Objectives:
Smoking exposure can be measured in a variety of ways including duration, intensity (cigarettes /day), and the product of these known as pack-years. How these variables should be combined to best model smoking as an exposure in cardiovascular (CVD) studies is not clear.
Methods:
The MESA study is a cohort of 6814 participants initially free of clinical heart disease at baseline in 2000-2002. MESA collected questionnaire data on smoking duration, intensity, and pack-years. Initial models were stratified by current versus former smoking status, with a second set of models that pooled all participants and controlled for status. Generalized additive models for time to incident CVD (myocardial infarction, stroke, and CVD death) were used to investigate the functional form of each aspect of smoking. Cox models were used to investigate the hazard ratio (HR) of each exposure definition of smoking on time to event. Models were compared which used intensity and duration alone, in combination, and expressed as pack-years. The interaction between duration and intensity was also tested. All models were adjusted for age, sex, and race/ethnicity. We summarized model performance using the pseudo- R
2
statistic and the area under the receiver-operator characteristic curve (AUC).
Results:
There were 2487 former and 887 current smokers at baseline, with 370 incident CVD events over a median of 8.5 years. There was no significant non-linearity detected for intensity, duration or pack-years with respect to CVD risk. Among the former smokers none of duration, intensity, and pack-years was significant. Among the current smokers, only intensity was significant (HR for cigarettes/day 1.03, p=0.005). There was no evidence of an interaction between intensity and duration. Among current smokers, the model including intensity alone had an R
2
of 0.24 and AUC of 0.697. The model adding duration had very similar results with an R
2
of 0.24 and AUC of 0.698. The model with pack-years had worse performance than the model with intensity alone, with an R
2
of 0.19 and AUC of 0.662.
In the model pooling all participants, former smoking and duration were non-significant, while current smoking (HR 2.0, p<.001), intensity (HR packs/day 2.0, p<.001) and the interaction of former smoking status and intensity (HR 0.57, p=0.005) were significant. The interaction indicates that intensity is only important for current smokers.
Other outcomes were also examined including coronary heart disease (CHD) and the findings were very consistent.
Conclusions:
For CVD risk, only current smoking intensity at baseline is significant. Adding duration to the model does not add information, and use of the pack-years variable results in a loss of information since it does not estimate the intensity effect as accurately. These results suggest that lowering or limiting current smoking would lower CVD risk, regardless of cumulative smoking exposure.