In this large prospective study, we found current smoking to be positively associated with VTE, among both men and women. Former smokers had essentially the same risk of VTE as never smokers, indicating that the mechanism of smoking’s effect on VTE is acute. We found that smoking more than 20 g day−1 among women and 30 g day−1 among men was associated with a 150–300% higher hazard of VTE. Lower rates of tobacco consumption were associated with only a 10% to 40% higher hazard of VTE compared with never smokers, indicating a threshold difference for both men and women (Fig. 2).
Strengths and limitations of our study
This prospective study included 641 VTE events, and was one of the largest prospective studies on VTE. Data on smoking habits were detailed, and data from follow-up after 5 years showed a high degree of concordance with the baseline information. The changes were generally toward decreased consumption of tobacco at the follow-up, which may have resulted in an underestimation of risk. All VTE events were validated by review of medical records, and only objectively verified VTE events were included. The hospital system in Denmark is financed by taxation and almost everybody with VTE symptoms will be admitted to and examined in a hospital. The Danish National Patient Registry has collected nationwide data on all somatic hospital admissions since 1977. Since 1995, discharges from emergency departments and out-patient clinics have also been included in the registry . The medical record includes information on visits to outpatient clinics as well as inpatients clinics. Obviously, there could be VTE events among participants that we missed. For example, because of a very low frequency of autopsy in Denmark (6% of all deaths in 2001) an unknown number of the participants died of PE. However, because smoking habits were unlikely to be associated with the chance of autopsy, any misclassification of fatal PE events was unlikely.
Smoking is an established risk factor for chronic obstructive lung diseases. Diagnosing PE is challenging, especially among patients with chronic obstructive lung disease, because perfusion/ventilation scintegraphy may not be conclusive in these patients. In addition, the most prominent symptom of PE, dyspnea, is normal for these patients. Dyspnea may be interpreted as a result of an exacerbation of the chronic obstructive lung disease, or as a result of pneumonia, which also occurs very often among these patients. The effect of smoking on PE may therefore be underestimated due to information bias.
Detailed information on a range of potential confounding factors was available for this study. However, adjusting for these factors in the statistical analyses had only a minor impact on the estimated hazard ratios. This indicates that residual confounding is not a likely explanation for the observed associations. However, Rosengren et al.  found a statistically significant positive association between PE and occupational class in men. Therefore we performed secondary analysis including adjustment for educational status and for years in primary school. The adjustments slightly weakened the association between smoking and VTE. Socioeconomic status may, however, directly influence smoking habits. Inclusion of socioeconomic status reduced the variation in smoking habits but also took into account potential confounding due to an association between socioeconomic status and smoking habits. However, as in all observational studies, we cannot completely rule out possible confounding by unknown factors.
Comparison with other studies
Our findings accord with the results from a number of other large-scale epidemiological studies. In a recent large case-control study, Pomp et al.  also found higher risk of VTE among smokers. Two prospective studies have found a positive association between smoking and venous thrombosis. Goldhaber et al.  investigated the association between smoking and PE among nurses. They reported 280 PE events during 16 years of follow-up. They found a significant positive association between smoking and PE among women who smoked more than 25 g day−1, and no association with lower tobacco doses, exactly as we did. Hansson et al.  also found a positive association between smoking more than 15 g day−1 of tobacco and the occurrence of VTE in a study of 851 men, followed for 30 years. The study included 65 VTE cases, with a substantial proportion, 21 of 65, diagnosed by autopsy, explaining why as many as 36 had PE. Also a substantial proportion (69%) was secondary VTE cases, of which 23 (35.5%) had cancer. After including comorbidity in a multivariate analysis, Hansson et al. were still able to find the association between smoking and VTE, in accordance with our findings.
In contrast, other prospective studies have not found an association between smoking and VTE. Tsai et al. combined data from two cohorts (the Atherosclerosis Risk in Communities study, ARIC, and the Cardiovascular Health study, CHS) in the Longitudinal Investigation of Venous Thromboembolism study (LITE). The ARIC cohort included men and women, 45–64 years of age, in which 130 VTE events occurred during follow-up. The CHS cohort included men and women above 65 years of age, in whom 85 VTE events occurred. Data in the LITE study included smoking status (never, former or current smoker). No association between VTE and smoking status was found. They also analyzed pack-years of smoking and found no association with VTE. The ARIC cohort data included current doses of tobacco. Analysis of this cohort showed a positive association between VTE and smoking more than 25 g day−1 of tobacco, compared with never smokers with adjusted hazard ratio of 1.68 (95% CI, 0.91–3.1) . Glynn et al. also found no association between smoking and VTE in the prospective Physicians’ Health study. This study included American male physicians and 358 VTE events. The information on smoking included data on smoking status (current, former or never smoker), but not on tobacco dose .
We analyzed the dose-response relationship in a spline model that included current smoking dose as a continuous variable. The hazard ratio of VTE among women increased steeply for smoking more tobacco than 20 g day−1. Among men, the hazard ratio of VTE did not increase substantially before the tobacco dose exceeded 30 g day−1 (Fig. 2). Most previous studies did not include tobacco dose, only smoking status (never, former or current smoker). However, studies that did include tobacco dose also found the strongest association for the highest doses of tobacco, in line with our results. Our data suggest thresholds for the effect of smoking, at different levels for men and women. This could be explained, in part, by the differences between genders in body volume or liver metabolism. Further studies are needed to confirm these findings.
In both genders, we found a strong positive association between the highest doses of tobacco and idiopathic VTE (i.e. 4.34 [95% CI, 2.10–8.96] in women and 1.89 [95% CI, 0.97–3.68] in men), underscoring that the effect of smoking on the risk for VTE may not only be mediated through secondary diseases caused by smoking. We reviewed the updated discharge history from the Danish National Patient Registry for the 20 heavy smokers who suffered from idiopathic VTE. Of the 20 patients, one woman had a diagnosis of lung cancer 1 year later than the date of VTE, another woman had a diagnosis of breast cancer 6 years later and a man had a diagnosis of lung cancer 6 years later. The remaining 17 patients did not have diagnosis of cancer before April 2008 (mean of follow-up, 6 years from the date of VTE diagnosis). In secondary analysis we excluded the woman with a cancer diagnosis 1 year later than the VTE event because she might have had the cancer at the date of VTE (1 year before the cancer diagnosis) and found a hazard ratio of 3.85 (95% CI, 1.80–8.25) for idiopathic VTE in women smoking more than 25 g of tobacco per day. Our findings suggest a direct effect of heavy smoking on the risk of VTE. These findings are biologically plausible because smoking increases the level of coagulation factors in the blood and it also promotes activation of the inflammatory system, both of which are found to be associated with venous thrombosis [16,17]. We can not rule out possible confounding by unknown factors; however, it is difficult to imagine a strong confounder that only occurs among the heaviest smokers and that can explain our findings. Steffen et al.  showed that a diet including more plant food and fish and less meat is associated with lower incidence of VTE and there might be an association between smoking habits and diet; however, we think it is unlikely that poor diet explains our findings. Further studies are needed on this topic.
Among women the highest doses of tobacco (more than 25 g day−1) were positively associated with PE (i.e. hazard ratio of 3.07 [95% CI, 1.45–6.54]), whereas among men an insignificant positive association with PE was found (i.e. hazard ratio of 1.33 [95% CI, 0.51–3.44]) with the highest doses of tobacco (more than 35 g day−1). This may indicate that smoking promotes coagulation but is not involved in the process of embolization.
In conclusion, we found a positive association between smoking and VTE. The hazard ratio of VTE was especially high at tobacco doses above thresholds of 20 g day−1 of tobacco for women and 30 g day−1 for men. The smoking effect seems to be mediated by an acute effect because former smokers have the same risk as those who never smoked.