05.27.2015
by Molly Walker
Contributing Writer
Women taking oral contraceptive pills had slightly higher risks of developing a clot, especially if the pill was a newer formulation, according to the results of two nested case-control studies from the U.K.
Yana Vinogradova, division of primary care, University Park in Nottingham, and colleagues reported any use of an oral contraceptive was associated with an increased risk of a venous thromboembolism (adjusted odds ratio 2.97, 95% CI 2.78-3.17) compared with women with no oral contraceptive use in the previous year.
But that risk seemed to vary by contraceptive formulation — newer third and fourth generation pills appear to have about a fourfold relative risk increase versus about 2.5 times increase for older, second generation pills, they wrote in The British Medical Journal.
Merle Myerson, MD, director, Center for Cardiovascular Disease Prevention at Mount Sinai St. Luke’s and Roosevelt hospitals in New York City, found this to be a well conducted analysis.
“This study appears to be very well designed and yields important information regarding risk for VTE associated with oral contraception use,” she said. “This is a very relevant clinical question.”
But MedPage Today clinical reviewer, F. Perry Wilson, MD, assistant professor of medicine at Yale University, took issue with some of the study’s statistical associations in a separate video commentary, calling the absolute risk very small and offering up a hypothesis for the increased risk from newer contraceptives.
“It’s conceivable that docs who prescribe these newer, fancier oral contraceptives are more vigilant when it comes to looking for clots and inflate their diagnosis rates,” he said.
The “older” group of contraceptives contained levonorgestrel (AOR 2.38, 95% CI 2.18-2.59), norethisterone (AOR 2.56, CI 2.15-3.06) and norgestimate (AOR 2.53, CI 2.17-2.96), while the “newer” group was comprised of desogestrel (AOR 4.28, CI 3.66-5.01), gestodene (AOR 3.64, 3.00-4.43), drospirenone (AOR 4.12, 3.43-4.96) and cyproterone (AOR 4.27, CI 3.57-5.11).
The number needed to harm ranged from six for levonorgestrel (95% CI 5-7) and norgestimate (CI 5-8) to 14 for desogestrel and cyproterone (CI 11-17, respectively) per 10,000 women ages 15-49 years.
When compared with existing studies, risks associated with gestodene were about 1.5 times higher than for levonorgestrel and about 1.8 times higher for desogestrel, drospirenone, and cyproterone.
Examining short-term compared with long-term use, only levonorgestrel was associated with clear increased risk (AOR 3.38, 95% CI 2.86-3.99 and AOR 2.15, CI 1.97-2.38, respectively, P<0.001), with other formulations only producing marginally higher or lower results the authors deemed “inconsistent.” Increased risk for levonorgestrel was significant because it was by far the most commonly prescribed contraceptive, with approximately 45% of the cases and 53% of the controls taking the drug.
Not surprisingly, other risk factors — including body mass index — played a role in the increased risk from any of these drugs, with body mass index increasing odds ratios a range of 7% to 10%. Risk factors included:
- Chronic conditions, such as cancer, congestive heart failure, varicose veins, and cardiovascular disease
- Traumatic events within 6 months before index data, such as acute infections or surgery
- Obesity
- Smoking
- Polycystic ovary syndrome
- Alcohol consumption
- Ethnic group
Combined, these risk factors increased odds ratios by percentages of 13% to 25% for any drug exposure.
Risks for VTE increased further when researchers examined women already taking anti-coagulation medication. They note, however, that these numbers may reflect “differential treatment” of patients based on known risk factors:
- 8% increase for northisterone
- 24% for levonorgestrel
- 40% for norgestimate and cyproterone
- 46% for desogestrel
- 48% for drospirenone
- 78% for gestodene
Sensitivity analyses only found significant heterogeneity for norgestimate (I2=89%, P=0.003), but the authors concluded there was no impact on the earlier estimates (combined odds ratio 2.49, 95% CI 1.56-3.57).
Researchers examined data from two large databases of primary care records in the U.K., QResearch and the Clinical Practice Research Datalink (CPRD), with women ages 15-49 with a first diagnosis of VTE from 2001 to 2013, matched with up to five controls by age, practice, and calendar year. VTE diagnoses were based on clinical Read codes. Final analysis included 5,062 VTE cases matched to 19,638 controls from CPRD and 5,500 VTE cases matched to 22,396 controls from QResearch. More than half of all VTE events from both databases (57%-58%) were classified as deep vein thrombosis only.
The median age of participants was 38 to 39 years across both samples, with nearly half of the cases containing established risk factors for VTE (47% each) compared with only a quarter of the controls (26%-27%). Smoking was more common among cases versus controls (27% versus 21% for both), as was obesity (BMI ≥30: 30% versus 17% for CPRD, 24% versus 14% for QResearch).
In the year prior to the study, around 30% of women had prescriptions for oral contraceptives in the cases group compared with 15% to 18% in the control group, and the vast majority of both cases (84%) and controls (77%-79%) were current users.
“Only Statistical Associations”
In a separate e-mail interview with MedPage Today, Vinogradova said she wanted to create a more comprehensive study on oral contraceptive use and VTE risk.
“Our aim was to make more accurate and reliable estimates of VTE risks associated with all combined oral contraceptive types prescribed in the U.K.,” she said. “Our findings are only statistical associations between use of different contraceptive drugs and VTE risks, but they represent more comprehensive and reliable information for doctors making evidence-based prescribing decisions.”
Vinogradova added a comment from her co-author, Professor Julia Hippisley-Cox, putting the findings in perspective.
“We have found a higher risk of thrombosis for some newer type of oral contraceptive pills in this research project [but] the risk is lower [than] risk of thrombosis which naturally occurs during pregnancy.”
Limitations noted by the authors included potential misclassification of exposure to oral contraceptives because it was based on prescription data, not actual use, as well as the fact that data from women getting their contraceptives from National Health Service (NHS) clinics would not be available in these databases. In terms of VTE diagnoses, they cited misclassification bias, since diagnostic tests to confirm VTE are not available in the primary care database.
The authors also commented on the higher odds ratios for cases with anticoagulant prescriptions — a limitation Wilson also addressed in his commentary.
“The best way to mess up a case control study is to choose the wrong controls,” he said, adding “Cases were excluded if they had received an oral anticoagulant within 42 days of entry into the cohort, whereas the controls were excluded if they had ever received an oral anticoagulant.”
While the authors concluded that they believe “this study has the statistical power and sufficient adjustment for relevant confounders to be regarded as an important clarifying study,” Vinogradova said separately that she would like to see a more thorough investigation into the role that contraceptives play in a woman’s risk for VTE.
“A comparable study to this one, which included more detail of actual use (our study measured prescriptions) and which could assemble a more complete and wider range of data for all possible factors affecting VTE risks and patient choice, may improve the accuracy and reliability of our findings,” she said.
The authors declared no competing interests.
Reviewed by Henry A. Solomon, MD, FACP, FACC Clinical Associate Professor, Weill Cornell Medical College and Dorothy Caputo, MA, BSN, RN, Nurse Planner
Primary Source
The British Medical Journal
Source Reference: Vinogradova Y, et al “Use of combined oral contraceptives and risk of venous thromboembolism: nested case-control studies using the QResearch and CPRD databases” BMJ 2015; DOI: 10.1136/bmj.h2135.