Hormone Replacement Therapy and Incidence of Acute Myocardial Infarction
A Population-Based Nested Case-Control Study
Background—Epidemiological studies suggest a decreased risk of coronary heart disease (CHD) in healthy women taking hormonal replacement therapy (HRT). Whether this effect is shared by oral and transdermal preparations is unknown.
Methods and Results—We conducted a population-based case-control study nested in a cohort of women 50 to 74 years of age without cardiovascular disease history in the United Kingdom. Among 164 769 women from the General Practice Research Database (January 1, 1991, to December 31, 1995), we identified 1242 first acute myocardial infarctions (AMI) and confirmed 1013 after medical record review. We randomly selected 5000 age-frequency–matched control subjects. AMI incidence was 1.6 per 1000 person-years; 13% and 17% of cases and control subjects used HRT within 6 months before the index date. Risk factor and comorbidity–adjusted OR of AMI for current-recent HRT users compared with nonusers was 0.72 (95% CI 0.59 to 0.89). The OR was similar within 30 days before the index date. The beneficial effect was present after 1 year of use (OR 0. 68; 95% CI 0.53 to 0.86), with no increase in risk within the first year. ORs for unopposed and opposed therapy were 0.52 (95% CI 0.35 to 0.78) and 0.79 (95% CI 0.59 to 1.08); 79% and 21% used oral and transdermal therapy. The protective effect was present at medium-high doses of estrogens with ORs for oral and transdermal therapy of 0.63 (95% CI 0.46 to 0.86) and 0.62 (95% CI 0.37 to 1.06) and ceased after 2 to 3 years since stopping HRT.
Conclusions—Results are consistent with those previously reported in women without CHD who were taking oral HRT and, although based on few users, suggest that transdermal therapy might have similar cardioprotective effects.
Aconsistent male-to-female ratio for coronary heart disease (CHD) death rates ranging from 2 to 5 in populations with very different heart disease rates and lifestyles has suggested that either sex or endogenous hormones have a significant influence on the vasculature.1 2 The apparent loss of this female advantage on CHD risk in women who had a premature menopause led early on to the hypothesis that endogenous estrogens are cardioprotective. There is now considerable evidence that estrogens exert indirect and direct antiatherogenic effects.3
Evidence from multiple observational studies suggested a marked reduction in the risk of CHD associated with postmenopausal estrogen use in primary prevention. A similar effect was observed when estrogens were opposed with progestins. Although several potential sources of bias might account for part of this beneficial effect, the latest meta-analysis of published studies suggested 30% and 34% reduction in the risk among ever-users of unopposed and opposed therapy, respectively, compared with never-users.4 However, all these studies were conducted in populations taking oral preparations, many of them from the US population. No studies have provided data on the effects of the transdermal route on the risk of CHD.
As part of a study on the effects of hormone replacement therapy (HRT) in postmenopausal women, we conducted a population-based cohort study with a nested case-control analysis to evaluate the association between HRT use and the incidence of first myocardial infarction in a European population taking oral and transdermal estrogens.5 Estimates of this association by duration, dose, regimen, and route were also calculated.
A cohort of women, 50 to 74 years of age, from the general population in the United Kingdom, was identified through registration status in the General Practice Research Database between January 1, 1991, and December 31, 1995. Women with a history of cardiovascular and cerebrovascular diseases, neoplasms, coagulopathies, vasculitis, and alcohol-related diseases were excluded. The resulting study cohort consisted of 164 769 women who were followed until the occurrence of the first recorded code of myocardial infarction, any of the exclusion criteria as listed above, reaching 75 years of age, death, or to December 31, 1995, whichever occurred first.
More than 2000 general practitioners (GPs) use the General Practice Research Database to register healthcare and medical information about their patients in a standardized manner.6 This research-oriented database covers a general population exceeding 3 million residents in England and Wales. The UK National Health System provides universal coverage, and no segment of the population is excluded from the database. The registered information includes demographic data, all medical diagnoses, consultant and hospital referrals, and a record of all prescriptions issued. Practitioners generate prescriptions directly from the computer, ensuring its automatic recording. Medication codes are based on the Prescription Pricing Authority. A modification of the OXMIS code is used to record diagnoses.7 Several validation studies have shown that >90% of information present in the manual medical records of the GPs, and >95% of newly prescribed drugs are recorded in the database.8 9 The GP may also record laboratory results and other medical data in a free text comment field.
Definition and Ascertainment of Cases
Potential cases were identified by an automated computer search with the use of specific and nonspecific OXMIS codes for myocardial infarction. We identified 1242 potential cases of first myocardial infarction. For each of them, a questionnaire was sent to their corresponding GPs requesting all available medical information related to the acute episode, including hospital discharge letters, ECGs, myocardial serum enzyme levels (CPKs), autopsy reports, and death certificates.
To validate cases, we adapted the international standardized diagnostic criteria for acute myocardial infarction.10 11 A woman was classified as a definite case of nonfatal acute myocardial infarction if she had ≥2 of the following criteria: characteristic chest pain, raised levels of CPKs, an ECG consistent with the diagnosis of myocardial infarction, an arteriogram documenting a recent coronary occlusion, or treatment with fibrinolytic therapy. A woman was classified as possible case of nonfatal acute myocardial infarction when this information was not available but a consultant cardiologist confirmed the diagnosis. The same criteria were applied to define fatal cases of myocardial infarction. For patients who died before reaching the hospital or for whom hospital records were not available, we used the diagnosis recorded on autopsy reports and death certificates, as available, to determine the cause of death. Fatal cases, whether sudden or not, were classified as (a) definite fatal myocardial infarction in which there was a definite myocardial infarction within 30 days of death or/and a fresh myocardial infarction and/or recent coronary occlusion described at necropsy report or (b) possible acute CHD death without a preceding nonfatal coronary event in which there was no good evidence for another cause of death clinically or at necropsy and a death certificate with acute CHD as primary code of death cause.
After the review of all available information by 2 of the investigators independently and blinded to therapy, a total of 1013 women were confirmed with a first myocardial infarction. Overall, 229 cases were excluded because the diagnosis of myocardial infarction was not established or an exclusion criterion was identified. For each event, the index date was defined as the earliest of the date of symptoms onset, when available, the date of hospital admission or the date of death.
Selection of Control Subjects
We used a time-incidence sampling method to select control subjects.12 A random date within the study period was generated for each of the women of the cohort. All women with a random date included in their person-time period of observation (from study entry to end of follow-up) were eligible as control subjects. This procedure takes into account the amount of person-time contributed by each woman. Thus, we ensure that our control selection will yield a comparison group whose relevant exposure characteristics accurately reflect those of the source population for cases. In addition, we introduced calendar time in the analysis to control for secular changes in both the incidence of cases and the prevalence of exposure. The same computer-based exclusion criteria as those for the cases were applied to all eligible control subjects, with each woman’s random date used as her index date. Finally, 5000 control subjects were randomly sampled and matched by the age-frequency of cases.
Exposure and Other Risk Factor Assessment
A list of all medications containing estrogens and/or progestogens recommended for HRT and available in the United Kingdom during the study period was extracted from the British National Formulary. These drugs were grouped in the following regimens: (a) oral estrogens (76% were conjugated equine estrogens; (b) transdermal estradiol; (c) estradiol implant; and (d) tibolone. In addition, oral estrogens and transdermal estradiol were classified as opposed or unopposed, depending on whether a progestogen was supplied along with estrogens. For each woman, the date of the last issued prescription for each of the regimens before the index date was identified. We defined a nonuser as a woman who never had a prescription for HRT recorded in the database, a current-recent user as a woman who used HRT at any time in the 6-month period before the index date, and a past user as a woman who stopped therapy >6 months before the index date. The time period corresponding to consecutive prescriptions was used to define duration of HRT use. The daily dose of estrogens, type of regimen, and route of administration was obtained from the last prescription. Duration was categorized as (a) ≤1 year of use; (b) between 1 year and 3 years; and (c) >3 years. Estrogen prescribed dose was categorized as (a) low dose: <0.625 mg of oral estrogens or 25 μg of transdermal estradiol; (b) medium dose: 0.625 mg of oral estrogens or 50 μg of transdermal estradiol; and (c) high dose: ≥1.25 mg of oral estrogens or 100 μg of transdermal estradiol.
Information on coronary risk factors and drug utilization was obtained from the database. History of hypertension, diabetes, hypercholesterolemia, smoking, and family history of CHD were considered present when this information was registered in the database. Overweight was defined as recorded history of obesity and/or individual body mass index (calculated from recorded height and weight) over 27 kg/cm2. Menopause was considered surgical when a bilateral oophorectomy was recorded with or without hysterectomy. Comorbidity was assessed for all cohort members and was present if there was ≥1 recorded code at any time before the index data of a disease in the following systems: nervous, cardiovascular, respiratory, renal, or osteoarticular or diabetes.
The distribution of risk factors according to HRT use was directly standardized by age with the use of the overall age distribution of control subjects. Person-time of follow-up was calculated for all cohort members by age and exposure categories and was used as the denominator to obtain incidence rates of myocardial infarction. Age-adjusted incidence rates were obtained by the direct method, with the overall age distribution of the cohort used as standard.13 To evaluate the potential clinical impact of using HRT, we calculated rate differences (the excess number of cases on average attributable to the nonuse of hormones per 1000 person-years).14
ORs were used as the measure of association. Adjusted ORs and 95% CIs were estimated with the use of unconditional multiple logistic regression by the LOGISTIC procedure of SAS release 6.08.15 Adjusted ORs were obtained by recent HRT use compared with nonuse. The effect of duration was assessed among current-recent users. The effect of oral and transdermal therapy and unopposed and opposed therapy was assessed in current-recent users with duration >1 year compared with nonusers. The effect of estrogen dose was also assessed in long-term users of oral estrogens and transdermal estradiol, unopposed or opposed, aggregated in a single group of users.
Overall prevalence of HRT was 22%. Use during the last 6 months was found in 133 (13%) cases and 855 (17%) control subjects .Women who had ever used HRT tended to be younger and have a better risk profile than those women who never used the therapy, except for smoking history (Table 1⇓). Use of HRT was inversely related to age. In the age group 50 to 59 years, 38% of control subjects used the therapy in the last 6 months. Among women ≥60 years of age, this pattern of use was found in 15% of control subjects.
Among cases, 791 were nonfatal myocardial cases (525 were definite and 266 possible) and 222 were fatal cases (103 were definite fatal myocardial infarctions and 119 possible acute CHD deaths). In our cohort, the absolute risk of all these events per 1000 person-years of follow-up was 1.6. The age-adjusted incidence rates per 1000 person-years were 1.8 in women who never used hormones and 1.2 in those who used hormones in the last 6 months, resulting in an absolute risk difference of 0.6 per 1000 person-years.
Overall incidence of fatal and nonfatal myocardial infarction and acute CHD deaths was positively associated with history of smoking, diabetes, hypertension, hypercholesterolemia, and family history of CHD (Table 2⇓). The age-adjusted OR and 95% CIs for myocardial infarction in women who used HRT during the last 6 months, as compared with nonusers, was 0.72 (95% CI 0.59 to 0.88) (Table 2⇓). After adjusting for potential confounding factors, that is, those mentioned above plus obesity, comorbidity, prophylactic use of aspirin (defined as current use of aspirin with duration >30 days), and surgical menopause, the OR remained unchanged. We performed a sensitivity analysis and estimated the OR by using different time-window definitions within the last 6 months before the index date. Narrowing this time period of use to 30 days before the index date yielded an OR of 0.73 (95% CI 0.59 to 0.91).
The protective effect was observed in users of >1 year of therapy (long-term users) with an adjusted OR of 0.68 (95% CI 0.53 to 0.86) (data not shown). The adjusted OR for >3 years of treatment duration was 0.59 (95% CI 0.42 to 0.85) (Table 3⇓). In addition, we subdivided the first year into shorter periods of time. During the first 4 months, the OR was 0.90 (95% CI 0.54 to 1.50) and thereafter was 0.72 (95% CI 0.46 to 1.15).
The protective effect observed among long-term users of HRT was stronger for fatal events than for nonfatal events, with adjusted ORs of 0.38 (95% CI 0.21 to 0.68) and 0.77 (95% CI 0.59 to 0.98), respectively.
The ORs associated with long-term current-recent use of HRT by regimen, route, and dose are displayed in Table 3⇑. The use of unopposed therapy was associated with a greater reduction in the risk of myocardial infarction than the use of opposed therapy, with adjusted ORs of 0.52 (95% CI 0.35 to 0.78) and 0.79 (95% CI 0.59 to 1.08), respectively. The majority of women used oral preparations (79%), whereas 21% used transdermal therapy. The adjusted OR for long-term use of oral estrogens was 0.66 (95% CI 0.50 to 0.88) and for transdermal estradiol use was 0.75 (95% CI 0.47 to 1.21) (Table 3⇑).
The effect of estrogens dose was examined among current-recent users of long-term oral estrogens and transdermal estradiol, either unopposed or opposed. Among them, 51% of cases and 61% of control subjects used medium doses. The corresponding ORs for low, medium, and high doses were 0.96 (95% CI 0.55 to 1.65), 0.59 (95% CI 0.42 to 0.82), and 0.75 (95% CI 0.48 to 1.19), respectively (Table 3⇑). The ORs associated with ≥0.625 mg of oral estrogens was 0.63 (95% CI 0.46 to 0.86) and with ≥50 μg of transdermal estradiol was 0.62 (95% CI 0.37 to 1.06).
The OR associated with past use of HRT was 0.73 (95% CI 0.51 to 1.03) (Table 2⇑). The benefit on the risk of myocardial infarction provided by HRT was lost ≈2 to 3 years after the cessation of use. Among women who had stopped HRT >2 years before the index date, the OR was 0.9 (95% CI 0.3 to 3.0).
A further stratified analysis was performed to determine whether the beneficial effect of the therapy on the risk of myocardial infarction varied in subgroups of women with different levels of coronary risk factors (data not shown). HRT was inversely related to the risk of myocardial infarction in all subgroups of women except in those who had history of hypercholesterolemia and/or used lipid-lowering drugs, as well in those with diabetes. However, numbers were limited in these 2 subgroups of women. Among diabetic women (11 exposed cases and 16 exposed control subjects), the OR was 1.4 (95% CI 0.6 to 3.3).
In this study, current-recent use of HRT was associated with an overall 28% reduction in the risk of first myocardial infarction when compared with never-use of this therapy. Baseline differences in risk factors for myocardial infarction, such as age, history of smoking, hypertension, diabetes, and obesity, family history, type of menopause, comorbidity, use of aspirin, and calendar period, did not explain this beneficial effect.
The effect was stronger after the first year of continuous use of HRT, with a 32% reduction, and was present at medium-high doses (low dose was not associated with a reduced risk). The protective benefit appeared to diminish 2 to 3 years after cessation of the therapy. We did not observe any increase in the risk of first myocardial infarction associated with short-term HRT use.
The lack of information on some risk factors in a small proportion of women could have introduced some bias because of incomplete control of confounding. Information on smoking history was missing in 29% of women and was correlated to the use of HRT. Thus, recording of smoking was higher among women who ever-used hormones, suggesting that GPs were more likely to record smoking history among women who were willing to start HRT and were most likely at high risk of having a heart attack. Also, prevalence of recorded history of hypercholesterolemia, either treated or not with lipid-lowering drugs, was very low (4%). Although our cohort of postmenopausal women was free of cardiovascular disease history, this low percentage most likely reflects only the most severe cases of hypercholesterolemia.
In our cohort of relatively healthy postmenopausal women, and consistent with other results, the beneficial effect of HRT was observed in women without coronary risk factors and in women with ≥1 risk factor but no diabetes.16 17 Although this finding is based on small numbers, it might suggest a potential effect modification by diabetes. The existence of diabetes in premenopausal women appears to abrogate the cardiovascular protective effects of endogenous estrogens.18 Although mechanisms by which diabetes obviates the cardiovascular-protective effects of female sex hormones are not well understood, enhanced platelet aggregation, oxidative stress, coagulation, and fibrinolytic activity, as well as lipoprotein abnormalities, endothelial dysfunction, and a possible interaction between hyperglycemia and estradiol in regulation of cardiovascular nitric oxide production, have been invoked.19 To date, few studies have specifically assessed the effects of HRT on the incidence of CHD in diabetic women showing inconclusive results.20 21 One of these studies observed a beneficial effect with >6 years of estrogen therapy.20 A reduced atherosclerosis measured by carotid wall thickness associated with HRT use in women with and without diabetes has been also reported.22 Further investigations are required to evaluate if women with diabetes mellitus may receive the same benefit from HRT as women without diabetes.
In our study, the beneficial effect on the risk of myocardial infarction was observed in users of either unopposed estrogens and estrogens opposed with progestogens, although unopposed therapy was associated with a greater risk reduction. To the best of our knowledge, this is the only study examining the association between the route of the HRT and the risk of CHD. Although the relative risk for oral therapy was somewhat lower than the one for transdermal therapy, this result is compatible with either no differences or differences in the protective effect by route. A similar cardioprotection conferred by oral and transdermal HRT was further supported by the results of the stratified dose-response analysis. However, relative risk estimators for transdermal preparations were based on small numbers, resulting in relatively wide confidence intervals.
Observational studies looking at the effect of HRT have the potential for selection bias. Several studies have shown that women who choose to use HRT are healthier than nonusers.23 24 It has been suggested that users of HRT had a better premenopausal cardiovascular risk factor profile than nonusers.25 26 This type of bias, if present in our study, would underestimate the risk of myocardial infarction among users of HRT, thus exaggerating its protective effect. However, the study was restricted to women without cardiovascular and other chronic diseases. Also, when we adjusted for recorded hypertension, hypercholesterolemia, diabetes, and other risk factors the effect remained practically unchanged. In addition, we estimated the number of visits of each woman to their GP within 1 year before the index date. If women who take hormones see a physician regularly, these visits may be associated with a decreased risk of CHD. Compared with nonusers, higher proportions of women taking HRT were seeing their GPs >10 times per year. However, when we adjusted the analysis by the number of visits, the estimate of risk associated with HRT was slightly decreased (OR 0.69; 95% CI 0.56 to 0.85), probably reflecting a high-risk profile of some of these women. In a prior study of the same population, no differences were observed in social class and self reported health indicators between HRT users and nonusers.27 Nonetheless, a potential healthy user effect cannot be completely discarded. Our results are consistent with those observed in the Nurses’ Health Study (NHS), a homogenous cohort of healthy postmenopausal women.17 After controlling for many risk factors, there was still a strong inverse association between current HRT and the risk of CHD. Compared with our cohort, the cohort of NHS was younger and correspondingly presented a lower age-adjusted incidence rate of CHD (1.3 per 1000 person-years in NHS versus 1.6 per 1000 person-years in our cohort).
In our study, women with a recorded history of cardiovascular disease, neoplasms, coagulopathies, vasculitis, and alcohol-related diseases were excluded. However, since many practices included in the General Practice Research Database only started entering data routinely at the beginning of the 1990s, historic medical information could be incompletely recorded for a number of patients. History of cardiovascular disease is an important risk factor for myocardial infarction, and it is possible that women with history were less or more likely to receive HRT. This might have influenced the results in a somewhat under or overestimated relative protective effect of HRT. However, only 7 (0.6%) potential cases were excluded as a result of a prior myocardial infarction after reviewing the medical records.
Women taking HRT might be subject to referral or diagnostic bias, which would overestimate the risk of myocardial infarction, but also among these women an early suspicion would likely result in an early diagnosis of an acute coronary event.23 Indeed, we observed a greater reduction on the risk of fatal cases than of nonfatal cases of myocardial infarction.
Nondifferential misclassification could be present if women not truly having myocardial infarction were included as cases, independent of their exposure status. However, in our study, a detailed validation process resulted in a high percentage of definite cases according to our diagnostic classification. When we used definite and possible events as separate end points, the estimates of the beneficial effect of HRT were similar (data not shown). Consequently, this level of misclassification is unlikely to have introduced bias.
In summary, the results from this large population-based study are consistent with those of prior epidemiological studies, conducted in different populations with different methodologies, which show a 70% to 20% reduction in the risk of acute CHD events among current users of HRT in primary prevention.4 Our results also suggest that oral and transdermal therapy might have a similar cardioprotective effect at medium-high doses.
In our cohort of women, 50 to 74 years of age who did not use replacement therapy, the age-adjusted risk of myocardial infarction was ≈1.8 per 1000 women per year. Among women who used HRT for >1 year, this absolute risk was reduced by 32%. On average, this would result in 7 prevented cases of myocardial infarction among 10 000 women taking HRT. However, the magnitude of the protective effect would be most pronounced in women with a higher baseline risk of disease, as observed by others.28
We thank the general practitioners for their collaboration and the Boston Collaborative Drug Surveillance Program for providing access to the General Practice Research Database. We also thank Marta Marcet for the data entry and the editing of this report.
- Received August 3, 1999.
- Revision received December 16, 1999.
- Accepted December 22, 1999.
- Copyright © 2000 by American Heart Association
Barrett-Connor E. Sex differences in coronary heart disease: why are women so superior? The 1995 Ancel Keys Lecture. Circulation. 1997;95:252–264.
Perez Gutthann S, Garcia Rodriguez LA, Castellsague J, et al. Hormone replacement therapy and risk of venous thromboembolism: a population based case-control study. BMJ. 1997;314:796–800.
Perry J, ed. OXMIS Problem Codes for Primary Medical Care. Oxford, UK: Oxmis Publications; 1978.
Jick H, Jick SS, Derby LE. Validation of information recorded on general practitioner based computerised data resource in the United Kingdom. BMJ. 1991;302:766–768.
World Health Organization Regional Office for Europe. Myocardial Infarction Community Registers. Copenhagen, Denmark: WHO; Public Health in Europe No. 5; 1976.
Walker AM. Observation, and Inference. Newton Lower Falls, Mass: Epidemiology Resources Inc; 1991.
Kahn HA, Sempos CHT. Statistical Methods in Epidemiology. New York, NY: Oxford University Press; 1989.
Rothman KJ, Greeland S. Modern Epidemiology. Philadelphia, Pa: Lippincott-Raven Publishers; 1998.
SAS System for Personal Computers, Release 6.08. Cary, NC: SAS Institute Inc; 1992.
Donahue RP, Barrett-Connor E, Orchard TJ, et al. Endogenous insulin and sex hormones in atherosclerosis and coronary heart disease. Arteriosclerosis. 1988;8:544–548.
Kaplan RC, Heckbert SR, Weiss NS, et al. Postmenopausal estrogens and risk of myocardial infarction in diabetic women. Diabetes Care. 1998;21:1117–1121.
Lawrenson RA, Leydon GM, Newson RB, et al. Coronary heart disease in women with diabetes: positive association with past hysterectomy and possible benefits of hormone replacement therapy. Diabetes Care. 1999;22:856. Letter.
Dubuisson JT, Wagenknecht LE, D’Agostino RB Jr, et al. Association of hormone replacement therapy and carotid wall thickness in women with and without diabetes. Diabetes Care. 1998;21:1790–1796.
Posthuma WF, Westendorp RG, Vanderbroucke JP. Cardioprotective effect of hormone replacement therapy in postmenopausal women: is the evidence biased? BMJ. 1994;308:1268–1269.
Matthews KA, Kuller LH, Wing RR, et al. Prior use of estrogen replacement therapy: are users healthier than nonusers? Am J Epidemiol. 1996;143:971–978.
Grodstein F. Invited commentary: can selection bias explain the cardiovascular benefits of estrogen replacement therapy? Am J Epidemiol. 1996;143:979–982.
Lancaster T, Surman G, Lawrence M, et al. Hormone replacement therapy: characteristics of users and non-users in a British general practice cohort identified through computerised prescribing records. J Epidemiol Community Health. 1995;49:389–394.