Human Reproduction, Vol. 14, No. 6, 1493-1499,
June 1999
© 1999 European Society of Human Reproduction and Embryology
The differential risk of oral contraceptives: the impact of full exposure history*
1 EPES Epidemiology, Pharmacoepidemiology and Systems Research, Berlin, Germany, 2 Department of Preventive Medicine, University of Kansas Medical Center, Kansas City, USA, 3 University of Potsdam, Potsdam, Germany, 4 Department of Preventive Medicine, University of Basel, Basel, Switzerland, 5 Centre of Epidemiology and Health Research (ZEG), Zepernick, Germany and 6 Department of Epidemiology and Biostatistics, McGill University, Montreal, Canada
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionAppendixReferences |
|---|
Previous discussions have indicated that the small increases of risk of venous thromboembolism (VTE) associated with newer combined oral contraceptives (third generation, containing desogestrel and gestodene) may be attributed to bias due to cohort effects. In a case-control analysis, this may produce an overestimate of risk of newer preparations. In 10 centres in Germany and the UK, the Transnational Study analysed data from 502 women aged 16–44 years with VTE, and from 1864 controls matched for 5-year age group and region. Information on lifetime exposure history from all subjects was added to the dataset used in previous analyses and entered into a Cox regression model with time-dependent covariates. Based on 17 622 continuous exposure episodes comprising 47 914 person-years of observation, the adjusted hazard ratio (equivalent to odds ratio, OR) of VTE for the comparison of current users of third-generation versus current users of second-generation (primarily levonorgestrel compounds) combined oral contraceptives was 0.8 (0.5 to 1.3). The OR obtained in standard case-control analysis had been 1.5 (1.1 to 2.1). Adjustment for past exposures includes more information and appears more valid than the standard cross-sectional analysis. Using this approach, the Transnational Study data show no evidence for an increased risk of VTE with third- compared with second-generation combined oral contraceptives.
Key words: case-control study/Cox regression analysis/epidemiology/oral contraceptive use/venous thromboembolism
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAppendixReferences |
|---|
A good deal of the development of new oral contraceptives (OC) was directed towards reducing the risk of thromboembolic events which were found to be associated with OC use soon after their introduction (Thorogood, 1993
). It therefore came as a surprise when studies published in 1995/1996 seemed to show an increased risk of venous thromboembolism (VTE) among users of newer OC compared with users of older OC (Bloemenkamp et al., 1995; Farley et al., 1995; Jick et al., 1995; Spitzer et al.; 1996; Farmer et al., 1997). This has engendered considerable discussions on the safety of newer (third-generation) OC in Europe which are still ongoing (Lewis and Heinemann, 1997; Spitzer, 1997a,b; Vandenbroucke and Rosendaal, 1997; Meirik, 1998). Because the differences found between users of newer (third-generation products containing <50 µg ethinyl oestradiol (EE2) and the progestins gestodene and desogestrel) and older (second-generation, containing <50 µg EE2 and primarily levonorgestrel) OC were small, ranging between 1.3 and 2.2 (Table I), and because the population impact of intervention is large, investigators were careful to address and explore sources of potential bias and confounding factors (Farley et al., 1995; Lewis et al., 1996b; Spitzer et al., 1996; Spitzer, 1997b). The biases addressed included physician diagnostic behaviour, preferential prescribing of newer OC to risk groups, and a phenomenon known as depletion or attrition of susceptibles, observed in a variety of epidemiological studies (Miettinen and Caro, 1989; Moride and Abenhaim, 1994; Posthuma et al., 1994).
|
Although the results of the Transnational Study (Spitzer et al., 1996
) were similar to others published at the same time, the first analyses showed that adjustment for duration of use of the current OC reduced the risk estimates for newer-generation OC by about 12% (Spitzer et al., 1996). Additional analyses showed that the magnitude of risk estimates for individual progestins were closely linked with the time of market introduction of the OC, with increasing risk estimates of VTE found for newer OC (Lewis et al., 1996b). Because it seemed unlikely that older, low-dose OC should consistently show a more beneficial risk profile than newer OC, and because it was found that a larger proportion of individuals using older OC were long-term users than of those using newer OC, a `healthy user' effect was postulated to be active within these studies. This is a cohort effect based on the notion that long-term users are individuals with good tolerance, whereas groups with shorter duration of use may constitute a different risk set. This results in a depletion of susceptibles in the group of older exposures and loading of problem cases in the group of newer exposures (Lewis et al., 1996b). Cross-sectional comparisons of two groups using medications with different market entry points are therefore likely to overestimate the risk associated with the most recently introduced medication, especially when an adverse event is not expected to occur immediately after exposure. Standard case-control analysis considers current risk factors and summary data on past exposures. The Cox proportional hazards model with time-dependent covariates permits multiple episodes of use and non-use of OC, including precise details of duration and nature of each episode, to be used in adjustment of the associations with current exposures. This approach becomes possible through the availability of the lifetime exposure history of all subjects included in our study. We present here the results of the analysis of a Transnational Study dataset, enhanced with information on all prior exposures of the study subjects.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAppendixReferences |
|---|
Study design and conduct
The Transnational Study on oral contraceptives and the Health of Young Women is a matched case-control study designed to assess the risk of oral contraceptives on deep vein thrombosis and pulmonary embolism (i.e. venous thromboembolism; VTE), arterial thrombotic stroke and myocardial infarction. Its methods are almost identical to those of the case-control study of the World Health Organization Human Reproduction Unit, and have been detailed elsewhere (Spitzer et al., 1993
; Poulter et al., 1995; Lewis et al., 1996a). The cases were women aged 16–44 years who had a diagnosis of VTE. Deep vein thrombosis was defined by pain and tenderness in the extremities, a precise record of knee circumference, and confirmation with imaging procedures. The diagnosis of pulmonary embolism was based on symptoms of pain in the chest or side, and confirmatory imaging procedures. The cases were accrued in hospital between early 1993 (first case index date 2 January 1993) and 20 October 1995 in centres in Germany and the UK. These were matched by centre and 5-year age-band to at least one community and one hospital control.
A full exposure history was documented by filling in a calendar sheet which identified exposure by month since menarche, indicating the first eligibility to take OC. Exposure periods of interest were the use of an OC and the brand of OC used, pregnancy, and periods of non-exposure. Current use of OC was defined as the OC used within a 3-month window at the time of the event (for a case), the hospital admission (for a hospital control) or the date of interview (for a community control). Prior use of OC was defined as any use that ended before this period. For this analysis, first-generation OC were defined as any preparation that contained 
50 µg of EE2, regardless of progestin content. Second-generation OC were defined as those which had 
35 µg of EE2 and contained a progestin other than gestodene or desogestrel. Preparations containing norgestimate were included with the second-generation products to retain consistency with other analyses. Third-generation OC were defined as combination products with low doses of EE2 (usually 30 µg or 20 µg) and either gestodene or desogestrel. Separate analyses for individual progestins were conducted. OC classification lists are available from the authors (Lewis et al., 1996a).
Data processing and analysis
All data were checked manually and by computer for eligibility and correct matching. The clinical data were coded twice and difficult or unreconciled diagnoses were arbitrated by local and international panels of clinical specialists. All other data were entered twice and verified. Plausibility checks of current and prior exposures were conducted by numerous procedures, including verification of age, periods of pregnancy, and presence of the indicated product on the market. The data on exposure histories from the pill calendar sheet were entered into a separate database and merged with the main Transnational Study dataset. The data were first modelled for the stroke component of the study. This dataset was chosen because it contained adequate exposures of second- and third-generation OC for cases and controls (Heinemann et al., 1997). This approach permitted an unbiased configuration and exploration of this complex dataset. The model derived from this work was then applied to the new VTE dataset.
Figure 1 shows the configuration of the new dataset for analysis using the pill calendar data. Ordinarily, cases and controls are captured within a 3-year study period, and their exposures are assessed for the time during which the event occurred. Standard analyses rarely account for prior exposures, which are key for the attrition of susceptibles. These prior exposures may constitute periods of challenge (high-risk periods). To adjust adequately for these periods, the data were analysed using a Cox regression model on a fixed time axis defined by the successive calendar months (Cox and Oakes, 1984; Clayton and Hills, 1993; Collet, 1994). This method is similar to the Poisson regression, but based on much finer subdivisions in time. In using this method, we computed the conditional probability that, given a failure occurred in a set of subjects, it had occurred in the case rather than in some other member of the risk set. The profile log likelihood of a cohort corresponded with the conditional log likelihood obtained for individually matched case-control studies (Breslow and Day, 1980; D'Agostino et al., 1990). In our approach, all study subjects were right-censored at the same time by virtue of their being accrued almost simultaneously within a case-control setting. The subject's risks were determined by different exposure periods and likelihood of exposure to various agents.
|
We defined a calendar axis from 0 to 458 months for the total period of observation. We left-censored the individual members at 9 years of age, this being the age of first exposure to risk of the subject with the earliest exposure to OC or pregnancy. The exposure episodes j of each subject i were arrayed along this time axis by month (1,..., k). Each subject was followed from cohort entry (t0i1) to case-control study entry (tmiki) by month through various exposure periods (tmij, where j = 1,..., k). Event status is defined by the outcome variable d, corresponding to case or control status. The time-dependent variable entered into the model was the exposure or non-exposure to specific OC at monthly intervals. We adjusted for non-time-dependent variables shown to influence risk, such as body mass index (BMI), alcohol use, index age and smoking status, as well as for current and previous duration of use by generation and for switching. The analyses were performed with STATA 5.0 (StataCorp, 1997
). We calculated adjusted hazard ratios expressed as odds ratios (OR; signifying their origin from a case-control study) and their 95% confidence intervals to estimate the risk of VTE associated with use of the various categories of OC investigated. |
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAppendixReferences |
|---|
Data on 2366 women aged 16–44 (502 cases and 1864 controls) were collected from 1 January 1993 to 20 October 1995 in the UK and Germany. Cases differed from controls in that a larger proportion of cases had a BMI >30 (17.7 versus 9.9%), were current smokers (45.8 versus 39.3%), and were current users of oral contraceptives (67.5 versus 41.8%) (see also Spitzer et al., 1996
). The stratification of current use and previous use by age in Table II shows a predominance of current users of third-generation OC in the age group 16–24 years among cases (41.4% versus 25.4% of controls). Use of third-generation OC is equally distributed among previous users (26.7% of cases and 30.3% of controls) in this age group. Current use of second-generation OC predominates in the age groups 25–44, whereas there is a distinct increase in the proportion of previous users of first-generation OC in the older age groups. Current use of OC decreases considerably with age in both cases and controls (Table II). The proportions of previous use are similar for cases and controls in each age group.
|
The average current duration of use of all OC increases with age with pronounced age-related differences in length of use by generation (Table III
). The total exposure times decline from first to third generation in all age groups for cases (mean times: for first generation 78, second 52, third 30 months) and controls (first generation 71, second 58, third 30 months). The proportion of cases who have used OC for >60 months is lowest in the group of third-generation OC users (12%) compared with cases using second- (20.1%) and first- (48.7%) generation OC. About 16% of cases and 22% of controls were never users, and 41.2% of cases and 42.2% of controls stayed on the same OC throughout. Defined switches from second- to third-generation OC took place in 9.6% of cases and 6.2% of controls. Multiple switches with no clear pattern were found in 26.3% of cases and 23.4% of controls.
|
A total of 17 622 separate exposure episodes covering 47 914 woman-years of observation were included in the Cox regression model with time-dependent covariates. The relative risk (hazard ratio; OR) for the model which included adjustments for exposure in each episode by generation, age, BMI, smoking, alcohol use, duration of current use by generation (first to third) and duration of previous use by generation (first to third) as linear variables, and `switching' by generation was 2.90 [95% confidence interval (C.I.): 2.06 to 4.09] for any OC use versus non-use, and 8.48 (3.02 to 23.86) for use of first generation versus non-use (Table IV
). The OR for second generation versus non-use was 2.85 (1.92 to 4.22) and for third generation it was 2.26 (1.46 to 3.50), all statistically significant. The direct comparison between users of third-generation versus second-generation OC resulted in a risk estimate of 0.79 (95% C.I. 0.50 to 1.26), which was not statistically significant. Estimates comparing users of various progestins with users of levonorgestrel-containing OC show no significant differences (Table IV).
|
There were no differences in the risk estimates when we stratified by 10-year age groups, nor was there a trend by time of market introduction overall or for any age group. Although there were absolute differences in risk estimates when comparing users with non-users in the UK (2.2; 1.48 to 3.3) and Germany (5.6; 2.9 to 0.8) due largely to the absence of current users of first-generation OC in the UK, there were no important differences in the adjusted comparisons of users of third-generation with users of second-generation OC between countries (UK: 0.85, 0.50 to 1.45; Germany: 0.76, 0.32 to 1.82). No important differences were found either between hospital and community controls when comparing current OC use against no current use (hospital controls: 2.42, 1.74 to 3.38; community controls: 2.19, 1.54 to 3.11), or when comparing users of third- with users of second-generation OC (hospital controls: 0.84, 0.52 to 1.35; community controls: 0.82, 0.51 to 1.33). The model was stable, in that results were altered only slightly as additional adjustment variables were introduced (unadjusted model: 0.84, 0.59 to 1.18; full adjustment: 0.79, 0.50 to 1.26). Concerning other variables which might potentially have influenced the occurrence of VTE, the most pronounced independent risk factors for VTE appeared to be a BMI >30 (2.72, 1.80 to 4.11), current smoking (1.42, 1.12 to 1.79) and diabetes (2.41, 1.07 to 5.42). The use of alcohol was associated with a decreased risk of VTE (0.50, 0.36 to 0.69, versus non-use of alcohol).
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAppendixReferences |
|---|
There are some indications that the studies on oral contraceptives do not meet one of the primary conditions of the case-referent approach, in that the processes which bring the cases to attention and by which controls are selected may not be independent of the exposure (Miettinen and Caro, 1989
). This may lead to confounded or biased results when the traditional cross-sectional methods of analysis are used (Lewis et al., 1996b; Suissa et al., 1997). This is expressed by a variety of inconsistent and implausible results which have led authors to postulate distortion due to heterogeneous exposure groups. A case-control study measures the instantaneous drop-out rate when only individuals (and their controls) who have been healthy up to the time of the index event are included. This approach is best suited for well-defined events which occur soon (preferably days or weeks) after a drug is taken. For an ill-defined and non-specific event such as VTE, this approach becomes problematic when the observed time window of exposure becomes large. The problem is further compounded if there have been previous exposure episodes which can be regarded as high-risk or challenge periods for VTE. For OC, exposure likelihoods additionally differ with age, in that older individuals are more likely to have been exposed to previous generation OC. Lastly, a comparison of older and newer generation OC becomes difficult because high-risk members of the older user group may have dropped out due to events or other intolerances and possibly switched to the apparently more tolerable newer products (Lewis et al., 1996b). There is no assurance that the cohort of exposed individuals used as the index population is actually comparable to the cohort of unexposed individuals used as the reference population. There are good reasons indeed to believe that they are not comparable, because the comparison groups for exposure are not homogeneous. In our study, these differences are reflected by differing proportions of users of specific types of OC in the various age groups (Table II
), by differences in the lengths of current and previous exposure episodes across age groups and generations (Table III), and by differences in switching patterns between cases and controls.
We adjust for these differences in an enhanced dataset which now includes past exposures and precise duration terms along a time axis. The model used is remarkably stable, and the results appear consistent and plausible. The main effect of the adjustment for prior exposure is a reduction of the risk estimate for users of third-generation OC from 4.8 in the original analysis to 2.3. The comparison of users of third- with users of second-generation OC shows that there is no evidence for an increased risk of third-generation OC (0.79, 0.50 to 1.26; previous estimate 1.5, 1.1 to 2.2) (Lewis et al., 1996b; Spitzer et al., 1996). The results indicate that the relevant effects of prior exposure for the comparisons between OC user groups are removed. This is demonstrated by the consistency of the results when stratifying by control group and by country, and by the absence of the dependence of risk estimates on the time of market introduction of the various progestins as shown in standard logistic regression (Lewis et al., 1996b).
A conventional case-control comparison cannot adequately account for those individuals who have not successfully passed their high-risk periods and who have become cases in the past. The present approach also cannot deal with this issue, but it does use the prior exposure data in as complete a manner as possible. Any estimates achieved with this approach will underestimate the risk in the users who have passed many high-risk periods (i.e. older users of older generation OC) compared with the risk in those who have not traversed as many high-risk periods (i.e. younger users and users of newer products). Recall bias can never be fully excluded in this form of study. However, validity checks of pregnancies and OC market introduction dates showed that women placed their OC brand in the correct time-frame, so that recall bias is unlikely to be an issue. A further limitation is that lifetime histories of other variables, such as BMI and smoking, were not collected, but only data on such variables at the time of entry of the subjects into the case-control study. Their influence cannot be estimated over time, but only in a cross-sectional fashion. However, it should be noted that the unadjusted model yielded virtually identical results, which argues against important distortion by these factors. The full adjustment for past exposures among these healthy women includes more information and provides more precision than the standard cross-sectional analysis.
|
Appendix |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAppendixReferences |
|---|
The Transnational Case-control Study
Overseeing Board and Advisory Groups
Scientific Reference Board: U.Bergman, Karolinska Institute; M.Breckwoldt, U.Freiburg; J.Collins, McMaster University; F.Kemper, U.Munster; J.Le Lorier, University of Montreal; S.MacLeod (Chair), McMaster University; K.MacRae, Charing Cross and Westminster Medical School (until May 1995); W.Ray, Vanderbilt University (since May 1995); J.Schlesselman, University of Pittsburgh.
Statistical Advisory Group: I.Guggenmoos-Holzmann, Freie Universität, Berlin; J.Hanley, McGill University; M.Lewis, Potsdam Institute of Pharmacoepidemiology and Technology Assessment and McGill University; K.MacRae, Charing Cross Hospital and Westminster Medical School; W.Ray, Vanderbilt University; J.Schlesselman, Pittsburgh University; S.Suissa, McGill University.
Ad hoc Peer Review Task Force: U.Bergmann, Karolinska Institute; J.Daling, Fred Hutchinson Cancer Research Center, University of Washington; M.Elstein, St Mary's Hospital, Manchester; J.Hanley, McGill University; S.Lock (Chair), Wellcome Trust; S.MacLeod, McMaster University; W.Ray, Vanderbilt University; L.Rosenberg, Boston University; S.Suissa, McGill University
British Steering Committee: V.Beral, Oxford University; N.Cherry, Manchester University (until June 1995); M.Elstein (Chair), St Mary's Hospital, Manchester; M.Harrison, UCL Medical School; C.Kay, Royal College of General Practitioners; M.Shipley, University of London.
German-Speaking Steering Committee: R.Brugger, AKS Bregenz; J.W.Dudenhausen, Freie Universität, Berlin; E.Fleck, German Heart Centre; I.Guggenmoos-Holzmann (Chair), Freie Universität, Berlin; U.Gundert-Remy, University of Göttingen; K.Kunze, Uniklinik Hamburg-Eppendorf; N.Victor, University of Heidelberg; C.Zippel, Klinik fur innere Medizin und Rehabilitation, Berlin.
The Transnational Case-control Study Group
Potsdam Co-ordinating Centre: Potsdam: V.Anning; A.Assmann; C.Klindworth; D.Kuhl-Habich; M.Lewis (Head of International Data Management); J.Von Rosen.
Senior Investigators: J.P.Boissel, Hopital Neuro-Cardiologique, Lyon; R.Bruppacher, University of Basel; L.Heinemann, Zentrum fur Epidemiologie und Gesundheitsforschung (Principal investigator from January 1996); M.Lewis, EPES Epidemiology, Pharmacoepidemiology and Systems Research, Berlin; K.MacRae (Principal Medical Statistician), Imperial College London; N.Poulter, UCL Medical School; W.O.Spitzer (Principal Investigator until December 1995), McGill University; M.Thorogood, London School of Hygiene and Tropical Medicine, University of London.
Centres
Austria: Bregenz: R.Brugger; H.Concin (Director); S.K.Fuchs; I.Michalek; C.Trotter. Graz: S.Hascher; S.Loibner; G.Stark (Director). Innsbruck: S.Alge; S.Hammerle: H.Jud; B.Kittinger; C.Pechlaner; K.Pfeiffer (Director), H.Ulmer. Salzburg: B.Gappmeyer; S.Humpeler; R.Weitgasser (Director); G.Scheurer.
France: Bordeaux: H.Barlet; M.Baumevieille; B.Begaud (Director); A.Chaslerie; F.Haramburu; F.Penouil. Lyon (national co-ordinating centre): B.Boissel (Director); C.Cornu; J.Gillet; A.Le Goff; B.Ravis; N.Strang.
Germany: Berlin (national co-ordinating centre): W.Barth; H.Behr; L.Heinemann (Director); T.Leonhard; B.Lorenz; F.Masius; K.-H.Mauritz; S.Moehner; B.Sykura; C.Thiel; O.J.Titlbach; C.Toussaint. Halle: J.Busse; W.Fleig; J.Haerting (Director until July 1994); M.Hagert; H.Klette; F.Lautenschlaeger (Director since July 1994); H.Liebal; H.Podhaisky; W.Teichmann; S.Zierz. Hamburg: A.Heinemann; U.Lockemann; K.Pueschel (Director). Jena: F.Hoffmann (Director); G.Grohmann; U.Merkel. Magdeburg: B.Bremer; D.Haase; W.Hellenbrand; C.Kalninsch; C.Listing; B.-P.Robra (Director); M.Schwarz. Nordhausen: M.Lustermann (Director). Schwerin: I.Schulzki (Director); H.Voigt. Zwickau: S.Boethig (Director); C.Lorenz.
Switzerland: Zurich: B.Candinas; F.Gutzwiller (Director); U.Kaser; N.Pua; J.Schilling.
United Kingdom: Glasgow: L.de Caestecker (Director); E.Hamilton; S.Mitchell; M.Soler-Lopez. London (national co-ordinating centre): D.Curle; J.Gibbs; H.Hemingway; G.Mein; M.Thorogood; N.Poulter; S.Smith. Manchester: P.Hannaford (Director since August 1995); Z.Evans; B.Faragher; S.Ferry; G.McHugh; C.McCollum (Director until August 1995); C.Sutcliffe. Southampton: A.Ascott; N.Dunn; R.Mann (Director); B.Weston.
|
Acknowledgments |
|---|
The Transnational Study was funded by a grant by Schering AG, Berlin. The analysis on full exposure history was funded by an unconditional grant from Organon NV. The participants of the Transnational Study are listed in the Appendix.
|
Notes |
|---|
*For the Transnational Study on Oral Contraceptives and the Health of Young Women — see appendix
7 To whom correspondence should be addressed at: EPES Epidemiology, Pharmacoepidemiology and Systems Research, Wulffstrasse 8, D-12165 Berlin, Germany 
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionAppendixReferences |
|---|
Bloemenkamp, K.W.M., Rosendaal, F.R., Helmerhorst, F.M. et al. (1995) Enhancement of factor V Leiden mutation of risk of deep-vein thrombosis associated with oral contraceptives containing a third-generation progestagen. Lancet, 346,1589–1593.[Web of Science][Medline]
Breslow, S.E. and Day, N.E. (1980) Statistical Methods in Cancer Research. World Health Organization. International Agency for Research on Cancer, Lyon.
Clayton, D. and Hills, M. (1993) Statistical Models in Epidemiology. Oxford University Press, Oxford, pp. 298–318.
Collet, D. (1994) Modelling Survival Data in Medical Research. Chapman & Hall, London.
Cox, D.R. and Oakes, D. (1984) Analysis of Survival Data. Wiley, New York.
D'Agostino, R.B., Lee, M.-L. and Belanger, A.J. (1990) Relation of pooled logistic regression to time dependent Cox regression analysis: the Framingham Heart Study. Stat. Med., 9, 1501–1515[Web of Science][Medline]
Farley, T.M.M., Meirik, O., Chang, C.L. et al. (1995) World Health Organization Collaborative Study of Cardiovascular Disease and Steroid Hormone Contraception. Effect of different progestagens in low oestrogen oral contraceptives on venous thromboembolic disease. Lancet, 346, 1582–1588.[Web of Science][Medline]
Farmer, R.D.T., Lawrenson, R.A., Kennedy, J.G. and Hambleton, I.R. (1997) Population-based study of risk of venous thromboembolism associated with various oral contraceptives. Lancet, 349, 83–88.[Web of Science][Medline]
Heinemann, L.A.J., Lewis, M.A., Thorogood, M. et al. (1997) Transnational Research Group on oral contraceptives and the health of young women. Case-control study of oral contraceptives and risk of thromboembolic stroke: results from international study on oral contraceptives and health of young women. Br. Med. J., 315, 1502–1504
Jick, H., Jick, S.S., Gurewich, V. et al. (1995) Risk of idiopathic cardiovascular death and nonfatal venous thromboembolism in women using oral contraceptives with differing progestagen components. Lancet, 346, 1589–1593.
Lewis, M.A. and Heinemann, L.A.J. (1997) Oral contraceptives and venous thromboembolism. Lancet, 349, 1621–1622.
Lewis, M.A., Assman, A., Heinemann, L. and Spitzer, W.O. (1996a) Transnational case-control study of oral contraceptives and health. Approved protocol revisions through September, 1995. Pharmacoepidemiol. Drug Safety, 5, 43–51.
Lewis, M.A., Heinemann, L.A.J., MacRae, K.D. et al. (1996b) The increased risk of venous thromboembolism and the use of third generation progestagens: role of bias in observational research. Contraception, 54, 5–13.[Web of Science][Medline]
Meirik, O. (1998) Risks of oral contraceptives. Lancet, 351, 521.
Miettinen, O.S. and Caro, J.J. (1989) Principles of nonexperimental assessment of excess risk, with special reference to adverse drug reactions. J. Clin. Epidemiol., 42, 325–331[Web of Science][Medline]
Moride, Y. and Abenhaim, L. (1994) Evidence of the depletion of susceptibles effect in non-experimental pharmacoepidemiologic research. J. Clin. Epidemiol., 47, 731–737.[Web of Science][Medline]
Posthuma, W.F.M., Westendorp, R.G.J. and Vandenbroucke, J.P. (1994) Cardioprotective effect of hormone replacement therapy in postmenopausal women: is the evidence biased? Br. Med. J., 308, 1268–1269.
Poulter, N.R., Chang, C.L., Farley, T.M.M. and Marmot, M.G. (1995) World Health Organization Collaborative Study of Cardiovascular Disease and Steroid Hormone Contraception. A multinational case-control study of cardiovascular disease and steroid hormone contraceptives. Description and validation of methods. J. Clin. Epidemiol., 48, 1513–1547.[Web of Science][Medline]
Spitzer, W.O. (1997a) Balanced view of risks of oral contraceptives. Lancet, 350, 1556–1557.[Medline]
Spitzer W.O. (1997b) The 1995 pill scare revisited: anatomy of a non-epidemic. Hum. Reprod., 12, 2347–2357.
Spitzer, W.O., Thorogood, M. and Heinemann, L. (1993) Trinational case control study of oral contraceptives and health. Pharmacoepidemiol. Drug Safety, 2, 21–31.
Spitzer, W.O., Lewis, M.A., Heinemann, L.A.J. et al. (1996) Transnational Research Group on Oral Contraceptives and the Health of Young Women. Third generation oral contraceptives and risk of venous thromboembolic disorders: an international case-control study. Br. Med. J., 312, 83–88.
StataCorp (1997) Stata Statistical Software: Release 5.0. College Station, Stata Corporation, Texas.
Suissa, S., Blais, L., Spitzer, W.O. et al. (1997) First-time use of newer oral contraceptives and the risk of venous thromboembolism. Contraception, 56, 141–146.[Web of Science][Medline]
Thorogood, M. (1993) Oral contraceptives and cardiovascular disease: an epidemiologic overview. Pharmacoepidemiol. Drug Safety, 2, 3–16.
Vandenbroucke, J.P. and Rosendaal, F.R. (1997) End of the line for the `third-generation-pill' controversy? Lancet, 349, 1113–1114.[Web of Science][Medline]
Submitted on August 13, 1998; accepted on January 29, 1999.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  O. Lidegaard, E. Lokkegaard, A. L. Svendsen, and C. Agger Hormonal contraception and risk of venous thromboembolism: national follow-up study BMJ, August 13, 2009; 339(aug13_2): b2890 - b2890. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. P. V. Gomes and S. R. Deitcher Risk of Venous Thromboembolic Disease Associated With Hormonal Contraceptives and Hormone Replacement Therapy: A Clinical Review Arch Intern Med, October 11, 2004; 164(18): 1965 - 1976. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. C G Skegg Oral contraceptives, venous thromboembolism, and the courts BMJ, September 7, 2002; 325(7363): 504 - 505. [Full Text] [PDF]
|
|
|
|
 |
|
 W. O. Spitzer, J. M. Faith, and K. D. MacRae Myocardial infarction and third generation oral contraceptives: aggregation of recent studies Hum. Reprod., September 1, 2002; 17(9): 2307 - 2314. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. M Kemmeren, A. Algra, and D. E Grobbee Third generation oral contraceptives and risk of venous thrombosis: meta-analysis BMJ, July 21, 2001; 323(7305): 131 - 131. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. P. Vandenbroucke, J. Rosing, K. W.M. Bloemenkamp, S. Middeldorp, F. M. Helmerhorst, B. N. Bouma, and F. R. Rosendaal Oral Contraceptives and the Risk of Venous Thrombosis N. Engl. J. Med., May 17, 2001; 344(20): 1527 - 1535. [Full Text] [PDF]
|
|
|
|
 |
|
 S. M. Bates, J. S. Ginsberg, S. E. Straus, H. Rekers, and D. L. Sackett Criteria for evaluating evidence that laboratory abnormalities are associated with the development of venous thromboembolism Can. Med. Assoc. J., October 1, 2000; 163(8): 1016 - 1021. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R D T Farmer, T J Williams, E L Simpson, and A L Nightingale Effect of 1995 pill scare on rates of venous thromboembolism among women taking combined oral contraceptives: analysis of General Practice Research Database BMJ, August 19, 2000; 321(7259): 477 - 479. [Abstract] [Full Text]
|
|
|
|
|
|
D. C G Skegg Third generation oral contraceptives BMJ, July 22, 2000; 321(7255): 190 - 191. [Full Text]
|
|
|
|
|
|
J. P Vandenbroucke, F. M Helmerhorst, F. R Rosendaal, R. Smith, and P. Hannaford Competing interests and controversy about third generation oral contraceptives BMJ, February 5, 2000; 320(7231): 381a - 381. [Full Text]
|
|
|
|
 |
|
 Oral contraceptives and cardiovascular risk DTB, January 1, 2000; 38(1): 1 - 5. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||