Background Studies of mid-aged adults provide evidence of a relationship between sitting-time and all-cause mortality, but evidence in older adults is limited. The aim is to examine the relationship between total sitting-time and all-cause mortality in older women.
Methods The prospective cohort design involved 6656 participants in the Australian Longitudinal Study on Women's Health who were followed for up to 9 years (2002, age 76–81, to 2011, age 85–90). Self-reported total sitting-time was linked to all-cause mortality data from the National Death Index from 2002 to 2011. Cox proportional hazard models were used to examine the relationship between sitting-time and all-cause mortality, with adjustment for potential sociodemographic, behavioural and health confounders.
Results There were 2003 (30.1%) deaths during a median follow-up of 6 years. Compared with participants who sat <4 h/day, those who sat 8–11 h/day had a 1.45 times higher risk of death and those who sat ≥11 h/day had a 1.65 times higher risk of death. These risks remained after adding sociodemographic and behavioural covariates, but were attenuated after adjustment for health covariates. A significant interaction (p=0.02) was found between sitting-time and physical activity (PA), with increased mortality risk for prolonged sitting only among participants not meeting PA guidelines (HR for sitting ≥8 h/day: 1.31, 95% CI 1.07 to 1.61); HR for sitting ≥11 h/day: 1.47, CI 1.15 to 1.93).
Conclusions Prolonged sitting-time was positively associated with all-cause mortality. Women who reported sitting for more than 8 h/day and did not meet PA guidelines had an increased risk of dying within the next 9 years.
Statistics from Altmetric.com
Based on 50 years of evidence, recent systematic reviews have confirmed an inverse dose–response relationship between physical activity (PA) and all-cause mortality in men and women, after adjustments for demographic and behavioural risk factors.1 ,2 In the last 10 years, behavioural epidemiologists have started to focus on the health risks of sedentary behaviours, which can occur independent of PA, since people can meet PA guidelines and yet spend most of their day sitting.3 ,4
Several cohort studies have shown that total sitting-time, as well as sitting in specific domains, such as watching TV or ‘screen time’, are associated with all-cause mortality, after adjusting for PA.5–11 However, most of these studies included mid-age adults; only one has provided results for participants aged 75 and over,9 and none has focussed specifically on older people. This is important because older people tend to spend more time sitting than young and mid-age adults,12 due to factors such as mobility limitations and fear of falling.13 ,14
The aim of this study was therefore to examine the relationship between total sitting-time and all-cause mortality in a population of older women who were 76–81 years old in 2002, followed for a period of up to 9 years.
The Australian Longitudinal Study on Women's Health is a prospective study of factors shaping the health and well-being of three cohorts of Australian women, born in 1973–1978, 1946–1951 and 1921–1926, recruited from the national Medicare health insurance database,15 with the focus of this paper on the 1921–1926 cohort. Participants completed mailed surveys in 1996, 1999, 2002, 2005, 2008 and 2011. At baseline, the women were largely representative of women aged 70–75 years in the Australian population, and analyses have shown that attrition has not resulted in bias for studies assessing mortality.15 ,16 More details can be found at http://www.alswh.org.au. The study was approved by the Universities of Newcastle and Queensland Ethic Committees and all participating women provided informed consent. As sitting-time was only assessed in the 2002 survey, this survey was taken as baseline for the current study. A total of 12 432 participants completed the first survey in 1996. In 2002, surveys were sent to 10 434 participants. Of these, 8647 participants (83%) returned the survey, 435 did not, 265 were too frail, 517 withdrew and 570 were deceased. Of those who returned the survey, the 6656 that provided complete data on sitting time formed the analysis sample for this survey.
For all variables, data were retrieved from the 2002 survey, except where indicated.
All-cause mortality was determined by matching participants’ identifying information to the Australian National Death Index to determine mortality status. Survival time was calculated as the number of months between the date of return of the 2002 survey and date of death or the censor point (7 December 2011).
Main explanatory variable
Sitting time was assessed by the question “Think about all the time you spend sitting EACH DAY while at home, at work, while getting from place to place or during your spare time. How many hours EACH DAY do you typically spend sitting down while doing things like visiting friends, driving, reading, watching television, or working at a desk or computer on (a) a usual week-day and (b) a usual weekend-day”.
This question is similar to that included in the International PA Questionnaire, which has been shown to have good reliability (correlation coefficient 0.77 for women) and moderate criterion validity (correlation coefficient 0.44 for women) against accelerometers (<100 counts/min).17 Sitting-time data were cleaned using protocols developed by van Uffelen et al,18 and categorised as 0–4, 4 to <8, 8 to <11 and ≥11 h/day, in line with previous studies.9
Other explanatory variables
Sociodemographic variables included age, area of residence, highest level of education and marital status categorised as shown in table 1.
Behavioural variables included smoking status, alcohol consumption and PA. PA scores were calculated from questions about time spent in walking, moderate and vigorous activity, using established protocols.19 (Score=walking+moderate min/week×3 MET)+(vigorous min/week×6 MET), with scores categorised as: not meeting (<450 MET×min/week) or meeting PA guidelines (≥450 MET×min/week). The PA measure has been shown to have acceptable measurement characteristics.20 ,21 Smoking status was categorised as ‘never smoked’, ‘ex-smoker’ or ‘current smoker’. Alcohol consumption was categorised as ‘never drink’, ‘rarely drink’, ‘less than once a week’ or ‘weekly’.
Health-related variables included: number of chronic conditions, categorised as none, one, two, three or more, from a list of 15 conditions reported to have been diagnosed by a doctor in the previous 3 years; self-reported health (assessed by the question ‘In general, would you say your health is…’ with responses ranging from ‘excellent’ to ‘poor’ on a five-point scale); receiving help with daily tasks for long-term illness or disability (assessed by the question, ‘Do you regularly need help with daily tasks because of long-term illness, disability or frailty?’ (yes/no); and body mass index (BMI; kg/m2); calculated using self-reported weight and height and categorised as underweight, BMI<18.5; normal weight, 18.5≤BMI<25; overweight, 25≤BMI<30; or obese, BMI≥30.22
The nature of the association between sitting-time and mortality risk was investigated using Cox proportional hazards models. A four-knot restricted cubic spline was first used to plot continuous sitting-time against unadjusted HRs for mortality. HRs were then calculated for each category of sitting, with <4 h/day as the reference category. Trends were calculated using the HRs for each category of sitting as a continuous variable. The unadjusted models were adjusted first for (1) sociodemographic variables (age, education, marital status and area of residence); with subsequent addition of (2) smoking and alcohol; (3) BMI; (4) PA; (5) and health-related variables (number of chronic conditions, self-rated health, receiving help with daily tasks for long-term illness or disability). These health-related variables, measured at baseline, were included to account for potential reverse causation. Selection of covariates was based on previous studies to enhance comparability of results.9 ,10 A product term (sitting×PA) was added to assess the potential interaction effect between sitting-time and PA categories. Sensitivity analyses omitting deaths within the first 2 years were conducted, to account for deaths due to underlying disease. Further, participants with poor self-rated health at baseline were excluded. All statistical analyses were conducted in SPSS V.20. p Values were based on two-sided tests and were considered statistically significant at p<0.05.
Of the 8647 women who returned the survey, 6656 provided sitting-time data. Of these 6656 women, 1903 had missing values for at least one of the covariates and were excluded from the adjusted analyses. There was no difference in sitting-time between those with and without missing values on any of the covariates (p=0.10). The proportion of missing data ranged from zero for age to 11% for BMI.
The number of deaths during the 9 years of follow-up was 2003; 254 women died within the first 2 years. Median follow-up time was 72.3 months.
Sociodemographic and behavioural data for the analysis sample are shown in table 1. The mean age of the women in 2002 was 78.2 (SD 1.45, range 74.3–82.2) years. More than half reported sitting for between 4 and 8 h/day (55%), and <5% for more than 11 h/day. Approximately, 38% met the current PA guidelines (interpreted as ≥450 MET×min/week), but only 4.4% were current smokers, and 60% never or rarely drank alcohol. Almost 90% had at least one chronic condition, but one-third rated their health as excellent or very good.
Sitting-time was significantly associated with all-cause mortality. Figure 1 shows a flat relationship at lower sitting times, with risk increasing from 7 to ≥11 h/day. Compared with participants who sat <4 h/day, those who sat 8–11 h/day had a 1.45 times higher risk of death (95% CI 1.26 to 1.66), and those who sat ≥11 h/day had a 1.65 times higher risk of death (95% CI 1.37 to 2.00) (HR for trend=1.06, 95% CI 1.04 to 1.07; table 2). Adding the demographic covariates and smoking/alcohol resulted in a slightly higher HR for the highest sitting category (HR 1.83, 95% CI 1.46 to 2.30), but sequential additional adjustment for BMI and PA did not markedly change the HRs from those in the unadjusted model. The HRs were however attenuated when the final adjustment for health variables (chronic conditions, self-rated health and assistance with daily tasks) was made, with the association between sitting ≥11 h/day and mortality no longer significant. Results from the sensitivity analyses in which deaths in the first 2 years were omitted, were similar to those of the original analyses; however, HRs were slightly attenuated (Web only file table 1). Excluding participants with poor self-rated health at baseline did not change the results.
The interaction between sitting-time and PA was significant (p=0.02). The data in table 3 show that mortality risk was significantly increased among participants who reported sitting for ≥8 (adjusted HR 1.31, 95% CI 1.07 to 1.61) or ≥11 h/day (adjusted HR 1.47, 95% CI 1.15 to 1.93) and not meeting PA guidelines. The increased risk of sitting was however no longer statistically significant when PA guidelines were met (see table 3).
The aim of this study was to assess the 9-year mortality risk associated with levels of daily sitting-time in a population of older Australian women. The results suggest prolonged sitting-time (>8 h/day) is associated with an increased risk of all-cause mortality in this population.
Our results are consistent with those of six recent studies that have assessed relationships between both screen/TV viewing time, and total sitting-time with all-cause mortality, in people of different ages, and where data are reported for women, in Canada, Australia, the UK and the USA.5–10 Data for this age group are sparse for both sitting-time and PA. van der Ploeg et al provide some comparability where the percentages in the sitting-time categories are similar for women and over 75s, respectively. Very few (if any) studies have reported PA data for Australian women of this age.
The data in figure 1 suggest that there may be a threshold of between 7 and 9 h of sitting per day, above which sitting is associated with an increased risk of all-cause mortality. Two other recent studies have reported similar findings. For example, earlier this year, van der Ploeg et al9 suggested a possible threshold of ≥7 h/day, and Matthews et al10 suggested a threshold of ≥9 h/day, for increased risk of all-cause mortality in populations aged 45 and older. One other study, from Japan, has also reported increased mortality risk in those who reported >8 h of ‘sedentary activities’, but the relationship was only observed in men.23 Differences in patterns of behaviours and/or differences in body fat levels, may explain the gender difference, as well as the contrasting findings in these studies of Western and Asian women.
There is debate about whether the reported associations between sitting and mortality are affected by confounding variables, although relatively consistent associations have remained in previous studies after adjustment for multiple confounders. To examine this issue, we carefully adjusted for a wide range of potential confounders. Adjustments for sociodemographic and behavioural variables, including both categorical PA and BMI, did not markedly affect the results. However, when the results were additionally adjusted for health-related variables, the increased risk of mortality was markedly reduced in participants who reported sitting ≥8 h/day. While most previous researchers have adjusted their analyses for chronic conditions and BMI, only one prior study included adjustment for self-rated health and help with daily tasks (but not for chronic conditions).9 In that study, the associations between sitting and mortality remained consistent after these adjustments, but the sample included a wider range of ages, and younger participants than in our study. One explanation for the observed marked reduction in HRs in our study after adjustment for the four health-related variables (table 2, model 6), may be reverse causation. Participants with more chronic illnesses, and those who are more frail and less mobile are likely to be the most in need of assistance and therefore spend more time sitting. This is further supported by the attenuated associations which were seen after exclusion of deaths in the first 2 years. In this age group, the most important confounders appear to be health related, and these should be accounted for in future research on sitting-time in older populations.
Despite the fact that the initial results were not greatly affected by adjustment for PA as a categorical variable, when the results were stratified for meeting or not meeting current PA guidelines, a different picture emerged. There was a significant interaction between sitting and PA among participants who did not meet PA guidelines; mortality risk was increased in those who sat for 8 h a day or more. However, for those who reported meeting guidelines, there were no significant associations between sitting and mortality, and for those sitting the most, the risk was reduced. These results may seem spurious given the small number of participants in the ≥11 h/day sitting group. However, further investigation found that the ≥11 h/day sitting group reported about 600 MET×min/week more PA, on average, than the other sitting-time groups (p=0.002). In this instance, the higher volume of activity appears to be of benefit, despite high amounts of sitting-time. Previous studies examining this issue have found that, although there were no statistically significant interactions, the stratified data showed greater risk for those who did not meet the PA guidelines, or for those who only minimally met the guidelines.5 ,9 ,10 The current and previous findings highlight the importance of both reducing sitting-time and maintaining activity levels in old age.
In contrast to recommendations in the UK and the USA, which suggest reducing or breaking-up extended time spent sitting,24 ,25 the current PA guidelines for older populations in Australia do not include any recommendation about reducing sitting time.26 In our view, there is currently insufficient evidence on which to base specific recommendations on the frequency and duration of breaking up sitting-time. However, building on the findings reported here, and as more data on the relationships between sitting and health outcomes emerge, it should soon be possible to confidently make broad recommendations about the amount of sitting that is detrimental for health, or about the amount of PA that might be required to offset the risks of sitting in older people.27
The strengths of this study include the large population based sample, long-term follow-up, national indexed identification of deaths and inclusion of many confounding variables. However, the limitations of self-reported measures of behaviour, such as those used in this study, are acknowledged. As is the case with PA, more accurate measures of sitting-time might produce more robust risk estimates. Indeed, using accelerometer data from the National Health and Nutrition Examination Survey, Koster et al28 have recently shown a 3.3 times greater risk of mortality in the most sedentary, compared with the least sedentary in that US study.28
In conclusion, we found that prolonged sitting-time was associated with all-cause mortality. Women who reported sitting for more than 8 h/day and did not meet PA guidelines had an increased risk of dying within the next 9 years.
What this study adds
Higher levels of daily sitting were positively associated with all-cause mortality in older women who did not meet physical activity guidelines, after adjustment for many potential confounding variables.
In this age group, health-related variables appear to be the most important confounders and should be accounted for in future sitting-time research in older populations.
The Australian Longitudinal Study on Women's Health (ALSWH), which was conceived and developed by a group of interdisciplinary researchers at the Universities of Newcastle and Queensland, is funded by the Australian Government Department of Health and Ageing. The funding source had no involvement in the research presented in this paper.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online table
Contributors All authors were involved in the planning, development and review of the manuscript. TP analysed the data, drafted the manuscript and is responsible for the overall content. GP and WB provided statistical advice and feedback on early drafts, and edited the final manuscript.
Funding TGP and GP were supported by an NHMRC program grant (#301200) at the University of Queensland, School of Human Movement Studies.
Competing interests None.
Ethics approval Universities of Newcastle and Queensland.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.