Cerebral sinus thrombosis and oral contraceptivesBMJ 1998; 317 doi: https://doi.org/10.1136/bmj.317.7157.483 (Published 22 August 1998) Cite this as: BMJ 1998;317:483
There are limits to predictability
- Jan P Vandenbroucke, Professor of clinical epidemiology.
Clinical review p 520
In short succession two recently published papers have shown not only a very high relative risk of cerebral sinus thrombosis in users of oral contraceptives but an even stronger effect of contraceptives among women who carry hereditary prothrombotic conditions. 1 2 This raises several questions: Why this sudden wave of publications? Why the very high relative risks? What are the lessons for clinical practice and prevention?
The field of investigation into oral contraceptive use and the occurrence of venous thrombosis has been undergoing a rapid paradigm shift over the past couple of years. That oral contraceptives may cause venous thrombosis has been known since the 1960s, but ever since there have been controversies about the size of the risk, the role of bias in observational research, the role of hormonal dosage, the type of hormones implicated, and—particularly—the glaring absence of any biochemical explanation. Some have even doubted the reality of the association. The discovery of first the factor V Leiden mutation and then the factor II mutation has changed the picture.
These prothrombotic mutations strongly enhance the risk of oral contraceptives causing venous thrombosis, which, firstly, proves that the association is real and, secondly, offers the beginning of an explanation. For instance, different coagulation tests have shown that oral contraceptive use leads to “acquired activated protein C resistance” of the same nature as the factor V Leiden mutation3–7 For studying the mechanisms of coagulation underlying venous thrombosis by epidemiological and clinical means, the rare cerebral sinus thrombosis is a purer model than deep venous thrombosis of the legs or pulmonary embolism. In these two conditions several other chronic risk factors (obesity, trauma, surgery, immobilisation, etc)may confuse the picture, whereas these are not risk factors for cerebral sinus thrombosis. Thus cerebral sinus thrombosis might be caused almost entirely by general “circulating” factors in the blood, possibly aided by factors affecting the vascular wall. This may explain why about 90% of previously healthy young women with cerebral sinus thrombosis in the recent case-control series from the Netherlands and from Italy used oral contraceptives, compared with a population use of about 45% in the Netherlands and 30% in Italy.
These differences give rise to surprisingly high relative risks: 10-fold to 20-fold increases. These increases in risk are much higher than one would have expected from older series of patients. This is puzzling, since the recent studies reflect the era of modern low dose contraception. Part of the reason might be that the newer studies are methodologically more rigorous: for the first time we now have sizeable patient series limited to women who were previously healthy, were premenopausal, and were not pregnant, in the postpartum, or using other hormones.
Although cerebral venous thrombosis is rare, it seems almost entirely associated with hormone use in healthy non-pregnant premenopausal women,7 with women with hereditary prothrombotic conditions at especially high risk. Given its poor prognosis, in terms not only of mortality but also of permanent neurological deficits, we should ask again whether screening for prothrombotic defects might be useful in women considering taking contraceptive hormones. Earlier, when discussing venous thrombosis of the legs and pulmonary embolism, we urged caution: widespread screening of asymptomatic people with no family history would necessitate biochemical investigations in hundreds of thousands of young women, and advice against the use of oral contraceptives in tens of thousands, to prevent a single death.8Nevertheless, a cost benefit analysis in Germany has shown that if the price of the screening test could be lowered to under 18 DM (£6) widespread screening would be completely cost effective from the point of view of a health insurer: the cost of the test would be balanced by the cost of days in hospital, anticoagulation treatment, post thrombotic complications, etc.9 Such analysis does not take into account, however, the “cost” of less desirable or less effective modes of contraception to the individual (although the advent of newer modes of contraception may change that picture) or the burden of knowing that one carries a potentially deleterious gene. Given the rarity of cerebral venous thrombosis, it is unlikely to sway any cost benefit calculations on screening for prothrombotic conditions dramatically.
On p 520 Laffan and Tuddenham argue that it is now clear that venous thrombosis is a polygenetic condition and that this might change our view about screening: we should investigate a whole array of risk factors, not only coagulation factor V and II, but also factor VIII, etc, and arrive at a risk profile for an individual, possibly by a computer model mimicking the coagulation cascade.10 Such a risk profile might more specifically pinpoint the individual at greatest risk. This proposal puts venous thrombosis due to oral contraceptive use in the brave new world of predictive genetics. In one way it is a good working case to think about since the issues might be less emotionally charged than with screening for “cancer genes” and much more is known about genetic and coagulation mechanisms. Nevertheless, putting aside the practicalities, costs, and ethics, we might still wonder whether there is a limit to predictability. For weather forecasts it is accepted that no useful day to day predictions can be made beyond seven days since the meteorological system is too complex. May not the system of coagulation, anticoagulation, fibrinolysis, vascular wall factors, and other circulating factors such as cytokines be like meteorology?In risk factor epidemiology, as in the prediction of coronary heart disease, we know that the combination of several strong risk factors yields beautifully smooth risk functions but in the end poor specificity for the individual.11Would it be too hazardous a prediction to suggest that a complex coagulation model will face similar problems?