A New Screening Tool (BAMSA) for Sleep Apnea in Male Professional Truck Drivers

1. Introduction

Obstructive sleep apnea (OSA) is a common disorder causing daytime sleepiness. In the general population, the prevalence of sleep apnea defined by AHI≥5 has ranged from 9-38 % in men and 1-17 % in women.^1,2 It is even more prevalent among professional truck drivers. In recent studies prevalence figures as high as 28-78% have been reported.^3–7 Daytime sleepiness increases the risk of traffic accidents. In Finland, driver sleepiness/fatigue is a contributing factor in 17-19% of fatal vehicle accidents.^8,9 In 2013 European driving license committees working group estimated that sleep apnea and narcolepsy together cause a relative risk elevation of 271 % (RR = 3.71) for traffic accidents, and most of it is due to sleep apnea.¹⁰

OSA is classified as mild, moderate, or severe according to apnea-hypopnea-index (AHI). European Union directives rule to assess daytime sleepiness yearly in professional drivers with at least moderate sleep apnea (AHI≥15). Obstructive sleep apnea syndrome is a treatable condition, e.g., by continuous positive airway pressure (CPAP).^11,12 That is why it is important to detect the possibility of OSA in occupational health care. Also, evidence shows that CPAP treatment reduces the risk of motor vehicle accidents. ^13,14

Previously different tools for predicting sleep apnea have been described. Some of those require oximetry data,^15–17 some require clinical evaluations of anatomy (for example Mallampati score) ¹⁸ among them DES-OSA Score. ¹⁹ Also automatic detection of OSA by speech signals has been studied.²⁰ Different prediction tools with questionnaires and some measurements, for example neck circumference or systolic blood pressure, but no clinical evaluation, have emerged. Among these are SAS Score, ²¹ No-Apnea, ²² STOP questionnaire, ²³ STOP-Bang,^24,25 NoSAS score ^26,27 and Four-Variable Screening Tool (Four-V). ²⁸

Of the prediction tools presented before, Berlin Questionnaire, ^29–31 a clinical prediction rule presented by Miranda Serrano et al. ³² and Multivariable Apnea Prediction (MAP) Index ³³ are more easily accomplished since they include only questions to be answered by the patient and the only measurement is body mass index (BMI).

Also Epworth Sleepiness Scale (ESS) has been suggested to predict the probability of sleep apnea and several studies with different outcomes have been published before. ^34,35 However, ESS did not appear to correlate with sleep-disordered breathing (SDB) in women, and prediction value in men was modest at best with an average ESS score of 9,1 for no SDB, and 10,1 for severe SDB. ³⁶ Several studies have compared the accuracy of different predictive tools and also meta-analysis has been done. ^37–39

There is still a need for an easy and cheap method to recognize the ones that should undergo sleep studies, especially in occupational health care. In this study we wanted to estimate the usefulness of various screening tools in detecting OSA and to establish a new, simple measure that does not require any clinical examination by a clinician, for screening of OSA.

2. Methods

A structured questionnaire was sent to 2066 professional truck drivers living in Southern Finland as described earlier.⁷ We selected the subjects from the Finnish Truck Drivers’ (Rahtarit ry) registry. 175 of these drivers had a clinical examination. The sleep studies included cardiorespiratory polygraphy as well as tests of daytime sleepiness, also maintenance of wakefulness test (MWT). Sleep studies were done successfully in 172 drivers. From these drivers we included in this BAMSA study only those, who had complete information so that we could compute also STOP-Bang, NoSAS, ESS and the BERLINscore for everyone. Altogether 160 men remained.

2.1. Questionnaire

Demographic variables, including age, weight, height and neck circumference, were asked. They were measured also during the medical visit. BMI-values were highly correlated (r = 0.9675). The BMI during the visit was a bit higher than BMI derived from the questionnaire (BMI(visit) = 2.36 + 0.913*BMI(questionnaire); R²=0.9386). The correlation of neck circumference (NC) was less good (r=0.733; NC(visit) = 18.47 + 0.548*NC(questionnaire); R²=0.537). We used BMI and neck circumference as they were measured because they were always available. Age at the time of visit was used.

All subjects filled in the questionnaire modified from the Basic Nordic Sleep Questionnaire (BNSQ). ⁴⁰ In addition, they answered questions about their driving history, sleepiness while driving and other driving related items.

The question on snoring (from the BNSQ) was: “Do you snore while sleeping (ask other people if you are not sure)?” History of subjective sleep apnea (from the BNSQ) was asked as: Have you had breathing pauses (sleep apnea) at sleep (have other people noticed that you have pauses in respiration when you sleep)?

The question on quality of snoring was: “How do you snore (ask other people about the quality of your snoring)?” Response alternatives were: 1) “I don’t snore”, 2) “my snoring sounds regular and it is of low voice”, 3) “it sounds regular but rather loud”, 4) “it sounds regular but it is very loud (other people hear my snoring in the next room)”, 5) “I snore very loudly and intermittently (there are silent breathing pauses when snoring is not heard and at times very loud snorts with gasping”.

The question on witnessed apneas was: “Have you had breathing pauses (sleep apnea) at sleep (have other people noticed that you have pauses in respiration when you sleep)?” Response alternatives were: 1) “never or less than once per month”, 2) “less than once per week”, 3) “on 1-2 nights per week”, 4) “on 3-5 nights per week”, and 5) “every night or almost every night”.

We also asked about daytime fatigue, tiredness and sleepiness. In addition, Epworth Sleepiness Scale was filled by the responders.

2.2. The Berlin Score, STOP-BANG, NoSAS and ESS criteria

The Berlin Questionnaire includes questions of three categories.²⁹ The first one includes questions of snoring and breathing pauses, the second one of tiredness and falling asleep while driving and the third one of high blood pressure and body mass index (BMI). The subject is categorized in high risk group when having at least two positive categories (this means answering positively in two questions of categories one and two, one question in category three or BMI>30).

The 8-item STOP-BANG score was computed. ²⁴ It includes questions of snoring, tiredness, breathing pauses, high blood pressure, BMI, age, sex and neck circumference with the options yes or no and the risk for sleep apnea increases with the higher scores. We have computed the results with older STOP-Bang criteria and also with the new one, which has different neck circumference criteria for male and female. ⁴¹

The NoSAS score is one of the previously used screening tools. ²⁶ It ranges from 0 to 17, includes questions on neck circumference, BMI, snoring, sex and age and with 8 points or more the individual is at high risk for OSA.

Epworth Sleepiness Scale (ESS) is also presented to be a possible predictor of OSA. ³⁴ In that questionnaire the subject is asked to evaluate the probability of falling asleep in eight different situations and total score ranges from 0 to 24.

2.3. Design of the BAMSA criteria

First, we computed sensitivity, specificity and the receiver operating characteristic (ROC) for detecting sleep apnea using different screening tools. We used also our empirical experience, since late 1970′s, in diagnosing patients with sleep apnea. We wanted to design a tool that is as simple as possible, yet sensitive and especially specific enough. We also wanted to have a screening tool that is based mainly on anamnestic information using semi-quantitative questions instead of dichotomic yes/no alternatives for occurrence of snoring and occurrence of sleep apnea.⁴⁰ The reason for this is that almost all humans snore at least sometimes if they are sleeping in supine position and if they have been drinking some alcohol. For this reason, “snoring - yes” differs from “snoring at least on one night per week”. For sleep apnea the cutpoint for “yes” was: having witnessed breathing pauses (apneas) at least on one night per month.

The new BAMSA screening tool consists of 5 questionnaire (interview) items: 1) BMI>30 kgm^-2 (no=0, yes=1); 2) Age>50 years (no=0, yes=1); 3) Gender (female=0, Male=1); 4) Snoring at least on one night per week (from the BNSQ; no=0, yes=1); 5) Having had at least sometimes breathing pauses (Apneas) during sleep (BNSQ response >1; no=0, yes=1).

2.4. Clinical examination

All subjects who were monitored in the sleep laboratory underwent a thorough medical examination. Weight and height were measured. The shape of the hard and soft palate was tabulated as well as size and shape of tonsils, tongue, jaw (micrognathia, retrognathia) and bite (overjet, signs of bruxism etc.). The clinician measured the neck and waist circumferences and tabulated the cricomental space in three classes: normal, slightly abnormal and abnormal. Pulse rate and blood pressure were measured and a neurological examination was performed.

2.5. Sleep recordings

All 160 subjects underwent full-night cardiorespiratory polygraphic sleep recordings as described earlier. ⁷ We recorded nasal airflow and thoracic and abdominal respiratory movements as well as snoring sounds, sleep position, pulse and oxygen saturation. We used a limit of ≥ 4% from pre-event baseline for oxygen desaturations as defined by AASM 1999 Type A definition / AASM 2.2 Criteria 1B for the detection of hypopnea. ^42,43 The 1999 rules with the 4% desaturation criteria has been reported to give essentially similar results as the current AASM rules. ⁴³ We categorized the severity of obstructive sleep apnea (OSA) into four groups by apnea-hypopnea index (AHI): AHI <5/h (normal/no sleep apnea), 5≤ AHI <15 (mild), 15≤ AHI <30 (moderate) and AHI≥30 (severe).

2.6. Statistics

We did statistical analyses with the STATA versions 15.1 and 17.0 (StataCorp, USA). We used medians, means, standard deviations (SD) and range to describe distributions. We tested statistical differences between groups using Student’s t-test, Mann-Whitney’s U-test or the Kruskall-Wallis test depending on the number of groups and nature of the distribution. We tested the normality of the distribution by the Shapiro-Wilk test ⁴⁴. Pearson’s chi-square test was computed from the cross-tabulations. Receiver operating characteristic (ROC) analysis and the area under the curve were computed when applicable. Sensitivity, specificity, positive predictive values (PPV) and negative predictive value (NPV), positive likelihood ratio (LR+) and negative likelihood ratio (LR-) were computed using the .diagt command of STATA. ⁴⁵ Odds ratios (OR) of a test with a specific cutpoint (LR+/LR-) were computed.

2.7. Ethical considerations

The ethical committee of Helsinki University Hospital approved this study. This study obeyed guidelines and instructions of the Declaration of Helsinki. Every subject fulfilled a written informed consent. All information was strictly confidential and the access to all the data required being an investigator in this study. If moderate or severe sleep apnea was diagnosed, the subjects had the opportunity to give permission to send the data to the public health care to receive the treatment for sleep apnea. The employers of the truck drivers did not get any data.

3. Results

3.1. Characteristics of the drivers

Altogether 160 subjects were men. Their mean age was 42.0 years (SD 9.3, median 41.4, range 20-58). 39 men (24.4%) were over 50 years old. Mean BMI was 28.3 kgm^-2 (SD 5.1, median 27.3, range 19.4 – 46.2). 49 men (30.6%) were obese with BMI>30 kgm^-2. Mean neck circumference was 41.2 cm (SD 3.1, median 41, range 35-52). 31 men (19.4%) had neck circumference >43 cm. The mean waist circumference was 101.4 cm (SD 15.2; median 99.5 cm, range 67.5-160). In 74 men (46.25%) waist circumference was >100 cm, and it was >110 cm in 36 men (23.1%). 27 men (16.9%) used antihypertensive medication and/or had a diagnosis of arterial hypertension.

Snoring at least on 3 nights per week was reported by 83 men (51.9%) and this also includes habitual, almost every night snoring, which was reported by 70 men (43.75%). 27 men (16.9%) reported that they never snore. Loudness (quality) of snoring associated with frequency of snoring history. All men reporting very loud and intermittent snoring (n=69) snored at least on 3 nights per week. 60 of them (87%) were habitual snorers.

Breathing pauses (apneas) at least sometimes were reported by 86 men (53.75%). Apneas at least on 3 nights per week were reported by 53 men (33.1%) and this also included those 36 men (22.5%) who reported having apneas almost every night. 34 (94.4%) of these 36 men reported snoring almost every night. 74 men (46.3%) reported that they have never had witnessed apneas.

Apnea-hypopnea index (AHI) was ≥ 5 in 70 drivers (43.75%), ≥15 in 17.5% (n=28), and ≥30 in 7.5% (n=12) of these 160 subjects. 53.9% of those snoring at least on three nights per week had AHI≥5. Only one of the drivers without any prior history of snoring had AHI≥30. His BMI was 32.7kg/m² and he was 57 years old. On the other hand, 33.3% of them had AHI≥5 and 7.4% (n=2) had AHI≥15 (one had AHI=52 and the other had AHI=18). In other words, a simple absence of snoring did not completely exclude severe sleep apnea.

Altogether 33 men (20.6%) reported tiredness on at least three days per week. The mean ESS was 8.7 (SD 3.8, median 9, range 1-19). ESS was >10 in 47 (29.4%) and >15 in 8 (5.0%). 28.6% of those who had AHI≥15 had ESS>10 and 29.6% of those who had AHI<15 had ESS>10. ESS was over 10 points in 31.4% of those who had AHI≥5 and in 27.8% of those who had AHI<5. ESS was on average 8.6 (SD 3.7, median 8) among those who had AHI≥5 and 8.7 (SD 3.8, median 8.5) among those who had AHI<15.

3.2. Performance of different screening tools

For presence of sleep apnea (AHI≥5) the ROC area for the BERLIN score was 0.660. The ROC area of the original STOP-BANG score was 0.724. The ROC area of the STOP-BANG score with the new neck circumference criteria (STOP-BANGnew) was 0.734. The ROC area for the NoSAS score was 0.698. The ROC area for ESS was 0.524 (for ESS≥11 the ROC area was 0.518). The ROC area of BAMSA for AHI≥5 was 0.741. The results for BAMSA were the best for detecting AHI≥5 (see Table 1). The ROC area of Epworth sleepiness score was close to 0.5.

For detecting AHI≥15 the sensitivity and specificity of BAMSA with a cutpoint of 3 were 100% and 47%, respectively. With a BAMSA cutpoint of 4 they were 85.7% and 78.8%, respectively (see Table 2).

Using AHI≥15 the overall ROC area for BAMSA was 0.862. The ROC area of the BAMSA score was statistically significantly better than that of the original STOP-BANG (P=0.085), the Berlin score (P=0.0005), ESS (P<0.0001) and NoSAS (P=0.027). There was no statistically significant difference between BAMSA and the modified STOP-BANGnew score with the new neck circumference criteria (P=0.342) (see Figure 1).

As all subjects were men, we analyzed the data by excluding gender from the formulas. The ROC areas of the different screening tools did not change. We wanted to keep the male gender in BAMSA questionnaire since it is a known risk factor for sleep apnea and common among professional truck drivers.

4. Discussion

Our aim was to create and validate a screening tool for professional drivers that is easy to use without any contact with the health care. We introduce BAMSA (0 to 5 points), a simple five-question tool that is shown to be at least as sensitive and more specific than the other tested screening tools to detect sleep apnea. It detects mild sleep apnea and especially moderate to severe sleep apnea very well. It includes four questions (no = 0 points, yes = 1 point) and measurement of BMI (> 30 kgm^-2 = 1 point), which is easy to measure at home by the subject himself. With this prediction tool it is also easy for a clinician to determine which patients to send for sleep laboratory studies. The ROC area of BAMSA for AHI≥5 was 0.741. BAMSA did even better in detecting moderate and severe sleep apnea and the ROC area of BAMSA for AHI≥15 was 0.862. When using the cutpoint of 4 points, the sensitivity of BAMSA in detecting AHI≥15 was 85.7% and the specificity was 78.8%.

STOP-BANG is one of the best validated and most commonly used screening tools. ²⁴Recently the criteria for STOP-BANG, particularly concerning the neck circumference criteria, have been updated. ⁴¹ It inquires about the measurement of neck circumference, and the criteria are different depending on the gender. The scoring of the original ²⁴ and also the new STOP-BANG ⁴¹ are somewhat more complicated than the scoring of BAMSA, and it has different options for evaluating the subject in high-risk group. The NoSAS Score ²⁶ includes sex, age, BMI, snoring and neck circumference and it also has a rather complicated scoring system. Berlin Questionnaire is another previously established screening tool, ²⁹ which does not require any measurements but again, the scoring is complicated. Epworth sleepiness scale is a widely used questionnaire to test for subjective daytime sleepiness. It has been suggested to be an option for predicting the risk of sleep apnea, ³⁴ but the results have not been promising. ³⁶ In fact, Lee et al. have suggested that ESS should not be used in clinical settings. ⁴⁶ Also, in our study ESS performed poorly and its predictive value for sleep apnea was close to tossing a coin.

A screening tool should be sensitive with a good negative predictive value to assure that there are not many false negatives. At the same time, it should be specific enough to avoid high rate of false positive suspicion of sleep apnea – to avoid unnecessary polygraphic sleep recordings, and to avoid causing unnecessary anxiety for the screened subjects. In this study, BAMSA showed a good sensitivity in detecting clinically relevant sleep apnea (AHI≥15), 86.2 % (cutpoint 4) or even 100 % (cutpoint 3). Specificity was also better than that of the other tested prediction tools (tables 1 and 2). This means that not only BAMSA detects almost all of the cases, but it also detects them more accurately than the other prediction tools.

The strengths of our study include the fact that our population of professional truck drivers consisted assumed healthy individuals as well as those with suspected sleep apnea. Our study population is rather large, which is a strength. The evaluation of occurrence of snoring is based on a quantitative time scale instead of purely subjective evaluation. ⁴⁰ The words such as “sometimes” and “often” may mean different things for different people, while for example “on 3 nights per week” means exactly the same for everybody. Many of the populations presented in previous studies include only the ones that have been addressed to sleep studies with suspected sleep apnea or other sleep disorder. ^{19,27,47–49} Also, as far as we know, this is the first screening tool to be used in professional drivers.

There are also limitations in our study. Our study population consisted of male truck drivers only. Therefore, our results cannot be generalized neither into female populations nor other populations than professional drivers. We have started to validate BAMSA also in other populations, including women. We can already conclude that BAMSA can be used as a screening tool for clinically significant sleep apnea in male populations. We used respiratory polygraphy instead of full polysomnography, which might have found more subjects with mild sleep apnea. However, we are confident that all subjects with clinically relevant sleep apnea were found.

BAMSA is a valid tool that may be used instead of, or in combination with other screening tools, for example STOP-BANG, in making inferences on prior probability of sleep apnea. Anyway, although BAMSA is working well in male populations, we need larger populations including women and younger subjects before we can conclude that it works well in all clinical settings.

5. Conclusions

In conclusion, we have created a sensitive and specific screening tool for sleep apnea in male professional drivers. BAMSA is easy to use and more simple than the other screening tools established before. BAMSA only consists of five yes or no questions. The cutpoint of 3 seems to be able to find all of the cases and, if using a cutpoint of 4, the specificity is high. Using BAMSA can help find professional male truck drivers that are in risk of having sleep apnea and especially the ones that should be sent to sleep laboratory tests.