Risk Factors for Venous Thromboembolism
Results From the Copenhagen City Heart Study
Background— Studies have suggested a link between risk factors for atherosclerotic disease and venous thromboembolism (VTE), but results are heterogeneous. We sought to identify risk factors for VTE with a focus on risk factors for atherosclerotic disease.
Methods and Results— Data were taken from the Copenhagen City Heart Study, a prospective cohort study of a random, age-stratified sample of people living in a defined area in Copenhagen, Denmark, started in 1976 with follow-up until 2007. First VTE (deep vein thrombosis and pulmonary embolism) diagnosis was retrieved from electronic national registries from study baseline to 2007. Of 18 954 subjects (median follow-up, 19.5 years) representing 360 399 person-years of follow-up, 969 subjects experienced at least 1 VTE, corresponding to a crude incidence rate of 2.69 (95% confidence interval [CI], 2.52 to 2.86) per 1000 person-years. The variables found to be significantly associated with VTE in a multivariable model adjusted for age and calendar time were as follows: body mass index (hazard ratio [HR] for ≥35 versus <20=2.10 [95% CI, 1.39 to 3.16]); smoking (HR for ≥25 g tobacco per day versus never smoker=1.52 [95% CI, 1.15 to 2.01]); gender (HR for men versus women=1.24 [95% CI, 1.08 to 1.42]); household income (HR for medium versus low=0.82 [95% CI, 0.70 to 0.95]); and diastolic blood pressure (HR for >100 versus <80 mm Hg=1.34 [95% CI, 1.08 to 1.66]). Other cardiovascular risk factors including total/high-density lipoprotein/low-density lipoprotein cholesterol levels, triglyceride levels, and diabetes mellitus were not associated with VTE.
Conclusions— Obesity and smoking were both found to be important risk factors for VTE whereas total/high-density lipoprotein/low-density lipoprotein cholesterol levels, triglyceride levels, and diabetes mellitus were not.
Received November 5, 2009; accepted March 9, 2010.
Venous thromboembolism (VTE) and atherosclerotic disease have for many years been considered 2 different disease entities with important differences in their pathways to thrombosis. Recent case-control studies, however, have shown that the 2 may be associated, with a high proportion of patients with VTE demonstrating endothelial dysfunction,1 preclinical atherosclerotic changes,2,3 and higher risk of subsequent development of atherosclerotic disease.4,5 In agreement with this, a recent meta-analysis concluded that the 2 share several risk factors, including obesity, hypertension, diabetes mellitus, low low-density lipoprotein (LDL) levels, and high triglyceride levels.6 However, for most risk factors the association was heterogeneous, and the hitherto largest prospective studies did not confirm the majority of these associations.1,7 In regard to smoking, which is one of the most important risk factors for cardiovascular disease, many clinicians assume a connection between smoking and VTE, but evidence of this is conflicting.1,2,8–11 Thus, the only cardiovascular risk factor consistently linked with VTE is obesity.7,9,10,12 The intriguing results from the Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin (JUPITER) trial, in which rosuvastatin reduced the occurrence of VTE,13 and a recent large study showing that microalbuminuria was associated with increased risk for VTE further highlight the gaps in our knowledge on risk factors for VTE.
Clinical Perspective on p 1903
Associations between risk factors for cardiovascular disease and risk of VTE have clinical implications with respect to risk assessment, risk factor modification, and medical treatment. We therefore studied associations between cardiovascular risk factors and subsequent VTE in a prospective population study comprising nearly 19 000 men and women with a median follow-up of >20 years, making this the largest prospective study in the field thus far.
Population and Design
We used data derived from the Copenhagen City Heart Study, a random, age-stratified sample of 19 698 individuals aged ≥20 years sampled in 1976 among 90 000 people living in a defined area in Copenhagen, Denmark. A total of 14 222 subjects (response rate, 72%) attended the first examination in 1976–1978. At follow-up in 1981–1983, 500 new subjects aged 20 to 25 years were invited; of these, 287 attended (response rate, 57%). In 1991–1993, 3000 new subjects were invited; of these, 1938 attended (response rate, 65%). Finally, in 2002–2003, 1040 new subjects aged 20 to 30 years were invited; of these, 517 attended (response rate, 50%). At every reexamination, all sampled persons were reinvited (if still alive), even if they did not attend a former examination, and thus the overall participation rate was 78%. More than 99% were white persons of Danish descent.
Identification of Venous Thromboembolic Events
The present analyses are based on all subjects attending ≥1 of the 4 examinations. Subjects were followed until December 31, 2006, for first fatal or nonfatal VTE event. A VTE event was defined as a deep venous thrombosis or pulmonary embolism corresponding to the International Classification of Diseases (ICD) codes (ICD-8 codes 451, 451.0, 451.9, 671, 450, 673.9 for 1977–1993 and ICD-10 codes I80.1, I80.2, I80.3, O22.3, O87.1, I26.0, I26.9 since 1994) obtained from the Cause of Death Registry and the National Patient Registry, an administrative registry of all in and out patient contacts in Denmark. The diagnosis of deep venous thrombosis in Denmark was performed in the last 10 to 15 years with the use of compression color duplex ultrasound and, before this, with x-ray phlebography. Pulmonary embolism is usually diagnosed with the use of perfusion/ventilation scintigraphy, although use of computer tomography is increasing. VTE events were divided into primary and secondary. A VTE was defined as secondary if the subject had cancer (besides basal cell carcinoma of the skin) within 5 years before or 2 years after an event or had a cerebrovascular event or a fracture requiring admission within 6 months before an event. All other events were classified as primary.
We followed all individuals from baseline until the occurrence of a first VTE event, death, emigration, or end of follow-up, whichever occurred first. Baseline was defined as the date at which the participant first attended the Copenhagen City Heart Study. For 14 222 participants, baseline was the 1976–1978 examination, for 1558 it was the 1981–1983 examination, for 2351 it was the 1991–1994 examination, and for 823 it was the 2002–2003 examination. We terminated any further follow-up on August 11, 2007, because this was the last date on which diagnostic information on end points could be obtained. Seventy-six participants emigrated during follow-up and were therefore censored at the emigration date. Follow-up is >99% complete.
Examinations included a self-administered questionnaire, a physical examination, and blood samples. Clinical and demographic data as well as data on nonresponders have been published previously.14,15
Variables of Interest
Diabetes mellitus was defined as self-reported diabetes mellitus or nonfasting glucose levels of ≥11.1 mmol/L (200 mg/dL). Arterial blood pressure was measured with the subject in a sedentary position after at least 5 minutes of rest. Hypertension was defined as blood pressure >140/90 mm Hg or answering “yes” to taking antihypertensive medication. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m2) and divided into 5 categories: <20, 20 to <25, 25 to <30, 30 to <35, and ≥35. Waist-hip ratio was divided into gender-specific quartiles to account for differences in distribution between genders. Waste-hip ratio was measured only in the third and fourth examinations. Blood lipids were taken with the subject in a nonfasting state. Values of triglycerides were measured only at the first, third, and fourth examinations, high-density lipoprotein (HDL) cholesterol only at the second, third, and fourth examinations, and LDL cholesterol only in the third and fourth examinations. All laboratory measures were analyzed in quartiles. Tobacco consumption was divided into 5 categories: never smokers, ex-smokers, and current smokers of 1 to 14, 15 to 24, and ≥25 g tobacco per day. Type of tobacco (cigarette, cheroot, cigar, pipe, or mixed) was recorded for current smokers. Current tobacco consumption was calculated by equating a cigarette to 1 g tobacco, a cheroot to 3 g tobacco, and a cigar to 5 g tobacco. Alcohol consumption was divided into 4 categories of increasing consumption. Physical activity in leisure time was divided into 2 categories as sedentary or moderate activity <4 hours per week versus moderate or intense activity >4 hours per week. Physical activity at work was divided into 2 categories as mostly sitting, standing, and walking or some lifting and heavy physical work. Educational level was divided into 3 categories: <8 years of schooling (completed primary school), 8 to 11 years, and >11 years. Household income level was divided into 3 categories. The use of hormone replacement therapy (HRT) was self-reported.
Categorical data were summarized as percentages, and differences were tested with the χ2 test. Continuous variables were summarized as medians and interquartile ranges or means and SDs, and differences were tested with the Student t test for normally distributed variables and with the Wilcoxon rank sum test for nonnormally distributed variables.
For baseline analyses only (Table 1), results refer to the first examination in which the value was measured. For survival analyses, the value for each variable was taken from the most recent examination that the subject attended; if a subject missed an examination, the value from the preceding examination was carried forward. This was done to account for changes in risk factors during the follow-up. For analyses of variables that were not measured at baseline (LDL cholesterol, HDL cholesterol, triglyceride levels, and waist-hip ratio), individuals’ observation time started at the examination in which the variable was first measured. For analyses of effect of waist-hip ratio, this meant that observation time started at the time of the third examination. Number of events and person-years under observation were thus reduced, leading to somewhat reduced power for these analyses. A Cox proportional hazards model was fitted on a multiple records per observation data set, allowing for time-dependent covariates. Results were reported as hazard ratios (HRs). Age was used as the underlying time scale, ensuring optimal adjustment for age. To account for calendar time as a confounder, a categorical variable counting 5-year periods from 1976 was entered into the model. All models were thus adjusted for both age and calendar time.
Initial analyses evaluated the contribution of each covariate in an age- and calendar time–adjusted gender-specific Cox regression model. The initial multivariable model was constructed containing the risk factors found to yield a univariable P<0.10. We forced gender, calendar time, and BMI into the model and removed the variables with the most insignificant P value, until only variables with P<0.05 were left in the model. After similar effects of covariates in men and women were ensured, further analyses were performed on pooled data with adjustment for the effect of gender. Tests for interaction between gender and each of the covariates in the model were performed with the log likelihood ratio test, and, similarly, all covariates in the model were examined for interaction with calendar time. Trends in HRs were examined by fitting a linear model to categories with increasing values.
Violation of the proportional hazards assumption was tested on the basis of Schoenfeld residuals. Stata 10 was used for statistical analyses (Stata 10 for Windows, StataCorp LP, College Station, Tex).
The study population comprised 18 954 subjects, representing 360 399 person-years of follow-up (median, 19.5 years; range, 0 to 30.8 years). During follow-up, 969 subjects experienced at least 1 VTE event, of which 713 were primary and 256 were secondary. The crude incidence rate was 2.69 (95% confidence interval [CI], 2.52 to 2.86) per 1000 person-years. The characteristics of the subjects at baseline divided into whether they later experienced a VTE or not are presented in Table 1. Subjects who later experienced a VTE were older and were characterized by a more adverse cardiovascular risk factor profile: They had higher prevalence of hypertension, higher systolic and diastolic blood pressure, higher BMI and waist-hip ratio, higher total and LDL cholesterol levels, and higher triglyceride levels, were more sedentary, and had a lower education level, whereas diabetes mellitus, alcohol consumption, and use of HRT did not differ.
Cardiovascular Risk Factors and VTE
Table 2⇓ summarizes gender-specific univariable HRs for all variables of interest adjusted for age and calendar time. Overall associations were stronger in women than in men, and we found a statistically significant interaction with gender in respect to educational attainment, but no further interactions with gender were found. BMI, smoking, diastolic blood pressure, and educational attainment were associated with VTE in both men and women. Waist-hip ratio, diabetes mellitus, triglycerides, and low HDL were associated with risk of VTE in women only. Hypertension or systolic blood pressure was not associated with increased risk of VTE, but, as noted, a diastolic blood pressure >100 mm Hg was associated with a significantly higher risk of VTE in both men and women. The strongest predictor of VTE among women was BMI, with a HR of 2.24 (95% CI, 1.42 to 3.53; P<0.005) for BMI ≥35 versus BMI <20 kg/m2. For men, the corresponding HR was 1.84 (95% CI, 0.78 to 4.30; P=0.16). Separate analyses of primary and secondary VTE yielded similar results (data not shown).
The effect of covariates on risk of VTE was similar in men and women; because of this, further analyses were done on the pooled data. As seen in the Figure, after multivariable adjustment, men had higher risk of VTE than women. The strongest association was seen for BMI: In the multivariable-adjusted analyses, HRs for BMI (5 categories) increased for each increase in category, with a highly statistically significant HR of 2.10 for all VTE (95% CI, 1.39 to 3.16; P<0.001) for BMI ≥35 versus BMI <20. For smoking, risk was not increased in ex-smokers versus never smokers (HR=0.99; 95% CI, 0.81 to 1.20; P=0.91) but increased with increasing consumption in current smokers to a HR of 1.52 (95% CI, 1.15 to 2.01; P=0.003) for ≥25 g per day of tobacco versus never smokers.
The higher risk seen for diastolic blood pressure >100 mm Hg was retained after multivariable adjustment, as was the association with household income. Having diabetes mellitus, low HDL cholesterol level, and low education level conferred a significantly increased risk of VTE when analyzed after univariable adjustment, but associations were attenuated and were no longer statistically significant when analyzed after multivariable adjustment.
Results were similar when secondary VTE was excluded. We also analyzed secondary VTE events separately and compared them with primary events, and the same trends were seen for the variables of interest, but much power was lost because of the relatively low number of events in this group (data not shown).
Risk of VTE increased over the almost 30-year study period in the multivariable-adjusted model and more so in women than in men (P for interaction between calendar time and gender=0.018). The HR for men versus women in the first period was 1.89 (95% CI, 1.15 to 3.09; P=0.011) and decreased to 0.86 (95% CI, 0.45 to 1.64; P=0.65) in the last period. The effect of all other covariates was stable over the observation period.
In this study, the largest prospective study of risk factors for VTE to date, we could not confirm that VTE and atherosclerotic disease uniformly share risk factors. Only male gender, obesity, current smoking, high diastolic blood pressure, and household income were found to be independent risk factors, whereas diabetes mellitus, hypertension, cholesterol levels, triglyceride levels, and physical inactivity were not. Thus, markers of dyslipidemia were not found to be risk factors for VTE.
Several studies have addressed the link between cardiovascular risk factors and VTE. Results were pooled in a recent meta-analysis, which concluded that obesity, hypertension, diabetes mellitus, triglycerides, and low HDL cholesterol were significant risk factors for VTE, with borderline associations found for smoking.6 However, the meta-analysis was based on both case-control studies and cohort studies, and for most risk factors there was considerable heterogeneity between studies. When we attempted to correct for this by removing studies of low quality, some of the reported effects disappeared. The main problem with interpretation of the results, however, was lack of adjustment for age and other cardiovascular risk factors. This was a particular problem with hypertension, HDL cholesterol, and triglycerides, which were reported to be significant risk factors, but all of these variables are closely related to BMI. In the present study, diabetes mellitus, low HDL cholesterol, and high triglycerides were associated with higher risk of VTE in initial univariable analysis, but this was no longer the case after adjustment for BMI. Thus, results from studies with insufficient adjustment should be interpreted with caution.
The only risk factor in concordance with the conclusion from the meta-analysis is obesity, which is already a well-established risk factor for VTE.7,9,10,12 Several prospective studies have addressed the associations between cardiovascular risk factors and VTE. The largest among these, with 329 526 person-years of follow-up and 358 VTE events, performed a competing risk analysis and found that coronary heart disease had a risk factor profile that differed widely from the profile for VTE.16
Another study with 148 054 person-years of follow-up and 215 VTE events7 found that high BMI, diabetes mellitus, and male gender were risk factors for VTE, whereas smoking, hypertension, dyslipidemia, physical inactivity, and alcohol consumption were not.7,17 A Swedish study of 855 men (65 VTE events) found that large waist circumference and smoking were risk factors for VTE, whereas high cholesterol and hypertension were not.9 In a recent Dutch study comprising 8574 subjects and 129 VTE events, BMI, microalbuminuria, and use of oral contraception in women were risk factors, whereas gender, diabetes mellitus, smoking, hypertension, and hyperlipidemia were not.1 Examining only pulmonary embolism, Goldhaber et al10 found that high BMI, smoking, and hypertension were risk factors. Many other smaller prospective and case-control studies corroborate high BMI as a risk factor.18–20
Smoking is an established risk factor for atherosclerotic disease, but its role in VTE is controversial. Some studies have shown smoking to be an independent risk factor,9–11,21,22 whereas others have failed to find any association between smoking and VTE.1,7 A recent large Danish study (641 VTE events) found smoking to be an independent risk factor with a dose-response relationship, similar to the present study.23 The studies not showing an association with smoking may have underestimated the risk because of pooling ex-smokers with current smokers and lack of distinction between light and heavy smokers. In addition, lack of adjustment for smokers quitting during follow-up, particularly if follow-up is prolonged, will cause risk associated with smoking to be underestimated. In the present study, participants were reexamined at regular intervals, and all covariates were updated.
Focusing on dyslipidemia as a risk factor for VTE is especially interesting in light of the recent results from the JUPITER trial showing significant reduction in risk of VTE in participants treated with rosuvastatin.13 In the present study, there was no association between lipids and subsequent VTE. Similarly, in the 2 other large prospective studies of risk factors for VTE, no association between this and HDL cholesterol or triglycerides was found.1,7,17 In addition, in the placebo-treated arm of the JUPITER trial, there was little evidence of higher risk among those with higher triglyceride or lower HDL levels.13 Thus, the results of the present and other large prospective studies indicate that the beneficial effect of statins on VTE risk is likely to be through pleiotropic effects.
We found diastolic blood pressure to be an independent risk factor, but the finding was only statistically significant when we compared the highest versus the lowest category. Hypertension analyzed as a dichotomous variable encompassing medication use and diastolic and systolic blood pressure was not significant, and this may be a chance finding. Low physical activity during either work or leisure time was not found to be a risk factor for VTE in this study, nor was this the case in 2 other major studies.7,16
Overall incidence of VTE was higher in this study than in most other studies. This is probably due to the higher age of our study population, the prolonged follow-up, and the complete event ascertainment due to almost 100% register coverage. Furthermore, we found that during the almost 30-year study period, the risk of VTE rose steadily and more so in women than in men. We ascribe this increase to the effect of higher vigilance, more sensitive and easily available diagnostic methods such D-dimer assays, compression color duplex ultrasound, and computed tomography, and a “healthy responder effect.”24 The reason that this effect was more pronounced in women than in men is much harder to explain, but a study from the Olmstead County cohort during 1966–1990 has reported a similar finding, with an increase in deep venous thrombosis events only in women during this period.25
Risk factors for VTE did not otherwise differ significantly between the 2 genders. We were not able to demonstrate HRT as a risk factor, although this has been the case in several studies, including a large randomized controlled trial.26 It is likely that our analyses underestimate the effect of HRT because of differential prescription and the fact that the result reflects the well-established discrepancy between the effect of HRT in observational and intervention studies.
Our study has several potential limitations. The study was well powered to observe any relevant effects in the pooled analysis, but for the gender-specific analysis, we may have lacked power. Because some variables were measured in only some of the examinations, this further decreased the power. The end points, which were taken from computerized registries, have not been validated, and some VTE events will have been undetected. However, it is unlikely that misclassification and incomplete registration of events would have been related to the risk factors in question and thus could have introduced a bias, but a weakening of the association cannot be ruled out.
The main strengths were the size of the study and the use of time-dependent covariates, thus allowing us to account for changes in variables such as smoking status and blood pressure during the follow-up. Furthermore, we adjusted for both calendar time and age, thus accounting for both as confounders.
Obesity and smoking were both found to be important risk factors for VTE, whereas total/HDL/LDL cholesterol levels, triglyceride levels, and diabetes mellitus were not. Thus, reports that cardiovascular disease and VTE share most major risk factors could not be confirmed in this large prospective, population-based study of risk factors for VTE.
Sources of Funding
This work was supported by the Foundation of 17-12-1981, Research Foundation of Bispebjerg Hospital, the Danish Heart Foundation, and the John and Birthe Meyer Foundation.
Becattini C, Agnelli G, Prandoni P, Silingardi M, Salvi R, Taliani MR, Poggio R, Imberti D, Ageno W, Pogliani E, Porro F, Casazza F. A prospective study on cardiovascular events after acute pulmonary embolism. Eur Heart J. 2005; 26: 77–83.
Ageno W, Becattini C, Brighton T, Selby R, Kamphuisen PW. Cardiovascular risk factors and venous thromboembolism: a meta-analysis. Circulation. 2008; 117: 93–102.
Samama MM, Dahl OE, Quinlan DJ, Mismetti P, Rosencher N. Quantification of risk factors for venous thromboembolism: a preliminary study for the development of a risk assessment tool. Haematologica. 2003; 88: 1410–1421.
Severinsen MT, Kristensen SR, Johnsen SP, Dethlefsen C, Tjønneland A, Overvad K. Anthropometry, body fat, and venous thromboembolism: a Danish follow-up study. Circulation. 2009; 120: 1850–1857.
Glynn RJ, Danielson E, Fonseca FA, Genest J, Gotto AM, Kastelein JJ, Koenig W, Libby P, Lorenzatti AJ, MacFadyen JG, Nordestgaard BG, Shepherd J, Willerson JT, Ridker PM. A randomized trial of rosuvastatin in the prevention of venous thromboembolism. N Engl J Med. 2009: 360: 1851–1861.
Glynn RJ, Rosner B. Comparison of factors for the competing risks of coronary heart disease, stroke, and venous thromboembolism. Am J Epidemiol. 2005; 162: 975–982.
Chamberlain AM, Folsom AR, Heckbert SR, Rosamond WD, Cushman M. High-density lipoprotein cholesterol and venous thromboembolism in the Longitudinal Investigation of Thromboembolism Etiology (LITE). Blood. 2008; 112: 2675–2680.
Jang MJ, Choi W, Bang S, Lee T, Kim Y, Ageno W, Oh D. Metabolic syndrome is associated with venous thromboembolism in the Korean population. Arterioscler Thromb Vasc Biol. 2009; 29: 311–315.
Deep vein thrombosis and pulmonary embolism are often viewed as 1 disease denoted venous thromboembolism (VTE), which is the cause of considerable morbidity and mortality. VTE and atherosclerotic disease such as acute myocardial infarction have for many years been considered 2 different disease entities with important differences in their pathways to thrombosis and hence in their preventive measures. Recent studies, however, have disputed this, among these a large clinical trial (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin [JUPITER]) showing an impressive lowering of VTE incidence with statin treatment. In this hitherto largest study, we studied common risk factors for atherosclerotic disease to determine whether these were also associated with VTE but found that only obesity and smoking were important risk factors for VTE, whereas total/high-density lipoprotein/low-density lipoprotein cholesterol levels, triglyceride levels, and diabetes mellitus were not. This would indicate that any beneficial effect of statins on VTE is through pleiotropic mechanisms and that treatment of dyslipidemia in itself may be of low value for preventing VTE. Conversely, management of obesity and smoking cessation are likely to be beneficial for both VTE and atherosclerotic disease.