Journal of Physical Activity and Health, 2015, 12, 1567 -1575 http://dx.doi.org/10.1123/jpah.2014-0251 © 2015 Human Kinetics, Inc.
ORIGINAL RESEARCH
Sleep Disorders, Physical Activity, and Sedentary Behavior Among U.S. Adults: National Health and Nutrition Examination Survey James L. Farnsworth, Youngdeok Kim, and Minsoo Kang Background: Disruptive sleeping patterns have been linked to serious medical conditions. Regular physical activity (PA) has a positive impact on health; however, few research have investigated the relationships between PA, body mass index (BMI), sedentary behaviors (SB), and sleep disorders (SD). Methods: Data from the 2005–2006 NHANES were analyzed for this study. Participants (N = 2989; mean age = 50.44 years) were grouped based upon responses to SD questions. Accelerometers were used to measure the average time spent in moderate or vigorous physical activity (MVPA) and SB. Multinomial logistic regression analyses were used to examine the associations between PA, SB, and SD after controlling for covariates and to explore potential moderation effects among common risk factors and the main study variables. Results: Among middle-aged adults, PA was significantly associated with SD [Wald χ2 (8) = 22.21; P < .001]. Furthermore, among adults in the highest tertile of SB, PA was significantly associated with SD [Wald χ2 (8) = 32.29; P < .001]. Conclusions: These results indicate that middle-aged adults who are less active may have increased likelihoods of SD. It is important for health care professionals to continue developing methods for increasing PA to decrease the risk of SD. Keywords: MVPA, sleep behavior
The American Sleep Association estimates that at least 40 million people in the United States suffer from chronic or longterm sleeping disorders (SD) and an additional 20 million people have experienced sporadic sleep and sleep-related problems.1 SD, which consists of more than 70 different symptoms across 3 general categories [ie, lack of sleep (insomnia), disturbed sleep (obstructive sleep apnea), and excessive sleep (narcolepsy)],2 are medical disorders that affect the sleeping patterns of individuals and have been identified as a significant risk factor associated with serious medical conditions including cardiovascular morbidities, daytime sleepiness, and impaired cognitive function.3–6 Because of its importance to improving individual and public health, it is crucial to understand SD-related risk factors that could effectively be modified to ameliorate the health consequences of SD. Of the several key risk factors for SD including but not limited to age, gender, race, craniofacial anatomy, familial and genetic predisposition,7 numerous epidemiological studies have indicated adiposity levels as one of the most prominent risk factors for the development of SD.8 In a population-based prospective study it was identified that a 1 standard deviation increase in body mass index (BMI) was associated with a 4-fold increase in the prevalence of SD.9 Although there are conflicting views in the nature in which adiposity level influences SD,10,11 intervention studies aimed at reducing the symptoms of SD have found success using surgical weight loss procedures.12,13 A number of studies have suggested that weight loss by any means (ie, surgery, and calorie restriction) can decrease the severity of SD in patients; however, the generalizability of these studies are limited due to small sample sizes and lack of appropriate control groups.14,15 Another potential risk factor that has been underestimated in relation to SD is physical activity (PA). PA is a well-known Farnsworth ([email protected]) and Kang are with the Dept of Health and Human Performance, Middle Tennessee State University, Murfreesboro, TN. Kim is with the Dept of Kinesiology and Sport Management, Texas Tech University, Lubbock, TX.
modifiable lifestyle factor associated with various aspects of individual health including but not limited to noncommunicable diseases such as cardiovascular diseases or cancers, and psychosocial wellbeing,16 and has been examined as a potential moderator in relation to SD among clinical populations. In a nationwide study of US adults with prediabetes, it was identified that adults with insomnia were less physically active when compared with adults without insomnia.17 Other population-based studies have found similar results among adults with obstructive sleep apnea.18,19 In a study examining midlife women with vasomotor symptoms, household activities such as laundry, vacuuming, and doing the dishes were associated with “good” sleep quality,20 suggesting that PA may play a role in moderating the effects of other SD-related risk factors. In addition, rising levels of sedentary behavior (SB) over the past 2 decades have been linked to a number of negative health outcomes such as an increased risk of type II diabetes, increased risk of metabolic syndrome, and increased waist circumference,21 and there has been some evidence suggesting that time spent in SB may be associated with SD. A study of US high-school students found that students who spent 2 or more hours on the computer for nonschool work related activities were less likely to get sufficient sleep each night.22 These data suggest that examination of SB in relation to SD may also help to improve our understanding of SDrelated risk factors. Despite the evidence of potential effects of PA and SB, in addition to BMI, on development of SD, there has been little research investigating the associations of these variables simultaneously among a general population of adults. Furthermore, considering that PA behaviors have been shown to prolong the natural decline in health associated with the aging process in adults, and the effects of PA on the aging process are more pronounced the earlier the individual begins to engage in PA,23 a more thorough investigation accounting for possible moderating effect of age on the relationships among these variables is warranted. This study therefore aimed to fill these gaps by examining (1) the associations of PA, SB, and BMI with SD while controlling for other common risk factors (eg, demographic characteristics, smoking status, etc.); and (2) the 1567
1568 Farnsworth, Kim, and Kang
moderating effect of age on the relationships among these variables among a national representative sample of US adults.
Methods
Downloaded by Washington Univ In St Louis on 09/16/16, Volume 12, Article Number 12
Study Population This study was conducted using data from the 2005–2006 National Health and Nutrition Examination Survey (NHANES). The NHANES is a cross-sectional survey of the civilian noninstitutionalized United States population that is released biannually through the National Center for Health Statistics (NCHS). All participants gave written informed consent. The protocol was approved by the NCHS Research Ethics Review Board. The 2005–2006 NHANES contains data for 10,348 individuals of all ages collected from January 2005 through December 2006 and is the first NHANES data including a complete questionnaire about physician-diagnosed sleep disorders. The sleep questionnaire and data are available for public use and can be downloaded from the NHANES website (http://wwwn.cdc.gov/nchs/nhanes/search/ nhanes05_06.aspx). A total of 6139 adult participants (age ≥ 20) provided responses to questions relating to sleeping habits and behaviors. Participants who did not provide responses to the main outcome variables (ie, diagnosis of SD and presence of sleeping problem) were excluded from analysis. Following exclusion the remaining sample included 4474 participants. To further clean the data, participants who did not provide responses to control variables (smoking status, marital status, etc.) were removed. This resulting in a final sample of 2989 adults (mean age = 50.44 ± 18.23 years).
Assessment of Sleeping Disorder The approach used to evaluate SD was modeled after a similar study examining relationships between sleep disordered breathing and depression.24 Participants’ sleeping habits and behaviors were assessed through the Computer-Assisted Personal Interviewing (CAPI) system. During the interview, participants were asked 2 items, which were used to identify participants with abnormal sleeping behaviors: “Have you ever told a doctor or other health professional that you have trouble sleeping?” and “Have you ever been told by a doctor or other health professional that you have a sleep disorder?” Possible responses included “yes,” “no,” “refused,” or “don’t know.” Participants were grouped, into 1 of 3 groups, based upon responses to SD questions (no sleeping problems, reported nondiagnosed sleeping problem, and reported diagnosis with sleeping disorder). Participants who responded with “don’t know” or “refused” were recoded as a “no” response.
Objectively Measured Physical Activity and Sedentary Behavior The Actigraph (model 7164; Actigraph, LLC. Pensacola, FL) accelerometer data were used to obtain the average duration of minutes spent in moderate and vigorous physical activity (MVPA) and SB. Participants capable of ambulatory movements were asked at the end of their examination to wear the accelerometer for 7 consecutive days.25 The device was worn over the right hip and secured with an elastic band. The accelerometers were set to begin collection data in 1-minute intervals beginning at midnight after the examination. Since the devices are not waterproof, no data collection occurred during water-based activities (eg, showering, swimming). In addition, participants were instructed to remove the monitor at bed time (no data were provided regarding on/off time for the devices).
Before analysis, PA data were compiled and analyzed using SAS macro programs available from the National Cancer Institute (NCI)26 to exclude participants with impossible data values, and consolidate the data. Nonwear time was defined as consecutive bouts of nonactivity (ie, counts of 0) for 60 minutes, with allowances for up to 2 consecutive minutes with activity counts less than 100. Valid days were defined as wearing of the device for 10 or more hours27 and participants with less than 4 valid days were excluded from the analysis. Time spent in MVPA and SB during wear time were obtained across valid days using the thresholds of ≥2020 counts per minute (cpm) and 423.50 min/day = highest tertile), respectively.
Body Mass Index Weight and standing height were collected from participants by health trained technicians during the examination. Weight was measured in pounds (later converted to kilograms (kg) using an automated system) using a Toledo digital scale. Participants were instructed to stand still in the center of the scale facing the recorder, with hands at their side, and looking straight ahead. Participants exceeding 440 pounds, were assessed using a similar protocol with 2 Seca digital scales. Standing height was measured in centimeters (cm) using a fixed stadiometer with vertical backboard and a moveable headboard. Participants were given instructions to move or remove hair ornaments, jewelry, buns, and braids from the top of their head. To obtain more consistent and reproducible results each participant was asked to take a deep breath and stand as tall as possible while the height was recorded. These anthropometric data were then used to compute BMI with the following formula: BMI = [kg / (cm / 100)2] The Centers for Disease Control and Prevention has specified the following categories for BMI measurements. Participants having a BMI < 18.5 = Underweight; BMI 18.5 to 24.9 = Normal; BMI 25.0 to 29.9 = Overweight; and BMI ≥ 30 = Obese.32 Due to the relative importance of obesity in regards to SD,8 and the small sample size for the underweight category (n = 84); the underweight, normal, and overweight categories were combined to create a 2-category variable for obesity (obese and nonobese).
Common Risk Factors Information was collected during the home interview component on covariates which included gender (male/female), age category (young adults 18–40, middle aged 41–65 and, older adults 65+), selfreported race/ethnicity (non-Hispanic white, non-Hispanic black, Mexican American, other/multiracial), marital status (single, married/living with partner), smoking status, and alcohol consumption. Smoking status and alcohol consumption were included because they have been reported to negatively affect sleeping quality.33,34 Participants were asked, “Do you now smoke cigarettes?” with possible responses including “every day,” “some days,” “not at all,” or “missing.” These responses were merged to create the categorical
JPAH Vol. 12, No. 12, 2015
Sleep Disorder, Physical Activity, and Sedentary Behavior 1569
variable current smoking status (“Yes” and “No”). Alcohol consumption was determined by the self-reported average number of alcoholic beverages consumed per week.
Downloaded by Washington Univ In St Louis on 09/16/16, Volume 12, Article Number 12
Statistical Analyses All statistical analyses were conducted using SAS v9.3 (SAS Institute, Inc., Cary, NC), to account for complex sampling design of the NHANES. The sample weights were recalculated based on age, gender, and racial/ethnic groups from the original NHANES mobile center weights to account for the selection bias by the inclusion criteria of this study. The crude associations between all study variables and SD were examined using Rao-Scott χ2 tests. Secondly, a multinomial logistic regression analysis was used to examine the associations of PA, SB, and BMI with SD, while controlling for other common risk factors (ie, age category, race, marital status, current smoking status, and alcohol consumption). A multinomial logit link function was applied to estimate the adjusted odds ratios (OR)
and associated 95% confidence intervals (CI) for the likelihoods of sleeping problems and diagnosed sleeping disorder compared with no sleeping problems. Lastly, cross-product interaction terms of age group with PA, SB, and BMI were included in the model to examine the moderating effects of age on the relationships of the main study variables (PA, SB, and BMI) with SD. We also examined the models with additional cross-product interaction terms between main study variables for exploratory purposes to identify possible moderating effects on the relationship with SD. Simple effects tests were conducted for significant interactions, and Bonferroni adjusted alpha level was used to estimate the associated CIs with respective OR.
Results The associations between SD and outcome variables are listed in Table 1. Among US adults an estimated 6.53% (95% CI: 5.57%– 7.49%) reported a diagnosed sleeping disorder, 17.34% (95% CI:
Table 1 Demographics for Percent of Population Characteristicsa (n = 2989) Variable Gender Male Female Age category Young adult (65) Race Non-Hispanic White Non-Hispanic Black Mexican American Other BMI Nonobese Obese Marital status Married/cohabiting Single Smoking status Nonsmoker Smoker MVPA 1st tertile (lowest) 2nd tertile 3rd tertile (highest) Sedentary behavior 1st tertile (lowest) 2nd tertile 3rd tertile (highest)
No sleeping problem
Nondiagnosed sleeping problem
Diagnosis with sleeping disorder
Total
79.15 ± 1.65 73.34 ± 1.25
12.59 ± 1.37 21.75 ± 1.17
8.26 ± 0.98 4.92 ± 0.58
48.10 ± 0.74 51.90 ± 0.74
83.92 ± 1.45 70.70 ± 1.51 73.07 ± 1.42
13.30 ± 1.47 20.08 ± 1.24 19.16 ± 1.25
2.78 ± 0.66 9.22 ± 0.88 7.77 ± 0.88
38.00 ± 1.64 44.87 ± 1.54 17.13 ± 1.51
73.50 ± 0.78 79.41 ± 1.61 86.45 ± 2.14 84.08 ± 3.12
19.60 ± 0.69 14.26 ± 1.06 9.07 ± 1.59 10.41 ± 2.08
6.90 ± 0.60 6.33 ± 1.29 4.48 ± 1.21 5.51 ± 1.9
71.86 ± 2.84 11.48 ± 1.98 7.97 ± 1.00 8.69 ± 1.13
78.98 ± 1.10 70.45 ± 1.46
17.36 ± 0.99 17.32 ± 1.54
3.69 ± 0.52 12.23 ± 0.83
66.82 ± 1.57 33.18 ± 1.57
77.62 ± 0.98 73.00 ± 1.60
15.59 ± 0.95 21.05 ± 1.34
6.79 ± 0.49 5.96 ± 0.79
67.88 ± 1.88 32.12 ± 1.88
76.80 ± 0.90 73.57 ± 2.36
17.02 ± 0.79 18.61 ± 1.58
6.19 ± 0.43 7.82 ± 1.41
79.42 ± 1.30 20.58 ± 1.30
66.25 ± 1.77 77.29 ± 1.05 81.53 ± 1.37
22.73 ± 1.41 16.90 ± 1.04 14.25 ± 1.40
11.02 ± 0.98 5.82 ± 0.61 4.22 ± 0.64
25.83 ± 0.95 34.20 ± 1.55 39.97 ± 1.65
78.88 ± 1.20 75.02 ± 1.36
16.10 ± 1.29 18.61 ± 1.25
5.02 ± 0.77 6.37 ± 0.82
33.84 ± 1.01 34.92 ± 1.11
74.39 ± 1.73
17.28 ± 1.58
8.33 ± 1.15
31.23 ± 1.33
a Values
Rao-Scott χ2 20.82
P-value < 0.01*
49.42
< 0.01*
35.83
< 0.01*
46.66
< 0.01*
10.39
< 0.01*
2.26
0.32
59.94
< 0.01*
7.51
0.11
represent the percentage of population characteristic. Percentages are rounded to the 100th place. * Indicates statistical significance P < .05. Note. Total population percentage for no sleeping problem, nondiagnosed sleeping problem, and diagnosis with a sleeping disorder were 76.13 (SE = 0.58), 17.34 (SE = 0.51), and 6.53 (SE = 0.49), respectively. Abbreviations: CI, confidence interval; BMI, body mass index; MVPA, moderate and vigorous physical activity. JPAH Vol. 12, No. 12, 2015
1570 Farnsworth, Kim, and Kang
16.34%–18.34%) reported sleep problems, and 76.13% (95% CI: 74.99%–77.27%) reported no sleep problems. Among adults in the highest tertile of physical activity an estimated 4.22% (95% CI: 2.96%–5.49%) reported diagnosed sleeping disorders compared with those in the lowest tertile where 11.02% (95% CI: 9.09%– 12.95%) reported diagnosed sleeping disorders. Among adults in the highest tertile of SB an estimated 8.33% (95% CI: 6.08%–10.57%) reported a diagnosed sleeping disorder compared with those in the lowest tertile where 5.02% (95% CI: 3.51%–6.53%) reported a diagnosed sleeping disorder. Adjusted odds ratios for all variables are provided in Table 2. SD were significantly associated with MVPA when controlling for covariates (Wald χ2 (4) = 38.6; P < .001). Using no sleep problems
as a reference category, participants in the highest tertile of PA were less likely to report sleeping problems (OR = 0.55; 95% CI: 0.36–0.86) and diagnosed sleeping disorders (OR = 0.39; 95% CI: 0.25–0.62), respectively, compared with those in the lowest tertile of PA. Participants in the middle tertile of PA were less likely to report sleeping problems (OR = 0.64; 95% CI: 0.47–0.88) and diagnosed sleeping disorders (OR = 0.46; 95% CI: 0.35–0.60) compared with those in the lowest tertile of PA. Among genders, females were more likely to report sleeping problems (OR = 1.63; 95% CI: 1.12–2.36), but less likely to report diagnosed sleeping disorders (OR = 0.49; 95% CI: 0.32–0.76), when compared with males. Follow-up analyses examined the data for all potential interaction effects between common risk factors (eg, age, gender, etc.) and
Downloaded by Washington Univ In St Louis on 09/16/16, Volume 12, Article Number 12
Table 2 Adjusted Odds Ratios for Sleeping Problems and Sleeping Disorders Variable Gender Male Female Age category Young adults (65) Race Non-Hispanic White Non-Hispanic Black Mexican American Other BMI Nonobese Obese Marital status Married/living with partner Single Smoking status Nonsmoker Smoker Average number of alcoholic beverages consumed per week MVPA 1st tertile (low MVPA) 2nd tertile 3rd tertile (high MVPA) Linear trend P-value SB 1st tertile (low SB) 2nd tertile 3rd tertile (high SB) Linear trend P-value
Nondiagnosed sleeping problem
Diagnosis with sleeping disorder
[Reference] 1.63 (1.12–2.36)*
[Reference] 0.49 (0.32–0.76)*
[Reference] 1.62 (1.16–2.26)* 1.05 (0.71–1.56)
[Reference] 3.05 (1.83–5.10)* 2.08 (1.13–3.81)*
[Reference] 0.56 (0.47–0.68)* 0.42 (0.28–0.64)* 0.44 (0.27–0.71)*
[Reference] 0.75 (0.44–1.30) 0.76 (0.44–1.33) 0.93 (0.51–1.71)
[Reference] 0.99 (0.70–1.40)
[Reference] 3.30 (2.24–4.85)*
[Reference] 1.52 (1.15–2.01)*
[Reference] 1.06 (0.76–1.48)
[Reference] 1.11 (0.76–1.62)
[Reference] 1.42 (0.84–2.38)
0.97 (0.91–1.04)
0.95 (0.86–1.04)
[Reference] 0.64 (0.47–0.88)* 0.55 (0.36–0.86)* 0.01*
[Reference] 0.46 (0.35–0.60)* 0.39 (0.25–0.62)* < 0.001*
[Reference] 1.03 (0.71–1.50) 0.83 (0.58–1.18)
[Reference] 1.12 (0.70–1.81) 1.04 (0.67–1.62)
0.30
0.88
Note. The above Adjusted Odds Ratios were calculated using multinomial logistic regression with “No Sleeping Problems” selected as the reference group. Abbreviations: CI, confidence interval; BMI, body mass index, MVPA, moderate and vigorous physical activity, SB, sedentary behavior. * Indicates statistical significance P < .05. JPAH Vol. 12, No. 12, 2015
Sleep Disorder, Physical Activity, and Sedentary Behavior 1571
Downloaded by Washington Univ In St Louis on 09/16/16, Volume 12, Article Number 12
main study variables (PA, SB, and BMI). A significant interaction effect was identified between age and SB (Wald χ2 (8) = 39.43; P < .001). No other interaction effects were found. Multinomial logistic regression analyses was performed for each domains of age and SB for simple effect tests (see Table 3 and 4 for simple effect tests by age and SB groups, respectively). Examination across domains of age revealed that Middle-aged adults in the highest tertile of PA were less likely to report sleeping problems (OR = 0.51; 98.33% CI: 0.28–0.93) and diagnosed sleeping disorders (OR = 0.33; 98.33% CI: 0.16–0.70) when compared with those in the lowest tertile of PA. Middle-aged adults in the middle tertile of PA were less likely to report diagnosed sleeping disorders (OR = 0.44; 98.33% CI: 0.25–0.75) when compared with those in the lowest tertile of PA. Examination across domains of SB indicate that for adults in the highest tertile of SB, PA was significantly associated with diagnosed sleeping disorders (Wald χ2 (4) = 18.53; P < .001). In addition, With the exception of those in the lowest tertile of SB (OR = 1.84; 98.33% CI: 0.71–4.79) adults who are obese are more likely to report diagnosed sleeping disorders when compared with nonobese adults.
Discussion In this nationally representative sample of adults we examined the relationship between objectively measured PA, SB, BMI, and SD when controlling for common risk factors. The results from this study suggest that for adults PA is an important risk factor for SD. This finding is in contrast to those found by Loprinzi and Cardinal35 who suggest that there were no associations between activity status and the likelihood of reporting a SD. The difference in findings is most likely the result of the different statistical models used by each study. While both models controlled for common risk factors (eg, age, race, smoking status, gender) our model included SB time as an additional variable. Inclusion of SB time into the model introduced an interaction effect between SB time and age. When examining the main study variables across age groups (ie, young adults, middle-aged adults, and older adults), we identified PA as significant risk factor for SD. Examination of risk factors for diagnosed sleeping disorders across age groups revealed that for middle-aged adults, low levels of PA were associated with an increased likelihood of diagnosed sleeping disorders. In the literature it has been reported that agerelated changes in the regulation of growth hormones as an individual transitions from early adulthood to later life stages have a strong influence on sleeping patterns and behaviors.36 Therefore, it is important when assessing sleep behavior that risk factors are examined independently for various life stages. It is also worth noting that, young adults typically had higher levels of PA and lower levels of SB (Figure 1) when compared with middle-aged and older adults. These trends are consistent with those found by Gordon-Larsen et al, which suggest a reciprocal relationship between PA, SB, and aging.37 As we age our PA levels decrease, and our SB levels increase. One particularly interesting finding in our study indicated among adults with low levels of SB, obesity and PA were not significantly associated with SD. This discovery suggest that the elevated risk of SD in obese adults may be moderated through decreasing SB which could have significant implications for future intervention strategies. In a cross-sectional examination of Australian adults it was identified that increased breaks in sedentary time were beneficially associated with diminished waist circumference, and fasting plasma glucose levels.38 In addition, previous research has suggested
that weight loss (surgical or otherwise) has been an effective treatment for SD.12–15 The decreased levels of SB would both directly and indirectly affect SD through weight loss and moderation of the effects of obesity on SD, respectively. The overall prevalence of SD from US adults were consistent with those that are reported in the literature.9,39 In men, 3.97% reported diagnosed sleeping disorders, while only 2.55% of women reported diagnosed sleeping disorders. Among genders, there were significant differences in the prevalence of sleeping problems and diagnosed sleeping disorders. Females were more likely to report sleeping problems (OR = 1.63; 95% CI: 1.12–2.36) but, nearly half as likely to report diagnosed sleeping disorders (OR = 0.49; 95% CI: 0.32–0.76). The results found in this study that woman are more likely than men to report sleeping problems may simply be an indicator that woman are more likely to report health issues compared with men as opposed to a true difference in sleep quality as the data suggest. This notion is supported by the results found by Benyamini et al,40 who reported that self-assessed health was a weaker predictor of mortality in woman than men. In addition, a study examining the sleep characteristics between genders found that females were more likely than males to report disruptive sleeping behaviors.41 When reporting negative health conditions, men tend to only report conditions which have significant negative health consequences. Women on the other hand, tend to report on a larger range of health problems (ie, sleeping problems). Similar to the findings of Bansil et al,42 there were no associations between race and diagnosed sleeping disorders. However, the likelihood of reporting sleeping problems was lower for nonwhite ethnic groups. These findings are in contrast with those found in other studies.43,44 Examination of adults using in-home polysomnography reported shorter sleep duration and lower sleep efficiency among Blacks.43 Furthermore, in the Sleep Heart Health Study frequent snoring, a predictor of sleeping disorders,45 was more common among Hispanic and Black women. The conflicting evidence may likely be the result of the methods used to collect the data. The data from this nationally representative sample were collected through self-report. Moreover, the presence of snoring does not directly indicate sleeping problems. A cross-sectional study of adults in a sleep laboratory identified waist circumference as one of the most prominent risk factors for development of SD with approximately 50% of obese adults being diagnosed with a SD.46 Reports from the National Sleep Foundation suggest that among older adults, 77% report some form of sleeping problems.47 The results of this study were consistent with those found in other studies indicating that approximately 63% of adults who report diagnosed sleeping disorders are obese. Obese adults demonstrate a much higher risk for development of SD when compared with nonobese individuals. This is particularly problematic in the United States where more than one-third of all adults are considered obese.48 Intervention studies involving lifestyle interventions have demonstrated that increasing PA levels in an obese population can significantly improve health outcomes. In a longitudinal study examining 313 obese adults, the use of a surgical weight loss procedure decreased the prevalence of obstructive sleep apnea from 33% to 2%.15 Promoting sufficient levels of MVPA among overweight and obese adults may help to reduce the prevalence rates of diagnosed sleeping disorders. It is interesting to note that while BMI was a significant predictor of diagnosed sleeping disorders, BMI had minimal associations with sleeping problems. This is in contrast to the findings from Turco et al,49 who reported that obese
JPAH Vol. 12, No. 12, 2015
1572
JPAH Vol. 12, No. 12, 2015
1.42 (0.70–2.88)
2.63 (0.92–7.49)
[Reference]
2.47 (0.82–7.46)
1.03 (0.25–4.23)
[Reference]
0.50 (0.12–2.07)
0.45 (0.13–1.56)
[Reference]
Diagnosed with sleeping disorder
0.85 (0.53–1.35)
[Reference]
0.94 (0.52–1.71)
1.34 (0.77–2.34)
[Reference]
0.51 (0.28–0.93)*
0.59 (0.32–1.09)
[Reference]
Nondiagnosed sleeping problem
3.48 (1.66–7.32)*
[Reference]
1.05 (0.48–2.28)
1.24 (0.55–2.79)
[Reference]
0.33 (0.16–0.70)*
0.44 (0.25–0.75)*
[Reference]
Diagnosed with sleeping disorder
Middle-aged adults (40–65)
0.83 (0.40–1.70)
[Reference]
0.67 (0.29–1.50)
0.79 (0.47–1.32)
[Reference]
0.97 (0.37–2.60)
0.74 (0.39–1.43)
[Reference]
Nondiagnosed sleeping problem
3.28 (1.78–6.06)*
[Reference]
0.34 (0.10–1.14)
0.56 (0.12–2.46)
[Reference]
0.68 (0.28–1.67)
0.40 (0.17–0.97)*
[Reference]
Diagnosed with sleeping disorder
Older adults (>65)
0.92 (0.49–1.73)
[Reference]
1.84 (0.71–4.79)
[Reference]
1.11 (0.59–2.10)
[Reference]
0.36 (0.19–0.70)*
0.66 (0.36–1.23)
[Reference]
Nondiagnosed sleeping problem
6.12 (2.40–15.60)*
[Reference]
0.21 (0.05–0.83)*
0.62 (0.27–1.40)
[Reference]
Diagnosed with sleeping disorder
Middle tertile
0.99 (0.52–1.90)
[Reference]
0.75 (0.29–1.92)
0.55 (0.28–1.07)
[Reference]
Nondiagnosed sleeping problem
2.94 (1.51–5.74)*
[Reference]
0.33 (0.13–0.79)*
0.26 (0.10–0.65)*
[Reference]
Diagnosed with sleeping disorder
Highest tertile
Note. The above Adjusted Odds Ratios were calculated using multinomial logistic regression with “No Sleeping Problems” selected as the reference group across domains of Sedentary Behavior. All covariates (excluding MVPA and BMI) included within the model have been excluded from the table. Covariates include: age, race, alcohol consumption, gender, smoking status, and marital status. Abbreviations: BMI, body mass index; MVPA, moderate and vigorous physical activity; SB, sedentary behavior. * Indicates statistical significance P < .0167 (Bonferroni adjusted alpha for multiple comparisons 0.05/3).
Obese
Nonobese
BMI
1.12 (0.31–4.00)
0.66 (0.32–1.39) 0.71 (0.38–1.33)
2nd Tertile
3rd Tertile (high)
1.48 (0.59–3.74)
[Reference]
Diagnosed with sleeping disorder
[Reference]
1st Tertile (low)
MVPA
Nondiagnosed sleeping problem
Lowest tertile
Domain—sedentary behavior tertile
Table 4 Adjusted Odds Ratios and 95% Confidence Intervals for the Effects of BMI and MVPA Sleeping Disorders by Tertiles of Sedentary Behavior
Note. The above Adjusted Odds Ratios were calculated using multinomial logistic regression with “No Sleeping Problems” selected as the reference group across domains of Age group. All covariates (excluding MVPA) included within the model have been excluded from the table. Covariates include: race, alcohol consumption, gender, smoking status, and marital status. Abbreviations: BMI, body mass index; MVPA, moderate and vigorous physical activity; SB, sedentary behavior. * Indicates statistical significance P < .0167 (Bonferroni adjusted alpha for multiple comparisons 0.05/3).
Obese
Nonobese
[Reference]
0.80 (0.36–1.80)
3rd Tertile (high)
BMI
0.75 (0.34–1.63)
2nd Tertile
1st Tertile (low)
[Reference]
0.49 (0.19–1.26)
SB
0.56 (0.25–1.27)
3rd Tertile (high)
[Reference]
2nd Tertile
1st Tertile (low)
MVPA
Nondiagnosed sleeping problem
Young adults (
ORIGINAL RESEARCH
Sleep Disorders, Physical Activity, and Sedentary Behavior Among U.S. Adults: National Health and Nutrition Examination Survey James L. Farnsworth, Youngdeok Kim, and Minsoo Kang Background: Disruptive sleeping patterns have been linked to serious medical conditions. Regular physical activity (PA) has a positive impact on health; however, few research have investigated the relationships between PA, body mass index (BMI), sedentary behaviors (SB), and sleep disorders (SD). Methods: Data from the 2005–2006 NHANES were analyzed for this study. Participants (N = 2989; mean age = 50.44 years) were grouped based upon responses to SD questions. Accelerometers were used to measure the average time spent in moderate or vigorous physical activity (MVPA) and SB. Multinomial logistic regression analyses were used to examine the associations between PA, SB, and SD after controlling for covariates and to explore potential moderation effects among common risk factors and the main study variables. Results: Among middle-aged adults, PA was significantly associated with SD [Wald χ2 (8) = 22.21; P < .001]. Furthermore, among adults in the highest tertile of SB, PA was significantly associated with SD [Wald χ2 (8) = 32.29; P < .001]. Conclusions: These results indicate that middle-aged adults who are less active may have increased likelihoods of SD. It is important for health care professionals to continue developing methods for increasing PA to decrease the risk of SD. Keywords: MVPA, sleep behavior
The American Sleep Association estimates that at least 40 million people in the United States suffer from chronic or longterm sleeping disorders (SD) and an additional 20 million people have experienced sporadic sleep and sleep-related problems.1 SD, which consists of more than 70 different symptoms across 3 general categories [ie, lack of sleep (insomnia), disturbed sleep (obstructive sleep apnea), and excessive sleep (narcolepsy)],2 are medical disorders that affect the sleeping patterns of individuals and have been identified as a significant risk factor associated with serious medical conditions including cardiovascular morbidities, daytime sleepiness, and impaired cognitive function.3–6 Because of its importance to improving individual and public health, it is crucial to understand SD-related risk factors that could effectively be modified to ameliorate the health consequences of SD. Of the several key risk factors for SD including but not limited to age, gender, race, craniofacial anatomy, familial and genetic predisposition,7 numerous epidemiological studies have indicated adiposity levels as one of the most prominent risk factors for the development of SD.8 In a population-based prospective study it was identified that a 1 standard deviation increase in body mass index (BMI) was associated with a 4-fold increase in the prevalence of SD.9 Although there are conflicting views in the nature in which adiposity level influences SD,10,11 intervention studies aimed at reducing the symptoms of SD have found success using surgical weight loss procedures.12,13 A number of studies have suggested that weight loss by any means (ie, surgery, and calorie restriction) can decrease the severity of SD in patients; however, the generalizability of these studies are limited due to small sample sizes and lack of appropriate control groups.14,15 Another potential risk factor that has been underestimated in relation to SD is physical activity (PA). PA is a well-known Farnsworth ([email protected]) and Kang are with the Dept of Health and Human Performance, Middle Tennessee State University, Murfreesboro, TN. Kim is with the Dept of Kinesiology and Sport Management, Texas Tech University, Lubbock, TX.
modifiable lifestyle factor associated with various aspects of individual health including but not limited to noncommunicable diseases such as cardiovascular diseases or cancers, and psychosocial wellbeing,16 and has been examined as a potential moderator in relation to SD among clinical populations. In a nationwide study of US adults with prediabetes, it was identified that adults with insomnia were less physically active when compared with adults without insomnia.17 Other population-based studies have found similar results among adults with obstructive sleep apnea.18,19 In a study examining midlife women with vasomotor symptoms, household activities such as laundry, vacuuming, and doing the dishes were associated with “good” sleep quality,20 suggesting that PA may play a role in moderating the effects of other SD-related risk factors. In addition, rising levels of sedentary behavior (SB) over the past 2 decades have been linked to a number of negative health outcomes such as an increased risk of type II diabetes, increased risk of metabolic syndrome, and increased waist circumference,21 and there has been some evidence suggesting that time spent in SB may be associated with SD. A study of US high-school students found that students who spent 2 or more hours on the computer for nonschool work related activities were less likely to get sufficient sleep each night.22 These data suggest that examination of SB in relation to SD may also help to improve our understanding of SDrelated risk factors. Despite the evidence of potential effects of PA and SB, in addition to BMI, on development of SD, there has been little research investigating the associations of these variables simultaneously among a general population of adults. Furthermore, considering that PA behaviors have been shown to prolong the natural decline in health associated with the aging process in adults, and the effects of PA on the aging process are more pronounced the earlier the individual begins to engage in PA,23 a more thorough investigation accounting for possible moderating effect of age on the relationships among these variables is warranted. This study therefore aimed to fill these gaps by examining (1) the associations of PA, SB, and BMI with SD while controlling for other common risk factors (eg, demographic characteristics, smoking status, etc.); and (2) the 1567
1568 Farnsworth, Kim, and Kang
moderating effect of age on the relationships among these variables among a national representative sample of US adults.
Methods
Downloaded by Washington Univ In St Louis on 09/16/16, Volume 12, Article Number 12
Study Population This study was conducted using data from the 2005–2006 National Health and Nutrition Examination Survey (NHANES). The NHANES is a cross-sectional survey of the civilian noninstitutionalized United States population that is released biannually through the National Center for Health Statistics (NCHS). All participants gave written informed consent. The protocol was approved by the NCHS Research Ethics Review Board. The 2005–2006 NHANES contains data for 10,348 individuals of all ages collected from January 2005 through December 2006 and is the first NHANES data including a complete questionnaire about physician-diagnosed sleep disorders. The sleep questionnaire and data are available for public use and can be downloaded from the NHANES website (http://wwwn.cdc.gov/nchs/nhanes/search/ nhanes05_06.aspx). A total of 6139 adult participants (age ≥ 20) provided responses to questions relating to sleeping habits and behaviors. Participants who did not provide responses to the main outcome variables (ie, diagnosis of SD and presence of sleeping problem) were excluded from analysis. Following exclusion the remaining sample included 4474 participants. To further clean the data, participants who did not provide responses to control variables (smoking status, marital status, etc.) were removed. This resulting in a final sample of 2989 adults (mean age = 50.44 ± 18.23 years).
Assessment of Sleeping Disorder The approach used to evaluate SD was modeled after a similar study examining relationships between sleep disordered breathing and depression.24 Participants’ sleeping habits and behaviors were assessed through the Computer-Assisted Personal Interviewing (CAPI) system. During the interview, participants were asked 2 items, which were used to identify participants with abnormal sleeping behaviors: “Have you ever told a doctor or other health professional that you have trouble sleeping?” and “Have you ever been told by a doctor or other health professional that you have a sleep disorder?” Possible responses included “yes,” “no,” “refused,” or “don’t know.” Participants were grouped, into 1 of 3 groups, based upon responses to SD questions (no sleeping problems, reported nondiagnosed sleeping problem, and reported diagnosis with sleeping disorder). Participants who responded with “don’t know” or “refused” were recoded as a “no” response.
Objectively Measured Physical Activity and Sedentary Behavior The Actigraph (model 7164; Actigraph, LLC. Pensacola, FL) accelerometer data were used to obtain the average duration of minutes spent in moderate and vigorous physical activity (MVPA) and SB. Participants capable of ambulatory movements were asked at the end of their examination to wear the accelerometer for 7 consecutive days.25 The device was worn over the right hip and secured with an elastic band. The accelerometers were set to begin collection data in 1-minute intervals beginning at midnight after the examination. Since the devices are not waterproof, no data collection occurred during water-based activities (eg, showering, swimming). In addition, participants were instructed to remove the monitor at bed time (no data were provided regarding on/off time for the devices).
Before analysis, PA data were compiled and analyzed using SAS macro programs available from the National Cancer Institute (NCI)26 to exclude participants with impossible data values, and consolidate the data. Nonwear time was defined as consecutive bouts of nonactivity (ie, counts of 0) for 60 minutes, with allowances for up to 2 consecutive minutes with activity counts less than 100. Valid days were defined as wearing of the device for 10 or more hours27 and participants with less than 4 valid days were excluded from the analysis. Time spent in MVPA and SB during wear time were obtained across valid days using the thresholds of ≥2020 counts per minute (cpm) and 423.50 min/day = highest tertile), respectively.
Body Mass Index Weight and standing height were collected from participants by health trained technicians during the examination. Weight was measured in pounds (later converted to kilograms (kg) using an automated system) using a Toledo digital scale. Participants were instructed to stand still in the center of the scale facing the recorder, with hands at their side, and looking straight ahead. Participants exceeding 440 pounds, were assessed using a similar protocol with 2 Seca digital scales. Standing height was measured in centimeters (cm) using a fixed stadiometer with vertical backboard and a moveable headboard. Participants were given instructions to move or remove hair ornaments, jewelry, buns, and braids from the top of their head. To obtain more consistent and reproducible results each participant was asked to take a deep breath and stand as tall as possible while the height was recorded. These anthropometric data were then used to compute BMI with the following formula: BMI = [kg / (cm / 100)2] The Centers for Disease Control and Prevention has specified the following categories for BMI measurements. Participants having a BMI < 18.5 = Underweight; BMI 18.5 to 24.9 = Normal; BMI 25.0 to 29.9 = Overweight; and BMI ≥ 30 = Obese.32 Due to the relative importance of obesity in regards to SD,8 and the small sample size for the underweight category (n = 84); the underweight, normal, and overweight categories were combined to create a 2-category variable for obesity (obese and nonobese).
Common Risk Factors Information was collected during the home interview component on covariates which included gender (male/female), age category (young adults 18–40, middle aged 41–65 and, older adults 65+), selfreported race/ethnicity (non-Hispanic white, non-Hispanic black, Mexican American, other/multiracial), marital status (single, married/living with partner), smoking status, and alcohol consumption. Smoking status and alcohol consumption were included because they have been reported to negatively affect sleeping quality.33,34 Participants were asked, “Do you now smoke cigarettes?” with possible responses including “every day,” “some days,” “not at all,” or “missing.” These responses were merged to create the categorical
JPAH Vol. 12, No. 12, 2015
Sleep Disorder, Physical Activity, and Sedentary Behavior 1569
variable current smoking status (“Yes” and “No”). Alcohol consumption was determined by the self-reported average number of alcoholic beverages consumed per week.
Downloaded by Washington Univ In St Louis on 09/16/16, Volume 12, Article Number 12
Statistical Analyses All statistical analyses were conducted using SAS v9.3 (SAS Institute, Inc., Cary, NC), to account for complex sampling design of the NHANES. The sample weights were recalculated based on age, gender, and racial/ethnic groups from the original NHANES mobile center weights to account for the selection bias by the inclusion criteria of this study. The crude associations between all study variables and SD were examined using Rao-Scott χ2 tests. Secondly, a multinomial logistic regression analysis was used to examine the associations of PA, SB, and BMI with SD, while controlling for other common risk factors (ie, age category, race, marital status, current smoking status, and alcohol consumption). A multinomial logit link function was applied to estimate the adjusted odds ratios (OR)
and associated 95% confidence intervals (CI) for the likelihoods of sleeping problems and diagnosed sleeping disorder compared with no sleeping problems. Lastly, cross-product interaction terms of age group with PA, SB, and BMI were included in the model to examine the moderating effects of age on the relationships of the main study variables (PA, SB, and BMI) with SD. We also examined the models with additional cross-product interaction terms between main study variables for exploratory purposes to identify possible moderating effects on the relationship with SD. Simple effects tests were conducted for significant interactions, and Bonferroni adjusted alpha level was used to estimate the associated CIs with respective OR.
Results The associations between SD and outcome variables are listed in Table 1. Among US adults an estimated 6.53% (95% CI: 5.57%– 7.49%) reported a diagnosed sleeping disorder, 17.34% (95% CI:
Table 1 Demographics for Percent of Population Characteristicsa (n = 2989) Variable Gender Male Female Age category Young adult (65) Race Non-Hispanic White Non-Hispanic Black Mexican American Other BMI Nonobese Obese Marital status Married/cohabiting Single Smoking status Nonsmoker Smoker MVPA 1st tertile (lowest) 2nd tertile 3rd tertile (highest) Sedentary behavior 1st tertile (lowest) 2nd tertile 3rd tertile (highest)
No sleeping problem
Nondiagnosed sleeping problem
Diagnosis with sleeping disorder
Total
79.15 ± 1.65 73.34 ± 1.25
12.59 ± 1.37 21.75 ± 1.17
8.26 ± 0.98 4.92 ± 0.58
48.10 ± 0.74 51.90 ± 0.74
83.92 ± 1.45 70.70 ± 1.51 73.07 ± 1.42
13.30 ± 1.47 20.08 ± 1.24 19.16 ± 1.25
2.78 ± 0.66 9.22 ± 0.88 7.77 ± 0.88
38.00 ± 1.64 44.87 ± 1.54 17.13 ± 1.51
73.50 ± 0.78 79.41 ± 1.61 86.45 ± 2.14 84.08 ± 3.12
19.60 ± 0.69 14.26 ± 1.06 9.07 ± 1.59 10.41 ± 2.08
6.90 ± 0.60 6.33 ± 1.29 4.48 ± 1.21 5.51 ± 1.9
71.86 ± 2.84 11.48 ± 1.98 7.97 ± 1.00 8.69 ± 1.13
78.98 ± 1.10 70.45 ± 1.46
17.36 ± 0.99 17.32 ± 1.54
3.69 ± 0.52 12.23 ± 0.83
66.82 ± 1.57 33.18 ± 1.57
77.62 ± 0.98 73.00 ± 1.60
15.59 ± 0.95 21.05 ± 1.34
6.79 ± 0.49 5.96 ± 0.79
67.88 ± 1.88 32.12 ± 1.88
76.80 ± 0.90 73.57 ± 2.36
17.02 ± 0.79 18.61 ± 1.58
6.19 ± 0.43 7.82 ± 1.41
79.42 ± 1.30 20.58 ± 1.30
66.25 ± 1.77 77.29 ± 1.05 81.53 ± 1.37
22.73 ± 1.41 16.90 ± 1.04 14.25 ± 1.40
11.02 ± 0.98 5.82 ± 0.61 4.22 ± 0.64
25.83 ± 0.95 34.20 ± 1.55 39.97 ± 1.65
78.88 ± 1.20 75.02 ± 1.36
16.10 ± 1.29 18.61 ± 1.25
5.02 ± 0.77 6.37 ± 0.82
33.84 ± 1.01 34.92 ± 1.11
74.39 ± 1.73
17.28 ± 1.58
8.33 ± 1.15
31.23 ± 1.33
a Values
Rao-Scott χ2 20.82
P-value < 0.01*
49.42
< 0.01*
35.83
< 0.01*
46.66
< 0.01*
10.39
< 0.01*
2.26
0.32
59.94
< 0.01*
7.51
0.11
represent the percentage of population characteristic. Percentages are rounded to the 100th place. * Indicates statistical significance P < .05. Note. Total population percentage for no sleeping problem, nondiagnosed sleeping problem, and diagnosis with a sleeping disorder were 76.13 (SE = 0.58), 17.34 (SE = 0.51), and 6.53 (SE = 0.49), respectively. Abbreviations: CI, confidence interval; BMI, body mass index; MVPA, moderate and vigorous physical activity. JPAH Vol. 12, No. 12, 2015
1570 Farnsworth, Kim, and Kang
16.34%–18.34%) reported sleep problems, and 76.13% (95% CI: 74.99%–77.27%) reported no sleep problems. Among adults in the highest tertile of physical activity an estimated 4.22% (95% CI: 2.96%–5.49%) reported diagnosed sleeping disorders compared with those in the lowest tertile where 11.02% (95% CI: 9.09%– 12.95%) reported diagnosed sleeping disorders. Among adults in the highest tertile of SB an estimated 8.33% (95% CI: 6.08%–10.57%) reported a diagnosed sleeping disorder compared with those in the lowest tertile where 5.02% (95% CI: 3.51%–6.53%) reported a diagnosed sleeping disorder. Adjusted odds ratios for all variables are provided in Table 2. SD were significantly associated with MVPA when controlling for covariates (Wald χ2 (4) = 38.6; P < .001). Using no sleep problems
as a reference category, participants in the highest tertile of PA were less likely to report sleeping problems (OR = 0.55; 95% CI: 0.36–0.86) and diagnosed sleeping disorders (OR = 0.39; 95% CI: 0.25–0.62), respectively, compared with those in the lowest tertile of PA. Participants in the middle tertile of PA were less likely to report sleeping problems (OR = 0.64; 95% CI: 0.47–0.88) and diagnosed sleeping disorders (OR = 0.46; 95% CI: 0.35–0.60) compared with those in the lowest tertile of PA. Among genders, females were more likely to report sleeping problems (OR = 1.63; 95% CI: 1.12–2.36), but less likely to report diagnosed sleeping disorders (OR = 0.49; 95% CI: 0.32–0.76), when compared with males. Follow-up analyses examined the data for all potential interaction effects between common risk factors (eg, age, gender, etc.) and
Downloaded by Washington Univ In St Louis on 09/16/16, Volume 12, Article Number 12
Table 2 Adjusted Odds Ratios for Sleeping Problems and Sleeping Disorders Variable Gender Male Female Age category Young adults (65) Race Non-Hispanic White Non-Hispanic Black Mexican American Other BMI Nonobese Obese Marital status Married/living with partner Single Smoking status Nonsmoker Smoker Average number of alcoholic beverages consumed per week MVPA 1st tertile (low MVPA) 2nd tertile 3rd tertile (high MVPA) Linear trend P-value SB 1st tertile (low SB) 2nd tertile 3rd tertile (high SB) Linear trend P-value
Nondiagnosed sleeping problem
Diagnosis with sleeping disorder
[Reference] 1.63 (1.12–2.36)*
[Reference] 0.49 (0.32–0.76)*
[Reference] 1.62 (1.16–2.26)* 1.05 (0.71–1.56)
[Reference] 3.05 (1.83–5.10)* 2.08 (1.13–3.81)*
[Reference] 0.56 (0.47–0.68)* 0.42 (0.28–0.64)* 0.44 (0.27–0.71)*
[Reference] 0.75 (0.44–1.30) 0.76 (0.44–1.33) 0.93 (0.51–1.71)
[Reference] 0.99 (0.70–1.40)
[Reference] 3.30 (2.24–4.85)*
[Reference] 1.52 (1.15–2.01)*
[Reference] 1.06 (0.76–1.48)
[Reference] 1.11 (0.76–1.62)
[Reference] 1.42 (0.84–2.38)
0.97 (0.91–1.04)
0.95 (0.86–1.04)
[Reference] 0.64 (0.47–0.88)* 0.55 (0.36–0.86)* 0.01*
[Reference] 0.46 (0.35–0.60)* 0.39 (0.25–0.62)* < 0.001*
[Reference] 1.03 (0.71–1.50) 0.83 (0.58–1.18)
[Reference] 1.12 (0.70–1.81) 1.04 (0.67–1.62)
0.30
0.88
Note. The above Adjusted Odds Ratios were calculated using multinomial logistic regression with “No Sleeping Problems” selected as the reference group. Abbreviations: CI, confidence interval; BMI, body mass index, MVPA, moderate and vigorous physical activity, SB, sedentary behavior. * Indicates statistical significance P < .05. JPAH Vol. 12, No. 12, 2015
Sleep Disorder, Physical Activity, and Sedentary Behavior 1571
Downloaded by Washington Univ In St Louis on 09/16/16, Volume 12, Article Number 12
main study variables (PA, SB, and BMI). A significant interaction effect was identified between age and SB (Wald χ2 (8) = 39.43; P < .001). No other interaction effects were found. Multinomial logistic regression analyses was performed for each domains of age and SB for simple effect tests (see Table 3 and 4 for simple effect tests by age and SB groups, respectively). Examination across domains of age revealed that Middle-aged adults in the highest tertile of PA were less likely to report sleeping problems (OR = 0.51; 98.33% CI: 0.28–0.93) and diagnosed sleeping disorders (OR = 0.33; 98.33% CI: 0.16–0.70) when compared with those in the lowest tertile of PA. Middle-aged adults in the middle tertile of PA were less likely to report diagnosed sleeping disorders (OR = 0.44; 98.33% CI: 0.25–0.75) when compared with those in the lowest tertile of PA. Examination across domains of SB indicate that for adults in the highest tertile of SB, PA was significantly associated with diagnosed sleeping disorders (Wald χ2 (4) = 18.53; P < .001). In addition, With the exception of those in the lowest tertile of SB (OR = 1.84; 98.33% CI: 0.71–4.79) adults who are obese are more likely to report diagnosed sleeping disorders when compared with nonobese adults.
Discussion In this nationally representative sample of adults we examined the relationship between objectively measured PA, SB, BMI, and SD when controlling for common risk factors. The results from this study suggest that for adults PA is an important risk factor for SD. This finding is in contrast to those found by Loprinzi and Cardinal35 who suggest that there were no associations between activity status and the likelihood of reporting a SD. The difference in findings is most likely the result of the different statistical models used by each study. While both models controlled for common risk factors (eg, age, race, smoking status, gender) our model included SB time as an additional variable. Inclusion of SB time into the model introduced an interaction effect between SB time and age. When examining the main study variables across age groups (ie, young adults, middle-aged adults, and older adults), we identified PA as significant risk factor for SD. Examination of risk factors for diagnosed sleeping disorders across age groups revealed that for middle-aged adults, low levels of PA were associated with an increased likelihood of diagnosed sleeping disorders. In the literature it has been reported that agerelated changes in the regulation of growth hormones as an individual transitions from early adulthood to later life stages have a strong influence on sleeping patterns and behaviors.36 Therefore, it is important when assessing sleep behavior that risk factors are examined independently for various life stages. It is also worth noting that, young adults typically had higher levels of PA and lower levels of SB (Figure 1) when compared with middle-aged and older adults. These trends are consistent with those found by Gordon-Larsen et al, which suggest a reciprocal relationship between PA, SB, and aging.37 As we age our PA levels decrease, and our SB levels increase. One particularly interesting finding in our study indicated among adults with low levels of SB, obesity and PA were not significantly associated with SD. This discovery suggest that the elevated risk of SD in obese adults may be moderated through decreasing SB which could have significant implications for future intervention strategies. In a cross-sectional examination of Australian adults it was identified that increased breaks in sedentary time were beneficially associated with diminished waist circumference, and fasting plasma glucose levels.38 In addition, previous research has suggested
that weight loss (surgical or otherwise) has been an effective treatment for SD.12–15 The decreased levels of SB would both directly and indirectly affect SD through weight loss and moderation of the effects of obesity on SD, respectively. The overall prevalence of SD from US adults were consistent with those that are reported in the literature.9,39 In men, 3.97% reported diagnosed sleeping disorders, while only 2.55% of women reported diagnosed sleeping disorders. Among genders, there were significant differences in the prevalence of sleeping problems and diagnosed sleeping disorders. Females were more likely to report sleeping problems (OR = 1.63; 95% CI: 1.12–2.36) but, nearly half as likely to report diagnosed sleeping disorders (OR = 0.49; 95% CI: 0.32–0.76). The results found in this study that woman are more likely than men to report sleeping problems may simply be an indicator that woman are more likely to report health issues compared with men as opposed to a true difference in sleep quality as the data suggest. This notion is supported by the results found by Benyamini et al,40 who reported that self-assessed health was a weaker predictor of mortality in woman than men. In addition, a study examining the sleep characteristics between genders found that females were more likely than males to report disruptive sleeping behaviors.41 When reporting negative health conditions, men tend to only report conditions which have significant negative health consequences. Women on the other hand, tend to report on a larger range of health problems (ie, sleeping problems). Similar to the findings of Bansil et al,42 there were no associations between race and diagnosed sleeping disorders. However, the likelihood of reporting sleeping problems was lower for nonwhite ethnic groups. These findings are in contrast with those found in other studies.43,44 Examination of adults using in-home polysomnography reported shorter sleep duration and lower sleep efficiency among Blacks.43 Furthermore, in the Sleep Heart Health Study frequent snoring, a predictor of sleeping disorders,45 was more common among Hispanic and Black women. The conflicting evidence may likely be the result of the methods used to collect the data. The data from this nationally representative sample were collected through self-report. Moreover, the presence of snoring does not directly indicate sleeping problems. A cross-sectional study of adults in a sleep laboratory identified waist circumference as one of the most prominent risk factors for development of SD with approximately 50% of obese adults being diagnosed with a SD.46 Reports from the National Sleep Foundation suggest that among older adults, 77% report some form of sleeping problems.47 The results of this study were consistent with those found in other studies indicating that approximately 63% of adults who report diagnosed sleeping disorders are obese. Obese adults demonstrate a much higher risk for development of SD when compared with nonobese individuals. This is particularly problematic in the United States where more than one-third of all adults are considered obese.48 Intervention studies involving lifestyle interventions have demonstrated that increasing PA levels in an obese population can significantly improve health outcomes. In a longitudinal study examining 313 obese adults, the use of a surgical weight loss procedure decreased the prevalence of obstructive sleep apnea from 33% to 2%.15 Promoting sufficient levels of MVPA among overweight and obese adults may help to reduce the prevalence rates of diagnosed sleeping disorders. It is interesting to note that while BMI was a significant predictor of diagnosed sleeping disorders, BMI had minimal associations with sleeping problems. This is in contrast to the findings from Turco et al,49 who reported that obese
JPAH Vol. 12, No. 12, 2015
1572
JPAH Vol. 12, No. 12, 2015
1.42 (0.70–2.88)
2.63 (0.92–7.49)
[Reference]
2.47 (0.82–7.46)
1.03 (0.25–4.23)
[Reference]
0.50 (0.12–2.07)
0.45 (0.13–1.56)
[Reference]
Diagnosed with sleeping disorder
0.85 (0.53–1.35)
[Reference]
0.94 (0.52–1.71)
1.34 (0.77–2.34)
[Reference]
0.51 (0.28–0.93)*
0.59 (0.32–1.09)
[Reference]
Nondiagnosed sleeping problem
3.48 (1.66–7.32)*
[Reference]
1.05 (0.48–2.28)
1.24 (0.55–2.79)
[Reference]
0.33 (0.16–0.70)*
0.44 (0.25–0.75)*
[Reference]
Diagnosed with sleeping disorder
Middle-aged adults (40–65)
0.83 (0.40–1.70)
[Reference]
0.67 (0.29–1.50)
0.79 (0.47–1.32)
[Reference]
0.97 (0.37–2.60)
0.74 (0.39–1.43)
[Reference]
Nondiagnosed sleeping problem
3.28 (1.78–6.06)*
[Reference]
0.34 (0.10–1.14)
0.56 (0.12–2.46)
[Reference]
0.68 (0.28–1.67)
0.40 (0.17–0.97)*
[Reference]
Diagnosed with sleeping disorder
Older adults (>65)
0.92 (0.49–1.73)
[Reference]
1.84 (0.71–4.79)
[Reference]
1.11 (0.59–2.10)
[Reference]
0.36 (0.19–0.70)*
0.66 (0.36–1.23)
[Reference]
Nondiagnosed sleeping problem
6.12 (2.40–15.60)*
[Reference]
0.21 (0.05–0.83)*
0.62 (0.27–1.40)
[Reference]
Diagnosed with sleeping disorder
Middle tertile
0.99 (0.52–1.90)
[Reference]
0.75 (0.29–1.92)
0.55 (0.28–1.07)
[Reference]
Nondiagnosed sleeping problem
2.94 (1.51–5.74)*
[Reference]
0.33 (0.13–0.79)*
0.26 (0.10–0.65)*
[Reference]
Diagnosed with sleeping disorder
Highest tertile
Note. The above Adjusted Odds Ratios were calculated using multinomial logistic regression with “No Sleeping Problems” selected as the reference group across domains of Sedentary Behavior. All covariates (excluding MVPA and BMI) included within the model have been excluded from the table. Covariates include: age, race, alcohol consumption, gender, smoking status, and marital status. Abbreviations: BMI, body mass index; MVPA, moderate and vigorous physical activity; SB, sedentary behavior. * Indicates statistical significance P < .0167 (Bonferroni adjusted alpha for multiple comparisons 0.05/3).
Obese
Nonobese
BMI
1.12 (0.31–4.00)
0.66 (0.32–1.39) 0.71 (0.38–1.33)
2nd Tertile
3rd Tertile (high)
1.48 (0.59–3.74)
[Reference]
Diagnosed with sleeping disorder
[Reference]
1st Tertile (low)
MVPA
Nondiagnosed sleeping problem
Lowest tertile
Domain—sedentary behavior tertile
Table 4 Adjusted Odds Ratios and 95% Confidence Intervals for the Effects of BMI and MVPA Sleeping Disorders by Tertiles of Sedentary Behavior
Note. The above Adjusted Odds Ratios were calculated using multinomial logistic regression with “No Sleeping Problems” selected as the reference group across domains of Age group. All covariates (excluding MVPA) included within the model have been excluded from the table. Covariates include: race, alcohol consumption, gender, smoking status, and marital status. Abbreviations: BMI, body mass index; MVPA, moderate and vigorous physical activity; SB, sedentary behavior. * Indicates statistical significance P < .0167 (Bonferroni adjusted alpha for multiple comparisons 0.05/3).
Obese
Nonobese
[Reference]
0.80 (0.36–1.80)
3rd Tertile (high)
BMI
0.75 (0.34–1.63)
2nd Tertile
1st Tertile (low)
[Reference]
0.49 (0.19–1.26)
SB
0.56 (0.25–1.27)
3rd Tertile (high)
[Reference]
2nd Tertile
1st Tertile (low)
MVPA
Nondiagnosed sleeping problem
Young adults (
