Cigarette smoking and the risk of venous thromboembolism: The Tromsø Study
Kristin F. Enga, Hematological Research Group (HERG), Department of Clinical Medicine, University of Tromsø, N-9037 Tromsø, Norway.
Tel.: +47 77620880; fax: +47 77646838.
Summary. Background: Conflicting findings have been reported on the association between smoking and the risk of venous thromboembolism (VTE).
Objectives: To conduct a prospective, population-based cohort study to investigate the association between cigarette smoking and the risk of incident VTE.
Patients/Methods: Information on smoking habits was assessed by self-administered questionnaires in 24 576 subjects, aged 25–96 years, participating in the fourth Tromsø Study in 1994–1995. Incident cases of VTE were registered until the end of follow-up at 1 September 2007.
Results: A total of 389 incident VTE events (1.61 per 1000 person-years) were registered during follow-up (median of 12.5 years). Heavy smokers (> 20 pack-years) had a hazard ratio (HR) of 1.46 (95% confidence interval [CI] 1.04–2.05) for total VTE, and and an HR of 1.75 (95% CI 1.14–2.69) for provoked VTE, as compared with never smokers. The risk of provoked VTE increased with more pack-years of smoking (P = 0.02). Smoking was not associated with risk of unprovoked VTE. The number of pack-years was associated with increased risk of cancer and myocardial infarction, whereas the association between pack-years of smoking and VTE disappeared when failure times were censored at the occurrence of cancer or myocardial infarction.
Conclusions: Heavy smoking was apparently a risk factor for provoked VTE in analyses with VTE events as the only outcome. The lack of association between smoking and risk of VTE in analyses censored at the occurrence of cancer or myocardial infarction may suggest that smoking-attributable diseases or other predisposing factors are essential for smoking to convey a risk of VTE.
Venous thromboembolism (VTE), including both deep vein thrombosis (DVT) and pulmonary embolism (PE), is a common disorder and a considerable cause of morbidity and mortality [1,2]. The annual incidence of VTE is 1–2 per 1000 person-years in developed countries [2–4], and it is the most common cause of preventable in-hospital death in the USA . Environmental factors such as age, obesity, cancer, certain medical conditions, surgery, trauma, immobilization, and the use of estrogens, along with inherited and acquired thrombophilias, are all established risk factors for VTE [1–3]. Nevertheless, 25–50% of VTE events occur in the absence of predisposing factors [1,5].
The harmful effects of cigarette smoking with regard to arterial cardiovascular diseases are well established [6,7]. However, the association between cigarette smoking and VTE is less clear. Several prospective studies have reported no association between smoking and the risk of VTE [8–11]. On the other hand, some observational studies have demonstrated an association between smoking status and VTE risk [12–15], whereas other prospective studies have reported that only heavy smoking is associated with a risk of future VTE [16–19].
Today, there are more than one billion smokers worldwide , and smoking represents the most important source of preventable morbidity and premature death [6,7]. Because of the existing conflicting results, it is important to further investigate the role of smoking in venous thrombosis. We used data from the Tromsø Study, a large, prospective cohort study with participants recruited from a general population, with detailed information on smoking habits and potential confounders, to assess the impact of cigarette smoking on the risk of incident VTE. Cox proportional hazard regression models were used to calculate hazard ratios (HRs) for VTE along with cause-specific hazard models to eliminate the impact of other smoking-related diseases, such as cancer and myocardial infarction (MI), on the risk of VTE.
Participants were recruited from the fourth survey of the Tromsø Study, carried out in 1994–1995. The Tromsø Study is a single-center, prospective, population-based study with repeated health surveys of the inhabitants of the municipality of Tromsø, Norway. All inhabitants older than 24 years were invited, and 27 158 (77% of the eligible population) participated. The regional committee of research ethics approved the study, and all subjects gave their written consent to participate. Subjects who did not give their consent to medical research (n = 201), and those not officially registered as inhabitants of Tromsø at the date of baseline examination (n = 44), were excluded from the study. Subjects with a known history of VTE (n = 47), a history of cancer (n = 750) or a history of MI (n = 608) at baseline were excluded, as cancer and cardiovascular disease (CVD) may be associated with smoking and VTE risk. Subjects with no information on smoking status (n = 622) were also excluded. Information on the duration and amount of smoking was available only for cigarette smokers. Participants who smoked cigars or pipes (n = 310) were therefore excluded. Accordingly, a total of 24 576 participants were included in the study, and were followed from the date of enrollment until the end of follow-up at 1 September 2007. Information on smoking status was available for all 24 576 participants, and number of pack-years was available for 24 530 participants.
Baseline information on lifestyle and cardiovascular risk factors was collected by means of physical examination, blood samples, and self-administered questionnaires. Information on smoking habits, including smoking status (current, former or never smoker), number of cigarettes per day, years of daily smoking, and time since smoking cessation, was collected with a self-administered questionnaire. With the same procedure, information on self-reported diabetes, higher education (education at university/college level) and current hormone therapy (current use of oral contraceptives or estrogen supplementation) was collected. Height and weight were measured with subjects wearing light clothes and no shoes. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters (kg m−2). Information on history of malignant disease was obtained from the Norwegian Cancer Registry , and cases of MI before baseline were identified from the CVD registry within the Tromsø Study .
Identification and validation of VTE
All incident events of VTE during follow-up were identified by searching the hospital discharge diagnosis registry, the autopsy registry and the radiology procedure registry at the University Hospital of North Norway, from the date of enrollment (1994–95) until 1 September 2007, as previously described .
The medical records for each potential VTE case were systematically reviewed. For subjects derived from the hospital discharge diagnosis registry and the radiology procedure registry, an episode of VTE was verified and recorded as a validated outcome when all four of the following criteria were fulfilled: (i) objectively confirmed by diagnostic procedures (compression ultrasonography, venography, spiral-computed tomography, perfusion–ventilation scan, pulmonary angiography, or autopsy); (ii) the medical record indicated that a physician had made a diagnosis of DVT or PE; (iii) signs and symptoms consistent with DVT or PE were present; and (iv) therapy with anticoagulants (heparin, warfarin, or a similar agent), thrombolytic therapy or surgery was required, unless contraindications to anticoagulant treatment were specified in the medical record. For subjects derived from the autopsy registry, a VTE event was recorded as an outcome when the autopsy registry indicated the VTE as the cause of death or as a significant condition.
A VTE event was further classified as unprovoked or provoked depending on the presence of provoking factors at the time of diagnosis. A VTE occurring without any provoking factors was defined as unprovoked, whereas a VTE occurring in the presence of one or more of the following factors was defined as provoked: recent surgery or trauma within 8 weeks before the event, acute medical conditions (acute MI, acute ischemic stroke, or major infectious disease), active cancer at the time of the event, marked immobilization (bed rest for > 3 days, confinement to a wheelchair, or long-distance travel exceeding 4 h within the last 14 days prior to event), or another potential provoking factor described by the physician in the medical record (e.g. intravascular catheter).
For each participant, person-years were accrued from the date of enrollment in the Tromsø Study (1994–1995) to the date when an incident VTE event was diagnosed, to the date when the participant died or moved from the municipality of Tromsø, or to the end of the study period (1 September 2007), whichever occurred first. Subjects who moved from the municipality of Tromsø (n = 3458) or died (n = 2158) during follow-up were censored from the date of migration or death.
Statistical analyses were performed with stata version 11.0 (Stata Corporation, College Station, TX, USA). Age-adjusted and sex-adjusted incidence rates were calculated by direct standardization, with the whole study population as the reference population, categorized into five age groups (≤ 30, 31–40, 41–50, 51–60 and > 60 years), and were expressed as number of events per 1000 person-years. Cox proportional hazard regression models were used to calculate the age-adjusted and sex-adjusted, and multivariable-adjusted hazard ratios (HRs) for VTE with 95% confidence intervals (CIs) for total VTE, provoked and unprovoked VTE by cigarette consumption; smoking status and pack-years. One pack-year was defined as smoking 20 cigarettes a day for 1 year. In the multivariable model, HRs were adjusted for age, sex, BMI, and level of education. Statistical interactions between smoking and age or sex were considered by including cross-product terms in the proportional hazard models. No statistical interactions were found. The proportional hazard assumption was verified by evaluating the parallelism between the curves of the log–log survival function for categories of pack-years, and a test of the proportional hazard assumption using Schoenfeld residuals was also conducted.
Cause-specific hazard analysis
Smoking is associated with risks of cancer and of CVD . Cancer is a strong risk factor for VTE , and a relationship between VTE and CVD has also been suggested [25,26]. Thus, it was pertinent to investigate the ‘pure’ association between smoking and the risk of VTE by using cause-specific hazard models, where the direct impact of smoking on VTE risk was assessed as though other events (i.e. cancer and MI) did not exist . The cause-specific hazard function represents the risk of an event at a moment in time, given that no other event has occurred so far . Information on incident cancer and MI during follow-up was obtained from the Norwegian Cancer Registry  and the CVD registry  within the Tromsø Study, respectively. Failure time was calculated from baseline inclusion to the date of a first event of either cancer (n = 1426), MI (n = 1113), or VTE (n = 238). Subjects who did not experience one of these events were censored from the date of death, migration, or the end of follow-up (31 December 2005). The data were prepared as a triplicate of the original dataset, giving each subject a separate observation for each outcome where two of the three outcomes were censored. Cox proportional hazard regression models, stratified by event type, including appropriate interaction terms, were used to assess the association between risk factors and outcomes .
There were 389 validated incident VTE events, of which 42.7% were considered to be unprovoked, during 268 518 person-years of follow-up (median of 12.5 years) (Table 1). The overall crude incidence rate was 1.45 per 1000 person-years (95% CI 1.31–1.60). Characteristics of the VTE events are shown in Table 1. There were no substantial differences between men and women with regard to the VTE localization (DVT/PE), distribution of clinical risk factors or provoking factors at the time of the event (data not shown).
Table 1. Characteristics of the venous thromboembolism (VTE) events (n = 389) evolving during follow-up
|Deep vein thrombosis||62.0 (241)|
|Pulmonary embolism||38.0 (148)|
|Unprovoked events||42.7 (166)|
| Clinical risk factors |
| Estrogen*†||15.3 (33)|
| Pregnancy/puerperium†||1.4 (3)|
| Heredity‡||3.4 (13)|
| Other medical conditions§||19.8 (77)|
| Provoking factors |
| Surgery||17.0 (66)|
| Trauma||6.7 (26)|
| Acute medical conditions||15.7 (61)|
| Cancer||21.3 (83)|
| Immobility¶||19.0 (74)|
| Other**||3.9 (15)|
The age-adjusted distribution of baseline characteristics across categories of pack-years is shown in Table 2. Heavy smokers had the highest mean age, and were predominantly men. The proportion of subjects with higher education and self-reported diabetes declined with increasing cigarette consumption. BMI was lower among current smokers than among never and former smokers. The age-adjusted proportion of estrogen users was highest among women who had smoked more than 20 pack-years (Table 2).
Table 2. Distribution of baseline characteristics across categories of pack-years of cigarettes*
|Number of observations||9098||6282||3784||3159||2204|
|Number of events||143||126||29||38||51|
|Age (years)||45.6 ± 15.9||49.5 ± 14.3||37.4.±11.4||45.3 ± 11.6||52.9.±10.6|
|Men||41.3 (3754)||54.1 (3398)||37.8 (1430)||45.0 (1421)||65.4 (1441)|
|BMI (kg m−2)||25.4 ± 4.0||25.6 ± 3.7||24.4 ± 3.6||24.5 ± 3.7||24.9 ± 3.8|
|Diabetes||1.1 (165)||1.1 (130)||0.9 (27)||0.6 (24)||1.0 (42)|
|Higher education†||39.1 (3710)||31.5 (1876)||17.4 (936)||15.5 (538)||20.4 (370)|
|Hormone therapy‡||20.1 (680)||19.4 (374)||19.6 (326)||15.8 (187)||21.2 (96)|
Among the study participants, 37.4% were current smokers, 25.6% were former smokers, and 37.0% were never smokers. Incidence rates and HRs for total, provoked and unprovoked VTE across categories of smoking status are shown in Table 3. The age-adjusted and sex-adjusted HR for current smokers vs. never smokers was 1.10 (95% CI 0.85–1.41) and the multivariable-adjusted HR was 1.21 (95% CI 0.93–1.56) (Table 3). Former smokers had a multivariable-adjusted HR of 1.13 (95% CI 0.88–1.45). Although none of the associations was statistically significant, the strongest HR (1.29) was found for current smoking and provoked VTE (Table 3).
Table 3. Incidence rates (IRs) and hazard ratios (HRs) with 95% confidence intervals (CIs) for venous thromboembolism (VTE) by smoking status (cigarettes only)
| Never smokers||98 217||143||1.46 (1.24–1.72)||Ref.||Ref.|
| Former smokers||69 209||126||1.55 (1.30–1.85)||1.12 (0.87–1.43)||1.13 (0.88–1.45)|
| Current smokers||101 092||120||1.39 (1.16–1.66)||1.10 (0.85–1.41)||1.21 (0.93–1.56)|
| Never smokers||97 785||82||0.83 (0.67–1.03)||Ref.||Ref.|
| Former smokers||68 835||70||0.90 (0.71–1.14)||1.08 (0.78–1.51)||1.10 (0.79–1.54)|
| Current smokers||100 715||71||0.83 (0.66–1.05)||1.16 (0.83–1.61)||1.29 (0.92–1.81)|
| Never smokers||97 664||61||0.64 (0.50–0.83)||Ref.||Ref.|
| Former smokers||68 685||56||0.67 (0.51–0.87)||1.16 (0.80–1.69)||1.16 (0.80–1.70)|
| Current smokers||100 564||49||0.57 (0.43–0.75)||1.02 (0.69–1.50)||1.10 (0.74–1.64)|
HRs for VTE across pack-years of smoking (Table 4) showed that subjects with > 20 pack-years had a 1.5-fold higher risk of VTE (HR 1.46; 95% CI 1.04–2.05) than never smokers. Risk estimates for VTE were close to 1.0 for subjects with ≤ 20 pack-years. Separate analyses for provoked and unprovoked VTE revealed that the increased risk of VTE caused by heavy smoking was confined to the risk of provoked VTE. Subjects who had smoked > 20 pack-years had a 1.8-fold increased risk of provoked VTE (HR 1.75; 95% CI 1.14–2.69). Number of pack-years was not associated with the risk of unprovoked VTE (multivariable-adjusted HR 1.09; 95% CI 0.62–1.92). Heavy smoking (> 20 pack-years) was associated with similar risk estimates for DVT and PE in separate analyses (data not shown). Extensive smoking duration (> 30 years of smoking) was also associated with increased risks of total and provoked VTE (multivariable-adjusted HR 1.36, 95% CI 0.99–1.86; and HR 1.54, 95% CI 1.03–2.31, respectively) (data not shown).
Table 4. Incidence rates (IRs) and hazard ratios (HRs) with 95% confidence intervals (CIs) of total, provoked and unprovoked venous thromboembolism (VTE) by categories of cigarette smoking (pack-years)*
| Never smokers||98 217||143||1.46 (1.24–1.72)||Ref.||Ref.|
| 0.1–10 pack-years||41 776||29||1.15 (0.80–1.66)||0.94 (0.63–1.43)||1.06 (0.70–1.62)|
| 10.1–20 pack-years||35 301||38||1.22 (0.88–1.67)||0.95 (0.66–1.37)||1.03 (0.71–1.51)|
| > 20 pack-years||23 451||51||1.51 (1.14–1.98)||1.33 (0.95–1.85)||1.46 (1.04–2.05)|
| P for trend||–||–||–||0.2||0.07|
| Never smokers||97 785||82||0.83 (0.67–1.03)||Ref.||Ref.|
| 0.1–10 pack-years||41 639||13||0.53 (0.31–0.91)||0.75 (0.41–1.37)||0.87 (0.47–1.59)|
| 10.1–20 pack-years||35 184||23||0.79 (0.52–1.19)||1.01 (0.63–1.62)||1.14 (0.70–1.86)|
| > 20 pack-years||23 334||34||0.96 (0.68–1.34)||1.56 (1.02–2.37)||1.75 (1.14–2.69)|
| P for trend|| – || – || – ||0.09||0.02|
| Never smokers||97 664||61||0.64 (0.50–0.83)||Ref.||Ref.|
| 0.1–10 pack-years||41 682||16||0.63 (0.39–1.03)||1.20 (0.67–2.13)||1.31 (0.73–2.35)|
| 10.1–20 pack-years||35 116||15||0.43 (0.26–0.72)||0.87 (0.49–1.54)||0.89 (0.49–1.63)|
| > 20 pack-years||23 212||17||0.56 (0.35–0.90)||1.02 (0.59–1.79)||1.09 (0.62–1.92)|
| P for trend|| – || – || – ||0.9||0.9|
Table 5 shows the relationship between pack-years and the risks of incident VTE, MI and cancer in a cause-specific hazard model. Number of pack-years was associated with increased risks of both MI and cancer. Smoking more than 20 pack-years yielded an HR of 2.60 (95% CI 2.17–3.11) for cancer and an HR of 2.32 (95% CI 1.97–2.72) for MI. However, in the cause-specific hazard model, there was no longer any association between number of pack-years and total VTE (multivariable-adjusted HR by > 20 pack-years vs. never smokers of 1.04; 95% CI 0.67–1.61) (Table 5).
Table 5. Incidence rates (IRs) and cause-specific hazard ratios (HRs) for venous thromboembolism (VTE), cancer, and myocardial infarction (MI) by categories of cigarette smoking (pack-years) with 95% confidence intervals (CI)*. The Tromsø Study 1994/1995–2005
|Person-years||84 360||36 257||29 996||19 436||–|
| HR (95% CI)†||Ref.||1.22 (0.77–1.93)||0.81 (0.50–1.31)||1.02 (0.66–1.57)||0.8|
| Multivariable HR (95% CI)‡||Ref.||1.26 (0.80–1.99)||0.75 (0.45–1.25)||1.04 (0.67–1.61)||0.8|
| HR (95% CI)†||Ref.||1.23 (0.95–1.58)||2.11 (1.75–2.55)||2.53 (2.12–3.02)||< 0.001|
| Multivariable HR (95% CI)‡||Ref.||1.23 (0.95–1.59)||2.22 (1.83–2.68)||2.60 (2.17–3.11)||< 0.001|
| HR (95% CI)†||Ref.||1.16 (0.93–1.45)||1.74 (1.46–2.06)||2.25 (1.92–2.64)||< 0.001|
| Multivariable HR (95% CI)‡||Ref.||1.19 (0.95–1.49)||1.81 (1.52–2.15)||2.32 (1.97–2.72)||< 0.001|
In the present study, heavy smoking (> 20 pack-years) was associated with a 1.5-fold increased risk of total VTE and a 1.8-fold increased risk of provoked VTE, whereas no association was found between heavy smoking and the risk of unprovoked VTE. However, in the cause-specific hazard analysis, heavy smoking was highly associated with increased risks of cancer and MI, but not with the risk of VTE in subjects free of MI and cancer, indicating that most of the effect of smoking on the incidence of VTE occurs in individuals with MI or cancer.
Smoking is strongly related to CVD and certain types of cancer , and ∼ 10% of all adult current or former smokers have smoking-attributable diseases such as chronic obstructive pulmonary disease , which is also related to the risk of VTE . In the present study, the lack of association between smoking and unprovoked VTE, and the lack of association between VTE and smoking in cause-specific hazard analyses, suggest that development of cancer or MI may be the actual cause of VTE in heavy smokers. The Iowa Women’s Health Study, which included a cohort study of older women , found an increased risk of VTE among current and former smokers. The risk was higher among heavy smokers (≥ 20 pack-years) and, in concordance with our findings, this association was largely driven by cancer-related VTE .
A number of studies have investigated the association between heavy smoking and VTE, but few other studies have taken the potential competing risk from other smoking-attributable diseases into account. In the Nurses’ Health Study , heavy smoking (> 25 cigarettes per day) was associated with both total and unprovoked VTE in a population free of cancer and CVD at baseline, but development of disease during the 16 years of follow-up were not considered. Heavy smoking (≥ 100 000 cigarettes ever) was associated with VTE among middle-aged women free of cancer at baseline (the MISS Study) , but incident cancer during follow-up was not accounted for, despite being noted to be an independent risk factor for VTE, in the analyses of smoking . In a prospective study of 18 954 participants in the Copenhagen City Heart Study, daily smoking was associated with dose-dependent increased risks of both total and unprovoked VTE in time-dependent analysis to account for changes in risk profile , but disease development during follow-up was not included in their model. Heavy smoking was associated with an increased risk of VTE in elderly men (≥ 15 g of tobacco per day) in ‘The study of men born in 1913’ , and the effect was still present after adjustments for cancer, MI, stroke and diabetes mellitus during follow-up. A Danish prospective study of 57 053 middle-aged men and women free of cancer at baseline  found a dose-dependent increased risk of total VTE among current smokers, with the same tendency for provoked non-cancer-related VTE. Furthermore, the risk of unprovoked VTE increased steeply for heavy smokers (> 25 g d−1 in women and > 35 g d−1 in men), suggesting a threshold effect .
Smoking status was not associated with an increased risk of VTE in our study. Accordingly, the Physicians’ Health Study , the Framingham Study  and the LITE Study  reported no risk of VTE or PE related to smoking status. However, recent results from the LITE Study showed an increased risk for smokers as compared with non-smokers in a time-dependent analysis where exposure variables, including smoking status, were updated during follow-up . The MEGA Study, a case–control study of 3989 cases and 4900 controls, in which subjects diagnosed with malignancy within 10 years before the index date were excluded, reported an increased risk of VTE among both current and former smokers . The divergence between studies that find and do not find an association between smoking status and VTE may, to some extent, be attributed to differences in the distribution of light and heavy smokers in the study populations. It is possible that a high proportion of light vs. heavy smokers may have contributed to the non-associative nature of smoking status observed in our study. The prevalence of heavy smokers was substantially higher in the MEGA Study  and in the Danish Cohort Study  than in our study. Thus, one cannot rule out the possibility that the observed lack of association between heavy smoking and unprovoked VTE in our study may be explained by the low prevalence of really heavy smokers.
The prospective design, the large number of participants recruited from a general population with a high attendance rate, the long follow-up time and validated VTE events are all strengths of the present study. One single hospital serves the entire Tromsø population, which enhances the completeness of the outcome registry. Other outcomes, such as incident events of MI and cancer, were available, and enabled the use of cause-specific hazard analyses. The study also has limitations. Smoking habits may change over time, so it is likely that some of the current smokers have stopped smoking during follow-up, and were misclassified as current smokers when they were actually former smokers. Similarly, individuals may have taken up smoking, and so have been misclassified as non-smokers for part of the follow-up. According to Statistics Norway, 33% of the Norwegian population between 16 and 74 years of age were daily smokers in the period 1995–1999, vs. 25% in 2003–2007 . A possible change in risk profile during follow-up may lead to misclassification and thereby an underestimation of the real association, owing to regression dilution . Confounding is a potential limitation of the cohort design, owing to the absence of random allocation. Smoking is related to socioeconomic variables such as occupational group, housing tenure, and marital status , which may be related to the risk of thrombosis [19,34]. In our study, educational level, as an indicator of socioeconomic status , was included in the multivariable model to adjust for confounding by socioeconomic variables. Moreover, further adjustments for self-reported diabetes and lipid levels did not change the risk estimates (data not shown). Self-reported information is another possible limitation. However, self-reporting of smoking habits has been proven to have high validity [36,37] and reliability . Unfortunately, we were not able to include cigar and pipe smokers (n = 306) in the study, owing to a lack of information on dose and duration of smoking. However, as cigar and pipe smokers represented only 3% of the smokers and 1% of the total population, exclusion of these had negligible effects. Moreover, we did not have verified baseline information on previous history of VTE for all study subjects. Hence, some of the subjects who were treated as healthy participants during follow-up could be prevalent VTE cases who should have been excluded from the study population. However, as these were likely to have been few in number, their inclusion or exclusion would, again, have had minimal effects.
In conclusion, the present population-based, prospective study found an increased risk of VTE caused by heavy smoking, and the increased risk was restricted to provoked VTE. However, the apparent association between heavy smoking and VTE disappeared when diagnoses of cancer or MI during follow-up were taken into consideration. This may suggest that the risk of VTE was mediated by the development of cancer or MI, as smoking was not associated with a risk of VTE in subjects who did not experience an MI event or cancer.
S. K. Brækkan and J. B. Hansen receive grants from the Northern Norway Regional Health Authority.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.