Oral Contraceptives and the Risk of Myocardial Infarction
Abstract & Commentary
Synopsis: Third-generation oral contraceptives may be associated with less cardiovascular morbidity than first- and second-generation OCs.
Source: Tanis BC, et al. N Engl J Med. 2001;345:1787-1793.
Because the literature has been contradictory regarding the risk of myocardial infarction (MI) and which progestin is contained in an oral contraceptive (OC), this group from The Netherlands performed a retrospective, case-control study, called the Risk of Arterial Thrombosis in Relation to Oral Contraceptives (RATIO).
RATIO enrolled women aged 18-49 years who presented to 1 of 16 centers with their first MI between January 1990 and October 1995. They were identified by searching hospital discharge databases for acute MI diagnosis codes. To qualify as an acute MI, the patient had to be symptomatic and have elevated cardiac enzymes and electrocardiographic changes. Three hundred twenty-one (321) women were identified. Of these 29 were excluded, 28 who died and 1 who could not participate. Of the remaining 292, 21 could not be found and 23 chose not to participate. This left 248 women.
The control group contained 925 women, stratified by age, who were recruited by random-digit dialing in 6 geographic areas corresponding to where the patients lived. Women who had a history of coronary, cerebral, or peripheral vascular disease were excluded. They were questioned about demographics, weight, height, reproductive history, menopausal status, level of education, family history of cardiovascular disease, hypertension, diabetes, hypercholesterolemia, alcohol use, tobacco use, and use of OCs. Tanis and colleagues grouped the OCs by generation and estrogen content. Because the gene mutations factor V (Leiden) and G20210A in the prothrombin gene have been associated with MI in young women, Tanis et al tested blood or buccal swab samples from 217 patients and 763 controls for these.
The patients were older than the controls (42.7 vs 38.1 years). As expected, the patients were more likely to be hypertensive, hypercholesterolemic, and diabetic. They smoked more and had a lower level of education. Their families were more likely to be afflicted with cardiovascular disease. There was no difference in menopausal status. After demographic analysis, 3 patients and 9 controls were excluded for using hormone replacement therapy or because the researchers could not be sure they had used OCs.
Of the remaining 245 patients, 99 (40%) had used OCs of any generation. Of the remaining 916 controls, 348 (38%) had used OCs. The odds ratio (OR) for MI among OC users, after adjustment, was twice that of nonusers. When the ORs were calculated based on which generation of OCs were used, the results were 2.8 (confidence interval [CI], 1.3-6.3) for first-generation, 2.4 (CI, 1.6-3.6) for second-generation, and 1.3 (CI, 0.8-2.3) for third-generation OCs.
The amount of estrogen did not seem to affect the results. When Tanis et al compared women who used second-generation OCs with levonorgestrel and either 35 µg or 50 µg of ethinyl estradiol, the ORs were 2.6 (CI, 1.6-4.2) and 2.0 (CI, 0.6-7.3), respectively, a non-significant difference.
The traditional cardiovascular risk factors of smoking (OR, 7.9; CI, 4.9-12.9), hypertension (OR, 5.1; CI, 2.9-8.8), hypercholesterolemia (OR, 3.3; CI, 1.6-6.8), diabetes (OR, 4.2; CI, 1.6-10.9), and obesity (OR, 3.4; CI, 2.2-5.3) all increased the chance of MI among nonusers. Among OC users, the ORs were even higher: smoking (13.6), hypertension (6.1), hypercholesterolemia (24.7), diabetes (17.4), and obesity (5.1). The presence of either gene mutation did not increase the risk.
Comment by Allan J. Wilke, MD
Almost all OCs have ethinyl estradiol as their estrogen component, except for a handful that contain mestranol. Estrogen-containing OCs are divided into 3 generations, based on their progestin content. First-generation OCs contain lynestrenol (not available in the United States) or norethindrone (for example, Brevicon®, Modicon®, Norinyl 1/35®, and Ortho-Novum 1/35®). Second-generation OCs contain norgestrel (Lo/Ovral®, Ogestrel-28®) or levonorgestrel (Alesse®, Tri-Levlin®, Triphasil®). Third-generation OCs contain desogestrel (Desogen®, Ortho-Cept®) or gestodene (not available in the United States). Some authorities include norgestimate (Ortho-Cyclen®, Ortho-Tricyclen®) as a third-generation OC, but Tanis et al did not. The third-generation OCs were marketed as being safer because of their favorable effect on lipid profiles. HDL cholesterol levels tended to be higher. In evidence-based medicine circles, this is known as disease-oriented evidence. Having a more favorable lipid profile may eventually prove to be a good thing, but it is premature to conclude that third-generation OCs are safer until we can look at "patient-oriented evidence that matters." This study is a first, hesitant step in that direction.
We have known about the association of OCs and MIs for almost 40 years. Over the years, attempts have been made to make them safer. Since we assumed that the main culprit was the estrogen component, we first reduced the amount of estrogen in each pill. As this study shows, reducing the estrogen content does not seem to reduce the risk; if anything, the ORs favored the higher 50-µg dose (statistically not significant). We then changed the progesterone component. The third-generation OCs have been associated with a small increase in the risk of venous thromboembolism (VTE). Indeed, Tanis et al just published a meta-analysis showing an increased risk of VTE with third-generation OCs.
As this is a retrospective, case-control study, we must examine its methods closely. The main concern is whether the patients and the controls were well matched. Are there unknown differences that distinguish the 2 groups? Other secondary concerns are the search strategy, the possibility of recall bias, and the definition of MI. Their search strategy potentially missed women who had MIs and either died or did not present to the hospital. Tanis et al relied on women’s memory of what OC they took (augmented by color photographs of available OCs). The definition of MI used in this study is strict; most practitioners in the United States would accept 2 of the 3 criteria as diagnostic. It could exclude diabetic patients who had a "silent" MI.
As with any report, it is prudent to compare your population with that of the study. These women were predominantly white (94% patients and 93% controls). Interestingly, they were not particularly well educated. Among the patients, 53% did not get beyond primary school (30% for controls) and only 10% of patients and 27% of controls had attended postsecondary school. This runs counter to my image of the Dutch. They were not extraordinarily obese (BMI = 25.7 for patients and 23.5 for controls), but they smoked like chimneys. Fully 84% of patients and 43% of controls were current smokers; Toledo, Ohio, my stomping grounds, is the smoking capital of metropolitan United States, and we are at only 31%.
I found it frustrating that Tanis et al did not "show their math" as they adjusted their data when calculating the ORs. The ratios may be legitimate, but I could not calculate them using the raw data in the tables. This makes accepting Tanis et al’s conclusions rather like a matter of faith.
Let us put things into clinical perspective. Among all people younger than 25 years in the United States, the risk of death from an acute MI is 0.1 per 100,000 population. For 25-to-34-year-olds it is 1.3, and for 35-to-44-year-olds it is 7.8. Compare this to risk of death related to pregnancy among American women. For women younger than 20 years old, it is 8.5 per 100,000 live births. Similarly, for 20-to-29-year-olds, 30-to-34-year-olds, and 35-to-39-year-olds, the numbers are 9.3, 11.9, and 21.1, respectively. In other words, even with a doubling of the MI rate (as purported by this article), it is much more dangerous to be pregnant than to use OCs, provided none of the traditional risk factors are present. In a study of causation, you want the ORs to be large. The ORs for MI between the various generations are small, 1.3-2.8. Byway of comparison, the ORs for OCs and smoking, hypercholesterolemia, and diabetes are greater by an order of magnitude.
The risk of MI is not the first thing I worry about when I prescribe an OC. When I prescribe these drugs, I am usually trying to prevent a pregnancy, normalize a menstrual cycle, or control acne. My risk-benefit considerations include the risk of pregnancy, the risk of deep vein thrombosis, and the risk of exacerbating an estrogen-sensitive neoplasm. I ask about migraine, hypertension, and family history. As Tanis et al argue, since OCs are essentially equivalent in their effectiveness, safety should drive our prescribing habits. Among OCs, the third generation appear to be the safest regarding MI. They may be less safe for venous thromboembolism. We will have to wait until later publications to learn about how safe they are in cerebral infarction and peripheral artery disease.
Dr. Wilke, Assistant Professor of Family Medicine, Medical College of Ohio, Toledo, OH, is Associate Editor of Internal Medicine Alert.