To determine effects of yoga and aerobic exercise compared with usual activity on objective assessments of sleep in midlife women.
Secondary analyses of a randomized controlled trial in the Menopause Strategies: Finding Lasting Answers for Symptoms and Health (MsFLASH) network conducted among 186 late transition and postmenopausal women aged 40–62 y with hot flashes. Women were randomized to 12 w of yoga, supervised aerobic exercise, or usual activity. The mean and coefficient of variation (CV) of change in actigraph sleep measures from each intervention group were compared to the usual activity group using linear regression models.
Baseline values of the primary sleep measures for the entire sample were mean total sleep time (TST) = 407.5 ± 56.7 min; mean wake after sleep onset (WASO) = 54.6 ± 21.8 min; mean CV for WASO = 37.7 ± 18.7 and mean CV for number of long awakenings > 5 min = 81.5 ± 46.9. Changes in the actigraphic sleep outcomes from baseline to weeks 11–12 were small, and none differed between groups. In an exploratory analysis, women with baseline Pittsburgh Sleep Quality Index higher than 8 had significantly reduced TST-CV following yoga compared with usual activity.
This study adds to the currently scant literature on objective sleep outcomes from yoga and aerobic exercise interventions for this population. Although small effects on self-reported sleep quality were previously reported, the interventions had no statistically significant effects on actigraph measures, except for potentially improved sleep stability with yoga in women with poor self-reported sleep quality.
Buchanan DT, Landis CA, Hohensee C, Guthrie KA, Otte JL, Paudel M, Anderson GL, Caan B, Freeman EW, Joffe H, LaCroix AZ, Newton KM, Reed SD, Ensrud KE. Effects of yoga and aerobic exercise on actigraphic sleep parameters in menopausal women with hot flashes. J Clin Sleep Med. 2017;13(1):11–18.
Women experiencing the menopausal transition and postmenopause often complain of sleep disturbance. Yoga and exercise have been suggested as useful nonpharmacologic treatments for midlife women experiencing sleep disturbance, but evidence regarding efficacy remains mixed and inconclusive. Several studies of yoga interventions for 8 to 16 w duration in women during and after menopause1–6 showed statistically significant within-group improvements in self-reported sleep outcomes.1–4 However, randomized controlled trials (RCTs) failed to show significant improvements in self-reported sleep with yoga compared to gentle exercise,5 moderate-intensity walking,6 or usual care.6 Only one yoga RCT of midlife women required participants to meet clinical diagnostic criteria for insomnia and reported significant improvement in insomnia symptoms versus usual care.3 However, this study failed to detect a significant difference in self-reported insomnia symptoms between yoga and passive stretching, and also found no differences among yoga, passive stretching, or usual care groups on objective polysomnographic sleep outcomes. Of four RCTs comparing 4 to 12 mo of aerobic exercise,6–10 only one reported significant improvement in self-reported sleep quality compared to nonexercise controls.10 Analyses of persons who improved in two of these studies showed that higher levels of exercise predicted greater improvement in sleep,7 but only among those exercising in the morning.9
Current Knowledge/Study Rationale: Yoga and exercise have been suggested as useful non-pharmacological treatments for midlife women experiencing sleep disturbance, but evidence regarding efficacy remains mixed and inconclusive. This study reports actigraphic sleep outcomes from a randomized controlled trial comparing 12 w of yoga, aerobic exercise, or usual care in women experiencing the menopausal transition or postmenopause with hot flashes.
Study Impact: Sleep disturbances were shown on baseline actigraphy sleep measures; however, yoga and aerobic exercise interventions did not significantly affect sleep measures. The limited body of current evidence does not support the efficacy of yoga or aerobic exercise for reducing sleep disturbance in women during the menopause transition or postmenopause with hot flashes. Future research should explore other approaches for improving sleep quality, such as cognitive behavioral therapy for insomnia.
The Menopause Strategies: Finding Lasting Answers for Symptoms and Health (MsFLASH) research group compared 12 w of yoga or aerobic exercise to usual activity on the primary outcomes of hot flashes in women in the late transition of menopause or postmenopause (N = 355).11,12 Although hot flash frequency and bother did not significantly improve in the intervention groups versus controls, secondary outcomes of self-reported sleep quality and insomnia severity were improved.11,12 Yoga was associated with improvement in the global sleep quality score on the Pittsburgh Sleep Quality Index (PSQI, pre to post change = −0.5 relative to usual activity, p = 0.049) and a reduction in the Insomnia Severity Index score (ISI, pre to post change = −1.3 relative to usual activity, p = 0.007).11 Similar changes in the PSQI (change = −0.8, p = 0.007) and ISI (change = −0.9, p = 0.025) were observed with aerobic exercise versus usual activity.12 These findings based on self-reported sleep provide modest support that yoga or exercise may improve sleep quality in midlife women.
Analysis of objective sleep outcomes provides a complementary perspective to the subjective findings on the efficacy of yoga or aerobic exercise for improving sleep in midlife women with vasomotor symptoms. Objective sleep data are available from the MsFLASH yoga and aerobic exercise trial on a subset of participants who wore an actigraph during baseline and prior to postintervention assessments (week 11–12).13 Subjective sleep measures may change without a corresponding improvement in objective measures. Evidence on objective sleep measures would inform the potential benefits of yoga and/or exercise for improving objective sleep quality midlife women with hot flashes. More importantly, multiday actigraphy data provide for analysis of sleep variability, a novel sleep outcome. Reduced night-to-night variability obtained from actigraphy could reflect improved sleep stability as an objective measure of sleep quality. As an example, an analysis of night-to-night variability showed significant differences in actigraphy measures between persons with no insomnia and those with chronic insomnia, suggesting that improved sleep stability may be an important outcome variable in insomnia trials.14 Therefore, we conducted secondary analyses to determine whether the mean values and variability of actigraphy-derived sleep variables differed with yoga or aerobic exercise interventions versus usual activity. The paucity of objective sleep data in trials of interventions for midlife women with vasomotor symptoms makes an analysis of actigraphy data from MsFLASH yoga and aerobic exercise study worthwhile.
MsFLASH study methods have been published,13,15 including the yoga and aerobic exercise trial.11,12 In brief, women in the late transition phase of menopause and postmenopause who qualified for participation were randomized to 12 w of yoga, aerobic exercise, or usual activity. The trial used a three-by-two factorial design, and simultaneously randomized participants from across the three behavioral intervention groups to take either an omega-3 fish oil oral supplement or placebo. No self-reported sleep outcome differences were found between omega-3 fish oil and placebo capsules,16 such that this intervention is not considered in the current analyses. Institutional Review Boards at each clinical site and the MsFLASH Data Coordinating Center approved the study. All participants provided written informed consent prior to completing any study procedures.
The yoga intervention was based on the Viniyoga style, and included a variety of poses (restorative, inverted, twists, lateral and forward bends) as well as a relaxing, meditative Yoga Nidra practice.11 The 12 weekly classes were 90 min long (attended once weekly) and were taught by yoga instructors with at least 5 y of experience and 500 hours of training. A staff member attended every yoga class to monitor adherence to the instruction protocol. Classes were taught on an ongoing basis, but used three differing sequences of the poses to maintain interest. Women joined the classes as they were recruited. The participants were instructed to practice at home for 20 min daily on nonclass days, alternating between poses one day and Yoga Nidra the next day. They were given written instructions, a DVD with the three yoga sequences, a CD guiding Yoga Nidra practice, and supportive yoga props (mat, blanket, strap, and bolster).
Participants in the aerobic exercise group attended 12 w of in-person training, with three weekly sessions.12 The sessions typically lasted 40–60 min; the duration was based on the work load needed to achieve the target energy expenditure goal, which progressed over the course of the program. The sessions were held at local fitness facilities and were supervised by a trained certified exercise trainer. Women self-selected the specific exercise (treadmill, elliptical trainer, or stationary bike) and exercised to a target heart rate based on heart rate reserve (50% to 60% in month 1, 60% to 70% thereafter). The aerobic exercise group was not asked to complete home practice sessions.
Eligibility, Screening, Randomization, Data Collection, and Blinding
Participant recruitment and data collection occurred at study sites in Indianapolis, Oakland, and Seattle. Eligibility criteria were: aged 40–62 y; in the menopausal transition or postmenopause or had hysterectomy with follicle stimulating hormone levels higher than 20 mIU/mL and estradiol levels 50 pg/mL or lower; generally in good health; experiencing 14 or more hot flashes/night sweats per week (on 2-w screening diaries); and hot flashes rated as bothersome or severe on four or more occasions/week. Exclusion criteria were: body mass index greater than 37 kg/m2; using hormonal medications (including contraceptives) during the past month; use of any medications to treat hot flashes in the past month; any unstable medical condition; current practice or use of one of the study interventions; contraindications for exercise or yoga; or major depressive episode in the past 3 mo. Subjects were not selected on the basis of insomnia or sleep disturbance symptoms.
Women were recruited by mass-mailing, using purchased lists and health plan enrollment files. Women who responded to the study invitation were screened by telephone, then completed screening questionnaires and 2-w vasomotor symptom diaries. Data collectors were blinded to intervention group allocation.
Following initial screening, participants completed an in-person baseline study visit. Data collection included a blood draw, physiological assessment, and baseline questionnaires. Following the baseline visit, participants wore an actigraph and completed a 1-w vasomotor symptom and sleep diary. Of the 355 participants enrolled and randomized in the parent study, a subset of 207 women wore an actigraph. There were a limited number of devices available at each study site for this ancillary study, and the number of participants wearing the actigraph significantly differed by site: Indianapolis (n = 87) versus Oakland (n = 63) and Seattle (n = 57) (p = 0.001). Those who wore an actigraph at baseline were more highly educated than those who did not (p = 0.036), but characteristics did not otherwise differ. Following the 1-w baseline data collection, participants returned to the clinic. Subjects whose hot flashes and night sweats did not decline by more than 50% over the prior 2-w screening period were allocated to yoga, aerobic exercise, or usual activity in a 3:3:4 ratio. Randomization was conducted by the MsFLASH Data Coordinating Center using a dynamic balancing randomization algorithm to maintain comparability between the study groups across the clinical sites. Assessments were repeated at week 11–12, the final week of the intervention. Participants were compensated $50 after each clinic visit for a possible total of $150.
Participants' demographic characteristics, select health data (including smoking history and alcohol use), and baseline symptoms were collected at the baseline visit, as described in a prior publication.13 In addition to basic demographic information, height and weight were measured to obtain body mass index. Baseline symptom assessment included the Generalized Anxiety Disorder 7-item scale,17 the Patient Health Questionnaire 8-item scale (PHQ-8)18 to assess depression, the ISI,19 and the PSQI.20 Global PSQI scores range from 0 to 21 with higher scores indicating poorer sleep quality. Cutoff scores of 5 and 8 have been reported to indicate poor sleep quality;20,21 the higher cutoff point was used in exploratory analyses because of previous studies suggesting that self-reported sleep disturbance is common among perimenopausal and postmenopausal women.22–24 Eighty percent of the women in this sample had a PSQI global score of greater than 5, and 49% had scores higher than 8.
The primary sleep outcomes in these analyses included percent change from baseline to 12 w in actigraphy-derived TST (total minutes asleep during time in bed [TIB]), WASO (time spent awake between initial sleep onset and final sleep offset), and number of long awakenings (LWAK = number of awakenings longer than 5 min). Secondary sleep outcomes that were also examined and included percent change from baseline to 12 w in sleep onset latency (time between bedtime and falling asleep) and sleep efficiency (percentage of time in bed spent asleep = TST / TIB × 100).
Participants wore an actigraph (Actiwatch 2, Philips Respironics, Bend, OR) on the nondominant wrist. Activity counts were summed in 1-min epochs. The device also recorded presence and density of white light and marker data (participants were instructed to press the marker button on the side of the device at bedtime and rise time) that were used in the scoring of actigraphy data. Actigraphy data were scored in Actiware 5 (Philips Respironics). Designation of the start and end of TIB was manually entered on each record by one of two trained technicians using a standard scoring algorithm codeveloped by one of the authors (DTB). Bedtimes and rise times were determined primarily from the participants' sleep diaries. However, the scoring algorithm provided a hierarchy for using actigraphlogged activity levels, actigraph-recorded light levels, and actigraph-recorded event markers to determine bedtimes and rise times when there was an observed mismatch between diary recorded bed/rise times and one or more of the actigraph-derived indicators. All scored records were visually inspected for accurate scoring by an investigator with expertise in actigraphy (DTB). The scorers and reviewer were blind to intervention group assignment. Each epoch was scored as sleep or wake using the automated algorithm in Actiware 5, and the software generated the sleep variables (TST, WASO, sleep onset latency, and sleep efficiency). Because LWAK is not included in the Actiware software package, this was calculated using locally developed SPSS (IBM, Armonk, NY) syntax.
Data were analyzed using SAS version 9.3 statistical software (SAS Institute Inc, Cary, NC). The mean and variability values of the actigraphy outcomes were calculated from each week of actigraphy data – baseline and the last week of the intervention (week 11–12). To provide consistency for the variability analyses,25 we included only women with 7 nights of actigraphy data at both assessment weeks: total n = 186, yoga n = 52, aerobic exercise n = 54, usual activity n = 80. The reduced number of participants was similar in each group (yoga n = 7, aerobic exercise n = 5, usual activity n = 6). Data loss occurred mainly from participant failure to wear the actigraph as directed or device malfunction.
Differences in intervention group characteristics at baseline were compared using analysis of variance or chi square (χ2) tests. Variability of actigraphy data was represented by the coefficient of variation (CV) of each sleep variable. The CV is calculated as the within-subject standard deviation divided by the within-subject mean multiplied by 100.26 This calculation allows comparison of variability between measures with different units; greater values indicate higher variability. Linear regression models were used to compare the primary outcomes and secondary outcomes for each intervention group relative to the usual activity group. All analyses were adjusted for baseline values on each outcome as well as clinical site. All variables except TST-mean were right-skewed and were log-transformed prior to analysis. A value of p < 0.05 was considered significant.
Baseline characteristics are shown in Table 1. The mean (± standard deviation) age was 54.9 ± 3.7 y. There were no significant differences in baseline characteristics between the intervention groups except that compared with usual activity, a lower proportion of participants in the aerobic exercise group had graduated college (p = 0.005) and mean alcohol consumption per week was lower in the yoga group (p = 0.033). Most of the women were postmenopausal (82.2%). In all three groups at baseline, mean PSQI scores exceeded the threshold for sleep disturbance (higher than 5),20 and mean ISI scores were above the cutoff considered to indicate at least mild insomnia (8 or higher).19
Baseline demographic and clinical characteristics by group.
Baseline demographic and clinical characteristics by group.
Actigraphic Sleep Outcomes
Baseline sleep measures were: TST-mean, yoga, 400.6 ± 65.2, aerobic exercise, 412.3 ± 46.5, usual activity, 408.8 ± 57.3 min; WASO-mean, yoga, 53.2 ± 19.9, aerobic exercise, 54.5 ± 23.5, usual activity, 55.7 ± 22.1 min; WASO-CV, yoga, 40.4 ± 19.1, aerobic exercise, 36.1 ± 18.2, usual activity, 37.2 ± 18.7; and LWAK-CV, yoga, 75.8 ± 37.4, aerobic exercise, 84.2 ± 47.1, usual activity, 83.5 ± 52.3.
Week 11–12 actigraphy sleep measures are shown in Table 2, along with change scores from baseline. Participants in all three groups slept on average about 6.75 hours each night, with nearly 1 h of WASO and two to three long awakenings per night. Sleep efficiency was similar across groups. The mean or CV values of these sleep measures at baseline did not differ significantly across study groups. Comparisons of yoga and aerobic exercise to usual activity revealed no significant differences in the means or CVs of any change in the sleep measures from baseline to week 11–12 (Table 2).
Baseline to week 11–12 change in actigraphy measures by treatment group.
Baseline to week 11–12 change in actigraphy measures by treatment group.
We conducted analyses to explore whether the findings differed among those who were most adherent to the intervention or those who had poor sleep at baseline (i.e., two groups most likely to experience improvement). Adherence was defined as completion of at least 80% of the sessions (which included home practice sessions in the yoga group). There were no differences on any outcome from baseline to week 11–12 between adherent yoga participants (n = 31, 60%) and adherent exercise participants (n = 45, 83%) versus the usual activity group.
Given that women were not selected on the basis of reported sleep disturbance or insomnia research criteria, we explored interactions between intervention group assignment and baseline sleep quality (PSQI) or insomnia symptoms (ISI) scores in linear regression analyses. No interactions of ISI with treatment group were statistically significant. In the yoga versus usual activity analyses, interactions between group and PSQI were statistically significant for LWAK-mean and TST-CV (p = 0.048 and 0.021, respectively). In the exercise versus usual activity analyses, interactions between group and PSQI were statistically significant for WASO-mean (p = 0.047).
We further explored these significant interactions in subgroup analyses of participants with baseline PSQI scores higher than 821 (yoga, n = 21; exercise, n = 29; usual activity, n = 42). In the yoga versus usual care analysis, TST-CV was significantly lower in the yoga group: % Change (95% confidence interval): −26.6 (−40.3, −9.7), p = 0.004 after adjusting for site and baseline TST-CV. None of the other PSQI analysis (LWAK-mean for yoga versus usual activity, WASO-mean for exercise versus usual activity) were statistically significant.
Our findings showed that 12 w of yoga or aerobic exercise had no effects on actigraph sleep parameters in women in late-transition menopause and postmenopausal women with hot flashes. In this MsFLASH trial, changes in actigraph sleep measures were not observed despite modest but statistically significant improvements in self-reported sleep quality and insomnia symptoms in the exercise and yoga intervention arms.11,12 Although women were not selected on the basis of sleep disturbance symptoms, mean baseline actigraph scores of sleep were beyond values often used to indicate poor sleep quality and at least mild insomnia, respectively. Nevertheless, change in mean actigraphy-derived sleep measures and values of sleep night-to-night variability were unaffected by the yoga or aerobic exercise interventions, each compared to usual activity. However, we found reduced variability in TST following yoga in women with baseline PSQI scores higher than 8, suggesting that the women with self-reported poor sleep experienced improvement in sleep stability.
Studies of midlife women with hot flashes commonly show self-reported sleep disturbance,27,28 which the current study corroborates with actigraphy data. The baseline actigraphy measures in this study showed no difficulty falling asleep, but moderate amounts of WASO (54–57 min), along with mildly reduced sleep efficiency (85% to 86%). A sleep efficiency less than 85% is used as an arbitrary indicator of insomnia and more than 30 min of WASO is used as one indicator of insomnia. Women were also sleeping, on average, less than the recommended minimum of at least 7 h/night for adults.29 Given that sleep disturbance was not an inclusion criterion for this study, the findings support the importance of assessing and addressing the symptoms of sleep disturbance and inadequate sleep in women during the menopause transition and postmenopause.
This study adds to the limited body of evidence on yoga for sleep in midlife women. With the increasing popularity of yoga as a mainstream form of exercise, it is important to evaluate as it is recommended for treatment of symptoms such as hot flashes. Overall, the evidence from prior studies of self-reported outcomes are mixed, tending to show within-group improvements with yoga,1,2,4 but no significant differences between yoga and control groups.4,6 To our knowledge, only one prior study of objective and subjective sleep outcomes from a yoga intervention for menopausal women reported reduced ISI scores in the yoga group compared to usual activity group, but no improvements on polysomnography-derived parameters.3
Prior evidence of effects of aerobic exercise is also uncertain. To our knowledge our study is among the first to report objective sleep outcomes from aerobic exercise in women in the late phase of the menopausal transition and postmenopause. As with yoga, findings from this MsFLASH trial showed small but significant improvements in self-reported sleep quality and insomnia symptoms with aerobic exercise versus usual activity.12 These findings are similar to another study that reported significant improvement in sleep symptoms,10 but improvements in objective measures did not reach statistical significance. Several additional studies of aerobic exercise for menopausal women evaluated only self-reported outcomes, without objective measures, and found no significant differences between the intervention and control groups on self-reported sleep outcomes.6–8
Overall, the current study adds objective evidence to prior studies using self-report, indicating that yoga and exercise interventions do not appear to improve sleep duration or sleep continuity in midlife women selected on the basis of hot-flash frequency, and who had an average of 7.3 to 8 hot flashes per day.11,12 We had hypothesized that the variability in the sleep measures would be reduced after 12 w of a yoga or exercise intervention, but we found that the CVs of the sleep measures were unchanged from baseline. Reduced TST variability was observed in women with poor baseline sleep (PSQI higher than 8), indicating a potential sleep-stabilizing effect of the yoga intervention.
High variability in actigraph-derived sleep duration has been associated with poorer subjective sleep quality and decreased well-being,26 and several studies have shown night-tonight variability in sleep as a characteristic of insomnia.14,30 The CV values at both baseline and week 11–12 in our study were similar to those reported in community-dwelling, relatively healthy middle-aged and older women and men.25,26 The focus of this analysis was not on examining the extent of night-to-night variability in relation to perceived sleep quality, hot flashes, or other baseline characteristics, but such an analysis would enhance understanding of the effect of sleep variability on symptoms and quality of life.
A limitation of this analysis could be that we did not account for the effects of hot flashes. However, prior analyses of the Ms-FLASH yoga and aerobic exercise trial showed no effects of these interventions on hot flashes, yet small improvements in self-reported sleep quality and insomnia symptoms. The authors interpreted these findings to suggest that complaints of poor sleep may be a function of age-related changes during the menopause transition.12 Another limitation is that the sample was restricted to women in late-transition menopause and postmenopausal women with hot flashes; thus, the effects of yoga and aerobic exercise on sleep may not be applicable to the broader population of midlife women without hot flashes. Because of our interest in studying possible changes in night-to-night variability we restricted the analyses to subjects with 7 nights of actigraphy data both on baseline and during the final intervention week. However, the small number of participants excluded from the final analyses was similar in the three intervention groups, suggesting that the loss of participants with incomplete data was unlikely to have introduced bias into the analyses. Finally, the algorithm we used for determining bedtimes and rise times on actigraphy has not been validated, but use of our algorithm provided a consistent approach in the scoring of actigraphy data.
Strengths of this study include the randomized controlled trial, multisite study design, and the careful characterization and selection of women restricted to the late phase of the menopausal transition and postmenopause when hot flashes and sleep disturbance symptoms are high. In addition, rigorous procedures were used in the processing and scoring of the actigraph data.
In summary, no significant effects of yoga or aerobic exercise on objectively measured sleep outcomes were observed among midlife women with hot flashes, although self-report improvements had been previously reported.11,12 There was some evidence of improved sleep stability with yoga in women with poor sleep quality on an exploratory analysis. At baseline, actigraphy showed that on average, midlife women with hot flashes slept less than the recommended amount and health consequences of disrupted and inadequate sleep are well known and of concern.31–33 Future research should explore other approaches for improving sleep quality, such as cognitive behavioral therapy for insomnia, to address the important problem of sleep disturbance in women during the menopausal transition and postmenopause.
This was not an industry supported study. This study was funded by the National Institutes of Health as a cooperative agreement issued by the National Institute on Aging, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Center for Complementary and Alternative Medicine, Office of Research on Women's Health, and grants U01AG032656, U01AG032659, U01AG032669, U01AG032682, U01AG032699, and U01AG032700 from the National Institute on Aging. At Indiana University, the project was partly funded by the Indiana Clinical and Translational Sciences Institute, grant UL1RR02571 from the National Institutes of Health, National Center for Research Resources, and Clinical and Translational Sciences Award. Diana Taibi Buchanan, Carol A. Landis, Katherine A. Guthrie, Julie L. Otte, Garnet L. Anderson, and Andrea Z. LaCroix received funding from the National Institutes of Health (NIH) during the conduction of the study. Ellen W. Freeman has received research support from Forest Laboratories Inc. Hadine Joffe has received grant support from Merck and Cephalon/Teva and is a consultant for Noven, Merck, and Mitsubishi Tanabe, and has done consulting for Sunovion. Andrea Z. LaCroix has consulted for Pfizer and Sermonix. Susan D. Reed has received research support from Bayer Pharmaceutical, holds a patent relating to methods and compositions for treating vasomotor symptoms (WO2014089019A1), and has received royalties from UpToDate and Scientific American Medicine. Katherine M. Newton received research support from NIH during the conduction of the study; she serves on the board of the North American Menopause Society. Kristine E. Ensrud received grants from NIH during the conduction of the study and is a consultant to a Data Monitoring Committee for Merck, Sharp, and Dohme. The other authors have indicated no financial conflicts of interest.