|
|
| |||||||||||||||||||||||||||||||||||||||||
Background Previous studies have linked the use of oral contraceptive agents to an increased risk of stroke, but those studies have been limited to oral contraceptives containing more estrogen than is now generally used.
Methods In a population-based, case–control study, we identified fatal and nonfatal strokes in female members of the California Kaiser Permanente Medical Care Program who were 15 through 44 years of age. Matched controls were randomly selected from female members who had not had strokes. Information about the use of oral contraceptives (essentially limited to low-estrogen preparations) was obtained in interviews.
Results A total of 408 confirmed strokes occurred in a total of 1.1 million women during 3.6 million woman-years of observation. The incidence of stroke was thus 11.3 per 100,000 woman-years. On the basis of data from 295 women with stroke who were interviewed and their controls, the odds ratio for ischemic stroke among current users of oral contraceptives, as compared with former users and women who had never used such drugs, was 1.18 (95 percent confidence interval, 0.54 to 2.59) after adjustment for other risk factors for stroke. The adjusted odds ratio for hemorrhagic stroke was 1.14 (95 percent confidence interval, 0.60 to 2.16). With respect to the risk of hemorrhagic stroke, there was a positive interaction between the current use of oral contraceptives and smoking (odds ratio for women with both these factors, 3.64; 95 percent confidence interval, 0.95 to 13.87).
Conclusions Stroke is rare among women of childbearing age. Low-estrogen oral-contraceptive preparations do not appear to increase the risk of stroke.
The first oral contraceptive agents marketed in the United States contained 150 µg of estrogen. Studies in the 1960s and early 1970s were based on the use of oral-contraceptive formulations that typically contained 80 or 100 µg of estrogen. Oral contraceptives now in widespread use in the United States contain 30 or 35 µg of estrogen. Previous large studies of stroke among users of oral contraceptives failed to provide information on the incidence of stroke. Moreover, the risk of stroke may be higher among users of oral contraceptives who are already at high risk of having a stroke for other reasons.23,24 In the United States, oral-contraceptive use has been largely restricted to women who are free of risk factors for cardiovascular disease.
We conducted a study of the relation of stroke to the use of oral contraceptives in a large health maintenance organization (HMO) in which the use of high-estrogen oral contraceptives (those containing >50 µg) was rare and among whose members risk factors for cardiovascular disease are likely to be detected. The population-based design allowed us to estimate the incidence of stroke.
Methods
This case–control study was conducted among the members of the Kaiser Permanente Medical Care Programs of Northern and Southern California. The study was approved by the relevant institutional review boards.
Ascertainment and Classification of Strokes
An attempt was made to identify all fatal and nonfatal strokes that occurred from May 1991 through August 1994 (for northern California) and from July 1991 through August 1994 (for southern California) in female members of the northern and southern California programs who were 15 through 44 years of age. Cases were identified through hospital admission and discharge records, emergency department logs, and records of payments for out-of-plan hospitalizations.
Stroke was defined as the new onset of rapidly developing symptoms and signs of loss of cerebral function that lasted at least 24 hours and had no apparent nonvascular cause. We excluded neurologic events due to subdural hematoma, brain tumor, infection, metabolic derangement, and multiple sclerosis.
Two physicians reviewed the records of potentially eligible patients and used defined criteria that included clinical symptoms and the results of computed tomography (CT) and magnetic resonance imaging (MRI) of the head, lumbar puncture, angiography, surgery, and autopsy to assess whether the event was a stroke and, if so, its type and subtype. A project neurologist adjudicated discrepancies between the assessments of these two physicians. All but two women with possible strokes that were not immediately fatal underwent CT or MRI (or both) of the head. Strokes were classified as venous, hemorrhagic, resulting from ischemic infarction, "other," or "unknown." Hemorrhagic strokes were subclassified as intraparenchymal, subarachnoid, or of mixed or uncertain subtype on the basis of the results of tests and procedures. Ischemic infarctions were subclassified as cardioembolic if the embolism had a cardiac source, as resulting from arterial dissection if angiography showed a dissection, or as "other ischemic infarction."
Controls
For each woman with stroke, three controls, matched for year of birth and location of the facility where care was received, were randomly selected from among female members of the program. Controls who could not be located, declined to be interviewed, or spoke neither English nor Spanish were replaced with other randomly selected controls until three controls had been enrolled for each woman with stroke or two replacement controls had been selected.
Sources of Information
Eligible women with stroke and controls were interviewed in person by trained interviewers who used a standardized questionnaire. The mean (±SD) interval between enrollment and the date of the interview was 80±56 days for women with stroke and 86±59 days for controls.
If a woman with stroke had died or was unable to communicate orally, an attempt was made to interview a family member or other proxy in her place. The first-choice respondent was the woman's husband or other live-in companion. If the woman had no husband or live-in companion or if the husband or companion could not provide accurate information, a daughter, mother, sister, or close friend of the woman was interviewed (with preference assigned in the order stated); proxy respondents were interviewed for 20.5 percent of the women with stroke.
Each study subject was assigned an "index date," which was the date of onset of symptoms for a woman with stroke and the same date for her matched controls. We used a calendar method that structures questions in relation to important life events to obtain information on contraceptive methods. A book with pictures of oral-contraceptive formulations marketed in the United States was used to help the subjects identify the drugs they had taken.
A woman was classified as having hypertension if she or the proxy respondent answered yes to a question about the use of medication for high blood pressure. A yes answer to a question about the use of insulin or pills for diabetes was used to classify women as having diabetes. The height and weight values were reported by the women or their proxies.
Statistical Analysis
Odds ratios were used to estimate relative risk. For the multivariate analyses, conditional logistic regression was used to estimate the odds ratio. In multivariate analyses, we adjusted for major risk factors for stroke: cigarette smoking, hypertension, and diabetes. We also adjusted for variables that affected the odds ratio for oral-contraceptive users as compared with nonusers if the adjustment altered the unadjusted estimate by 10 percent or more. Variables that directly or indirectly measured socioeconomic status (education, race or ethnic group, and income) were highly correlated. In the final models, we adjusted for race or ethnic group because the proportion of subjects for whom data on income were missing was high and because there is an established association between race or ethnic group and the risk of stroke.25 The incidence of stroke was estimated with the total number of girls and women 15 to 44 years of age who were members of the HMO as the denominator.
Results
A total of 408 confirmed strokes occurred in 1.1 million girls and women 15 through 44 years of age during 3.6 million woman-years of observation (Table 1). The incidence of stroke was thus 11.3 per 100,000 woman-years of observation.
|
Interviews were completed for 357 of the 395 eligible women or their proxies (90.4 percent). This analysis is based on the 295 women with ischemic infarction or hemorrhagic stroke who were interviewed; we excluded women with venous strokes (n = 6) and strokes of unknown type (n = 3), women who were pregnant at the time of the stroke (n = 11), those who had undergone a hysterectomy, bilateral oophorectomy, or both (n = 35), women for whom there were no controls (n = 2), and women for whom proxy respondents were interviewed and information on oral-contraceptive use was incomplete (n = 5). We excluded women with venous strokes in order to focus on arterial vascular disease. One of the five women with venous strokes for whom information on oral-contraceptive use was available was a current user. We excluded women who had undergone hysterectomy or oophorectomy because none were current users of oral contraceptives and they therefore contributed no information about the relation of stroke to the current use of such agents. Furthermore, because hysterectomy and oophorectomy are related to the risk of disease, the statistical model would have been even more complex had these women been included. Controls who had undergone hysterectomy or bilateral oophorectomy were also excluded.
We compared information on the use of oral contraceptives from the medical records of eligible women who were interviewed (or whose proxies were interviewed) and those who were not in order to assess whether response bias was present. In the two years before the index date, 12.5 percent of the nonrespondents and 12.9 percent of the respondents had received a prescription for oral contraceptives. We could not determine with certainty whether oral contraceptives had been used within one month of the index date for about 50 percent of the subjects. Medical records were not useful for determining the extent of oral-contraceptive use, if any, in the distant past. The medical records did not contain reliable and complete information on smoking, race or ethnic group, or income. Because there was no evidence of response bias with respect to oral-contraceptive use within two years of the index date, because medical records were not useful for determining the extent of past use of oral contraceptives, if any, and because information on important confounding variables was not available for nonrespondents, we based our main analysis on the information gathered in the interviews.
We compared the information obtained from the women with stroke with that from proxy respondents and found no statistically significant differences in the distribution of any variables except past use of oral contraceptives and no lifetime use of such agents. The proportion of proxy respondents who reported that the women with stroke had never used oral contraceptives was higher and the proportion reporting past use was lower than the corresponding proportions of the women with stroke who were interviewed directly (data not shown). A review of the medical records showed that 13.2 percent of the women with stroke whom we interviewed and 13.1 percent of those interviewed by proxy had received a prescription for oral contraceptives in the two years before the index date. On the basis of this analysis, data from proxy respondents were excluded from the main analyses of past oral-contraceptive use and any oral-contraceptive use but were included in the analyses of current as compared with former use of oral contraceptives or no use of such agents.
Table 2 shows the characteristics of the women with ischemic infarction and hemorrhagic stroke, along with the distribution of oral-contraceptive use among all the controls. Among controls, factors associated with current use or nonuse of oral contraceptives were age, cigarette smoking, body-mass index (the weight in kilograms divided by the square of the height in meters), income, race or ethnic group, alcohol use, and marital status. Among the controls, only a small percentage of current users of oral contraceptives had hypertension, and none had diabetes.
|
|
|
After the exclusion of women for whom proxy respondents were interviewed, the adjusted odds ratio among current users of oral contraceptives as compared with former users and those who had never used oral contraceptives was 1.07 (95 percent confidence interval, 0.46 to 2.49) for ischemic stroke and 1.13 (95 percent confidence interval, 0.50 to 2.51) for hemorrhagic stroke. When we conducted an analysis based on medical records, the odds ratios for oral-contraceptive use within the past two years were 1.45 (95 percent confidence interval, 0.65 to 3.27) for ischemic stroke and 1.07 (0.58 to 1.97) for hemorrhagic stroke.
Table 5 shows the odds ratios among current users as compared with former users and those who had never used oral contraceptives according to smoking status, age, and progestogen type and for women who did not have hypertension. The small number of current users with hypertension precluded the examination of the odds ratio in women with this condition. For hemorrhagic stroke, but not ischemic stroke, there was statistical evidence of a positive interaction between current oral-contraceptive use and smoking (P = 0.04). The odds ratios for ischemic or hemorrhagic stroke associated with current oral-contraceptive use were not substantially higher for older women than for younger women. There was no statistical evidence of an interaction between current oral-contraceptive use and age (P>0.05 for both types of stroke). The adjusted odds ratio for all types of stroke among current oral-contraceptive users was 1.16 (95 percent confidence interval, 0.72 to 1.88; data not shown).
|
Several studies have found a relation between the use of oral contraceptive agents with relatively high estrogen content and a higher risk of ischemic stroke and venous thromboembolism.23,26,27,28 Women already at high risk for vascular disease may be at even greater risk for vascular disease when they use oral contraceptives.9,10,11,12,13,23,24 In the population we studied, 96 percent of current users of oral contraceptives used formulations containing less than 50 µg of estrogen. There were no users of formulations containing more than 50 µg of estrogen. Fewer than 5 percent of controls with treated hypertension were current oral-contraceptive users, no controls who had diabetes were current users, and fewer than 10 percent of current users were 40 years of age or older.
The virtually exclusive use of oral contraceptives containing less than 50 µg of estrogen and the selective use of oral contraceptives in young women without hypertension or diabetes is a plausible explanation for the lack of elevation in the risk of stroke in our study population. However, there are some limitations to our study. The analysis relied on information reported by the respondents. The adjusted odds ratios for oral-contraceptive use within two years, based on data from the medical records, were 1.45 (95 percent confidence interval, 0.65 to 3.27) for ischemic stroke and 1.07 (95 percent confidence interval, 0.58 to 1.97) for hemorrhagic stroke. Thus, we cannot entirely rule out recall or response bias as an explanation for our results. Furthermore, the confidence intervals for our estimates of the risk of stroke among current users of oral contraceptives were wide, and in this study we cannot reliably differentiate between a true null effect and a true small or moderate increase in risk.
A Danish case–control study of "thromboembolic attack" found an odds ratio of 1.8 (95 percent confidence interval, 1.1 to 2.9) among current users of low-estrogen oral contraceptives (those containing less than 50 µg),21 as compared with the combined group of former users and those who had never used oral contraceptives. This figure is compatible with our estimates. Taken together, our study and the Danish study suggest that the true relative risk of ischemic stroke among users of low-estrogen oral contraceptives, as compared with nonusers, is not more than 2.5.
Three other recent case–control studies determined odds ratios for subarachnoid hemorrhage among current oral-contraceptive users as compared with former users and those who had never used oral contraceptives.20,22,23 In a study of fatal subarachnoid hemorrhage by Thorogood et al.,20 the odds ratio in current oral-contraceptive users was 1.1 (95 percent confidence interval, 0.7 to 1.9). Longstreth et al.22 reported an odds ratio for subarachnoid hemorrhage of 0.89 (95 percent confidence interval, 0.22 to 3.61). In an analysis from the Royal College of General Practitioners' Oral Contraception Study, which encompassed cases of stroke occurring from 1968 through 1990,23 the relative risk of subarachnoid hemorrhage among current oral-contraceptive users was 1.5 (95 percent confidence interval, 0.6 to 3.7); for intraparenchymal hemorrhage, the relative risk was 1.1 (95 percent confidence interval, 0.2 to 7.1). Our findings support the conclusion that current oral-contraceptive use does not increase the risk of hemorrhagic stroke overall.
In the Collaborative Study of Stroke in Young Women,24 the odds ratio for hemorrhagic stroke among current oral-contraceptive users, as compared with the combined group of former users and those who had never used oral contraceptives, was 1.8 for women with normal blood pressure, 2.8 for women with borderline hypertension, 8.4 for those with moderate hypertension, and 25.7 for those with severe hypertension. The Royal College of General Practitioners' study also reported an interaction between hypertension and current oral-contraceptive use with respect to hemorrhagic stroke.23 Our study included so few current users of oral contraceptives who had hypertension that the odds ratio for hemorrhagic stroke among current oral-contraceptive users in the subgroup of women with hypertension could not be estimated. Given all the evidence, hypertension should be considered a contraindication to the use of oral contraceptives.
Studies of myocardial infarction have reported interactions of current oral-contraceptive use with current smoking and age over 35 years.9,10,11 We found a positive interaction of smoking with current oral-contraceptive use with respect to hemorrhagic stroke. Our data on a possible interaction of oral contraceptive use with age are difficult to interpret. Our results neither establish nor rule out an interaction of age and current oral-contraceptive use with respect to the risk of stroke. The 95 percent confidence intervals for our odds ratios were wide for all subgroup analyses.
Studies in Great Britain in the late 1970s and early 1980s suggested that the risk of arterial vascular disease may be related to the type of progestogen contained in oral contraceptives.29,30 The World Health Organization Collaborative Study reported an increased risk of venous thromboembolism among current users of oral contraceptive formulations containing desogestrel or gestodene.31 In our study, there were no known users of formulations containing these progestogens. The small number of current oral-contraceptive users limits the power of our study to detect differences in risk between formulations containing norethindrone-type progestogens and those containing norgestrel.
In our study, during 3.6 million woman-years of observation, there were 195 ischemic and 201 hemorrhagic strokes in girls and women 15 through 44 years of age. The incidence of ischemic infarction was 5.4 per 100,000 woman-years of observation, and the incidence of hemorrhagic stroke was 5.6 per 100,000 woman-years. Our study documents the low incidence of stroke among young women and provides a basis for interpreting the small increases in the risk of stroke that our results do not rule out.
A drug is deemed safe when the magnitude of its benefits appears to outweigh that of its risks.32 The use of oral contraceptives has a number of established beneficial effects, including most prominently the prevention of pregnancy. This study establishes the low incidence of stroke among women of childbearing age. Even if the small increase we observed was due directly to the use of oral contraceptives, the number of excess cases of stroke in healthy women would be small. We found no association between the past use of oral contraceptives and an increased risk of stroke. We conclude that, as used by the women in this study, currently available low-estrogen oral contraceptives are generally safe with respect to the risk of stroke.
Supported by a contract (N01-HD-3108) with the National Institute of Child Health and Human Development.
We are indebted to Teresa Picchi, Luisa Hamilton, Mona Ackerman, and Nancy Lusk, who oversaw field operations, and to Kimberly Tolan, who did the computer programming for this analysis.
Source Information
From the Kaiser Permanente Medical Care Program, Southern California, Pasadena (D.B.P., S.W., H.K.Z.), and Northern California, Oakland (S.S., A.B., C.Q.).
Address reprint requests to Dr. Petitti at Research and Evaluation, SCPMG, 393 E. Walnut St., Pasadena, CA 91188.
References
| |||||||||||||||||||||||||||||||||||||||||
Related Letters:
Stroke in Users of Low-Dose Oral Contraceptives
Carolei A., Marini C., Petitti D. B., Sidney S., Tolan K.
Extract |
Full Text
N Engl J Med 1996;
335:1767-1768, Dec 5, 1996.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
