Obstructive Sleep Apnea and Airway Dimensions in Chronic Obstructive Pulmonary Disease

To the Editor:

Chronic obstructive pulmonary disease (COPD) and obstructive sleep apnea (OSA) are prevalent conditions. Current estimates suggest that from 10% to 15% of patients with COPD have OSA, depending on the population studied (1, 2). COPD–OSA overlap confers heightened risk of sleep desaturation, poor sleep efficiency, COPD exacerbation, and mortality compared with COPD alone (2, 3). Quantitative multidetector computed tomography (MDCT) is used in research studies to identify phenotypes among those with COPD. Thickened airway walls visualized by MDCT are believed to be due to intermittent hypoxemia and airway inflammation and have been associated with increased COPD exacerbations and decreased quality of life (4, 5). Few have studied MDCT characteristics in COPD–OSA (6–8), and understanding these characteristics could help elucidate mechanisms driving heightened risk for adverse outcomes in this population. The hypothesis of this study is that those with COPD–OSA overlap have increased thickness of airways compared with those who have COPD alone.

Methods

Current or former smokers aged 40 years or older were enrolled in a convenience sample from outpatient pulmonary clinics at a single center. Participants had COPD defined as follows: post-bronchodilator forced expiratory volume in 1 second (FEV₁)/forced vital capacity (FVC) less than 0.70, FEV₁ percent predicted 30–70%, and diffusing capacity of the lung for carbon monoxide greater than or equal to 35% predicted as defined by the parent study. In addition, they had no COPD exacerbations for 6 weeks before enrollment and were not receiving supplemental oxygen. Polysomnogram (PSG) and MDCT data from 53 participants were analyzed. The Johns Hopkins Institutional Review Board approved this study, and written informed consent was obtained. Participants completed an in-laboratory full-montage PSG. Sleep parameters included apnea–hypopnea index (AHI), arousal index (AI), and mean oxygen saturation. AHI is the combination of apneas and hypopneas per hour of sleep. Apnea is defined as total cessation of airflow for at least 10 seconds. Hypopnea is defined as a decrease in flow greater than or equal to 30% from baseline for at least 10 seconds with a 4% oxygen desaturation or an arousal. AI is the number of arousals per hour. Each participant underwent whole-lung inspiratory and expiratory MDCT (1-mm slice thickness) using settings comparable to those in other multicenter trials (9). Measurements included multilevel airway luminal and wall dimensions, percentage emphysema (percentage voxels less than −950 Hounsfield units [HU] at total lung volume), and percentage air trapping (percentage voxels less than −856 HU at residual volume) using VIDA software (VIDA Diagnostics). Airways were arbitrarily designated as small (<4 mm) or large (≥4 mm) by diameter. Up to 19 airways per individual were identified for analysis. The airway walls were analyzed using wall area fraction (WAF) of the total airway area (the proportion of airway wall area as a fraction of the total airway area [Ao − Ai/Ao], where Ao is the total airway area and Ai is the luminal area).

Demographic and clinical data were compared by AHI category (<15 vs. ≥15 events/h) using t tests and chi-square tests. Sleep and MDCT parameters were logarithmically transformed for normality and analyzed cross-sectionally using unadjusted and adjusted linear regression for percentage emphysema and percentage air trapping, and generalized estimating equations were used to account for clustering of multiple measurements of airways per individual. Models were adjusted for age, sex, race, FEV₁ percent predicted, pack-years smoked, total lung volume, body mass index (BMI), and any respiratory inhaler use in the 2 weeks before enrollment. Analyses were completed using Stata version 13 software (StataCorp).

Results

The participants’ median age was 62 years (interquartile range [IQR], 55–72), 59% were male, and 40% were African American. They had a median 53 (IQR, 37–69) pack-year smoking history, and most were current smokers (79%). Median BMI was 25 kg/m² (IQR, 22–31), FEV₁ percent predicted was 54.0 (IQR, 45–61), and 40% of participants had an AHI greater than or equal to 15. AHI was correlated with AI (r = 0.857). Median time between computed tomography (CT) and PSG was 65 days (IQR, 36–145). Median percentage emphysema and percentage air trapping for the cohort were 5 (IQR, 1–14) and 31 (IQR, 17–52), respectively. There were no differences in clinical characteristics or MDCT measures (Table 1) by AHI category. Thirty-three individuals used at least one inhaled respiratory medication, and none had received oral corticosteroids in the 2 weeks before enrollment.

Table 1.

Participant characteristics, by apnea–hypopnea index dichotomized at 15 events per hour

	n (%) or Median (IQR)
	AHI <15/h (n = 32)	AHI ≥15/h (n = 21)
Age, yr	60 (55–72)	64 (56–72)
Male sex	15 (50)	14 (74)
Race
African American	13 (41)	8 (38)
White	19 (59)	13 (62)
Body mass index, kg/m2	24.6 (21.9–31.7)	26.2 (21.7–31.1)
Cardiovascular disease*	11 (34)	6 (29)
Current smoker	25 (78)	17 (81)
Smoking pack-years	43 (33–64)	55 (42–83)
FEV1, % predicted	53 (41–61)	56 (52–62)
Apnea–hypopnea index	6 (4–9)	29 (20–48)
Inhaler use†	22 (69)	11 (52)
Percentage emphysema	6.49 (0.56–17.25)	3.07 (0.87–11.62)
Percentage air trapping	36.43 (15.75–55.71)	24.34 (16.74–37.22)
Airway diameter‡, mm	7.26 (6.84–8.17)	7.66 (7.12–8.41)
Large airway WAF, mm	0.57 (0.49–0.64)	0.58 (0.49–0.65)
Small airway WAF, mm	0.71 (0.68–0.74)	0.72 (0.70–0.75)

Definition of abbreviations: AHI = apnea–hypopnea index; FEV₁ = forced expiratory volume in 1 second; IQR = interquartile range; WAF = wall area fraction, the proportion of airway wall area as a fraction of the total airway area.

^*Cardiovascular disease includes reported heart failure, stroke, diabetes, heart attack, angina, or hypertension.

^†Inhaler use includes reported use of any respiratory inhaler in the 2 weeks preceding enrollment.

^‡Airways were designated as small (<4 mm) or large (≥4 mm) by diameter.

In multivariable analysis, sleep parameters were associated with WAF in small airways (Table 2). No associations of sleep parameters with WAF in large airways were observed. In addition, there were no associations of sleep parameters with percentage emphysema or percentage air trapping on MDCT (data not shown).

Table 2.

Multivariable linear regression of wall area fraction on polysomnogram measurements in chronic obstructive pulmonary disease (N = 53)

	Airway Wall Area Fraction
	Predicted % Change per 10% Increase	95% CI
Large airways (≥4-mm diameter)
Apnea–hypopnea index	0.038	(−0.086 to 0.153)
Arousal index	0.048	(−0.057 to 0.143)
Mean oxygen saturation	5.021	(−4.590 to 15.600)
Small airways (<4-mm diameter)
Apnea–hypopnea index	0.095	(0.019 to 1.682)
Arousal index	0.048	(0.019 to 0.143)
Mean oxygen saturation	−5.161	(−9.048 to −1.100)

Definition of abbreviation: CI = confidence interval.

Airway wall area fraction and polysomnogram measurements were log transformed for normality. Predicted percentage changes in all parameters for small airways were statistically significant (P < 0.05). Model was adjusted for age, sex, race, percent predicted forced expiratory volume in 1 second, pack-year smoking history, lung volume on computed tomography, body mass index, and inhaler use. Inhaler use includes reported use of any respiratory inhaler in the 2 weeks preceding enrollment.

Discussion

This analysis demonstrates that in COPD, OSA is associated with a distinct phenotype of increased small airway WAF. In addition, a dose–response association was found between sleep fragmentation and nocturnal hypoxemia with small airway WAF. Although overall airway size is smaller in subjects with COPD than in healthy control subjects (7), increased proportion of wall area is associated with poor COPD outcomes, including exacerbations and symptoms (4, 5). In the present analysis, among individuals with COPD, overall airway diameter was not different between those with and without OSA, whereas WAF was found to be higher in those with concomitant OSA. These findings linking CT measures with OSA may help identify mechanisms underlying increased risk for adverse outcomes in individuals with COPD–OSA overlap (3).

Individuals with COPD–OSA overlap may have increased inflammation leading to changes in airway morphology. Local inflammation could be caused by the presence of alveolar hypoxia in COPD intensified by a partially closed airway in COPD–OSA overlap. In this study, participants were awake and supine during CT; therefore, it is likely that these observed differences were due to chronic changes rather than being temporally related to obstruction. Upper airway obstruction was not measured during CT; however, this would be an interesting measure for future studies.

Alternatively, airway remodeling in the small airways may be due to increased regional cytokine release from abdominal visceral adipocytes, near the diaphragm in those with central obesity, which have been shown to be elevated in OSA (10). These findings may be limited to small airways rather than large airways, owing to differences in the ability to fully inspire in the setting of central obesity, leading to decreased ability to fully open peripheral airways. Participants with and without OSA had similar BMI, which is unexpected; however, individuals with COPD often have lower weight as disease progresses, which may explain this result. Furthermore, we did not perform additional measures of central adiposity, which may be increased among those with OSA in the setting of a normal or overweight BMI.

The association between sleep fragmentation, measured as AI, and WAF may have implications unrelated to OSA and could indicate that the combination of respiratory events, limb movements, and insomnia may influence WAF. However, based on the strong correlation between AHI and AI, the association is likely driven largely by respiratory events in this cohort.

This study of the association between OSA and WAF in COPD is novel and, to our knowledge, has not been reported previously. Previously reported differences in WAF in COPD have shown similar effect size changes when people with and without bronchitis (mean WAF, 0.63 vs. 0.62) (11) have been compared and also when frequency of exacerbations has been compared (mean WAF, 0.62 vs. 0.63) (5), supporting the hypothesis that the difference presented here is meaningful, and this adds to the current literature on relevant radiographic phenotypes within COPD. Recently reported associations between OSA and percentage emphysema or percentage air trapping (8) were not replicated, possibly because of differences in sex, BMI, and percentage emphysema between study populations. This cohort had relatively low percentage emphysema and may have had more chronic bronchitis, influencing WAF changes. Availability of full PSG and spirometry data are strengths of this study. Use of inhaled medications was collected via questionnaire; however, adherence to medications was not measured. Limitations include small sample size, cross-sectional design limiting our ability to understand directionality, lack of a comparator group, and delayed time between some MDCT and PSG. Information on medications such as sedatives or analgesics were not collected in this cohort, and is a limitation of this analysis. In contrast to findings in the Sleep Heart Health Study that showed no association between mild COPD and OSA in a community cohort (2), there is a high prevalence of COPD–OSA overlap in the present COPD cohort that may limit generalizability.

Individuals with COPD and OSA have increased small airway WAF visualized by MDCT, which represents an airway phenotype not explained by poor lung function alone. This intriguing finding, though unlikely to change clinical practice, adds to existing literature on airway changes in COPD based on MDCT imaging and is worthy of further study. This evidence does not point to a new modality to screen for OSA; however, gaining an understanding of how airway measurements change with COPD and OSA is important for potential treatment with positive pressure. If OSA heightens WAF in COPD, understanding the underlying mechanisms is important, given the implications for treatment strategies and potential for improved outcomes in this high-risk population.