‘A study of correlations between sleep recording data and three sleep questionnaires: Epworth, Pittsburgh, Berlin’

Introduction

Obstructive sleep apnea-hypopnea syndrome (OSA) affects approximately one billion people worldwide with a rate exceeding 50% in some countries [1]. Its prevalence increases almost linearly in adults up to the age of 65 years [2]. In a general population cohort from Lausanne, Switzerland, the prevalence of moderate to severe OSA in 3043 participants was 23.4% in women and 49.7% in men [3]. OSA is an independent risk factor for mortality, mainly of cardiovascular origin [4]. It also increases the risk of car accidents [5]. The early diagnosis of this pathology, therefore, appears to be a goal to be pursued. Polysomnography remains the ‘gold standard’ for the diagnosis of OSA but it is a long examination, performed in a hospital (one or two nights in hospital) in most cases, and requires specific equipment and qualified personnel. Ventilatory polygraphy is also widely used, requiring less equipment and can be performed at the patient’s home. However, it remains less sensitive than polysomnography because it underestimates respiratory events.

There are several questionnaires, some of which assess the quality of sleep and some of which attempt to quantify the probability of having OSA, with the aim of selecting patients for sleep recording. It is repeatedly reported in the literature and in lectures to students that the most common symptom in OSA is excessive daytime sleepiness, which is nevertheless a common symptom in many other sleep disorders. However, in a study of a general population aged 40 to 65 years, approximately one in five subjects had moderate to severe OSA, but the majority were neither symptomatic nor sleepy and had no alertness disorder [6]. The Epworth questionnaire (ESS) represents the most widely used validated questionnaire currently used in clinical routine to assess subjective daytime sleepiness. Subjective daytime sleepiness is considered to exist if the score is greater than or equal to 11 [7]. The Pittsburgh Sleep Quality Index (PSQI) assesses sleep quality and disturbances over a one-month time interval. A PSQI score >5 classifies the patient as a ‘poor sleeper’ [8].

The Berlin Questionnaire (BQ) was first developed at the Conference on Sleep in Primary Care in Berlin, Germany in 1996 [9]. Three categories such as snoring, sleepiness on awakening or fatigue, and presence of obesity or hypertension and a total of 10 items were selected. Each category was assessed independently, and positive subjects with two or more categories are accepted as being at high risk for OSA [9].

Several studies have been conducted to investigate the correlation between the scores of these questionnaires and the apnea-hypopnea index (AHI) obtained by polysomnographic recordings, with the aim of finding out whether these questionnaires can predict patients at risk of OSA [10–12]. In the majority of papers, one or two questionnaires are used and sometimes authors have tried to combine two questionnaires in order to increase the probability of predicting the presence of OSA [12]. These questionnaires are poor predictors of OSA [10–13]. The aim of our study was to investigate the correlation of sleep recording data obtained by ventilatory polygraphy and the three questionnaires of Epworth, Pittsburgh, and Berlin.

Material and method

One hundred and sixty adult patients with a sleep recording by ventilatory polygraphy in the ENT and Cervicofacial Surgery Department at the Nancy University Hospital from 1 January 2019 to 31 December 2021 were included in this study. The indications for sleep recording were: suspicion of OSA on subjective or objective criteria (presence of anatomical factors predictive of OSA during ENT examination), patients consulting for ronchopathy; patients at high risk of OSA (after treatment of ENT cancer [14], epistaxis). All patients had a prior ENT consultation. Sleep recording was performed on an outpatient basis by ventilatory polygraphs (NOX T3 Portable Sleep Monitor, Resmed, USA) including measurements of nasal-oral airflow, oximetry, heart rate, and chest and abdominal movements. Patients completed the Epworth, Pittsburgh, and Berlin questionnaires on the same day of sleep recording. Patients with insufficient recording time or incompletely completed forms were excluded from the study.

Data collection

The following data were collected: age, weight, height, sex, effective recording time calculated by the device, AHI, oxygen desaturation index (ODI), mean SpO2, number of desaturations during recording, mean pulse rate, plethysmography signal drop, Epworth, Pittsburgh, and Berlin scores.

Statistical analysis

Quantitative variables are presented as mean ± standard deviation. Qualitative variables are presented as frequency and percentage. The Chi 2 test or the Fisher exact test (when the theoretical numbers were less than 5) were used to compare categorical variables between groups. We defined the groups: presence of OSA (AHI ≥ 5), absence of OSA (AHI < 5); moderate/severe OSA (AHI ≥ 15) and absence of moderate/severe OSA (AHI < 15). An Epworth score >10 and a Pittsburgh score >5 was considered positive. For the Berlin score, the score is positive if there are 2 or more positive categories. Pearson correlation coefficients were calculated to look for a correlation between quantitative variables. To assess the predictors of moderate/severe OSA, a multivariate logistic regression model was performed. Questionnaire variables as well as BMI and age were included, in a forced manner, in the multivariate logistic regression model, which was performed with bottom-up stepwise selection. Analyses were performed using SAS v9.1 software (SAS Inst., Cary, NC). The p value < .005 was considered significant, p values < .05 suggestive.

Results

A total of 160 patients (103 men [64.4%] and 57 women [35.6%]) were included in this study. The mean age was 53.4 ± 15.1 years. Most of the study subjects were overweight, with a mean BMI of 27.8 ± 5 kg/m2. The prevalence of OSA in the study population was 75.3%. OSA was severe in 33.6% of OSA patients. The mean AHI was 20.2 ± 19.3/h. The mean oxygen desaturation index was 16.8 ± 16/h. The means at ESS, Pittsburgh of the total population were 8.5 ± 4.7 and 7.1 ± 3.5, respectively (Table 1). The Berlin score was calculated by categories with a mean of 1.4 ± .7 for positive categories.

Table 1. The means and medians of questionnaire scores, sleep recording data, and patient characteristics.

	Mean	Median
Epworth	8.5 ± 4.8	8.0
Pittsburgh	7.1 ± 3.5	6.0
Berlin^a	1.4 ± 0.7	1.0
AHI	20.2 ± 19.3	13.2
ODI	16.7 ± 16	12.1
Pulse	65.5 ± 9.7	65.0
Plethysmography	39.2 ± 18.4	37.1
BMI	27.8 ± 5.6	27.0
Age	53.4 ± 15.1	55

^aBerlin scores were calculated by positive categories. AHI: apnea-hypopnea index; ODI: oxygen desaturation index; BMI: body mass index.

Table 2 shows the distribution of patients according to AHI values and pathological thresholds of questionnaire scores as well as BMI. The presence of OSA (AHI ≥ 5/h) is found among 24% of patients with a pathological Epworth score, 46.8% of patients scoring a pathological Pittsburgh score, and 33.8% of patients being at high risk of OSA according to the Berlin score. Interestingly, 52% of the patients with OSA had a non-pathological Epworth sleepiness score. In our study, 60% of patients with OSA and 33.1% of patients with moderate/severe OSA had a BMI <30kg/m2. When the scores of the questionnaires are considered pathological, the presence of moderate/severe OSA is found in only 14.3% of patients according to the ESS, 23.8% according to the Pittsburgh, and 20.1% according to the Berlin.

Table 2. Distribution of patients according to AHI and pathological thresholds of scores and BMI (Chi 2 test). An Epworth score > 10 and a Pittsburgh score > 5 are considered pathological. For the Berlin score, the score is positive if there are 2 or more positive categories.

		AHI < 5	AHI ≥ 5	P	P AHI ≤ 15	AHI > 15	p
Epworth	≤10	26(16.9%)	80(52%)	.82	59(38.3%)	47(37%)	.86
	>10	11(7.1%)	37(24%)		26(16.9%)	22(14.3%)
Pittsburgh	≤5	11(7.7%)	40(28%)	.46	21(16.7%)	30(21%)	.012
	>5	25(17.5%)	67(46.8%)		58(40.6%)	34(23.8%)
Berlin (positive categories)	<2	23(14.9%)	65(42.2%)	.48	49(31.8%)	39(25.3%)	.74
	≥2	14(9.1%)	52(33.8%)		35(22.7%)	31(20.1%)
BMI	< 30 kg/m2	28(17.8%)	94(59.9%)	.49	70(44.6%)	52(33.1%)	.22
	≥ 30 kg/m2	10(6.4%)	25(15.9%)		16(10.2%)	19(12.1%)

There was no correlation between questionnaire scores and sleep recording data such as AHI, ODI, pulse, and plethysmography (Table 3). We find the same findings for the correlation between BMI, age, and sleep recording data.

Table 3. Pearson’s correlation coefficients ® between scores, BMI, age and sleep recording data.

	Pittsburgh	Berlin	AHI	ODI	Pulse	Pleth	BMI	Age
ESS	.30	.44	−.026	.02	.10	.19	.06	−.14
	p = .0003	p < .0001	p = .74	p = .82	p = .22	p = .02	p = .42	p = .08
Pittsburgh		.25	−.11	−.13	.15	−.006	−.0001	−.03
		p = .003	p = .17	p = .12	p = .08	p = .94	p = .10	p = .76
Berlin			.15	.17	−.009	.16	.13	−.01
			p = .05	p = 0.04	p = .91	p = .06	p = .10	p = .23
AHI				.95	.11	0.11	.18	.34
				p < .0001	p = .17	p = .17	p = .02	p < .0001
ODI					.16	.1	.25	.35
					p = .05	p = .23	p = .0015	p < .0001
Pulse						.17	.21	−.05
						p = .04	p = .01	p = .6
Pleth							.15	−.30
							p = .06	p = .0002
BMI								−.09
								p = .28

ESS: Epworth sleepiness scale; AHI: apnea-hypopnea index; ODI: oxygen desaturation index; Pleth: plethysmography; BMI: body mass index.

Based on multivariate analyses and using the logistic regression model, when scores are pathological, the only questionnaire that represents a predictive factor for moderate/severe OSA is the Pittsburgh (OR = 2.7; CI95% [1.23–6.04]). Scores considered pathological on the ESS and Berlin do not predict moderate/severe OSA (OR = .88; 95% CI [.38–2.06], and OR = .79; 95% CI [.36–1.73]). Similarly, obesity (BMI >30kg/m2) and age were not predictive factors for moderate/severe OSA in our patients.

Discussion

Sleep apnea-hypopnea syndrome is a common pathology in the world, and carries its share of cardiovascular complications, an increased accidental risk, and also a mortality risk related to the severity of OSA [15]. Intermittent nocturnal hypoxia is the main determinant. It is the result of a collapse of the upper airway during sleep.

Several questionnaires have been developed and used in our current practice in order to screen patients at risk of OSA. To our knowledge, this is the first study comparing three sleep questionnaires commonly used in clinical practice with sleep recording data. Several studies had previously shown an uncertain or even poor relationship between ESS scores and AHI [16–17]. In the study by Hesselbacher et al. [10], the sensitivity and specificity of the ESS for the diagnosis of OSA were 54% and 57%, respectively; the positive and negative predictive values were 64% and 47%, respectively. Regarding Pittsburgh, a study in 2015 shows a lack of significant correlation between PSQI and AHI, desaturation index, minimum SpO2, and total sleep time < 90% [18]. The 2013 study by Scarlata et al. indicated that ESS and PSQI global scores were of little value in identifying patients with OSA [11]. In addition, the sensitivity, specificity, positive and negative predictive value of the PSQI score for the diagnosis of OSA were insufficient in this study. For the Berlin score, there are variable correlations in studies with AHI, but there seems to be a good correlation in some studies with severe OSA [19–20].

Our study shows little or no correlation between ESS, PSQI, and Berlin scores and AHI as well as other data recorded during recording by ventilatory polygraphy such as mean pulse or plethysmography drop value. According to our results, there are 52% of patients with OSA who have an ESS score considered non-pathological. It appeared that the Pittsburgh questionnaire was better at predicting OSA, especially moderate/severe OSA. These questionnaires were initially developed to assess excessive sleepiness of multifactorial origin (ESS) or to assess sleep quality (Pittsburgh). Therefore, a weak correlation or lack of correlation is not surprising. The Berlin questionnaire was developed to identify patients at high risk for OSA. However, this questionnaire is not a relevant tool according to the results of our study.

Therefore, it is interesting to develop a simple, reproducible questionnaire in order to identify more precisely patients at high risk of OSA.

In our study, we did not find an association between BMI, age and the presence of OSA. This may be explained by the fact that sleep recordings by ventilatory polygraphy were performed in patients with anatomical abnormalities in the upper airway found during an ENT examination. These anatomical abnormalities result in narrowing of the retrovelar and retrobasilingual epiglottic spaces, which eventually obstruct the pharynx during sleep. This may explain the high number of patients without excessive daytime sleepiness in our series (69% of patients).

Limitations

Our study has some limitations. First, a recruitment bias related to patients selected in ENT consultation for suspected OSA. Moreover, we used sleep recordings by ventilatory polygraphy, with the limitations that can be associated with it, in particular the absence of certainty on the beginning and end of sleep. Therefore, there may be over or underestimations of sleep duration. It is also known that this type of sleep recording underestimates the AHI and cannot identify micro-awakenings. One can also ask the question of other parameters measured in polygraphy rather than the usual parameters (including AHI) which could possibly be better correlated with the clinical questionnaires. Further studies are needed to demonstrate this. On the other hand, some questionnaires have shown interesting correlations in a few studies, such as the STOP-BANG questionnaire, the No-Apnea score or the No-SAS score.

Conclusion

This study shows a weak correlation, or even an absence of correlation, between the scores of the three questionnaires: Epworth, Pittsburgh and Berlin, and the sleep recording data. This could mean that it is not possible to predict the diagnosis or the stage of severity of OSA based on these questionnaires.

Disclosure statement

No potential conflict of interest was reported by the author(s).