Hormonal contraceptives—oral

History

The fact that some women could develop thromboembolic complications as a result of taking oral contraceptives first emerged in 1961, although at that time the evidence was anecdotal and poorly quantified. The first reasonably quantified investigation conducted on a sufficient scale to merit conclusions was published in 1967 by the UK Medical Research Council [ ]. This and other large studies conducted during the early years (and considered in older volumes in this series) concluded that women using oral contraceptives ran a greater risk than non-users of developing deep venous thrombosis, pulmonary embolism, cerebral thrombosis, myocardial infarction, and retinal thrombosis. Later papers and case reports described deep venous thrombosis, portal venous thrombosis, and pulmonary embolism [ ]. The Boston Collaborative Drug Surveillance Program follow-up study of more than 65 000 healthy women in 1980–82 showed a positive association between current oral contraceptive use and venous thromboembolism (rate ratio 2.8); there was also a positive association between current oral contraceptive use and stroke or myocardial infarction [ ]. A UK study using data from 1978, by which time lower-dose products were increasing in use, pointed to an approximate doubling of the risk of thromboembolism compared with controls [ ]. The early 1980s were nevertheless marked by a series of critical papers that sought to question the entire concept of there being a link.

Much of the work on both sides of the argument was less than watertight. Some studies failed to consider the confounding effects of other susceptibility factors (notably smoking) or the likelihood of detection bias (particularly for venous thromboembolism, which is much more common in young women than myocardial infarctions or strokes). The results of some studies were also confounded by uncertainties in the history of drug exposure. A landmark paper to resolve the issue concluded that the link with venous thromboembolism in subjects without predisposition had been consistently observed in case-control and cohort studies [ ]. However, the evidence regarding myocardial infarctions, various types of strokes, and cardiovascular mortality was less consistent. By 1990, an international Consensus Development Meeting reached agreement on the following statement regarding the relation between oral contraceptive use and cardiovascular disease [ ]:

“The majority of epidemiological studies strongly suggest an association between current oral contraceptive use and certain cardiovascular deaths. Although the relative risk is increased, the absolute risk is small. Because the risk of myocardial infarction is apparent in current users, disappears on cessation of use, and is not associated with duration of use, there is no epidemiologic support for the hypothesis that risk of cardiovascular diseases is of atherogenic origin. Whether particular formulations or progestogens have qualitative advantages or disadvantages merits further study. Estrogens and progestogens interact at many levels, and in epidemiologic studies of users of combined oral contraceptives it is difficult to assign a risk to either component separately. Moreover, it is physiologically unsound to do so. Alterations in plasma lipid, carbohydrate, and hemostasis variables are of major importance for the development of cardiovascular diseases, and their concentrations can be influenced by sex steroids, including artificial steroids contained in oral contraceptives. The pharmacodynamic responses are dependent on not only the type and dosage of sex steroids, but also on intra- and interindividual variability in pharmacokinetics.”

Frequency

It has been confirmed that the incidence and mortality rates of thrombotic diseases among young women are low [ ]. However, the risk is increased by oral contraceptives, and there is variation in the risk, depending on the accompanying progestogen. The spontaneous incidence of venous thromboembolism in healthy non-pregnant women (not taking any oral contraceptive) is about five cases per 100 000 women per year of use. The incidence in users of earlier formulations is about 15 per 100 000 women per year. The incidence in users of third-generation formulations is about 25 per 100 000 women per year: this excess incidence has not been satisfactorily explained by bias or confounding. The risks of venous thromboembolism increase with age and are likely to be increased in women with other known factors for venous thromboembolism, such as obesity. The risk in pregnancy has been estimated at 60 cases per 100 000 pregnancies.

The Medicines Commission of the UK reviewed all currently available relevant data and found an incidence of venous thromboembolism of about 25 per 100 000 women per year of use [ ].

The incidence of venous thromboembolic disease in about 540 000 women born between 1941 and 1981 and taking oral contraceptives was 4.1–4.2 cases per 10 000 woman-years [ ].

In another study, the figures ranged from 1895 events per 100 000 women-years when norgestimate was used to 3969 per 100 000 women-years when desogestrel was used [ ]. Although the authors did not find the difference statistically significant, it runs parallel to findings from other work regarding a higher risk when third-generation progestogens are used.

In women aged 15–29 years who used oral contraceptives containing third-generation progestogens, venous thromboembolism was twice as common as arterial complications. In women aged 30–44 years of age the number of arterial complications exceeded the number of venous complications by about 50%. However, in women under 30 years, deaths from arterial complications were 3.5 times more common than deaths from venous complications and in women aged 30–44 years 8.5 times more common. Women over 30 years of age who take oral contraceptives containing third-generation progestogens may have a lower risk of thrombotic morbidity, disability, and mortality than users of second-generation progestogens. However, a weighted analysis such as this does not result in any consistent recommendation of a particular progestogen type.

Nevertheless, some groups have produced data from their own systems that fail to confirm this. Some of these studies, including unpublished data circulated to experts for purposes of special pleading, have used selected material, and one can only consider them flawed. On the other hand, Jick et al. may be entirely right in their finding that, insofar as the special risk of idiopathic cerebral hemorrhage is concerned, no material difference in risk has been demonstrated between products of the second generation and those of the third generation [ ].

In another study users of oral contraceptives with second-generation progestogens had a 30% greater increased risk of thrombotic diseases, a 260% greater increased risk of thrombotic deaths, and a 220% greater increased risk of post-thrombotic disability than users of oral contraceptives with third-generation progestogens [ ].

Cyproterone acetate in combination with ethinylestradiol is indicated for the treatment of women with severe acne and moderately severe hirsutism. This product has been associated with a greater risk of venous thromboembolism than oral contraceptives. However, in a rigorous case-control study the risk of venous thromboembolism with cyproterone acetate + ethinylestradiol was not significantly greater than the risk in women who took conventional oral contraceptives [ ].

Effects of dosage and formulation

As noted above, the early recognition of thromboembolic complications had repercussions for the formulation of the oral contraceptives; the progestogen and estrogen contents were both progressively reduced, and in 1969/70 mestranol was replaced by ethinylestradiol, on the grounds that the dose could thereby be halved (although it is not at all certain that this reduced the estrogenic contribution to thromboembolic events). By the late 1980s, when a major cohort study based in Oxford examined the problem [ ], products containing estrogen 50 micrograms accounted for about 70% of the woman-years of tablet use, most of the remainder being accounted for by lower doses.

By the mid-1990s it was reasonably well proven that progressive dose reduction had reduced the risk of thrombotic complications [ ]. Epidemiological studies showed that users of low-dose combined products had small, and often statistically non-significant, rises in the risks of myocardial infarction [ , ], thrombotic stroke [ ], venous thromboembolism [ ], and subarachnoid hemorrhage [ , , ]. Several of these studies compared the risks presented by different doses and found somewhat higher rates for products containing more than 50 micrograms of estrogen but somewhat lower rates among women currently using the lower-dose formulations [ , , ].

A group in The Netherlands has stressed the fact that even though the risk of venous thrombosis is small in absolute terms, oral contraceptives form the major cause of thrombotic disease in young women. The risk is higher during the first year of use (up to one per 1000 per year), among women with a prothrombotic predisposition, and with third-generation progestogens [ ].

Presentation

The commonest presentation is deep venous thrombosis in the leg, which can lead to pulmonary embolism. Fatal pulmonary embolism has even been reported after intravenous injections of conjugated estrogens [ ]. Despite the improvement noted with reduced doses, incidental case reports of severe cardiovascular events during the use of low-dose products have continued to appear. They include incidents of cerebral venous thrombosis and subarachnoid hemorrhage, fatal central angiitis, sinus thrombosis, and cerebral ischemia. In one series of 22 cases of cerebral infarction involving either arteries or veins, all the oral contraceptives that had been used contained a low dose of ethinylestradiol [ ]. Thromboembolism in other veins, such as the hepatic vein, that is Budd–Chiari syndrome, has occasionally been reported [ ]; the first 10 such cases were reported as long ago as 1972 and the increasing number since then has at times raised some concern [ , ]. A case of renal vein thrombosis has also occurred [ ]. Incidental reports continue to appear of thrombotic incidents in relatively unusual forms, including a further case of mesenteric thrombosis leading to intestinal necrosis [ ] and a report of fatal pulmonary embolism following intravenous injections of conjugated estrogens [ ].

The incidence of hepatic veno-occlusive disease in 249 consecutive women treated with norethisterone who underwent allogenic hemopoietic stem cell transplantation was 27% compared with 3% in women without this treatment [ ]. One-year survival rates were 17% and 73% in patients with (n = 24) or without veno-occlusive disease (n = 225) respectively. Because of this adverse reaction, norethisterone should not be used in patients undergoing bone-marrow transplantation. Heparin prophylaxis does not affect the risk of death from veno-occlusive disease.

Susceptibility factors

By 1980 it was considered clear that the risk of thromboembolic events was further increased under particular conditions. It was higher in smokers, in older women, and in obese people, and appropriate warnings were issued. The fact that these warnings to a large extent eliminated the high-risk individuals who had formed part of the early population of oral contraceptive users means that data from the early period cannot be used to provide a valid historical comparison with later findings [ , ].

Ethnicity

Much of the evidence on the occurrence of thromboembolic complications with oral contraceptives or hormone replacement therapy has been gathered from European or American populations, and it can be helpful to identify data from other parts of the world, where factors such as body weight, climate, or diet could affect the incidence. When a Japanese group sought to obtain national data by mailing questionnaires to a large number of institutes monitoring these forms of treatment, 771 (71%) of 1083 institutes responded [ ]. Follow-up questionnaires were sent to 39 institutions that reported having experienced in all 53 cases of thromboembolism during hormone therapy; 29 cases related to oral contraceptives and 13 to hormone replacement therapy, while 11 had taken other forms of hormone treatment. Of the 29 patients taking contraceptives, eight had developed arterial thromboembolism (including two with myocardial infarctions).

Age

The increase in risk with age is clear [ ], although the underlying risk of cardiovascular disease also rises as age progresses. The US Food and Drug Administration has concluded that the benefits may outweigh the harms in healthy non-smoking women over age 40, and it has in most countries been common for over two decades to advise reticence in the use of oral contraceptives after the age of 35.

Obesity

Obesity has repeatedly been shown to play a role, and its relevance to particular types of complication has been demonstrated. The Oxford Family Planning Association’s 1987 data showed that the risk of myocardial infarction or angina increased significantly with weight [ ].

Smoking

Smoking has been very clearly incriminated as a susceptibility factor for thromboembolism and arterial thrombosis in women taking the oral contraceptive, and its apparently synergistic role has been well defined and quantified [ , ]. The 1989 case-control analysis of the RCGP cohort study estimated the relative risk of myocardial infarct during current oral contraceptive use at only 0.9 for non-smokers, but at 3.5 for women smoking under 15 cigarettes per day, and as much as 21 for users of more than 15 cigarettes per day [ ]. Smoking increases not only the risk of myocardial infarction among oral contraceptive uses [ ], but also the risk of angina pectoris [ ], thrombotic stroke [ , ], and subarachnoid hemorrhage, and it can double or treble mortality [ ]. There has also been further confirmation that the effect is dose-related: light smokers have twice the risk of coronary heart disease and heavier smokers have up to four times the risk, compared with non-smokers; cessation of smoking is accompanied by a reduction in risk of coronary heart disease to the level prevalent among non-smokers within 3–5 years [ ].

Smoking contributes to effects on the procoagulation process in young women [ ]. The effects of oral contraceptives on the coagulation system are much greater in smokers than non-smokers [ ]. Oral contraceptive users who were smokers generally have significantly lower fibrinolytic activity than non-smokers [ ], but not consistently [ ].

Other factors

Women of blood group O have less of a risk of thromboembolism [ ]. The risk of thromboembolic complications may be greater where there is a history of diabetes, hypertension, and pre-eclamptic toxemia. In some studies there has been an association with type II hyperlipoproteinemia, hypercholesterolemia, and atheroma [ ]. Hypertension may be an additional susceptibility factor when considered in relation to oral contraceptive use.

Mechanisms

Study of the mechanisms that might underlie the link between oral contraceptives and thromboembolic events is of importance in developing safer formulations, but also in identifying, if possible, individuals at particular risk who should be advised to change to alternative contraceptive methods.

The 1990 international consensus statement cited above noted that: “Oral contraceptives induce alterations in hemostasis variables. There are changes in the concentrations of a large number of specific plasma components of the coagulation and fibrinolytic systems, although usually within the normal range. It is conceivable that these effects are estrogen-mediated because they have not been demonstrated in progestogen-only preparations. There is a dose-dependent relationship in the case of estrogen, although in combination tablets, the progestogens might exert a modifying effect. Further attention should be given to changes in factor VIIc and fibrinogen induced by oral contraceptives and also to the association between carbohydrate metabolism and fibrinolysis.”

Quantification of coagulation factors is notoriously difficult, because of the interrelations among the various components of the coagulation cascade, the broad range of normal values, and considerable inter-laboratory variability [ ]. This variability is illustrated by a WHO study of users of combined oral contraceptives, conducted on several continents, which showed statistically significant differences among clinical centers in prothrombin time, fibrin plate lysis, plasminogen, and activated partial thromboplastin time [ ]. Effects also vary between different populations, users of different doses, users of different products, and tests performed at different periods of the medication cycle [ , ].

The term “hypercoagulability” has been used to describe a supposed pre-thrombotic state, identifiable by certain changes in the hemostatic system, but to date there is no broad-spectrum laboratory test for assessing the risk of thrombosis in a given individual, although coagulation changes in vitro have sometimes been regarded as proof of a thrombotic state. Deviations in laboratory data from patients with thromboses have often been interpreted as demonstrating the cause of the thrombosis, whereas they may simply be a consequence.

Despite the variations that are found, the overall conclusion is that oral contraceptives cause an increase in coagulation factors I (fibrinogen), II, VII, IX, X, and XII, and a reduction in antithrombin III concentrations, which would be expected to predispose to venous thromboembolism, especially if not counterbalanced by an increase either in fibrinolytic activity or of other inhibitory proteins of the coagulation, such as protein C [ ].

There is also fairly strong evidence that immunological mechanisms play a role in thrombotic episodes associated with oral contraceptives, especially when they occur in the absence of susceptibility factors for vascular disease [ ], although this has been contested [ ]. In one series of reports on cerebral infarction, circulating immune complexes and/or specific antihormone antibodies were found in 15 of 20 patients [ ]. In a large series of women with venous or arterial thrombosis, anti-ethinylestradiol antibodies were absent in nonusers but present in 72% of users; they were also present in 33% of healthy oral contraceptive users without thrombosis [ ]. In half of the cases there were both anti-ethinylestradiol antibodies and a history of smoking, which were found jointly in half of the cases.

There is a significant rise in fibrinogen concentrations during the early months of oral contraceptive use, and concentrations return to baseline after withdrawal [ ]. Prolonged use of oral contraception also seems to lower concentrations of antiaggregatory prostacyclin [ ].

Some work that was considered to show severe acquired plasma resistance to activated protein C among users of third-generation (as opposed to second-generation) products was re-examined by a French group [ ]. In their view the technical measures used to demonstrate the effect of activated protein C introduced a bias of interpretation and hence false results; they further argued that such a test cannot demonstrate the presence of a raised thromboembolic risk in asymptomatic women taking these contraceptives, since it is nonspecific and subject to changes in the plasma concentrations of many coagulation factors that are themselves increased or decreased by estrogens and progestogens. They pointed, for example, to protein S [ ], changes in which account for the differential effect of oral contraceptives on Rosing’s assay [ ], but which are in their view irrelevant to issues of thromboembolic risk with oral contraceptives; the androgenic potential of the progestogen may further counteract the effect of estrogens in the test. More generally, in such a complex situation in which there is a “modification of the modification”, there is no hemostasis-related test that provides a risk indicator for thrombosis. This argument is sound, but it naturally remains theoretical; the question of thromboembolism with the third-generation products must, as pointed out above, be resolved on the basis of epidemiological data, and certainly those data strongly point to an increased risk.

Relative roles of estrogens and progestogens

The risk of thrombosis is closely associated with the estrogen component [ ] for both arterial and venous events and with the progestogen for arterial events; however, if a particular progestogen is metabolized to estrogen or raises estrogen concentrations, it will make a contribution to venous complications. Estrogen alone has after all been incriminated as a cause of thromboembolism when given to men [ ]; the risk of puerperal thromboembolism after estrogen inhibition of lactation has been shown in several studies [ ]; non-contraceptive estrogens clearly increase the risk of acute myocardial infarction in women under 46 years of age [ ]. Changes in coagulation factors appear to be related to the estrogen dose [ , , , ]. Progestogen-only formulations do not have any significant effects on the coagulation system.

Prognosis

Despite the vast literature on thromboembolic complications of oral contraceptives, little attention has been paid to factors that may determine the ultimate prognosis and risk of death. Data from the Swedish Adverse Reactions Monitoring Bureau and other sources have been used to study this question in regard to pulmonary embolism, as well as estimating the incidence [ ]. Over 36 years (during which the spectrum and usage of oral contraceptives naturally changed) 248 cases of suspected pulmonary embolism were reported. The presence of thromboembolism was confirmed in all fatal cases and 83% of non-fatal cases. The medical records showed that the presence of nausea or abdominal pain, age above 35 years, concomitant treatment with other drugs that increase the risk of thromboembolism, vein or lymph vessel malformations, and a deep vein thrombosis above the knee were positively associated with a fatal outcome. Chest pain and previous use of a combined oral contraceptive were negatively associated with a fatal outcome. Using pharmacy records to estimate sales, the incidence of verified pulmonary embolism was calculated as 1.72 per 100 000 treatment years; the figure for fatal cases was 0.25 per 100 000 treatment years.

Third-generation oral contraceptives and thromboembolism

As noted above, there are reasons to doubt whether the third-generation oral contraceptives are indeed safer than their predecessors in respect to thromboembolism and substantial grounds for believing that they present greater risks.

The first reason is theoretical. The demonstrated effects of the new substances and combinations on lipids and carbohydrates do not have any major relevance to the thromboembolic process. The latter is linked primarily to changes in the hemostatic system and blood coagulation, involving platelet aggregation, coagulation factors, fibrinogen concentrations, and blood viscosity.

The second reason is kinetic. It is true that the dose of estrogen (probably the main instrument in inducing thromboembolism) has been kept to a minimum, but the third-generation progestogen gestodene tends to accumulate in the system with continued use, and the concentrations of ethinylestradiol increase simultaneously; this increase is due to the ability of gestodene to inhibit cytochrome P450 and therefore to inhibit the breakdown of estrogen, as well as its own metabolism [ ]. Similar findings emerged with desogestrel, although they were somewhat less marked [ ].

The third reason is hematological. The third-generation contraceptives have greater adverse effects on the clotting system than those of the second generation. In particular, women using the third-generation products have a greater resistance to activated protein C [ ], a shift that is associated with a higher risk of thrombosis.

The fourth reason is epidemiological. During the period 1987–88, when the third-generation products were relatively new, anecdotal reports of thromboembolic events appeared, including at least one death, and partly for this reason a series of large controlled studies were set up. The findings of three such studies (British, European, and global) became available to the drug control authorities late in 1995 and were subsequently published. The UK Committee on Safety of Medicines, considering all three, concluded that the risk of venous thromboembolic events in these third-generation oral contraceptives was about double that in users of the previous generation of products using the older progestogens (30 as opposed to 15 per 100 000 woman-years, the risk in healthy women being only five per 100 000). Despite the different populations studied, the individual studies produced broadly consistent results. The global study on four continents found a relative risk of 2.6 when comparing the desogestrel/gestodene products with the older variety, while the European study found a relative risk of 1.5–1.6. Various later papers pointed in the same direction. In Denmark, there was an increase in hospital admissions for primary venous thromboembolism in young women coinciding with the introduction of the third-generation products [ ]. Papers from The Netherlands have confirmed the main trend [ , ].

Authoritative reviews and editorials have further confirmed the correctness of the above findings. There has been some criticism of the individual studies on various points of detail, but it is difficult to see that this in any way undermines what is now very consistent evidence that the third-generation oral contraceptives increase the risk of thromboembolic events to a substantially greater degree than previous products. Some work that has been advanced as pointing to the safety of third-generation products [ ] proves to relate primarily to the second-generation combinations, with only a few late entrants using the more recent oral contraceptives, and other work was performed on a very small scale.

The study of adverse reactions generally relates to current and emerging issues. However, now and again it can be instructive to look back into recent history. When a drug problem has been fairly clearly defined, and particularly when it has for a time been the subject of debate and even frank controversy, one can learn something from the processes involved. How did the facts become known? Why did the controversy emerge? And could the risk have been detected and eliminated earlier?

Since their appearance in the late 1950s, oral contraceptives have gone through several stages of development. What are now in retrospect referred to as first-generation oral contraceptives were high-dose combinations of progestogens (more particularly norethynodrel, norethisterone, and lynestrenol in doses of 2.5 mg or more) and the estrogen mestranol 75 micrograms. A decade later a second generation emerged, with substantially lower doses, commonly half of those used earlier and some new progestogens, notably the more potent levonorgestrel. Finally, in the early 1980s some manufacturers introduced so-called third-generation products, a particular characteristic of which was the use of entirely new, very potent progestogens, among them desogestrel and gestodene. Clinical studies of contraceptives that contained gestodene and desogestrel suggested that they are very similar to one another, although differences in dosage and potency could account for reports that products that contain gestodene provide better cycle control [ ].

Almost from the earlier years, the risk of thromboembolic complications among users of “the pill” was recognized, and by the mid-1960s it was well documented [ , ]. Progressive reductions in dosage, in particular that of the estrogenic component, during the period that first- and second-generation products held sway were widely regarded as having reduced this risk to manageable proportions, although it was not eliminated. The relative risk with first-generation products was highly variable (2–11), but the best work in the UK and the USA fairly consistently reached an estimate of 4–6 [ ]. With the second-generation products the relative risk of thromboembolic complications was again variously estimated, but a large cohort study published in 1991 set it at 1.5 with products containing the lowest doses of estrogen, and 1.7 with products containing intermediate doses of estrogen [ ].

The fact that both prescribers and users of medicines are likely to anticipate that new drugs will be in some way better than those that have gone before means that both groups are in principle receptive to potentially spurious claims and suggestions. By the time the third-generation oral contraceptives were marketed, this type of contraception had been around for a quarter of a century; the risk of thromboembolism, the most widely publicized problem in the field, seemed by that time to have receded with progressive reductions in dosage. There was every reason to hope that it would recede further with the newest generation of products. That expectation was further nurtured by the even lower doses latterly attainable. It also seems to have been fostered by some of the suggestive promotion that appeared, although that in fact related as a rule merely to an improved lipid spectrum, which in turn raised the theoretical possibility, also discussed but not documented by some clinical investigators [ ], that arterial and cardiac risks might be less.

What in fact happened was that by 1989 alarm bells began to ring in Germany, where the regulatory authorities were alerted to the submission of an unusually high number of spontaneous reports of thromboembolic complications thought to be associated with the new products. Cases continued to accumulate, long-term studies already begun were completed, and in 1995 Britain’s then Committee on Safety of Medicines made a public statement to the effect that the risk of thromboembolic complications among hitherto healthy users of third-generation products was approximately twice than that seen with second-generation products [ , ]. The studies in question, including work by the World Health Organization and others were subsequently published and confirmed that conclusion, as did later work [ ]. It was further reinforced by others [ ], who worked on a smaller scale but provided well-documented evidence that while a factor V Leiden mutation or a biased family history could increase the risk in individual cases, they did not explain the higher thrombosis risk seen with a product based on desogestrel than with contraceptives that incorporated levonorgestrel, norethisterone, or lynestrenol.

Why was the particular risk of the third-generation contraceptives identified so late? These third-generation products had been in development since the late 1970s and the first had been marketed in 1981–82, some 14 years before the Committee on Safety of Medicines issued its statement. Could society not have done better and thereby reduced the risks to which women were exposed? There are two principal answers, both of them at least partly in the affirmative.

The first is that products of this type could well have been entered at an earlier date into large studies of oral contraception and their effects. A series of university centers around the world, as well as bodies such as Britain’s Royal College of Physicians and Royal College of General Practitioners, have throughout the oral contraceptive era either sponsored or participated in prolonged cohort and case-control studies of these products. Experience with data on thromboembolism suggests that significant data are likely to be obtainable in a cohort study of manageable size within some 5–7 years. The use of third-generation products may have been small in the early years, but they were aggressively promoted in major oral contraceptive markets to ensure rapid growth, in all probability sufficient to provide adequate recruitment. One would hesitate to argue that such studies should be a universal condition of the marketing of drugs, but when the products concerned have immense social significance and considerable potential for good and harm, as the oral contraceptives do, and when the compounds involved are entirely new, there is at least a sound medical reason for such work in every case. That work was performed with successive forms of the earlier oral contraceptive products, in which dosages were progressively reduced, and there was particular reason to set it in motion on the introduction of products that contained new chemical components with some significant structural and pharmacological differences from the older progestogens. A straightforward statistical calculation shows that an early cohort study involving some 30 000–50 000 women taking a third-generation product could within 2 years have shown the degree of increase in the thrombotic risk, which was actually not elicited until much later.

The second answer with respect to the earlier acquisition of risk data must come from the laboratory. Not from animal studies, which in this field are of very restricted value, but from biochemical and particularly hematological work. When during the 1990s various groups began to examine in detail the effects of the third-generation contraceptives on processes related to the clotting system, they identified a series of properties that could very well explain an increased incidence of thrombosis. The first of these was an increase in circulating concentrations of factor VII produced by the desogestrel plus estrogen combination, which was some 20–30% higher than that seen with a second-generation product based on levonorgestrel [ ]. The methods used to carry out this work were available before 1988 [ ], and it is not at all clear from the published material whether there was a failure to compare the two generations in this respect at an early date, or whether such work was performed and either overlooked or misinterpreted.

A second finding related to the effects of activated protein C on thrombin generation in low-platelet plasma via the intrinsic or extrinsic clotting pathways. Using a method developed on the basis of work first published in 1997 [ ], a Dutch group in Maastricht found that all types of combined oral contraceptives induced acquired resistance to activated protein C. With the third-generation contraceptives, however, the effect was significantly more marked than with those of the second generation: in other words, these drugs significantly reduced the ability of activated protein C to down-regulate the formation of thrombin [ ]. However, this work only became feasible in the late 1990s.

A third underlying mechanism seems to involve a reduction in concentrations of free protein S, again more pronounced with third-generation products. When protein S falls, the antifibrinolytic effect of the so-called thrombin-activated fibrinolysis inhibitor is increased; in other words, fibrinolysis is impeded, with an increased risk of clotting problems [ ]. Again, however, these methods were not available when the third-generation products were launched.

The laboratory findings therefore suggest that a greater thrombosis-inducing effect of the third-generation oral contraceptives can be explained and even anticipated on the basis of known mechanisms. Not all the relevant methods were available in the early years, but that relating to factor VII most certainly was. It is unfortunate, to say the least, that such work was either not performed or not properly interpreted.

All in all, had a combination of hematological methods and field studies been initiated sufficiently soon, the increased risk of thromboembolism with the third-generation oral contraceptives could have been detected some years earlier, sufficient for society to take decisions on the benefit-to-harm balance of these drugs before so much needless injury was incurred.

The third-generation oral contraceptives: a judicial assessment

It was extraordinary to find a major epidemiological dispute regarding drug safety being handled by the High Court in England in late July 2002, when the Court handed down its decision regarding thromboembolic events induced by the third-generation oral contraceptives [ ]. Essentially, a group of women who claimed to have been injured as a consequence of having using this latest version of “the pill,” based on two new progestogens, had sought to reclaim extensive damages from the manufacturers, since in their view the product did not possess the degree of safety which, in the words of European law, the user was legitimately entitled to expect. Since the safety achieved with the widely used products of the second generation was so widely regarded as acceptable, the Court had to decide whether the newer products had significantly failed to meet that standard. Faced with a long procession of expert epidemiological witnesses from both sides, and with some flat contradictions, the judge was obliged to rule on their arguments.

However, that it was an English court in which the issue came to be debated was not surprising, for it was in England that the Committee on Safety of Medicines had written to prescribers in 1995 stating that three unpublished studies on the safety of combined oral contraceptives in relation to venous thromboembolism had indicated about “a two-fold increase in the risk of such conditions” compared with the preceding generation of products. This issue of a “two-fold increase” became crucial to the case. “For reasons of causation,” as the Judge put it, the claimants had accepted the burden of proving that the increase in risk was not less than two-fold.

In fact, the English authorities, having rejected a vigorous defence of these products by the manufacturers, were by 1999 speaking more precisely of an increase in risk, compared with the earlier products, “of about 1.7–1.8 after adjustment”, which was “not fully explained by bias or confounding”; appropriate label warnings were therefore imposed. These new warnings, summarized, said that an increased risk associated with combined oral contraceptives generally was well established, but was smaller than that associated with pregnancy (60 cases per 100 000 pregnancies). In healthy non-pregnant women who were not taking any combined contraceptive it was about five cases per 100 000 woman-years; in those taking the second-generation products it was about 15; and for third-generation products it was about 25. By September 2001, the European Union’s Committee on Proprietary Pharmaceutical Products had formed its own view, and here too it was concluded that the “best estimate of the magnitude of the increased risk is in the range of 1.5–2.0.”

In Court to support the claimants, Professor Alexander Walker assessed the relative risk of the third-generation products at 2.2, Dame Margaret Thorogood at 2.1, and Professor Klim McPherson at about 1.9. The experts for the defendants took the view that the relative risk was well below two, and could well be zero. As Mr Justice Mackay noted, having listened to these experts: “… the debate between them has been unyielding, at times almost rancorous in tone, and with a few honourable exceptions … devoid of willingness to countenance that there may be two sides to the question. So, science has failed to give women clear advice spoken with one voice.”

There was also fundamental disagreement on confidence intervals when calculating relative risks in such matters: “The Defendants say that to establish causation in the individual, and therefore a relative risk which is greater than two, there must be seen not just a point estimate but also a lower confidence interval which is greater than two in order for the result to be significantly different from two.”

The Court was faced with “a series of studies with different point estimates and largely overlapping confidence intervals. Time after time experts have had their attention drawn to point estimates from studies that appear, to the layman’s eye, to be very different. Almost invariably they have dismissed those apparent differences by reference to the overlapping confidence intervals, saying that the figures are statistically compatible and there is no significant difference.” Confronted with such material, the Court chose to set aside as inexact and theoretical much of the statistical rhetoric. Having done that, the Judge felt himself in a position to emerge “from that forest into broader more open country where the simpler concept of the balance of probabilities rules.” Constructing his judgement in that way, Mr Justice Mackay advanced in the course of 100 pages to the conclusion that the claimants had failed to demonstrate a doubling of the risk. In his view, “the most likely figure to represent the relative risk is around 1.7.”

This extraordinary and wise judgement merits most careful reading by anyone concerned to understand the safety issues surrounding oral contraceptives. First, because of the insight that it demonstrated into the manner—not always edifying—in which evidence in this vital matter has been adduced, interpreted, and argued over in the course of more than a decade. Secondly, because it arrived, through a process of tight reasoning, at what was at the time the most reliable conclusion we had. It seems beyond all possible doubt that the third generation of oral contraceptives is primarily characterized by an increased risk of thromboembolic complications. Whether that risk is great enough to warrant financial compensation is a matter for lawyers to decide. However, given the lack of any tangible benefit to the user, the risk is clearly significant in human terms, and it is hard to see that there is any valid reason at all for continuing to use these products.

Frequency

West German data over the period 1955–80 showed no community-wide increase in the incidence of ischemic heart disease, cerebral vascular embolism, or pulmonary embolism, despite the rapid growth in oral contraceptive use (and the prevalence of high-dose products) during much of that period [ ]. The Oxford Family Planning Association’s 1989 paper on its cohort study, which had followed up more than 17 000 women for an average of nearly 16 years, found no significant overall effect of oral contraceptive use on mortality, with a relative risk of 0.9 [ ]. Mortality from diseases of the circulatory system had slightly increased; the relative risk of death from ischemic heart disease in current or past oral contraceptive users was 3.3 (95% CI = 0.9–17.9), while data on fatal cerebrovascular disease were too few to be interpreted.

Similarly, a massive Finnish mortality study, covering 1 585 000 women-years of oral contraceptive use and two million women-years of copper-bearing intrauterine device use, showed no increase in relative risk among oral contraceptive users for myocardial infarction or cerebral hemorrhage deaths; however, there might have been an increased risk of death from pulmonary embolism among users of oral contraceptives [ ].

An analysis of the cardiovascular mortality risk associated with low-dose oral contraceptives (under 50 micrograms of ethinylestradiol) in the USA showed that among non-smokers and light smokers the mortality among current oral contraceptive users was likely to be lower than the mortality due to pregnancy [ ]. Only among heavier smokers over 30 years old did the risk of oral contraceptive use exceed the risk of pregnancy. The researchers noted that in countries with higher maternal mortality rates than the USA, even older women who are both heavy smokers and oral contraceptive users would have a lower mortality risk than that associated with pregnancy.

A 5-year case-control study involving all Danish hospitals has once more quantified the thromboembolic risks associated with oral contraceptives as a whole; the risk with third-generation products was some 30% higher than with second-generation products (RR = 1.3; CI = 1.0, 1.8) [ ]. However, data on cerebral thrombosis from the same study showed that with third-generation products the mean risk was some 40% lower than with second-generation products (RR = 0.6; CI = 0.4, 0.9) [ ].

Susceptibility factors

The effect of smoking has been investigated on a large scale in Denmark, where the incidence of smoking among women is much higher than in many other countries [ ]. Evidence has emerged that the combination of smoking with oral contraceptive use may have a synergistic effect on the risks of acute myocardial infarction and cerebral thromboembolism (but not of venous thromboembolism), particularly among users of high doses (50 micrograms). The authors therefore suggested that the very low-dose products should be preferred in smokers.

For current or potential users of oral contraceptives the question arises whether it is not wise to examine the individual’s possible predisposition to thromboembolism (thrombophilia) before deciding for or against this form of birth control. It has been suggested that in teenage users who might prove to be carriers of the Factor V Leiden mutation, routine screening would not be economically justified [ ]. It was instead the author’s view that clinicians can use thoughtful screening questions to identify potentially high-risk patients for thrombophilia and consider testing for inherited susceptibility factors case by case.

Estrogens versus progestogens

As far as the estrogens are concerned, the original estrogen, mestranol, was abruptly replaced by ethinylestradiol in most or all products after a wide-scale panic relating to risk of thrombosis in 1969. The motive lay entirely in the fact that the ethinylestradiol was about twice as potent, so that the dose could be halved. Whether this in fact led to any reduction in cardiovascular thromboembolic risk was never specifically examined. The choice of estrogen might still be relevant to the extent that (physiological) 17-β-estradiol in microcrystalline form later became available for oral use and in theory might prove safer in some respects than semisynthetic estrogens, since it seems to have less effect on the fibrinolytic system.

The question of the progestogens came to the fore as a result of the debate regarding the third-generation progestogens gestodene and desogestrel. At the time of marketing, no specific claim appears to have been made that they would present a lesser degree of risk of thromboembolism. However, emphasis was placed on experimental findings that might indicate a better cardiovascular prognosis. The older oral contraceptives produce an increase in low density lipoprotein and a reduction in high density lipoprotein and cholesterol; those changes may have some relevance to the occurrence of arterial disease. By avoiding them, and having the lowest possible estrogen content, the new combinations were expected to be safer, although this had in no sense been proven. However, the manner in which this information was presented could have introduced a degree of confusion in the minds of prescribers and users, and the belief appears to have arisen that these third-generation oral contraceptives might prove less likely than their predecessors to cause thromboembolic disorders. That this is not so in relation to venous thromboembolism is discussed above. The theoretical argument that the risk of atherosclerosis and its associated complications might in the long run be less is so far entirely unproven, and with the fall in use of these products following these unfavorable findings [ ], it seems uncertain whether the data necessary to examine that hypothesis will ever be accumulated.

Persistence of risk after discontinuation

Most, but not all, studies have concluded that the effect of oral contraception on the risk of cardiovascular disorders disappears after withdrawal [ , , , ]. The Nurses’ Health Study found no differences in either incidence of or mortality from various cardiovascular diseases between never-users and past-users, regardless of the duration of use, or the time since last use [ ]. The nested case-control analysis of the RCGP study showed that, although stroke risk was higher among current oral contraceptive users regardless of smoking status, former users had an increased risk only if they were current smokers. Such conclusions are supported most strongly by a 1990 meta-analysis of published studies on the relation between past use of oral contraceptives and myocardial infarction, which produced an adjusted relative risk estimate of 1.01, suggesting no association [ ]. Thus, there does not appear to be a long-term mechanism at work, such as atherosclerosis, but rather an effect confined to current use, such as thrombosis [ , ]. However, there have been only a few studies of stroke in relation to previous oral contraceptive use, and those studies have produced inconsistent results; thus, meta-analysis of stroke data is precluded.

With some of the newer drugs, the effects may (as seems to be the case with the older drugs) persist for a number of weeks and justify the withdrawal of these products some time before surgery or other risks. A study of changes in hemostasis after withdrawal of the newer combined oral contraceptives (ethinylestradiol 30 micrograms plus either desogestrel or gestodene) [ ] showed that several weeks elapsed before plasma concentrations of fibrinogen, factor X, and antithrombin III returned to baseline.

It should be noted that a few studies have produced deviant conclusions, for reasons that are not clear. Some found that the risk of fatal myocardial infarction was similar for current and past users [ ], whereas others actually reported a lower risk of cerebral thromboembolic attack in former users compared with never users [ ].

In summary, the final chapter in the story of oral contraceptives and arterial lesions has not yet been written. However, in the light of long experience there is reason for optimism. There is no reasonable evidence that oral contraceptives, whatever their biochemical effects, actually do increase the risk of atherosclerosis; as to the occurrence of acute arterial events, these are explained largely or entirely by the presence of risk factors other than oral contraception.