Figure 1

Tenth survival percentile differences, differences in months by which ten percent of the cohort has died as function of total physical activity. Data were fitted using a Laplace regression model with restricted cubic splines with 3 knots (36, 41 and 48 MET-hrs/day) of the distribution of total physical activity. The estimates were adjusted for baseline age, body mass index, alcohol consumption, smoking status, and educational level. Dashed lines represent 95% confidence limits. The reference value of total physical activity is the median of the bottom quartile (36.5 MET-hrs/day). Tick marks represent position of men who died. The histogram is the distribution of total physical activity in the cohort.