Air Pollutants Are Associated With Obstructive Sleep Apnea Severity in Non-Rapid Eye Movement Sleep

Abstract

Study Objectives:

The relationship between seasonal variation of obstructive sleep apnea and ambient temperature and pollutants has been inconsistent in previous studies. It is also unknown whether the seasonal variation in apnea-hypopnea index influences continuous positive airway pressure treatment dose. This study aims to examine the seasonality of obstructive sleep apnea and continuous positive airway pressure treatment, and the association between air pollutants and apnea-hypopnea index in adults with different sleep apnea severity during different sleep stages.

Methods:

Polysomnography of 5,413 patients referred to one sleep center during 2008–2015 were examined retrospectively. Ambient conditions and air pollutants levels were collected from the official air condition surveillance database. Cosinor analysis was used to examine seasonal variances. The general linear model was used to examine associations between air conditions and apnea-hypopnea index adjusted for seasonality. Models for apnea-hypopnea index in different sleep stages, sex groups, and obstructive sleep apnea severity groups were analyzed separately.

Results:

Seasonal variations for continuous positive airway pressure treatment were not significant. Particulate matter less than or equal to 10 μm, ozone, sulfur dioxide, and relative humidity were associated with apnea-hypopnea index only in patients with severe obstructive sleep apnea. The association was significant only in non-rapid eye movement sleep.

Conclusions:

An adjustment for continuous positive airway treatment dose by season is not warranted. Protection for air pollutant-vulnerable groups should be provided. The exact mechanism of the associations between apnea-hypopnea index and air conditions only in non-rapid eye movement sleep must be clarified.

Citation:

Cheng WJ, Liang SJ, Huang CS, Lin CL, Pien LC, Hang LW. Air pollutants are associated with obstructive sleep apnea severity in non-rapid eye movement sleep. J Clin Sleep Med. 2019;15(6):831–837.

BRIEF SUMMARY

Current Knowledge/Study Rationale: It has been shown that the severity of obstructive sleep apnea has seasonal patterns, but the association with ambient conditions is inconclusive. It is also unknown whether continuous positive airway pressure treatments have seasonal patterns.

Study Impact: We found that air pollutants and relative humidity were related to the apnea–hypopnea index in women and patients with severe obstructive sleep apnea but only during non-rapid eye movement sleep. Seasonal variations for continuous positive airway pressure treatment dose were not significant; therefore, an adjustment for continuous positive airway treatment dose by season is not warranted. This study identified populations vulnerable to air pollutant-related sleep apnea exacerbation and suggested that AHI in non-rapid eye movement sleep is more affected by ambient factors than AHI in REM sleep.

INTRODUCTION

Obstructive sleep apnea (OSA) has been found to have a seasonal pattern in both children and adults.^1–5 A Google Trends study revealed a seasonal pattern in internet search queries for snoring and sleep apnea, and such searches peaked in winter and early spring.⁶ Some risk factors for OSA fluctuate with season, including upper airway infections,⁷ body weight,^8,9 asthma,¹⁰ alcohol consumption,¹¹ and tobacco use.¹² Seasonal variation of atmospheric conditions has also been suggested to be one of the mechanisms. Apnea-hypopnea index (AHI) was found to be inversely correlated with ambient temperature, atmospheric pressure, relative humidity, carbon monoxide (CO),⁵ particulate air matter less than or equal to 10 and 2.5 μm in aerodynamic diameter (PM10 and PM2.5),¹³ nitrogen dioxide,^14,15 ozone,^16,17 and indoor biomass fuel pollution.^18,19 However, the results were contradicting in study areas with different temperature and pollutant variations.²⁰

It has been proposed that correcting for season with polysomnography (PSG) will improve treatment decision making (eg, adenotonsillectomy in children).¹ Continuous positive airway pressure (CPAP) is the standard treatment for adult patients with high AHI. However, noncompliance remains a major issue for treatment failure. One common reason for non-compliance is the use of high pressure, which results in nasal dryness and discomfort.²¹ Whether a CPAP titration study performed in winter could lead to a higher pressure than that in other seasons has not been studied.

Differences between AHI in rapid eye movement (REM) and non-rapid eye movement (NREM) sleep in patients with OSA has been explored. Previous research suggests that sleep apnea during REM sleep is more associated with upper airway anatomical structure characteristics and pharyngeal muscle responses.²² However, it was also noted that almost half of patients with OSA had a higher AHI in NREM than in REM sleep.²³ Nonanatomical traits may also play important roles.²⁴ One possible explanations is that during NREM sleep, loop gain is higher than that in REM sleep, thus leading to an oversensitive ventilator control.²⁵ Otherwise, in phasic REM sleep, the arousal threshold was found to be higher and AHI was lower than those in slow-wave sleep.²⁶ Whether seasonal patterns of AHI and its associations with air pollutants differ between REM sleep and NREM sleep has not been studied. The difference may provide insight into the mechanism of seasonal variations in OSA and the effect of air pollutants on OSA.

This study examined the following: (1) the seasonality of AHI and CPAP treatment dose, (2) the association between air pollutants and AHI in adults with different sleep apnea severities, and (3) the association between air pollutants and REM and NREM AHI.

METHODS

Participants

The study population comprised patients referred to the Sleep Center of the China Medical University Hospital for PSG diagnostic study by physicians between 2008 and 2015. The center is located in the city of Taichung, Taiwan, at a latitude of 23.5° north. All patients underwent full-night PSG and provided their demographic data before the study (ie, age, sex, and smoking status). Neck circumference, body weight, and body height were measured by sleep technicians. Smoking status was categorized into five groups: no smoking, fewer than 10, 11–20, 21–30, and more than 31 cigarettes per day. Body mass index was calculated as the weight in kilograms divided by the square of the height in meters (kg/m²). The PSG results and demographic data were reviewed retrospectively in 2017. This study was approved by the Institutional Review Board of China Medical University Hospital (CMUH103-REC2-082).

Polysomnography and CPAP Titration

The patients were asked to arrive at the sleep center before 11:00 pm, and the following parameters were measured: six-channel electroencephalography; electrooculography; electrocardiography; nasal air pressure transducer; oronasal thermistor; thoracic and abdominal movements using inductance plethysmography; electromyography with submental and shin leads; finger oxygen saturation (SpO₂); sound recordings; piezoelectric sensor for snore detection; and videotaping. Only patients who were older than 18 years and had a total time in bed greater than 6 hours and a total sleep time greater than 3.5 hours in the PSG study were included in this study. Sleep staging and respiratory events coding were accomplished by certificated sleep technicians. PSG data from 6,186 patients were collected.

OSA severity was measured using AHI, which was calculated as the average number of apneas and hypopneas per hour. Obstructive apnea was defined as cessation of airflow through the nose with paradoxical chest and abdominal movements, and hypopnea was defined as a ≥ 30% reduction in nasal pressure with paradoxical chest and abdominal movements resulting in desaturation of at least 4% of SpO₂. Central apnea and obstructive apnea were calculated separately.

We also identified 262 patients who participated CPAP titration study between 2008 and 2015. The patients had standard PSG, and CPAP pressures were titrated by sleep technicians until AHI was less than 5 during the REM stage or the patient was in a supine position. Sleep specialists further determined the suggested CPAP treatment dose according to the titration reports.

Air Pollutants

Hourly air quality data recorded by the Environmental Protection Administration are available at http://taqm.epa.gov.tw/taqm/tw/YearlyDataDownload.aspx. Levels of CO, ozone, nitrogen oxides (NOx), sulfur dioxide (SO₂), PM2.5, PM10, nonmethane hydrocarbons (NMHC), relative humidity, and ambient temperature from the monitoring station geographically closest to the sleep laboratory were averaged to obtain daily data.²⁷ We retrieved the data collected from 2008 to 2015.

Data Analysis

The daily values of AHI, CPAP treatment pressure, ambient temperature, and air pollutants were averaged across 2008– 2015 to yield data for 365 days (ie, from January 1 to December 31). Cosinor analysis was employed to test if there were significant seasonal variations in these parameters. Cosinor analysis fits a sinusoid to an observed time series and estimates the amplitude, acrophase, and midline statistic of rhythm (mesor).²⁸ We tested the seasonal curve of AHI for patients with an obstructive apnea index lower than the central apnea index (n = 414), which did not show a significant seasonal pattern (P = .53). It suggested that central sleep apnea does not have a seasonal pattern. Hence, only patients with predominantly OSA were included in the following analysis (n = 5,413). Cosinor analyses were performed using the R software suite (version 3.4.4).

Next, the 12 months of a year were categorized into four seasons. We defined summer as June to August because the acrophase of ambient temperature was in July. The other seasons were defined as spring (March to May), autumn (September to November), and winter (December to February). We examined the differences in demographic and physiological characteristics and PSG study parameters between the four seasons using chi-square test for categorical variables and analysis of variance for continuous variables.

Generalized linear models were used to examine the relationship between personal characteristics, air pollutants, and AHI. The dates of PSG measurements were linked to daily air condition data. The cosinor models were used as part of a generalized linear model to adjust for seasonal variations of OSA severity. We adopted daily sine and cosine from the null cosinor models because the inclusion of the seasonality parameter showed the best model fit statistics, (ie, F statistics and adjusted R²), compared with models without seasonal parameters. Collinearity of air pollutants was examined using the variance inflation factor. In the models with all air pollutants, we yielded the largest condition index of 42.5 and a variance inflation factor above 8 for PM2.5, PM10, NOx, NMHC, and CO levels. Because of the high collinearity of air pollutants, they were added into the generalized linear models one at a time. We examined the models in all participants, participants with an AHI greater than 30, and male and female groups. We also examined the effect of personal characteristics and air pollutants on AHI in REM and NREM sleep.

RESULTS

The differences between the four seasons were significant for sex and smoking (Table 1). A higher percentage of men visited the sleep center in winter than in summer. Total AHI, AHI in REM and NREM sleep, stage N1 sleep percentage, sleep efficiency, and time with a SpO₂ less than or equal to 90% were also higher in winter than in summer. Conversely, the lowest SpO₂ and the percentage for REM sleep was higher in summer than in winter.

Table 1.

Demographic characteristics of study participants and their polysomnography parameter distributions over four seasons.

The cosinor analysis results showed that seasonal variation was significant for total AHI but was nonsignificant for CPAP treatment dose (Table 2). The adjusted mean (mesor) of AHI was 22.07 events/h. The acrophase occurred on February 16. Both REM AHI and NREM AHI showed significant seasonal patterns, but NREM AHI had a better model fit than REM AHI (adjusted R² = .051 and .044). The acrophase of NREM AHI was closer to overall AHI than was REM AHI. All air pollutants showed significant seasonal variations. Except for ozone, all pollutants had an acrophase that occurred in January, with the lowest point in July. By contrast, ozone showed a reverse pattern, with the acrophase occurring on August 5.

Table 2.

Seasonal patterns of apnea-hypopnea index, CPAP treatment dose, and air conditions (n = 5,413).

In the regression analysis for all patients, AHI was significantly predicted by older age, male sex, smoking, higher neck circumference, higher body mass index, and relative humidity (Table 3). Ambient temperature and air pollutants were not significantly associated with AHI. In models without adjustment for seasonal variations for AHI, (ie, daily sine and cosine), ambient temperature was significantly associated with AHI (β coefficient = −0.245, P < .001). However, PM10, ozone, SO₂, and relative humidity were significantly associated with AHI in patients with severe OSA. In this group, younger age and female sex were significantly associated with AHI. Ozone and relative humidity were found to be associated with AHI in the female group but not the male group. In the severe OSA group, we found that PM10, ozone, SO₂, and relative humidity were significantly associated with NREM AHI, but not REM AHI (Table 4).

Table 3.

Multivariable linear models for apnea-hypopnea index.

Table 4.

Multivariable linear models for apnea-hypopnea index in rapid eye movement and non-rapid eye movement sleep in patients with AHI > 30 events/h (n = 1,559).

DISCUSSION

This is the first study to examine the seasonal variation of CPAP treatment dose and the associations between air pollutants and AHI in patients with varying OSA severities. We found that although CPAP treatment dose was higher in winter than in summer, its seasonal variation was not significant. However, seasonal variations for AHI and air conditions were signifi-cant. With seasonal variations for AHI adjusted, we still found several factors that were associated with AHI. In female patients and those with severe OSA, PM10, ozone, SO₂, and relative humidity were associated with AHI. Furthermore, the association was significant for AHI in only NREM sleep but not in REM sleep. Several risk factors for OSA did not show significant effect in our models. For example, neck circumference was not associated with AHI in men and the AHI > 30 group, after adjusting for BMI. In the AHI > 30 group, women were associated with higher AHI. These findings suggest a more sophisticated investigation in the pathology of OSA in different groups.

The seasonal variation of AHI may be partially due to a higher percentage of men referred to PSG studies in winter than in summer. The sex differences between seasons was consistent with previous findings,⁵ but the reason was unknown. Despite the significant seasonal variation of AHI, the difference in means was as trivial as 2.5 events/h between winter and summer. The difference in suggested CPAP dose between seasons was also as small as 1.2 cmH₂O. This explains the lack of significant seasonal variation of CPAP treatment dose. Furthermore, CPAP titration results and treatment responsiveness were influenced by complicated factors other than AHI. Mathematical equations used to predict CPAP treatment dose in the literatures often include body mass index, mean oxygen saturation, and neck circumference.^29,30 Laryngopharyngeal anatomy such as tongue position and hyoid-mental distance was also found to be associated with CPAP treatment dose.³¹ Nasal surgery also reduces the required treatment pressure.³² Hence, we concluded that CPAP treatment does not vary significantly between seasons.

Several studies have examined the associations between AHI and temperature. One experimental study showed that a room temperature of 24°C was associated with lower AHI than that of 16°C, and a study in Brazil also showed that lower temperature was associated with higher AHI.^5,33 However, two community studies found a reverse relationship,^13,16 one of which the average temperature was 13°C.¹⁶ Our study participants had an average AHI higher than that in the latter two studies, and the average temperature was 10°C higher. Therefore, the inverse directions of associations in the studies suggest that the relationship between temperature and AHI may not be linear. Additional studies are warranted to determine the optimal temperature for patients with OSA. An association between relative humidity and AHI found in a study in Brazil, although the coefficient was low (β = 0.006), was inconsistent with our findings.⁵ An experimental study on patients with mild to moderate OSA revealed no effect of relative humidity on AHI.³⁴ Many other studies found that heated humidifiers did not improve AHI but did relieve nasal symptoms in CPAP users.^35–37 However, some researchers have found that humidification benefits patients with inadequate physiological levels of humidity, (eg, mouth breathing, nasal congestion, and high CPAP treatment dose).^38,39 The functions of heating and humidifying inspired air by nose varies between different ethnic groups, and the function was weaker in East Asians than in Northern Europeans.⁴⁰ This may lead to vulnerability to lower relative humidity, as seen in our study. Nevertheless, the inverse relationship between relative humidity and AHI may be confounded because its seasonal pattern was not in concordance with that of most air pollutants. In summary, there is still inadequate research to conclude the effect of relative humidity on AHI.

The associations between air pollutants and AHI were also inconsistent in previous studies.²⁰ The reasons for these inconsistencies include variations in participant characteristics (from either treatment-seeking patients or community population) and AHI assessment (by either standardized PSG in laboratories or by single channel at home). For example, ozone was associated with higher AHI in one study,¹⁶ but no association was found in another.⁵ In our study, ozone was associated with higher AHI only in patients with severe OSA and in women. Moreover, solar radiation-induced photochemical reaction is key to the formation of ozone.⁴¹ Variations in sun exposure to each study area may lead to different results, but the exact ozone level was not available or comparable in units in previous studies. In our study, PM10 levels were twice as high as in previous studies. Exposure to high PM10 was still associated with AHI only in patients with severe OSA. One study found that short-term variations in PM10 were significantly associated with AHI only in summer, when ambient ozone level was high.¹³ However, another female-predominant study with particulate matter levels similar to those of our study found a significant association between AHI and PM2.5 but not PM10.⁴² A low variability and high mean level of PM2.5 may have obscured its effect on AHI variations. Nevertheless, the effects of particulate air matter size fractions in mixture and interactions with season were complex and unclear.⁴³ The finding that SO₂ was associated with AHI was not consistent with findings from a previous study.⁵ Although the association between SO₂ and AHI is less studied, SO₂ exposure has been found to induce subjective upper airway symptoms and respiratory diseases.^44,45 In this study, we can only conclude that air pollutants affect certain groups (ie, female and patients with severe OSA). This conclusion is consistent with previous studies, which evidenced that patients with severe OSA were more susceptible to air pollution-associated blood pressure changes than those with an AHI less than 30.¹⁵ Likewise, women have been found to be more vulnerable than men to air pollutant-induced cardiovascular and respiratory diseases,^46,47 partly related to their susceptibility to autoimmune disorders.

A new finding in our study was that AHI was associated with air pollutants only during NREM sleep. Seasonal variation of AHI in NREM was more similar to AHI than was AHI in REM. Because sleep apnea in NREM sleep was less related to upper airway anatomical features than in REM sleep,²⁴ our results suggest that air pollutant-related sleep apnea induced sleep physiological changes rather than anatomical changes. However, it has not been studied whether air pollutants affect ventilator control system sensitivity, or arousal threshold. Air pollutants PM10 and PM2.5 have been found to cause inflammation and oxidative stress.⁴⁸ In addition to the well-known exposure consequences of respiratory and cardiovascular diseases, particulate matter is associated with neuroinflammation, brain oxidative stress, stroke, and neurodegenerative disorders.⁴⁹ However, no study has directly examined the effect of particulate matter on sleep physiology. Evidence for the influence of ozone and SO₂ on sleep is even more scarce.

This study has several limitations. First, the variance of AHI was retrieved from aggregated data from all patients visiting the sleep center at different times rather than from changes between two visits by the same individual. The results for seasonal patterns were confounded by other factors with seasonal variations, such as exercise levels, respiratory illness, and alcohol consumption. Second, the air condition data were collected from a surveillance station closest to the sleep center. Although most patients visiting the sleep center are local residents, they may be exposed to different air conditions at the places they live and work. We were not able to control indoor air conditions, either. Portable air condition monitoring devices are needed to collect more precise exposure data. Air condition data were limited by surveillance purpose of the Environmental Protection Agency. For example, atmospheric pressure was found to be associated with AHI,⁵ but we were unable to examine the effect. Furthermore, it has been reported that air conditions vary diurnally.⁵⁰ Future studies are needed to examine the temporal relationship of air condition exposure and changes of AHI. Last, the generalizability of this study should be limited to places with similar climate and air conditions.

In conclusion, this study used standard PSG study parameters and official air condition surveillance data to examine the seasonal variance and the effect of air condition on AHI in a subtropical city with warm climate and moderate level air pollution. We replicated previous findings that AHI had a seasonal pattern, but it was associated with air pollutants PM10, ozone, SO₂, and relative humidity only in patients with severe OSA and in women. Therefore, protective measures should be taken early for these vulnerable populations. We also determined that the association was preserved for AHI in NREM sleep and suggested that AHI in NREM sleep is more affected by ambient factors than AHI in REM sleep. The exact mechanism remains to be clarified.

DISCLOSURE STATEMENT

All authors have seen and approved the final manuscript. This study was supported by a research grant from China Medical University Hospital, Taiwan (DMR-99-113). This study was supported in part by the Taiwan Ministry of Health and Welfare Clinical Trial Center (MOHW107-TDU-B-212-113004). The funder had no role in the study design, the collection, analysis and interpretation of the data, the writing of the report, or the decision to submit the paper for publication. All authors report no conflicts of interest.

ABBREVIATIONS

AHI

apnea-hypopnea index

CPAP

continuous positive airway pressure

CO

carbon monoxide

NMHC

nonmethane hydrocarbons

NOx

nitrogen oxides

NREM

non-rapid eye movement

OSA

obstructive sleep apnea

PM10

particulate air matter less than or equal to 10 μm in aerodynamic diameter

PM2.5

particulate air matter less than or equal to 2.5 μm in aerodynamic diameter

PSG

polysomnography

REM

rapid eye movement

SO₂

sulfur dioxide

SpO₂

oxygen saturation