BACKGROUND: In 1995, 3 independent studies found that OCs containing the progestogens desogestrel or gestodene (third-generation OCs) doubled the risk of VTE over OCs containing levonorgestrel. A subsequent warning issued by the Committee on Safety of Medicines of England altered prescribing and use patterns accordingly, but no reduction of VTE incidence occurred, suggesting that the original research was flawed.
POPULATION STUDIED: Subjects were selected from more than 3 million patient records contained in the United Kingdom General Practice Research Database, which includes personal characteristics, drugs prescribed, and clinical diagnoses. The investigators identified women aged 15 to 39 years who were users of either third-generation OCs (n=361,724) or OCs containing levonorgestrel (n=979,052) and whose records had been contained within the database for at least 1 year. A total of 106 cases of idiopathic VTE occurred within this cohort. These cases were matched with 569 controls for the case-control analysis. Patients were excluded if their VTE was due to an apparent proximate cause, such as pregnancy, recent surgery or lower limb injury, cancer, or recent severe trauma.
STUDY DESIGN AND VALIDITY: Two study designs were employed, a cohort design and a nested case-control design. Data for both were analyzed separately for 2 time periods from January 1993 to October 1995 (before the safety warning) and from January 1996 to December 1999 (following the warning). For the cohort study, person-time at risk was estimated from the date of first prescription of oral contraceptive until either the OC was stopped; a different study OC was prescribed; the woman died, left the practice, or became a case; or the study period ended. Summary incidence rates for the 2 periods were adjusted for age and compared between the 2 types of OC users. Each patient with idiopathic VTE (case group) was then matched with up to 6 patients without VTE (control group) by age, practice, and index date (both using OCs on the date of the diagnosis of the case). Odds ratios were calculated for both study periods individually and combined, with adjustments for body mass index, smoking history, duration of use of OCs, and whether a change in OC had occurred. The authors describe significant measures to assure the quality of the General Practice Database and made blinded assessments of outcome. By excluding women with predisposing risk factors for VTE they appropriately limited the study results to average risk individuals. Since 1,340,776 women contributed only 361,300 person-years of observation over 7 years of observation, significant variation in length of OC use must have occurred. In fact, though more women took third-generation OCs, these contributed fewer person-years to the analysis. Several known confounders (smoking, obesity, duration of OC use) were controlled for in the case-control analysis.
OUTCOMES MEASURED: The first case of idiopathic VTE as measured by careful chart review was the outcome of interest. The chart review was conducted by blinded investigators.
RESULTS: Both the cohort and matched case-control analyses of these data demonstrated an approximately two-fold increase in risk of VTE among users of third-generation OCs compared with OCs containing levonorgestrel (age-adjusted relative risk [RR]=1.9; 95% confidence interval [CI], 1.3-2.8 and odds ratio [OR] =2.3, 95% CI, 1.3-3.9). No difference in the age-adjusted incidence of VTE was apparent between the 2 study periods, though the proportion of users of third-generation OCs dropped significantly after the warning notice was issued. VTE was also independently associated with higher body mass index and smoking.
This observational study supports earlier studies by demonstrating a two-fold association of third-generation oral contraceptives (those containing the progestins gestodene or desogestrel) with venous thromboembolism compared with OCs containing levonorgestrel. Since the overall incidence of VTE is still quite low for either group of OCs (3.8 vs 2.2 per 10,000 person-years, respectively), the evidence is not compelling enough to mandate switching current users of third-generation OCs. However until more definite evidence as from a randomized controlled trial becomes available, clinicians should favor OCs that do not contain either of these progestins when initiating or changing therapy.