Oral contraceptives and venous thromboembolism: a
quantitative discussion of the uncertainties
T. KOSTER,* R.-A. SMALL,| F. R. ROSENDAAL*t & F. M. H E L M E R H O R S T t
From the * Department of Chmcal Epidemiology, the i Department of Raematology, and the J Department of Obstetncs, Gynaecology and Reproductive Mediane, Umversity Hospital Leiden, leiden, The Netherlands
Abstract. Koster T, Small R-A, Rosendaal FR, Heimerhorst FM (Department of Clinical Epidemiology, Department of Haematology and Department of Obstetrics, Gynaecology and Reproductive Medicine, University Hospital Leiden, Leiden, The Netherlands). Oral contraceptives and venous thromboembolism: a quantitative discussion of the uncertainties. J Intern Med 1995; 238: 31-37.
Objectives. The majority of post-thrombotic women are barred from using oral contraceptives. We evaluated this policy for its clinical relevance. Design. A meta-analysis of controlled studies between 1960 and 1993.
Setting. A Medline Computer search, from 1966 to 1993, in multiple languages, with the following Index terms: thrombosis, thrombopheblitis, vein, venous, pulmonary embolism, contraceptives, oestro-gen, oral.
Study selection. A total of 588 articles or abstracts were reviewed for controlled studies, in which an index group was compared with a control group. Included were one randomized trial, six follow-up studies and eight case-control studies.
Main outcome measures. Summary thrombosis risk for oral contraceptive users, number needed to discontinue oral contraceptives to prevent one (recurrent) thrombosis, comparison of additional
unwanted pregnancies and postpartum thrombosis between alternative birth-control methods.
Results. The studies proved highly heterogeneous with regard to size and direction of the risk estimate. The summary relative risk of first thrombosis during oral contraceptive use was 2.9 (95% CI, 0.5-17). Since the risk of thrombosis recurrence is not well known, we estimated alternatives, making various hypothetical assumptions, wherein women would continue to take oral contraceptives after a first episode of thrombosis, or stop and switch to use of an intra-uterine device, condom or the progestogen-only pill. Depending on the assumptions with regard to recurrence risk and the existence of possible subgroups with genetic coagulation defects, the cost-benefit ratio of advising against the use of oral contraceptives after a first thrombosis varied tremen-dously.
Conclusions. Our analysis shows that we lack the necessary data for recurrence risk of venous throm-bosis during continuing use of oral contraceptives, or after switching to other modes of contraception. This reflects the clinical uncertainties that result in highly contradictory advice to young women who have experienced a first thrombosis. Only follow-up studies on recurrence risk will settle the issue.
Keywords: benefit-risk ratio, meta-analysis, oral contraceptives, venous thrombosis.
Introduction
Since the first report on the occurrence of pulmonary embolism in a young woman who used a combined oral contraceptive (the oestrogen mestranol and the progestogen norethynodrel) was published in 1961 [1], many studies on the relationship between oral contraceptive use and venous thromboembolism
have been reported. Concern about recurrent venous thrombosis caused many physicians to advise women to cease use of oral contraceptives after a first thrombotic episode. This implies that the woman involved is forever barred from using oral contra-ceptives. This policy has major consequences, and should be evaluated for its clinical relevance, i.e. its absolute effect and benefit-risk ratio. Two issues will
32 T. KOSTER et al.
be considered here. First, the size of the absolute risk and the effect of advice against the use of oral contraceptives: how many women should discon-tinue oral contraceptive use in order to prevent one thrombosis? Secondly, since oral contraceptives are the most reliable, easily reversible method of con-traception, the risk of alternative methods should be taken into account. These alternative methods will result in more pregnancies and undoubtedly more pregnancy terminations, äs well äs puerperia, which in itself increases the risk of deep vein thrombosis. Therefore, the eure may be worse than the ailment. In this meta-analysis we have attempted to quantify all these aspects, in order to reach conclusions that may assist the clinician in rational decision-making. Given that the recurrence risk of thrombosis is not well known, we must rely on a sensitivity analysis that tests various hypothetical risks.
Methods
Data collection and deflnitions
We searched for studies, published in English, French or German, in which the risk of venous thrombosis for women who used oral contraceptives äs a method of birth control had been estimated. The studies were identified by using a computerized literature search (Medline (l January 1966 to 31 December 1993), with the following index terms: thrombosis, throm-bopheblitis, vein, venous, pulmonary embolism, contraceptives, oestrogen, oral), and by cross-referencing. A first selection was made on the basis of the abstracts, after which the entire manuscripts were reviewed. The analysis in this report is limited to controlled studies, in which an index group is compared to a control group. Excluded were case series and studies controlled by general popula-tion data, and reports in which data for venous thrombosis could not be differentiated from data for arterial thrombosis. In some reports, parts of the data were published more than once. From these articles we selected only those reporting the original or final results. Overall data on venous thrombotic events which also included flgures for rarely occurring (neurological, ophthalmic or mesenteric) thromboses were included.
Study characteristics, such äs type of study, oestrogen dose, and use of objective methods to diagnose venous thromboembolism, were
indepen-dently scored by two authors. In the event of disagreement, the other two authors were consulted, and a consensus could be reached in all cases.
Number needed to discontinue oral contraceptive use Whereas a relative risk (RR) is useful for establishing a causal association, it provides little Information about the clinical relevance, which is dependent on the absolute risk, i.e. the probability for an individual patient. From the absolute risk one may, by simple calculation, estimate the number of women that should discontinue the use of oral contraceptives in order to avoid one thromboembolic event [2]. Ob-viously, if a drug increases the risk from 1/1000 to 10/1000 (risk difference 9/1000), a total of 1000/9 (1/risk difference) = 111 women should discontinue this drug in order to avoid one thromboembolic event.
To perform this analysis, we need to estimate the incidence of thrombosis in those women who use oral contraceptives, and in those who do not use them. First, we estimated the Overall incidence (I) of venous thrombosis in women of child-bearing age, by dividing the number of first, objectively confirmed episodes of deep vein thrombosis in 15- to 44-year-old women over a 5-year period (data from the Leiden Thrombophilia Study (LETS) [3]) by the total number of women-years of this age group in the same well-defined greater Leiden area. The incidence of thrombosis among oral contraceptive users (Ie) and non-users (I0) can be estimated from this overall incidence (I) by the following equation:
(Proportion of pill-users χ pooled relative risk χ Ι0)
+ (proportion non-users χ I0) = I.
Ie can be determined by multiplying I0 by the
pooled risk estimate. Age-specific data for oral con-traceptive use for women aged 15 to 44 years were obtained from the Dutch Central Bureau of Statistics
[4]·
mated to be l per 100 women [5]. The failure rate (accidental pregnancy risk) among typical couples ('daily life') was considered to be 3 per 100 user years for oral contraceptives, 3 per 100 user years for intra-uterine devices, and 12 per 100 user years for condom use [6]. We took the lowest expected failure rates among couples who use the method perfectly to be 0.1 per 100 user years for oral contraceptives, 0.8 per 100 user years for intra-uterine devices, 2 per 100 user years for condom use, and 0.5 per 100 user years for the progestogen-only pill [6].
Staüsücal analysis
Prior to pooling, we tested for homogeneity between the studies analysed, i.e. we tested the hypothesis that the differences between the reported risks were only due to random error around the 'true' RR [7]. The test-statistic was computed by calculating G =
Σwl(k-l~~kt,)z, in which wt is defined äs the inverse
squared Standard error of the natural logarithm of the relative risks, /q is the estimate of the log relative risk of each study, and ke is the pooled estimate of the log relative risk. Under the null hypothesis, G follows a Chi-square distribution with n — l degrees of freedom (df) for n studies. Despite the reported insensitivity of this test, the studies still proved highly heterogeneous (G = 80.3, P < 0.001; df 14). There-fore, the pooled estimate was estimated äs the unweighted average of natural logarithmic trans-formed point estimates, whereas the Standard de-viation was calculated in the Standard manner for n (number of studies) observations.
For each study the relative risks with their 95% confidence intervals (CI) were extracted or recalcu-lated, if necessary. In some articles only stratum-specific relative risks were given. Whenever two-by-two tables were provided, the outcomes within these studies were combined by the Mantel-Haenszel method. Otherwise, the natural logarithm of the reported risks was recalculated to one relative risk by using inverse variance weights. In order to estimate whether publication bias existed, a funnel plot was constructed [8].
Results
A total of 588 papers and abstracts were examined for possible inclusion in the study. We reviewed 206 studies in detail and identified 18 controlled studies.
Three of these were excluded: in two studies only the risk of fatal pulmonary embolism was investigated, which did not fall within the scope of our study [9, 10] and in another study the controls were women with suspected pulmonary embolism that could not be confirmed by objective tests and, more significantly, an appropriate age-matched control group for women who used oestrogens for birth control only (and not for post-menopausal hormone replacement) could not be obtained from that study [11]. The general results of two of these excluded papers showed the absence of a relationship between (fatal) thrombosis and the use of oral contraceptives [10, 11]. The 15 remaining studies consisted of eight case-control studies [12-19], six follow-up studies [20-25] and one randomized clinical trial [26]. Figure l shows the reported relative risks with 95% confidence intervals. It clearly illustrates that the results differ substantially in direction and effect. The unweighted summary relative risk was 2.9 (95% CI, 0.5-17). On average, the reported risks were higher in case-control studies (RR 4.2; 95% CI, 1.3-14) than in follow-up studies (RR 2.1; 95% CI, 0.3-16). The randomized clinical trial yielded no increase in thrombosis risk for oral contraceptive users (RR 1.1; 95% CI, 0.4-2.9).
In these studies, objective diagnostic methods were either not explored at all, or were only explored in some of the patients. In none of the studies was objective testing performed in more than 90% of diagnoses. Four studies included data for oral contra-ceptives containing less than 50 μg oestrogen per pill
[18, 23-25], and in only one was a ' dose-response' relationship between the amount of oestrogen and thrombosis risk present [25]. However, the majority of the patients in these studies were using oral contraceptives containing ^ 50 μg oestrogen. Five studies reported the use of oral contraceptives con-taining ^ 50 μg [17, 20-22, 26], and in six studies the dosages were not stated [12-16, 19].
34 T. KOSTER et al. 100 Γ-ΙΟ - τ W XI •σ O 0.1
—
-f-FUP -f-FUP RCT CC CC -f-FUP CG CC -f-FUP CG CG -f-FUP CG -f-FUP CG ι ι ι ι ι ι ι ι ι ι ι ι ι ι ι l 85 82 7l 87 78 74 89 67 78 75 69
Publication year
86 69 78 79
Fig. l Funnel plot of the relative risks and 95% confidence intervals ranked by Standard error. FUP denotes follow-up study, CC denotes case-control study and RCT denotes randomized clinical trial. (A) represents the pooled unweighted relative risk; (B) represents the classical precision weighted (using inverse variance weights) pooled relative risk.
equation, we obtain (0.42 χ 2.9 χ Ι0) + (0.58 χ Ι0) =
2/10000. Therefore, the incidence for non-oral contraceptive users was calculated to be 1.1 per 10000, and the incidence for oral contraceptive users was calculated to be 3.3 per 10000. If oral contraceptive use increases the risk of venous throm-bosis from 1.1 to 3.3 per 10000 (risk difference 2.2/10000), a total of 4545 asymptomatic women should refrain from using oral contraceptives in order to prevent a first episode of deep vein thrombosis within l year. At present, the recurrent risk of venous thrombosis, after a woman has stopped using oral contraceptives because of this thrombosis, is not known. It might be higher than the baseline risk, since these women might have other risk factors, but it must be lower than the primary risk during oral contraceptive use, since an important risk factor has been removed. Therefore, we have tentatively used a relative risk of two for recurrent thrombosis in women who have stopped taking oral contraceptives. Consequently, their risk of thrombosis recurrence is 2.2/10000 (the additional recurrence risk is there-fore 1.1/10000). Equally unknown is the recurrence risk when women continue to use oral contraceptives
after their thrombotic episode. It will be at least äs
high äs the primary risk, and may be higher if there are women who are particularly sensitive to the thrombogenic effect of oral contraceptives. By way of sensitivity analysis we have set absolute risks of thrombosis recurrence at 4.4/10000 (primary risk for oral contraceptive users (3.3/10000) + additional recurrence risk of 1.1/10000), 9.9/10000 and 33/10000 (representing relative risks of 3 and 10, i.e. complementary to the risk of other factors). The latter risks may appear to be high. However, they approach the risks that are observed in subgroups with genetic coagulation deficiencies [27, 28]. On the basis of these absolute risks, 4545, 1299 or 325 women, respectively, should discon-tinue the use of oral contraceptives in order to avoid one second thrombotic episode per year.
Table l Companson of thrombosis outcome m three cohorts of post-thiombotic women Incidencc of recuirent VT.]: Cases of VT Failure rate Unwanted pregnancies Frequency of VT dunng
pregnancy and puerpenum Additional cases of VT§ Totdl number of VT Oral contraceptive (lOOOOOpyt) 44/10000 44 003(0001*) 3000 (100*) 001 30(1*) 74(45*)
Intra-utenne device Condom (100000 py) (100000 py) 9 9/10000 99 003 (0001*) 3000 (100*) 001 30(1*) 129 (100) 33/10000 330 0 03 (0 001*) 3000(100*) 001 30(1*) 360(331*) 2 2/10000 22 0 03 (0 008*) 3000 (800*) 001 30(8*) 52 (30*) 2 2/10000 22 0 1 2 ( 0 0 2 * ) 12000(2000*) 001 120 (20*) 142 (42*)
*Data when the lowest expected failure lates were apphed fpy, person years
i VT, venous thrombosis.
§ Calculated by: (number of pregnancies) χ (thrombosis risk, pregnancy and puerpenum)
recurrence risk, among typical couples ('daily life') the number of unexpected pregnancies äs well äs the
sum of venous thrombotic episodes will be highest among condom users. The use of intra-uterine devices will result in a reduction of at least 30% in the number of venous thrombotic episodes, and in an equal number of accidental pregnancies when com-pared to use of oral contraceptives. When we apply the lowest expected failure rates ('perfect use'), both intra-uterine device and condom use lead to a decrease in thrombosis outcome. However, this reduction in thrombosis will be accompanied by an 8- to 20-fold increase in unwanted pregnancies, because intra-uterine devices and condoms have higher failure rates than oral contraceptives, even when the methods are used perfectly.
Discussion
The results of this overview suggest that oral contraceptives are associated with an almost three-fold increase in risk of venous thrombosis. Given the low absolute thrombosis risk in 15- to 44-year-old women, the effort that must be made to prevent recurrent thrombosis is highly dependent on assump-tions about recurrence risk and contraceptive failure rates.
The selected studies were not homogeneous, i.e. the results showed more Variation around the summary estimate than could be explained by chance alone. The studies are therefore difficult to Interpret äs a whole. The pooled summary estimate of case-control studies (RR 4.2) was higher than the pooled summary estimate of follow-up studies (RR 2.1). In
both study designs, biases may explain at least some of the results, äs was nicely illustrated by Katerndahl et al. [29]. Interestingly, in the only randomized controlled trial, which will not have suffered from the biases described above, no difference was found in thrombosis risk between 4965 oral contraceptive users (aged 15 to 40 years) and a group of 4933 control women who were using vaginal barrier methods (RR 1.1; 95% CI, 0.4-2.9) [26].
Publication bias, caused by under-reporting smaller studies with no significant effect, must always be considered äs a possible explanation of the outcome in a critical review. However, ' funnel plot' analysis of the distribution of the various study outcomes ranked by their Standard errors did not suggest that such a bias was present (Fig. 1) [30].
The 'true' underlying thrombotic risk also varies according to the type of oral contraceptive used. The overall results of the 15 studies were dealing with higher-dose oestrogen-containing pills, which appear to be associated with a higher thrombosis risk than currently marketed 'sub-fifty' formulations [31-34], Furthermore, women with leg Symptoms who are using oral contraceptives are more likely to be diagnosed äs having venous thrombosis than women with the same leg Symptoms who are not current users of oral contraceptives. This detection and referral bias will result in an overestimation of the relative risk, the more so since, in the studies selected, for many patients the diagnosis was not confirmed by means of an objective test, which will without doubt have resulted in numerous false-positive diagnoses [35, 36], especially for women using oral contra-ceptives [37].
36 T. KOSTER et al
The use of condoms of intra-uterine devices äs an alternative to oral contraceptives will lead to more unwanted pregnancies. Under the lowest assump-tions for thrombosis risk among typical couples this might give an increase in the occurrence of venous thrombosis (Table 1). The decision of whether to continue using oral contraceptives after a thrombotic event or to switch to an intra-uterine device or progestagen-only puls is a more difficult one. The risk of thrombosis is decreased, but the risk of unwanted pregnancies can be increased considerably compared to that associated with oral contraceptive use, dependmg on which efficacy criterion is used.
The data are based on the total number of women with a first thrombosis. For specific subgroups, i.e. those with genetic coagulation disorders, the risk will be higher. Evidence supporting this view is already available for deficiencies of protein C, protein S and antithrombin [38]. However, these genetic risk factors have a very low prevalence in the general population, and hitherto played only a minor role in considerations when advising women who had experienced a first thrombosis during oral contra-ceptive use. However, this Situation might change completely with the recent discovery of genetic resistance to activated protein C, which is present in 3% of the general population and in 20% of unselected consecutive thrombosis patients [39]. It leads to a sevenfold increase in thrombosis risk, which might be further enhanced by oral con-traceptive use [27-28]. For these women, who represent a significant proportion of young women with deep venous thrombosis, future follow-up studies will describe the magnitude of the recurrent risk. Even so, in all of these women with genetic coagulation disorders, the consequences of prescrib-ing less reliable birth-control methods should be carefully considered, given the additionally increased percentages of pregnancy- and post-partum throm-bosis [40].
Our attempt to quantify the difficult clinical decision of whether or not to discontinue the use of oral contraceptives after a first thrombosis has led to the startling realization that we lack the necessary data. Depending on the assumptions made, one can arrive at totally opposite conclusions. This appears to be a reflection of the uncertainties that exist in current clinical practice, where highly contradictory advice is given to young women. In the end, the relative subjective fears of either thrombosis or
pregnancy, and the relative acceptability of different forms of contraception for a woman and her partner, might be the decisive factors.
Acknowledgements
We thank Professor Jan P. Vandenbroucke, MD PhD (Department of Clinical Epidemiology, University Hospital Leiden) for critical reading of the manu-script, and for advice.
References
1 Jordan WM Pulmonary embohsm Lancet 1961, i: 1146-7. 2 Decidmg on the best therapy In Sackett DL, Haynes RB, Guyatt GH, Tugwell P, eds Chmcal Epidemiology A Basic Science for Chmcal Medicme Boston Little, Brown and Company, 1991, 187-248
3 Koster T, Rosendaal FR, de Ronde H, Briet E Vandenbroucke JP, Bertina RM Venous thrombosis due to poor anticoagulant response to activated protein C Leiden Thrombophiha Study Lancet 1993, 342: 1503-6
4 Central Bureau of Statistics Statistical Yearbook 1993. Gravenhage SDU, 1993, 476
5 Carter C, Gent M, Leclerc JR The epidemiology of venous thrombosis In. Colman RW, Hirsch J, Marder VJ, Salzman EW, eds Hemostasis and Thrombosis Basic Prmciples and Chmcal Practice, 2nd edition Philadelphia Lippincott, 1987, 1185-98
6 Trussel J, Hatcher RA, Cates W Jr, Hance Stewart F, Kost K A guide to interpretmg contraceptive efficacy studies Obstet Gynecol ]990, 76: 558-67
7 Greenland S Quantitative methods m the review of epidem-lologic hterature. Epidemiol Rev 1987, 9: 1-30
8 Vandenbroucke JP Passive smokmg and lung cancer a publication bias? Br Med J 1988; 296: 391-2
9 Inman WHW, Vessey MP Investigation of deaths from pulmonary, coronary, and cerebral thrombosis and embohsm m women of child-beanng age Br Med ] 1968, 2: 193-9 10 Thorogood M, Mann J, Murphy M, Vessey M. Risk factors of
fatal venous thromboembolism m young women. a case-control study Int J Epidemwl 1992, 21: 48-52
11 Qumn DA, Thompson BT, Terrm ML, Thrall JH, Athanasoulis CA, McKusick KA et al A prospective mvestigaüon of pulmonary embohsm m women and men JAMA 1992, 268: 1689-96
12 Royal College of Practitioners Oral contraception and thrombo-embolic disease / R Coi! Gen Pract 1967, 13: 267-79
13 Sartwell PE, Masi AT, Arthes FG, Greene GR, Smith HE Thromboembolism and oral contraceptives an epidemiologic case-control study Am ] Epidemwl 1969, 90: 365-80 14 Vessey MP, Doll R Investigation of relation between use of
oral contraceptives and thromboembolic disease. a further report Br Med ] 1969, 2: 651-7.
15 Greenblatt DJ A retrospective case-control sludy of diseases associated with oral contraceptive use. Am Heart J 1975, 89: 677-8
contracepüves, smoking and other faclors m relation to nsk ofvenousthromboembolic disease AmJEpidemiol 1978 108: 480-5
l 7 Mdguire MG Tonascid ] Sartwell PE Stolley PD Tockman MS Increased nsk of thrombosis due to oral contraceptives, d further report Am J bpidemiol 1979 110: 188-95 18 HelmnchSP Rosenberg L Kaufman DW Strom B Shapiro S
Venous thi omboembohsm m relation to oral contraceptive use Ohtet Gynecol 1987 69:91-5
l 9 WHO Collabordtive Study Cardiovascular disease and use of oral contraceptives WHO Bulletin OMS 1989 67:417-23 20 Grounds M Anovulants thrombosis and other associated
changes Med ] Aust 1974 2:440-6
21 Royal College of General Prdcütioneis Oral contraceptives, venous thrombosis dnd vancose vems / R Coll Gen Prati 1978 28:393-9
22 Diddle AW Gardner WMH Wilhamson PJ, Johnson JR Hemphill JL, Godwm CW Oral contraceptive steroids and thrombophlebitis / Tenn Med Assoc 1978 71: 22-6 23 Porter JB, Hunter JR, Damelson DA Tick H, Stergachis A Oral
conlrdceptives and nonfatal vascular disease recent expen-ence Obstet Gynecol 1 982 59:299-302
24 Porter JB Hunter JR Jick H Stergachis A Oral contraceptives and nonfatal vascular disease Obstet Giinecol 1985 66: 1-4 25 VesseyM MantD Smith A Yeates D Oral conlraceptives and
venous thromboembohsm findmgs in a laige prospective study Br Med] 1986 2 9 2 : 5 2 6
26 Fuertes de la Haba A, Curet JO Pelegrma I Bangdiwala I Thrombophlebitis among oral and nonoral contraceptive users O&siei Gyntcol 1971 38:259-63
27 Vandenbroucke JP, Koster T Briet E ReitsmaPH Bertina RM Rosendaal FR Increased nsk of venous thrombosis m oral-contraceptive users who are carners of factor V Leiden mutaüon Lancet 1994, 344: 1453-7
28 RosendaalFR Koster T Vandenbroucke JP ReitsmaPH High nsk oi thrombosis m patients homozygous for factor V Leiden (APC-iesistance) Blood 1995 85:1504-08
29 Katerndahl DA Realini J Cohen PA Oral contraceptive use and cardiovascular disease is the reldtionship real or due to study bias? / Γαιη Pract 1992 35: 147-57
30 Light RS, Pillemer DB Summini? l/p the Science of Reviewmg Research Cambridge, Massachusetts Harvard Umversity Press 1984, 65-6
31 Gerstman BB Piper JM, Freiman JP, Tomita DK Kennedy DL Ferguson WJ, Bennett RC Oral contraceptives oestrogen and progestin potencies and the mcidence of deep venous thrombo-embohsm Int] Epidemiol 1990 19: 931-6
32 Gerstman BB Piper JM Tomita DK Ferguson WJ Stadel B V Lundin FE Oral contraceptive estrogen dose and the nsk of deep venous thromboembolic disease Am J Epidemwl 1991
133: 32-7
33 Lmdegaard 0 Oral contraception dnd nsk of a cerebral thromboembolic altack results of a case-control study Br Med} 1993 306: 956-63
34 Bottiger LE, Boman G Eklund G Westerholm B Oral contraceptives and thromboembolic disease effects of low-ermg oestrogen Content Lancet 1980 1: 1097-101 35 Huisman MV Buller HR, Ten Cate JW, Vreeken J Senal
impedance plethysmography for suspected deep venous throm bosis m outpatients The Amsterdam general practitioner study NEnglJMed 1986, 314: 823-8
36 LensmgAWA Prandom P Brandjes D Huisman PM VigoM Tomasella G et al Accurate detection of deep-vem thrombosis by real-time B-mode ultrasonography N Engl ] Med 1989 320: 342-5
37 Barnes RW KrapfT Hoak JC Erroneous clinical diagnosis of leg vein thrombosis in women on oral contraceptives Obstet Giinecol 1978 51: 556-8
38 Pabmger I Schneider B Thrombotic nsk of women with hereditary antithrombin III, protem C- and protem S deficiency taking oral contraceptive medication Thiomb Haenwit 1994, 71: 548-52
39 Bertina RM Koeleman RPC, Koster T Rosendaal FR Dirven RJ Ronde H de et al Mutation m blood coagulation factor V associated with resistance to activated protem C Natwe 1994,369: 64-7
40 Conard J Horellou MH van Dreden P Lecompte T Samama M Thrombosis and pregnancy m congemtal deficiencies in AT III protem C or protem S studv of 78 women (letter) Thromb Haemost 1990 63:319-20
Received 7 August 1994, dccepted 3 January 1995
Correspondence F M Heimerhorst MD Department of Obstetncs, Gynaecology and Reproductive Medicme, Umversity Hospital Leiden, P 0 Box 9600, 2300 RC Leiden The Netherlands
