Ratio of geometric means and 95% confidence intervals (CIs). LDL indicates low-density lipoprotein; HDL, high-density lipoprotein; Lp(a), lipoprotein(a); apoA-I, apolipoprotein A-I; apoB, apolipoprotein B; PAI-1, plasminogen activator inhibitor type 1; and HbA1c, glycosylated hemoglobin.
Manning PJ, Allum A, Jones S, Sutherland WHF, Williams SM. The Effect of Hormone Replacement Therapy on Cardiovascular Risk Factors in Type 2 DiabetesA Randomized Controlled Trial. Arch Intern Med. 2001;161(14):1772-1776. doi:10.1001/archinte.161.14.1772
Postmenopausal women with diabetes are at high risk for cardiovascular disease, compared with their nondiabetic counterparts. Combined continuous hormone replacement therapy (HRT) is associated with improvements in serum lipoprotein levels in nondiabetic women; however, the effect in women with diabetes has not been determined. We evaluated the effect of combined continuous HRT on lipoprotein and coagulation factor concentrations and glycemic control in postmenopausal women with type 2 diabetes mellitus.
In a randomized controlled crossover study, 61 subjects received combined continuous HRT or placebo. Each treatment phase was of 6 months' duration, with an 8-week washout phase between treatment phases.
Total cholesterol concentration decreased by 7% (95% confidence interval [CI], 4%-11%) during HRT. For low-density lipoprotein concentration, the mean decrease with HRT was 12% (95% CI, 6%-17%). Apolipoprotein B levels decreased in keeping with the reduction in low-density lipoprotein cholesterol concentrations. There were no significant changes in concentrations of high-density lipoprotein, its subfractions, or triglycerides. Lipoprotein(a) and fibrinogen concentrations were reduced by 21% (95% CI, 10%-31%) and 8% (95% CI, 2%-13%), respectively, with HRT. Fructosamine concentrations declined by 5% (95% CI, 2%-9%) during HRT.
In postmenopausal women with type 2 diabetes mellitus, combined continuous HRT has beneficial effects on lipoprotein concentrations and improves some markers of coagulation and glycemic control.
CARDIOVASCULAR disease accounts for more than 60% of the deaths among diabetic patients.1 In population studies, female patients with diabetes have 3 times the coronary heart disease mortality rate of age- and sex-matched control subjects.2 Nonfatal cardiovascular events follow a similar pattern. This excess in coronary artery disease is partly explained by abnormalities in lipoprotein metabolism. Typically the diabetic lipid profile consists of hypertriglyceridemia, raised levels of very low-density lipoprotein (VLDL), and reduced levels of high-density lipoprotein (HDL).3 In nondiabetic women, menopause has been shown to alter lipoprotein levels significantly, most frequently in the form of an increase in low-density lipoprotein (LDL) cholesterol concentrations4; however, these changes may simply reflect the effect of aging.5
Extensive and consistent observational evidence exists that estrogen use reduces the risk for coronary heart disease in nondiabetic postmenopausal women by at least 35%.6 However, a recent randomized controlled trial has suggested that this beneficial effect may not be afforded to women with established coronary artery disease.7 Unopposed estrogen decreases LDL levels by approximately 15% and increases HDL levels by a similar amount; however, the effect on HDL may be less pronounced with combined hormone replacement therapy (HRT), depending on the androgenicity of the progestin used.8 The presumed beneficial effect of HRT on the cardiovascular system may also be mediated by nonlipid mechanisms. In addition, estrogen has been shown to improve vessel wall function,9 and oral, but not transdermal, estrogen reduces plasminogen activator inhibitor type 1 (PAI-1) concentrations in nondiabetic postmenopausal women.10
Although there are numerous studies examining the effect of HRT on the concentrations of lipoproteins in nondiabetic postmenopausal women, there have been few studies on their diabetic counterparts. The studies to date have been of short duration and have used estrogen-only treatment regimens.11,12 The aim of this study was to examine the effect of a 6-month combined, continuous regimen of HRT on lipoprotein variables and markers of fibrinolysis and glycemic control in postmenopausal women with type 2 diabetes.
Female patients with a diagnosis of type 2 diabetes mellitus presenting to the diabetes clinic of the Department of Medicine, Dunedin Hospital, Dunedin, New Zealand, were eligible for entry into the study if they were postmenopausal (defined as absence of menstrual periods for >2 years) and had no contraindication to HRT (a history of estrogen-dependent cancer, undiagnosed vaginal bleeding, uncontrolled hypertension, or severe liver dysfunction). Subjects were excluded if they had poorly controlled diabetes (glycosylated hemoglobin [HbA1c] level, >10%), a concomitant significant medical disorder, or a myocardial infarction or unstable angina within the past 3 months, or if they met the New Zealand criteria for the use of statin therapy at the time of recruitment (with no previous coronary artery disease, total cholesterol level of >347 mg/dL [>9 mmol/L]; with established coronary artery disease, total cholesterol level of >270 mg/dL [>7 mmol/L]).
One investigator (P.J.M.) obtained a detailed medical history and performed an examination of each subject. Weight was measured using electronic scales with the subject in light clothing. Blood pressure was measured in the right arm in the seated position. Blood samples were drawn after an overnight fast on 2 successive days for measurement of concentrations of total cholesterol, HDL cholesterol (including subfraction analysis), LDL cholesterol, total triglycerides, lipoprotein(a) (Lp[a]), and apolipoproteins A-I (apoA-I) and B (apoB). The mean of the results for the 2 successive days was used as the value for that time in the study. Blood samples were also drawn for measures of glycemic control (levels of fasting glucose, HbA1c, and fructosamine) and concentrations of PAI-1 and fibrinogen.
Subjects who met the criteria for entry into the study entered an 8-week run-in phase, during which time there was no alteration to their usual diabetes management. Subjects were requested to continue with the dietary instructions they had previously received as part of their care from the diabetes service. At 8 weeks, subjects were randomized in a double-blind fashion to receive HRT or placebo by means of a random number-generating process, and were allocated the appropriate treatment by 1 investigator (S.J.) who did not interact with the subjects at any other time during the study. The HRT preparation consisted of conjugated equine estrogen, 0.625 mg (Premarin; Wyeth-Ayerst, Philadelphia, Pa), and medroxyprogesterone acetate, 2.5 mg (Provera; Upjohn, Kalamazoo, Mich). The estrogen and progestin medications were combined in a single capsule that was identical in appearance to the placebo capsule. To minimize acute adverse effects of HRT, study medication therapy was titrated upward for 4 weeks. At the end of this time, subjects were receiving 1 capsule per day of placebo or HRT.
Subjects were seen every 3 months for measurement of weight and blood pressure, inquiry of adverse events, blood sampling, and analysis of compliance by capsule counting. After 6 months, the study medication was withdrawn, and subjects entered an 8-week washout phase before changing to the other treatment arm. Subjects were then followed up for an additional 6 months until the completion of the study. Patients were withdrawn from the study if an adverse reaction to study medication developed or if they experienced a serious concurrent illness contraindicating HRT or met the current national criteria for commencing therapy to lower lipid levels during the study. The study was approved by our local ethics committee.
Sample size was calculated from the expected baseline lipoprotein concentrations of the study group and the expected change in LDL cholesterol levels seen with HRT in nondiabetic postmenopausal women.8 Because we used a crossover design, a sample of 48 subjects had a 90% power to detect a significant difference at the level of P = .05. To allow for a 20% dropout rate, we increased the sample size to 60.
Venous blood was collected into EDTA-treated or plain tubes. Plasma and serum were immediately separated by means of low-speed centrifugation at 4°C. Plasma VLDL was separated by ultracentrifuging EDTA-treated plasma according to the protocol of the Lipid Research Clinics Program.13 Concentration of HDL cholesterol was measured in the supernatant after precipitating apoB-containing lipoproteins with dextran sulfate sodium and magnesium chloride.14 Concentration of HDL3 cholesterol was measured in the supernatant after precipitation of lipoproteins using polyethylene glycol.15 Cholesterol and triglyceride levels were measured in plasma fractions by using commercial enzymatic kits (Boehringer Mannheim, Mannheim, Germany). Plasma LDL cholesterol concentration was calculated by subtracting HDL cholesterol concentration from the cholesterol concentration in the density (d) >1.006 g/mL plasma fraction. Serum apoA-I and apoB levels were measured by means of immunoturbidimetry.16 Plasma Lp(a) level was measured using a 2-site radioimmunoassay kit (Pharmacia, Uppsala, Sweden).
Fibrinogen level was measured using a derived fibrinogen method using a commercially available reagent (Thomborel S; Behring Diagnostics, Kanata, Ontario).17
The PAI-1 antigen assay was performed using a commercially available kit (Thrombonostika PAI-1; Organon Teknika, Boxtel, the Netherlands). Blood glucose level was determined using a commercial glucose oxidase method (Boehringer Mannheim) and HbA1c level was determined using a method based on the turbidimetric inhibition immunoassay for hemolyzed whole blood (Tinaquant; Boehringer Mannheim). Fructosamine level was measured according to the method of Johnson et al.18
The laboratory measurements were logarithm transformed before the data were analyzed. Geometric means and ranges are presented for each subject at the time of randomization. The data collected at the end of each treatment period were used to compare HRT with placebo. They were analyzed as a crossover trial using methods described by Senn19 with a factor for period included in the model. The results are presented as ratios of geometric means with 95% confidence intervals (CIs). There was no evidence of a carryover effect.
Of the 70 patients undergoing screening for the study, 61 (87%) met the entry criteria and agreed to participate; 9 (13%) had preexisting cardiovascular disease. Subjects were randomized to active treatment (n = 29) or placebo (n = 32) for the first 6 months of the study. Nine subjects withdrew while receiving HRT (intolerant to HRT [n = 4], cardiovascular event [n = 1], personal reasons [n = 2], and commenced therapy to lower lipid levels [n = 2]), and 2 withdrew while receiving placebo (cancer of the bowel [n = 1] and cerebrovascular event [n = 1]). Two subjects were withdrawn during the 8-week washout phase (cardiovascular event [n = 1] and peripheral vascular event [n = 1]), both of whom had received placebo during the first 6-month treatment phase. Forty-eight subjects completed the entire study. All subjects who continued in the study were compliant with study medication (defined as a tablet count of >80%).
The medians and interquartile ranges for baseline data for the study sample are presented in Table 1. The baseline characteristics of those who did not complete the study were compared with those of subjects who did using the Student t test. No statistically significant differences were found.
After 6 months, a significant decrease in total cholesterol level was found during treatment with HRT compared with placebo (Table 2). This reduction in total cholesterol level (ratio of geometric means, 7%) was due to a significant reduction in LDL cholesterol concentrations (12%) with HRT treatment (Figure 1). In keeping with the reduction in LDL cholesterol level, apoB concentrations decreased with HRT treatment. Serum apoA-I concentrations increased significantly, and there was a nonsignificant trend toward an increase in levels of HDL cholesterol and HDL subfractions during treatment with HRT. Total triglyceride, VLDL triglyceride, and total VLDL levels did not change significantly during HRT treatment compared with placebo. Lipoprotein(a) concentrations decreased significantly during treatment with HRT.
Significant reductions were noted in plasma fibrinogen levels with HRT treatment. Although a trend toward a reduction in plasma PAI-1 concentrations was noted, this was not significant.
Fasting plasma glucose and HbA1c levels were not significantly different between treatment periods; however, fructosamine concentrations were lower during treatment with HRT. There were no significant changes in weight or body mass index (calculated as the weight in kilograms divided by square of height in meters) throughout the study. With HRT, there was a nonsignificant reduction of 2.84 mm Hg in systolic blood pressure (95% CI, −3.60 to 9.28 mm Hg) and an increase in diastolic blood pressure of 0.74 mm Hg (95% CI, −3.50 to 2.02 mm Hg). Of the 62 women, 20 had undergone hysterectomies. The effect of treatment was not significantly different for those with and without hysterectomies for all the variables in Table 2.
To our knowledge, this is the first study to describe the effect of combined continuous HRT on lipid and hemostatic risk factors for cardiovascular disease in postmenopausal women with type 2 diabetes mellitus. In addition, to our knowledge, it is the only trial that reports the effect of HRT for a duration of greater than 3 months on cardiovascular risk factors and glycemic control. Our study shows that the daily administration of 0.625 mg of conjugated estrogens in combination with 2.5 mg of medroxyprogesterone acetate reduces levels of total and LDL cholesterol, Lp(a), fibrinogen, and fructosamine and increases plasma apoA-I levels in postmenopausal diabetic women.
Estrogen has been shown to directly increase hepatic LDL uptake via up-regulation of LDL receptors.20 The 12% reduction in LDL cholesterol levels in the present study is comparable to that seen in other studies in women with diabetes using estrogen-only regimens, indicating that medroxyprogesterone does not attenuate the ability of estrogen to lower LDL levels in postmenopausal diabetic women.11,12 This degree of reduction in LDL cholesterol level is also similar to that seen in nondiabetic women using the same preparation of HRT.8 The reduction in concentration of apoB, a structural protein for LDL, simply reflects the reduction in circulating LDL concentration with HRT.
Unlike studies using estrogen-only regimens, our study did not show an increase in levels of HDL cholesterol or its subfractions, although an increasing trend was evident. This finding probably reflects the addition of medroxyprogesterone to the estrogen therapy. In the Postmenopausal Estrogen/Progestin Interventions Trial, in which HRT was given to nondiabetic women, the addition of medroxyprogesterone diminished the increase in HDL levels seen with estrogen.8 Progestins increase the activity of hepatic lipases, increasing the metabolism of HDL cholesterol, and the extent of this effect is related to the androgenic properties of the progestin used.20 Cross-sectional studies in nondiabetic women, however, have shown no effect from the addition of a progestin on HDL concentrations.21 The present study was primarily powered to detect a change in LDL cholesterol levels in diabetic women using HRT, and the lack of an increase in HDL levels may simply reflect an inadequate sample size. The increase in apoA-I levels seen in our study supports the belief that HRT has a favorable effect on HDL concentrations. Apolipoprotein A-I is located on the surface of HDL and plays a central role in the transfer of free cholesterol from tissues and other lipoproteins.20 Apolipoprotein A-I also appears to protect lipoproteins from potentially atherogenic oxidative damage.
In nondiabetic women, oral HRT increases plasma triglyceride levels primarily by increasing the production of large VLDL. In diabetic women, however, combined HRT in the present study and estrogen-only therapy in 2 previous studies11,12 did not alter plasma triglyceride concentrations. The mechanism for this lack of effect on triglyceride levels is unclear; however, it may be related to an improvement in insulin sensitivity. This would lead to a reduction in the flux of VLDL to the liver, which is the predominant cause of raised triglyceride levels in diabetic patients. In a cross-sectional study of 694 women with diabetes, of whom 10% were current users of HRT, a trend toward increasing triglyceride concentrations in those women currently receiving estrogen was noted, although this effect was less marked in those using combined estrogen and progestin therapy.22 However, there have been case reports of significant hypertriglyceridemia developing in diabetic patients when HRT was commenced.23 It is therefore reassuring that no significant alteration in plasma triglyceride levels was noted in the present study.
Hemostatic factors have an important role in determining cardiovascular risk. Our results show that combined HRT reduces fibrinogen and Lp(a) concentrations in postmenopausal diabetic women. Fibrinogen, PAI-1, and Lp(a) concentrations have all been shown to be independent risk factors for cardiovascular disease. Plasminogen activator inhibitor type 1 and Lp(a) exert this effect by interfering with the process of fibrinolysis. Lipoprotein(a) also binds to macrophages in the vessel wall, promoting the formation of foam cells. The level of Lp(a) is increased in patients with type 2 diabetes, adding to their increased cardiovascular risk. Hormone replacement therapy decreases Lp(a) levels by 17% to 23% in nondiabetic women, and this effect is not altered by the addition of a progestin.8 Our study demonstrated a 21% lowering of Lp(a) levels when women were receiving HRT. To our knowledge, this is the first report of this effect of HRT on Lp(a) levels in women with type 2 diabetes. Our study revealed a nonsignificant trend in reduction of PAI-1 levels. These levels have been shown to be significantly reduced by estrogen-only or combined HRT in nondiabetic women10 and with estrogen-only HRT in women with diabetes.12 The addition of a progestin may account for the failure to show a significant reduction in PAI-1 levels in our study. The fibrinogen levels increase after menopause, perhaps due to the effects of age, and decline with estrogen use in nondiabetic women. Our results indicate that the same holds true for women with diabetes. This effect in women with diabetes has not previously been described. The changes in hemostatic factor concentrations measured in our study would appear to be favorable and therefore would not explain an increase in cardiovascular events described during the first year of combined HRT in the Heart and Estrogen/Progestin Replacement Study (HERS) trial.
Our study showed a significant reduction in fructosamine levels and a trend to a reduction in fasting glucose and HbA1c levels in those women receiving HRT. These results suggest that HRT may have a beneficial effect on glycemic control in postmenopausal women with type 2 diabetes. Improvements in glucose homeostasis have been documented in women receiving short-term estrogen-only HRT.11,12 Presumably, this effect is mediated by an increase in insulin sensitivity, which has been documented by euglycemic clamp methods.11
Of our original 61 subjects, 13 (21%) did not complete the study. Most of these (n = 9) dropped out while receiving HRT, and the reasons were primarily due to intolerance to medication, which was due to either breast tenderness or nausea. This dropout rate is similar to that noted in other trials using HRT for a prolonged period and was allowed for when estimating sample size before recruitment. This trial was completed before the publication of the HERS trial7 and was not powered to determine the effect on cardiovascular events; however, no excess in cardiovascular events was noted during the HRT phase. Although the analysis of this study was not intention to treat, the baseline characteristics of those who dropped out did not differ from those of the group as a whole. Compliance, as measured by tablet counting, was greater than 80% in those subjects continuing in the study, and blinding was not broken on any subject throughout the study.
Patients with diabetes are at high risk for cardiovascular disease. There is compelling evidence that HRT exerts a beneficial effect on a number of cardiovascular risk factors in nondiabetic women, and this study shows that these beneficial changes also are observed in women with diabetes. However, given the results of the HERS trial,7 there is insufficient evidence at present to suggest that HRT should be used for the prevention of coronary heart disease. This study demonstrates the need for trials of HRT in diabetic women for primary prevention of coronary heart disease using cardiovascular events as the end point.
Accepted for publication January 18, 2001.
This study was supported by grants 96/331 (Dr Manning) and 99/500 (Ms Williams) from the Health Research Council of New Zealand, Auckland.
The authors are very grateful to the patients who participated in this study. We also thank J. I. Mann, DSc, for helpful comments in the preparation of this manuscript.
Corresponding author and reprints: Patrick J. Manning, MBChB, MMedSc, FRACP, Department of Medicine, Dunedin Hospital, 201 Great King St, Dunedin, New Zealand (e-mail: PatrickManning@clear.net.nz).