Reallocating Time to Sleep, Sedentary Time, or Physical Activity: Associations with Waist Circumference and Body Mass Index in Breast Cancer Survivors

Obesity is associated with poorer prognoses in females diagnosed with breast cancer (1). Weight gain after diagnosis, in particular, has been associated with greater comorbidity and poorer quality of life (1). The majority of breast cancer survivors gain weight during and after treatment (2), so it is important to identify strategies that can prevent this from occurring.

Research indicates that higher levels of moderate-to-vigorous intensity physical activity (MVPA) are associated with improved weight control and body composition in breast cancer survivors (3, 4). However, MVPA comprises a small fraction of each day; studies using objective measures have found MVPA accounts for 1% to 3.5% of waking hours in breast cancer survivors (5–8). Few studies have examined behaviors that account for much larger portions of the day, such as sleep, sedentary behavior, and light-intensity physical activity (LPA; ref. 9). Recent research in breast cancer survivors has found sedentary time accounts for between 57% and 79% of waking hours, LPA accounts for between 20% and 40%, (5–8), and daily sleep duration is approximately 7 hours (10). One previous study has examined how sedentary behavior and LPA are associated with body composition. This study suggested higher levels of sedentary time and lower levels of LPA were significantly associated with higher body mass index (BMI) and greater waist circumference; however, these associations were largely attenuated after adjustment for MVPA (8).

Sleep, sedentary, and physically active behaviors are interdependent, because the finite number of hours in a day means that increasing time in one of these behaviors decreases time spent in others. The way sedentary time is accumulated throughout the day may also impact its influence on health outcomes. Previous research indicates that prolonged, unbroken bouts of sedentary time have a greater detrimental effect on cardiometabolic outcomes than shorter bouts (11, 12). No previous studies in breast cancer survivors have examined sedentary time accumulation patterns or taken into account the interdependent nature of sleep, sedentary, and active behaviors.

In this study, we used isotemporal substitution modeling to account for the full 24-hour day to investigate the interdependent associations of self-reported sleep and objectively assessed sedentary time, LPA, and MVPA with waist circumference and BMI in breast cancer survivors.

The ACCEL-Breast study was a cross-sectional study that was conducted in Western Australia (WA) in 2013. The full methods have been described previously (6). Briefly, participants were between 2 and 4 years postdiagnosis, between 18 and 80 years of age, and residing in WA at the time of diagnosis, and had previously taken part in a case–control study (13). Breast cancer survivors who were currently receiving chemotherapy or radiotherapy, had experienced a recurrence of their breast cancer, or had been diagnosed with another cancer were ineligible.

Participants were asked to wear an Actigraph GT3X+ accelerometer (ActiGraph Corporation) on their right hip during waking hours for 7 consecutive days and to complete a written questionnaire assessing demographic information and patient-reported outcomes. A total of 552 eligible breast cancer survivors were invited to take part in the study, 340 of whom agreed to participate in the study and were sent the study questionnaire and an accelerometer. Of those 340 participants, 274 (50% of the 552 eligible breast cancer survivors invited) completed the study. The participants and nonparticipants did not differ in terms of age, socioeconomic status, breast cancer grade, or time since diagnosis (6). Eighteen participants were missing data on one or more of the variables used in this analysis, leaving 256 participants in the current study. A further 18 participants did not report their current height and/or weight, so the BMI analysis contains only 238 participants. Written informed consent was obtained from all participants, and the study was approved by Human Research Ethics Committees at the WA Department of Health and The University of WA (Western Australia, Australia).

The accelerometer data were summarized using 60-second epochs. Commonly applied cut-off points were used to derive daily time spent sedentary [<100 counts per minute (CPM)], in LPA (100–1,951 CPM) and in MVPA (≥1,952 CPM; refs. 14, 15). Time spent sedentary was split into prolonged sedentary time (≥20 minutes, with no allowance for an interruption) and nonprolonged sedentary time, as sedentary bouts of ≥20 minutes have been shown to adversely affect cardiometabolic biomarkers (16). Non-wear time was defined as intervals of at least 60 consecutive minutes of zero counts (with allowance for ≤2 minutes of observations of <50 CPM). To be considered valid, days of data collection required at least 10 hours of wear time and no excessive counts (>20,000 CPM).

Average sleep duration in the last month was assessed using the following question from the Pittsburgh Sleep Questionnaire Index (PSQI): “During the past month, how many hours of actual sleep did you get at night (this may be different than the number of hours you spent in bed)?” (17). The PSQI has acceptable validity and reliability when assessing sleep quality and duration in the general population and in people with cancer (17, 18).

Waist circumference was measured (by each participant) using an elastic tape measure. Participants were asked to stretch out the tape measure, wrap it around their (exposed) waist at navel-level, then breathe out and measure (to the nearest 0.1 cm). They were asked to repeat this three times. The average of the three measurements was used in these analyses. Participants were also asked to self-report their current height and weight.

Breast cancer stage and time since diagnosis were determined from data obtained from the WA Cancer Registry. Information about various demographic, behavioral, and medical factors, including highest level of education, current working status, smoking status, comorbidities, breast cancer treatment, and hormone therapy use, was obtained via self-report from the participants (6). Area-level socioeconomic status was assessed using the Index of Relative Socio-Economic Disadvantage at the postcode level (19).

All exposure variables in isotemporal models must be in the same metric, so daily time spent sleeping and in prolonged sedentary, nonprolonged sedentary, LPA, and MVPA were all converted to units of 30 minutes (e.g., 15 minutes = 0.5, 30 minutes = 1). We also created a “total time” variable, which was the sum of these five activities, and converted this to the same 30-minute metric. We chose to use 30-minute units to aid interpretation of the results; as it is assumed the relationship between the exposure variables and the outcomes is linear, the choice of metric does not impact the results. The 30-minute unit variables were used in all analyses.

Three linear regression models were used to estimate the associations of the activity exposures (sleep duration, prolonged sedentary time, nonprolonged sedentary time, LPA, and MVPA) with waist circumference and BMI: a single effects model, a partition model, and an isotemporal substitution model. Age, socioeconomic status, comorbidity, and smoking status were included as confounders in all models. Education, working status, breast cancer stage at diagnosis, time since diagnosis, breast cancer treatment type, and current hormone therapy use were also considered as potential confounders but were not included in the final models as their inclusion had minimal effect on the results.

A full description of, and rationale behind, the single effects, partition and isotemporal substitution models have been described by others (20, 21). In brief, the single effects model estimates the association between an individual activity and the outcome, without taking into account the other activities (i.e., they are not included in the model), whereas in the partition model, all the activities are mutually adjusted. The variance inflation factors for each of the exposure variables in the partition models for waist circumference and BMI were less than four, suggesting absence of problematic multicollinearity.

The isotemporal model includes “total time” and all of the activities, except the activity of interest. The coefficient from the regression analysis in the isotemporal models for each of the included activities is an estimate of the mean effect on the outcome of reallocating 30 minutes of the omitted activity with 30 minutes of each included activity, while holding time spent in the other activities constant. It is important to note that these activity reallocations generate the predicted change in outcome at the population level rather than at the individual level, and that these are cross-sectional associations rather than causal associations of individuals reallocating time between activities.

We also conducted two sensitivity analyses to investigate the issues of non-wear time and the integration of self-reported sleep duration with objectively measured sedentary and active time. In the first, we limited the isotemporal substitution analyses to only those participants with fewer than 21 hours of “total time” (i.e., sleep, sedentary, and active values that summed to less than 21 hours). Just over one third of the participants had fewer than 21 hours of “total time.” In the second sensitivity analysis, we reran the original isotemporal substitution analyses with only the accelerometer variables (i.e., dropping the sleep variable).

A two-sided P < 0.05 was considered statically significant. Stata 14.1 (StataCorp) was used for all analyses.

The participants were between 36 and 84 years of age and were 2.3 to 3.7 years postdiagnosis (Table 1). Three quarters of the participants had been diagnosed with a stage I or II breast cancer. Mean waist circumference was 92.2 cm, and mean BMI was 26.7 kg/m². Most participants (86%) had a waist circumference greater than 80 cm, indicating an increased risk of metabolic complications (22), and most participants (57.5%) were classified as being overweight or obese based on their BMI.

Prolonged sedentary time was significantly correlated with LPA (Pearson r = −0.75), nonprolonged sedentary time (−0.29), and MVPA (−0.27), whereas LPA and MVPA were also significantly correlated (0.24). All other correlations between the different activities were low (between −0.10 and 0.07) and not statistically significant.

MVPA was the only activity significantly associated with waist circumference in the single effects model, with an increase of 30 minutes of MVPA associated with lower waist circumference [2.22 cm; 95% confidence interval (CI), −4.04 to −0.40 cm; Table 2]. When all activities were mutually adjusted (i.e., the partition model), increasing amounts of sleep, nonprolonged sedentary time, and MVPA were significantly associated with lower waist circumference and lower BMI. Removing prolonged sedentary time (which was significantly correlated with LPA, nonprolonged sedentary time, and MVPA) from the partition models had minimal effect on the results (data not shown).

In the isotemporal substitution models (Table 3; Figs. 1 and 2), reallocating 30 minutes to MVPA was associated with lower waist circumference when allocated from sleep (−2.50 cm; 95% CI, −4.45 to −0.56 cm), prolonged sedentary time (−2.51 cm; 95% CI, −4.38 to 0.64 cm), or LPA (−2.71 cm; 95% CI, −4.72 to −0.69cm). A similar pattern was observed for BMI, although the only statistically significant difference was for replacing prolonged sedentary time with MVPA (−0.93 kg/m²; 95% CI, −1.75 to −0.10 kg/m²). Reallocating 30 minutes of prolonged sedentary time to nonprolonged sedentary time was significantly associated with lower waist circumference (−0.94 cm; 95% CI, −1.79 to −0.10 cm) and lower BMI (−0.41 kg/m²; 95% CI, −0.79 to −0.03 kg/m²). Interchanging 30 minutes of sleep, prolonged sedentary time or LPA with each other were not significantly associated with waist circumference or BMI differences.

The results observed in both of the sensitivity analyses were similar to those from the original analyses (Supplementary Tables S1 and S2).

In this study, we found that reallocating 30 minutes of sleep, prolonged sedentary time, or LPA to 30 minutes of MVPA was associated with significantly lower waist circumference and/or BMI in breast cancer survivors. Reallocating 30 minutes of prolonged sedentary time to 30 minutes of nonprolonged sedentary time was also associated with significantly lower waist circumference and BMI.

Increasing MVPA was most strongly associated with body composition in this study, which is consistent with randomized controlled trials of physical activity interventions conducted in breast cancer survivors (3, 4) and reinforces the important role that MVPA plays in breast cancer survivorship. Higher levels of MVPA have also been associated with improved prognosis, higher quality of life, and lower fatigue (23). We did not observe any evidence that allocating time to sleep from sedentary or active time was associated with lower waist circumference or BMI in this study. However, we did find that longer sleep duration was associated with lower waist circumference and lower BMI in the partition models. This is consistent with previous research in the general population, which indicates short sleep duration (i.e., fewer than 6 hours) is associated with higher BMI (24). Although previous studies have found that sleep duration and/or sleep quality are associated with health-related quality of life, fatigue, depression, and anxiety in breast cancer survivors (10, 25, 26), we are not aware of any previous studies that have investigated the association between sleep duration and body composition in this population.

Longitudinal studies conducted among females in the general population have found that higher levels of sedentary behavior were associated with higher BMI and/or waist circumference (27, 28), and the only previous (cross-sectional) study to be conducted among breast cancer survivors reported a similar finding, although the association attenuated when adjusted for MVPA (8). Our study provides more context to these findings; in keeping with previous studies conducted in the general population (11, 12, 29–31), we found that the way sedentary time is accumulated may impact its association with body composition. Previous studies in blue collar workers (32, 33), older adults (34), and the general population (35) have all found that spending greater time in shorter bouts of sedentary time is associated with either lower or no change in waist circumference and/or BMI, while also finding that greater time in longer bouts of sedentary time is associated with higher waist circumference and/or BMI (32–35). Our results are also similar to those observed in a recent study of overweight and obese people with type II diabetes, which found that replacing prolonged sedentary time with an equivalent amount of nonprolonged sedentary time was associated with lower waist circumference and BMI (36). These results lend some support to the notion that decreasing time spent in prolonged, unbroken sedentary bouts (by taking breaks from prolonged periods of sitting) may be an avenue for improving body composition in clinical populations.

Strengths of this study include objective measurement of sedentary time and physical activity, assessment of the full 24-hour day (i.e., sleep, sedentary, and active time), and use of a statistical technique that allowed these activities to be investigated simultaneously. Although allocating time to MVPA is likely to be the most potent behaviour for improving a wide range of health outcomes (20), there is clearly a limit to how much time can realistically be spent doing MVPA. It is therefore important to examine the associations between other activities that take up larger portions of the day. The isotemporal methods used in this study allow us to examine these associations, as well as how they may change depending on the activity they replace. However, it has recently been argued that a compositional data analysis approach may be better than isotemporal methods for understanding the effects of reallocating time between sleep, sedentary, and active behaviours (37).

A limitation of this study is the sample size, which may have resulted in low power to detect modest associations. As such, although we did not observe any significant associations for longer sleep duration or higher levels of LPA in the isotemporal models, it is not possible to rule out a beneficial effect for these behaviors. It is also important to note that body composition is just one outcome, and that reallocating time from sitting to sleep and/or LPA has been associated with obesity and metabolic outcomes in the general population (20, 37, 38). Examining the interdependent associations between sleep, sedentary, and active behaviors with other metabolic outcomes in breast cancer survivors may be an avenue for future research. Other limitations include the use of self-reported sleep duration, BMI, and waist circumference, which may have resulted in some measurement error. Finally, the use of a waist-worn accelerometer means the sedentary time estimates are likely to have included periods of stationary standing, and, as the cut-off points used to differentiate MVPA from LPA in this study were calibrated to measure ambulatory activities, such as walking and running (14), it is also possible that some less ambulatory moderate-to-vigorous activities were misclassified as LPA.

In conclusion, we found that replacing 30 minutes of prolonged, unbroken sedentary time with 30 minutes of nonprolonged sedentary time or 30 minutes of MVPA was associated with significantly lower waist circumference and BMI in breast cancer survivors. These results reinforce the important role that MVPA plays in breast cancer survivorship and provide some evidence that decreasing time spent in prolonged, unbroken sedentary bouts may be an avenue for improving body composition in this population.

No potential conflicts of interest were disclosed.

The funding agencies had no involvement in any aspect of the study.

Conception and design: T. Boyle, J.K. Vallance, B.M. Lynch

Development of methodology: T. Boyle, B.M. Lynch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Boyle

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Boyle, J.K. Vallance, M.P. Buman

Writing, review, and/or revision of the manuscript: T. Boyle, J.K. Vallance, M.P. Buman, B.M. Lynch

Study supervision: T. Boyle

We would like to acknowledge Emily Ransom for her assistance in data collection and study management and Jessica Occleston for her assistance in data processing. We also sincerely thank the people who took the time to participate in this study.

This work was supported by Breast Cancer Research Centre - Western Australia (grant to T. Boyle, J.K. Vallance, and B.M. Lynch), Cancer Council Western Australia (Award to T. Boyle); the Australian National Health and Medical Research Council (Fellowship #1072266 to T. Boyle); the Canadian Institutes of Health Research (Fellowship #300068 to T. Boyle); the Michael Smith Foundation for Health Research (Trainee Award #5553 to T. Boyle); Killam Trusts (Postdoctoral Research Fellowship to T. Boyle); the Canada Research Chairs (to J.K. Vallance); Alberta Innovates – Health Solutions (Population Health Investigator Award to J.K. Vallance); NCI (grant #1R01CA198971 to M.P. Buman); and the Australian National Breast Cancer Foundation (Fellowship to B.M. Lynch).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.