Epidemiology of venous thromboembolism: the need for large (including prospective) studies and meta-analyses


Gordon D. O. Lowe, Institute of Cardiovascular and Medical Sciences, University of Glasgow, 126 University Place, Glasgow, UK.
Tel.: +44 141 330 3325; fax: +44 141 330 7394.
E-mail: gordon.lowe@glasgow.ac.uk


See also Rosendaal FR. Etiology of venous thrombosis: the need for small original studies. This issue, pp 2189–90 and El-Galaly TC, Kristensen SR, Overvad K, Steffensen R, Tjønneland A, Severinsen MT. Interaction between blood type, smoking and factor V Leiden mutation and risk of venous thromboembolism: a Danish case–cohort study. This issue, pp 2191–3.

In this issue of the Journal, El-Galaly et al. [1] report on the interactions of two genetic risk factors (non-O blood group and factor V Leiden) for venous thromboembolism (VTE) and two lifestyle risk factors (tobacco-smoking and obesity), using data from a large national prospective study in Denmark. They observed positive interactions of non-O blood group with factor V Leiden, and with heavy smoking, but not with severe obesity. This report extends our knowledge on interactive risk factors for VTE, and highlights the need for further large (including prospective) studies and meta-analyses, to provide scientists, clinicians and their patients with reliable information.

It is generally accepted that, like arterial thrombosis, VTE is a multifactorial disease, to which many genetic and environmental factors contribute about equally [2]. These risk factors interact: about half of acute VTE events are precipitated by recognized acute environmental exposures such as trauma, surgery, other illnesses such as infections, or drugs such as oral estrogens. At these times, such prothrombotic interactions are postulated to exceed a threshold at which thrombosis occurs [2].

It is increasingly realized that venous and arterial thrombosis share multiple risk factors [3], including precipitation by acute infections, trauma, surgery [4] or oral estrogens [5]. After coronary heart disease (CHD) and stroke, VTE is the commonest cause of cardiovascular death and disability. It is therefore timely and appropriate to compare the historical epidemiological approaches to establishing reliable estimates of the contributions of postulated risk factors (and their interactions) to VTE and to arterial thrombosis (CHD and stroke).

To address the epidemic of CHD, which emerged in developed countries in the second half of the 20th century, several prospective studies of risk factors were established in North America, Europe and Australasia. Within 20 years, interactive risk factors, including age, male sex, tobacco-smoking, blood pressure and serum cholesterol, were identified in studies such as the Framingham Study, and observed to be common to stroke and peripheral arterial disease, as well as CHD [6]. Possible chance differences between findings in individual studies were addressed by systematic reviews and meta-analyses of individual participant data (IPD) by the Prospective Studies Collaboration [7,8]. Following the identification of random measurement error in estimates of an individual’s usual levels of risk factors as a major cause of under-estimation of their effects on risk, prospective studies of serial measurement of risk factors, such as the British Regional Heart Study, established that approximately 90% of inter-individual differences in risk of CHD were attributable to just three risk factors: tobacco-smoking, blood pressure and cholesterol [9]. Collaborative meta-analyses of IPD from randomized trials of blood pressure reduction and serum cholesterol reduction [10] have shown that reduction in levels of these risk factors results in reductions in risk of CHD and stroke, which are consistent with those predicted from prospective epidemiological studies, and consistent with causality for these risk factors. In parallel, prospective studies of national healthcare statistics have generally observed that about half of the decline in the epidemic of CHD and stroke is attributable to control of risk factors, while the other half is attributable to improved management [11].

Although a parallel epidemic of venous and arterial thrombosis in the second half of the 20th century was clearly apparent when major prospective studies of risk factors for coronary heart disease and stroke were established, few of these studies initially studied risk factors for VTE. From a synthesis of studies in older persons, Rosendaal et al. [12] reported an exponential increase in risk with age, and a higher risk in men. Recent interest in the association between arterial and venous thrombosis, which is probably due to common risk factors [3], led Ageno et al. [13] in 2008 to perform a meta-analysis of ‘arterial’ risk factors: tobacco-smoking, blood pressure, cholesterol and diabetes. The small number of VTE cases (for example, < 1000 in studies reporting smoking) resulted in non-significant associations, with wide confidence intervals for odds ratios, for smoking and cholesterol.

Clearly, the only way to establish such associations with confidence was to perform larger studies in generally healthy populations, to increase the number of VTE cases. A large case–control study (the MEGA study in the Netherlands), which included almost 4000 VTE cases and a similar number of controls, showed a dose- and time-dependent increase in risk with smoking habit [14], as well as a dose-dependent association with obesity [15].

Recent large prospective studies from the UK [16] and Denmark [17] have confirmed the association with smoking, while the Danish study has confirmed the association with obesity [18]. So, recent large studies are increasing the limited literature on the associations of common lifestyle risk factors for VTE, and it may be timely to update the 2008 meta-analysis of these associations [13]. Furthermore, it is timely to perform updated meta-analyses of non-modifiable risk factors such as age and sex [12], as recently performed for seasonality [19].

Both case–control and prospective cohort studies have their advantages and disadvantages [20]. Case–control studies are quicker and cheaper, can study many risk factors simultaneously and require smaller sample sizes. However, they do not involve as accurate a time sequence; while this is immaterial for studies of genetic risk factors, they cannot demonstrate causality for lifestyle risk factors. Survival bias may have little effect on deep vein thrombosis (DVT) where mortality is low, but may have a greater effect on pulmonary embolism where mortality is high. Other biases are possible and case–control studies can provide only approximate estimates of relative risk. The choice of control groups is also a consideration for environmental VTE risk factors [21].

While prospective studies are more expensive to establish and maintain, their long-term advantages for study of risk factors for VTE are a sound study design, coupled with an increasingly accurate ascertainment of VTE endpoints at a population level (routine objective confirmation of DVT and PE by non-invasive diagnostic methods), which are available in national databases. Thus, recent publications from large cohort studies established for studies of diet and health [17,18], general practice [16] and hormone replacement therapy [22,23] have generated valuable data. The value of studying the time sequence of VTE in large prospective cohort studies is illustrated by the Million Women Study in the UK. After surgery, 5419 women were admitted to hospital for VTE; the risk was greatest in the first 6 weeks, but persisted for 12 weeks and varied with the type of surgery [22]. In addition to studying risk factors over the time sequence of precipitating factors for VTE such as surgery, injury, medical illness or exposure to drugs such as estrogens or antipsychotics [16], prospective studies of changes in lifestyle risk factors for VTE, such as smoking, obesity, blood pressure and cholesterol, have high potential for estimation of causality, as previously reported for arterial thrombosis [9].

Large prospective studies, such as the Danish study, are important not only for establishing with confidence the associations of individual risk factors for VTE, but also for establishing with confidence their interactions [1] and for establishing and validating with confidence multifactorial risk scores for VTE in the general population [16]. For example, the UK Million Women Study has recently established that VTE risk not only increases with increasing body mass index, but that the associated excess risk is much greater after surgery than without surgery [23]. Also, the UK routine general practice data have been used to develop and validate a risk prediction score, based on simple clinical variables, which could be used to risk-assess patients before hospital admission, or before starting medication which might increase risk of VTE [16].

Finally, no matter how large an individual study is, the best estimate of the size and shape of an association of a risk factor with VTE, or of the interactions of multiple risk factors, will arise from collaborative meta-analyses, in which individual participant data are pooled. It is timely to progress such international collaborations, following the examples of collaboration on blood pressure [7], cholesterol [8,10] plasma fibrinogen [24] and other emerging risk factors [25] in arterial thrombosis. Such collaborations have recently been productive in risk prediction for recurrent VTE [26]. The time has come for international collaborative meta-analyses on prediction of first VTE.

Disclosure of Conflict of Interests

The author states that he has no conflict of interest.