oral contraceptives

A huge observational study was done in France looking at women on oral contraceptives (OCs), looking at the relative dose of the hormones and the risk of pulmonary embolism (PE), ischemic stroke and MI (http://www.bmj.com/content/353/bmj.i2002 ). details:

--4,945,088 women aged 15-49 with at least one reimbursement for OCs and no prior history of cancer, PE, ischemic stroke or MI, between 2010-2012

--mean age 28, 34% used oral contraceptive with 20 mg estrogen,

--risk factors used in models: age, socioeconomic data (SES), hypertension, diabetes, and some indirect measures of smoking (eg, getting nicotine replacement therapy) though smoking was not directly assessed in the database. the incidence of hypertension, diabetes or being prescribed nicotine replacement therapy were each <2%.

--results:

    --5,443,916 women-years of OC use

    ​--1800 PEs (33/100K women-years), 1046 ischemic strokes (19/100K women-years), 407 MIs (7/100K women-years)

    --women >35 yo had 26.1% of all OC usage and 56.7% of these serious adverse events

--women in the lowest SES had 1.4-fold higher risk of PE, 1.5-fold increase in stroke and 2.5-fold increase in MI

    --by estrogen dose (ethinyl-estradiol), adjusting for progestogen and risk factors, and compared to levonorgestrel:

        ​--low dose (20 mg vs 30-40 mg): 25% decrease PEs [RR 0.75 (0.67-0.85)]; 18% decrease ischemic strokes [RR 0.82(0.70-0.96)]; 46% decrease in MI [RR 0.56 (0.39-0.79)]

    --by progestin type, adjusting for estrogen dose and risk factors:

        --desogestrel associated with 116% higher risk for PE [RR 2.16 (1.93-2.41)]; gestodene 63% higher risk [RR 1.63 (1.34-1.97)]; no significant difference for MI or stroke

    --levonorgestrel combined with 20 mg estrogen was statistically better than levonorgestrel combined with 30-40 mg​ estrogen

this is an observational study, but a really large one, with several inherent limitations:

--short followup period of max 2 years, 3 months. likely that the arterial events (MI and stroke) take longer to manifest themselves than PE (which might suggest that longer followup may have produced even more bad events with the higher doses estrogen/use of desogestrel or gestodene)

--no start dates  (a concern since the incidence of PE is highest early on, esp in the first 3 months). so, unclear when people were started on the low dose estrogens, and in fact these were available in France only more recently. however, when analysis was limited to postpartum women who have a pretty clear start date, the results were similar to the overall group

--possible confounding by indication: did doctors choose different OCs depending on BMI, or family predisposition to venous/arterial disease? or those they felt to be at higher cardiovascular risk? this bias would, however, lead to more use of lower dose OCs and would therefore likely decrease the magnitude of the protective effect found with low-dose OCs. 

--one of the biggest deficiencies is the lack of clear smoking data, but again these women were likely to be put on the lowest dose OCs, which would tend to understate their risk (of note, other studies have not found that smoking modified the risk much)

--also, they only looked at PE, not other venous thrombotic events (eg, DVTs)

--so, it seems that the stars continue to align on this (and, see blog from last year on another large study finding that the newer OCs had much higher rates of venous thromboembolic disease: http://gmodestmedblogs.blogspot.com/2015/06/new-vs-old-ocps-and-thromboembolism.html  . to me, it makes sense to use OCs with levonorgestrel unless there is a pretty compelling reason not to, and with whatever progestogen, use the lowest dose of estrogen one can.