Etiology of venous thrombosis: the need for small original studies


Frits R Rosendaal, Department of Clinical Epidemiology, Leiden University Medical Center, Leiden, the Netherlands.
Tel.: +31 71 5264037.


See also Lowe GDO. Epidemiology of venous thromboembolism: the need for large (including prospective) studies and meta-analyses. This issue, pp 2186–8 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, El-Galaly and co-workers show increased risks of venous thrombosis for individuals with non-0 blood groups and factor V Leiden, particularly when associated with smoking and obesity. My counterpart to this small debate, Dr Lowe, used these data to sound a cri de coeur for large prospective studies and meta-analyses into the epidemiology of venous thrombosis. While agreement may be inappropriate in a debate, I do not disagree with his laudatory words for the study of El-Galaly, nor with the idea that large studies or even meta-analyses may be useful, but I question whether these are the type of studies that are needed most and will advance our field most.

Ever since Virchow postulated his three groups of causes of thrombosis, namely stasis, changes in the composition of the blood and changes in the vessel wall, more and more risk factors for venous thrombosis have been discovered, mainly in the last few decades. It is illustrative to compare Virchow’s knowledge of the causes of thrombosis with the risk factors we know today. In his Gesammelte Abhandlungen zur Wissenschaftlichen Medicin (1856) [1] he describes the main groups of causes, which are (i) marantic thrombosis, as related to various kinds of serious illnesses; (ii) compression thrombosis, for instance from a malignant or tuberculotic mass; (iii) dilatation thrombosis, which includes thrombus formation in aneurysms and varicose veins; (iv) traumatic thrombosis, amongst which the most common was thrombosis after bloodletting, followed by thrombosis after amputation; (v) newborn’s thrombosis, mainly concerning the umbilical vein or the placenta; (vi) puerperal thrombosis, related to blood loss or uterine infection; and (vii) thrombosis secondary to infection of the vessel wall, particularly when pus entered the vessel.

These causes differ quite remarkably from those that are established today, of which the top 50 are listed in Table 1. Largely, these fall into several overlapping categories, such as trauma, surgery and illness on the one hand and factors affecting coagulability on the other, or environmental vs. genetic causes. Only major environmental causes, such as surgery, trauma and cancer, as well as pregnancy and puerperium, have been known for more than 50 years, and were already mentioned by Virchow. A prothrombotic effect of exogenous female hormones was reported first in the early 1960s, and all other causes were identified only in the last two decades.

Table 1.   Fifty causes of venous thrombosis
  1. PC, Protein C; PS, Protein S.

AgeChemotherapyNon-0 blood group
Major surgeryPsychotropic drugsAT deficiency
NeurosurgeryThalidomidePC deficiency
Orthopedic surgeryExogenous femalePS deficiency
ProstatectomyHormonesFactor V Leiden
TraumaAir pollutionProthrombin 20210A
Prolonged bed restHigh FIIFactor XIII val34leu
Venous catheterHigh FVIIISERPINC1 (rs2227589)
Plaster castHigh FIXFXI (rs2289252)
Long-haul travelHigh FXIFV (rs4524)
PregnancyHigh VWFGP6 (rs1613662)
MalignancyLow TFPIFXI (rs2036914)
HyperthyroidismHigh PCIVWF (rs1063856)
CushingHigh TAFISTXBP5 (rs1039084)
ObesityDysfibrinogenemiaHIVEP1 (rs169713)
SmokingLupus anticoagulant 
No alcoholHypofibrinolysis 

How did this remarkable progress come about, and by which methods were these causes identified? Well, not in large collaborative prospective studies or meta-analyses. Almost all were found in small focused association or family studies, a handful in genome-wide association studies and none at all in meta-analyses. Let’s examine a few of the most important findings more closely: ABO-blood group, APC-resistance and the synergistic effect of factor V Leiden and oral contraceptive use.

Non-0 blood group was the first genetic factor that was discovered as a risk factor for venous thrombosis, and is on a population level the most important. In the 1960s, a drug surveillance programme was run in the Boston area, and accidentally a deficit of individuals with blood group 0 was observed in patients using anticoagulant drugs. In a subsequent study, ABO blood group was studied in a series of patients with venous thrombosis, and the distribution was compared with control data from the population. Those with non-0 blood group had twice the risk of venous thrombosis of those with blood group 0, which was reported by Hershel Jick [2] in The Lancet in 1969. This was later confirmed, refined to a risk-enhancing effect of the A1 and B-alleles and shown to be mediated through levels of von Willebrand factor and factor VIII [3–5]. Since a majority of the population worldwide has non-0 blood groups, this is a risk factor responsible for over a third of all venous thrombotic events (population-attributable risk, PAR).

The discovery of APC-resistance was based on inexplicable laboratory findings in a single patient, as recounted by Björn Dahlbäck [6] in a historical sketch in the Journal of Thrombosis and Haemostasis. In a complex in-house functional protein C assay consistently abnormal results were found in this patient, which appeared dependent on the dilution of the plasma. In a sometimes hilarious story, Dahlbäck tells how he gave little priority to the study of this finding, which went on, with some intermissions, for several years. After a refreshing sabbatical in 1990, he obtained new blood samples from the patient, and shortly after the experiments were carried out that showed a defective anticoagulant response to activated protein C, after which ‘APC-resistance’ was reported in a hallmark paper in 1993 [7]. Soon after, it was shown that this prothrombotic abnormality was highly prevalent (3–5% of Caucasians), increased the risk of thrombosis several-fold, and was responsible for 15% of all venous thrombotic events [8].

The use of oral contraceptives increases the risk of venous thrombosis 4-fold. As tens of millions of women worldwide use oral contraceptives, this is of obvious public health importance. The association was reported by Jordan in 1961 [9], based on a single case, who was a nurse for whom he had prescribed the then recently licensed preparation for complaints related to endometriosis, and who developed pulmonary embolism immediately after initiation of therapy. Shortly after the finding that APC-resistance was caused by a mutation in factor V (factor V Leiden), we wished to estimate the risk of homozygous carriers. The estimate that was published of an 80-fold increased risk was based on seven homozygous individuals found in the Leiden Thrombophilia Study (LETS), a case–control study of 474 patients with venous thrombosis and 474 controls [10]. Remarkably, of the seven homozygous patients, six were women, and of the five women of reproductive age, three used oral contraceptives (60%, compared with around one-third of controls) [10]. This led Jan Vandenbroucke [11] to do a formal analysis of the joint effect of oral contraceptive use and factor V Leiden on the risk of venous thrombosis. The result, which became a textbook example of the analysis of synergy of risk factors, showed that while heterozygous carriership of factor V Leiden increased the risk of thrombosis 7-fold and oral contraceptive use increased the risk 4-fold, the joint presence led to a risk 35-fold higher than in those with neither risk factor.

These three examples show that major discoveries come from shrewd observations, sometimes in only one patient or a handful of patients. This is by no means restricted to research in thrombosis, but essentially true in all medical research. It has therefore been proposed that there are two ‘hierarchies’ of clinical research [12]. One is the hierarchy of evidence-based medicine, which refers to the strength of evidence of therapeutic interventions. Here, randomized controlled trials and meta-analyses of such trials confer the most confidence that an association is truly causal. However, such large studies are performed to prove, or refute, effects of interventions that were previously found in smaller studies, and not to make new discoveries. Insights into etiology and mechanisms come from studies with an opposite hierarchy, with case reports, case series and case–control studies at the top. The large collaborative studies and meta-analyses will only serve to obtain more precise estimates of the effects found in such studies.

It is fair to ask what the insights into etiology and pathogenesis have brought to clinical care. While the studies listed above have increased our insight into the development of thrombosis, they have not benefitted patients. At most, they have led to a flurry of diagnostic testing for prothrombotic abnormalities, at a major expense but with arguable clinical benefits [13]. This will not change with larger studies yielding more precise estimates or the discovery of genetic variants having negligible effects in individuals. Paradoxically, the major progress in understanding thrombosis occurrence with the discovery of all the risk factors listed in Table 1, has led to a widening of the gap between knowledge and applicability. The challenge for the next decades will be to bridge this gap, for which we do not so much need larger studies, but new study designs to integrate knowledge for individual prediction.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.