3
May 16, 2012
To the Editor, British Medical Journal:
Lidegaard et al.1 describe a retrospective database study designed to assess the risk of venous thrombosis in current users of non-oral hormonal contraception in Danish non pregnant women ages 15-49 followed from 2001-2010. The authors suggest an increased risk of venous thromboembolism (VTE) in users of NuvaRing® (etonogestrel/ethinylestradiol containing vaginal ring) compared with users of levonorgestrel-containing combined oral contraceptives (LNG-COCs). Methodological issues common to retrospective database studies of VTE risk and contraceptive use call into question the validity of these findings.
Well-designed prospective cohort studies of new users of contraceptives can address many of the limitations inherent in retrospective database studies such as the study by Lidegaard et al1. The risk of VTE associated with NuvaRing® use has been studied in a large, controlled, prospective, open-label observational trial, Transatlantic Active Surveillance on Cardiovascular Safety of NuvaRing® (TASC) funded by Merck, (known as MSD outside the US and Canada). In this study, incidence rates of VTE were similar for the NuvaRing® and COC cohorts. (Dinger J and Pineda AA, Risk of VTE in Users of an Etonogestrel-Containing Vaginal Ring and Combined Oral Contraceptives, presented at the American College of Obstetricians and Gynecologists (ACOG) Annual Clinical Meeting May 7, 2012).
One fundamental limitation of the Lidegaard study is that comparisons were made among current contraceptive users. The study evaluates the risk of VTE among current users of contraceptives together, combining both continuous and new users into a single risk group. This is problematic, because the risk of VTE is not uniform among new and continuous users. The risk of VTE is highest in the first year of use of any combined hormonal contraceptive (CHC). Women who are new users of a CHC will be at higher risk of VTE while women with continuous use of a CHC will be at lower risk. By combining these groups, the authors fail to account for differential risk due to time period of exposure to the CHC, resulting in a potential bias against a newer product. This bias is observed especially when one CHC has been available for a relatively long time (more continuous users) and the second CHC for a shorter period of time (more new users). NuvaRing® had been marketed in Denmark since October 2002 compared with LNG-COCs available since the 1970s. When comparing the incidence rates of VTE among NuvaRing® and LNG-COC users, the minor differences (7.75 and 6.22 per 10,000 woman years respectively) might easily be accounted for by the differences in introduction dates and the fact that continuous users of LNG-COCs at the study window start date of 1-1-2001 were included in the database and likely had a lower risk of VTE. Any early VTEs in this group were not included in the database, because the events occurred before the study window.
The authors also failed to adequately adjust for a number of factors known to be associated with an increased risk of VTE: smoking, body weight, family history of venous thromboembolism, and the duration of current and past hormonal contraception use. Only age, calendar year (as a proxy for body weight) and education (as a proxy for smoking status) were accounted for in the analysis, resulting in unexplainable and inconsistent adjustment effects across the comparisons.
The TASC study enrolled more than 33000 women in the US and Europe who were followed for up to 4 years (2007-2011), and was designed to compare the short-term and long-term risks of venous thromboembolism in new users of NuvaRing® compared to new users of combined oral contraceptives (COCs) during standard clinical practice. The study design allowed for documentation of all relevant risk factors, exclusion of existing long-term use before study start; differentiation between starter, switcher, and recurrent user status; documentation of all patient-reported outcomes with attending physicians; blinded adjudication of all VTEs; and comprehensive follow-up procedures to minimize loss to follow-up. Incidence rates of VTE, similar for the NuvaRing® and COC cohorts, are displayed in the Table.
The Table displays the incidence of VTE associated with NuvaRing® and COCs and adjusted rate ratios in the Lidegaard and TASC studies
Table. VTE Incidence (per 10,000 exposure years) and
Adjusted Rate Ratios
Study / TASC / LidegaardNuvaRing® / 8.3 / 7.75
COCs / 8.5* / 6.22**
Adjusted relative risks/hazards for NuvaRing v. COCs (CI) / 0.9 (0.4-2.0)* / 1.9 (1.3-2.7)**
* COCs excluding desogestrel, gestodene, and drospirenone
** LNG-COCs
CI = 95% confidence interval
Methodological issues inherent in retrospective database studies examining VTE incidence and CHC use limit interpretability of findings. In the Lidegaard study, failure to a) limit comparison of VTE rates to new users and b) to account for key risk factors for VTE increase the potential for bias in the results. The discrepant findings between the Lidegaard and TASC studies illustrate the biases that are likely to result from retrospective studies.
Reference
1. Lidegaard Ø, Hougaard Nielsen L, Skovlund CW, Løkkegaard E. Venous thrombosis in users of non-oral hormonal contraception: follow-up study, Denmark 2001-10, BMJ 2012;344:e2990.