By Jeffrey T. Jensen, MD, MPH
Synopsis: A review of incident cases of venous thrombosis in women using oral contraceptives (OCs) in a U.S. insurance database found that use of drospirenone-containing OCs containing 20 μg of ethinyl estradiol (EE) was associated with a two-fold increase in risk compared to levonorgestrel pills. This risk was not increased with use of 30 μg EE drospirenone pills.
Source: Bird ST, et al. Drospirenone and non-fatal venous thromboembolism: Is there a risk difference by dosage of Ethinyl-Estradiol? J Thromb Haemost 2013; Apr 11 [Epub ahead of print].
Several prior database studies have demonstrated an increased risk of venous thromboembolism (VTE) with third-generation and drospirenone (DRSP)-containing combined oral contraceptive (COC) pills when compared to levonorgestrel (LNG)-containing COCs, but the expected estrogen dose-response relationship has not been consistent. To determine whether DRSP and ethinyl-estradiol 20 μg (DRSP/EE20) had a lower VTE risk than DRSP and ethinyl-estradiol 30 μg (DRSP/EE30), the authors performed a retrospective cohort analysis using the IMS Life-Link Health Plan Claims Database from the United States. The cohorts were assembled using the claims database to identify women aged 18-46 years taking DRSP- or LNG-containing COCs. The diagnosis of VTE was defined using ICD-9-CM coding and required anticoagulation. The VTE relative risk (RR) was calculated using a Cox proportional hazards model comparing DRSP and LNG users, and the cohorts were further stratified by EE dose and user-type (new/prevalent). A propensity score was used to control for baseline comorbidities such as BMI and other potential confounders.
This large database included 238,683 DRSP and 193,495 LNG users. Overall, the incidence of VTE was 18/10,000 women-years (WY) among DRSP users, and 8.9/10,000 WY among women using LNG COCs (relative risk [RR], 1.90; 95% confidence interval [CI], 1.51-2.39). However, when EE dose was considered, the risk was confined to new users of DRSP/EE20 pills (RR, 2.35; 95% CI, 1.44-3.82) compared to new users of LNG/ EE20. In contrast, an increased risk for DRSP/EE30 new users relative to LNG/EE30 new users was observed only among women initiating COCs between 2001-2006 (RR, 2.51; 95% CI, 1.12-5.64), with a reversal of the point estimate seen during 2007-2009 (RR 0.76, 95% CI; 0.42-1.39). A decrease in the risk of VTE with a reduction in EE dose was not observed: In direct comparison DRSP/ EE20 had an elevated VTE risk relative to DRSP/EE30 (RR, 1.55; 95% CI, 0.99-2.41), but this was not statistically significant.
The DRSP VTE controversy may seem like old news, and the issue has been extensively covered in previous issues of OB/GYN Clinical Alert. Although the issue may seem resolved for those of us in the United States with new product labeling for DRSP-containing pills, the issue is still contentious in Europe. The controversy was recently reignited by highly publicized cases of VTE in young women using the cyproterone acetate (CPA)-containing COC (Diane 35®) in France. Although this pill (not available in the United States) has been on the market for more than 30 years, a cluster of reports of rare VTE events led to public outcry and a national investigation that resulted in the withdrawal of Diane from the French market and to recommendations for further restrictions on COC prescribing. This new firestorm has been powered by the recent publication of case-control and database studies supporting an increased VTE risk associated with use of third-generation progestogens, CPA, and DRSP compared to LNG. Although highly consistent, the risk estimates that come from these studies are controversial, as similar results were not observed in the true prospective study conducted by Dinger et al.1 The wind that fans the fire blows directly from Denmark in the form of the large Danish database studies published by Lidegaard.2,3 More about the European story will come in a Special Feature in our next issue, but it seems appropriate to comment on the Bird study now.
Since VTE is a rare (defined as an incidence of < 1/ 1000) event in otherwise healthy young women using the pill, large numbers of individuals must be studied to obtain statistically valid comparisons. For this reason, it is easy to see the attraction and rationale behind the use of large national insurance claims databases to address rare serious adverse events. Unfortunately, the Achilles’ heel of claims database studies is the inability to accurately evaluate the presence of important baseline confounders. Confounders are factors that are associated both with inclusion in a cohort and with the outcome of interest. In modern database studies, complex schemes such as propensity scores are constructed to adjust for the presence of missing confounders that are simply not available in the record. Unfortunately, this is subjective, and the technics used can artificially influence the results. As inconvenient as it may be, there is no substitute for real prospective data. This makes interpretation of risk estimates smaller than 2-3 extremely hazardous.
Personally, I don’t find the evidence convincing that the type of progestogen in a COC modifies the risk of VTE. One of the basic issues I struggle with is the lack of biologic plausibility, as no definitive mechanism for a progestogen-effect on coagulation has been demonstrated. In contrast, the biologic response of the coagulation system to estrogen is well established. This response to estrogen is a physiologic and life-saving evolutionary tactic to prevent hemorrhage during childbirth. A reduction in VTE risk in response to a decrease in EE dose has been observed for more than 40 years.4 Not surprisingly, the resolution of epidemiologic studies has proven insufficient to determine whether a continued decrease in the dose of EE below 35 μg further reduces VTE risk. In fact, a paradoxical increase in risk with 20 μg compared to 30 μg desogestrel-containing pills was observed in several studies during the third-generation progestogen pill scare of the 1990s.5-7 Taken together, these observations led many experts to conclude that the third-generation progestogen scare was unwarranted, and the data seemed seriously compromised by preferential prescription of newer pills to high-risk women. Denying this suggests a fantasy world contrary to our current understanding of the coagulation system. The inverse dose-response reported by Bird et al in the current manuscript is consistent with these earlier reports and equally unlikely to be true.
Although at first glance the overall elevated risk associated with DRSP compared to LNG pills reported in the Bird study seems to support the findings of earlier studies, the data actually provide evidence that the increased risk association reported by Lidegaard and others may in fact be spurious. The drop in risk for the 30 μg DRSP pill over time demonstrates both a healthy user effect and a likely shift of high-risk users to both LNG pills and 20 μg DRSP pills.8 Taken together, this suggests that the observed association of an increased risk associated with DRSP pills in all database studies is an artifact attributable to preferential prescribing to high-risk women. Look for more about this in the August Special Feature.
1. Dinger JC, et al. Contraception 2007;75:344-354.
2. Lidegaard O, et al. BMJ 2012;344:e2990.
3. Lidegaard O, et al. BMJ 2011;343:d6423.
4. Inman WH, et al. Br Med J 1970;2:203-209.
5. Farmer RD, et al. Hum Reprod Update 1999;5:688-706.
6. Effect of different progestagens in low oestrogen oral contraceptives on venous thromboembolic disease. Lancet 1995;346:1582-1588.
7. Spitzer WO, et al. BMJ 1996;312:83-88.
8. Farmer RD, et al. Eur J Contracept Reprod Health Care 1996;1:31-37.