To the Editor:
Both obesity and obstructive sleep apnea (OSA) are common and evolving global health problems. OSA is estimated to affect ∼1 billion individuals globally (1). Obesity is a strong risk factor for OSA (2). However, not all individuals with obesity develop OSA, which likely reflects individual physiological differences (3). Prior work using invasive measurements indicated that differences in upper airway collapsibility and the responsiveness of upper airway dilator muscles explain why only some individuals with obesity develop OSA (3, 4). Our goal was to validate prospectively these findings using a noninvasive technique that estimates mechanisms underlying OSA from routine polysomnography (5–7).
Methods
The aim of the SLIM-OSA (Underlying Mechanisms of Obesity-induced Obstructive Sleep Apnea) trial (NCT 04793334; institutional review board #191948) is to elucidate the mechanisms that explain why weight loss alleviates OSA in some but not all people with class 2 (body mass index [BMI] ≥35 to <40 kg/m2) and class 3 (BMI ≥ 40 kg/m2) obesity. For this single-center trial, we enrolled participants 18–65 years of age with BMIs ≥35 kg/m2, who were nonsmokers, without major comorbidities, who planned to undergo sleeve gastrectomy. Participants underwent in-laboratory polysomnography before and after surgery. For this cross-sectional analysis, we included data from participants who completed presurgical polysomnography (an analysis of changes in upper airway anatomy/function will be the subject of a subsequent paper). As before (3), OSA was defined as an apnea–hypopnea index (AHI) >15 events/h, with hypopneas based on reductions in flow of ≥30% for ≥10 seconds associated with a ≥3% desaturation or arousal.
Physiological mechanisms underlying each person’s OSA (“traits”) were estimated from polysomnography using published techniques (5–7). To quantify upper airway collapsibility, we used primarily Vmin, which estimates flow as a percentage of eupneic ventilation at nadir drive. Vmin correlates well (r = −0.54) with passive closing pressure, which is arguably the gold standard to assess upper airway collapsibility (5, 8). Vpassive, which estimates flow at eupneic drive, was used as a secondary measure of upper airway collapsibility (a lower Vmin:Vpassive ratio denotes worse collapsibility). To assess the responsiveness of upper airway dilator muscles, we quantified Vactive (flow at elevated drive, specifically the arousal threshold), Vcomp (change in flow from passive to active conditions [Vactive − Vpassive]), and upper airway gain (UAG; the ratio of the change in flow to the change in drive, as drive increases from the eupnea to the arousal threshold, which might arguably be the best noninvasive surrogate for upper airway muscle responsiveness invasively measured during a previous investigation [3]). Higher values of Vactive, Vcomp, and UAG denote better upper airway dilator muscle function. Ventilatory instability (“loop gain”) was estimated as the magnitude of the ventilatory drive response to a prior reduction in ventilation in the time frame of 1 minute (range, 0 to infinity; higher values denote more instability). The arousal threshold was estimated as the degree of respiratory drive causing arousals (lower values denote easier arousability).
We compared traits across participants with versus without OSA using Wilcoxon rank sum tests and used linear regression to adjust for baseline covariates. To assess the relative contributions of traits to OSA pathogenesis, we performed covariate-adjusted logistic regression with scaled traits as predictors and OSA status as the outcome. The primary outcome was passive upper airway collapsibility (Vmin) compared across participants with versus without OSA (without adjustment), using a P value <0.05 to indicate statistical significance. Analyses were performed using R version 4.3.2.
Results
Of 55 participants, 7 were excluded (1 unable to sleep during polysomnography and 6 in whom we were unable to estimate traits because of insufficient flow-limited events). This analysis included 48 patients. Participants were mostly women (90%), with a median age of 45 years; 35% were non-White, and 52% reported Hispanic/Latino ethnicity. The OSA group had a median AHI of 37 events/h, indicating severe OSA (Table 1).
Table 1.
Patient demographics and univariable comparisons of traits
| Characteristic | Overall (n = 48) | OSA (n = 26) | No OSA (n = 22)* | P Value† |
|---|---|---|---|---|
| Female sex | 43 (90) | 21 (81) | 22 (100) | 0.054 |
| Body mass index, kg/m2 | 40.4 (38.1–42.6) | 40.7 (39.1–42.3) | 39.6 (37.7–42.6) | 0.4 |
| Age, yr | 45 (37–53) | 48 (40–55) | 43 (35–47) | 0.2 |
| Race | 0.3 | |||
| American Indian/Alaska Native | 1 (2.1) | 1 (3.8) | 0 (0) | |
| Asian | 1 (2.1) | 1 (3.8) | 0 (0) | |
| Black or African American | 3 (6.3) | 2 (7.7) | 1 (4.5) | |
| More than one race | 8 (17) | 2 (7.7) | 6 (27) | |
| Native Hawaiian or other Pacific Islander | 1 (2.1) | 0 (0) | 1 (4.5) | |
| Unknown/not reported | 3 (6.3) | 1 (3.8) | 2 (9.1) | |
| White | 31 (65) | 19 (73) | 12 (55) | |
| Ethnicity | 0.049 | |||
| Hispanic or Latino | 23 (48) | 16 (62) | 7 (32) | |
| Not Hispanic or Latino | 25 (52) | 10 (38) | 15 (68) | |
| Apnea–hypopnea index, events/h | 16 (10–38) | 37 (27–48) | 8 (2–11) | <0.001 |
| Time beneath 90%, min | 2 (0–12) | 6 (2–18) | 0 (0–2) | <0.001 |
| Hypoxic burden, %min/h | 20 (12–79) | 73 (43–123) | 9 (2–16) | <0.001 |
| Vmin, % | 72 (63–80) | 66 (57–73) | 77 (72–85) | 0.001 |
| Vpassive,‡ % | 74 (64–83) | 68 (58–75) | 83 (74–88) | <0.001 |
| Vactive, % | 101 (91–104) | 99 (77–103) | 102 (94–105) | 0.047 |
| Vcomp, % | 4 (1–9) | 4 (0–10) | 4 (2–8) | >0.9 |
| UAG, % | 1.2 (0.2–2.3) | 1.0 (0.3–1.8) | 1.3 (0.2–2.8) | 0.5 |
| Loop gain | 0.54 (0.46–0.72) | 0.55 (0.47–0.67) | 0.53 (0.44–0.73) | 0.8 |
| Arousal threshold, % | 116 (105–131) | 124 (115–147) | 108 (102–119) | 0.001 |
| Number of windows | ||||
| Vmin/Vpassive/Vactive/Vcomp | 67 (36–104) | 89 (73–121) | 35 (15–55)§ | <0.001 |
| Loop gain | 46 (28–80) | 63 (47–93) | 27 (12–42)ǁ | <0.001 |
| Arousal threshold | 41 (24–53) | 50 (45–74) | 24 (10–33)¶ | <0.001 |
Definition of abbreviations: AHI = apnea–hypopnea index; OSA = obstructive sleep apnea; UAG = upper airway gain; Vactive = is flow at elevated drive, specifically the level of the arousal threshold; Vcomp = the change in flow from passive to active conditions (V active-V passive); Veupnea = is eupneic ventilation; Vpassive = is flow at eupneic drive; Vmin = estimates the flow as a percentage of eupneic ventilation (%Veupnea) at nadir drive (lowest decile).
Data are expressed as n (%) or median (interquartile range).
Seven participants in the no OSA group had AHIs <5 events/h, whereas the rest had AHIs of 5–15 events/h.
Fisher exact test; Wilcoxon rank sum test.
Vpassive was transformed for normality as in prior studies: Vpassive > 100 was set to 100, then Vpassive was square-root transformed as 1 − (1 − Vpassive/100)0.5.
Participants with AHIs of 5–15 events/h had significantly more valid windows for trait estimation than those with AHIs <5 events/h (45 [35–56] vs. 14 [8–16]; P = 0.001).
Participants with AHIs of 5–15 events/h had significantly more valid windows for trait estimation than those with AHIs <5 events/h (34 [27–49] vs. 11 [7–13]; P = 0.002).
Participants with AHIs of 5–15 events/h had significantly more valid windows for trait estimation than those with AHIs <5 events/h (28 [24–34] vs. 10 [5–11]; P < 0.001).
Compared with participants without OSA, those with OSA had significantly more collapsible upper airway anatomy on the basis of lower Vmin (P < 0.001; Figure 1, Table 1). Results were similar when adjusting for baseline covariates or when using Vpassive instead (Table 2). With accumulation of respiratory drive, patients with OSA achieved lower ventilation immediately before arousal than those without OSA on the basis of lower Vactive (P = 0.047), even though their drive at the threshold of arousal was significantly higher (P < 0.001; Figure 1). When adjusting for covariates, results for the arousal threshold were similar but attenuated for Vactive. Vcomp, UAG, and loop gain were similar across both groups (P ≥ 0.5). Sensitivity analyses excluding seven patients with AHI < 5 events/h from the “no OSA” group or excluding the five men from the “OSA” group had slightly attenuated significance, but similar effect estimates suggested robustness of results (data not shown).
Figure 1.
Box-and-whisker plots comparing traits in patients with (red) versus without OSA (blue). For numerical data, see Table 1. OSA = obstructive sleep apnea.
Table 2.
Results from multivariable models
| Model 1 (Age + Body Mass Index + Ethnicity) |
Model 2 (Model 1 + Sex) |
|||||
|---|---|---|---|---|---|---|
| Linear Regression: Trait ∼ OSA Status | β* | (95% CI) | P Value | β* | (95% CI) | P Value |
| Vmin | −12.2 | (−23.1 to −1.2) | 0.03 | −10.2 | (−22.0 to 1.7) | 0.09 |
| Vpassive† | −13.9 | (−26.2 to −1.7) | 0.03 | −10.3 | (−23.3 to 2.7) | 0.12 |
| Vactive | −11.2 | (−25.3 to 2.8) | 0.11 | −7.9 | (−23.0 to 7.2) | 0.30 |
| Vcomp | −3.0 | (−8.7 to 2.7) | 0.29 | −2.9 | (−9.1 to 3.3) | 0.36 |
| UAG | 0.1 | (−2.3 to 2.4) | 0.97 | 0.14 | (−2.4 to 2.7) | 0.92 |
| Loop gain | −0.03 | (−0.14 to 0.08) | 0.59 | −0.04 | (−0.16 to 0.09) | 0.56 |
| Arousal threshold | 21.2 | (5.7 to 36.7) | 0.01 | 12.4 | (−2.8 to 27.6) | 0.11 |
| Logistic Regression: OSA Status ∼ Traitscaled | OR‡ | (95% CI) | P Value | OR‡ | (95% CI) | P Value |
|---|---|---|---|---|---|---|
| Vmin | 0.36 | (0.10 to 0.85) | 0.02 | 0.39 | (0.11 to 0.96) | 0.04 |
| Vpassive† | 0.35 | (0.11 to 0.83) | 0.02 | 0.40 | (0.13 to 1.01) | 0.052 |
| Vactive | 0.48 | (0.16 to 1.07) | 0.08 | 0.60 | (0.20 to 1.44) | 0.25 |
| Vcomp | 0.69 | (0.33 to 1.36) | 0.28 | 0.79 | (0.27 to 2.04) | 0.64 |
| UAG | 1.02 | (0.53 to 2.00) | 0.95 | 1.05 | (0.55 to 2.17) | 0.88 |
| Loop gain | 0.83 | (0.41 to 1.65) | 0.58 | 0.84 | (0.40 to 1.73) | 0.64 |
| Arousal threshold | 4.47 | (1.55 to 21.7) | 0.003 | 3.12 | (1.01 to 15.0) | 0.048 |
Definition of abbreviations: CI = confidence interval; OR = odds ratio; OSA = obstructive sleep apnea; UAG = upper airway gain.
For the linear regression analyses, traits were the outcomes, whereas OSA status was the outcome for the logistic regression models. P values in boldface type denote statistical significance.
β reflects the mean difference in the trait in participants with versus without OSA.
Vpassive was transformed for normality as in prior studies: Vpassive > 100 was set to 100, then Vpassive was square-root transformed as 1 − (1 − Vpassive/100)0.5.
For a 1 standard deviation change in the trait, the odds of having OSA change by (OR − 1) × 100%.
On the basis of multivariable logistic regression, an increase in Vmin by 1 standard deviation (i.e., a less collapsible upper airway) decreased the odds of having OSA by 64% (odds ratio, 0.36 [95% confidence interval, 0.10–0.85]; P = 0.02; Table 2). In exploratory, covariate-adjusted analyses, Vmin:Vpassive ratio and Vactive were inversely associated with both AHI and hypoxic burden (P < 0.05). Lower Vcomp was associated with greater hypoxic burden (P = 0.04) but not AHI (P = 0.48).
Discussion
Our findings validate and extend results from previous work (3, 4). This prospective study, which included substantially more women and had a higher BMI than previous cohorts, confirms that in patients with class 2 and 3 obesity, OSA is likely caused primarily by a more collapsible upper airway both during passive/hypotonic (Vmin) and active (Vactive) conditions (3). Consistent with prior observations, we found no difference in ventilatory control stability but an increase in arousal threshold across groups which is likely a consequence rather than a cause of OSA (9). Counter to our expectations, we did not find a difference in Vcomp or UAG, but we note that these measures are composites of other trait estimates and thus have larger variance (i.e., less precision), and confidence intervals were wide, including biologically important effects. Furthermore, the similarity of our findings to previous results from invasive measurements (3) also provides important evidence for the construct validity of noninvasive trait estimates of upper airway collapsibility, arousal threshold, and loop gain (10, 11).
Strengths and Limitations
Limitations of this study include its preliminary nature with a modest sample size, cross-sectional analyses, small number of male participants (precluding the assessment of effect modification by sex), and the inability to quantify traits in some participants with very low AHIs (<3 events/h). Furthermore, estimates in patients with AHI < 5 events/h have not been validated, but on the basis of exploratory analyses, we found a similar width of bootstrapped 95% confidence intervals in participants with AHIs <5 versus >5 events/h for all traits except loop gain (data not shown). This finding suggests that estimates for most traits have similar precision in subjects with AHIs <5 versus >5 events/h, and results from a sensitivity analysis excluding participants with AHI < 5 events/h were overall similar, but more research is needed to validate trait estimates from patients with few respiratory events. Moreover, we did not assess end-expiratory lung volume, which might be reduced in patients with obesity and thus contribute to a more collapsible upper airway (12, 13). However, BMI was similar across groups, and results were similar when adjusting for BMI.
Conclusions
Future research including dynamic magnetic resonance imaging of the upper airway, and longitudinal assessments after weight loss, might help pinpoint the exact location responsible for the upper airway compromise in patients with obesity and OSA (e.g., tongue fat) (14). Ultimately, this line of inquiry might allow better identification of patients with OSA benefiting from weight loss interventions as part of a precision medicine approach for OSA.
Footnotes
Supported by U.S. Department of Veterans Affairs Clinical Science Research and Development grant CDA-2 IK2CX002524-01A2 and the National Institutes of Health (NIH) Loan Repayment Program (B.N.); Division of Intramural Research grant R01HL148436 (A.M.); and American Heart Association grant CDA#940501, NIH grant K23HL161336, and American Academy of Sleep Medicine Foundation grant 277-JF-22 (C.N.S.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the American Heart Association, NIH, or American Academy of Sleep Medicine Foundation. The project described was partially supported by NIH grant UL1TR000100 of Clinical and Translational Science Award funding before August 13, 2015, and grant UL1TR001442 of Clinical and Translational Science Award funding beginning August 13, 2015, and beyond. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Clinical trial registered with www.clinicaltrials.gov (NCT 04793334).
Author disclosures are available with the text of this letter at www.atsjournals.org.
References
- 1. Benjafield AV, Ayas NT, Eastwood PR, Heinzer R, Ip MSM, Morrell MJ, et al. Estimation of the global prevalence and burden of obstructive sleep apnoea: a literature-based analysis. Lancet Respir Med . 2019;7:687–698. doi: 10.1016/S2213-2600(19)30198-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Peppard PE, Young T, Barnet JH, Palta M, Hagen EW, Hla KM. Increased prevalence of sleep-disordered breathing in adults. Am J Epidemiol . 2013;177:1006–1014. doi: 10.1093/aje/kws342. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Sands SA, Eckert DJ, Jordan AS, Edwards BA, Owens RL, Butler JP, et al. Enhanced upper-airway muscle responsiveness is a distinct feature of overweight/obese individuals without sleep apnea. Am J Respir Crit Care Med . 2014;190:930–937. doi: 10.1164/rccm.201404-0783OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Schwartz AR, Gold AR, Schubert N, Stryzak A, Wise RA, Permutt S, et al. Effect of weight loss on upper airway collapsibility in obstructive sleep apnea. Am Rev Respir Dis . 1991;144:494–498. doi: 10.1164/ajrccm/144.3_Pt_1.494. [DOI] [PubMed] [Google Scholar]
- 5. Vena D, Taranto-Montemurro L, Azarbarzin A, Op de Beeck S, Marques M, Vanderveken OM, et al. Clinical polysomnographic methods for estimating pharyngeal collapsibility in obstructive sleep apnea. Sleep . 2022;45:zsac050. doi: 10.1093/sleep/zsac050. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Terrill PI, Edwards BA, Nemati S, Butler JP, Owens RL, Eckert DJ, et al. Quantifying the ventilatory control contribution to sleep apnoea using polysomnography. Eur Respir J . 2015;45:408–418. doi: 10.1183/09031936.00062914. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Sands SA, Edwards BA, Terrill PI, Taranto-Montemurro L, Azarbarzin A, Marques M, et al. Phenotyping pharyngeal pathophysiology using polysomnography in patients with obstructive sleep apnea. Am J Respir Crit Care Med . 2018;197:1187–1197. doi: 10.1164/rccm.201707-1435OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Carberry JC, Jordan AS, White DP, Wellman A, Eckert DJ. Upper airway collapsibility (Pcrit) and pharyngeal dilator muscle activity are sleep stage dependent. Sleep . 2016;39:511–521. doi: 10.5665/sleep.5516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Loewen A, Ostrowski M, Laprairie J, Atkar R, Gnitecki J, Hanly P, et al. Determinants of ventilatory instability in obstructive sleep apnea: inherent or acquired? Sleep . 2009;32:1355–1365. doi: 10.1093/sleep/32.10.1355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Younes M, Schwab R. Con: can physiological risk factors for obstructive sleep apnea be determined by analysis of data obtained from routine polysomnography? Sleep . 2023;46:zsac158. doi: 10.1093/sleep/zsac158. [DOI] [PubMed] [Google Scholar]
- 11. Sands SA, Edwards BA. Pro: can physiological risk factors for obstructive sleep apnea be determined by analysis of data obtained from routine polysomnography? Sleep . 2023;46:zsac310. doi: 10.1093/sleep/zsac310. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Jordan AS, White DP, Owens RL, Eckert DJ, Rahangdale S, Yim-Yeh S, et al. The effect of increased genioglossus activity and end-expiratory lung volume on pharyngeal collapse. J Appl Physiol (1985) . 2010;109:469–475. doi: 10.1152/japplphysiol.00373.2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Squier SB, Patil SP, Schneider H, Kirkness JP, Smith PL, Schwartz AR. Effect of end-expiratory lung volume on upper airway collapsibility in sleeping men and women. J Appl Physiol (1985) . 2010;109:977–985. doi: 10.1152/japplphysiol.00080.2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Wang SH, Keenan BT, Wiemken A, Zang Y, Staley B, Sarwer DB, et al. Effect of weight loss on upper airway anatomy and the apnea-hypopnea index: the importance of tongue fat. Am J Respir Crit Care Med . 2020;201:718–727. doi: 10.1164/rccm.201903-0692OC. [DOI] [PMC free article] [PubMed] [Google Scholar]