable in our cohorts. Third, despite extensive control for confounding, residual and unmeasured confounding were still possible given the observational study design. Lastly, the homogeneities of our study population (which comprised only health care professionals and were predominantly of European ancestry) may limit generalizability of our findings to underrepresented groups.
Conclusion
In summary, our prospective findings suggest that a healthier diet, particularly one with anti-inflammatory potential, may be associated with lower risk of OSA.
Supplementary Material
Acknowledgements
We thank all participants and staff of the Nurses’ Health Study, Nurses’ Health Study II, and Health Professionals Follow-up Study at Brigham and Women's Hospital and Harvard T.H. Chan School of Public Health for their contributions to this research.
The authors’ responsibilities were as follows—YL and TH: conceived and designed the study, had full access to all the data in the study, and take responsibility for the integrity of the data and the accuracy of the data analysis; YL: acquired, analyzed, and interpreted the data, drafted the manuscript, and performed the statistical analysis; FKT, MJS, SR, and TH: critically revised the manuscript for important intellectual content; SR and TH: obtained funding; MJS and TH: provided administrative, technical, or material support; TH: provided supervision; and all authors: read and approved the final manuscript.
Notes
Supported by NIH grants R35 HL135818 (to SR) and K01 HL143034 (to TH). The Nurses’ Health Study (NHS), NHS II, and Health Professionals Follow-up Study were supported by NIH grants UM1 CA186107, U01 CA176726, and U01 CA167552. The funders had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; and in the preparation, review, or approval of the manuscript.
The content is solely the responsibility of the authors and does not necessarily represent the views of the NIH.
Author disclosures: The authors report no conflicts of interest.
Supplemental Figures 1–4, Supplemental Tables 1–5, and Supplemental Methods are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.
Address correspondence to TH (e-mail: tih541@mail.harvard.edu).
Abbreviations used: AHEI, Alternative Healthy Eating Index 2010; AHI, Apnea-Hypopnea Index; EDIP, Empirical Dietary Inflammatory Pattern; EDS, excessive daytime sleepiness; HPFS, Health Professionals Follow-up Study; HT, hormone therapy; NHS, Nurses’ Health Study; OSA, obstructive sleep apnea; RRR, reduced rank regression; SSB, sugar-sweetened beverage.
Contributor Information
Yue Liu, Channing Division of Network Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA; Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA.
Fred K Tabung, James Cancer Hospital and Solove Research Institute, The Ohio State University Comprehensive Cancer Center, Columbus, OH, USA; Division of Medical Oncology, Department of Internal Medicine, College of Medicine, The Ohio State University Wexner Medical Center, Columbus, OH, USA; Division of Epidemiology, College of Public Health, The Ohio State University, Columbus, OH, USA; Department of Nutrition, Harvard TH Chan School of Public Health, Boston, MA, USA.
Meir J Stampfer, Channing Division of Network Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA; Department of Nutrition, Harvard TH Chan School of Public Health, Boston, MA, USA; Department of Epidemiology, Harvard TH Chan School of Public Health, Boston, MA, USA.
Susan Redline, Division of Sleep Medicine, Harvard Medical School, Boston, MA, USA; Division of Sleep and Circadian Disorders, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA.
Tianyi Huang, Channing Division of Network Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA; Division of Sleep Medicine, Harvard Medical School, Boston, MA, USA.
Data availability
Data described in the article, code book, and analytic code are available upon request pending approval by the Channing Division of Network Medicine at Brigham and Women's Hospital and Harvard Medical School, and Harvard TH Chan School of Public Health. Further information including the procedures to obtain and access cohort data is described at https://www.nurseshealthstudy.org/researchers (contact e-mail: nhsaccess@channing.harvard.edu) and https://sites.sph.harvard.edu/hpfs/for-collaborators/. The data are not publicly available owing to privacy or ethical restrictions.
References
- 1. Peppard PE, Young T, Barnet JH, Palta M, Hagen EW, Hla KM. Increased prevalence of sleep-disordered breathing in adults. Am J Epidemiol. 2013;177(9):1006–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Yaggi HK, Concato J, Kernan WN, Lichtman JH, Brass LM, Mohsenin V. Obstructive sleep apnea as a risk factor for stroke and death. N Engl J Med. 2005;353(19):2034–41. [DOI] [PubMed] [Google Scholar]
- 3. Gottlieb DJ, Yenokyan G, Newman AB, O'Connor GT, Punjabi NM, Quan SFet al. . Prospective study of obstructive sleep apnea and incident coronary heart disease and heart failure: the Sleep Heart Health Study. Circulation. 2010;122(4):352–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Huang T, Lin BM, Stampfer MJ, Tworoger SS, Hu FB, Redline S. A population-based study of the bidirectional association between obstructive sleep apnea and type 2 diabetes in three prospective U.S. cohorts. Diabetes Care. 2018;41(10):2111–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Huang T, Lin BM, Stampfer MJ, Schernhammer ES, Saxena R, Tworoger SSet al. . Associations of self-reported obstructive sleep apnea with total and site-specific cancer risk in older women: a prospective study. Sleep. 2021;44(3):zsaa198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Javaheri S, Barbe F, Campos-Rodriguez F, Dempsey JA, Khayat R, Javaheri Set al. . Sleep apnea: types, mechanisms, and clinical cardiovascular consequences. J Am Coll Cardiol. 2017;69(7):841–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Varraso R, Chiuve SE, Fung TT, Barr RG, Hu FB, Willett WCet al. . Alternate Healthy Eating Index 2010 and risk of chronic obstructive pulmonary disease among US women and men: prospective study. BMJ. 2015;350:h286. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Neelakantan N, Koh WP, Yuan JM, van Dam RM. Diet-quality indexes are associated with a lower risk of cardiovascular, respiratory, and all-cause mortality among Chinese adults. J Nutr. 2018;148(8):1323–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Hanson C, Rutten EP, Wouters EF, Rennard S. Influence of diet and obesity on COPD development and outcomes. Int J Chron Obstruct Pulmon Dis. 2014;9:723–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. McNicholas WT. COPD-OSA overlap syndrome: evolving evidence regarding epidemiology, clinical consequences, and management. Chest. 2017;152(6):1318–26. [DOI] [PubMed] [Google Scholar]
- 11. Godos J, Grosso G, Castellano S, Galvano F, Caraci F, Ferri Ret al. . Association between diet and sleep quality: a systematic review. Sleep Med Rev. 2021;57:101430. [DOI] [PubMed] [Google Scholar]
- 12. Chaput JP. Sleep patterns, diet quality and energy balance. Physiol Behav. 2014;134:86–91. [DOI] [PubMed] [Google Scholar]
- 13. Reid M, Maras JE, Shea S, Wood AC, Castro-Diehl C, Johnson DAet al. . Association between diet quality and sleep apnea in the Multi-Ethnic Study of Atherosclerosis. Sleep. 2019;42(1):zsy194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Beebe DW, Miller N, Kirk S, Daniels SR, Amin R. The association between obstructive sleep apnea and dietary choices among obese individuals during middle to late childhood. Sleep Med. 2011;12(8):797–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Jeznach-Steinhagen A, Okręglicka K, Nitsch-Osuch A, Czerwonogrodzka-Senczyna A, Barnaś M. Nutritional status and dietary patterns in adults with severe obstructive sleep apnea. Adv Exp Med Biol. 2020;1279:71–9. [DOI] [PubMed] [Google Scholar]
- 16. Melaku YA, Reynolds AC, Appleton S, Sweetman A, Shi Z, Vakulin Aet al. . High quality and anti-inflammatory diets and a healthy lifestyle are associated with lower sleep apnea risk. J Clin Sleep Med. 2022;18(6):1667–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. VanderWeele TJ. Can sophisticated study designs with regression analyses of observational data provide causal inferences?. JAMA Psychiatry. 2021;78(3):244–6. [DOI] [PubMed] [Google Scholar]
- 18. Tabung FK, Smith-Warner SA, Chavarro JE, Wu K, Fuchs CS, Hu FBet al. . Development and validation of an empirical dietary inflammatory index. J Nutr. 2016;146(8):1560–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Lopez-Garcia E, Schulze MB, Fung TT, Meigs JB, Rifai N, Manson JEet al. . Major dietary patterns are related to plasma concentrations of markers of inflammation and endothelial dysfunction. Am J Clin Nutr. 2004;80(4):1029–35. [DOI] [PubMed] [Google Scholar]
- 20. Nettleton JA, Steffen LM, Mayer-Davis EJ, Jenny NS, Jiang R, Herrington DMet al. . Dietary patterns are associated with biochemical markers of inflammation and endothelial activation in the Multi-Ethnic Study of Atherosclerosis (MESA). Am J Clin Nutr. 2006;83(6):1369–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Huang T, Tobias DK, Hruby A, Rifai N, Tworoger SS, Hu FB. An increase in dietary quality is associated with favorable plasma biomarkers of the brain-adipose axis in apparently healthy US women. J Nutr. 2016;146(5):1101–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Tabung FK, Smith-Warner SA, Chavarro JE, Fung TT, Hu FB, Willett WCet al. . An empirical dietary inflammatory pattern score enhances prediction of circulating inflammatory biomarkers in adults. J Nutr. 2017;147(8):1567–77. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Lee DH, Li J, Li Y, Liu G, Wu K, Bhupathiraju Set al. . Dietary inflammatory and insulinemic potential and risk of type 2 diabetes: results from three prospective U.S. cohort studies. Diabetes Care. 2020;43(11):2675–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Li J, Lee DH, Hu J, Tabung FK, Li Y, Bhupathiraju SNet al. . Dietary inflammatory potential and risk of cardiovascular disease among men and women in the U.S. J Am Coll Cardiol. 2020;76(19):2181–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Tabung FK, Liu L, Wang W, Fung TT, Wu K, Smith-Warner SAet al. . Association of dietary inflammatory potential with colorectal cancer risk in men and women. JAMA Oncol. 2018;4(3):366–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Dempsey JA, Veasey SC, Morgan BJ, O'Donnell CP. Pathophysiology of sleep apnea. Physiol Rev. 2010;90(1):47–112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Unnikrishnan D, Jun J, Polotsky V. Inflammation in sleep apnea: an update. Rev Endocr Metab Disord. 2015;16(1):25–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Huang T, Goodman M, Li X, Sands SA, Li J, Stampfer MJet al. . C-reactive protein and risk of obstructive sleep apnea in four US cohorts. Chest. 2021;159(6):2439–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Lopes TVC, Borba MES, Lopes RVC, Fisberg RM, Paim SL, Teodoro VVet al. . Association between inflammatory potential of the diet and sleep parameters in sleep apnea patients. Nutrition. 2019;66: 5–10. [DOI] [PubMed] [Google Scholar]
- 30. Colditz GA, Manson JE, Hankinson SE. The Nurses’ Health Study: 20-year contribution to the understanding of health among women. J Womens Health. 1997;6(1):49–62. [DOI] [PubMed] [Google Scholar]
- 31. Colditz GA, Hankinson SE. The Nurses’ Health Study: lifestyle and health among women. Nat Rev Cancer. 2005;5(5):388–96. [DOI] [PubMed] [Google Scholar]
- 32. Harvard TH Chan School of Public Health . Health Professionals Follow-up Study. [Internet]. Boston (MA): Harvard TH Chan School of Public Health. [Accessed 2022 Sep 30]. Available from: https://sites.sph.harvard.edu/hpfs/. [Google Scholar]
- 33. Hu FB, Rimm E, Smith-Warner SA, Feskanich D, Stampfer MJ, Ascherio Aet al. . Reproducibility and validity of dietary patterns assessed with a food-frequency questionnaire. Am J Clin Nutr. 1999;69(2):243–9. [DOI] [PubMed] [Google Scholar]
- 34. Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi Jet al. . Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol. 1985;122(1):51–65. [DOI] [PubMed] [Google Scholar]
- 35. Yuan C, Spiegelman D, Rimm EB, Rosner BA, Stampfer MJ, Barnett JBet al. . Relative validity of nutrient intakes assessed by questionnaire, 24-hour recalls, and diet records as compared with urinary recovery and plasma concentration biomarkers: findings for women. Am J Epidemiol. 2018;187(5):1051–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Hu FB, Stampfer MJ, Rimm E, Ascherio A, Rosner BA, Spiegelman Det al. . Dietary fat and coronary heart disease: a comparison of approaches for adjusting for total energy intake and modeling repeated dietary measurements. Am J Epidemiol. 1999;149(6):531–40. [DOI] [PubMed] [Google Scholar]
- 37. Bernstein AM, Rosner BA, Willett WC. Cereal fiber and coronary heart disease: a comparison of modeling approaches for repeated dietary measurements, intermediate outcomes, and long follow-up. Eur J Epidemiol. 2011;26(11):877–86. [DOI] [PubMed] [Google Scholar]
- 38. Chiuve SE, Fung TT, Rimm EB, Hu FB, McCullough ML, Wang Met al. . Alternative dietary indices both strongly predict risk of chronic disease. J Nutr. 2012;142(6):1009–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. McCullough ML, Feskanich D, Stampfer MJ, Giovannucci EL, Rimm EB, Hu FBet al. . Diet quality and major chronic disease risk in men and women: moving toward improved dietary guidance. Am J Clin Nutr. 2002;76(6):1261–71. [DOI] [PubMed] [Google Scholar]
- 40. Shi N, Aroke D, Jin Q, Lee DH, Hussan H, Zhang Xet al. . Proinflammatory and hyperinsulinemic dietary patterns are associated with specific profiles of biomarkers predictive of chronic inflammation, glucose-insulin dysregulation, and dyslipidemia in postmenopausal women. Front Nutr. 2021;8:690428. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Jin Q, Shi N, Aroke D, Lee DH, Joseph JJ, Donneyong Met al. . Insulinemic and inflammatory dietary patterns show enhanced predictive potential for type 2 diabetes risk in postmenopausal women. Diabetes Care. 2021;44(3):707–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Huang T, Lin BM, Markt SC, Stampfer MJ, Laden F, Hu FBet al. . Sex differences in the associations of obstructive sleep apnoea with epidemiologic factors. Eur Respir J. 2018;51(3):1702421. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Huang T, Lin BM, Redline S, Curhan GC, Hu FB, Tworoger SS. Type of menopause, age at menopause, and risk of developing obstructive sleep apnea in postmenopausal women. Am J Epidemiol. 2018;187(7):1370–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Liu Y, Yang L, Stampfer MJ, Redline S, Tworoger SS, Huang T. Physical activity, sedentary behaviour and incidence of obstructive sleep apnea in three prospective US cohorts. Eur Respir J. 2022;59(2):2100606. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Lin DY, Fleming TR, De Gruttola V. Estimating the proportion of treatment effect explained by a surrogate marker. Stat Med. 1997;16(13):1515–27. [DOI] [PubMed] [Google Scholar]
- 46. Tobias DK, Stuart JJ, Li S, Chavarro J, Rimm EB, Rich-Edwards Jet al. . Association of history of gestational diabetes with long-term cardiovascular disease risk in a large prospective cohort of US women. JAMA Intern Med. 2017;177(12):1735–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Veronese N, Li Y, Manson JE, Willett WC, Fontana L, Hu FB. Combined associations of body weight and lifestyle factors with all cause and cause specific mortality in men and women: prospective cohort study. BMJ. 2016;355:i5855. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Kechribari I, Kontogianni MD, Georgoulis M, Lamprou K, Mourati I, Vagiakis Eet al. . Associations between red meat intake and sleep parameters in patients with obstructive sleep apnea. J Acad Nutr Diet. 2020;120(6):1042–53. [DOI] [PubMed] [Google Scholar]
- 49. Burgos-Sanchez C, Jones NN, Avillion M, Gibson SJ, Patel JA, Neighbors Jet al. . Impact of alcohol consumption on snoring and sleep apnea: a systematic review and meta-analysis. Otolaryngol Head Neck Surg. 2020;163(6):1078–86. [DOI] [PubMed] [Google Scholar]
- 50. Wood AM, Kaptoge S, Butterworth AS, Willeit P, Warnakula S, Bolton Tet al. . Risk thresholds for alcohol consumption: combined analysis of individual-participant data for 599 912 current drinkers in 83 prospective studies. Lancet. 2018;391(10129):1513–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51. Lai Y-J, Hu H-Y, Lee Y-L, Ko M-C, Ku P-W, Yen Y-Fet al. . Frequency of alcohol consumption and risk of type 2 diabetes mellitus: a nationwide cohort study. Clin Nutr. 2019;38(3):1368–72. [DOI] [PubMed] [Google Scholar]
- 52. Li X-H, Yu F-f, Zhou Y-H, He J. Association between alcohol consumption and the risk of incident type 2 diabetes: a systematic review and dose-response meta-analysis. Am J Clin Nutr. 2016;103(3):818–29. [DOI] [PubMed] [Google Scholar]
- 53. Pimenta E, Stowasser M, Gordon RD, Harding SM, Batlouni M, Zhang Bet al. . Increased dietary sodium is related to severity of obstructive sleep apnea in patients with resistant hypertension and hyperaldosteronism. Chest. 2013;143(4):978–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Kasai T, Arcand J, Allard JP, Mak S, Azevedo ER, Newton GEet al. . Relationship between sodium intake and sleep apnea in patients with heart failure. J Am Coll Cardiol. 2011;58(19):1970–4. [DOI] [PubMed] [Google Scholar]
- 55. Yumino D, Redolfi S, Ruttanaumpawan P, Su M-C, Smith S, Newton GEet al. . Nocturnal rostral fluid shift: a unifying concept for the pathogenesis of obstructive and central sleep apnea in men with heart failure. Circulation. 2010;121(14):1598–605. [DOI] [PubMed] [Google Scholar]
- 56. Redolfi S, Yumino D, Ruttanaumpawan P, Yau B, Su M-C, Lam Jet al. . Relationship between overnight rostral fluid shift and obstructive sleep apnea in nonobese men. Am J Respir Crit Care Med. 2009;179(3):241–6. [DOI] [PubMed] [Google Scholar]
- 57. Perger E, Jutant EM, Redolfi S. Targeting volume overload and overnight rostral fluid shift: a new perspective to treat sleep apnea. Sleep Med Rev. 2018;42:160–70. [DOI] [PubMed] [Google Scholar]
- 58. da Silva BC, Kasai T, Coelho FM, Zatz R, Elias RM. Fluid redistribution in sleep apnea: therapeutic implications in edematous states. Front Med (Lausanne). 2017;4:256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59. Du Y, Duan X, Zheng M, Zhao W, Huang J, Lao Let al. . Association between eating habits and risk of obstructive sleep apnea: a population-based study. Nat Sci Sleep. 2021;13:1783–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60. Spadola CE, Guo N, Johnson DA, Sofer T, Bertisch SM, Jackson CLet al. . Evening intake of alcohol, caffeine, and nicotine: night-to-night associations with sleep duration and continuity among African Americans in the Jackson Heart Sleep Study. Sleep. 2019;42(11):zsz136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61. Liang N, Kitts DD. Role of chlorogenic acids in controlling oxidative and inflammatory stress conditions. Nutrients. 2015;8(1):16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62. van Dijk AE, Olthof MR, Meeuse JC, Seebus E, Heine RJ, van Dam RM. Acute effects of decaffeinated coffee and the major coffee components chlorogenic acid and trigonelline on glucose tolerance. Diabetes Care. 2009;32(6):1023–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63. Elnour MAA, Saleh AA, Kalantan MM, Mirghani HO. The relationship between coffee intake, obstructive sleep apnea risk, and type 2 diabetes glycemic control, in Tabuk City, The Kingdom of Saudi Arabia: a case–control study. BMC Res Notes. 2019;12(1):798. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 64. Vgontzas AN, Bixler EO, Chrousos GP. Sleep apnea is a manifestation of the metabolic syndrome. Sleep Med Rev. 2005;9(3):211–24. [DOI] [PubMed] [Google Scholar]
- 65. Huang T, Sands SA, Stampfer MJ, Tworoger SS, Hu FB, Redline S. Insulin resistance, hyperglycemia, and risk of developing obstructive sleep apnea in U.S. men and women. Ann Am Thorac Soc. 2022; Apr 6 (Epub ahead of print; doi: 10.1513/AnnalsATS.202111-1260OC). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66. Tishler PV, Larkin EK, Schluchter MD, Redline S. Incidence of sleep-disordered breathing in an urban adult population: the relative importance of risk factors in the development of sleep-disordered breathing. JAMA. 2003;289(17):2230–7. [DOI] [PubMed] [Google Scholar]
- 67. Won CHJ, Reid M, Sofer T, Azarbarzin A, Purcell S, White Det al. . Sex differences in obstructive sleep apnea phenotypes, the multi-ethnic study of atherosclerosis. Sleep. 2020;43(5):zsz274. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 68. Edwards BA, Wellman A, Sands SA, Owens RL, Eckert DJ, White DPet al. . Obstructive sleep apnea in older adults is a distinctly different physiological phenotype. Sleep. 2014;37(7):1227–36A. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data described in the article, code book, and analytic code are available upon request pending approval by the Channing Division of Network Medicine at Brigham and Women's Hospital and Harvard Medical School, and Harvard TH Chan School of Public Health. Further information including the procedures to obtain and access cohort data is described at https://www.nurseshealthstudy.org/researchers (contact e-mail: nhsaccess@channing.harvard.edu) and https://sites.sph.harvard.edu/hpfs/for-collaborators/. The data are not publicly available owing to privacy or ethical restrictions.