SEARCH

SEARCH BY CITATION

Keywords:

  • Berlin Questionnaire;
  • epidemiology;
  • polysomnography;
  • screening;
  • sensitivity and specificity;
  • sleep apnea syndromes

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References
  10. Supporting Information

The Berlin Questionnaire (BQ) is a widely used screening tool for obstructive sleep apnea (OSA), but its performance in the general population setting is unknown. The prevalence of OSA in middle-aged adults is not known in Norway. Accordingly, the aims of the current study were to evaluate the utility of the BQ for OSA screening in the general population and to estimate the prevalence of OSA in Norway. The study population consisted of 29 258 subjects (aged 30–65 years, 50% female) who received the BQ by mail. Of these, 16 302 (55.7%) responded. Five-hundred and eighteen subjects were included in the clinical sample and underwent in-hospital polysomnography. Screening properties and prevalence were estimated by a statistical model that adjusted for bias in the sampling procedure. Among the 16 302 respondents, 24.3% (95% confidence interval (CI) = 23.6–25.0%) were classified by the BQ to be at high-risk of having OSA. Defining OSA as an apnea–hypopnea index (AHI) ≥5, the positive predictive value of the BQ was estimated to be 61.3%, the negative predictive value 66.2%, the sensitivity 37.2% and the specificity 84.0%. Estimated prevalences of OSA were 16% for AHI ≥ 5 and 8% for AHI ≥ 15. In conclusion, the BQ classified one out of four middle-aged Norwegians to be at high-risk of having OSA, but the screening properties of the BQ were suboptimal. The estimated prevalence of OSA was comparable to previous estimates from general populations in the USA, Australia and Europe.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References
  10. Supporting Information

The identification of subjects with obstructive sleep apnea (OSA) is crucial for the prevention of premature mortality (Punjabi et al., 2009; Young et al., 2008), cardiovascular morbidity (Somers et al., 2008), traffic accidents (Rodenstein, 2008) and work disability (Sivertsen et al., 2008) associated with OSA. Still, the proportion of undiagnosed subjects has been estimated to be 82% in men and 93% in women of the general population (Young et al., 1997). Several questionnaires and prediction rules have been developed (Harding, 2001; Pang and Terris, 2006; Rowley et al., 2000), but we have not identified community-based validation studies of screening instruments for OSA. Knowledge of properties of screening instruments applied in population-based samples is therefore needed.

To date, the Berlin Questionnaire (BQ) is the only screening instrument for OSA that has been validated in general practice, hospital and sleep clinic samples (Ahmadi et al., 2008; Chung et al., 2008; Gami et al., 2004; Gus et al., 2008; Netzer et al., 1999; Sharma et al., 2006; Weinreich et al., 2006). However, evaluations of the screening properties of the BQ have provided divergent results, with sensitivities ranging from 63% to 86% and specificities from 49% to 95%. The BQ is widely used in clinical practice and has been used to assess the risk of OSA in two large, representative US telephone surveys (Hiestand et al., 2006; Kapsimalis and Kryger, 2009).

An apnea–hypopnea index (AHI) ≥ 5 plus daytime symptoms or an AHI ≥ 15 is necessary to diagnose OSA (American Academy of Sleep Medicine, 2005). The prevalence of OSA is estimated to be 4–10% in middle-aged general populations of predominantly Caucasian origin, with higher prevalences with rising age and male gender (Bearpark et al., 1995; Bixler et al., 1998, 2001; Duran et al., 2001; Plywaczewski et al., 2008; Young et al., 1993). However, 20-year-old estimates from middle-aged populations in Sweden and Denmark have been lower than estimates from other Caucasian populations (1.3–1.9% in males, 0.9% in females; Gislason et al., 1988; Jennum and Sjol, 1992). The prevalence of OSA in Norway is not known. Accordingly, the aims of the current study were to evaluate screening properties of the BQ in a Norwegian, general population sample, and to estimate overall age- and gender-specific prevalences of OSA using polysomnography with cut-off values of 5 and 15 on the AHI as reference standards.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References
  10. Supporting Information

Screening sample

The Akershus Sleep Apnea Project (ASAP) is a cross-sectional, two-phase study of subjects residing in three counties in central eastern Norway. A questionnaire and a letter describing the study were mailed twice to a sample of 30 000 age- and sex-stratified subjects. The sample consisted of 2000 males and 2000 females aged 30, 35, 40, 45, 50, 55 and 60 years, in addition to 1000 males and 1000 females aged 65 years that were randomly drawn from the National Population Register. Questionnaires could be answered by returning the paper version or via a web-based solution. The study population consisted of 29 258 subjects (aged 30–65 years, 50% female) who received the questionnaire by mail (Fig. 1). Of these, 16 302 subjects responded (55.7%). The response rate to the screening sample was significantly higher for subjects aged 50 years and more (62.0%) than for subjects less than 50 years of age (50.1%), and was higher for females (59.0%) than males (52.4%). Males aged 30 years used the web solution most (22.3%), and females aged 65 years the least (1.4%). Age- and gender-stratified analyses of differences between web and paper responses revealed no systematic differences.

image

Figure 1.  A flow diagram of subjects in the study. F, females; M, males.

Download figure to PowerPoint

Estimation sample and random draws

Questionnaires received more than 4 months after the first mailing (= 41) or without a contact telephone number (= 729), and 672 questionnaires that were administratively misplaced were regarded as not eligible (= 1442). Thus, the estimation sample consisted of 14 860 questionnaires. The BQ (described in detail in a subsequent section) was used for risk classification, but missing of one or more BQ item occurred in 43.8% of these questionnaires. After careful consideration, missing was recoded to ‘zero’ to establish risk status of all respondents. In addition, BQ low-risk questionnaires with missing items (= 5541) were made unavailable for the subsequent draws to avoid misclassification among subjects recruited to the clinical part of the project.

The draws were primarily organized to fill 32 strata based on age, gender and BQ risk status. The 16 BQ high-risk strata were additionally substratified by self-reported previous otitis media surgery (= 242), diabetes (= 155), and a stratum with no previous otitis media surgery, myocardial infarction or diabetes (= 581). In addition to the 32 strata, an age- and gender-independent BQ high-risk stratum with subjects reporting previous myocardial infarction and no diabetes (= 107) was drawn. Substratification in the BQ high-risk strata was performed because we a priori hypothesized that these conditions would be associated with increased likelihood of OSA (Gozal et al., 2008; Somers et al., 2008).

Clinical sample

Drawn subjects were invited by mail and approached by telephone. Invited subjects were excluded after three unsuccessful attempts of contact (= 202), use of continuous positive airway pressure (= 10), pregnancy (= 9), lack of Norwegian language skills (= 5) or severe physical impairment (= 4). A total of 585 subjects declined the invitation and 422 drawn subjects were never invited, leaving a total sample of 535 subjects included in the overall project. In addition, 17 subjects were excluded from analysis according to poor sleep quality (Young et al., 1993). The clinical sample thus comprised of 518 subjects (38.4% of 1350 invited by mail). BQ high-risk subjects were substratified into 91 with otitis media surgery, 44 with diabetes, 190 with no diabetes, otitis media surgery or myocardial infarction, and 40 with myocardial infarction.

Questionnaires

The one-page questionnaire consisted of a Norwegian translation of the BQ (Netzer et al., 1999) and the Epworth Sleepiness Scale (ESS) (Johns, 1991), as well as questions assessing other possible predictors of OSA in the general population (Young et al., 2002). A BQ high-risk subject was defined as one in whom at least two of the following three positive risk categories were present: (I) snoring; (II) daytime somnolence; and (III) obesity, defined as a body mass index (BMI) > 30 kg m−2 or self-reported history of hypertension. Definitions of each category have previously been published (Chung et al., 2008). Translation of the BQ was done according to international guidelines, with forward and backward translation by two pairs of bilingual researchers.

The BQ item 3, snoring frequency, was dichotomized into snoring every night or almost every night versus other frequencies to assess habitual snoring (Partinen and Gislason, 1995). The ESS (Johns, 1991) sum score was calculated and excessive daytime sleepiness was defined as an ESS score >10 (Johns, 1991).

Clinical study procedure

Participants were admitted to Akershus University Hospital, Department Stensby, between June 2006 and January 2008 after providing written informed consent. The polysomnography registrations included two-channel electroencephalography (C4/A1 and C3/A2), two-channel electrooculography, one-channel submental electromyography, leg electromyography (tibialis), measurement of SaO2, assessment of breathing movements (Respitrace; Ambulatory Monitoring, Ardsley, NY, USA), nasal and oral air flow assessment (Protech, Woodinvill, WA, USA), and body position monitoring. All electrophysiological signals were preamplified, stored and subsequently scored at 30-s epochs using the Somnologica 3.2 software package (Flaga-Medcare, Buffalo, NY, USA) according to the Rechtshaffen and Kales scoring manual (Rechtshaffen and Kales, 1968) by two US board-certified polysomnography technicians who were blinded to the result of the BQ. Arousals were documented and classified (American Sleep Disorders Association, 1992). Obstructive apneas were scored when airflow dropped below 10% of the reference amplitude for more than 10 s. Hypopneas were scored when airflow dropped below 70% for more than 10 s with a subsequent oxygen desaturation ≥4%. Few central apneas or hypopneas were registered, and no distinction was made between central and obstructive events. The AHI was calculated as the average of the total number of apneas and hypopneas per hour of sleep. OSA was defined using AHI cut-off values of 5 and 15 to enable comparison of the results of the present study with those of previous questionnaire validation studies.

The study protocol was approved in 2005 by the Regional Committee for Medical Research Ethics in eastern Norway, the National Data Inspectorate and the Norwegian Social Science Data Services.

Statistical analysis

Comparisons between groups were performed using the chi-square test for dichotomous variables, Student’s t-test for normally distributed continuous variables and the Mann–Whitney U-test for not normally distributed continuous variables. Two-sided P-values <0.05 were considered statistically significant. Age- and gender-specific prevalences were estimated by calculating stratum-specific probabilities obtained from a logistic regression model. Those probabilities were combined by weighted averages and adjusted for the clinical sample participation status of the 1772 subjects randomly drawn. Standard formulas for stratified sampling were used to calculate 95% CI (Cochran, 1963). The adjusted prevalence estimates were extrapolated to represent the estimation sample by simulations from 100 random iterations. These iterations were finally used to obtain cut-off values on estimated probabilities needed to estimate screening properties of the BQ, such as sensitivities, specificities, likelihood ratios and predictive values (see online supplement for further details). Posttest probabilities for OSA were calculated in accordance with Bayes Theorem. Calculations were performed using Microsoft Office Excel 2003 and the Statistical Package for the Social Sciences (spss, Inc., Chicago, IL, USA), version 16.0.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References
  10. Supporting Information

Sample characteristics

Demographic variables, self-reported sleep characteristics and polysomnography characteristics of the screening, estimation and clinical samples are reported in Table 1. In the screening sample, simple snoring was reported by 69.3% of males and 49.8% of females, and habitual snoring in 32.8% of males and 20.0% of females. The mean ESS score was 7.1 standard deviation (SD) 4.0 in males and 6.7 (SD, 4.0) in females, and the mean BMI was 26.8 kg m−2 (SD 3.9 kg m−2) in males and 25.3 kg m−2 (SD 4.5 kg m−2) in females. The frequency of obesity defined by a BMI >30 kg m−2 was 16.0% for males and 13.8% for females. All gender-specific differences in the screening sample were highly significant (< 0.01). The differences between the estimation sample and the screening sample reported in Table 1 were statistically significant only in the three variables of subjective sleep. Also, differences between the clinical sample and the subjects randomly drawn but not included (= 1254) were small and also limited to variables assessing subjective sleep. However, the clinical sample differed substantially from the screening and estimation samples in all variables except age. The median AHI (25th, 75th percentiles) for males and females in the clinical sample were 10.1 (3.2, 28.4) and 3.7 (0.8, 10.7), respectively.

Table 1.   Demographic variables and polysomnography characteristics of the screening, estimation and clinical samples
Demographic variablesScreening sample n = 16 302Estimation sample n = 14 860Clinical sample n = 518
  1. AHI, apnea–hypopnea index; BMI, body mass index; ESS, Epworth Sleepiness Scale; SWS, slow-wave sleep; TST, total sleep time, SD, standard deviation.

Mean age (SD)47.8 (10.7)47.8 (10.7)48.3 (11.2)
Male gender (%)46.846.854.8
Daily smoking (%)  26.9
Married or cohabitant (%)  84.0
Higher education (%)  32.2
In regular work (%)  62.6
Mean BMI (SD)26.0 (4.3)26.0 (4.3)27.9 (4.8)
BMI ≥ 30 (%)14.814.932.0
Subjective sleep variables   
Simple snoring (%)58.959.589.6
Habitual snoring (%)26.226.748.2
Mean ESS (SD)6.9 (4.0)6.9 (4.0)8.8 (4.6)
Polysomnography variables   
Median AHI (25th, 75th percentiles)  6.4 (1.7, 18.3)
Mean minutes TST (SD)  418.0 (75.6)
Mean SWS as a % of TST (SD)  22.9 (8.1)

Screening properties of the BQ

The characteristics of BQ high- and low-risk subjects of the screening and clinical samples are presented in Table 2. The proportion of subjects classified as BQ high-risk in the screening sample was 24.3% (95% CI = 23.6–25.0%), 28.8% in males and 20.3% for females. Differences between BQ high-risk subjects and BQ low-risk subjects were significantly different in all variables in Table 2 for both samples. A total of 5056 subjects (31%) reported waking feeling tired or fatigued on ≥3 days of the week. This symptom was three times more frequent in BQ high-risk subjects (63.5%) than in BQ low-risk subjects (20.6%). The median AHI (25th, 75th percentiles) of BQ high-risk subjects and BQ low-risk subjects in the clinical sample were 7.8 (2.4, 22.1) and 3.2 (0.7, 11.2; < 0.01), respectively.

Table 2.   Empirical characteristics of the BQ in the screening sample and the clinical sample
VariableScreening sample n = 16 302Clinical sample n = 518
BQ high-riskBQ low-riskBQ high-riskBQ low-risk
  1. *< 0.05, **< 0.01 for differences between high-risk subjects in the two samples.

  2. #< 0.05, ##< 0.01 for differences between low-risk subjects in the two samples.

  3. BMI, body mass index; BQ, Berlin Questionnaire; EDS, excessive daytime sleepiness defined by Epworth Sleepiness Scale >10; ESS, Epworth Sleepiness Scale; Habitual snoring, snoring every night or almost every night of the week; HT, self-reported hypertension; MI, myocardial infarction; Obesity, BMI > 30.

n (% of total)3960 (24.3)12 342 (75.7)365 (70.5)153 (29.5)
Mean age (SD)48.8 (10.3)47.5 (10.8)48.6 (11.1)47.6 (11.4)
Male sex (%)2198 (55.5)5431 (44.0)207 (56.7)77 (50.3)
BQ risk categories and related items
I: BQ snoring (%)3654 (92.3)3011 (24.4)332 (91.0)61 (39.9)##
 Simple snoring (%)3657 (92.3)5952 (48.2)338 (92.6)126 (82.4)##
 Habitual snoring (%)2192 (57.7)1468 (14.5)211 (60.5)31 (20.3)#
II: BQ somnolence (%)2498 (63.1)1574 (12.8)255 (69.9)*18 (11.8)
 EDS (%)1336 (33.7)1547 (12.5)156 (42.7)**20 (13.1)
 Mean ESS (SD)8.8 (4.5)6.3 (3.7)9.8 (4.5)**6.6 (3.9)
III: BQ HT or obesity (%)2624 (66.3)1374 (11.1)240 (65.8)15 (9.8)
 HT (%)1454 (36.7)845 (6.8)134 (36.7)8 (5.2)
 BMI > 30 kg m−2 (%)1699 (42.9)715 (5.8)156 (42.7)10 (6.5)
 Mean BMI (SD)29.1 (5.2)25.0 (3.4)29.0 (5.0)25.4 (2.9)
Substratification variables
Diabetes (%)274 (7.1)272 (2.2)44 (12.3)**2 (1.4)
Previous MI (%)150 (3.9)194 (1.6)46 (12.9)**2 (1.3)
Otitis media surgery (%)634 (31.5)1511 (27.4)106 (52.0)**12 (18.8)

Because missing responses to items were scored as zero, we conducted a sensitivity analysis to assess whether this coding practice could have biased the proportion of subjects classified as BQ high-risk in the screening sample. All missing BQ items in the daytime somnolence (II) and obesity/history of hypertension (III) categories were recoded using maximum values. Questions in the snoring (I) category were recoded as maximum if a negation was not stated on the question ‘do you snore?’ After this procedure, the proportion of high-risk subjects in the screening sample increased from 24.3% to 27.7%.

Estimated properties of the BQ and total prevalence of OSA are reported in Table 3. Among subjects classified as high-risk, the estimated proportion of subjects with an AHI ≥ 5 (i.e. a positive predictive value) was 61.3% (95% CI = 59.7–62.9%). Conversely, among subjects classified as low-risk, the estimated proportion of subjects with an AHI < 5 (i.e. a negative predictive value) was 66.2% (95% CI = 65.3–67.1%). The corresponding estimated values for test sensitivity and specificity were 37.2% (95% CI = 36.0–38.4%) and 84.0% (95% CI = 83.2–84.7%). These values yielded a positive likelihood ratio of 2.3 (95% CI = 2.2–2.5) and a negative likelihood ratio of 0.8 (95% CI = 0.7–0.8). The corresponding values for an AHI cut-off of ≥15 as the definition of OSA are detailed in Table 3. When defined as an AHI ≥ 5, the probability of OSA increased from a pretest probability (= population prevalence) of 16% (95% CI = 13–20%) to 31% (95% CI = 25–38%) in subjects with a positive BQ result. Changes in pretest (1 − prevalence) and posttest probabilities among test-negative subjects and for OSA defined as an AHI ≥ 15 are depicted in Fig. 2.

Table 3.   Estimated* screening properties of the BQ and total prevalence of OSA among the Norwegian general population
 AHI ≥ 5 Estimate (95%CI)AHI ≥ 15 Estimate (95%CI)
  1. *Estimates were based on a statistical model that adjusted for oversampling in the BQ high-risk group and selection of snorers to the low-risk group.

  2. AHI, apnea–hypopnea index; CI, confidence interval; LR, likelihood ratio; OR, odds ratio.

Prevalence0.16 (0.13–0.20)0.08 (0.06–0.11)
OR3.1 (2.9–3.4)3.0 (2.7–3.2)
Positive predictive value (%)61.3 (59.7–62.9)33.5 (32.0–35.0)
Negative predictive value (%)66.2 (65.3–67.1)85.5 (84.8–86.1)
Sensitivity (%)37.2 (36.0–38.4)43.0 (41.2–44.8)
Specificity (%)84.0 (83.2–84.7)79.7 (79.0–80.5)
LR positive2.3 (2.2–2.5)2.1 (2.0–2.3)
LR negative0.8 (0.7–0.8)0.7 (0.7–0.7)
image

Figure 2.  Estimated probabilities of having OSA and not having OSA before and after completing the Berlin Questionnaire. AHI, apnea–hypopnea index; BQ, Berlin Questionnaire. [*Correction added on 15 July 2010 after first online publication on 16 June 2010. The following lines on the y-axis of Figure 2 were missing and have now been added: Posttest probability of AHI ≥ 5 among BQ positive, Posttest probability of AHI < 5 among BQ negative, Pretest probability of AHI ≥ 15, Pretest probability of AHI < 15].

Download figure to PowerPoint

Estimated age- and gender-specific prevalences of OSA are presented in Table 4.

Table 4.   Estimated* gender- and age-specific prevalences of OSA among the Norwegian general population
 AHI ≥ 5 Estimate (95% CI)AHI ≥ 15 Estimate (95% CI)
  1. *Estimates were based on a statistical model that adjusted for oversampling in the BQ high-risk group and selection of snorers to the low-risk group.

Male0.21 (0.16–0.26)0.11 (0.07–0.15)
Female0.13 (0.09–0.17)0.06 (0.03–0.09)
Total <50 years0.13 (0.09–0.18)0.06 (0.03–0.09)
Male <50 years0.18 (0.11–0.25)0.09 (0.03–0.14)
Female <50 years0.10 (0.05–0.15)0.04 (0.01–0.07)
Total >50 years0.20 (0.15–0.24)0.10 (0.06–0.14)
Male >50 years0.23 (0.16–0.31)0.13 (0.07–0.20)
Female >50 years0.16 (0.10–0.22)0.08 (0.03–0.13)

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References
  10. Supporting Information

The novel findings of this study are that the BQ was observed to classify 24.3% of middle-aged Norwegians to be at high-risk of OSA, while the prevalences of OSA in the overall general population were estimated to be 16% for AHI ≥ 5 and 8% for AHI ≥ 15. However, further analyses of weighted BQ risk status and simulated OSA cases in contingency tables revealed suboptimal screening properties of the questionnaire.

The BQ risk classification properties presented in Table 2 are comparable to previous BQ risk classifications conducted in the US population (Hiestand et al., 2006; Kapsimalis and Kryger, 2009). In the US study from 2005, 56% of participants 30 years or older were classified as high-risk by the BQ. The subjects in that study were more obese (25% had a BMI > 30 kg m−2), which may have influenced the risk classification. The US study from 2007 was conducted only in females of the general population, and the estimated proportion of BQ high-risk was 25.4% compared with our estimate of 20.3%. The frequency of a BMI > 30 kg m−2 was considerably larger (23.6%) than in females in our study (13.8%). The proportions of subjects reporting simple snoring or daytime somnolence in our study were similar to the US studies. Accordingly, the difference between the BQ risk classification of our study and those of previous general population studies can be ascribed mainly to differences in the prevalence of age and obesity.

The BQ has also been used to indicate risk of OSA in numerous clinical samples defined more or less stringently. The prevalence of OSA in the general population is much lower than the prevalence observed in the majority of these samples. Netzer et al. initially found a proportion of 37.5% BQ high-risk subjects in a general practitioner sample (Netzer et al., 1999), and later a proportion of 32.3% in a large US and European general practitioner sample (Netzer et al., 2003). The prevalence of subjects with an AHI ≥ 5 in the first of these samples was 66%, which is more than four times higher than our estimates for the general population.

Although the response rate of 55.7% in our study was better than response rates of 20% and 22%, respectively, in the two US studies (Hiestand et al., 2006; Kapsimalis and Kryger, 2009), response bias might have influenced the characteristics of BQ risk classification. We believe the most likely factor of systematic selection (if any) between responders and non-responders to our questionnaire were subjective sleep complaints. We therefore also compared sleepiness in the screening sample (Table 1) with a representative, Norwegian study assessing sleepiness by the same questionnaire (Pallesen et al., 2007). The mean ESS score in this study was 6.9, which was identical to the ESS score found in our study.

The proportion of missing BQ items in the questionnaire (43.8%) could also have influenced the risk classification of the BQ. We did not find any recommendations in the existing literature on how to deal with this problem, and we therefore conducted a sensitivity analysis to estimate the maximum impact of the potential bias introduced by our strategy of scoring missing values as zero. This analysis suggested that it had only a minor effect.

Estimated screening properties of the BQ

We have not identified any other studies that have estimated the screening properties of the BQ in a general population sample. In contrast, the BQ has been validated against polysomnography for a variety of clinical samples, but with divergent results. For instance, screening properties have been reported to be substantially better for general practice (Netzer et al., 1999), general internal medicine outpatient settings (Sharma et al., 2006) and cardiology outpatient settings (Gami et al., 2004; Gus et al., 2008) than for surgical patients (Chung et al., 2008), sleep clinic patients (Ahmadi et al., 2008) and pulmonary rehabilitation patients (Weinreich et al., 2006). The divergent results across different study populations are likely due to both differences in the distribution of important confounding factors, such as snoring, sleepiness, hypertension, BMI, age, gender, ethnicity, and in the prevalence of OSA per se. Prevalences of OSA were not reported in all studies cited above, but reported proportions of AHI ≥ 5 ranged from 41% (Ahmadi et al., 2008) to 66% (Netzer et al., 1999), values that are 2.5–4 times higher than the estimated prevalence in our study. Thus, the estimated screening properties of the BQ in our study should only be generalized to other studies that have used the BQ in samples with low-prevalence of OSA.

The screening properties of the BQ presented in Table 3 suggest that the BQ applied with a scoring algorithm that identifies a single cut-off criterion is a suboptimal instrument for detecting cases among the general population. Sensitivities of 37.2 (AHI ≥ 5) to 43.0 (AHI ≥ 15) imply that less than 50% of subjects with OSA among the general population will be identified by the questionnaire. Our estimated specificities of 79.7–84.0 suggest that subjects without OSA among the general population are most likely to be true negatives. Nonetheless, Fig. 2 illustrates that BQ negativity was only associated with a modest change in the probability of having OSA (likelihood ratios of 0.7–0.8). In contrast, BQ positivity was associated with a doubling of the probability of having OSA (likelihood ratios of 2.1–2.3) compared with the population prevalences. However, posttest probabilities for OSA were still below 50%. These modest effects of a dichotomized scoring algorithm based on traditional risk factors for OSA are consistent with studies of other prediction formulas than the BQ (Harding, 2001; Pang and Terris, 2006; Rowley et al., 2000). We have therefore with interest registered two studies that have reported significant improvement of screening properties for OSA in low-prevalence samples when risk classification was graded and conducted in two phases (Gurubhagavatula et al., 2004; Takegami et al., 2009).

Estimated prevalences of OSA

The optimal diagnostic criterion for OSA remains a matter of contention and differs between studies. In the current study, prevalence was defined solely according to the AHI to enable comparison between our results and previous prevalence estimates of OSA (Lee et al., 2008), as well as six out of seven of previous BQ validation studies (Ahmadi et al., 2008; Chung et al., 2008; Gami et al., 2004; Gus et al., 2008; Netzer et al., 1999; Sharma et al., 2006; Weinreich et al., 2006). We chose not to calculate the proportion of AHI ≥ 5 with additional sleepiness as recent studies have indicated that self-reported sleepiness is not a risk factor for OSA in the general population (Bixler et al., 2005; Gottlieb et al., 1999; Young et al., 2002). The difference between BQ high-risk subjects regarding the BQ daytime somnolence category (63.1%) and excessive daytime sleepiness defined by an ESS > 10 (33.7%) in Table 2 further illustrates the problem of using measures of subjective sleep in a diagnostic criterion. Finally, neither BQ daytime somnolence nor ESS sleepiness were significant predictors in the statistical model used to estimate prevalence in our study. However, we emphasize that the estimated prevalences may have differed if another diagnostic criterion, for example one that included subjective complaints, had been used.

The exclusion of 10 cases treated with continuous positive airway pressure may have resulted in an underestimation of the proportion of subjects with OSA. However, sensitivity analyses show that adding these individuals to the clinical sample and counting them as OSA cases changed distributions across strata only marginally. Unadjusted prevalences across strata did not differ by more than 1%. Because the combination of distributions across strata and unadjusted prevalences were used in calculating weighted prevalences, we do not expect any impact of the exclusion of these 10 subjects on final prevalence estimates.

The gender- and age-specific prevalence estimates of OSA in the general Norwegian population generated by our statistical model lay approximately equal to and slightly higher than previous estimates from the Wisconsin Sleep Cohort and Pennsylvania State Cohort (Bixler et al., 1998, 2001; Young et al., 1993), but much higher than previous Scandinavian estimates (Gislason et al., 1988; Jennum and Sjol, 1992). Our estimates were slightly lower than previous estimates from Spain, Poland and Australia (Bearpark et al., 1995; Duran et al., 2001; Plywaczewski et al., 2008). Unfortunately, in most of the previous studies the BMI has not been reported. However, according to official statistics from the USA and Norway, the proportion of adults with obesity has increased by more than 10% since the first of these studies was published in 1988 (http://www.nih.gov, http://www.fhi.no). It is therefore not surprising that the estimated prevalence of OSA in the current study is higher than that noted in older US and Scandinavian estimates. The similarity between our model-based estimates of prevalences of OSA and previous estimates from the literature also strengthens the generalizability of the estimated screening properties derived from the same model.

Strengths and limitations

A strength of this study was the use of in-hospital polysomnography as the reference standard for the clinical sample. However, the use of only one single, sleep registration is a potential limitation of the study (Stepnowsky et al., 2004). Ideally, sleep should be measured twice, but in accordance with most previous studies that have assessed screening properties of the BQ or prevalence of OSA, we chose to rely on a single measure of sleep. The reliability of sleep data was further optimized by implementing the same criteria for sleep quality as was conducted in the Wisconsin sleep study (Young et al., 1993). Another concern regarding reliability of the reference standard was the use of a clinical scoring and the use of two scorers. Inter-rater reliability between these two raters was not assessed, but we believe that the most unreliable registrations were removed by excluding sleep studies of poor quality.

An important limitation of this study was the multi-stage sampling procedure and substratifications in the BQ high-risk group. Also the decision to make 5541 BQ low-risk respondents with missing BQ unavailable from the subsequent random draws introduced systematic bias to the clinical sample. To adjust for these selection mechanisms, advanced statistical simulation models were required. These models were based on the estimation sample, and we therefore assumed that this sample was representative of the screening sample. These assumptions were strengthened by the lack of differences between these samples as presented in Table 1. We also assumed that associations between self-report items used in the statistical model and OSA pathology were constant through the severity specter of the AHI. This assumption can of course be questioned, but we argue that the internal validation procedures described in the online supplement and the external validation discussed above regarding the estimated prevalences are sufficient to indicate the soundness of the model that was developed to adjust for this bias.

The final limitation of this study was the cross-sectional design that restricted the clinical interpretation of BQ high-risk status. Thus, the finding of suboptimal screening properties of the BQ against cut-off values of the AHI does not necessarily mean that the BQ high-risk classification is of low value. This question can only be answered in studies with a longitudinal design.

Conclusion and clinical implications

The BQ classified one out of four middle-aged Norwegians to be at high-risk of having OSA, but we estimated that the BQ performed suboptimally as a screening test for OSA in this low-prevalence population. Although test positivity was associated with an approximate doubling of the probability of OSA, we cannot recommend the BQ with the current scoring algorithm as a screening tool in the general population. However, this does not disqualify the instrument for use in the practitioner’s office or clinic with higher estimates of OSA prevalence. Given the burden of diseases associated with OSA, it should be the goal of future investigations to develop and validate new instruments or refine existing instruments to enhance our ability to identify subjects with OSA in the general population.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References
  10. Supporting Information

This study was supported economically from South-Eastern Norway Regional Health Authority (grant number 2004219) and the University of Oslo. We also want to thank Akershus University Hospital for providing research facilities and Faculty Division Akershus University Hospital for providing technical support (Anita Fjellum and Gunn Seim Eikeland). We finally acknowledge the staff at Akershus University Hospital, Department Stensby.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References
  10. Supporting Information
  • Ahmadi, N., Chung, S. A., Gibbs, A. and Shapiro, C. M. The Berlin questionnaire for sleep apnea in a sleep clinic population: relationship to polysomnographic measurement of respiratory disturbance. Sleep Breath., 2008, 12: 3945.
  • American Academy of Sleep Medicine. International Classification of Sleep Disorders: Diagnostic and Coding Manual, 2nd edn. American Academy of Sleep Medicine, Westchester, IL, 2005.
  • American Sleep Disorders Association. EEG arousal: scoring rules and examples. Sleep, 1992, 15: 174184.
  • Bearpark, H., Elliott, L., Grunstein, R. et al. Snoring and sleep apnea. A population study in Australian men. Am. J. Respir. Crit. Care Med., 1995, 151: 14591465.
  • Bixler, E. O., Vgontzas, A. N., Ten, H. T., Tyson, K. and Kales, A. Effects of age on sleep apnea in men: I. Prevalence and severity. Am. J. Respir. Crit. Care Med., 1998, 157: 144148.
  • Bixler, E. O., Vgontzas, A. N., Lin, H. M. et al. Prevalence of sleep-disordered breathing in women: effects of gender. Am. J. Respir. Crit. Care Med., 2001, 163: 608613.
  • Bixler, E. O., Vgontzas, A. N., Lin, H. M., Calhoun, S. L., Vela-Bueno, A. and Kales, A. Excessive daytime sleepiness in a general population sample: the role of sleep apnea, age, obesity, diabetes, and depression. J. Clin. Endocrinol. Metab., 2005, 90: 45104515.
  • Chung, F., Yegneswaran, B., Liao, P. et al. Validation of the Berlin questionnaire and American Society of Anesthesiologists checklist as screening tools for obstructive sleep apnea in surgical patients. Anesthesiology, 2008, 108: 822830.
  • Cochran, W. G. Sampling Techniques, 2nd edn. Wiley, New York, NY, 1963.
  • Duran, J., Esnaola, S., Rubio, R. and Iztueta, A. Obstructive sleep apnea-hypopnea and related clinical features in a population-based sample of subjects aged 30 to 70 year. Am. J. Respir. Crit. Care Med., 2001, 163: 685689.
  • Gami, A. S., Pressman, G., Caples, S. M. et al. Association of atrial fibrillation and obstructive sleep apnea. Circulation, 2004, 110: 364367.
  • Gislason, T., Almqvist, M., Eriksson, G., Taube, A. and Boman, G. Prevalence of sleep apnea syndrome among Swedish men – an epidemiological study. J. Clin. Epidemiol., 1988, 41: 571576.
  • Gottlieb, D. J., Whitney, C. W., Bonekat, W. H. et al. Relation of sleepiness to respiratory disturbance index: the Sleep Heart Health Study. Am. J. Respir. Crit. Care Med., 1999, 159: 502507.
  • Gozal, D., Kheirandish-Gozal, L., Capdevila, O. S., Dayyat, E. and Kheirandish, E. Prevalence of recurrent otitis media in habitually snoring school-aged children. Sleep Med., 2008, 9: 549554.
  • Gurubhagavatula, I., Maislin, G., Nkwuo, J. E. and Pack, A. I. Occupational screening for obstructive sleep apnea in commercial drivers. Am. J. Respir. Crit. Care Med., 2004, 170: 371376.
  • Gus, M., Goncalves, S. C., Martinez, D. et al. Risk for obstructive sleep apnea by Berlin Questionnaire, but not daytime sleepiness, is associated with resistant hypertension: a case-control study. Am. J. Hypertens., 2008, 21: 832835.
  • Harding, S. M. Prediction formulae for sleep-disordered breathing. Curr. Opin. Pulm. Med., 2001, 7: 381385.
  • Hiestand, D. M., Britz, P., Goldman, M. and Phillips, B. Prevalence of symptoms and risk of sleep apnea in the US population: results from the national sleep foundation sleep in America 2005 poll. Chest, 2006, 130: 780786.
  • Jennum, P. and Sjol, A. Epidemiology of snoring and obstructive sleep apnoea in a Danish population, age 30–60. J. Sleep Res., 1992, 1: 240244.
  • Johns, M. W. A new method for measuring daytime sleepiness: the Epworth sleepiness scale. Sleep, 1991, 14: 540545.
  • Kapsimalis, F. and Kryger, M. Sleep breathing disorders in the U.S. female population. J. Womens Health, 2009, 18: 12111219.
  • Lee, W., Nagubadi, S., Kryger, M. H. and Mokhlesi, B. Epidemiology of obstructive sleep apnea: a population-based perspective. Expert. Rev. Respir. Med., 2008, 2: 349364.
  • Netzer, N. C., Stoohs, R. A., Netzer, C. M., Clark, K. and Strohl, K. P. Using the Berlin Questionnaire to identify patients at risk for the sleep apnea syndrome. Ann. Intern. Med., 1999, 131: 485491.
  • Netzer, N. C., Hoegel, J. J., Loube, D. et al. Prevalence of symptoms and risk of sleep apnea in primary care. Chest, 2003, 124: 14061414.
  • Pallesen, S., Nordhus, I. H., Omvik, S., Sivertsen, B., Tell, G. S. and Bjorvatn, B. Prevalence and risk factors of subjective sleepiness in the general adult population. Sleep, 2007, 30: 619624.
  • Pang, K. P. and Terris, D. J. Screening for obstructive sleep apnea: an evidence-based analysis. Am. J. Otolaryngol., 2006, 27: 112118.
  • Partinen, M. and Gislason, T. Basic Nordic Sleep Questionnaire (BNSQ): a quantitated measure of subjective sleep complaints. J. Sleep Res., 1995, 4: 150155.
  • Plywaczewski, R., Bednarek, M., Jonczak, L. and Zielinski, J. Sleep-disordered breathing in a middle-aged and older Polish urban population. J. Sleep Res., 2008, 17: 7381.
  • Punjabi, N. M., Caffo, B. S., Goodwin, J. L. et al. Sleep-disordered breathing and mortality: a prospective cohort study. PLoS. Med., 2009, 6: e1000132.
  • Rechtshaffen, A. and Kales, A. A Manual of Standardized Terminology and Scoring System for Sleep Stages of Human Sleep. University of California at Los Angeles, Brain Information Service/Brain Research Institute, Los Angeles, CA, 1968.
  • Rodenstein, D. Driving in Europe: the need of a common policy for drivers with obstructive sleep apnoea syndrome. J. Sleep Res., 2008, 17: 281284.
  • Rowley, J. A., Aboussouan, L. S. and Badr, M. S. The use of clinical prediction formulas in the evaluation of obstructive sleep apnea. Sleep, 2000, 23: 929938.
  • Sharma, S. K., Vasudev, C., Sinha, S., Banga, A., Pandey, R. M. and Handa, K. K. Validation of the modified Berlin questionnaire to identify patients at risk for the obstructive sleep apnoea syndrome. Indian J. Med. Res., 2006, 124: 281290.
  • Sivertsen, B., Overland, S., Glozier, N., Bjorvatn, B., Maeland, J. G. and Mykletun, A. The effect of OSAS on sick leave and work disability. Eur. Respir. J., 2008, 32: 14971503.
  • Somers, V. K., White, D. P., Amin, R. et al. Sleep apnea and cardiovascular disease: an American Heart Association/American College Of Cardiology Foundation Scientific Statement from the American Heart Association Council for High Blood Pressure Research Professional Education Committee, Council on Clinical Cardiology, Stroke Council, and Council On Cardiovascular Nursing. In collaboration with the National Heart, Lung, and Blood Institute National Center on Sleep Disorders Research (National Institutes of Health). Circulation, 2008, 118: 10801111.
  • Stepnowsky, C. J., Jr, Orr, W. C. and Davidson, T. M. Nightly variability of sleep-disordered breathing measured over 3 nights. Otolaryngol. Head Neck Surg., 2004, 131: 837843.
  • Takegami, M., Hayashino, Y., Chin, K. et al. Simple four-variable screening tool for identification of patients with sleep-disordered breathing. Sleep, 2009, 32: 939948.
  • Weinreich, G., Plein, K., Teschler, T., Resler, J. and Teschler, H. [Is the Berlin questionnaire an appropriate diagnostic tool for sleep medicine in pneumological rehabilitation?]. Pneumologie, 2006, 60: 737742.
  • Young, T., Palta, M., Dempsey, J., Skatrud, J., Weber, S. and Badr, S. The occurrence of sleep-disordered breathing among middle-aged adults. N. Engl. J. Med., 1993, 328: 12301235.
  • Young, T., Evans, L., Finn, L. and Palta, M. Estimation of the clinically diagnosed proportion of sleep apnea syndrome in middle-aged men and women. Sleep, 1997, 20: 705706.
  • Young, T., Shahar, E., Nieto, F. J. et al. Predictors of sleep-disordered breathing in community-dwelling adults: the Sleep Heart Health Study. Arch. Intern. Med., 2002, 162: 893900.
  • Young, T., Finn, L., Peppard, P. E. et al. Sleep disordered breathing and mortality: eighteen-year follow-up of the Wisconsin sleep cohort. Sleep, 2008, 31: 10711078.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References
  10. Supporting Information

Supplement. Online supplement to a Norwegian population based study on the risk and prevalence of OSA.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
JSR_861_sm_supplement.doc79KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.