Association Between Various Sedentary Behaviours and All-Cause, Cardiovascular Disease and Cancer Mortality

The Multiethnic Cohort Study

Yeonju Kim; Lynne R Wilkens; Song-Yi Park; Marc T Goodman; Kristine R Monroe; Laurence N Kolonel

Disclosures

Int J Epidemiol. 2013;42(4):1040-1056. 

In This Article

Methods

Study Population

The design and characteristics of the Multiethnic Cohort Study (MEC) have been described in detail elsewhere.[17,18] In brief, the MEC was established between 1993 and 1996 in Hawaii and California, USA, to examine the association of lifestyle factors with the risk of cancer and other chronic diseases in diverse populations. More than 215 000 men and women, aged 45–75 years, comprising mainly five racial/ethnic groups (African American, Latino, Japanese American, Native Hawaiian and White) were enrolled by completing a 26-page mailed questionnaire. The study was approved by the institutional review boards of the University of Hawaii and the University of Southern California.

In the current analyses, we excluded 13 989 participants who did not self-identify as one of the five major racial/ethnic groups, and 8263 participants with implausible dietary data based on total energy intake or its components. We further excluded subjects with missing data on hours spent sleeping (n = 6308), spent in any physical activity (n = 10 461) or spent in two or more of the sitting activities (n = 3578). We also excluded subjects with missing data on height or weight (n = 2318) or smoking history (n = 5149). Additionally, men and women who reported a personal history of cancer, heart attack or stroke at baseline (n = 30 590), and participants who died within the 1st year after cohort entry (n = 530) were excluded. Finally, a total of 134 596 men and women were included in this analysis.

The excluded subjects were very similar to the analytical sample on median follow-up time, body mass index (BMI), energy intake, hours spent sleeping per day and prevalence of smoking. On the other hand, the excluded subjects were older (by 4.9 years on average), consumed somewhat more fruits and vegetables, drank an average of 2.0 fewer grams of alcohol per day, were more likely to have a history of hypertension and/or diabetes and were less likely to have a college education. These factors are adjusted for in our analysis.

Ascertainment of Mortality

Deaths among MEC participants were identified through linkage to death certificate files in Hawaii and California, which were augmented through periodic linkages with the National Death Index in the USA. Causes of death were classified according to the International Classification of Diseases, Ninth Revision (ICD-9) and Tenth Revision (ICD-10). Specific causes of death were grouped into three categories: cardiovascular disease (ICD-9 codes 390–434, 436–448; ICD-10 codes I00–I78), cancer (ICD-9 codes 140–208; ICD-10 codes C00–C97), and all other causes. Up to 31 December 2007, a total of 19 143 deaths were identified during an average 13.7 years of follow-up.

Measures of Time Spent in Sitting and Physical Activity

In our baseline questionnaire, we ascertained sedentary activities (sleeping and five types of sitting) on a 24-h basis, and three categories of physical activity on a weekly basis. The full questionnaire can be viewed at (http://www.crch.org/multiethniccohort/index.htm). Sitting activities included: 'sitting in a car or bus', 'sitting at work', 'sitting at meals', 'sitting watching television' and 'other leisure sitting activities (such as reading, playing cards, sewing)'. Each sitting activity was asked in seven categories: 'never', '<1 hour', '1–2 hours', '3–4 hours', '5–6 hours', '7–10 hours', and '11 hours or more' per day. Total daily sitting was the sum of the midpoints of the specific sitting categories, using 0 for 'never', 0.5 for '<1 hour', 1.5 for '1–2 hours', 3.5 for '3–4 hours', 5.5 for '5–6 hours', 8.5 for '7–10 hours' and 11 for '11 hours or more'. The analysis was also performed using the Pareto Curve[19] to assign the midpoint for the open-ended category of 11 h or more, to account for differences in patterns by sitting activities. Results were unchanged and are not shown.

Vigorous and moderate physical activities included: 'strenuous sports (such as jogging, bicycling on hills, tennis, racquetball, swimming laps, aerobics)', 'vigorous work (such as moving heavy furniture, loading or unloading trucks, shovelling, weight lifting or equivalent manual labour)' and 'moderate activity (such as housework, brisk walking, golfing, bowling, bicycling on level ground, gardening)'. Each type of physical activity was asked in eight categories: 'never', '30 minutes–1 hour', '2–3 hours', '4–6 hours', '7–10 hours', '11–20 hours', '21–30 hours' and '31 hours or more' per week. Light physical activity was estimated by subtracting the total time spent in all activities (sitting, physical activity and sleeping) from 24 h. The metabolic equivalents (METs) for physical activity per week were computed using the following formula:[20] [(Number of hours in moderate activities × 4.0) + (Number of hours in vigorous work and in strenuous sports × 7.2)]. In a validation study, the correlation for energy expenditure based on the MEC activity questionnaire and the gold standard of doubly-labelled water was reasonable (r = 0.31)[21] and comparable to those reported in the literature using other instruments.[22,23]

Statistical Analysis

Cox proportional hazard models, with age as the time metric, were used separately for men and women. Observation started at cohort entry, and exit time was defined as the date of death or at the end of follow-up (31 December 2007), whichever occurred earlier. Sitting variables were parameterized as indicator variables representing three categories of duration. Trend tests were conducted by inclusion of a continuous variable in the model assigned the median value for the appropriate category of sitting. Because smoking is a strong risk factor for early death, we carefully adjusted for this variable using a complex time-dependent model previously developed for a study of tobacco use and lung cancer incidence.[24]

Multiple imputation was used for hours of sitting activities when only one type was missing, so that total sitting time could be computed for these individuals. The percentage missing data for sitting behaviours was: 1.6% for sitting watching TV; 1.6% for sitting in other leisure activities; 2.1% for sitting in a car or bus; 6.6% for sitting at work; and 0.3% for sitting at meals. The missing values were imputed by the Markov Chain Monte Carlo method based on education, age, ethnicity and smoking status and intensity. Five imputation data sets were created and the results were aggregated across data sets using standard multiple imputation methods.[25] All hazard ratios presented in this report were derived from the multiple imputation method.

We began with a minimally adjusted model consisting of race/ethnicity, age at cohort entry (5-year age groups) to further account for any cohort effects, and educational level (less than college, and college graduate or higher) as strata variables, in addition to the adjustment for smoking. In the fully adjusted model, we made additional adjustment for the following potential confounders: prevalent diabetes and/or hypertension at baseline (yes or no), energy intake divided at the median (2184 Kcal/day for men and 1747 Kcal/day for women), alcohol intake divided at the median (2.8 g/day for men and 0 g/day for women) and physical activity (METs/week by sex-specific quartiles). Fully adjusted models for duration of sitting by type also mutually adjusted for hours spent in all other sitting behaviours as trend variables. Global and pairwise comparisons from competing risk models were used to examine whether the hazard ratios associated with sitting were similar across causes of death.[26] Sensitivity analysis was performed by excluding men and women who died within 5 years from the initial date of follow-up. We also ran the analysis for individual sitting activities, including all cohort members with data, to determine the influence of the exclusions which produced similar results (data not shown).

We further stratified the analysis of sitting and mortality by several factors: age, racial/ethnic group, education level, diabetes and/or hypertension, smoking status, BMI, vigorous and moderate physical activity, light physical activity, fruit and vegetable intake and sleep duration. As in other similar studies,[5,9] BMI, which may be a mediator of the effect of prolonged sitting on mortality, was not included as an adjustment variable. Because detailed current employment status was not asked for in the baseline questionnaire, to minimize the effect of classification bias of employment status, we restricted the' sitting at work' analyses to subjects who reported at least some time spent sitting at work, which included 45 115 men and 51 180 women.

The effects on mortality for different types of sitting activities were compared in a model including total hours of sitting per day, as well as the proportion of total sitting contributed by each of five sitting activities. This model is called the differential effects model in this paper. A global Wald test was used to compare the parameters for proportion in each sitting activity, allowing us to compare different sitting activities, adjusting for total sitting time. The joint effects of different sitting activities were then examined in pairwise fashion. Interaction was assessed by a Wald test of the cross-product terms of the proportions contributed by the two sitting activities being considered, grouped into three categories. This avoided the detection of an interaction between sitting activities due solely to the increase in total sitting time.

Analyses were conducted using SAS version 9.2 (SAS Institute, Cary, NC, USA).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....