Cardiovascular risk at age 53 years in relation to the menopause transition and use of hormone replacement therapy: a prospective British birth cohort study

Authors


Professor D. Kuh, MRC National Survey of Health and Development, Department of Epidemiology and Public Health, Royal Free and University College Medical School, 1-19 Torrington Place, London WC1E 6BT, UK.

Abstract

Objectives  To investigate cardiovascular risk factors and changes in risk factor levels in relation to menopausal stage, hysterectomy status and hormone replacement therapy use in a cohort of women aged 53 years with prospective data on smoking, lifetime socio-economic circumstances, and blood pressure and obesity at age 43 years.

Design  A prospective study.

Setting  England, Scotland and Wales.

Population  A cohort of women from the Medical Research Council Survey of Health and Development.

Methods  A total of 1303 women, aged 53 years, from a UK birth cohort study with measures of cardiovascular risk factors were classified by five menopausal status groups (premenopause, perimenopause, postmenopause, hysterectomy and hormone replacement therapy user). Body mass index, glycosolated haemoglobin, blood pressure, high density lipoprotein, low density lipoprotein and total cholesterol measurements were taken, and analysed within the groups taking confounding variables into account. Changes in body mass index and blood pressure measurement in the same women obtained when 43 years of age were also compared.

Main outcome measures  Body mass index, glycosolated haemoglobin, blood pressure, high density lipoprotein, low density lipoprotein and total cholesterol.

Results  At 53 years, body mass index, waist circumference, total and low density lipoprotein cholesterol, and glycosolated haemoglobin (HbA1c) varied by menopausal status group, but blood pressure did not. Levels of total cholesterol and HbA1c increased across the natural menopause transition, before and after adjustment for body mass index, smoking and lifetime socio-economic circumstances. After adjustment for confounders, levels of risk factors for hysterectomised women were similar to those of naturally postmenopausal women. Women on hormone replacement therapy had lower levels of total and low density lipoprotein cholesterol, HbA1c, and were less obese than postmenopausal women. The lower obesity levels were partly due to these women already being less obese at age 43 years.

Conclusions  This study showed that naturally postmenopausal or hysterectomised women had higher levels of metabolic risk factors compared with premenopausal or perimenopausal women of the same age. The long term stability of these differences and their translation into variations in incidence of cardiovascular disease remain to be seen. The lower levels of metabolic risk factors for women on hormone replacement therapy may protect against future cardiovascular disease or may be overwhelmed by other adverse, and as yet unknown, effects of hormone replacement therapy.

INTRODUCTION

The role of the natural menopause on cardiovascular disease is unclear,1,2 and may be confounded by smoking status.3–5 Several studies have reported adverse effects of the menopause on cardiovascular risk factors such as lipid profiles, blood pressure and obesity,6 but the evidence is inconsistent. Prospective studies,7,8 which follow women through the menopausal transition, indicate that the effects of menopause on lipid levels are the most convincing. However, even in these longitudinal studies, it is difficult to distinguish a discrete effect of the menopause on cardiovascular risk from age-related effects. We examined cardiovascular risk factors in relation to the natural menopause transition in a cohort of women all born in one week in March 1946. Thus, the major advantage of this study is that all women were the same age (53 years) at examination, so there are no age or time period effects to add complexity to the analysis. Measures of cardiovascular risk available were blood pressure, central and total obesity, total cholesterol, high density lipoprotein and low density lipoprotein cholesterol, and glycosolated haemoglobin (HbA1c). HbA1c has been recommended as a reliable estimate of average plasma glucose when studying the long term effects of menopause or hormone replacement therapy use.9,10

Most studies of the effect of the menopausal transition on cardiovascular risk are unable to allow for premenopausal levels of cardiovascular risk factors and thus cannot disentangle the direction of association. A second advantage of our study is that measures of blood pressure and central and total obesity were also available at age 43 years (before most women had reached menopause). This allowed us to test whether changes in these cardiovascular risk factors were associated with the menopausal transition.

A sizeable proportion of women in western countries do not experience a natural menopause because of hysterectomy and/or oophorectomy. Studies have generally found that women with bilateral oophorectomy have an increased risk of heart disease.1,4 Results for women who have had a hysterectomy without bilateral oophorectomy are more mixed.11–13 Hysterectomy is more common in women who are less educated, of lower socio-economic position and higher body mass index and these potential confounders are not always taken into account.14

Use of hormone replacement therapy during the menopause transition is common in many western countries,15–18 and further complicates the relationship of cardiovascular risk and the menopause transition. The role of hormone replacement therapy for cardiovascular risk is unclear19 and is confounded by menopausal or hysterectomy status, social and behavioural factors, as well as prior risk factor levels and disease history, which are associated with overall hormone replacement therapy use and/or choice of preparation.2,20 Hormone replacement therapy has a generally favourable impact on lipoproteins, depending on preparation type, route and dose.21–23 It has inconsistent effects on blood pressure and glucose metabolism, although not all studies dealt adequately with potential confounding. In contrast to observational studies, most clinical trials show either no effect or adverse effects of hormone replacement therapy on cardiovascular disease.24–27

The third advantage of this study is its nationally representative sample with information on hysterectomy status, hormone replacement therapy use and potential confounders. The purpose of this article is to examine whether cardiovascular risk factors and changes in risk factor levels are related to menopausal stage, hysterectomy status and hormone replacement therapy use at age 53 years, taking into account lifetime social class and smoking behaviour, and current body size as appropriate.

METHODS

The Medical Research Council National Survey of Health and Development is a socially stratified sample of 2547 females and 2815 males followed up since their birth in England, Scotland and Wales in the first week of March 1946.28–30 During a follow up at age 53 years, contact was attempted with 3386 study members and 1566 women and 1469 men (90% of those contacted) provided information. No attempt was made to contact the remaining 1976 (37%) members of the original cohort; these study members had either previously refused to take part (12%), were living abroad (11%), were untraced (5%) or were known to have died (9%). The adult sample remains broadly representative in socio-economic terms of the cohort born in England, Scotland and Wales.30

At the home visit at 53 years and the previous visit at age 43 years, nurses measured height (cm), weight (kg), waist (cm) and hip circumference (cm) according to a standardised protocol.31 Body mass index (kg/m2) was used as a measure of total obesity and waist–hip ratio (calculated as a percentage) and waist circumference (cm) were used as measures of central obesity. Blood pressure in the brachial artery of the upper left arm was taken twice by the nurse with the participant sitting and after 5 minutes of rest at 43 and 53 years. The second reading was used in the analyses.32

A non-fasting venous blood sample was taken by the nurse according to standardised protocol at the same home visit at age 53 years. As described elsewhere,33 total cholesterol and high density lipoprotein cholesterol were analysed using standard laboratory procedures and low density lipoprotein cholesterol was calculated using the Friedewald formula. Samples were analysed for glycosylated haemoglobin with the Tosoh A1C 2.2 Plus Analyzer (Tosoh, Tokyo, Japan) using high performance liquid chromatography. Those reporting the use of anti-diabetic medication or doctor-diagnosed diabetes (n= 35) and two women with values of HbA1c well outside the normal range (4–8%) were excluded from the analyses.

Most of the women with a measure of cardiovascular risk had also participated in an annual postal study of the menopausal transition between 47 and 54 years34 or provided comparable information at the home visit at age 53 years. From these data, menopausal or hysterectomy status and use of hormone replacement therapy at age 53 years were derived. Five main menopausal status groups were defined. The first three groups denoted the stage of the natural menopause transition (premenopausal, perimenopausal and postmenopausal) for women not on hormone replacement therapy, based on the method used in the Massachusetts Women's Health Study.35 Women were classified on the basis of reported last menstrual period, as postmenopausal if periods had ceased for more than 12 months, perimenopausal if periods had ceased for between 3 and 12 months or were reported to be less regular and premenopausal if they still had periods and reported no change in regularity. A fourth menopausal status group included women who experienced a hysterectomy or bilateral oophorectomy but were not taking hormone replacement therapy. The fifth group included all women taking hormone replacement therapy and was further divided in some analyses according to the women's menopausal status before the start of hormone replacement therapy.

A history of cigarette smoking by age 53 years (lifelong non-smoker, ex-smoker and current smoker), adulthood and childhood social class (retaining all six classes of the British Registrar General's social class classification) were included as potential confounders. Childhood social class was based on the father's occupation, and adult social class was based on own current or last occupation as this was more strongly related to the cardiovascular risk factors than a social class based on partner's occupation. Previous studies of this cohort and others have shown that smokers have an earlier menopause,14,36 an increased risk of cardiovascular disease37,38 and are more likely to take hormone replacement therapy.15,39 Similarly, women from a less advantaged childhood or adult social class have adverse levels of cardiovascular risk factors,31,40–42 and experience earlier menopause than their more advantaged peers.43,44 In older cohorts14 (although not in this cohort so far15), hormone replacement therapy use is associated with more advantaged adult socio-economic circumstances. Body mass index was used as an additional confounder in analyses with central obesity, blood pressure, cholesterol and HbA1c as outcomes.

Of the 1566 women who provided information at 53 years, 263 women without measured height and weight, or whose menopausal status group was unknown because of insufficient information or because period cessation was due to chemotherapy or other treatment were excluded from the analyses. The remaining 1303 women (83%) who had their height, weight, waist and hip circumference measured and could be assigned a menopausal status group formed the main sample for analysis. Almost all these women (96%) had their measurements made within a seven-month period when they were 53 years old. Of this main sample, 1157 (89%) had complete data on confounders. Somewhat fewer women had blood pressure measurements (n= 1284, and n= 1095 in the restricted sample with information also on confounders), lipid measurements (n= 1087 and n= 1009 in the restricted sample) or a measure of glycosolated haemoglobin (n= 1104, n= 1024 in the restricted sample).

First, mean levels of body mass index, waist circumference, waist–hip ratio, systolic and diastolic blood pressure, total cholesterol, high density lipoprotein and low density lipoprotein cholesterol and HbA1c at ages 53 (for all outcomes) and 43 years (for body weight and shape and blood pressure only) were examined in relation to the five menopausal status groups in the total samples available. A test of heterogeneity (F test) across all five menopausal status groups was carried out, followed by a test for linear trend across the three natural menopause groups. Then using multiple regression models, the mean differences in the levels of these cardiovascular risk factors by menopausal status groups were examined in the restricted samples with complete data, first unadjusted, then adjusted for potential confounders and finally adjusted for measures taken at age 43 (where available). In these models, postmenopausal women were taken as the baseline group. Further regression analyses were undertaken to identify any differences among hormone replacement therapy users who had started hormone replacement therapy before or after the menopause or had undergone hysterectomy. All analyses were carried out using the Statistical Package for the Social Sciences (SPSS).

RESULTS

Of the 1303 women in the main sample, 62% were not taking hormone replacement therapy and these included 6% who were still premenopausal, 16% who were perimenopausal, a third (33%) who were postmenopausal and just 8% who had undergone hysterectomy and were not on hormone replacement therapy. Thirty-eight percent of the sample were on hormone replacement therapy and this included 20% who had started hormone replacement therapy before the menopause, 5% who were already postmenopausal before starting hormone replacement therapy and 13% who were taking hormone replacement therapy after a hysterectomy. The cardiovascular risk factor levels of those excluded from the analyses did not differ at the 5% level from those included, except for their lower levels of total and low density lipoprotein cholesterol (P < 0.01 and P < 0.001, respectively).

At age 53 years, the mean body mass index was 27.4 kg/m2, the mean waist circumference was 85.7 cm and the mean waist–hip ratio was 80.6%. Mean values of body mass index and waist circumference, but not waist–hip ratio, at ages 43 and 53 years, varied by menopausal status groups (Fig. 1). Perimenopausal women had higher values than either premenopausal or postmenopausal women. Women on hormone replacement therapy had lower values, similar to premenopausal women. Women who had undergone hysterectomy had the highest values (Fig. 1).

Figure 1.

Mean values of body mass index, waist circumference and waist–hip ratio at 43 years (n= 1231) and 53 years (n= 1303) by menopausal status groups.

These associations were also observed in a regression model with body mass index at age 53 as the outcome, using the restricted sample with complete data and postmenopausal women as the baseline group (Table 1, model a). Further analyses (not shown) distinguishing the prior menopausal status of hormone replacement therapy users showed that women who started hormone replacement therapy before the menopause had a significantly lower body mass index than hysterectomised women taking hormone replacement therapy.

Table 1.  Mean differences in body mass index and waist circumference at age 53 years by menopausal status groups, unadjusted and adjusted for potential confounders. Baseline group is postmenopausal women not on hormone replacement therapy (n= 1157).
 Regression coefficient (95% CI)
PremenopausalPerimenopausalPostmenopausal
Body mass index, model a−1.31 (−2.68 to 0.07)0.61 (−0.32 to 1.54)0.00
Body mass index, model b−1.02 (−2.38 to 0.34)0.72 (−0.20 to 1.64)0.00
Body mass index, model c−0.69 (−1.48 to 0.11)−0.22 (−0.76 to 0.32)0.00
Waist circumference, model a−2.94 (−6.23 to 0.34)1.05 (−1.15 to 3.25)0.00
Waist circumference, model b−1.85 (−5.09 to 1.39)1.44 (−0.73 to 3.62)0.00
Waist circumference, model c−1.04 (−3.11 to 1.02)−0.52 (−1.91 to 0.88)0.00
 Regression coefficient (95% CI)P (all groups)P (trend)+
HysterectomyHormone replacement therapy user
  • Model a unadjusted. Model b adjusted for childhood and adult social class and smoking.

  • Model c as model b and adjusted additionally for body mass index or waist circumference at 43 years.

  • +

    Trend across premenopausal, perimenopausal and postmenopausal groups.

Body mass index, model a1.34 (0.14 to 2.55)−0.90 (−1.63 to −0.17)<0.0010.41
Body mass index, model b1.06 (−0.13 to 2.25)−0.69 (−1.41 to 0.03)0.0020.68
Body mass index, model c0.07 (−0.63 to 0.77)−0.44 (−0.87 to −0.02)0.160.09
Waist circumference, model a2.50 (−0.35 to 5.35)−2.15 (−3.88 to 0.42)0.0010.36
Waist circumference, model b1.92 (−0.89 to 4.73)−1.72 (−3.44 to −0.01)0.0090.81
Waist circumference, model c1.14 (−0.65 to 2.93)−1.26 (−2.35 to −0.17)0.050.26

Adjusting for lifetime social class and smoking reduced most of the estimates (compare models a and b in Table 1). Further adjustment for body mass index at 43 years tested whether the change in body mass index between 43 and 53 years, for a given body mass index at 43 years, varied by menopausal status (Table 1, model c). The increased body mass index estimates at age 53 years that had been apparent in the previous models for perimenopausal and hysterectomised women were no longer evident, suggesting that these women were already more overweight or obese than postmenopausal women at age 43 years (as seen in Fig. 1). The estimates of lower body mass index at age 53 years for premenopausal women and hormone replacement therapy users compared with postmenopausal women were halved. This suggests that at least some of the effects were due to these women already having a lower body mass index at age 43 years (as seen in Fig. 1). Similarly, the difference in body mass index between women who started hormone replacement therapy before the menopause and hysterectomised women taking hormone replacement therapy was no longer significant after adjustment for body mass index at 43 years (not shown).

Comparable findings were observed using waist circumference at age 53 as the outcome and adjusting for the same measure at age 43 years (Table 1, models a–c). Results for waist–hip ratio were in the same direction but weaker (not shown).

There were no differences in systolic or diastolic blood pressure, either at 43 or 53 years, by menopausal status group (Fig. 2). Similarly, there were no blood pressure differences at 53 years by menopausal status group in the restricted sample before or after adjustment for confounders at age 53, and body mass index and blood pressure at 43 years (not shown).

Figure 2.

At age 53 years, the mean levels (mmol/L) of total cholesterol, high density lipoprotein cholesterol and low density lipoprotein cholesterol were 6.12, 1.83, and 3.54, respectively, and the mean level (%) of glycosolated haemoglobin was 5.58. Mean values of total and low density lipoprotein cholesterol, and HbA1c at age 53 varied by menopausal status group (Fig. 3). Total cholesterol and HbA1c increased across the natural menopause groups with premenopausal women having the lowest values, perimenopausal women having intermediate values and postmenopausal women the highest values (Fig. 3). The trend for low density lipoprotein cholesterol was similar but not significant. Women on hormone replacement therapy had low levels of total and low density lipoprotein cholesterol and HbA1c. Women who had undergone hysterectomy had the highest levels of total and low density lipoprotein cholesterol but these levels were only slightly higher than for postmenopausal women. There were no significant differences in high density lipoprotein cholesterol across the menopausal status groups (Fig. 3).

Figure 3.

Mean values of total, high density lipoprotein and low density lipoprotein cholesterol (mmol/L) (n= 1087) and HbA1c (%) (n= 1109) by menopausal status groups.

The associations between these groups and total and low density lipoprotein cholesterol and HbA1c were also observed in the restricted sample with complete data (Table 2, model a). The estimates were hardly affected by adjusting for lifetime social class, smoking behaviour and current body mass index (Table 2, model b). Further analyses (not shown) revealed that hysterectomised women taking hormone replacement therapy had higher total and high density lipoprotein cholesterol levels than other hormone replacement therapy users, and that these differences persisted after adjustment for confounders.

Table 2.  Mean differences in metabolic factors at age 53 years by menopausal status groups, unadjusted and adjusted for potential confounders. Baseline group is postmenopausal women not on hormone replacement therapy (n= 1009 for lipid measures and n= 1024 for HbA1c).
 Regression coefficient (95% CI)
PremenopausalPerimenopausalPostmenopausal
Total cholesterol, model a−0.29 (−0.59 to 0.00)−0.10 (−0.30 to 0.10)0.00
Total cholesterol, model b−0.28 (−0.57 to 0.01)−0.13 (−0.33 to 0.07)0.00
High density lipoprotein cholesterol, model a−0.02 (−0.15 to 0.11)−0.01 (−0.10 to 0.08)0.00
High density lipoprotein cholesterol, model b−0.05 (−0.17 to 0.07)−0.01 (−0.09 to 0.07)0.00
Low density lipoprotein cholesterol, model a−0.17 (−0.45 to 0.10)0.00 (−0.19 to 0.19)0.00
Low density lipoprotein cholesterol, model b−0.15 (−0.42 to 0.12)−0.02 (−0.21 to 0.17)0.00
HbA1c (%), model a−0.25 (−0.37 to −0.14)−0.10 (−0.18 to −0.03)0.00
HbA1c, model b−0.20 (−0.32 to −0.09)−0.10 (−0.18 to 0.02)0.00
 Regression coefficient (95% CI)P (all groups)P (trend)+
HysterectomyHormone replacement therapy users
  • Model a unadjusted. Model b adjusted for childhood and adult social class, smoking and body mass index.

  • +

    Trend across premenopausal, perimenopausal and postmenopausal groups.

Total cholesterol, model a0.05 (−0.21 to 0.32)−0.40 (−0.57 to −0.25)<0.0010.05
Total cholesterol, model b0.02 (−0.24 to 0.28)−0.42 (−0.58 to −0.26)<0.0010.04
High density lipoprotein cholesterol, model a−0.04 (−0.16 to 0.07)0.04 (−0.03 to 0.11)0.220.73
High density lipoprotein cholesterol, model b−0.01 (−0.12 to 0.09)0.01 (−0.05 to 0.08)0.200.45
Low density lipoprotein cholesterol, model a0.04 (−0.21 to 0.29)−0.40 (−0.55 to −0.25)<0.0010.33
Low density lipoprotein cholesterol, model b0.01 (−0.23 to 0.25)−0.40 (−0.55 to −0.25)<0.0010.35
HbA1c (%), model a−0.07 (−0.17 to 0.03)−0.21 (−0.28 to −0.15)<0.001<0.001
HbA1c, model b−0.08 (−0.18 to 0.02)−0.20 (−0.26 to −0.14)<0.001<0.001

DISCUSSION

In a representative sample of British women aged 53 years, the natural menopause transition was more strongly associated with metabolic factors than with obesity or blood pressure. The unique feature of this study is that all women were from the same birth cohort and all but 4% were measured within a seven-month period so these findings are not confounded by age or time period effects.

Most studies do not find that body weight or body mass index increase across the natural menopause transition.7,45 The Massachussetts Women's Health Study found that for a given weight at baseline, women making the transition from pre- to perimenopause were heavier than those remaining premenopausal, but the transition to postmenopause was not associated with any significant weight change.45 In our study, the higher levels of body mass index of perimenopausal women were due to their higher levels of body mass index 10 years earlier. Change in body mass index showed a weak trend across the three stages of the natural menopause transition. A greater body mass index may delay menopause rather than menopause affecting body mass index.46 While women in our study who were perimenopausal at age 53 years may have a later menopause because they were heavier than women who were already postmenopausal, this explanation does not explain why women still premenopausal at that age were relatively thin. We are currently examining age at menopause in relation to lifetime weight trajectories.

There is some evidence that progress through the natural menopause transition is associated with increases in waist–hip ratio.47 Wing et al.48 showed differences cross sectionally but not longitudinally in the Massachusetts Women's Health Study. Our findings provide no support for a greater effect of menopause on central rather than on total obesity (body mass index).49 Measures of central obesity based on whole body composition using dual energy X-ray absorptiometry may be needed to pick up any gradual and subtle changes in body fat distribution with menopause.50,51

In keeping with most previous studies,7,52 we did not find mean blood pressure levels rise across the natural menopause transition. However, some studies show greater prevalence of hypertension in postmenopausal compared with perimenopausal women53 and there is evidence that the increase in blood pressure, like the increase in blood glucose levels, becomes more apparent with time since menopause.54

The increase in mean levels of total and low density lipoprotein cholesterol across the natural menopause transition confirms the results of both cross sectional and longitudinal studies which demonstrate more adverse lipid levels in postmenopausal women.3,22 This is despite a lack of evidence that lipid levels are associated with reproductive hormones.3 Total cholesterol is thought to increase at the time of menopause although low density lipoprotein cholesterol changes are more variable.22 Changes in lipid levels have been shown to occur more rapidly during the late perimenopausal period or the first year of menopause than in subsequent postmenopausal years,54 and higher levels are established within the first five years after menopause,52 with little subsequent increase.

Mean levels of HbA1c also increased across the natural menopause transition. Higher levels in postmenopausal women have been shown in studies of French55 and Chinese50 women. Findings from studies of the natural menopause in relation to insulin and glucose levels are less consistent than those for lipoproteins. Some studies find no association between menopause and plasma glucose or insulin levels.7,52 Others find a decline in fasting insulin levels across the menopause transition56 or lower levels in postmenopausal women,57 but no relationship with glucose levels. Changes in fasting glucose may become more apparent in the years after menopause54; one study58 showed that time since menopause is a significant predictor of impaired glucose tolerance after controlling for body mass index and other cardiovascular risk factors. HbA1c provides a composite estimate of long term glucose status and increases with deterioration of glucose tolerance in non-diabetics.55 Single measures of fasting glucose and insulin or postchallenge glucose may capture aspects of glucose homeostasis, which have different associations with age and time since menopause.9 One study comparing several measures of glucose tolerance showed that HbA1c, but not fasting or postchallenge glucose, was significantly higher in postmenopausal compared with premenopausal women, after accounting for body mass index and age.50 HbA1c has also been shown to be a better predictor of cardiovascular disease in non-diabetic older women than fasting or postchallenge plasma glucose.59

In this study, women with a hysterectomy and not on hormone replacement therapy had a poor cardiovascular risk profile. In the unadjusted analyses they were heavier, had the worst total and low density lipoprotein cholesterol levels and relatively high HbA1c levels compared with women in other menopausal status groups. In the multivariable analyses, the cardiovascular risk factor levels of these women differed little from those of naturally menopausal women not on hormone replacement therapy whose profiles were more adverse in terms of total and low density lipoprotein cholesterol and HbA1c than women at an earlier stage of the menopause transition or on hormone replacement therapy. Women with bilateral oophorectomy have been shown to have higher total and low density lipoprotein cholesterol levels after surgery than women who had hysterectomy only, despite no difference in levels before surgery.13 Recent findings from the Women's Health Initiative11 showed that women with hysterectomy without ovarian conservation had an increased cardiovascular risk (in terms of the Framingham Risk Score) compared with women with intact uteri. Most of this excess risk was attributable to higher systolic blood pressure levels in these women. Whether the association with the Framingham Risk score was independent of obesity was unclear as this was not included as a confounder in the multivariable analysis. Whether the higher blood pressure levels in the sample with hysterectomy and bilateral oophorectomy than in the sample with intact uteri were due to ethnicity, higher rates of obesity or hormone replacement therapy use (which increased systolic blood pressure in the Women's Health Initiative11,55) was not established. Another study12 found higher rates of body mass index, diastolic blood pressure and hypertension in women with ovarian preservation compared with women with an intact uterus, regardless of hormone replacement therapy use. The blood pressure differences remained after controlling for body mass index and other confounders, unlike initial differences in cholesterol levels. In contrast, we found effects on lipids but no significant difference on blood pressure between women with hysterectomy and any other menopausal status group, before or after taking account of obesity or hormone replacement therapy use.

In our study, women taking hormone replacement therapy had a lower body mass index and waist circumference than all except the premenopausal group even after adjusting for social class and smoking, which concurs with a number of other studies.9,60 We showed that about half of this effect was due to these women having a lower body mass index than the other groups 10 years earlier, before the majority of them had started hormone replacement therapy. Any remaining effect may be due to these women using exercise and dietary regimes more than other women in order to limit weight gain, rather than any effects of hormone replacement therapy per se. The few longitudinal studies suggest that hormone replacement therapy use is unrelated to weight change.45,61 The proportion of women on hormone replacement therapy in these cohorts was smaller than in our cohort and they may not have had the power to detect small differences in weight changes. In the PEPI trial, mean weight change from baseline was greatest among women in the placebo group and significantly greater than weight gain among women treated with unopposed oestrogen.23

Our findings that women taking hormone replacement therapy had more favourable lipid profiles than postmenopausal women support findings from previous observational studies and clinical trials.23,25,26,62 Favourable high density lipoprotein cholesterol levels were restricted to hysterectomised women on hormone replacement therapy; the PEPI trial showed that increases in high density lipoprotein cholesterol were highest in women taking oestrogen only preparations.23 It has been shown that a marked heterogeneity in response occurs depending on route, type and dose.21,22,63 So far, clinical trials do not show that these favourable lipid profiles associated with hormone replacement therapy translate into lower incidence of cardiovascular disease, perhaps because proinflammatory effects of hormone replacement therapy adversely affect the development of cardiovascular disease.2,19 Inflammatory markers were not measured in this cohort.

Our finding that women on hormone replacement therapy had lower HbA1c levels than perimenopausal, postmenopausal or hysterectomised women adds to the growing evidence for this association.9,10,64 For example, a study of postmenopausal women in the UK9 showed that hormone replacement therapy use was also associated with lower HbA1c levels in a study of postmenopausal women independent of age, body mass index, waist–hip ratio, parity, family history, hysterectomy status, smoking history and education. The effects of hormone replacement therapy on fasting glucose and insulin or postchallenge glucose are more conflicting, even in clinical trials.23

There was little evidence that hormone replacement therapy affects blood pressure levels which supports findings from observational studies,62,65 and some clinical trials.23 In contrast, the Women's Health Initiative25 reported an increase in systolic blood pressure for those assigned to treatment rather than placebo.

The main disadvantage of this study is that cardiovascular risk factor levels were not measured for each individual at every stage of the menopause transition, although measures of body mass index, waist circumference, waist–hip ratio and blood pressure taken at age 43 years did allow assessment of change in these risk factors over a 10-year period. Our findings are based on non-fasting levels of the metabolic factors and may under-estimate the true associations with menopause and hormone replacement therapy use. Under these circumstances, values of total and high density lipoprotein cholesterol are likely to be the most reliable,66 and are acceptable for screening purposes and guiding primary prevention treatment decisions.67,68

CONCLUSIONS

In a representative sample of British women aged 53 years, the natural menopause transition was more strongly associated with an increase in levels of total and low density lipoprotein cholesterol and HbA1c than with levels of obesity or blood pressure. The stability of these differences in cardiovascular risk factor levels over time once the premenopausal and perimenopausal women reach menopause, and any subsequent benefits in terms of lower cardiovascular disease incidence remain to be seen. The lower levels of metabolic risk factors for women on hormone replacement therapy may protect against future cardiovascular disease or may be overwhelmed by other adverse, and as yet unknown, effects of hormone replacement therapy.

Acknowledgements

HK received additional funding from the Netherlands Organisation for Scientific Research.

Accepted 5 August 2004

Ancillary