Do Third-Generation Oral Contraceptives Have a Higher Risk of Venous Thrombosis?

Abstracts & Commentary

Synopsis: There continues to be confusion over whether third-generation oral contraceptives are associated with a higher risk of venous thromboembolism compared with older products.

Sources: Kemmeren JM, et al. BMJ. 2001;323:131; Jick H, et al. BMJ. 2001;321:1190-1195.

Kemmeren and associates from the university Medical Centre in Utrecht in The Netherlands performed a meta-analysis of the studies assessing the risk of venous thromboembolism (VTE) associated with oral contraceptives (OCs). Of 114 identified studies, 13 were accepted for inclusion. The analysis concluded that the odds ratio (OR) for the risk of VTE was 1.7 (confidence interval [CI], 1.4-2.0) comparing users of desogestrel- and gestodene-containing OCs with users of levonorgestrel-containing products. Among first-time users, the OR was 3.1 (CI, 2.0-4.6) comparing the newer products to the older OCs.

Jick and associates from Boston University used the General Practice Research Database from general practices in the United Kingdom to perform additional case-control and cohort studies of the risk of VTE in users of OCs. The analyses compared users of levonorgestrel-containing OCs with users of desogestrel- and gestodene-containing OCs (so-called third-generation OCs). Because use of third-generation OCs dramatically declined after the reports in 1995-1996 that these products had higher risks of venous thrombosis (VT), Jick et al assessed the incidence of VTE in a cohort population before and after the "pill scare." In both periods of time, the risk of VTE was twice as great associated with third- generation products compared with pills containing levonorgestrel. Jick et al concluded that because there was no change before and after the pill scare, the difference reflects an inherent difference between products. In the case-control analysis, each case was matched with women of the same age who went to the same clinician and were using pills at the same time. Adjustments were made for body mass index, smoking, duration of pill use, and switching of pills. The OR for VTE in third- generation pill users compared with levonorgestrel users was 2.3 (CI, 1.3-3.9). In contrast to previous studies, smoking was found to be associated with a significant increased risk of VTE.

Comment by Leon Speroff, MD

This story goes back and forth. The original reports uniformly concluded that OCs containing gestodene or desogestrel had twice the risk of VTE compared with levonorgestrel-containing products.1-5 These results were immediately challenged. The studies were believed to reflect 2 consistent problems: preferential prescribing and the healthy user effect. Clinicians were swayed by advertising claims and believing the new products to be safer more often than prescribed third-generation products to new users and women perceived to be at higher risk. Individuals who do well with a product tend to remain with that product; therefore, individuals on an older product will be relatively healthy and free of side effects. Thus, comparing users of older and newer products involves 2 groups of women that are not identical.

Studies accounted for the above criticisms by focusing on first-time users and adjusting for duration of use. These studies failed to demonstrate a difference between third-generation products and levonorgestrel-containing OCs.6-9 Nevertheless, the meta-analysis summarized above reported the difference between third-generation and older products was still present in first-time users. They fail to report a clear analysis of the effect of correcting for duration of use. Preferential prescribing is dismissed by referring to "2 studies" (both from the same investigators) that, in fact, analyzed the effect of referral bias not preferential prescribing.10,11 I am not impressed with the quality of this meta-analysis.

The story is even more complicated because the latest Jick analysis reported above and the analysis by Farmer and associates used the exact same database and arrived at different conclusions!12 Of course, Jick et al believe their analysis is more refined and accurate. The editorial a month after the Jick article supports the conclusion that third-generation OCs have a higher risk of VT, citing a study that found acquired resistance to activated protein C to be more pronounced in users of these products.13 There was no mention of the fact that these results could not be confirmed by other investigators.14,15

Perplexed by these disagreements, I called Organon (accused in an editorial by Professor David Skegg from the University of Otago in New Zealand that analyses of the General Practice Research Database sponsored by the pharmaceutical company have been withheld, implying that the results were unfavorable toward the third-generation products).16 The major inconsistency in the studies that was pointed out to me was a quantitative and qualitative difference in the cases that were excluded from analysis. They further pointed out that all of their sponsored studies have not been withheld but have been published.

So what are we to think? I believe that the meta-analysis reviewed the same evidence I have reviewed and reached a different conclusion. I have no problem with that because meta-analyses of case-control and cohort studies are very subjective—actually only opinions because they reflect the interpretation of published literature by the authors. My opinion and your opinion are as valid as theirs. The difference in the analyses of the UK database is due to differences in the methods and the problem of trying to assess a clinical problem that has a low incidence; hence a small number of cases easily influenced by methods of analysis.

I remain convinced that the apparent differences associated with third-generation OCs were due to: 1) the marketing and preferential prescribing of new products; and 2) the characteristics of the patients for whom the new products were prescribed. There is a 3-to-4 fold increased risk of VTE associated with all low-dose OCs, a risk that is approximately one third that associated with pregnancy. Most studies have found smoking to be associated with arterial disease, not venous thrombosis. Because 99.85% of women who would test positively when screened for an inherited susceptibility for clotting will not have a clinical event, it is not cost-effective to perform routine screening prior to prescribing. However, screening for inherited disorders should be pursued in women with a previous episode of idiopathic VTE or a close positive family history (parent or sibling) of venous thrombosis.


1. World Health Organization. Lancet. 1995;346: 1575-1582.

2. World Health Organization. Lancet. 1995;346: 1582-1588.

3. Jick H, et al. Lancet. 1995;346:1589-1593.

4. Bloemenkamp KW, et al. Lancet. 1995;346:1593-1596.

5. Spitzer WO, et al. BMJ. 1996;312:83-88.

6. Suissa S, et al. Contraception. 1997;56:141-146.

7. Lidegaard O, et al. Contraception. 1998;57:291-301.

8. Farmer RD, et al. Lancet. 1997;349:83-88.

9. Lewis MA, et al. Hum Reprod. 1999;14:1493-1499.

10. Bloemenkamp KW, et al. Lancet. 1995;346:1593-1596.

11. Bloemenkamp KW, et al. Arch Intern Med. 1999;159: 65-70.

12. Farmer RD, et al. BMJ. 2000;321:477-479.

13. Skegg DC. BMJ. 2000;321:190-191.

14. Schramm W, Heinemann LA. Br J Haematol 1997;98: 491-492.

15. Heinemann LA, et al. Contraception. 1998;58:321-322.

16. Skegg DC. BMJ. 2000;321:1171-1172.