University of Birmingham Undiagnosed obstructive sleep apnoea and physical activity in older manual workers

Cardiovascular disease (CVD) is a negative health outcome of Obstructive Sleep Apnoea 2 (OSA). Risk factors associated with OSA development include low physical activity (PA), 3 high body mass index (BMI), and increasing age (>50 years) and weight loss is usually 4 recommended as treatment. This cross-sectional study examined the association between PA, 5 BMI and OSA severity in manual workers. Fifty-five participants, (23 females, 32 males) 6 mean age 55.2, were examined for OSA and completed a PA and anthropometric assessment. 7 On average, OSA severity was mild, PA levels were moderate and 32% of the sample was 8 classified as obese. PA was negatively associated with OSA severity, but BMI strongly 9 independently predicted OSA severity, with no evidence of mediation. As both PA and BMI 10 were significantly associated with OSA in older manual workers, increasing PA should also 11 be a focus of treatment for OSA. 12 13 16 17 18 19

excessive daytime sleepiness and morning headaches may present their own problems during 1 the course of the working day (Banno & Kryger, 2005), such that work efficiency may be 2 compromised and could result in sickness presenteeism (attending work whilst ill) (Guertler, 3 Vandelanotte, Short, Alley, Schoeppe, & Duncan, 2015). Sleep loss owing to OSA becomes 4 a real problem in the workplace in terms of production and even health and safety (Gaultney 5 & Collins-McNeil, 2009). This is particularly true of older workers in manual occupations, 6 where non-health related outcomes of OSA such as difficulty in staying awake, may lead to 7 an increased risk of workplace injury (Heaton, Azuero, & Reed, 2010). Occupations that 8 involve driving carry a high risk of road traffic accidents where OSA is concerned (Lemos,9 Marqueze, Sachi, Lorenzi-Filho, & de Castro Moreno, 2008). Road traffic accidents relating 10 to OSA are costly to industry and to lives and have been evidenced to warrant that American 11 commercial drivers in Philadelphia are screened before being deemed fit for work 12 (Gurubhagavatula, Nkwuo, Maislin, & Pack, 2008). 13 OSA-related sleepiness might be even more of a problem in older manual workers, 14 who are already likely to experience age-related fatigue (Kiss, De Meester, & Braeckman, 15 2008). As a result of daytime sleepiness, adults aged over 50 years old with OSA are 16 unlikely to practice good health behaviours such as regular PA, a lack of which has been 17 shown to be associated with moderate to severe OSA (Peppard & Young, 2004;Simpson et 18 al., 2015). Chennaoui, Arnal, Sauvet, and Leger (2015)  It is also important to bear in mind that the ability to perform manual work 8 competently without negative effect to health requires cardiorespiratory fitness, which may 9 be achieved through regular leisure-time PA (Leino-Arjas, Solovieva, Riihimaki, Kirjonen, & 10 Telama, 2004). As such, strenuous workloads that regularly exceed a worker's ability due to 11 lack of fitness will increase the risk of long-term sickness absence and the development of 12 CVD. On the contrary, increased leisure-time PA may decrease these risks (Holtermann,13

Participants and Design 4
Participants were recruited by the researcher through advertising in an original population of 5 225 manual workers. The group was comprised of (healthcare workers (e.g. nurses, 6 paramedics and healthcare assistants), Physical Education teaching teachers,e.g. Physical 7 Education (PE) and craft/labourerss) aged over 50 years old (mean = 57.11, standard 8 deviation (SD) = 5.62 years) who were recruited in 2015 as part of a study on presenteeism in 9 manual workers from various industries (Thogersen-Ntoumani, Black, Lindwall, Whittaker, 10 & Balanos, 2017). Those who already had a diagnosis of and were receiving treatment for 11 OSA, or were under current investigations for OSA, or had a history of cardiorespiratory 12 conditions were excluded. Participants who met the inclusion criteria i.e. had no previous 13 diagnosis of OSA, who had no history of cardiorespiratory conditions, were otherwise 14 healthy and were not receiving treatment for OSA were included and invited for testing. 15 Fifty-five participants (healthcare workers, PE teaching teachers and craft/labourers) willing 16 to undertake further investigation were invited to complete a questionnaire to measure daily 17 levels of physical activity, and to wear a home sleep screening device overnight to identify 18 the presence of obstructive sleep apnea. Participants were told that the ApneaLink would 19 indicate the degree of OSA but was not an official diagnosis. Results of the screening were 20 given to the participants in confidence via a telephone call or in person. Participants were 21 encouraged to see their GP if our screening was suggesting that OSA was present. The 22 University of Birmingham STEM ethics committee approved the study and written informed 23 consent was given by all participants. 24

Measures 25
Physical Activity Questionnaire. The Baecke Questionnaire (Baecke, Burema, & Frijters, 1 1982) was used to measure habitual levels of physical activity. The Baecke Questionnaire 2 measures physical activity across three contexts: sport/exercise, occupational (referred to as 3 work), and leisure-time physical activity in the past month, and provides total and domain-4 specific scores. The questionnaire includes a rating of amount (hours per week, months per 5 year) and type (swimming, jogging etc.) of activity. It has been tested in a group of workers 6 over one month with a test-retest reliability of 0.71 (Philippaerts & Lefevre, 1998). Participants wore the device at home and were scored for OSA based on an 1 apnoea/hypopnoea index (AHI). As per the ApneaLink TM default settings, apnoea was 2 defined as a reduction in flow by 20% of normal for at least 10 seconds and hypopnoea as a 3 reduction in flow by 70% of normal for at least  HoMedics Group Ltd., Kent, UK) was used to measure weight, body fat percentage (through 12 bioelectrical impedance) and BMI (kg/m 2 ). A Bosch PLR 30c Laser Measure (Robert Bosch 13 GmbH, Germany) was used to measure height. 14

Procedures 15
A participant information sheet with details on the use of the ApneaLink TM device was 16 provided alongside a demonstration on how to wear it. Participants were shown by the 17 researcher how to operate the device and given a paper illustration for reference. Participants 18 were required to wear the device on a "normal/regular" night and to go to sleep at their 19 habitual bedtime. Wear time was required overnight for one night and the device was 20 collected the following day. All participants wore the ApneaLink TM for the minimum 21 computer generated summary report with automatic scoring for interpretation of AHI score 23 and presence of OSA. Graphical readouts from the ApneaLink TM were checked by trained 24 researchers to ensure that they reflected the AHI score given. 25

Data Analysis 1
Data analysis was conducted using IBM Statistical Package for the Social Sciences (SPSS) 2 version 22. Firstly, correlations and ANOVAs were performed to explore potential 3 associations between OSA severity, PA, socio-demographic variables and BMI. Further 4 exploration of the associations between OSA, PA and BMI was carried out using linear 5 regression analysis. Change in R-squared is reported as the effect size from the regression 6 analyses. Mediation analyses were conducted using the PROCESS macro (Hayes, 2013) 7 based on simple linear regression modelling. 8

Descriptive Statistics 10
Descriptive statistics for all participants are shown in Table 1. The mean (SD) age of the 11 participants was 55.2 (4.21) years, and the sample included slightly more males (58.2%) than 12 females. The mean AHI score was 8.9 (11.75) indicating undiagnosed levels of mild OSA. 13 Mean Physical Activity (PA) levels were reported as moderate, and 32% of the sample was 14 classified as obese with a BMI > 30 kg/m 2 . PA levels were further broken down to reveal 15 high levels of work PA (heavy lifting and walking), moderate leisure PA and lower levels of 16 sport and exercise (e.g. running, swimming, cycling) (Table 1) 14 19 participants had a lower BMI than the group mean of 28.50 kg/m 2 . Further, the lowest BMI 20 for those with mild OSA was 21 kg/m 2 . There were no significant gender differences for 21 OSA severity or PA. Age was also unrelated to OSA severity and PA. 22

23
OSA severity and physical activity. Correlation revealed that PA was significantly and 24 negatively associated with OSA severity, r(54) = -0.29, p = 0.04, such that those who engaged in less physical activity had greater OSA severity. A follow-up ANOVA between 1 the four OSA severity groups (no, mild, moderate, severe) revealed no significant overall 2 main effect of group, F(3,53) = 1.98, p = 0.13, however, pairwise post-hoc comparisons 3 demonstrated significantly lower levels of PA in those with moderate OSA compared to those 4 without OSA, p = 0.037 (Figure 1). There were no other significant group differences for 5 PA. 6 OSA severity, physical activity and BMI. The correlation between OSA severity and BMI 7 was significant, r(54) = 0.51, p < 0.001, such that those with higher BMI had higher OSA 8 severity. BMI, as would be expected, was also significantly different between the OSA 9 severity groups, F(3,50) = 7.7, p < 0.001, η 2 = 0.316. Post-hoc analysis demonstrated 10 significant differences between the OSA severity groups for BMI ( Figure 2). Further, BMI 11 and PA were significantly positively associated, r(54) = 0.29, p = 0.03. Interestingly, when 12 BMI was added into a regression model predicting OSA severity from PA, the significant 13 negative association between PA and OSA severity (β = -0.29, p = 0.04, ΔR 2 = 0.08) became Tenhunen, Lyytikainen, Wiklund, Cong, Saarinen, Tarkka, Partinen, and Cheng (2015) who 5 recently found that low levels of PA were not associated with OSA in individuals aged over 6 50 years old. However, the present study also showed BMI to be a stronger indictor of 7 undiagnosed OSA than PA, implying that BMI may be a higher risk for OSA than low PA. 8 That BMI was the stronger (and only significant) predictor in the model that included PA 9 could be due to two possibilities. First, that there was reduced power to find effects given the 10 moderate sample size and the correlation between these variables, or second that BMI is a 11 multi-faceted variable, and contains variance not determined by PA, which may contribute 12 through different (non-PA related) mechanisms to OSA severity. However, In observational 13 research it is difficult to separate these aspects. However, this finding corresponds with 14 However, the present study also found levels of currently undiagnosed mild OSA in those 16 with a lower BMI than that classified as obese (>30 kg/m 2 ) suggesting that BMI may not be 17 the sole trigger for the development of OSA. 18 Additionally, participants in the present study without OSA reported significantly higher 19 levels of PA than those with moderate OSA. Arguably, there is evidence to suggest that BMI 20 is a risk factor for OSA (Koyama et al., 2012) and previous research has shown that OSA 21 severity increases with body weight (Butner et al., 2013). Additionally, it has been suggested 22 that the only effective method of reducing OSA severity is through weight-loss (Young, 23 Peppard, & Gottlieb, 2002). Interestingly, the capacity to exercise is thought to be associated 24 with age and weight and therefore, those who are older and have higher BMI may find it 25 harder to exercise (Butner et al., 2013). In contrast, it is also logical that those who do not 1 engage in much physical activity might also develop a higher BMI (Hankinson, Daviglus, 2 Bouchard, Carnethon, Lewis, Schreiner, Liu, & Sidney, 2010). The present study partially 3 confirms the theory of Butner et al. (2013) in that BMI was found to be associated with levels 4 of PA, although the direction of causality between these variables obviously cannot be 5 inferred. It might be assumed that low PA may play a part in the development of OSA, but 6 its role in the development or prevention of OSA is unclear. Previous evidence has 7 demonstrated a lack of clarity in the ability of PA to reduce OSA severity in those aged over 8 50 years old with or without weight-loss (Schobersberger, 2013; Sengul, Ozalevli, Oztura, 9 Itil, & Baklan, 2011). The question as to whether PA or BMI form the strongest risk factor of 10 OSA severity is difficult to entangle given the expected and observed correlation between the 11 two, and this is illustrated by a recent review, which explored the capacity of PA to reduce 12 It is possible that other health behaviours such as diet, smoking, and alcohol 4 consumption that were not included in the present study may have had a bearing on the 5 results. However, Franklin and Lindberg (2015) suggest that smoking is not an established 6 risk factor of OSA and the role of alcohol is unclear. Thus, with the exception of BMI, there 7 was little evidence to suggest that confounding variables were an important consideration in 8 measuring the association of OSA and PA in the participant group. 9 The present study is not without limitations. First, there is the obvious issue of 10 reverse causation, making it difficult to infer whether low PA predicts OSA or vice versa, 11 similarly with the link between BMI and PA, however, given that PA engagement is self-12 selected, only PA intervention randomized trials could start to shed light on the potential 13 causal direction of effects, which is a future direction for this research. Second, the sample 14 size is relatively small, making it possible that the association between PA and OSA severity 15 was attenuated by BMI due to low power and reduced degrees of freedom. Indeed, the lack 16 of evidence of mediation suggests this, and in fact makes it probable that both PA and BMI 17 are independent contributors to OSA severity, although this would need to be confirmed in a 18 larger study. Third, the majority of the participants demonstrated moderate levels of PA and 19 mostly none or mild levels of OSA severity. Thus, it is possible that there were not sufficient 20 participants with moderate to severe OSA in the group to provide a clearer outcome of the 21 association between PA and OSA. However, a strength of the study is that all participants 22 (healthcare, e.g. nurses, teaching, e.g. PE, and craft/labour) were in manual occupations 23 (classed as manual or involving a substantial physical component e.g. PE teaching) and all 24 were older than 50 years old, which is a neglected group; given the growing age of the 25 workforce, older workers' health is becoming of paramount importance (Billett et al., 2011). 1 Fourth, self-report measures of PA are limited due to risk of over-reporting and recall bias. 2 However, the instrument used to assess PA in the present study is validated to be one of the 3 best methods of self-report PA measures and provides detailed information on three areas of 4 PA. Future research may benefit from using objective measures of PA. 5 Evidence in respect of the association between undiagnosed OSA severity and PA is 6 limited. Only a few studies have used participants aged over 50 years old and of those, OSA 7 has already been diagnosed or the sample size is very small (Redolfi,

Conclusion 17
It is clear from the results that undiagnosed OSA exists in the group of older manual 18 workers examined in the present study. Indeed, low levels of PA were found to be associated 19 with increased severity of OSA. It should also be noted that BMI in the present study was a 20 stronger indicator of OSA severity than PA and therefore future intervention guidance may 21 prove confusing. However, given the negative health outcomes associated with OSA and the 22 benefits that regular PA offers to physical health and mental wellbeing, as well as a method 23 of reducing weight and thereby BMI, it is worthwhile considering low PA as a risk factor for 24 OSA in older workers. Further, the outcome of the present study warrants the consideration 25 of PA as a cost effective management treatment of OSA through aiding weight loss and 1 promoting healthy cardiovascular function in this age group. ApneaLink for the screening of sleep apnea: a novel and simple single-channel 10