**Exhaled Breath Analysis in Obstructive Sleep Apnea Syndrome: A Review of the Literature**

#### **Panaiotis Finamore <sup>1</sup> , Simone Scarlata 1,\* , Vittorio Cardaci <sup>2</sup> and Ra**ff**aele Antonelli Incalzi <sup>1</sup>**


Received: 27 June 2019; Accepted: 22 August 2019; Published: 27 August 2019

**Abstract:** *Background and Objectives:* Obstructive sleep apnea syndrome (OSAS) represents an independent risk factor for cardiovascular, metabolic and neurological events. Polysomnography is the gold-standard for the diagnosis, however is expensive and time-consuming and not suitable for widespread use. Breath analysis is an innovative, non-invasive technique, able to provide clinically relevant information about OSAS. This systematic review was aimed to outline available evidence on the role of exhaled breath analysis in OSAS, taking into account the techniques' level of adherence to the recently proposed technical standards. *Materials and Methods:* Articles reporting original data on exhaled breath analysis in OSAS were identified through a computerized and manual literature search and screened. Duplicate publications, case reports, case series, conference papers, expert opinions, comments, reviews and meta-analysis were excluded. *Results:* Fractional exhaled Nitric Oxide (FeNO) is higher in OSAS patients than controls, however its absolute value is within reported normal ranges. FeNO association with AHI is controversial, as well as its change after continuous positive airway pressure (C-PAP) therapy. Exhaled breath condensate (EBC) is acid in OSAS, cytokines and oxidative stress markers are elevated, they positively correlate with AHI and normalize after treatment. The analysis of volatile organic compounds (VOCs) by spectrometry or electronic nose is able to discriminate OSAS from healthy controls. The main technical issues regards the dilution of EBC and the lack of external validation in VOCs studies. *Conclusions:* Exhaled breath analysis has a promising role in the understanding of mechanisms underpinning OSAS and has demonstrated a clinical relevance in identifying individuals affected by the disease, in assessing the response to treatment and, potentially, to monitor patient's adherence to mechanical ventilation. Albeit the majority of the technical standards proposed by the ERS committee have been followed by existing papers, further work is needed to uniform the methodology.

**Keywords:** obstructive sleep apnea; inflammation; FeNO; exhaled breath condensate; volatile organic compounds

#### **1. Introduction**

Obstructive sleep apnea syndrome (OSAS) is a highly prevalent sleep breathing disorder characterized by intermittent reduction (hypopnea) and/or cessation (apnea) of airflow due to upper airways collapse and represents an independent risk factor for cardiovascular [1,2], metabolic [3], neurological diseases [4,5], and motor vehicle accidents [6]. The disease is also common in children, with a prevalence of 1–4%, and associates with behavioral and cognitive deficits [7,8]. The exact mechanism underpinning these detrimental effects is still unknown, however the pro-inflammatory state and the oxidative stress likely due to the intermittent hypoxia are deemed to play a key role [9]; indeed, the use of a continue positive airways pressure ventilation (C-PAP) has demonstrated to be effective in reducing the airways collapse, minimizing the endothelial stress and, consequently, the pro-inflammatory state [10]. Given the severity of the complications, a correct diagnosis is warranted and the gold-standard is represented by polysomnography (PSG) [11] that, however requires specialized personnel and devoted setting which limits a wide use of the tool and compels to screen the population to refer to the specialist. Questionnaires are validated screening tools, however up to 45% of patients referred with the suspicion of OSAS are not confirmed by PSG [11,12], thus new approaches in identifying patients affected by OSAS need to be identified.

Exhaled breath is abundant in volatile organic compounds (VOCs), part of which are endogenous and produced by cellular metabolism. Exhaled breath analysis, proved to detect the metabolic changes induced by OSAS, can be applied as a non-invasive tool able to shed light on the pathways modified by the disease, and also to provide a more rapid and economic instrument for diagnosis, monitoring and, eventually, characterization of the disease. Systematic reviews in this field of research are already available in literature [13,14], but, recently, several studies have been published that have enriched the available amount of evidence; furthermore, all the available reviews preceded the recently published European Respiratory Society (ERS) statement about the technical standards to follow in the exhaled breath analysis published in 2017 [15] and is therefore unclear, at the moment, to which extent the previous works adhered such methodological standards.

The aim of this systematic review is therefore to outline the newly available evidences on the exhaled breath analysis role in OSAS, taking into account whether they conform to the proposed ERS technical standards.

#### **2. Materials and Methods**

We performed a computerized and manual literature search on PubMed, limited to English language articles published up to May 2019, to identify articles reporting original data on exhaled breath analysis in obstructive sleep apnea. We entered the following MeSH terms: Obstructive Sleep Apnea; Obstructive Sleep Apneas Syndrome; OSA; OSAS; in combination with: volatile organic compounds; VOC; electronic nose; gas chromatography mass spectrometry; spectrometry; exhaled breath condensate; EBC; nitric oxide; FeNO. Two authors (P.F. and S.S.) performed the literature search and assessed the eligibility of identified publications independently. All studies that evaluated exhaled breath analysis in OSAS were screened. Duplicate publications, case reports, case series, conference papers, expert opinions, comments, reviews and meta-analysis were excluded. The selection process is summarized in Figure 1. The literature search has been integrated with other relevant studies about methodological and clinical issues.

**Figure 1.** PRISMA diagram showing the flow of information through the different phases of the reviewing process.

#### **3. Results**

The thirty-six studies included in the review encompass the three main domains of exhaled breath analysis: the fractional exhaled nitric oxide (FeNO), the exhaled breath condensate (EBC) and the exhaled VOCs. The characteristics of the main studies included in the review are summarized in Tables 1–3.

#### *3.1. FeNO and Exhaled Carbon Monoxide (eCO)*

Nitric oxide (NO) is a gaseous molecule produced by nitric oxide synthase (NOS) enzymes from L-arginine and oxygen. There are three isoforms of NOS, two are constitutively produced (endothelial NOS–eNOS– and neuronal NOS–nNOS–) and one is inducible (iNOS), increasing during inflammation [16], as that characterizing airways in asthmatic patients. Indeed, the FeNO in the gas phase emerged in the last decade of the last century as an innovative diagnostic marker of asthma [17,18]. Being non-invasive and easy to perform, FeNO raised a wide interest, allowing a deeper understanding of mechanisms underpinning its production and addressing technical issues related its measurement. Nowadays, FeNO is considered a marker of T-helper 2 cell-type inflammation, rather than a marker of asthma per se, and a marker of response to corticosteroid treatment in those patients [19].

The study of FeNO in the diagnosis of OSAS has led to contradictory findings. Indeed, while some studies described a raising of FeNO level in OSAS [20–25], the majority did not confirm the finding [26–30] or just showed a higher concentration in OSAS patients when compared with non-obese healthy controls [31–33]. Besides, even considering only those studies with a positive finding, the FeNO level, albeit statistically higher than healthy controls, did not reach a clinical significance. Indeed, in all studies the mean FeNO expressed in part per billion (ppb) was below 30 ppb, which means that OSAS patients are classified in the group of individuals without airway inflammation (or without eosinophilic inflammation) or in the grey zone between 25 and 50 ppb according to the ATS guidelines [19], the same groups of healthy controls. One possible explanation of the low level of FeNO despite the inflammatory state can be the different location of the process, closer to the alveoli than the airways or in the opposite, as the result of a topical, mechanically induced inflammation at the level of the upper airway caused by snoring and apnea associated mechanical stress [34,35]. Indeed, international guidelines suggest to use a flow of 50 mL/s for the measurement of FeNO, however it is not high enough to allow the collection of the alveolar portion of NO [36]. Albeit some studies have found a statistically significant higher concentration of exhaled nitric oxide (eNO) at a flow of 250 mL/s or more in association with an elevated concentration of NO in the gas phase of Alveoli (CaNO) [22,25,37], Fortuna and colleagues reported a lower CaNO in OSAS patients than healthy controls [23] and Foresi and colleagues did not find a difference in CaNO between normotensive OSAS patients and controls [30]. The more validated hypothesis is that the increased inflammation damages the alveolar endothelium reducing the expression of the eNOS and the diffusion of NO [38]. Mechanisms of inflammation induced by OSAS are reproduced in Figure 2.

**Figure 2.** Principal inflammatory pathways induced by OSAS.

Furthermore, it is still unclear whether an overnight change in the production of eNO exists or not. While some studies reported an overnight increase in FeNO [20,24] and in the concentration of nitric oxide exhaled by the nose (nasal nitric oxide–nNO–) and by the mouth (orale nitrix oxide–oNO–) [39] in OSAS patients [20,39], other studies failed to confirm the evidence [21], or they found an overnight increase limited to subgroups of OSAS, such as obese OSAS patients [29] or children with mild OSAS but not moderate/severe [28], or healthy controls [39].

Finally, eNO has been proposed as a marker to monitor the efficacy of C-PAP therapy. Indeed, evidence in literature suggests that one-to-three month C-PAP treatment is effective in reducing FeNO [22–24] and increasing CaNO [23]. The effect should also be time-dependent, at least for

FeNO, since a single or 2-nigth treatment with C-PAP increases CaNO [30,40] but do not reduce FeNO [30]. This suggest that C-PAP, normalizing oxygen saturation, reduces inflammation and oxidative stress, promoting alveolar endothelial function and therefore candidates CaNO as a marker of endothelial function.

Even the association of the eNO with the apnea-hypopnea index (AHI) is controversial. Indeed, while some studies found a strong and positive correlation between FeNO and AHI, with a *r* of 0.8–0.9 [23,33], or oNO and AHI (*r*: 0.46) [32] and a negative one between CaNO and AHI, with a *r* of 0.9 [23], this was not confirmed by other studies [20–22,27,28,39,41].

Knowledge about exhaled carbon monoxide (eCO) in OSAS is more limited than FeNO. To the extent possible, eCO has been reported higher only in severe OSAS [42], it has a weak correlation with AHI [42] and it is not normalized after one-month of C-PAP [22], probably because it needs a longer period to be normalized.

#### *3.2. Exhaled Breath Condensate*

The alveolar and airway lining fluids (ALF) contain hydrophobic and hydrophilic nonvolatile and volatile compounds which are continuously released into the environment as droplets created during breathing. In contrast to bronchoalveolar lavage, EBC is a noninvasive way to sample these compounds by directing the exhaled breath through a cooling device. The sample, mostly composed by water vapour, can be stored or immediately analyzed. Albeit noninvasive, EBC composition is highly influenced by the collection and the condenser procedure, which undermine the reliability of the achieved results. Principles of functioning of exhaled breath condensate technology is summarized in Figure 3.

**Figure 3.** Principles of functioning of exhaled breath condensate technology.

#### 3.2.1. EBC pH

Given the inflammatory and pro-inflammatory state characterizing OSAS, EBC pH in OSAS was expected to be lower than healthy controls. The hypothesis has been confirmed by all the studies carried out so far, with the exception of that by Greulich and colleagues [43], with a mean absolute value of EBC pH in OSAS around 7.4, by far below the first quartile of EBC pH distribution in healthy subjects and equal to the fifth percentile [44]. pH has shown a negative correlation with AHI (*r*: −0.66), sleep time with a SaO2 < 90% (*r*: −0.62) and neck circumference (*r*: −0.63) [31], but also with body-mass index (BMI) (*r*: −0.54). Although Petrosyan and colleagues demonstrated that OSAS EBC pH is lower than controls, even if obese [22], the finding has not been confirmed by Carpagnano et al. [31], raising doubts about the association between EBC acidity and OSAS. Albeit it is not possible to exclude that obesity, rather than OSAS, reduces EBC pH, probably by increasing the likelihood to have gastro-esophageal reflux, it seems that EBC acidity is due to OSAS. Indeed, after the treatment with C-PAP EBC pH increases [22], becoming closer to normal reference values. A change of the EBC pH after C-PAP or surgical treatment has not been confirmed by other studies [43,45], however in both cases the EBC pH value of OSAS patients was already normal at baseline. No significant difference has been found between OSA smokers and non-smokers [46]. To conclude, all studies analyzing EBC pH performed de-aeration before the analysis, but did not performed the analysis in real time or immediately after collection without freezing or storing EBC, as suggested by international guidelines [15]. OSAS seems to increase EBC acidity, however exist a variability in the EBC pH that compels to investigate the effect of other factors.

#### 3.2.2. EBC Cytokines

EBC cytokine level has been studied in OSAS patients. As expected, all studies confirmed that the concentration of IL-6, TNF-α, IL-8 and ICAM-1 is higher than healthy controls, while IL-10 concentration, which has anti-inflammatory properties, is lower [46–49]. However, there is a wide range of cytokine concentrations among the studies: indeed, while the mean EBC IL-6 concentration was in the order of decades of pg/mL in some studies [47,48], it was below the unit in other studies [46,50], notwithstanding the concentration was expressed in the same unit of measurement. Similarly, the concentration of TNF-α in the studies of Li and colleagues [48,51] was ten times the concentration of TNF-α in the study of Antonopoulou and colleagues [46]. Hence, even pro-inflammatory cytokines seem elevated in OSAS and anti-inflammatory cytokines reduced, sampling procedure should be revised, because confounding factors, as dilution, seem to have affected the absolute value. Other confounding factors to take into account are obesity and smoking. Indeed, while some studies do not report a difference in IL-6 level between smoking and non-smoking OSAS patients [46], other studies suggest a pro-inflammatory effect of smoking [48]. Noteworthy, no doubts are on the pro-inflammatory role of obesity, with all studies confirming an elevated concentration of EBC IL-6, IL-8 and ICAM-1 in obese than normal weight individuals [47,49]. Being inflammation in OSAS closely related with intermittent hypoxia, it is not surprising that AHI was positively correlated with EBC IL-6 (*r*: 0.6−0.8) [47,48], ICAM-1 (*r*: 0.7) [49] and TNF-α (*r*: 0.85) [48] and negatively correlated with EBC IL-10 (*r*: −0.63) [51]. As expected, EBC IL-6 also positively correlated with the neck circumference (*r*: 0.5) [47]. EBC cytokines are stable over time if patients do not start a treatment [51], while effective treatment reduces their concentration. Indeed, even with different absolute values, two studies demonstrating the effectiveness of C-PAP therapy [50,51], but also the positive role of oral appliances and surgery in abating inflammation and thus EBC cytokine concentration [51].

#### 3.2.3. EBC Oxidative Stress

The EBC concentration of 8-isoprostane, a product of the lipid peroxidation of arachidonic acid and marker of oxidative stress, has been repeatedly found elevated in adult patients affected by OSAS [22,46–48,52,53], and in children [28]. The mean value in OSAS patients is heterogeneous, ranging from 6 to more than 30 pg/mL, and overlaps with the mean values observed in healthy controls [46,48]. Smoking seems to affect the marker concentration [48], while the role of obesity is conflicting. Indeed, while Petrosyan and colleagues found a higher level of 8-isoprostante in healthy non obese than obese individuals, both were significantly lower than OSAS patients [22], Carpagnano and colleagues observed exactly the opposite, also failing to discriminate OSAS from obese controls by 8-isoprostane concentration [47]. 8-isoprostane has shown a positive correlation with AHI, with a *r* of 0.4−0.5, [22,28,46–48,52,53] and neck circumference (*r*: 0.5−0.6) [47,52]. Interestingly, the concentration of 8-isoprostane is higher in the morning than in the evening in OSAS patients, with the latter similar to the concentration of healthy controls [52]. C-PAP therapy is effective in reducing the concentration of 8-isoprostane, but it is also reduced by oral appliances and surgery [50–52].

More limited evidence exists on the EBC concentration of hydrogen peroxide (H2O2). To the extent possible, H2O2 seems elevated in OSAS [22,54], regardless of patient's BMI [22]. Noteworthy, obesity is associated with an increase in the H2O2 concentration in healthy controls [22]. H2O2 is also positively associated with the AHI, with the same correlation of 8-isoprostante [22], and thus with the severity of the disease, being higher in patients with moderate to severe than mild OSAS [54]. This marker is not modified by one month of C-PAP therapy [22]. While Petrosyan and colleagues clearly recommended the use of a filter on the inspiratory valve to avoid an environmental conditioning [22], it is not clear whether Malakasoti and colleagues did the same [54]. Both studies did not perform the measurement of H2O2 immediately after the collection, as suggested by the ERS guidelines [15].

#### 3.2.4. Other EBC Markers

Other markers assessed in the EBC of OSAS patients are: urates, leukotrienes and leptin. EBC concentration of acid uric, which has antioxidant capacity, has been studied in children and resulted significantly higher than healthy controls [55], probably having a role in contrasting the increased oxidative stress driven by the disease. Similarly, leukotrienes (leukotriene B4, which is also associated with the severity of the disease [22,56] and leukotriene C4/D4/E4), lipid mediators prompting inflammation, are elevated in OSAS, even though with a completely different absolute value in pg/mL among studies. Indeed, the concentration found in one study in OSAS patients completely overlaps with that found in healthy controls in another study [22,56]. Contrary to the expectations, prostaglandins (PGE2) did not show any difference between children affected by OSAS and controls [56]. Furthermore, no role seems to have leptin as an EBC biomarker of OSAS. Indeed, while obese OSAS patients have higher concentration than controls, non-obese OSAS and obese controls have the same concentration, suggesting, together with a strong and positive correlation with BMI, that obesity rather than OSAS affects the concentration of this mediator [57].

#### *3.3. Volatile Organic Compounds: Spectrometry and Electronic Nose*

Exhaled breath is abundant in VOCs with very low concentration, most of which are undetectable by the human nose. These molecules in part originate from the endogenous metabolism and human gut and airway microbiome [58], thus their study might provide information about any diseases threatening the internal homeostasis and thus help address their diagnosis, disease severity stratification and prognosis, as already demonstrated in other respiratory diseases [59]. To date, there exist two main approaches to the study of VOCs: the first aims to identify single biomarkers related to the disease in the mixture of molecules and it is based on the use of spectrometry, often coupled with separation techniques as gas-chromatography; the second is aimed to identify a pattern in the mixture able to discriminate, through the use of a pattern-recognition approach, the disease from other conditions and it is based on the use of electronic-noses. Both have been applied in the study of OSAS, either alone or in association.

The use of analytical techniques have demonstrated a good accuracy in discriminating OSAS patients from healthy controls [60], even if obese [61]. However, no study has so far identified a single molecule able to discriminate OSAS from controls, thus discrimination is based on a set of VOCs. Greulich and colleagues reported in their study an increase in OSAS of 2-methylfuran, 2-(methylthio)-ethanol and hexanal and a reduction in 3-methylbutanal or 3-methylbutyraldehyde and acetone [60]. Interestingly, an increase in 2-methylfuran in serum and pharyngeal wash of those patients was also reported. However, none of the compounds described by Greulich were also identified by Dragonieri and colleagues, who reported a good discriminative capacity between OSAS and obese controls basing on the following compounds: tetrachloroethene, 2,3,5-trimethylhexane, β-pinene, 1,3,5-trimethylbenzene, 9-methylacridine, tetradecane, 6,10-dimethyl-5,9-undecadien-2-one and β-ionone [61]. Besides, Aoki and colleagues found that although almost all the aromatic and satured hydrocarbons are more expressed in the exhaled breath of severe OSAS patients, only isoprene is always elevated in OSAS, regardless the severity of the disease [62].

A good discriminative accuracy in discriminating OSAS from normal weight controls and chronic obstructive pulmonary disease (COPD) patients has also been demonstrated by the use of electronic noses, which showed a lower accuracy in discriminating people affected by the disease from healthy obese controls [43,63–65]. As already observed for other exhaled breath markers (e.g., 8-isoprostane), the breath pattern changed overnight in OSAS patients but not in controls, likely due to the inflammation and oxidative stress promoted by the intermittent hypoxia; indeed there was a difference in breath pattern between OSAS and controls only in the morning. Noteworthy, the difference is still present after the exclusion of patients suffering from gastro-esophageal reflux and COPD [66]. The finding is in line with that of Olopade and colleagues who reported a higher concentration of oral pentane in the morning than in the evening [39]. While some studies found a positive correlation between the breath pattern and the AHI [43], other studies failed to confirm the finding [66]. Albeit apparently contradictory, it is possible that the association between AHI and breath pattern is mediated by patients' comorbidities, as suggested by Incalzi and colleagues [67]. Breath-pattern is sensitive to the effects of the C-PAP therapy, indeed concentrations of isoprene and acetone decrease [62] and it is possible to discriminate treated and untreated patients with good accuracy [68]; even a single night treatment is associated with a change in the breath pattern. Interestingly, the breath pattern change does not have the same characteristics in all OSAS patients, with two different types of response being distinguished depending on the comorbidities of those individuals [67]. Noteworthily, almost all the studies did not perform an external validation of the discriminative model, hence it is not possible to exclude an overfitting of the models, even though minimized by the use of internal cross-validation. Technical and operative descriptions of these approaches have been summarized in Figure 4 and discussed in detail elsewhere [69,70].











hypoxemic; non-hypo: non hypoxemic;

✔: technical standard satisfied; **X:** technical standard not satisfied.

**Figure 4.** Measure chain of an e-nose based sensor system.

#### **4. Discussion**

This updated systematic review confirms the promising role of exhaled breath analysis in the understanding of the mechanisms underpinning disease and its clinical relevance in identifying individuals affected by OSAS. Besides, in addition to previous reviews of the field, it shows that, although the majority of the technical standards proposed by the ERS committee have been followed, more research is needed to stadardize the methodology and hence reduce the variability in the results.

OSAS is characterized by an endothelial dysfunction, arterial stiffening and elevated levels of inflammatory markers as an effect of the intermittent hypoxia caused by the upper airways collapse [71] which increase the risk to develop cardiovascular, metabolic or neurological events. Indeed, hypoxia increases the production of reactive oxygen species (ROS) and thus the oxidative stress, which impairs the phosphorylation of NOS [72,73], reduces the release of nitric oxide and promotes the endothelial dysfunction. Results of studies on FeNO are in line with this notion. Indeed, overall the concentration of FeNO measured at a flow of 50 mL/s is below the 50 ppb, identified by the American Thoracic Society (ATS) as a threshold of the presence of eosinophil airway inflammation. Moreover, the reduced CaNO in the studies of Fortuna and Foresi and its elevation after effective treatment support the existence of an alveolar damage in the disease [23,30]. Furthermore, intermittent hypoxia fosters the development of a chronic inflammation, and this is confirmed by the studies carried out on the EBC. Indeed, pro-inflammatory cytokines increase while anti-inflammatory cytokines decrease in the breath of those patients, and the markers of oxidative stress are elevated in the morning [39,52], as demonstrated also by the studies on the breath pattern [66]. Moreover, inflammatory cells were increased in the muscular layer of patients with OSAS, with CD4+ and activated CD25+ T cells (both increased approximately threefold) predominating. Inflammation was also present in upper airway (UA) mucosa, but with a different pattern consisting of CD8+ (2.8-fold increase) and activated CD25+ (3.2-fold increase) T cell predominance, suggesting that inflammatory cell infiltration affects not only the mucosa, but also the UA muscle of patients with OSAS, this potentially leading to a systemic pro-inflammatory spillover of cytokines and mediators that could promote and amplify chronic inflammatory response [35]. Indeed, these proposed mechanisms are still far from being confirmed and further research is needed to confirm this pathophysiologic mechanism.

Although all the techniques studying volatile and non-volatile compounds are able to discriminate OSAS patients from controls, EBC and the study of volatile organic compounds seem more promising than FeNO for a clinical use. However, efforts are needed to address some the technical and non-technical issues that are hindering the applicability of breath analysis in clinical practice. The role of smoking in increasing inflammation, as well as that of obesity, should be deeper investigated in the studies about OSAS. Besides, issues as the dilution of the EBC [74] or the lack of external validity in most of the studies about volatile organic compounds need to be addressed to increase the reliability of the techniques.

Breathprint analysis of VOCs might have practical applications and could act as a valuable instrument in OSAS management in the next future: considering the high prevalence of OSAS in the general population and its dramatic impact on health status, any effort should be made in order to detect and treat it as soon as possible. Breathprint analysis might complement, or even replace questionnaires in the screening process and, consequently, improve the cost/effectiveness ratio of polysomnography. Furthermore, VOCs analysis could be used to monitor the response to, and the adherence with C-PAP ventilation [57]. Finally, the breath print analysis could help better understanding of the heterogeneity of OSAS phenotypes [69] and define their prognosis, as in other respiratory diseases [75].

#### **5. Conclusions**

To conclude, in the era of precision medicine breath analysis, being non-invasiveness, rapid and economic, might play a key role in the understanding of the pathways underpinning OSAS and in the clinical management of the patients affected by the disease.

**Author Contributions:** P.F., S.S. and R.A.I. participated in the study concept and design. P.F. and S.S. performed the literature search and assessed the eligibility of identified publications independently, R.A.I. and V.C. reviewed the manuscript for important intellectual content. All the authors fulfil authorship criteria, have revised the final version of the manuscript and gave their consent to publication.

**Funding:** The present study has not received any funding.

**Conflicts of Interest:** Authors deny any conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Brief Report*

### **Risk Assessment for Self Reported Obstructive Sleep Apnea and Excessive Daytime Sleepiness in a Greek Nursing Sta**ff **Population**

#### **Alexia Alexandropoulou 1, Georgios D. Vavougios 2, Chrissi Hatzoglou 1,3, Konstantinos I. Gourgoulianis <sup>3</sup> and Sotirios G. Zarogiannis 1,3,\***


Received: 15 June 2019; Accepted: 5 August 2019; Published: 12 August 2019

**Abstract:** *Background and objectives*: The risk assessment of Obstructive Sleep Apnea (OSA) and Excessive Daytime Sleepiness (EDS) in specific occupational populations is important due to its association with morbidity. The aim of the present study was to identify the risk of OSA development and EDS in a Greek nursing staff population. *Materials and Methods*: In this cross-sectional study a total of 444 nurses, 56 males (age = 42.91 ± 5.76 years/BMI = 27.17 ± 4.32) and 388 females (age = 41.41 ± 5.92 years/BMI = 25.08 ± 4.43) working in a Greek secondary and tertiary hospital participated during the period from 18 January 2015 to 10 February 2015. The participants completed the Berlin Questionnaire (BQ), concerning the risk for OSA and the Epworth Sleepiness Scale (ESS), concerning the EDS. The work and lifestyle habits of the participants were correlated with the results of the questionnaires. *Results*: According to the BQ results 20.5% (*n* = 91) of the nursing staff was at high risk for OSA. Increased daytime sleepiness affected 27.7% (*n* = 123) of the nurses according to ESS results. Nurses at risk for Obstructive Sleep Apnea Syndrome (OSAS), positive for both BQ and ESS, were 7.66% (*n* = 34). Out of the nurses that participated 77% (*n* = 342) were working in shifts status and had significant meal instability (breakfast *p* < 0.0001, lunch *p* < 0.0001, dinner *p* = 0.0008). *Conclusions*: The population at high risk for OSA and EDS in the nursing staff was found to be 20% and 28% respectively. High risk for OSAS was detected in 7.66% of the participants. The high risk for OSA and EDS was the same irrespective of working in shift status. In specific, nursing population age was an independent predictor for high risk for OSA and skipping lunch an independent predictor of daytime sleepiness.

**Keywords:** Berlin Questionnaire; Epworth Sleepiness Scale; nursing staff; Obstructive Sleep Apnea Syndrome; risk assessment

#### **1. Introduction**

The World Health Organization (WHO) indicates that Obstructive Sleep Apnea Syndrome (OSAS) is a preventable lung disease [1]. Most patients with this syndrome exhibit no detectable respiratory dysfunction when awake while OSAS appears in all age groups. However, in the adult population the incidence of this syndrome increases with age and is clearly linked with excessive daytime sleepiness (EDS) [1–3].

The prevalence of this syndrome is probably higher than the one presumed due to underdiagnosis. Thus, OSAS constitutes an important public health issue [4–7]. It is estimated that 26% of the worldwide adult population is at high risk for developing the syndrome [4]. Epidemiological studies

indicate that the exact determination of OSAS prevalence is difficult due to different methodological approaches [3,8,9]. Studies from USA, Australia, India, China and Korea report that the prevalence in the general adult population spans from 3 to 7% in men and 2 to 5% in women to more than 49% depending on age and gender [5–7,9–15]. This reported non-uniformity of the prevalence in 4 different continents strengthens the notion that the disease is common but the prevalence in the community needs to be studied more rigorously in order to avoid underdiagnosis [4,8].

The investigation of OSAS in the context of specific occupations is of high interest given that its occurrence may be associated with working conditions that do not only induce the disease but also affect the job performance and overall health [16–18]. A study on American and Canadian police officers showed that this specific population has at least one sleep disorder. Moreover, one third of the study population, suffered from OSAS (33.6%) [19]. Another study in young male Korean soldiers using the Berlin Questionnaire (BQ) reported a prevalence of 8.1% of OSA [20]. In a similar study conducted in the staff of an Iranian hospital again with the BQ tool, it was found that 6.9% were at high risk for OSAS. Finally, a study in 21 nurses, showed that according to the BQ, 24% were at high risk for OSA, but subsequent polysomnography revealed that 43% of them were diagnosed with OSAS [16].

No studies exist in the Greek population regarding the risk assessment of self-reported OSA and EDS in specific occupational groups. Given that the nursing staff work several times in shift status, which has been implicated in the induction of OSAS [21,22], we hypothesized that this population would be under risk for developing OSA and EDS, due to their sleep fragmentation. Thus, the aim of the present study was to identify the nursing population at high risk for OSA and EDS in a secondary and tertiary hospital in Greece.

#### **2. Materials and Methods**

#### *2.1. Study Population*

The study population consisted of 444 nurses working in the University Hospital and the General Hospital of Larissa during the period 18 January 2015 to 10 February 2015 who volunteered to participate in the study. In total, 530 questionnaires were distributed by the primary author in personal communication with the potential participants. The potential participants were given a week to complete the questionnaires and were asked to put them in an un-named envelope and hand them to a designated administrative officer of each hospital sector (Medical, Surgical and Intensive Care) for collection by the primary author. Out of the 530 questionnaires 449 were returned to the primary author. Out of the 449 questionnaires, 5 were not fully completed and were thus excluded from the study, leading to a final number of 444 questionnaires included in the study. The study involved 56 male (12.6%) and 388 female (87.4%) nurses, regardless of educational level and work experience, from all nursing departments. All participants provided demographic information such as gender, age, height, weight, smoking habits, skipping meals and whether they worked under night shifts. The Ethics Committee of the University Hospital of Larissa approved the research protocol (Protocol number: 1/14-1-2015).

#### *2.2. OSAS and EDS Assessment Tools*

In order to assess the risk of OSAS the Greek version of the Berlin Questionnaire (BQ) was used [23]. The BQ contains 10 questions that are divided in 3 categories. In the first category the questions aim at identifying the self-reported snoring behavior along with witnessed apneas during sleep by the partner. The second category assesses self-reported fatigue after sleep and the third assesses the presence of obesity or history of hypertension. If two of the categories of the BQ are positive, then the participant is assigned as being at high risk for OSAS.

In order to assess the excessive daytime sleepiness, the Greek version of the Epworth Sleepiness Scale (ESS) was used [24]. ESS aims at the quantification of daytime sleepiness though a set of self-reported incidents of dozing in eight different setting during the day and the scoring spans from 0 to 24. A participant in high risk for daytime sleepiness has a score of 10 or higher.

#### *2.3. Statistical Analysis*

Statistical analysis was performed by EpiInfo v. 7.0 (CDC, Atlanta, GA, USA), the SPSS 24.0 Software (IBM Corporation, New York, NY, USA) and GraphPad Prism v. 8.1 (San Diego, CA, USA). Fisher's exact was used to assess differences among proportions. The Mann Whitney test was used to assess differences between two groups. Multivariate logistic regression was performed as in other similar studies [25]. The Forward Conditional Logistic Regression Model was used to perform multivariate analyses of the effect of univariate predictors on the likelihood of belonging to the high OSA probability (based on BQ) or high daytime sleepiness (Based on Epworth Scale) groups, while controlling for potential confounders. Values are expressed as mean ± S.D. A *p* value of less than 0.05 was deemed significant.

#### **3. Results**

#### *3.1. Study Population*

Out of the 530 questionnaires that were distributed, 444 were completed and collected, providing a responsiveness rate of 83.8%. The demographics of the study population along with lifestyle habits are shown in Table 1.


**Table 1.** Characteristics of the participants in the study.

#### *3.2. BQ and ESS Questionnaire Results*

The results of the BQ questionnaire showed that 20% (*n* = 91) of the participants were found to be at high risk for OSAS as opposed to 80% (353) that were found to be at low risk for OSAS (Figure 1A). With regards to the ESS questionnaire 28% (*n* = 123) of the participants were found to be at high risk for EDS as opposed to the 72% (*n* = 321) that were found at low risk (Figure 1B).

More importantly a fraction of these two groups mounting to 8% (*n* = 34) were found to be at concomitant high risk for both OSA and EDS, thus at OSAS risk.

There were no differences in the proportions of male and female nurses that were positive in BQ (*p* > 0.99) or ESS (*p* = 0.75). Working under night shift status did not result in a higher proportion of nurses to test positive in BQ (*p* = 0.69), but resulted in a higher proportion of nurses testing positive in ESS (*p* = 0.005).

Another significant finding of our study was that the nursing staff working on shift work status reported skipping meals significantly more than the nurses not under shift status that had a greater stability in maintaining the three main meals of the day as shown in Figure 2.

**Figure 1.** (**A**) Results of the Berlin Questionnaire (BQ) showing 20% of the participants at high risk for Obstructive Sleep Apnea (OSA). (**B**) Results of the Epworth Sleepiness Scale (ESS) showing 28% of the participants at high risk for Excessive Daytime Sleepiness (EDS).

**Figure 2.** Significant meal instability (skipping of a meal) in the nursing staff working on shift work status regarding (**A**) breakfast, (**B**) lunch and (**C**) dinner.

The Forward Conditional Binary Logistic Regression model was subsequently used in order to determine the effects of age, sex, alcohol consumption, education level, smoking status and breakfast/lunch/dinner skipping on the likelihood of belonging to the (a) high OSA risk and (b) high daytime sleepiness groups. Age was the single independent predictor of belonging to the high risk OSA group [OR: 0.959 (95% CI: 0.922–0.998), *p*-value = 0.038], whereas lunch skipping [OR: 1.631, (95% CI: 1.060–2.509), *p* = 0.026] independently predicted higher daytime sleepiness.

#### **4. Discussion**

The aim of the present investigation was the identification of the risk for OSA and EDS in the nursing population of a secondary and tertiary hospital in Greece using the standard questionnaires. Our results showed that the nursing population at high risk for OSA was 20%, while that of EDS was 28%. The fraction of the study population that was at high risk for both, and therefore at high OSAS

risk, was 8%. Moreover, according to our results, working in shift status did not directly affect the risk for OSA. However, it significantly worsened EDS in the nursing population. It has to be taken into account that the sensitivity and specificity of BQ for OSA diagnosis in the Greek population has been shown to be 76% and 40%, respectively, while the Greek version of ESS had also proved a useful tool for the identification of EDS in Greece [23,24]. In the studied population after multivariate logistic regression it was shown that age was a significant predictor of high risk for OSA, while skipping lunch was an independent prognosticator for high EDS risk. Aging is a known risk factor for OSA development so our result is in line with the literature [26]. As far as lunch skipping is concerned, a study that focused on the daytime sleepiness of subjects during the Ramadan intermittent fasting showed that this intentional prolonged daytime abstinence from food intake, induced an increase in the objective and subjective daytime sleepiness of the subjects [27]. Thus, based on the ESS scores of the participants of our study, we are in agreement with the notion that daytime food abstinence increases the propensity for subjective daytime sleepiness.

There is lack of studies on the risk assessment for OSA and EDS in nursing populations in Greece, so our results cannot be compared to the published literature. However, a similar study performed in the USA in nurses working in shift status using the BQ, showed that 24% of participants were at risk for OSA [16]. Although our study had a sample size nearly 20-fold bigger the above-mentioned results are comparable to ours that showed that 20% of the population was at high risk for OSA. On the other hand, in the study of Geiger-Brown et al., after polysomnography 43% of the participating nurses were diagnosed with sleep-disordered breathing, therefore if we extrapolate these findings to our study, we should expect a significantly higher number of nurses with OSAS in our sample. This was a limitation of our study but is the topic of a new investigation currently underway. A previous Greek study has shown that in subjects that underwent polysomnography, OSAS was diagnosed five times more in men than women, nevertheless using the BQ we did not detect such a difference between genders [28].

Data stemming from other occupational groups have reported comparable results. A cross-sectional and prospective cohort study in police officers in North America that involved a 10-fold greater population (*n* = 4957) than ours indicated that 40.1% of the police staff had at least one sleep disorder [29]. The most important of these disorders that was observed in 33.6% was OSAS. A total of 28.5% of the police staff also had EDS and the significant possibility to sleep during driving (once a month).

The failure of sleep replenishment during the day and the abnormality of melatonin levels affects the daily fatigue and reduces the quality of life [18]. In our study, 28% of the nurses were found to have EDS according to the ESS results. This finding is alerting under the rationale that sleep deprivation can lead to reduced concentration and productivity, and also in increased traffic and occupational accidents and injuries, as well as chronic diseases (cardiovascular and metabolic) and reduced quality of life [16–18,21,22]. Although no relevant data exists in the literature regarding Greece in order to be able to compare our findings, we are in good agreement with studies performed in nurses in New Zealand (that reported 33.75% of positive ESS) and Sweden (that reported 32.5% of positive ESS with a cut-off of 9 that was different from the one in our study that was 10) [30,31]. Our results were higher than a recent study performed in China that reported 16.1% positive ESS but the cut-off the authors used was 14, therefore these results are not directly comparable with the current study [32]. Another important finding of our study was that the nursing population working in shift status had significant instability in all three main meals of the day as determined by the self-report of participants of meal skipping. It has been reported that the instability of meals induces increases in body weight and thus BMI [18]. Obesity is a known risk factor for OSAS and furthermore a specific type of sleep apnea is observed in this population, the Obesity Hypoventilation Syndrome (OHS) [3,16]. Indeed, in our study the participants that were found to be in the high risk for OSA based on BQ results were significantly heavier that the ones in the low-risk group. Moreover, shift status has been implicated in the induction of OSAS [21,22]. Shift-work disrupts the expression of circadian genes and sleep patterns,

deregulates metabolic processes and can cause sleep apnea and several disorders linked to OSAS like cardiovascular disorders and obesity [33].

There were some limitations in the current study. The population of our study was young since most nurses were in their early forties predominantly. Additionally, our sample comprised predominantly of females and this may have diluted the significance of our results. Finally, all participants were originating from a single geographic area and thus further multicenter studies are needed.

#### **5. Conclusions**

In conclusion, we found that the risk for OSA and EDS in a nursing population of a secondary and tertiary hospital in Greece was 20% and 28% respectively. At high risk for OSAS were 8% of the participants (positive BQ and ESS simultaneously). Moreover, we found that nurses that work under night shift status had significant meal instability, which is a risk factor of obesity, which is in turn linked to OSAS development. However, we detected no differences in OSAS risk between these two groups of the population assessed. Further study of the population under high risk for OSAS of this study involving polysomnography assessment is needed.

**Author Contributions:** Conceptualization, S.G.Z.; data curation, A.A. and S.G.Z.; formal analysis, A.A.; investigation, A.A.; methodology, G.D.V., C.H., K.I.G. and S.G.Z.; project administration, S.G.Z.; resources, C.H., K.I.G. and S.G.Z.; supervision, S.G.Z.; visualization, A.A. and S.G.Z.; writing—original draft preparation, A.A. and S.G.Z.; writing—review and editing, A.A., G.D.V., C.H., K.I.G. and S.G.Z.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Medicina* Editorial Office E-mail: medicina@mdpi.com www.mdpi.com/journal/medicina

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18