Assessment of Obstructive Sleep Apnea in Association with Severity of COVID-19: A Prospective Observational Study.

INTRODUCTION: OSA has been postulated to be associated with mortality in COVID-19, but studies are lacking thereof. This study was done to estimate the prevalence of OSA in patients with COVID-19 using various screening questionnaires and to assess effect of OSA on outcome of disease.
METHODOLOGY: In this prospective observational study, consecutive patients with RT-PCR confirmed COVID-19 were screened for OSA by different questionnaires (STOPBANG, Berlin Questionnaire, NoSAS, and Epworth Scale). Association between OSA, outcome (mortality) and requirement for respiratory support was assessed.
RESULTS: In study of 213 patients; screening questionnaires for OSA [STOPBANG, Berlin Questionnaire (BQ), NoSAS] were more likely to be positive in patients who died compared to patients who survived. On binary logistic regression analysis, age ≥ 55 and STOPBANG score ≥ 5 were found to have small positive but independent effect on mortality even after adjusting for other variables. Proportion of patients who were classified as high risk for OSA by various OSA screening tools significantly increased with increasing respiratory support (p < 0.001 for STOPBANG, BQ, ESS and p = 0.004 for NoSAS).
CONCLUSION: This is one of the first prospective studies of sequentially hospitalized patients with confirmed COVID-19 status who were screened for possible OSA could be an independent risk factor for poor outcome in patients with COVID-19.

Introduction

The beginning of 2020 saw the evolution of COVID-19 into a global pandemic. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was discovered in December 2019 and is responsible for the causation of coronavirus disease 2019 (COVID-19) [1]. As of February 3rd 2021, over 100 million cases of COVID-19 have been registered around the world resulting in 2,264,968 deaths [2]. As the world struggles to cope with this pandemic, researchers across the world are curetting mechanistic pathways that would possibly explain the disease severity in certain population.

Mortality in COVID-19 disease was mainly seen in the subgroup of patients who developed severe respiratory failure owing to acute interstitial pneumonia involving both lungs and acute respiratory distress syndrome (ARDS) [3].While the understanding of pathogenesis is still a topic of research, few studies have pointed clinical association between mortality and older age, male sex, hypertension, diabetes, obesity, and cardiovascular diseases [4-6]. Interestingly obstructive sleep apnea (OSA) is also commonly associated with these similar comorbidities [7, 8].

OSA is a pro-inflammatory state involving mediators like IL-6, TNF-alpha, MCP-1, etc. [9]. IL-6 levels have also got prognostic implications in COVID-19 disease indicating that OSA might lead to worsening hypoxemia and cytokine storm. Furthermore, it has been hypothesized that SARS-CoV-2 infects humans through ACE-2 receptor which again has increased expression in obese OSA patients [10]. Obesity causes impaired respiratory mechanics leading to decreased FEV1, FVC, diaphragmatic excursion which in turn worsens outcome in patients with respiratory failure [11]. OSA can cause hypoxemia leading to poorer outcome in patients of COVID-19 pneumonia.

In accordance with these facts, it was observed that almost one-third of the COVID-19 cases requiring intensive care unit (ICU) admission had pre-existent OSA [12]. Similarly in two small studies on patients of severe COVID-19 pneumonia it was observed that one-quarter of the patients were known case of OSA [13, 14].

Since OSA and patients developing COVID-19 ARDS have so much in common, it is worthwhile to look into the association between these two. Thus, we need prospective studies to see whether people with OSA are at greater risk to develop COVID-19 related complications. Association, if any will further help in early triaging of complication prone population and possibly in prevention of complications. This study was planned to estimate the proportion of COVID-19 patients who have OSA based on various standard screening questionnaires and to explore if there is any association of being at high risk for OSA and severity of COVID-19 including mortality.

Methods

Setting and Design

This was a single center prospective observational study done at All India Institute of Medical Sciences Bhopal, India between 10th August and 22nd September 2020 on consecutive COVID-19 positive patients admitted to intensive care as well as isolation wards of hospital.

Participants and Procedures

All consecutive patients with confirmed positive report of COVID-19 RT-PCR on nasopharyngeal and oropharyngeal swab were enrolled. Patients were either admitted to intensive care or isolation wards as per clinical decision. Inclusion and exclusion criteria were as follows.

Inclusion Criteria

Patients positive for COVID-19 by RT-PCR of nasopharyngeal and oropharyngeal swab.

Age > 18 years.

Patients whose spouse or bed partner was willing for confirmation of history and sleeping pattern.

Patients and attendants who gave written informed consent for participating in study.

Exclusion Criteria

Patients/attendants who refused to cooperate or to give written informed consent.

Sleeping partner not available to confirm history given by patient.

Patient who were already intubated and were on mechanical ventilation.

Patients who were not in a state to answer questions.

Any recent surgery in last one month.

After obtaining informed consent from patient or bed partner of patient, demographic details, medical history including comorbidities were noted at the time of admission. As general practice, height and weight is measured for all patients in our ICU and ward at the time of admission. BMI was calculated according to other recorded weight and height data. Height and weight were measured using Seca®213 portable stadiometer and Seca®803 electronic flat scale, respectively. Neck circumference was measured at the level of cricothyroid membrane in sitting position using girth measuring tape.

Patients were asked for NOSAS [15], STOP BANG [16], BERLIN [17], and Epworth sleepiness scores [18] within one day of admission. History given by patient was reconfirmed by the bed partner of same patient.

NoSAS score category was classified as probable OSA if score was ≥ 8. STOP BANG was classified further as high risk of OSA for score ≥ 5. Berlin score was calculated in all three categories and classified into high risk if there were two or more categories with score ≥ 2 and low risk if there was only one category or no category where score was ≥ 2. ESS scores ≥ 10 was used to classify ESS into high or low risk categories for daytime sleepiness.

Patients were observed for their maximum oxygen and ventilatory requirement during their stay in hospital and were further divided into four groups as follows:

‘No oxygen’ group: patients who did not require oxygen during their hospital stay.

‘Oxygen only’ group: patients who required oxygen through facemask, venturi mask or nasal prongs and their fraction of oxygen requirement was less than 0.5. These patients never required invasive mechanical ventilation (IMV) or non-invasive ventilation (NIV) or high flow nasal cannula (HFNC) during their hospital stay.

‘NIV’ group: patients who required either NIV or HFNC anytime during their stay. These patients were never intubated during their hospital stay.

‘IMV’ group: patients who were intubated.

Patients were followed up for a period of 28 days after discharge or death during that period.

Data Analysis

We have used R software version 3.6.1 with gtsummary, ggplot2, and finalfit packages for data analysis [19-22]. Nominal variables are summarized as count and percentages while numerical variables as mean and standard deviation. Proportion of patients with high risk for OSA as defined by different questionnaires were analyzed and compared for baseline characteristics, types of respiratory or ventilatory requirement and mortality. Difference in distribution of nominal variables across groups was tested by chi-squared test and in numerical variables by Wilcoxon rank sum test. Logistic regression models were fitted separately to estimate effect age, gender, neck circumference, history of diabetes, hypertension, coronary artery disease, and STOP BANG score categories. Then to multivariable logistic regression model was fitted to test effect of variables which had p < 0.25 in univariable analysis. Model assumptions and goodness-of-fit was also tested by standard procedures.

Ethics and Permissions

Institutional Human Ethics Committee of AIIMS Bhopal reviewed and approved study protocol with approval letter number IHEC-LOP/2020/IM0309. Eligible patients and attendants were provided participant information sheet in native language for explaining purpose of study, procedures, and expectations from participants.

Results

During the study period, 250 patients with RT-PCR positive for COVID-19 got admitted in our hospital (Fig. 1). After applying inclusion and exclusion criteria, 213 patients (144 male and 69 female) were finally enrolled and were prospectively followed till final outcome (28 days post-discharge/death). Out of 213 patients, 57 succumbed to COVID-19 ARDS (Table 1). Patients who died were elderly (p < 0.001) and were more likely to have hypertension (p = 0.007) and/or Diabetes mellitus (p = 0.001). Screening questionnaires for OSA [STOPBANG, Berlin Questionnaire (BQ), NoSAS] were more likely to be positive in patients who died compared to patients who survived. Similarly, Epworth sleepiness scale (ESS) score was significantly higher in deceased group {12 (9.0–13.0) v/s 9.0 (5.2–11.0)} (p < 0.001).

Fig. 1

Flowchart of patients admitted during study period

Table 1

Baseline characteristics (demographic, OSA screening tool scores and clinical) among survived and deceased patients

Characteristic	Overall (N = 213)	Survived (n = 156)	Deceased (n = 57)	p value
Age	55 (44, 64)	53 (40, 62)	60 (52, 69)	< 0.001
Gender				> 0.9
Male	144 (68%)	105 (67%)	39 (68%)
Female	69 (32%)	51 (33%)	18 (32%)
Diabetes	93 (44%)	59 (38%)	34 (60%)	0.007
Hypertension	94 (44%)	58 (37%)	36 (63%)	0.001
CAD	25 (12%)	14 (9.0%)	11 (19%)	0.067
CKD	5 (2.3%)	1 (0.6%)	4 (7.0%)	0.019
BMI	26.7 (23.6, 29.5)	26.3 (23.6, 29.4)	28.0 (24.1, 30.1)	0.2
Neck circumference	40.0 (38.0, 42.0)	40.0 (37.0, 42.0)	41.0 (38.0, 43.0)	0.007
STOP BANG	3.00 (2.00, 4.00)	3.00 (2.00, 4.00)	4.00 (3.00, 5.00)	< 0.001
STOP BANG category				< 0.001
Low to intermediate	160 (77%)	129 (84%)	31 (55%)
High	49 (23%)	24 (16%)	25 (45%)
Berlin category				0.002
Low risk	108 (51%)	90 (58%)	18 (33%)
High risk	102 (49%)	65 (42%)	37 (67%)
NoSAS	9.0 (6.0, 13.0)	9.0 (6.0, 11.0)	11.0 (8.0, 15.0)	< 0.001
NoSAS category				0.020
No OSA	69 (33%)	58 (38%)	11 (20%)
OSA	140 (67%)	95 (62%)	45 (80%)
ESS score	9.0 (6.0, 12.0)	9.0 (5.2, 11.0)	12.0 (9.0, 13.0)	< 0.001
ESS category				< 0.001
High risk	98 (47%)	60 (39%)	38 (68%)
Low risk	112 (53%)	94 (61%)	18 (32%)

Similar findings were seen when baseline characteristics of 213 patients were compared according to across various modes of respiratory support (without oxygen, oxygen only, NIV group or IMV groups) (Table 2).

Table 2

Baseline characteristics (demographic, OSA screening tool scores, and clinical) across various modes of respiratory support

Characteristic	No oxygen, N = 71	Oxygen, N = 43	NIV, N = 37	IMV, N = 62	p value
Age	46 (32, 62)	57 (50, 66)	53 (50, 60)	59 (51, 67)	< 0.001
Male	48 (68%)	30 (70%)	23 (62%)	43 (69%)	0.9
Female	23 (32%)	13 (30%)	14 (38%)	19 (31%)
Diabetes	11 (15%)	24 (56%)	22 (59%)	36 (58%)	< 0.001
Hypertension	13 (18%)	22 (51%)	22 (59%)	37 (60%)	< 0.001
CAD	4 (5.6%)	6 (14%)	5 (14%)	10 (16%)	0.2
CKD	0 (0%)	0 (0%)	1 (2.7%)	4 (6.5%)	0.035
Neck circumference	39.0 (37.0, 41.0)	39.5 (37.0, 42.0)	40.0 (39.0, 42.0)	40.0 (38.0, 43.0)	0.067
STOP BANG score	2.00 (1.00, 3.00)	4.00 (2.00, 4.00)	4.00 (3.00, 4.00)	4.00 (3.00, 5.00)	< 0.001
STOP BANG group
Low to intermediate	65 (92%)	33 (79%)	27 (77%)	35 (57%)	< 0.001
High	6 (8.5%)	9 (21%)	8 (23%)	26 (43%)
Berlin category
Low risk	57 (80%)	18 (42%)	12 (33%)	21 (35%)	< 0.001
High risk	14 (20%)	25 (58%)	24 (67%)	39 (65%)
NoSAS	8.0 (3.5, 11.0)	9.0 (7.0, 13.0)	11.0 (7.5, 12.5)	11.0 (8.0, 14.0)	< 0.001
NoSAS category
No OSA	35 (49%)	12 (29%)	9 (26%)	13 (21%)	0.004
OSA	36 (51%)	30 (71%)	26 (74%)	48 (79%)
ESS score	6.0 (4.0, 10.0)	9.0 (8.0, 11.0)	9.5 (8.0, 12.2)	12.0 (9.0, 13.0)	< 0.001
High ESS (> 10)	20 (28%)	20 (48%)	18 (50%)	40 (66%)	< 0.001
Low ESS (< 10)	51 (72%)	22 (52%)	18 (50%)	21 (34%)

Proportion of patients who were classified as high risk for OSA by various OSA screening tools significantly increased with increasing respiratory support (Table 2 and Fig. 2) (p < 0.001 for STOPBANG, BQ, ESS and p = 0.004 for NoSAS).

Fig. 2

Proportion of patients classified as high risk for OSA by various OSA screening tools across different modes of respiratory support

On univariate analysis, age, hypertension (HTN), diabetes mellitus (DM), and OSA were found to be significant. Since median age was 55 years in our sample, so cut off of 55 years was used for multivariate analysis. On multivariate analysis, STOP BANG was used, since it is most commonly used screening tool for OSA screening both from clinical and research point of view. On binary logistics regression analysis, only age ≥ 55 and STOPBANG score ≥ 5 was found to be determinants of mortality.

We fitted a logistic model to predict outcome (mortality) with age group, STOP BANG score, the presence of diabetes, hypertension, coronary artery disease, and neck circumference. The model's explanatory power is moderate (Tjur's R2 = 0.14). Within this model, the effect of age group [≥ 55 years] is positive and can be considered as small and significant (beta = 0.74, SE = 0.37, 95% CI [0.04, 1.48]), while the effect of STOP BANG score in high risk group is also positive and can be considered as small and significant (beta = 0.91, SE = 0.42, 95% CI [0.08, 1.74]). The effects observed for older age and higher STOP BANG score were adjusted for neck circumference as well as history of comorbidities. Odds ratio for these variables which are exponentiated coefficients of logistic model are also presented. Odds of mortality were 2.10 (1.04–4.37, p = 0.042) among age more than 55 years compared to those with age less than 55 years. Participants classified in high risk category on STOP BANG score were having odds of 2.48 (1.09–5.69, p = 0.031) for mortality compared to those classified as low to intermediate risk for OSA. (Table 3).

Table 3

Results of binary logistics regression analysis for determinants of mortality

Dependent: deceased	Survived	Deceased	OR (univariable)	OR (multivariable)
Age group
< 55 years	84 (84.0)	16 (16.0)	–	–
≥ 55 years	72 (63.7)	41 (36.3)	2.99 (1.57–5.90, p = 0.001)	2.10 (1.04–4.37, p = 0.042)
Gender
Male	105 (72.9)	39 (27.1)	–	–
Female	51 (73.9)	18 (26.1)	0.95 (0.49–1.80, p = 0.878)	–
BMI
Mean (SD)	26.6 (4.8)	27.5 (4.9)	1.04 (0.98–1.11, p = 0.214)	–
Neck circumference
Mean (SD)	39.6 (5.1)	41.0 (3.0)	1.07 (1.00–1.16, p = 0.092)	1.04 (0.97–1.13, p = 0.272)
STOPBANG
Low to intermediate	129 (80.6)	31 (19.4)	–	–
High	24 (49.0)	25 (51.0)	4.33 (2.20–8.66, p < 0.001)	2.48 (1.09–5.69, p = 0.031)
DM
No	97 (80.8)	23 (19.2)	–	–
Yes	59 (63.4)	34 (36.6)	2.43 (1.31–4.56, p = 0.005)	1.61 (0.78–3.31, p = 0.192)
HTN
No	98 (82.4)	21 (17.6)	–	–
Yes	58 (61.7)	36 (38.3)	2.90 (1.56–5.50, p = 0.001)	1.27 (0.56–2.86, p = 0.567)
CAD
No	142 (75.5)	46 (24.5)	–	–
Yes	14 (56.0)	11 (44.0)	2.43 (1.01–5.71, p = 0.043)	1.60 (0.59–4.28, p = 0.351)

Number in data frame = 213, number in model = 209, missing = 4, AIC = 228.7, C-statistic = 0.738, H&L = Chi-sq(8) 11.21 (p = 0.190)

Discussion

This is one of the first prospective studies of sequentially hospitalized patients with confirmed COVID-19 status who were screened for possible OSA in a questionnaire-based format. This study shows that OSA could be an independent risk factor for poor outcome in patients with COVID-19.

Studies have highlighted association of severity of COVID-19 with older age, obesity, male sex, and comorbidities like DM, HTN, Coronary Artery Disease, and Chronic Kidney Disease [4]. Few researchers have observed a possible association with obstructive sleep apnea retrospectively. Cade et al. analyzed electronic health record data and observed 443 of 4668 participants with sleep apnea had increased mortality rate of 11.7% as compared to controls (6.9%) with an odds ratio of 1.79 [23]. In a study involving 700 patients, 124 had pre-diagnosed OSA; of all the patients requiring ICU care, 29% patients had pre-existent OSA in the same study [12]. Two small case series focusing critically ill patients of COVID-19 had shown 20–25% patients having OSA [13, 14]. In the CORONADO study, 144/1189 patients were already known case of OSA. They have also found OSA to be independent risk factor for poor outcome in COVID-19 related illness [24].

Age and neck circumference are integral component of most OSA screening questionnaire (STOPBANG and NoSAS) and the presence of hypertension is included as question in STOPBANG and Berlin Questionnaire. Thus, multivariate analysis was done to find independent association of OSA with mortality. In-fact in our study, only age and OSA were found to be significant factor for mortality in multivariate analysis. This was in contrast to most of previous studies, in which HTN & DM were found to be important factors for mortality.

Cardiac morbidity in COVID patients seems to be high and could prove fatal. Cardiac complications in SARS-CoV 2 infection includes myocarditis, cardiomyopathy, acute myocardial infarction, heart failure, venous thromboembolism, and arrythmias [25, 28] which might get accentuated in the presence of OSA. OSA leads to a procoagulant state and increased risk of DVT and PE. DVT and PE are known risk factors for mortality in COVID-19 [26]. The presence of procoagulant state is a breeding ground for COVID related coagulopathy. OSA has been associated with obesity, HTN, DM, CAD, arrythmia, chronic kidney disease, dyslipidemia, metabolic syndrome, and pulmonary embolism [27-33]. All these factors were almost consistently associated with poor prognosis in patients with COVID-19 in various studies [34, 35].

STOPBANG, BQ, and NoSAS have comparable sensitivity and specificity to polysomnography (PSG) for the diagnosis of OSA [15, 36, 37]. In our study, it was shown that patients with higher respiratory requirements had significantly higher probability of having OSA and this was consistently seen in all questionnaires for increasing severity of respiratory support.

Strength of this study is that screening tools were asked to both patients and sleeping partner. It is known fact that questions like history of apnea and history of snoring are more reliably answered by patients sleeping partner rather than patients. So, if there was any discrepancy in the answers for apnea or snoring, then answer from sleeping partner was considered final. Those patients who were not in a state to answer questions were excluded from our study. Second, we screened for OSA using multiple questionnaires. Most importantly, this is the first study in which patients being admitted for COVID-19 were screened for OSA by different questionnaires. All previous studies on possible association between OSA and COVID-19 related ARDS were done in already diagnosed cases of OSA.

Important limitations were: (1) It was a single center study conducted at a tertiary care hospital where relatively more sick patients were admitted. This poses possibility of selection bias and, therefore, results of the study should be interpreted in this context and not be generalized to all COVID-19 patients. (2) Our diagnosis of OSA was based on screening questionnaires and gold standard PSG could not be done for confirmation of OSA. However, we have evaluated risk for OSA by multiple questionnaires and results were coherent. (3) We have tried to be precise in our results by limiting to patients whose spouse or bed partner was willing to confirm history and sleeping pattern. Thereby some selection bias may have been encountered. Similarly, excluding patients already intubated leads to similar selection bias as it eliminates the sickest patients. However, we have tried to present our results with greater precision thereby compromising on some selection bias. (4) Our study had limited sample size. The results may not have reflected some common factors associated with poor outcome of COVID-19 as reflected in larger studies.

Conclusions

This study shows that OSA might be an independent risk factor for poor outcome in COVID-19 related illness.