Grey Journal Sept 2009 – No association between BMI and OCP effectiveness… or is there?
This month’s Grey Journal brings us a bunch of great articles, but here’s one that caught my eye:
I thought been pretty well established that oral birth control pills have somewhat lower effectiveness in patients with higher BMIs (30+), and then this study comes around and throws a wrench in the works: In 112,659 women years of exposure and 545 unplanned pregnancies, there was no statistical association between patient BMI and contraception failure.
An unexpected result, for certain. So here’s how they studied the questionThis study is a secondary analysis of the EURAS-OC study, a large European prospective cohort study intended to investigate the potential association between the new progestin drosperinone and cardiovascular events, created by Bayer at the demands of European drug authorities. This study created a huge dataset of women on oral contraceptives, including demographic data and long term clinical outcomes. In order to investigate the effects of BMI on contraceptive failures, our compared rates of unintended pregnancy in different groups of women within the EURAS-OC study. Given the large dataset, the authors were able to stratify risk of unintended pregnancy not only by BMI, but also by progestin.
After data analysis, overall failure rates were strikingly low – Year 1 failure rate was 0.75%, year 2 1.33%, year 3 1.53%, and year 4 1.67%. Furthermore, for all progestins except for chlormadinone (which is not available in the US), there was no association between BMI and failure rate. With chlormadinone there was a 3x risk of failure rate at BMI > 30, which is thought to be due to the highly lipophilic nature of that particular progestin, and thus a higher relative volume of distribution compared to other progestins.
So why is that all worth talking about? – Its because this data is just so different that what we have seen before –
Holt et al Green Journal 2005 – case control of unintended pregnancies on OCPs, odds ratio of failure 1.58 – 2.22 for BMIs > 27, depending on level of obesity and consistency of OCP use.
Brunner et al Ann Epidemiol 2006 – case control of unintended pregnancies on OCPs in South Carolina, odds ratio of failure 2.54 – 2.82 for obese women.
Brunner et al Matern Child Health 2005 – retrospective cohort of 18,445 women, adjusted OR for unintended pregnancy in obese women 1.73 – 1.75.
Furthermore, the overall failure rate of around 1% is much different than the 9% first year failure rate reported in the last iteration of the US National Survey of Family Growth.
When we see such different results of studies that seem to ask the same question, one of the following must be true: 1) One of the studies is biased, or both are 2) The studies are looking at different populations or 3) The studies may be defining their outcomes differently. In this case, I think all three are happening.
The biggest difference that I see between this study and the American studies is the way patients were recruited, and think this difference could have a strong effect on the outcomes. This study was a prospective cohort with data gathered over many years. Each patient studied was approached after they had decided to use OCPs. If they decided to enter the study, they were closely monitored for years for a large number of outcomes, and had medication compliance frequently measured. The American studies were either case control studies looking at contraceptive failures compared to matched controls, or in the case of the second Brunner study, a retrospectively constructed cohort. So why does this matter? It goes back to an old engineering saying: Anything you measure you will improve. The patients in the European study knew they were in a study, and that they would be asked to report their OCP compliance, making them more likely to be compliant with medications than someone who does not know their outcomes would be analyzed. 2 American case control studies suggested that the effects of obesity on pill efficacy is pronounced in women who are not consistent with their pills. As such, it is possible that this huge study misses the real life effect of obesity by producing a preposterous population of peristently perfect pill popping patients!
Another difference between the European study and the previous American data is the type of obesity seen in the study populations. Unfortunately for America, we have a lot more obesity than Europe. This study had a mean BMI of around 22, and sadly the current mean BMI in the United States is 26.4. This translates to a larger number of superobese ( > 35 ) in American study populations, which are not as well addressed in this large data set composed of less obese people.
Finally, overall failure rate is calculated much differently in the European study that in the large American population studies. In the American studies women-years are counted in month by month increments, but only when women are sexually active. The European study counts a woman-month irrespective of whether or not the woman has had sex in that month. It is arguable which of these methods is more appropriate, but inarguable that it has a profound effect on the absolute value of the failure rate. As we (the Americans) are counting months only when there is intercourse during the month, there will assuredly be a higher failure rates than if we excluded the non-exposed months, which increase the denominator(exposure) without possibly increasing the numerator(failures.) As such, the overall failure rate of around 1% is difficult in this study is hard to compare to the American failure rate of 9%, as it could represent decreased sexual activity in Europe relative to America as much as it could represent different contraceptive efficacy.
So what do we take from all this? I think this new study shows us that if people are highly compliant with OCPs, failure rates are very low and are not strongly affected by BMI – but given the potential biases created by study design and interpretation, I’m not ready to say that OCPs work just as well in obese and superobese women as they do in lower BMI women.
So what do you think?
Dinger JC, Cronin M, Möhner S, Schellschmidt I, Minh TD, Westhoff C. Oral contraceptive effectiveness according to body mass index, weight, age, and other factors. Am J Obstet Gynecol. 2009 Sep;201(3):263.e1-9. Epub 2009 May 30.