Lipoprotein Particle Concentrations May Explain the Absence of Coronary Protection in the Women’s Health Initiative Hormone Trials
Objective— The Women’s Health Initiative randomized hormone trials unexpectedly demonstrated an increase in early coronary events. In an effort to explain this finding, we examined lipoprotein particle concentrations and their interactions with hormone therapy in a case–control substudy.
Methods and Results— We randomized 16 608 postmenopausal women with intact uterus to conjugated estrogens 0.625 mg with medroxyprogesterone acetate 2.5 mg daily or to placebo, and 10 739 women with prior hysterectomy to conjugated estrogens 0.625 mg daily or placebo, and measured lipoprotein subclasses by nuclear magnetic resonance spectroscopy at baseline and year 1 in 354 women with early coronary events and matched controls. Postmenopausal hormone therapy raised high-density lipoprotein cholesterol and particle concentration and reduced low-density lipoprotein cholesterol (LDL-C; all P<0.001 versus placebo). In contrast, neither unopposed estrogen nor estrogen with progestin lowered low-density lipoprotein particle concentration (LDL-P).
Conclusions— Postmenopausal hormone therapy–induced reductions in LDL-C were not paralleled by favorable effects on LDL-P. This finding may account for the absence of coronary protection conferred by estrogen in the randomized hormone trials.
The Women’s Health Initiative randomized, controlled hormone trials unexpectedly demonstrated increased stroke risk with postmenopausal hormone therapy, and no coronary protection.1,2 Among postmenopausal women with intact uterus, the primary efficacy outcome of myocardial infarction/coronary death (CHD) was more frequent among women randomized to active estrogen with progestin (E+P), with hazard ratio 1.81 in year 1, 1.34 in year 2, 1.27 in year 3, and 1.25 in year 4.3 Among women with prior hysterectomy, unopposed conjugated estrogens (CEE) neither increased nor decreased CHD overall, with hazard ratio 1.11 in year 1, 1.20 in year 2, 0.89 in year 3 and 0.79 in year 4.4
See accompanying article on page 1582
Subgroup analyses failed to identify demographic or clinical characteristics which predicted who might safely take estrogen for vasomotor symptoms.3,4 Lipoprotein particle concentrations predict coronary risk in older men and women5,6 and may reflect hormone effects more accurately than lipoproteins measured by conventional enzymatic methods.7 We measured lipoprotein particle concentrations by nuclear magnetic resonance (NMR) spectroscopy in a case–control substudy and evaluated their association with CHD.
The Women’s Health Initiative included 2 randomized controlled trials of postmenopausal hormone therapy: the Estrogen Plus Progestin trial in women with intact uterus and the Estrogen Alone trial in women with prior hysterectomy. Eligibility criteria, recruitment methods, baseline data collection, randomization, and follow-up procedures for the trials have been previously reported.8–10 The protocol and consent forms were approved by the institutional review boards of participating institutions, and written informed consent was obtained from all participants. The Estrogen Plus Progestin trial randomized 16 608 women, aged 50 to 79 years with intact uterus, to 0.625 mg of oral conjugated equine estrogen plus 2.5 mg of medroxyprogesterone acetate (Prempro, Wyeth) daily or to placebo. The Estrogen Alone trial randomized 10 739 women with prior hysterectomy to CEE 0.625 mg daily (Premarin, Wyeth) or to placebo.
Participants reported emergency room visits, overnight hospital stays, and outpatient coronary revascularization procedures semiannually. Medical records for all overnight hospitalizations and outpatient coronary revascularization procedures were scrutinized for potential outcomes of interest. Outcomes were classified by central physician adjudicators on the basis of medical record review. Myocardial infarction was defined using an algorithm which included symptoms, electrocardiographic findings, and cardiac enzymes.10
In the nested case–control study, for each of the hormone trials, cases of CHD occurring during the first 4 years after randomization were matched to controls on age, randomization date, and prevalent coronary disease. Using fasting blood samples collected at baseline and 1 year, lipid profiles were performed at Medical Research Laboratories (Highland Heights, Ky). Lipoprotein particle concentrations were assessed by NMR spectroscopy (Liposcience, Raleigh, NC).11
Baseline characteristics of cases and controls were compared by t test and χ2 (Table 1). Baseline biomarker values were log-transformed; differences from baseline to year 1 were analyzed on the original scale. Baseline and on-treatment lipoprotein levels are presented as medians and interquartile ranges; the association with CHD was compared in logistic regression models adjusted for treatment assignment (Table 2). Change in lipoprotein levels was compared between treatment groups by F test (Table 3). We evaluated the log-linearity relationship between lipoprotein levels and CHD in generalized additive models12 adjusting for age, body mass index, current smoking, treated diabetes, and hypertension. For lipoprotein variables found to be nonlinear, a quadratic function was added to subsequent conditional logistic regression models evaluating the association between lipoproteins and CHD risk (Table 4). Logistic models were adjusted for treatment assignment, age, current smoking at baseline, diabetes, alcohol use, and hypertension. Models evaluating the change of biomarker from baseline to year 1 included the (log-transformed) baseline biomarker. Interactions between treatment assignment and lipoproteins were evaluated by adding an interaction term to logistic regression models. From the 48 models, we would expect 2 probability values <0.05 by chance. Analyses were carried out using SAS System for Windows v 9.1. The authors had full access to the data and take responsibility for its integrity.
Baseline characteristics for CHD cases and controls are shown separately for the 2 hormone trials (Table 1). CHD cases and controls were matched on age and prevalent cardiovascular disease at baseline. Smoking (P<0.001 for E+P and 0.003 for CEE) and hypertension (P=0.003 for E+P and 0.02 for CEE) were more prevalent, whereas Calcium/Vitamin D trial participation was less frequent (P=0.004 for E+P and 0.04 for CEE) among women with on-trial CHD events. CHD cases had greater waist circumference (P<0.04 for E+P and 0.003 for CEE).
Median levels and interquartile ranges for lipoproteins are shown separately for the 2 trials by case–control status (Table 2). After adjustment for treatment arm, high-density lipoprotein cholesterol (HDL-C) and particle concentration (HDL-P) were negatively associated with CHD (both P<0.001), whereas low-density lipoprotein cholesterol (LDL-C) and low-density lipoprotein particle concentration (LDL-P; both P<0.001), very low–density lipoprotein particle concentration (VLDL-P; P=0.005), and triglycerides (P<0.001) were positively associated with CHD. On-treatment, the associations with HDL-C and HDL-P remained strong. On-treatment LDL-C was not associated with CHD, whereas the association of LDL-P with CHD persisted. For none of the lipoproteins was change from baseline to year 1 significantly associated with CHD after adjustment for treatment arm (data not shown).
Changes in lipoproteins are shown by treatment group assignment (Table 3). For each trial, baseline lipoprotein levels were similar in the active treatment and placebo groups (data not shown). From baseline to year 1, HDL-C and HDL-P rose (P<0.001 for both) among women assigned to active E+P or CEE. LDL-C fell 16% among women assigned to active E+P and 12% among those assigned to active CEE (P<0.001 for both). In contrast, neither LDL-P nor LDL-P+VLDL-P were reduced by E+P or CEE.
Associations between lipoproteins and CHD after multivariable adjustment are shown separately for the 2 trials (Table 4). In the E+P trial at baseline, HDL-C was an independent negative predictor of CHD (OR 0.69, 95% CI 0.52 to 0.93, P=0.01). Positive predictors included LDL-P (OR 1.40, 95% CI 1.08 to 1.83, P=0.01), VLDL-P (OR 1.49, 95% CI 1.11 to 2.00, P=0.009), and triglycerides (1.39, 95% CI 1.08 to 1.79, P=0.01). At year 1, the associations of CHD with HDL-C, VLDL-P, and triglycerides persisted.
In the CEE trial at baseline, HDL-P was negatively associated with CHD (OR 0.64, 95% CI 0.44 to 0.93, P=0.02). LDL-C was an independent positive predictor of CHD (OR 1.61, 95% CI 1.13 to 2.30, P=0.008), but LDL-P and VLDL-P were not. At year 1, none of the lipoprotein levels were independently associated with CHD.
The interaction between postmenopausal hormone therapy and lipoprotein levels with regard to CHD was evaluated by formal interaction testing across the 4 treatment groups (data not shown). Interactions with randomized treatment assignment were identified for baseline HDL-P (P for interaction=0.05) and LDL-C (P for interaction=0.02). No significant interactions were identified between treatment assignment and lipoprotein levels at year 1 or with change in lipoprotein levels from baseline to year 1.
Postmenopausal hormone therapy significantly reduced LDL-C and raised HDL-C and HDL-P, but did not alter LDL-P. In the E+P trial, on-treatment HDL-C was negatively associated, whereas triglycerides and VLDL-P were positively associated with CHD. In the E Alone trial, no on-treatment lipoprotein was an independent predictor of CHD.
Strengths of this analysis include the randomized, controlled design, large number of women in the randomized trials, and long duration of follow-up, diversity of the cohort, wide range of covariates available for analysis, and prospective collection of hard clinical end points. The predominant limitation is the number of CHD cases, which limited our ability to detect subtle associations, to evaluate combinations of lipoproteins and interactions with treatment assignment. No future randomized hormone trial, however, is likely to have a larger number of available cases. For example, the Women’s International Study of long Duration estrogen after Menopause (WISDOM) was designed as a 22 300 woman trial with 15-year follow-up, but was terminated when the Women’s Health Initiative results were announced after accruing only 7 CHD events.13
At baseline, participation in the Calcium/Vitamin D trial was less common among CHD cases than controls (50% versus 64%, P=0.004 in the E+P trial and 48% versus 60%, P=0.04 in the E Alone trial). This may be attributable to enrollment bias, as women with concurrent health issues may have been reluctant to join the Calcium/Vitamin D trial when invited at their first annual visit. In the randomized trial of calcium/vitamin D supplementation, we found no treatment effect on CHD.14
The goal of this case–control study was to explore mechanisms underlying the early CHD risk with postmenopausal hormones. Lipid modulating agents which raise HDL-C15 or lower LDL-C,16 and have parallel effects on lipoprotein particle concentrations,17,18 have reduced CHD events in randomized trials. In the Women’s Health Initiative randomized trials, postmenopausal hormone therapy raised HDL-C and lowered LDL-C, but did not reduce CHD.3,4 This may be attributable to the fact that postmenopausal hormone therapy-induced changes in lipoprotein cholesterol concentrations did not consistently reflect changes in particle concentrations. For example, E+P reduced median LDL-C by 16% whereas LDL-P increased by 3%. Similarly, unopposed CEE reduced LDL-C by 12%, while increasing LDL-P by 5%. This dissociation is consistent with prior reports in nonrandomized cohorts.7
While LDL-C and LDL-P at baseline independently predicted CHD in our trials, we would expect estrogen’s effects on CHD to be reflected in on-treatment levels. The association with LDL-C disappeared altogether on-treatment, and that with LDL-P was weaker. A possible explanation for this observation is that proinflammatory and thrombotic effects of postmenopausal hormone therapy19 supersede the combination of favorable and unfavorable lipoprotein effects. Supporting this possibility is the apparent potentiation by progestins of the interleukin (IL)-6–mediated rise in C-reactive protein induced by estrogen.20 The rise in CHD risk after initiating estrogen with progestin is prompt, with hazard ratio 1.81 in year 1 of the E+P trial3 and relative hazard of 2.3 in the first 4 months of treatment in the Heart & Estrogen/progestin Replacement Study,21 consistent with a fairly rapid mechanism.
Characteristics of women with incident CHD in our case–control study reflected typical predictors such as cigarette smoking and waist circumference. However, even the controls demonstrated some CHD risk characteristics such as body mass index above 25 kg/m2, C-reactive protein above 3 mg/L, and LDL-P above 1500 nmol/L. These observations underscore the intrinsic limitations of estimating CHD event rates from risk factors, because we know the overall CHD event rates were low, 0.56%/yr for the E Alone trial and 0.33%/yr for the E+P trial.3,4 Ongoing genomic and proteomic studies from Women’s Health Initiative samples are seeking novel biomarkers to improve current prediction models.
Because vasomotor symptoms remain a significant clinical issue for menopausal women, we carried out the interaction analyses to identify predictors of lower or higher CHD risk with hormones. Individuals perceived at higher risk could then be advised against taking exogenous estrogen for menopausal symptoms, whereas women lacking these characteristics could be reassured that their absolute and relative CHD risk with estrogen was low. In view of the possibility that the interactions we detected were attributable to chance, we cannot be confident that they distinguish women who can more safely take hormones.
This case–control study provides a plausible explanation for the lack of coronary protection with postmenopausal hormone therapy. The observed dissociation between hormone-induced changes in LDL-C and LDL-P supports the view that measures of atherogenic particles, either by NMR spectroscopy or apolipoprotein B levels, may be preferable to LDL-C as surrogate markers for CHD risk.
Sources of Funding
The Women’s Health Initiative program is funded by the National Heart, Lung, and Blood Institute, National Institutes of Health, US Department of Health and Human Services through contracts N01WH22110, 24152, 32100-2, 32105-6, 32108-9, 32111-13, 32115, 32118-32119, 32122, 42107-26, 42129-32, and 44221.
Dr Hsia is employed by and owns stock in AstraZeneca. Dr Otvos is employed by and owns stock in Liposcience. Dr Robinson has research grants from Abbott, Bristol–Myers Squibb, Merck/Schering-Plough, Merck, AstraZeneca, Hoffman LaRoche, Takeda, and Pfizer, is a member of the Merck/Schering-Plough speakers’ bureau and advisory boards for AstraZeneca and Merck/Schering-Plough.
Original received May 15, 2008; final version accepted June 22, 2008.
Kuller L, Arnold A, Tracy R, Otvos J, Burke G, Psaty B, Siscovick D, Freedman DS, Kronmal R. Nuclear magnetic resonance spectroscopy of lipoproteins and risk of coronary heart disease in the cardiovascular health study. Arterioscler Thromb Vasc Biol. 2002; 22: 1175–1180.
Blake GJ, Otvos JD, Rifai N, Ridker PM. Low-density lipoprotein particle concentration and size as determined by nuclear magnetic resonance spectroscopy as predictors of cardiovascular disease in women. Circulation. 2002; 106: 1930–1937.
Otvos JD, Jeyarajah EJ, Bennett DW, Krauss RM. Development of a proton nuclear magnetic resonance spectroscopic method for determining plasma lipoprotein concentrations and subspecies distributions from a single, rapid measurement. Clin Chem. 1992; 38: 1632–1638.
Hastie TJ, Tibshirani RJ. Generalized Additive Models. New York: Chapman and Hall, 1990.
Vickers MR, MacLennan AH, Lawton B, Ford D, Martin J, Meredith SK, DeStavola BL, Rose S, Dowell A, Wilkes HC, Darbyshire JH, Meade TW, for the WISDOM team. Main morbidities recorded in the women’s international study of long duration oestrogen after menopause (WISDOM): a randomized controlled trial of hormone replacement therapy in postmenopausal women. BMJ. 2007; 335: 239.
Hsia J, Heiss G, Ren H, Allison M, Dolan NC, Greenland P, Heckbert SR, Johnson KC, Manson JE, Sidney S, Trevisan M, for the WHI Investigators. Calcium/vitamin D supplementation and cardiovascular events. Circulation. 2007; 115: 846–854.
Robins SJ, Collins D, Wittes JT, Papdemetriou V, Deedwania PC, Schaefer EJ, McNamara JR, Kashyap ML, Hershman JM, Wexler LF, Rubins HB. Relation of gemfibrozil treatment and lipid levels with major coronary events: VA-HIT: a randomized controlled trial. JAMA. 2001; 285: 1585–1591.
Colhoun HM, Betteridge DJ, Durrington PN, Hitman GA, Neil HAW, Livingstone SJ, Thomason MJ, Mackness MI, Charlton-Menys V, Fuller JH. Primary prevention of cardiovascular disease with atorvastatin in type 2 diabetes in the Collaborative Atorvastatin Diabetes Study (CARDS): multicentre randomized placebo-controlled trial. Lancet. 2004; 364: 685–696.
Otvos JD, Collins D, Freedman DS, Shalaurova I, Schaefer EJ, McNamara JR, Bloomsfield HE, Robins SJ. Low-density lipoprotein and high-density lipoprotein particle subclasses predict coronary events and are favorably changed by gemfibrozil therapy in the Veterans Affairs High-Density Lipoprotein Intervention Trial. Circulation. 2006; 113: 1556–1563.
Soedamah-Muthu SS, Colhoun HM, Thomason MJ, Betteridge DJ, Durrington PN, Hitman GA, Fuller JH, Julier K, Mackness MI, Neil HAW. The effect of atorvastatin on serum lipids, lipoproteins and NMR spectroscopy defined lipoprotein subclasses in type 2 diabetic pateints with ischaemic heart disease. Atherosclerosis. 2003; 167: 243–255.
Hulley S, Grady D, Bush T, Furberg C, Herrington D, Riggs B, Vittinghoff E; Heart and Estrogen/progestin Replacement Study (HERS) Research Group. Randomized Trial of Estrogen Plus Progestin for Secondary Prevention of Coronary Heart Disease in Postmenopausal Women. JAMA. 1998; 280: 605–613.