OUP user menu

Hormone therapy and risk of myocardial infarction: a national register study

Ellen Løkkegaard, Anne Helms Andreasen, Rikke Kart Jacobsen, Lars Hougaard Nielsen, Carsten Agger, Øjvind Lidegaard
DOI: http://dx.doi.org/10.1093/eurheartj/ehn408 2660-2668 First published online: 30 September 2008


Aim To assess the risk of myocardial infarction (MI) as a result of hormone therapy (HT), with focus on the influence of age, duration of HT, various regimens and routes, progestagen type, and oestrogen dose.

Methods and results All healthy Danish women (n = 698 098, aged 51–69) were followed during 1995–2001. On the basis of a central prescription registry, daily updated national capture on HT was determined. National Registers identified 4947 MI incidents. Poisson regression analyses estimated rate ratios (RRs). Overall, we found no increased risk [RR 1.03 (95% CI: 0.95–1.11)] of MI with the current HT compared with women who never used HT; age-stratified RR among women aged 51–54, 55–59, 60–64, and 65–69 years were 1.24 (1.02–1.51), 0.96 (0.82–1.12), 1.11 (0.97–1.27), and 0.92 (0.80–1.06), respectively. An increasing risk with longer duration was found for younger women, which was not observed with older age groups. In all age groups, the highest risk of MI was found with continuous HT regimen. No increased risk was found with unopposed oestrogen, cyclic combined therapy, or tibolone. Significantly lower risk was found with dermal route than oral unopposed oestrogen therapy (P = 0.04). No associations were found with progestagen type or oestrogen dose.

Conclusion In a National cohort study, we found that HT regimen and route of application could modify the influence of HT on the risk of MI.

  • Hormone therapy
  • Hormone replacement therapy
  • Myocardial infarction
  • Coronary heart disease
  • Ischaemic heart disease
  • Oestrogen


Postmenopausal use of hormones was widely used in the western world until 2002, when the largest randomized clinical trial (RCT), the Woman’s Health Initiative (WHI), investigating the health effects of continuous combined hormone therapy (HT) was prematurely terminated due to overall increased morbidity with HT.1 This finding was unexpected, as a primary preventive effect of HT on cardiovascular diseases was predicted to outbalance the perceived increased risk of breast cancer and venous thrombo-embolism. These expectations were based on observational studies;2 however, the WHI found an increased risk of both coronary heart disease and stroke.1,3 These findings were in accordance with an earlier RCT that tested the effect of HT on re-event after coronary heart disease (the HERS study4).

Following the study termination, part of the WHI testing unopposed oestrogen vs. placebo among women without uterus was also prematurely stopped, as no cardio-protective was observed, and an increased risk of stroke was instead found.5 The discrepant findings from the RCT and observational literature were the topic of much debate. The observational studies could be influenced by a ‘healthy user’ bias,2 and the WHI was criticized for not being applicable for healthy younger perimenopausal women.6,7 Recently, the RCT and observational results have been found to be more in agreement if time since HT initiation was controlled for these studies was found to be in closer agreement.8

The WHI trial tested two hormone therapies: oral continuous combined therapy with conjugated equine oestrogen 0.625 and 2.5 mg/day medroxyprogesterone acetate vs. placebo for women with an intact uterus, which HERS also tested, and oral unopposed equine oestrogen 0.625 mg/day vs. placebo for women who had a hysterectomy. After termination of the WHI, no other randomized studies testing other hormone therapies have been initiated, despite positive results with lower oestrogen dosages from a pilot study.9 Studies exploring the overall increased risk of venous thrombo-embolism with hormone therapy indicate a lower risk with unopposed oestrogen therapy10 and potentially with dermal application.11

In the observational literature, little focus has been placed on the significance of various potentially important factors concerning the risk of myocardial infarction (MI) associated with hormone therapy, i.e. regimens as oestrogen monotherapy or combined oestrogen–progestagen therapy in cyclic or continuous combination; oral, dermal, or vaginal route of administration; chemical structure of the progestagen and dosages of oestrogen and progestagen.

In studies in Danish populations, we previously found that women using HT could not be characterized as ‘healthy users’,12,13 and in the Danish nurse cohort study, we found no overall protective effect of HT on MI, but instead found a harmful interaction between diabetes and hormone therapy.12

The purpose of this study was to assess the risk of MI associated with HT using the National registry information on all Danish women; specifically, we assessed the influence of duration of use, various regimens, routes of administration, progestagen types, and oestrogen dose, and also investigated the potential interactions between HT and register-recorded risk factors for cardiovascular diseases.


The Danish Sex Hormone Register Study (DaHoRS) is based on five National registers that are merged through an individual personal registration number given to Danish citizens at birth or at immigration, and hereafter replaced by a random number to ensure anonymity. The registers include the following: (i) the Civil Registration System (CRS) that registers all Danish inhabitants’ age and address, (ii) the National Register of Patients (NRP) that collects diagnoses from all hospitalizations in Denmark, (iii) the Cause of Death register, which has information from death certificates, (iv) The National Register of Medicinal Product Statistics, which records all prescriptions reimbursed on Danish pharmacies, and (v) Statistics Denmark that delivers information about the individual’s education.

Study base

In the CRS, a National cohort of all Danish women aged at least 51 years by 1 January 1995 or reaching 51 years during the period from 1 January 1995 to 31 December 2001 were identified. In order to focus the analysis on postmenopausal women, we used a cut off age of 51 years, as this was the average age at menopause in Denmark. Women were excluded from the cohort when they turned 70 years old. This open cohort included 748 324 women.

Exclusion criteria

We aimed to establish a cohort of healthy women; consequently, women recorded in the NRP with cardiovascular diseases or hormone-related cancers prior to entrance were excluded. The NRP has collected discharge diagnoses and surgical codes on all hospitalized patients since 1976, coded according to WHO’s international classification of diseases (version ICD-8 until end of 1993 and ICD-10 from 1 January 1994). The specific diseases leading to exclusion were previous ischaemic heart disease (ICD-8: 410–414; ICD-10: DI20–25), stroke (ICD-8: 430–31; ICD-10: 433–434; 436/DI60–64), venous thrombo-embolism (ICD-8: 450–4; ICD-10: DI26, DI80–82), breast cancer (ICD-8: 174; ICD-10: DC50), cancer of female genitals (ICD-8: 180, 182–184; ICD-10: DC53–57), colorectal cancer (ICD-8: 153–54; ICD-10: DC180–211) and haematological malignancy (ICD-8: 200–207; ICD-10: DC81–85, DC88, DC90–96).

In total, 23 657 and 25 342 women were excluded due to previous cardiovascular and malignant diseases, respectively, 1135 due to both, and 92 women were excluded due to only one day of observation, leaving 698 098 women in the cohort. These women were followed until the end of 2001, corresponding to 2 987 068 woman-years of observation.

A woman was excluded from the study if diagnosed with any of the diseases (except MI, which was considered an event) during the study period. Additionally, women were excluded upon emigration or death from reasons other than MI, or at turning 70 years of age.


Exposure to HT was recorded from the National Register of Medicinal Product Statistics (NRM), which has collected data on redeemed prescriptions by Danish citizens since January 1994, and is considered complete as of 1 January 1995.

In NRM, all prescriptions on hormone products were recorded by Anatomical-Therapeutical-Chemical (ATC) codes. The date the prescription was redeemed, pack size, number of packs, the defined daily doses, and administration form were available. The included ATC codes are described in detail in earlier publications;14 briefly, HT was categorized into six main groups according to regimen, 25 sub-groups according to chemical compounds, and 45 detailed groups according to the route of administration, and type and dose of oestrogen and progestagen.

For the cardiovascular analyses, women were grouped based on the type of hormone used most recently. Detailed information is available at www.dachre.dk. The exposures of these hormones were updated daily for each individual through the study period, and at the expiration of the prescription, the women used the hormones for four additional months to account for individual variation in prescription pattern and diagnostic delay in the NRP. HT exposure was thereafter considered a time-varying covariate in the statistical model. Exposure to hormones before age 51 but within the 6 year study period was also recorded and used in the calculation of duration of HT.

End points

The first event of MI was recorded (ICD-10 code DI21–22) in either the NPR or cause of death registry receiving information from death certificates. In total, 4947 events of MI were identified during the follow-up period.


Ages were calculated from birth dates, which were extracted from the individual person's registration number.

Information on education was recorded from the Statistics Denmark integrated database for labour market research. Potential confounders included the most recently completed education recorded at the start of study period in 1995: (i) elementary school/high school, (ii) occupational practice, (iii) short-term/middle term/longer education, or (iv) unknown. Information on surgical procedures, oophorectomy, and hysterectomy were determined from the NRP. Actual address at study entry was determined from the CRS and categorized into four regions: East Zealand and Bornholm (capital area), the remaining Zealand, Funen and Southern Jutland, and the remaining Jutland.

From the NRM, four time-varying indicator variables were recorded as positive upon the prescription of a minimum of 100 defined daily doses for one of four medical conditions: diabetes [A10A (insulin)/A10B (oral anti-diabetics)], cardiac arrhythmia [C01 (anti-arrhythmic)], hypertension [C02 (antihypertensive)/C03 (diuretics)/C07 (beta-blockers)/C08 (calcium antagonists)/C09 (drugs affecting renin–angiotensin system)] and hypercholesterolaemia [C10 (lipid lowering)]. These were used as updated confounders in the main analyses.

In total, 4388 had missing values on one of the confounders, leaving 693 710 women for analyses.

Statistical analysis

Data were analysed according to eight pre-specified time-varying hormone exposure definitions: (a) the hormone status: never used HT, previous HT, and current HT; (b) length of current therapy: short term (<1 year), middle term (1–4 years), long term (>4 years); (c) hormone regimen: oestrogen only therapy, cyclic combined oestrogen/progestagen therapy, long-cycle combined oestrogen/progestagen therapy (i.e. simultaneous redemption of 7–14 times more DDD oestrogen than DDD progestogen), continuous combined oestrogen/progestagen therapy, tibolone, and raloxifene; (d) route of administration: oral oestrogen, oral combined oestrogen/progestagen, dermal oestrogen, dermal combined oestrogen/progestagen, hormone-IUD, hormone-IUD, and oral oestrogen, hormone-IUD and dermal oestrogen, and local oestrogen; (e) type of progestagen: Norethisterone acetate (NETA), Medroxyprogesterone (MPA), Levonorgestrel (Lng), Cyproterone acetate (CPA); (f) dose of progestagen; cyclic combined, continuous low dose (0.5 mg NETA or 2.5 mg MPA), continuous high dose (≥1 mg NETA or ≥5 mg MPA); (g) type of oestrogen: conjugated equine oestrogen, non-conjugated oestrogen; (h) dose of oestrogen (only non-conjugated): low (<1 mg), middle (1–2 mg), and high (>2 mg).

When data were recorded, the person years and events with various levels of (f) progestagen dose and (g) oestrogen type were too few to determine the estimates for these pre-specified HT definitions. In addition, the following categories of variables had too little exposure and few MI to determine estimates: Raloxifene in (c) hormone regimen, hormone-IUD combinations in (d) route of administration, and CPA in (e) type of progestagen.

In the analysis of the axes, (a) hormone status and (b) length of therapy women never on systemic HT in the same age band was the reference group. Analysing the axis, the (c) hormone regimen, no HT (including vaginal treatment) group was used as the reference since local treatment was one of the levels in this axis.

Included confounders were crude model including age and calendar year, the adjusted models, including additionally education and four geographical areas, and the fully adjusted model, including also medication variables on diabetes, anti-arrhythmic, anti-hypertensive, and lipid-lowering medicine. Finally, duration adjusted models in which the various levels of exposure in the (c–h) definitions mentioned above were additionally sub-categorized according to the duration of the therapy.

Data was analysed by Poisson regression analysis on a data set consisting of risk time (women-years) and number of MI events for each combination of exposure axis, age band, and included confounders. Age was used as the timescale in the analyses, and women were divided into 5 year age bands (51–54, 55–59, 60–64, and 65–69 years), assuming constant risk of MI within each band.

As a model control, each model was checked for significance of an interaction between age (51–54, 55–59, 60–64, and 65–69 years) and exposure-axis as well as between age and each of the confounders. To eliminate random findings due to multiple testing, we lowered the P-value to 0.01 in the interaction testing. In some of the models, the interaction between age and exposure definitions was found to be significant in simple adjusted models, and consequently the results concerning these axes are presented in 5 year age bands.

Interactions between HT exposure and concomitant use of other medications (anti-diabetics, anti-arrhythmic, anti-hypertensive, and lipid-lowering medicine) were calculated.

Rate ratio estimates and 95% confidence intervals were calculated for each model. The statistical software used for the analysis was SAS, version 8.2.


The 698 098 women in the cohort resulted in 2 952 635 women-years of observation; 74% did not use HT during the observation period, 7% were previous users, and 19% were current users of hormones at censoring.

The risk of MI was associated with age, lower education, anti-hypertensive, and anti-diabetic medication, with taking anti-arrhythmic and lipid-lowering medicine (Table 1). The use of HT was positively associated with antihypertensive medication and gynaecological surgery, and inversely associated with the use of anti-diabetic medicine (Table 1).

View this table:
Table 1

Distribution of person years among healthy Danish women aged 51–70 years observed from 1995 to 2000

Year of birthWomen-years%MIRateWomen 1 January 2000CurrentPreviousNever
Age1925–1929250 8388.48563.4****
1930–1934610 73720.517402.895 52413.97.179.0
1935–1939728 70724.412211.7114 92519.310.170.6
1940–1944919 42830.88470.9150 29323.212.464.4
1945–1949477 35916.02830.6149 09320.311.068.7
EducationElementary School1 570 92152.634542.2249 73817.410.272.4
Occupational practice901 30430.210711.2164 00721.410.867.8
Further education458 30115.33190.786 88123.610.565.9
Unknown56 5421.91031.8920916.710.672.7
Geographical areaMetropolitan and Bornholm1 040 25734.814891.4178 28222.411.366.3
Other Zealand327 63411.06642.055 28217.910.471.7
Funen and Southern Jutland537 58118.08511.691 20720.510.768.8
Other Jutland1 081 59636.219431.8185 06417.49.673.0
MedicationNo lipid lowering2 946 89098.747201.6499 72119.810.569.7
Lipid lowering40 1781.42275.610 11416.811.471.8
No anti-arrhythmic2 950 83798.844891.5503 68819.410.570.1
Antiarrhythmic36 2311.245812.6614720.310.968.8
No anti-hypertensive2 235 80074.920360.9367 76918.59.871.7
Anti-hypertensive751 26825.229113.9142 06623.012.264.8
No anti-diabetic2 922 30797.844661.5497 40520.010.569.5
Anti-diabetic64 7612.24817.412 43011.48.879.8
Gynaecological operationNo hysterectomy2 727 94691.346191.7451 27817.59.972.6
Hysterectomy259 1228.73281.358 55737.414.847.8
No oophorectomy2 921 89997.848501.7496 67318.810.470.8
Oophorectomy65 1702.2971.513 16254.315.130.6
  • Myocardial infarctions (absolute and rate per 1000 women years) and hormone therapy (%) at 1 January 2000 according to the various background variables in analyses.

  • *Not included in the cohort 1 January 2000 due to inclusion of women aged 51–70 years.

Hormone status and risk of myocardial infarction

Compared with women who never used HT, the relative risk of MI with current use of HT among women aged 51–70 was 1.03 (95% CI: 0.95–1.11) and with past use was 0.81 (95% CI: 0.71–0.93). The risk associated with the use of hormones varied across age groups, as there was a significant interaction between age groups and HT status in the crude model (P = 0.005); however, this interaction was not significant in the fully adjusted model (P = 0.10). In the age group 51–54 years, the current use of hormones was associated with an increased risk of MI [RR 1.24 (1.02–1.51)]. In the older age groups, the relative risk was 0.92 (0.80–1.06) (Table 2). In women of 60–69 years of age, the previous use was associated with a decreased risk of MI. Exclusion of women who had an oophorectomy did not change the risk in women of 51–54 years of age [RR 1.26 (1.04–1.53)], nor did it influence the estimates of the older women (data not shown).

View this table:
Table 2

Risk of myocardial infarction (MI) in the various hormone therapy (HT) categories

HT StatusAgeWomen-yearsMIRate per 1000 women yearCrude RR95% CIAdjusted RR95% CI
Systemic HT status (P < 0.0001)
Never51–54610 8803740.611.001.00
55–59569 3316601.161.001.00
60–64510 79611102.171.001.00
65–69488 40915983.271.001.00
Previous51–5466 689380.570.950.681.330.840.601.18
55–5970 228761.080.960.761.220.940.741.19
60–6443 800671.530.720.570.930.740.570.94
65–6927 338642.340.730.570.940.770.600.99
Current51–54177 3401430.811.321.091.601.241.021.51
55–59192 1032071.080.930.801.090.960.821.12
60–64120 2472742.281.060.931.211.110.971.27
65–6975 4732112.800.860.750.990.920.801.06
Duration systemic HT (P < 0.0001)
<1 year51–5454 291420.771.210.881.671.180.861.63
55–5941 516421.010.820.601.130.840.611.15
60–6423 297692.961.271.001.631.331.041.70
65–6915 717503.180.910.691.210.950.721.27
1–4 years51–54101 337780.771.281.001.631.200.941.53
55–59108 2211151.060.930.761.140.960.791.17
60–6464 5111482.291.070.901.281.130.951.35
65–6940 5471112.740.850.701.030.910.751.11
>4 years51–5421 672231.061.811.192.771.591.042.44
55–5942 366501.181.060.791.071.070.801.44
60–6432 439571.760.850.651.110.890.681.16
65–6919 209502.600.830.631.110.890.671.19
Regimena (P < 0.0001)
Never any HT2 082 27735961.731.001.00
Oestrogen179 7422881.600.970.861.090.940.831.06
Long cycle combined26 097341.301.980.701.371.070.761.50
Cyclic combined220 1212441.110.850.750.970.920.811.05
Continuous combined118 1352442.
Tibolone19 457241.230.700.471.050.800.541.20
Routea (P < 0.0001)
Never any HT2 082 27735961.731.001.00
Oral oestrogen148 3882641.781.020.901.160.980.671.12
Dermal oestrogen31 354240.770.610.410.910.620.420.93
Oral combined358 6155231.461.010.921.111.080.981.19
Dermal combined25 196230.910.820.541.230.950.631.43
Vaginal68 723691.000.540.420.680.560.440.71
Oestrogen dosea (P < 0.0001)
Never any HT2 082 27735961.731.001.00
Unopposed low35 979511.420.910.691.200.910.691.20
Unopposed medium114 3812021.771.030.901.191.000.871.16
Unopposed high17 463251.430.870.591.300.820.551.21
Cont comb low6195101.611.140.612.131.300.702.42
Cont comb medium111 9392342.
Cont comb high21 116391.851.350.981.851.360.991.86
Progestagen typea (P < 0.0001)
Never any HT2 082 27735961.731.001.00
NETA continuous118 1342442.
NETA cyclic131 1671541.170.890.761.050.950.811.12
MPA cyclic42 905461.070.900.671.210.980.731.31
LEVO cyclic32 451331.020.720.511.010.780.561.10
  • The crude analyses were adjusted for age and calendar year. The adjusted analyses are additionally adjusted for education, employment status, habitation, and medication for hypertension, heart conditions, hyperlipidaemia, or diabetes. P-values for the significance of the various hormone therapy categories are in brackets. Significant results are in bold.

  • NETA, Norethisterone acetate; MPA, Medroxy progesterone acetate; LEVO, Levonorgestrel.

  • aThe reference group was never any systemic HT (excluding vaginal administration).

Duration of hormone therapy

Compared with women who never used HT, the risk according to the duration of HT was 1.06 (0.92–1.23), 1.03 (0.93–1.14), and 0.99 (0.85–1.16) for short-term (<1 year), middle term (1–4 years), and long-term (>4 years) use, respectively (P = 0.016). In the younger age groups, we observed an increased risk of MI with increasing duration of systemic HT, which was not observed in the older age groups (Table 2). There was significant interaction between age groups and duration of HT in the crude model (P = 0.007); this interaction, however, was not significant in the fully adjusted model (P = 0.12). In the analyses of the various axes categorized according to the duration of therapy, no consistent association with duration was found (data not shown).

Hormone regimen

The highest risk of MI was found with continuous combined therapy when compared with women who never used HT [RR 1.35 (1.18–1.53)] (Table 2), whereas the risk associated with cyclic combined regimen was 0.92 (0.81–1.05) and with tibolone was 0.80 (0.54–1.20). The difference in risk between women on continuous combined therapy vs. women on cyclic combined therapy, and vs. women on tibolone was significant, with P-values of <0.001 and 0.007, respectively. Unopposed oestrogen was not associated with the risk of MI and was not significantly different from the risk with cyclic combined therapy (P = 0.39).

The P-value for the interaction between age and HT regimen in the fully adjusted model was 0.09. For all age groups, the highest risk was found with continuous combined regimens (Figure 1).

Figure 1

The age-stratified risk of myocardial infarction with various hormone therapy regimens from multivariable model. Rate ratios (RR) with 95% confidence intervals (95% CI) are presented.

Route of administration

There was a significantly decreased risk of MI [RR 0.62 (0.42–0.93)] with dermal unopposed oestrogen compared with women who never used HT (Table 2). The risk associated with dermal unopposed oestrogen was significantly lower than for oral unopposed oestrogen use (P = 0.04). In women on combined therapy, no difference was detected between whether the treatment was administered orally or dermal (P = 0.33).

Vaginal oestrogen was associated with a significantly decreased risk of MI [RR 0.56 (0.44–0.71)].

Oestrogen dose

There was no overall indication of an increased risk of MI with increasing oestrogen dose (Table 2).

Progestagen type

Norethisterone acetate was the only progestagen administered with the continuous combined regimen. Consequently, NETA-containing regimens were subdivided as to whether they were administered in a continuous or cyclic combined regimen. For cyclic combined regimens, no indication of a differential effect with various progestagen types was detected (Table 2).

Interaction between hormone therapy and other medications

There were no significant interactions between the use of hormones and concomitant medications for diabetes, anti-arrhythmic, anti-hypertensives, or lipid-lowering medicine (Figure 2).

Figure 2

Rate ratios (RR) with 95% confidence intervals (95% CI) of the risk of myocardial infarction with current and previous use of hormone therapy compared with no HT use stratified by concomitant use of various medications. P-values from interaction terms between hormone status and medical variable was, respectively, 0.07 for anti-hypertensive, 0.51 for anti-diabetic, 0.17 for anti-arrhythmic, and 0.17 for lipid-lowering in full model adjusting for all significant covariates (education, habitation, and calendar year).


On the basis of national observational data, overall we found no association between HT and MI. We found an increased risk of MI among younger women on HT, which was correlated to the duration of use; no such correlation was found in older women on HT.

For all age groups, the highest risk was found with combined regimens. For the regimen equivalent to the WHI regimen, we found comparable estimates despite differences in design. The hazard ratio for unopposed oestrogen in the WHI was 0.95 (0.79–1.16); our estimate 0.94 (0.83–1.06). For combined regimens, WHI found an HR of 1.24 (1.00–1.54), while we calculated an RR of 1.35 (1.18–1.53).

Our age-stratified estimates were higher for the youngest age groups compared with the WHI results. Of note, our study had no information on menopausal status, although the majority in the young age group were postmenopausal due to the cut off at 51 years. However, the group who never used HT includes premenopausal women, and as late menopause possibly protects against ischaemic heart disease, suggesting that the risk estimate associated with HT in this group will be elevated.

Women oophorectomized at a young age have early menopause and are therefore at an increased risk of MI. To what extent HT decreases the risk of MI in these women is not known. Excluding these women from the group who never used HT should increase our risk estimate. On the other hand, excluding them from the current user group should not necessarily decrease the estimates, as HT may counteract the increased risk of MI in these oophorectomized women. We conducted analyses in which we excluded the oophorectomized women and found no change in these elevated risks among younger women. With these considerations, our data do not support the timing hypotheses that perimenopausal hormone therapy is associated with a reduced risk of cardiovascular disease.15,16

We found lower risk with cyclic combined than with continuous combined therapy. Cyclic regimens were not previously tested in randomized designs, and previous observational studies had limited power to test the differences between various combined regimens.12,17 Unfortunately, the available data did not allow us to test whether this is due to the differential dose of progestagen or whether it is based on the regimen per se. When cyclic regimens were considered, we found no significant difference in risk between HT with MPA, used in the USA, and NETA, which is mostly used in the Scandinavian countries. We found lower risk associated with dermal application, especially of unopposed oestrogen, in accordance with other studies.18 This interesting finding may possibly be explained by reduced activation of the haemostatic system due to the avoidance of the first pass hepatic effect.1921 We found a surprisingly low risk of MI with vaginal oestrogen, which should have little or no systemic effect. A biological effect here may be possible. Alternatively, this may be caused by residual confounding. This finding should be confirmed from other studies before clinical recommendations are given.

We were not able to test the effect of conjugated estrogens vs. 17 β-estradiol, as conjugated estrogens are infrequently used in Denmark. We found no clear association between oestrogen dose and risk of MI.

We do not believe that a healthy user bias was in effect for several reasons. First, previous studies have no evidence of such a bias.12,13 Secondly, we adjusted for education, and thirdly, women taking hormones were more frequently on medication against hypertension and cardiac arrhythmia, although more seldom on anti-diabetics.

We found no indication that women with pre-existing medically treated diabetes, hypercholesterolemia, hypertension, or heart arrhythmics had increased risk with HT, in contrast to our previous finding from the Danish Nurse Cohort Study in which diabetic women had a higher relative risk with hormones than those without diabetes.12 However, our data did not include the potential important confounders, weight and body fat distribution, both are potential confounders.22

Among the strengths of our study were the National unselected data. Denmark has free access to medical care and, to some extent, public refunds of medical expenses. Consequently, HT is generally not associated with a healthy user lifestyle.13 The data were collected before the results from the WHI were published, implying stable exposure during the study period. We had daily updated information on HT exposure and complete records of all hospitalized events from the NRP with high validity.23

Several limitations should be noted. Information on HT exposure is based on whether prescriptions are redeemed. In a previous validation study, we found high agreement between self-reported HT use and redeemed prescriptions.24 In contrast to classical cohort studies, the time window used in this study could result in exposure misclassification due to truncation of the database in 1995; this allowed older women who used HT in their 50s to be misclassified as having never used HT instead of as classification as previous users of HT. However, the influence of HT on coronary heart disease seems to be quite immediate.25 Consequently, this circumstance should have minimal influence on our cardiovascular analyses. Also, no information was available on individual risk factors such as physical activity, smoking, and alcohol habits, which could result in residual confounding. However, we do have individual information on education level and habitation, as well as the use of other medication for medical conditions that under other circumstances might be considered intermediate variables.


Our study found risk estimates of MI comparable with estimates in randomized clinical studies. Our data suggest a lower risk with cyclic combined than with continuous combined therapy, and low risk with dermal or vaginal application of oestrogen.


Copenhagen County University hospital (9870471).

Conflict of interest: none declared.


View Abstract