Never smoking, high levels of physical activity, and normal weight at age 50 and over appears to be not only associated with increases in total life expectancy, but also with a greater number of years lived free of CVD. The effects on total life expectancy (both sexes) and years lived free of CVD (men) appeared greatest for never smoking; although for women the increase in years free of CVD differed less between the three health hearty behaviors. Normal weight was the only behavioral factor in our study found to also reduce the time lived with CVD.
Never smoking (men), high levels of physical activity, and normal weight were each associated with lower risks of developing CVD as compared to low physical activity, current smoking and obesity, respectively. Never smoking and high levels of physical activity also reduced the risks of dying in those with and without a history of CVD, but normal weight did not. The greater number of years lived free of CVD among those with high levels of physical activity or never smoking is due to their lower incidence of CVD, in combination with lower mortality without a history of CVD. The greater number of years free of CVD and reduction in time lived with CVD among those with a normal weight is due to lower incidence of CVD, not being accompanied with a reduction in mortality.
The HRs we found support existing evidence that non-smoking, physical activity and normal weights reduces incidence of CVD [2, 4, 8]. The reductions in mortality among persons without cardiovascular disease associated with never smoking and physical activity, are also in line with documented effects for a large set of fatal diseases, including respiratory diseases and several cancers [2, 4]. Our results also confirm that non-smoking and physical activity reduces mortality in persons with a history of CVD [25, 26]. The effects of BMI on mortality are more disputed, and appear to depend on the age considered . The lack of a harmful effect of overweight and obesity on mortality in elderly persons with CVD, is in agreement with the study of Kalanar-Zadey. It is noteworthy that our findings on mortality in persons with and without CVD are not in contrast with existing evidence that obesity increases overall mortality. In fact, our findings suggests that higher total mortality among obese persons is mainly due to higher rates of onset of CVD, which, in combination with higher mortality risks among persons with CVD, results in higher total mortality, and not due to worse survival.
Our results show smaller differences in total life expectancy at age 50 associated with BMI compared to prior analyses also based on the Framingham Heart Study [14, 29] Additional analyses comparing the studies in more detail (data available on request) suggested that the majority of the difference was explained by the fact that whilst the current analysis assessed the effect of BMI in a population age 50 and over and updated BMI information every 12 years, the prior analyses assessed BMI only at age 40-50, and followed this cohort without updating this information to more recent years and ages. The effects of BMI on mortality have been shown to be smaller in elderly populations . We found also smaller differences associated with smoking as compared to prior work based on the Framingham Heart Study , which compared never and always smokers, instead of never, current and former smokers in the present study.
Some limitations of our study must be considered. This is a life table analysis based on a prospective observational study where no intervention was performed. Therefore it has the inherent weaknesses of all cohort studies and lacks the strength of causality that randomized trials could offer. Reverse causation, which means that lower physical activity levels, BMI levels, and quitting smoking are caused by ill health and not the other way around, is an important issue to consider since it could introduce bias in the evaluation of the effect of these health behaviors. While owing to the longitudinal design we fulfill the temporality criteria of Bradford and Hill  that the effect (CVD onset or death) has to occur after the cause (unhealthy behavior), we can not rule out that the health behavior is affected by the health of the individual. Different approaches exist to reduce the effect of reverse causation but there is no method to eliminate it completely. To correct for reverse causation we adjusted our analyses for co-morbidities at baseline instead of excluding the subjects with disease at the start of follow-up, since we are interested in the effect of health behaviors in the general elderly population and not in selected healthy populations. Finally we examined whether excluding the first two years of follow up did reduce the HRs, but found minimal effect (data not shown).
Another limitation is that during the Framingham Heart Study, physical activity and smoking were evaluated by self-report, which might be affected by recall bias. In contrast height and weight were measured in the study. Differences in measurement of these behaviors would hamper our comparison if the resulting misclassification affects the transitions differently (e.g. affect transition from cardiovascular disease to death, but not from non-CVD to cardiovascular disease, or death). However, misclassification, if present, is likely to be non-differential and to affect each transition similarly.
Participants included in the FHS were mainly white; therefore, our results may not apply to other ethnic groups. Furthermore, the historical character of the FHS, limits the extrapolation of findings to today's populations. Nonetheless by focusing on those over age 50, and updating BMI, smoking and physical activity information every 12 years, we have automatically put more weight on contemporary data (as older age in the cohort means more recent years). Life expectancy at age 50 in our study is very similar to the life expectancy at age 50 in the US in 2000-2004 (life expectancy at age 50: men: FHS: 27.3, US 28.4; women: 33.7 and 32.2 years respectively). Also, our study focuses on comparisons between different risk factor categories (e.g. normal weight vs. obesity).
The key strengths of our study include using a well-organized historic cohort that has been followed biannually for decades (we used 36 years), and provides upgraded information on health behaviours, covariates and outcomes. Further, this is the first comparative life table analysis of the effects of the main heart-healthy behaviours on life expectancy and life expectancy with and without CVD.