Device-measured physical activity, adiposity and mortality: a harmonised meta-analysis of eight prospective cohort studies

Abstract

Background: The joint associations of total and intensity-specific physical activity with obesity in relation to all-cause mortality risk are unclear.

Methods: We included 34 492 adults (72% women, median age 62.1 years, 2034 deaths during follow-up) in a harmonised meta-analysis of eight population-based prospective cohort studies with mean follow-up ranging from 6.0 to 14.5 years. Standard body mass index categories were cross-classified with sample tertiles of device-measured total, light-to-vigorous and moderate-to-vigorous physical activity and sedentary time. In five cohorts with waist circumference available, high and low waist circumference was combined with tertiles of moderate-to-vigorous physical activity.

Results: There was an inverse dose-response relationship between higher levels of total and intensity-specific physical activity and mortality risk in those who were normal weight and overweight. In individuals with obesity, the inverse dose-response relationship was only observed for total physical activity. Similarly, lower levels of sedentary time were associated with lower mortality risk in normal weight and overweight individuals but there was no association between sedentary time and risk of mortality in those who were obese. Compared with the obese-low total physical activity reference, the HRs were 0.59 (95% CI 0.44 to 0.79) for normal weight-high total activity and 0.67 (95% CI 0.48 to 0.94) for obese-high total activity. In contrast, normal weight-low total physical activity was associated with a higher risk of mortality compared with the obese-low total physical activity reference (1.28; 95% CI 0.99 to 1.67).

Conclusions: Higher levels of physical activity were associated with lower risk of mortality irrespective of weight status. Compared with obesity-low physical activity, there was no survival benefit of being normal weight if physical activity levels were low.

Keywords: body mass index; exercise; observational study.

Figure 1

Joint associations of physical activity or sedentary time and body mass index (BMI) with risk of all-cause mortality. n=34 492, deaths=2034. Individuals self-reporting prevalent cardiovascular disease (CVD) or cancer are excluded. Analysis of sedentary time includes n=33 552; 1920 deaths as there were no cases in the reference category in the British Regional Heart Study. Models are adjusted for age, sex (when applicable), socioeconomic status, smoking and the covariates included in each study’s published final multivariable-adjusted model (see online supplemental eTable 1 for details). Additionally, models using moderate-to-vigorous physical activity were adjusted for sedentary time (continuous form) with sedentary time and light physical activity adjusted for moderate-to-vigorous physical activity (continuous). Median activity and sedentary time in exposure categories are shown in online supplemental eTable 2. PA, physical activity.

Figure 2

Joint associations of moderate-to-vigorous physical activity (MVPA) and waist circumference with risk of all-cause mortality. Top panel: excluding individuals with self-reporting prevalent cardiovascular disease (CVD) or cancer (n=17 773; 1443 deaths). Bottom panel: as top panel but further restricted to never smokers (n=9136; 608 deaths). Both models are adjusted for age, sex (when applicable), socioeconomic status, smoking (top panel only), sedentary time (continuous) and the covariates included in each study’s published final multivariable-adjusted model (see online supplemental eTable 1 for details).

Figure 3

Joint associations of physical activity or sedentary time and body mass index (BMI) with risk of all-cause mortality in never smokers. n=17 485, deaths=864. Analysis of sedentary time includes n=17 072; 820 deaths as there were no cases in the reference category in the British Regional Heart Study. Individuals self-reporting prevalent cardiovascular disease (CVD) or cancer are excluded. Models are adjusted for age, sex (when applicable), socioeconomic status and the covariates included in each study’s published final multivariable-adjusted model (see online supplemental eTable 1 for details). Additionally, models using moderate-to-vigorous physical activity were adjusted for sedentary time (continuous form) with sedentary time and light physical activity adjusted for moderate-to-vigorous physical activity (continuous). PA, physical activity.