Objective To examine whether the associations between sedentary behaviours (ie, daily sitting/TV-viewing time) and mortality from cardiovascular disease (CVD) and cancer differ by different levels of physical activity (PA).

Design Harmonised meta-analysis of prospective cohort studies. Data on exposure variables were harmonised according to a predefined protocol and categorised into four groups for sedentary behaviours and into quartiles of PA (MET-hour/week).

Data sources PubMed, PsycINFO, Embase, Web of Science, Sport Discus and Scopus.

Eligibility criteria for selecting studies Individual level data on both sedentary behaviours and PA and reported effect estimates for CVD or cancer mortality.

Results Nine studies (n=850 060; deaths=25 730) and eight studies (n=777 696; deaths=30 851) provided data on sitting time and CVD and cancer mortality, respectively. Five studies had data on TV-viewing time and CVD (n=458 127; deaths=13 230) and cancer (n=458 091; deaths=16 430) mortality. A dose–response association between sitting time (9%–32% higher risk; p for trend <0.001) and TV time (3%–59% higher risk; p for trend <0.001) with CVD mortality was observed in the ‘inactive’, lowest quartile of PA. Associations were less consistent in the second and third quartiles of PA, and there was no increased risk for CVD mortality with increasing sedentary behaviours in the most active quartile. Associations between sedentary behaviours and cancer mortality were generally weaker; 6%–21% higher risk with longer sitting time observed only in the lowest quartile of PA.

Conclusion PA modifies the associations between sedentary behaviours and CVD and cancer mortality. These findings emphasise the importance of higher volumes of moderate and vigorous activity to reduce, or even eliminate these risks, especially for those who sit a lot in their daily lives.