IJE Advance Access originally published online on January 4, 2006
International Journal of Epidemiology 2006 35(3):731-738; doi:10.1093/ije/dyi285
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Article |
Hormone replacement therapy and acute myocardial infarction: a large observational study exploring the influence of age
1 Medical Statistics Unit, Department of Epidemiology and Population Health, London School of Hygiene & Tropical Medicine, London WC1E 7HT, UK
2 Non-communicable Disease Epidemiology Unit, Department of Epidemiology and Population Health, London School of Hygiene & Tropical Medicine, London WC1E 7HT, UK
* Corresponding author. E-mail: joseph.kim{at}lshtm.ac.uk
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Background We examined whether the discrepancy between observational studies and randomized clinical trials (RCTs) on the effect of hormone replacement therapy (HRT) on myocardial infarction (MI) could be explained by differences in age of participants at the time of either HRT initiation or the MI event.
Methods A matched case–control study was performed using the General Practice Research Database. Cases (n = 22 225) had a first diagnosis of MI between 1987 and 2001; up to six controls (n = 144 085) were matched to each case based on age, sex, and practice. Conditional logistic regression was performed adjusting for traditional cardiovascular risk factors, as well as, cardiovascular drug use and consultation rate.
Results HRT users had a lower overall risk of MI compared with never users [odds ratio (OR): 0.76; 95% confidence interval (95% CI) 0.68–0.86); results were similar for opposed (OR: 0.73; 95% CI 0.68–0.86) and unopposed (OR: 0.79; 95% CI 0.66–0.96) therapy. This apparent benefit increased with older age at HRT initiation(18–44 years: OR = 0.99; 95% CI 0.74–1.33; 45–54 years: OR = 0.84; 95% CI 0.71–1.00; 55–64 years: OR = 0.67; 95% CI 0.54–0.82; 65–74 years: OR = 0.50; 95% CI 0.34–0.74; >75 years: OR = 0.55; 95% CI 0.23–1.32). Moreover, the cardioprotective effect was greater for MIs occurring at older ages (P-value for interaction = 0.003).
Conclusion The effect of HRT on MI becomes more pronounced with older age at initiation and at older ages at the time of an MI event. These findings contradict the hypothesis that the apparent protection seen in observational studies was due to the inclusion of younger participants.
Keywords Acute myocardial infarction, coronary heart disease, hormone replacement therapy, pharmacoepidemiology
Accepted 11 November 2005
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
There has been much controversy concerning the use of hormone replacement therapy (HRT) in the prevention of coronary heart disease (CHD).1–3 The evidence supporting a beneficial effect of HRT on clinical endpoints for CHD has arisen entirely from observational studies4 whereas contrasting evidence of no beneficial effect, and perhaps an early harmful effect, has come from randomized clinical trials (RCTs).5 The prospective cohort Nurses' Health Study6 found a 39% reduction in the risk of CHD among current users of HRT and a 45% reduction among women using combined HRT (i.e. oestrogen plus progesterone). The two large RCTs of HRT—the Woman's Health Initiative (WHI)7 and the Heart and Estrogen/progestin Replacement Study (HERS) trial8—showed no evidence of benefit on CHD and, in fact, suggested a possible harmful effect within the first few years of use.
A leading explanation put forward to explain this discrepancy is the presence of selection bias and uncontrolled confounding in the results of observational studies consequent to the absence of randomization of exposure. The issue of time dependency between the onset of HRT therapy and the occurrence of CHD events has also been explored.9 Another possible explanation—one which has not been fully investigated—is the age of participants at the time of HRT initiation as well as at the time of the CHD event.2,10 Given that women enrolled in the WHI were substantially older (aged 50–79 years) than their counterparts in the Nurses' Health Study (aged 30–55), it would be of interest to explore the possible influence of age on the HRT–CHD relationship.
We present here findings from a large nested case–control study—involving 166 310 women selected from the UK General Practice Research Database (GPRD)—to explore whether differences in the age of participants between observational studies and RCTs help explain the HRT controversy.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The GPRD is a population-based longitudinal database containing detailed demographic and medical information on over 3 million people in the UK.11 The GPRD constitutes the largest source of continuous data on individuals' illnesses and prescribing habits in the UK.12 Participating practices broadly represent UK general practices. The GPRD aims to capture data on all drug prescriptions, consultations, and diagnoses recorded by a participating general practitioner including all hospital diagnoses. The diagnosis of myocardial infarction (MI) has been previously validated in the GPRD, where a diagnosis of MI was confirmed in >90% of computerized cases after medical record review.13,14 The GPRD prescribing data have been shown to be over 95% complete.15
The source population in this study were all women registered with a participating GPRD practice from June 1987 to December 2001. Eligible cases were women with a first diagnosis of MI (identified using OXMIS and READ codes), which occurred during the study period. All subjects were continuously registered with their GPRD practice for at least 1 year prior to the index date (i.e. defined as the date of MI occurrence). Subjects with a previous history of MI were excluded from the study. For each case, up to six controls with no record of MI were matched on age and practice. Controls were alive and registered with the GPRD on the index date of their matched case.
HRT exposure definitions were matched as closely as possible to those of the WHI. HRT use was classified into two types reflecting treatment in the two WHI trials16,17: oestrogen only (unopposed therapy) and combined HRT (oestrogen plus progestogen; opposed therapy). In the WHI oestrogen only trial, patients were randomized to 0.625 mg per day of conjugated equine oestrogen or to placebo, while in the combined HRT trial, patients were given the same oestrogen plus 2.5 mg per day medroxyprogesterone acetate or placebo. In our study, combined HRT was defined as the use of a single product that included oestrogen plus progestogen or the use of a concurrent prescription of progestogen given within 6 months of oestrogen. We also restricted attention to ‘current’ users of HRT (defined as HRT use within the 6 months prior to the index date). Age of HRT initiation was calculated using the first recorded instance of HRT use within the time frame of observation. Duration of use was not included as part of the analysis owing to the large volume of missing data on prescription days.
Patients with a history of tibolone use were excluded from analysis. Topical formulations of HRT were also excluded, including both vaginal and transdermal formulations. In total, 811 (3%) cases and 4958 (3%) controls were dropped from analysis because of both these restrictions. In the specific analysis of opposed and unopposed HRT therapy, we only considered patients who never switched between different HRT formulations. Only those prescriptions given prior to the index date were included in the analysis.
Potential confounders present at the time of the index date considered for adjustment were history of angina, hyperlipidaemia, hypertension, other atheromatous disease, peripheral vascular disease, atrial fibrillation, stroke, heart failure, diabetes, smoking, and alcohol use, as well as, consultation rate, body mass index (BMI), and cardiovascular drug use (including aspirin use). Consultation rate was considered for adjustment since women who receive HRT may have higher rates of consultation, which may also increase the opportunity for recording other diagnoses. Consultation rate was calculated by dividing the total number of clinic visits prior to the index date by the length of each patient's follow-up. BMI used weight in kilograms divided by the square of height in metres.18 Patients were categorized according to height and weight measures available nearest to the index date: non-overweight (BMI < 25 kg/m2), overweight (25 
BMI < 30 kg/m2), obese (30 BMI < 36 kg/m2), and severely obese (>36 kg/m2). Cardiovascular drug use was defined as the prescribing of any drug listed in Chapter 2 of the British National Formulary.19
Patient characteristics were evaluated using a two-sample t-test for quantitative variables and the chi-squared test for proportions. Multivariable analysis was performed using conditional logistic regression adjusting for hyperlipidaemia, hypertension, atheroma, angina, atrial fibrillation, peripheral vascular disease, stroke, heart failure, diabetes, smoking, alcohol, BMI, aspirin use, cardiovascular drug use, and consultation rate. All statistical analyses were performed using Stata version 8.2 (Stata Corporation, College Station, TX, USA).
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
From the population of women registered with general practices participating in the GPRD, 22 225 women with a first recorded MI and 144 085 age and practice matched controls were identified. Among cases and controls, 2.1% (475/22 225) and 2.5% (3590/144 085) of women, respectively, had ever used HRT. Of the study population, 5% (8137/166 310) were aged 45–54 years on the index date, of whom 20% (1622/8137) had a history of HRT use. This is in contrast to women aged 65 and older who comprised 78% (128 970/166 310) of the population, where only 0.4% (575/128 970) of women in this age group had used HRT. The mean age at MI was 73 years (SD 11 years).
The distribution of available risk factors for MI is shown in Table 1. Results show that compared with non-users, HRT users were more likely to have a history of hyperlipidaemia but less likely to be diabetic or have a history of angina, atrial fibrillation, stroke, and heart failure. They were also more likely to smoke, be overweight, be consistent drinkers, and have more frequent general practice consultations.
|
The association of patient factors with risk of MI is shown in Table 2. As expected, the majority of measured risk factors were positively related to MI occurrence, while alcohol use and cardiovascular drug use were found to be protective against MI. Missing data were common for BMI, smoking and alcohol use: cases were more likely to have missing data on BMI (53% vs 51%, P < 0.001) but were less likely to have missing data on smoking (32% vs 37%, P < 0.001).
|
Table 3 shows an overall 18% reduction in the crude odds for MI among HRT users (OR = 0.82; 95% CI 0.74–0.91); adjustment for covariates slightly enhanced this association (OR = 0.76; 95% CI 0.68–0.86). The table also shows that the results for women who only received opposed therapy were similar to those for any HRT use. The association between HRT and MI was somewhat less pronounced among those who were only prescribed unopposed therapy.
|
The cardioprotective benefit of HRT appears to increase with older age at initiation, where the effects become significant after the age of 55 years (Table 4). The largest benefit was observed in current users of HRT who began therapy between 65 and 74 years of age (OR = 0.50; 95% CI 0.34–0.74); women in the oldest category experienced a similar reduction in risk (OR = 0.55; 95% CI 0.23–1.32). Results for women who only received opposed or unopposed therapy were similar to those for any HRT use. Tests for linear trend with age for all three HRT types were statistically significant (P < 0.01).
|
Table 5 shows that the cardioprotective effect of HRT appears to increase with a woman's age at the time of the MI event (P-value for interaction = 0.003). HRT provided no benefit for women who experienced an event before the age of 45; however, it conferred up to nearly 50% reduction in the odds of MI among women in the highest age categories.
|
We also found that the influence of HRT on MI seems to depend neither on the calendar year HRT therapy was initiated nor on the year of the MI event (results not shown). There were no significant differences found in a repeat analysis of this study comparing opposed vs unopposed therapy.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
We found in this large observational study that the apparent beneficial influence of HRT on risk of MI becomes more pronounced with older age at the initiation of HRT therapy as well as with older age at the time of the MI event. Thus, these findings suggest that the age gradient in risk is in the opposite direction from what was to be expected based on current hypotheses—including the one mentioned in a recent issue of the International Journal of Epidemiology10—which attempt to explain the discrepancy between observational studies and RCTs by arguing that the apparent protective association seen in observational studies could have occurred because of a stronger protective effect at younger ages. Moreover, these findings do not support the notion that the failure to find protection in the RCTs was a result of studying older women; in fact, this study showed that the apparent protective effect is accentuated in older women. We propose, therefore, that the most plausible explanation for the discrepancy between observational studies and RCTs is possible selection bias and unmeasured confounding affecting all age groups regardless of HRT type (i.e. opposed or unopposed therapy).
The use of a large dataset, such as the GPRD, allowed us to study the potential modifying effects of age by providing sufficient numbers of patients within each age group. Adjustment for available potential confounding factors made little difference to the observed association in our study. Unmeasured confounding still remains a crucially important problem. The apparently large protective effect seen in older HRT users (in direct contradiction to the results of large RCTs) suggests that some uncontrolled confounding persisted. For example, our findings of protection against MI may be explained in part by our inability to fully control for socioeconomic status. Lawlor et al.20 recently found higher socioeconomic status to be positively related to HRT use. Higher socioeconomic status is known to be associated with a lower risk of CHD. Thus incomplete adjustment for socioeconomic status in observational studies, including ours, may partly account for the apparent protective effect observed. Some previous observational studies that fully adjusted for socioeconomic status did not find a protective association of HRT on CHD.21 However, the authors of the Nurses Health Study (an observational study that found a protective association between HRT use and risk of CHD) claim that their results were not affected because nurses, for the most part, are a socially homogeneous group.22 This study also has substantial missing data on smoking and BMI, which were significantly related to both caseness and exposure status; therefore, we cannot exclude the possibility of residual confounding in our study results.
The discrepancies between observational studies and randomized studies in whether HRT has a protective effect against CHD do not have an unequivocal indisputable explanation. Overall, observational studies and RCTs of HRT have found some consistency, for example in seeing an increased risk of breast cancer and a reduction in colorectal cancer, although these latter effects were not seen in the oestrogen-only arm of the WHI in women with a hysterectomy. The observational studies23,24 found little difference between oestrogen alone and combined HRT; though, some studies, notably the Million-Women Study,25 found an important difference between them in breast cancer. It has been argued that observational studies are more reliable in looking at harms than benefits, perhaps, because the harms are usually unanticipated associations.26,27
A limitation of this study was that age of HRT initiation was calculated using available data within the time frame of study observation. Therefore, it is likely that at least some individuals with a first recorded use of HRT—particularly those over the age of 65—began HRT therapy prior to the observation period. In addition, a large volume of data on prescription days were missing, which precluded further analysis by duration of treatment.
The findings of this study may be explained by selection effects that were stronger among older women, who chose to be on HRT therapy, who also had a lower risk of MI. In general, younger women tend to be on HRT primarily for symptomatic relief of hot flushes, whereby the health selection effects associated with such treatment are likely to be minimal. However, once through the symptomatic phase of menopause, the main reason for starting HRT treatment or continuing to take HRT is likely to be the potential for longer-term health benefits. The scope for selection bias is, therefore, much greater among older women; and hence, the observed stronger protective effect of HRT at older ages.
The appropriate interpretation of our study is affected heavily by the results of RCTs, which are less subject to selection bias than observational studies owing to the random allocation of treatment.28 The lack of a protective association between HRT use and MI in RCTs has previously been attributed to the older age of women in the RCTs compared with observational studies. It has been argued that the association between HRT use and MI found in observational studies among younger women represents a direct causal protective effect of HRT. However, this logic is directly contradicted by our finding of a protective association between HRT use and MI that was stronger among older women. Prentice29 has suggested that the observational studies are poor at studying either early or cumulative effects of HRT because they include few new users of HRT. This is partially true of our study also, and we do not have the full range of possible confounding factors available to Prentice.
Our observational study is the largest so far demonstrating that women who take HRT have a lower risk of MI than others of the same age. This finding is well beyond the realms of chance, exists over a broad range and is not explained by potential confounders available in GPRD. Nevertheless, the quite different results from RCTs suggest that it is inappropriate to conclude that HRT actually causes a reduced risk of MI. The most plausible explanation is selection bias and unmeasured confounding. That is, those women who choose to take HRT (with the help of their doctors) tend to be at lower risk of developing CHD.
Current evidence from RCTs using older women suggests that HRT has an adverse effect on the risk of cardiovascular disease. It has been proposed that this adverse effect may not be generalizable to younger women. However, the findings from this study suggest that the misleading results of observational studies with regard to CHD are not because they have studied younger women. Appropriate caution is required when interpreting the apparent beneficial effects of HRT, whatever the age of women.
KEY MESSAGES
|
|
Acknowledgments |
|---|
The authors of this paper thank Mike Bennett for providing database management support.
Conflicts of Interest
The authors have declared no conflict of interest.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
1 Vandenbroucke JP. Commentary: the HRT story: vindication of old epidemiological theory. Int J Epidemiol 2004;33:456–57.
2 Lawlor DA, Davey Smith G, Ebrahim S. Commentary: the hormone replacement-coronary heart disease conundrum: is this the death of observational epidemiology? Int J Epidemiol 2004;33:464–67.
3 Grodstein F. Invited commentary: can selection bias explain the cardiovascular benefits of estrogen replacement therapy? Am J Epidemiol 1996;143:979–82; discussion 983–4.
4 Hu FB, Grodstein F. Postmenopausal hormone therapy and the risk of cardiovascular disease: the epidemiologic evidence. Am J Cardiol 2002;90:26F–29F.[CrossRef][ISI][Medline]
5 Beral V, Banks E, Reeves G. Evidence from randomised trials on the long-term effects of hormone replacement therapy. Lancet 2002;360:942–44.[CrossRef][ISI][Medline]
6 Grodstein F, Stampfer MJ, Manson JE et al. Postmenopausal estrogen and progestin use and the risk of cardiovascular disease. N Engl J Med 1996;335:453–61.
7 Manson JE, Hsia J, Johnson KC et al. Estrogen plus progestin and the risk of coronary heart disease. N Engl J Med 2003;349:523–34.
8 Hulley S, Grady D, Bush T et al. Randomized trial of estrogen plus progestin for secondary prevention of coronary heart disease in postmenopausal women. Heart and Estrogen/progestin Replacement Study (HERS) Research Group. JAMA 1998;280:605–13.
9 Prentice RL, Langer R, Stefanick ML et al. Combined postmenopausal hormone therapy and cardiovascular disease: toward resolving the discrepancy between observational studies and the Women's Health Initiative clinical trial. Am J Epidemiol 2005;162:404–14.
10 Stampfer M. Commentary: hormones and heart disease: do trials and observational studies address different questions? Int J Epidemiol 2004;33:454–55.
11 Garcia Rodriguez LA, Perez Gutthann S. Use of the UK General Practice Research Database for pharmacoepidemiology. Br J Clin Pharmacol 1998;45:419–25.[CrossRef][ISI][Medline]
12 Smeeth L, Hubbard R, Fletcher AE. Cataract and the use of statins: a case–control study. QJM 2003;96:337–43.
13 Jick H, Derby LE, Gurewich V, Vasilakis C. The risk of myocardial infarction associated with antihypertensive drug treatment in persons with uncomplicated essential hypertension. Pharmacotherapy 1996;16:321–26.[ISI][Medline]
14 Jick H, Vasilakis C, Derby LE. Antihypertensive drugs and fatal myocardial infarction in persons with uncomplicated hypertension. Epidemiology 1997;8:446–48.[CrossRef][ISI][Medline]
15 Walley T, Mantgani A. The UK General Practice Research Database. Lancet 1997;350:1097–99.[CrossRef][ISI][Medline]
16 Rossouw JE, Anderson GL, Prentice RL et al. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results From the Women's Health Initiative randomized controlled trial. JAMA 2002;288:321–33.
17 Patterson L. Effects of conjugated equine estrogen in postmenopausal women with hysterectomy. The Women's Health Initiative randomised controlled trial. JAMA 2004;291:1701–12.
18 Criqui MH, Klauber MR, Barrett-Connor E, Holdbrook MJ, Suarez L, Wingard DL. Adjustment for obesity in studies of cardiovascular disease. Am J Epidemiol 1982;116:685–91.
19 British National Formulary. London, UK: British Medical Association and the Royal Pharmaceutical Society, 2001.
20 Lawlor DA, Davey Smith G, Ebrahim S. Socioeconomic position and hormone replacement therapy use: explaining the discrepancy in evidence from observational and randomized controlled trials. Am J Public Health 2004;94:2149–54.
21 Humphrey LL, Chan BK, Sox HC. Postmenopausal hormone replacement therapy and the primary prevention of cardiovascular disease. Ann Intern Med 2002;137:273–84.
22 Grodstein F, Manson JE, Colditz GA, Willett WC, Speizer FE, Stampfer MJ. A prospective, observational study of postmenopausal hormone therapy and primary prevention of cardiovascular disease. Ann Intern Med 2000;133:933–41.
23 Nelson HD, Humphrey LL, Nygren P, Teutsch SM, Allan JD. Postmenopausal hormone replacement therapy: scientific review. JAMA 2002;288:872–81.
24 Stampfer MJ, Colditz GA. Estrogen replacement therapy and coronary heart disease: a quantitative assessment of the epidemiologic evidence. Prev Med 1991;20:47–63.[CrossRef][ISI][Medline]
25 Beral V. Breast cancer and hormone-replacement therapy in the Million Women Study. Lancet 2003;362:419–27.[CrossRef][ISI][Medline]
26 Vandenbroucke JP. When are observational studies as credible as randomized trials? Lancet 2004;363:1728–31.[CrossRef][ISI][Medline]
27 Miettinen OS. The need for randomization in the study of intended effects. Stat Med 1983;2:267–71.[Medline]
28 Pocock SJ, Elbourne DR. Randomized trials or observational tribulations? N Engl J Med 2000;342:1907–09.
29 Prentice RL. Chronic disease prevention: public health potential and research needs. Stat Med 2004;23:3409–20.[Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  G. M. Sare, L. J. Gray, and P. M.W. Bath Association between hormone replacement therapy and subsequent arterial and venous vascular events: a meta-analysis Eur. Heart J., July 3, 2008; (2008) ehn299v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. J Stampfer Commentary: Hormones, heart disease, and the definition of hormone 'initiation' Int. J. Epidemiol., June 1, 2006; 35(3): 738 - 739. [Full Text] [PDF]
|
|
|
|
|
|
J. Kim, S. Evans, L. Smeeth, and S. Pocock Response: Response to the Stampfer commentary Int. J. Epidemiol., June 1, 2006; 35(3): 740 - 740. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||