disease, but the mechanism remains obscure. Table 1 lists the major determinants of arterial and venous thrombosis.
The large number of determinants of thrombosis known today has led to the concept of thrombosis as a multigenic and multicausal disease, i.e. disease will only develop in the presence of several interacting determinants [15,16]. Therefore, in reviewing female hormones as risk factors of thrombotic
disease, we will also review their effect in the presence of other risk factors.

Oral contraceptives
Composition and mechanism of contraception
Most oral contraceptives contain an estrogen and a progestogen (indicated as ‘combination preparation’) and the same combination is taken every day for at least 21 days each cycle (indicated as ‘monophasic preparation’). In biphasic and triphasic preparations the dose of both compounds varies over the cycle of administration. There are single preparations that only contain a progestogen, which can also be used for parenteral depot, subcutaneous, intrauterine and intravaginal administration.
In addition to this, transdermal combined contraceptives have recently become available.
The contraceptive action is brought about by the progestogen, which acts by suppressing luteinizing hormone (LH). Nowadays the estrogen is added to prevent blood loss during the cycle (‘breakthrough bleeding’), which often occurs in progestogenonly preparations or combination preparations with a very low estrogen content. The failure rate with good compliance is very low.Combination preparations are by far the most used. Since oral
contraceptives first became licensed in 1959, the estrogen dose has been reduced. Early preparations contained 150 mg mestranol or ethinylestradiol, the latter of which is the estrogen in virtually all combination preparations currently used. Ethinylestradiol is a synthetic, slightly altered version of the naturally occurring estradiol, which is inactive when taken orally. The dose of ethinylestradiol has been stepwise reduced over the years, and nowadays most preparations contain either 50 mg or 30 mg ethinylestradiol, while some recently introduced brands contain even less, i.e. 20 mg or 15mg ethinylestradiol.
The progestogen content has also changed over time, but in this case the concern was over the chemical composition of the progestogen rather than the dose [17]. Early preparations contained lynestrenol or norethynodrel, derivatives of norethindrone or nethisterone that are also known as first-generation progestogens or estranes, and are not used much in oral contraceptives anymore. The estrane steroids, derivatives of testosterone, are constructed out of the basic steroid skeleton of 17 carbons (three 6-carbon rings and one 5-carbon ring, also called the gonane structure) and an extra carbon on the 13-position.
Second-generation progestogens are gonanes, and include norgestrel and levonorgestrel, of which evonorgestrel is widely used. Third-generation progestogens, which have the same basic steroid molecule as the gonanes, are desogestrel and gestodene. These are both also widely used, despite their
increased thrombogenicity. Finally, some progestogens are difficult to classify: norgestimate is categorized as a thirdgeneration gonane, but because after uptake it is in part rapidly converted to levonorgestrel, it may metabolically belong more among the second-generation progestogens.
Two substances used in oral contraceptives in combination with an estrogen need to be treated separately. First, the antiandrogen cyproteronacetate (a pregnane progestogen derived from progesterone), which is used for treatment of acne vulgaris, seborrhea or mild hirsutism, and which has a similar antiovulatory
action as a progestogen [18–20]. Secondly, drospirenone, which is an antimineralocorticoid, with anti-ldosterone
and anti-androgenic effects, which also inhibits ovulation [21,22].

Oral contraceptives and venous thrombosis
The first controlled study showing that oral contraceptives increase the risk of venous thrombosis was performed by the Royal College of General Practitioners in the UK [23]. From this case-control study, a 3-fold increased risk was reported for oral contraceptive users vs. non-users. Subsequent case-control studies confirmed the elevated risk, with estimates for the relative risk of users vs. non-users between four and 11 [24–27], as did large prospective follow-up studies in the 1970s [28–30]. Overall, all studies performed until 1990 showed a 3-fold increased risk [31]. An important finding from these studies was that the risk was not cumulative with longer use, i.e. the risk brought about by oral contraceptives was immediate and only lasted as long as oral contraceptives were taken.

The most recent studies, including a large multinational study by the World Health Organization, still show similar risks of venous thrombosis caused by oral contraceptives [32–35]. Even though the risk is clearly increased, at 2- to 6-fold, it is the absolute risk that is most relevant to women using oral contraceptives.
Since oral contraceptives, in contrast to postmenopausal hormonal replacement therapy, are used by young
women, the baseline risk (among non-users) is low, and the risk in oral contraceptive users, although increased, remains small in absolute terms. The absolute risk of venous thrombosis in women of reproductive age is less than one per 10 000 per year [32]. In oral contraceptive users, it becomes two to three per 10 000 per year [32,36]. For most women, these risks are so low that they accept them, and consider them less important
than the benefits of oral contraceptives, i.e. a highly reliable method of contraception. Still, because of the very large number of women who use oral contraceptives thrombotic events do occur, often followed by a post-thrombotic syndrome [37], and are sometimes fatal. Therefore, the search for the safest oral contraceptive remains highly relevant.
The risk of venous thrombosis is highest during the first year of use, reaching an absolute risk of 12 per 10 000 women per year (for second-generation progestogen containing oral contraceptives) [38,39]. However, there is no ‘duration of use’ effect, in the sense that risk accumulates with prolonged use. On the contrary, the effect is immediate and reversible, i.e. it returns to baseline shortly after discontinuation.
Nowadays, the most commonly used brands contain 30 mg ethinylestradiol or less. The use of very low doses of ethinylestradiol of 20 or even 15 mg is accompanied by poor cycle control in some women, i.e. spotting and breakthrough bleeding [40]. There is not much evidence supporting a reduced risk of venous thrombosis with these ultra-low dose formulations, and there are no clear time-trends of the risk going down: there was
a 4- to 6-fold increased risk in the earliest, and in the newest reports. Head-to-head comparison showed lower risk for 30 mg vs. 50 mg ethinylestradiol containing oral contraceptives in one study [41], but not in another study [42]. In the most recent comparison, in our ongoing Multiple Environmental and Genetic Assessment of risk factors for venous thrombosis (MEGA) study, including 612 women with a venous thrombosis before
age 50, 75% of whom were taking oral contraceptives at the time of the event, we observed a clearly, about 2-fold, lower risk for oral contraceptives containing 30 mg ethinylestradiol than for those with 50 mg ethinylestradiol.
Following the first three reports in 1995 which showed a higher risk of venous thrombosis in oral contraceptives containing the third-generation progestogens desogestrel or gestodene [42–44], a series of studies have confirmed a 2-fold higher risk for these progestogens compared with oral contraceptives containing
the second-generation progestogen levonorgestrel [1,45]. The risk appeared especially high in starters, who had a risk of over 30 per 10 000 per year when using an oral contraceptive with a third-generation progestogen [39].
Recently, it has been shown that oral contraceptives that contain cyproteronacetate confer a substantially increased risk of venous thrombosis, with an even higher risk than for thirdgeneration oral contraceptives. In a small group of women with idiopathic venous thrombosis, oral contraceptives containing cyproteronacetate conferred a 4-fold higher risk than those containing levonorgestrel [46]. In the MEGA study, the risk was increased 18-fold compared with non-users.
As yet, no data are available on oral contraceptives containing drospirenone. However, a series of reported cases of venous thrombosis after the introduction of an oral contraceptive containing this anti-ovulatory antimineralocorticoid has led Dutch general practitioners to caution against its use in the absence of sufficiently proven safety [47].
Whilst there are differences between brands of oral contraceptives with regard to thrombotic risk, there are also differences between women. While obesity increases the risk of thrombosis about 2-fold for body mass index (BMI)>30 kgm2 [48,49], overweight (BMI>25 kgm2) and obese (BMI>30 kgm2) women have a 10-fold increased risk of thrombosis when they use oral contraceptives [49]. In familial thrombophilia caused by
deficiencies of protein C, protein S or antithrombin, oral contraceptives greatly enhance the risk of thrombosis in carriers of one of these defects [50,51]. While these deficiencies are rare in the general population (0.020.2%), factor V (FV) Leiden and prothrombin 20210A are common (26%), although they are slightly weaker risk factors (relative risk 3–8). Heterozygous FV Leiden carriers have a 20–30-fold increased risk of thrombosis when they use oral contraceptives [32], and in homozygous carriers oral contraceptives confer an even higher risk [52,53]
(Table 2). Prothrombin 20210A carriers who use oral contraceptives were reported to have a 16-fold increased risk compared with non-users with the wild-type genotype [54]. High levels of several procoagulant factors (FII, FVIII, FIX, FXI, FX) confer a 2- to 3-fold increased risk when levels exceed the 90th percentile of the distribution in the population [55–59]. When data of women of reproductive age were re-analyzed in the Leiden Thrombophilia Study, the combination of oral contraceptive use and high levels (compared with non-users with
normal levels of the clotting factor), showed a 10-fold increased risk for high levels of FII and FXI, while for the other procoagulant factors the combined risk did not appear to exceed the separate risk of oral contraceptives and high levels [60].

Oral contraceptives and the risk of arterial
disease Many studies have confirmed the initial case reports on an association between oral contraceptive use and the occurrence of MI [7,61–68]. The most recent data are from the World Health Organization, reporting a 5-fold increased risk of MI, as well as a 3-fold increased risk of ischemic stroke [69,70]. In the recent RATIO (Risk of Arterial Thrombosis In relation to Oral contraceptives) study, it was shown that oral contraceptive users have a 4-fold increased risk of peripheral artery disease (PAD) [71].
There is little evidence that the lowering of the dose of the estrogen in oral contraceptives has led to a lower risk of arterial thrombosis. In a large population-based study on arterial disease in young women (including 248 women with MI, 203 women with ischemic stroke, 152 women with PAD and 925 controls, all aged less than 50), little or no difference was observed for oral contraceptives containing 50 mg or 30mg ethinylestradiol[71–73]. Third-generation progestogens (desogestrel and gestodene) have a favorable effect on the lipid profile [74]. It was therefore hoped that they would confer a lower risk of arterial thrombosis than oral contraceptives containing a second-generation progestogen, e.g. levonorgestrel. Some initial small studies showed a reduced risk, while others did not [69,75, 76]. Two recent large studies, the MICA study from the UK [77] and the RATIO study from the Netherlands [73], also yielded different results, with one showing a 2-fold increased risk for
oral contraceptives containing a third-generation progestogen relative to second-generation progestogens [77], and the other showing the opposite [73]. Although both studies were large, neither could exclude the absence of a difference. Since there were no obvious biases in either study that could explain this difference, it is most likely that the difference in risk of myocardial infarction between oral contraceptives with a third- or a
second-generation progestogen, if any, is small. In the RATIO study, there was no difference between oral contraceptives with a second- or a third-generation progestogen with regard to the risk of ischemic stroke, or the risk of peripheral artery disease [71,72].
The risk of myocardial infarction is much higher in oral contraceptive users who smoke or have hypertension than in those without such conventional risk factors [1,63]. In theWorld Health Organization study, no increased risk was found in women who had had a blood pressure screening before prescription [69], which corroborates a pooled analysis of two studies on the west coast of the USA that reported no increased risk [78]. In the RATIO study, the overall risk of MI for oral contraceptive users was increased 2-fold. However, it was much higher for women who also smoked (14-fold), had hypertension (6-fold), had hypercholesterolemia (25-fold), had diabetes (17-fold) or were obese (5-fold) [73]. For ischemic stroke, with an overall 2-fold increased risk, these interactive effects with conventional risk factors were also seen, although they were less striking: women who also had hypertension had an 8-fold increased risk, those who also smoked a 4-fold increased risk, and those with hypercholesterolemia a 11-fold increased risk [72]. Almost all young women with PAD are smokers [71], and
the combination of smoking and oral contraceptive use led to a 36-fold increased risk in the RATIO study [71]. Similarly, very high relative risks were found for the combination of oral contraceptive use with hypercholesterolemia (50-fold increased risk) and with diabetes (40-fold increased risk).
Contrary to the case of venous thrombosis, coagulation abnormalities play a minor role in arterial disease [79–82]. Women with FV Leiden or prothrombin 20210A do not have an increased risk of myocardial infarction when they start using oral contraceptives, beyond the risk in women without these mutations [73]. While in the RATIO study an increased risk of MI was observed for women with elevated levels of FVIII or FIX (2- to 3-fold increased risk for levels exceeding the 90th percentile of the distribution in the population), this risk was not
enhanced more than expected by oral contraceptive use (B.C.Tanis, unpublished data).

Hormone replacement therapy
Content and routes of administration
Hormonal replacement therapy has been prescribed to postmenopausal women for three reasons: relief from menopausal complaints, reduction of the progression of osteoporosis, and lowering of the risk of cardiovascular disease. Unopposed estrogens increase the risk of endometrial cancer, and therefore, except in women who have had a hysterectomy, most preparations contain an estrogen and a progestogen. The estrogens in oral preparations are usually conjugated estrogens retrieved from pregnant mare urine, or micronized estradiol. The progestogen mostly used in combination preparations is medroxyprogesterone acetate, a pregnane progestogen. Besides oral administration, the hormones can also be administered transdermally (by patches) and subcutaneously.
There is no doubt that estrogens relieve symptoms of the menopause, predominantly by reducing hot flushes, which,however, are also reduced by 50% with placebo [83]. It also is established that prolonged use of estrogens reduces the progression of osteoporosis and may increase bone density [84–86]. It is less clear whether this leads to a reduced incidence of fractures, although observational studies showed clear effects
[87,88]. Two recent randomized trials, however, yielded con-flicting results. In the Women’s Health Initiative (WHI), over 16 000 women received either estrogens and progestogens or placebo for over 5 years, and the risk of hip fractures was reduced by one-third [13]. In over 2000 women followed for more than 8 years, half of whom received placebo, in the Heart and Estrogen/progestin Replacement Study (HERS), no effect
on fractures was observed [89]. The main reason to prescribe postmenopausal hormone replacement therapy has been to reduce risk of cardiovascular disease, after several observational studies showed a strong favorable effect [87,88,90–94].
These studies have been criticized because women who chose to use hormones had a better cardiovascular risk profile than those who did not, e.g. they smoked less and exercised more [95–97].
Therefore, randomized placebo-controlled trials were initiated.

Hormone replacement therapy and risk of venous thrombosis
Because of an almost mythical belief in estrogens as a panacea for the postmenopausal woman, it took until 1996 before serious studies into thrombotic evidence were undertaken.
Then, several studies in rapid succession established a clearly increased risk of venous thrombosis for users of hormonal replacement therapy, with a 2- to 4-fold increased risk compared with non-users [10–12,98–103] (Fig. 1).
It has been postulated that the absence of a first-pass effect through the liver with transdermal administration might lead to less risk, but an increased risk has been shown for patches 1374 F. R. Rosendaal et al