Venous thrombosis in users of non-oral hormonal contraception: follow-up study, Denmark 2001-10BMJ 2012; 344 doi: https://doi.org/10.1136/bmj.e2990 (Published 10 May 2012) Cite this as: BMJ 2012;344:e2990
All rapid responses
In the Danish study the fundamental principle that exposure groups should be compared over the same time intervals is violated. The investigators state that follow-up was from 2001 to 2010. In fact, for women who were contraceptive users as of 2001, follow-up of COCs containing levonorgestrel (LNG) could actually have commenced as early as 1994, whereas for patch (EVRA) users follow-up could only have commenced in 2003, and for vaginal ring (NuvaRing) users in 2001 (the dates of introduction in Denmark). Only the occurrence of VTE was determined from 2001 to 2010.
For these reasons, with LNG-containing COCs as the reference, the duration-adjusted relative risks (RRs) of venous thromboembolism (VTE) for the patch (RR, 2.31) and vaginal ring (RR, 1.90) are not valid. Between 1994 and 2010, there were major advances in the detection and diagnosis of VTE (e.g., Duplex ultrasound), as well as increasing awareness of VTE risk among COC users. Moreover, women who had used LNG from as far back as 1994 without developing VTE would by 2001 have been at exceedingly low risk (depletion of susceptibles). Still further, the risk of VTE is highest in the first 3 months of COC use, followed by a decline to a lower but still elevated level thereafter1-3. Hence, in a valid comparison of different COCS it is essential to compare women who have not used COCs for at least 6 months (‘new users’) – and preferably, to compare women who have never used a COC previously (‘first-time-ever users’). And further still, the markedly reduced incidence of VTE in non-users of hormonal contraceptives (2.1 per 10 000 woman years) indicates that VTE was underascertained1-3, further augmenting the likelihood of bias.
The increased risk of VTE is a class effect dependent on the estrogen dose3, and it was predictable that a comparison of women who could have commenced use 7 or more years apart, and who also differed in the proportions of new and first-time-ever users, would have resulted in spuriously elevated RRs for the most recently introduced products.
The failure to control major confounding further invalidates the evidence. Body mass index and age are powerful determinants of VTE risk, and their combined effects are more than additive1,2. A family history of VTE is also a powerful determinant. The absence of data on these variables indicates that the Danish registry is an unsuitable resource for the evaluation of VTE risk.
1. Dinger JC Shapiro S. Combined oral contraceptives, venous thromboembolism, and the problems of interpreting large but incomplete datasets. J Fam Plann Reprod Health Care 2012;38:2-6.
2. Dinger JC, Heinemann LAJ, Kuhl-Habisch D. The safety of a drospirenone-containing contraceptive: final results of the European Active Surveillance study on oral contraceptives based on 142 475 women-years of observation. Contraception 2007;75:344-354.
3. Reid RL, Weshoff C, Mansour D, et al. Oral contraceptives and venous thromboembolism. Consensus opinion from and international workshop held in Berlin, Germany in December 2009. J Fam Plann Reprod Health Care 2010;36:117-122.
Competing interests: The author presently consults and in the past has consulted with manufacturers of hormonal contraceptives, including NuvaRing. In the past he has consulted with the manufacturer of EVRA.
Studies based on this Danish registry have already been extensively criticised (1) and the latest analysis suffers the same deficiencies, including failure to compare like with like, residual uncontrolled confounding, and biologically nonsensical results. First, one must compare new users with new users, as it is well-established that the risk of VTE is highest in the first 6 months of use of oestrogen-containing contraceptives. NuvaRing was launched in Denmark in late 2001 / early 2002, while the EVRA patch was launched there in September 2003. By contrast combined pills containing levonorgestrel have been in use since the 1970s, and those containing norgestimate since the mid 1980s. Thus, since the study period began in 2001, all users of NuvaRing and EVRA must be new users, and more likely to be first time users of hormonal contraceptives / women with risk factors. Meanwhile, the users of the comparator COCs are more likely to be long term users and therefore at lower risk, (attrition of susceptibles). The effect of duration of use is most clearly seen for NuvaRing in Table 4, where the relative risk is enormously lower going from less than one year of use, to 1-4 years of use. From Table 4 it is also evident that the numbers in each category are very small, leading to statistical instability (e.g., a total of 5 and 6 confirmed VTE for the implant and patch, respectively). Furthermore, the table shows an obvious artefact of the analysis: for the levonorgestrel IUS, the long-term contraceptive method par excellence, no risk estimate was calculated for use over 4 years, because “women …were censored after three years” due to missing data on exposure. Instead the long-term users were incorrectly adjudicated to the 1 to 4 year category.
Lidegaard et al again claim that their study is not materially affected by having no information on smoking, family history of VTE and BMI. However, whenever a new contraceptive comes onto the market, there is a tendency for contraceptive starters to be preferentially prescribed it, as well as women who have had problems with established products and those with risk factors for cardiovascular disease. For example, it has been shown that the combined pill, Yasmin, has been preferentially prescribed to obese women (2). In the EURAS study (1,2) adjustment for age, BMI, duration of current use, family history of VTE and the interaction between age and BMI reduced the VTE relative risk by 27%, as compared with an analysis that adjusted only for age.
Not surprisingly, neither of the two progestogen-only methods under study, was associated with a significantly increased risk of VTE; progestogen-only methods have not been implicated in VTE risk, since this is a class effect of oestrogen (3). Indeed, progestogen only methods are advised (in preference to oestrogen containing contraceptives) for women with risk factors for VTE (4). However, in the abstract, the authors misleadingly state that ‘the relative risk was increased in women who used subcutaneous implants’ and yet their relative risk of 1.4 was not statistically significant and thus compatible with chance. Paradoxically, users of the levonorgestrel intrauterine system had a statistically significant 40% reduction in the risk of VTE compared to women not using hormonal contraception, a finding totally lacking biological plausibility. Such an obvious artefact further illustrates the lack of credibility of this database analysis.
1. Dinger JC, Shapiro S. Combined oral contraceptives, venous thromboembolism, and the problem of interpreting large but incomplete datasets J Fam Plann Reprod Health Care 2012;38:2–6. doi:10.1136/jfprhc-2011-100260
2. Dinger JC, Heinemann LAJ, Kühl-Habich D. The safety of a drospirenone-containing oral contraceptive: final results from the European Active Surveillance study on oral contraceptives based on 142,475 women-years of observation. Contraception 2007;75:344-54.
3. Reid RL, Westhoff C, Mansour D, de Vries C, Verhaeghe J, Boschitsch E,et al. Oral Contraceptives and Venous Thromboembolism: Consensus Opinion from an International Workshop held in Berlin, Germany in December 2009 J Fam Plann Reprod Health Care 2010; 36(3): 117–122
4. Blanco-Molina MA, Lozano M, Cano A, Cristobal I, Pallardo LP, Lete I. Progestin-only contraception and venous thromboembolism Thrombosis Research 129 (2012) e257–e262
Competing interests: AS has received honoraria, consultancy fees and conference sponsorship from Bayer Schering (formerly Schering HealthCare), Schering Plough (formerly Organon), Janssen-Cilag and Wyeth. DM has received honoraria, consultancy fees and conference sponsorship from Bayer Schering (formerly Schering HealthCare), Schering Plough (formerly Organon), Janssen-Cilag, HRA Pharma and Wyeth.