Ischemic Stroke and Use of Estrogen and Estrogen/Progestogen as Hormone Replacement Therapy
Background and Purpose—Information about the risk of stroke in current postmenopausal hormone users is limited.
Methods—In this case-control study, women aged 45 to 74 years hospitalized with a fatal or nonfatal stroke in any of 10 Northern California Kaiser Permanente facilities during the period November 1991 to November 1994 were identified as cases. Controls were selected at random from female Health Plan members. Data regarding use of estrogen plus progestogen or estrogen alone were obtained in interviews.
Results—The analysis was based on nonproxy responses from 349 cases of ischemic stroke and 349 matched control subjects. After adjustment for confounders, the odds ratio for ischemic stroke in current hormone users was 1.03 (95% confidence interval, 0.65 to 1.65). The odds ratios for ischemic stroke in current hormone users showed no clear trend of increasing or decreasing risk in relation to duration of hormone use. The odds ratio for ischemic stroke in past hormone users was 0.84 (95% confidence interval, 0.54 to 1.32).
Conclusions—In this study postmenopausal hormone use was not associated with an increase or decrease in the risk of ischemic stroke, a finding that is consistent with the body of literature on this topic.
A number of epidemiological studies have examined the relationship between stroke and use of postmenopausal hormones.1 2 3 4 5 6 7 8 9 10 11 We identified only two studies1 2 presenting data on the postmenopausal use of combined estrogen and progestogen therapy (HRT). We identified only three studies1 3 4 presenting information on stroke risk specifically in relation to current use of estrogen alone (ERT). Studies of stroke and hormone use are difficult to interpret in aggregate because of differences in the types of strokes included and variation in the extent of statistical control for socioeconomic and demographic variables.
The Women’s Health Initiative is a large randomized trial that will examine the relationship of ERT and HRT with a variety of health outcomes. The results of this study will not be available for several years.
Our study was designed to estimate the relative risk of stroke in relation to current use of ERT and HRT in a population with a relatively high prevalence of hormone use. The risk of stroke in relation to duration of use of postmenopausal hormones is also presented because this is a topic that has not been examined closely in prior studies.
Subjects and Methods
This was an epidemiological case-control study. The study was reviewed and approved by the Institutional Review Board of the Kaiser Permanente Medical Care Program of Northern California.
Case Ascertainment and Classification
An attempt was made to identify all fatal and nonfatal strokes in women aged 45 to 74 years in 10 medical centers of the Kaiser Permanente Medical Care Program, Northern California region during the period November 1991 to November 1994. Start and finish dates differed by facility. Thus, the number of months of case ascertainment differed at each facility; it averaged 33.4 months. Sources of case identification included hospital admission and discharge records, emergency department logs, and payments for out-of-plan hospitalizations.
Stroke was defined as the new onset of rapidly developing symptoms and signs of loss of cerebral function with no apparent cause other than that of vascular origin. Specifically excluded were neurological events due to subdural hematoma, brain tumor, infection, metabolic derangement, and multiple sclerosis. Case subjects with a prior history of cerebrovascular disease based on medical record review at the time of their event were also excluded.
To establish the stroke diagnosis, records of all potential stroke events were reviewed by two physicians (D.B.P. and S.S.) with adjudication of discrepancies in their assessment by a project neurologist (A.B.). The review used a defined protocol. Strokes were subclassified as being hemorrhagic or due to ischemic infarction, also according to a defined protocol that included review of records by two physicians (D.B.P. and S.S) with adjudication of discrepancies by a project neurologist. A stroke was considered to be hemorrhagic if blood was found on CT or MRI scan, if a nontraumatic lumbar puncture revealed blood and/or xanthochromia, or if intracerebral blood was found at autopsy, on cerebral angiography, or at surgery. Remaining strokes were considered ischemic infarctions.
For each case subject, one control subject, matched on birth year and facility of care, was randomly selected from female members of the Health Plan. Control subjects who could not be located, declined interview, or spoke other than English or Spanish were replaced with control subjects selected at random until one control subject was obtained for each case subject or two “replacement” control subjects had been selected.
Eligible case and control subjects were interviewed in person by trained interviewers using a standardized instrument. If a case subject had died or was unable to communicate verbally, an attempt was made to interview a proxy.
Interview questions were asked relative to an index date, which was the date of symptom onset for case subjects and the same date for the matched control subjects. The calendar method, in which hormone replacement use information is gathered by structuring questions in relation to significant life events, was used to obtain information on all hormone replacement preparations ever used. A visual aid, consisting of actual pills for commonly used preparations and a color chart with a picture of all hormone replacement preparations ever marketed in the United States, was used to facilitate recall of the formulation of hormone replacement preparations used currently and in the past.
Hypertension was defined as a “yes” to a question about use of medication for high blood pressure. Diabetes was defined as a “yes” to a question asking about use of insulin or pills for diabetes. A study subject was defined as a nonsmoker if she answered “no” to the question, “Have you ever smoked cigarettes?” If she answered “yes” to this question, she was categorized as a current regular smoker on the basis of her answer to the question, “On (index date) were you still smoking regularly?” (“regularly” was defined as at least five cigarettes per week, almost every week). Body mass index (kilograms per meter squared) was determined with self-reported height and weight. Exercise was assessed by self-report. A study subject was defined as having a prior self-reported history of stroke or TIA if she stated that a physician had told her she had had a stroke or TIA.
The analysis was restricted to postmenopausal women. A study subject was categorized as postmenopausal if she had undergone bilateral oophorectomy or she was nonhysterectomized and reported that she had ceased having menstrual periods. Hysterectomized women who have undergone a unilateral oophorectomy would not know whether they were premenopausal or postmenopausal based on menstrual function. For this reason, we also excluded hysterectomized women less than 55 years of age because they could not be classified as postmenopausal with certainty. These exclusions were applied to both case and control subjects.
The vast majority of hysterectomized women (98%) used ERT, and most nonhysterectomized women (76%) used HRT. For this reason, estimates of the ORs for stroke in users of ERT who have not had a hysterectomy and the ORs for stroke in users of HRT who have had a hysterectomy are highly unstable statistically. Thus, we excluded from the main analysis hysterectomized women who used HRT and nonhysterectomized women who used ERT. Women who used only progestogen were also excluded from both case and control subjects because the number of women in this group was very small, and valid estimates of risk in relation to exposure to progestogen only could not be derived. These exclusions were applied to both case and control subjects.
Of the women who used ERT, 90% used conjugated equine estrogen; of these, 81% used 0.625 mg/d. Of women who used HRT, 95% used medroxyprogesterone acetate. The regimens of HRT use were varied, and no single regimen had a large number of women. For these reasons, we did not attempt to assess the ORs for stroke in relation to specific doses of estrogen, types of progestogen, or regimens of HRT.
We compared information on hormone use in the 2 years before interview documented in medical records with information provided by proxies to assess response bias and found that use of hormones was substantially underreported by proxy respondents. For this reason, proxy responses were also excluded from the analysis. We did not use information from medical records for the analysis because neither hormone use on a particular date nor lifetime past use could be ascertained from available medical records.
The exposure OR was used to estimate relative risk. Conditional logistic regression was used in the main analysis. For the main analysis, we defined an exposure variable—current hormone use—as “yes” for hysterectomized woman who used ERT and nonhysterectomized woman who used HRT. Because of the way hormone exposure is defined, the OR for hormone use in the main analysis pertains to current use of ERT for hysterectomized women and to current use of HRT for nonhysterectomized women. Duration (years) of hormone replacement therapy in current hormone users was examined by replacing the hormone replacement treatment variable (ERT for hysterectomy, HRT for no hysterectomy) with years of use of either. In the multivariate analyses, we adjusted for major established risk factors for ischemic stroke—cigarette smoking, hypertension, and diabetes—and for prior history of stroke or TIA. In addition, we adjusted for other variables that were apparent confounders in these data, including the variable in the model if there was an appreciable change in the exposure coefficients with the addition of the potential confounder.
We also performed an analysis that examined the ORs for stroke in relation to current use of ERT and HRT separately. This analysis was not matched and adjusted for hysterectomy as well as other confounders.
There were 885 possible stroke cases identified by our surveillance effort during the period of case ascertainment; 758 cases were confirmed stroke cases. Of these 758 confirmed cases, 550 were cases of ischemic stroke, 201 were cases of hemorrhagic stroke, and 7 could not be classified as ischemic or hemorrhagic. Of the 550 cases of ischemic stroke, exclusions because of patient or physician refusal, proxy interviews, and other reasons (Table 1⇓) left 349 case/control sets for the analysis. The same exclusions for cases of hemorrhagic stroke left only 83 case/control sets for analysis. Because of the small number of cases of hemorrhagic stroke that remained after exclusions, we have chosen not to present detailed data on hemorrhagic stroke.
Table 2⇓ shows the characteristics of ischemic stroke case and control subjects along with the age-adjusted ORs for each characteristic. Age-adjusted ORs were elevated in relation to smoking, hypertension, diabetes, high body mass index, low education, low income, and African-American ethnicity. Women who drank alcohol were at a lower risk of ischemic stroke. The risk of ischemic stroke in women with high cholesterol was only slightly elevated.
When we took into account only the matching variables, the OR for ischemic stroke in current ERT users compared with never hormone users was 0.81 (95% CI, 0.52 to 1.26); the OR in current HRT users compared with never hormone users was 0.62 (95% CI, 0.35 to 1.11). The OR for past hormone use compared with never hormone use was 0.77 (95% CI, 0.54 to 1.11).
Table 3⇓ shows results adjusted for established risk factors for ischemic stroke (age, hypertension, diabetes, smoking, body mass index, prior history of stroke or TIA) and for socioeconomic variables (education, ethnicity). Alcohol use did not confound the relationship between hormone use, and stroke and was not included in the adjustment model. After adjustment for confounders, the OR for ischemic stroke in current hormone users compared with never hormone users was 1.03 (95% CI, 0.65 to 1.65). The adjusted ORs for ischemic stroke in current hormone users showed no clear trend of increasing or decreasing risk in relation to duration of hormone use. The adjusted OR for ischemic stroke in past hormone users was 0.84 (95% CI, 0.54 to 1.32).
We also did an analysis in which women with hysterectomy who used HRT and women without hysterectomy who used ERT were not excluded. There were 374 case subjects and 360 control subjects. After the adjustment for hysterectomy as well as the variables shown in Table 3⇑, the OR in current users of HRT was 0.60 (95% CI, 0.31 to 1.16). For current ERT use, the adjusted OR was 1.04 (95% CI, 0.60 to 1.10).
After adjustment for smoking, hypertension, diabetes, body mass index, ethnicity, and education, the OR for hemorrhagic stroke in current hormones users compared with never hormone users was 0.33 (95% CI, 0.12 to 0.96), based on 83 case/control sets.
Our data show no significant increase or decrease in the risk of ischemic stroke in current users of ERT or HRT. The upper and lower bounds of the CI provide an assessment of the power of the study. The study had a 95% chance of detecting a true OR of ischemic stroke in current hormone users as low as 0.65 and as high as 1.65, and it does not rule out ORs for ischemic stroke between these two values. Firm conclusions about hemorrhagic stroke in current users of ERT and HRT cannot be drawn from our study because the large number of excluded cases raises concerns about the representativeness of the subjects available for analysis.
Psaty et al12 summarized the published literature on hormone use and stroke in 1993 and concluded, based on available data then, that there was little if any association with stroke. This conclusion was based on the lack of a consistent relationship between stroke and hormone use. Grady et al,13 in their comprehensive literature review on hormone replacement therapy, derived a summary estimate of the relative risk of stroke in ERT users of 0.96 (95% CI, 0.82 to 1.13) but pointed out there was statistical evidence of heterogeneity for studies of stroke and ERT. There were few data on stroke in relation to use of HRT at the time these two reviews were done.
Recent data on stroke in relationship to current use of both ERT and HRT come from the Nurse’s Health Study.1 This analysis was based on 285 cases of ischemic stroke and 155 cases of subarachnoid hemorrhage. These two types of strokes were combined in the analysis because the relationship between hormone use and risk was the same for both stroke types. After adjustment for age, age at menopause, body mass index, diabetes, hypertension, high cholesterol, smoking, past use of oral contraceptives, parental history of myocardial infarction before age 60, and type of menopause, the relative risk of all stroke in current users of ERT compared with never hormone users was 1.27 (95% CI, 0.95 to 1.69); it was 1.09 (95% CI, 0.66 to 1.80) in current users of HRT compared with never hormone users.
We were able to identify only two other published studies that have examined the relationship specifically between current postmenopausal hormone use and stroke.3 4 Both studies addressed only ERT. In a report based on 11 cases of subarachnoid hemorrhage and 23 cases of other stroke, Petitti et al3 reported that the relative risk of subarachnoid hemorrhage in current users of ERT was 1.6 (95% CI, 0.7 to 3.8), and the relative risk of other stroke (virtually all cases were ischemic stroke) was 0.9 (95% CI, 0.4 to 1.8) after adjustment for confounders. Rosenberg et al4 reported an estimated relative risk of ischemic stroke in current users of ERT of 1.16 (95% CI, 0.75 to 1.77) based on a case-control study of 198 nonfatal ischemic strokes. Other published studies of stroke in users of hormone replacement therapy have examined ever use of hormones. Ever use of hormones is difficult to interpret as an exposure because women ever exposed to hormones are a heterogeneous group of current and past users, with varying durations of use and varying periods of time since last use among past users. Consideration of our results along with the results of other studies of stroke in relationship to current use of ERT or HRT1 4 5 leads to a conclusion that current postmenopausal hormone use does not increase or decrease the risk of ischemic stroke.
It would be useful to have more information on hemorrhagic stroke before drawing a similar conclusion about the absence of an increased or decreased risk in current users of ERT and HRT. Although ischemic and hemorrhagic stroke share in common an association with hypertension and smoking, the associations of diabetes, body mass index, and alcohol are different for hemorrhagic and ischemic stroke. It would be a mistake to generalize the results of studies of hormone use and ischemic stroke to hemorrhagic stroke. Future studies of hemorrhagic stroke need to take into account the high proportion of women who cannot provide information about their hormone use and the lack of validity of proxy responses about hormone use.
Grodstein et al1 also examined stroke risk in relation to past hormone use in a way that is comparable to our study. Their findings are consistent with ours and suggest that any effect of hormone use on stroke risk does not persist after cessation of use.
Our study has many limitations. Major concerns are recall bias and nonresponse bias. To assess these biases as explanations for our results, we examined information about hormone use in the 2 years before index date as recorded in medical records of a systematic sample of all women (case and control subjects) eligible for the study. To assess recall bias as a possible explanation for our results, we compared self-reports about current use of hormones from respondents with medical record information for the same subjects. The percentages were similar—30% and 32%. Thus, we do not believe that recall bias explains the findings of our study. To assess nonresponse in control subjects as a possible explanation for our results, we compared medical record information on hormone use within 2 years of the index date for respondent and nonrespondent control subjects. These percentages were 32% and 34%, respectively. We conclude that response bias in control subjects does not explain our findings.
We also used information from the medical record review to assess the possibility that exclusion of proxy responses among case subjects might explain our results. The proportion of hormone users within 2 years of the index date, after differences in the age of ischemic stroke cases in the analysis and ischemic stroke cases whose medical records were reviewed are taken into account, was 27% for direct respondents and 22% for proxy respondents. Since 14% of ischemic stroke cases were excluded as a result of proxy response, we conclude that excluding proxies did not bias our findings for this outcome. Concern about bias due to exclusion of proxy data for cases of hemorrhagic stroke is the reason we did not present detailed analysis for this outcome.
Stroke is an important cause of morbidity and mortality in women. Grady et al13 estimated that a 50-year-old woman has a 20% lifetime probability of developing a stroke and an 8% lifetime probability of dying from stroke. If hormone use caused either an increase or a decrease in the risk of stroke, it would have important implications for the assessment of the overall effect of hormone use on life expectancy and quality of life. Our study joins a growing body of literature that suggests that hormone use does not affect the risk of ischemic stroke.
Selected Abbreviations and Acronyms
|ERT||=||estrogen replacement therapy|
|HRT||=||combined estrogen and progestogen replacement therapy|
|TIA||=||transient ischemic attack|
This research was supported by a grant (R01-HL-47043) from the National Heart, Lung, and Blood Institute. Teresa Picchi and Luisa Hamilton oversaw field operations. Kimberly Tolan did the computer programming for this analysis.
Reprint requests to Dr Diana B. Petitti, Research and Evaluation, SCPMG, 393 E Walnut St, Pasadena, CA 91188.
- Received July 31, 1997.
- Revision received September 22, 1997.
- Accepted October 8, 1997.
- Copyright © 1998 by American Heart Association
Bush TL, Barrett-Connor E, Cowan LD, Criqui MH, Wallace RB, Suchindran CM, Tyroler HA, Rifkind BM. Cardiovascular mortality and noncontraceptive use of estrogen in women: results from the Lipid Research Clinics Program Follow-up Study. Circulation. 1987;75:1102–1109.
Pfeffer RI, van Den Nort S. Estrogen use and stroke risk in postmenopausal women. Am J Epidemiol. 1976;103:445–456.
Boysen G, Nyboe J, Appleyard M, Sorensen PS, Boas J, Sommier F, Jensen G, Schnor P. Stroke incidence and risk factors for stroke in Copenhagen, Denmark. Stroke. 1988;19:1145–1153.
Grady D, Rubin SM, Petitti DB, Fox CS, Black D, Ettinger B, Ernster VL, Cummings SR. Hormone therapy to prevent disease and prolong life in postmenopausal women. Ann Intern Med. 1992;117:1016–1037.