Postmenopausal Estrogens and Risk of Myocardial Infarction in Diabetic Women NICHOLAS L. SMITH, PHD KATHERINE M. NEWTON, PHD BRUCE M.

Transcription

1 E p i d e m i o I o g y / H e a 11 h S e r v i c e s / P s y c h o s o c i a I R e s e a r c h N A L A R T I C L E Postmenopausal Estrogens and Risk of Myocardial Infarction in Diabetic Women ROBERT C. KAPLAN, MS SUSAN R. HECKBERT, MD, PHD NOEL S. WEISS, MD, DRPH PATRICIA W. WAHL, PHD NICHOLAS L. SMITH, PHD KATHERINE M. NEWTON, PHD BRUCE M. PSATY, MD, PHD OBJECTIVE The effect of hormone replacement therapy on the risk of myocardial infarction in diabetic women has not been well studied. We conducted a case-control study of postmenopausal estrogen use and risk of incident myocardial infarction (Ml) in pharmacologically treated diabetic women enrolled at Group Health Cooperative of Puget Sound, a large health maintenance organization in the state of Washington. RESEARCH DESIGN AND METHODS Case subjects (n = 212) were all postmenopausal women with treated diabetes who sustained an incident fatal or nonfatal MI between July 1986 and December Control subjects (n = 122) were treated diabetic women drawn from a stratified random sample of postmenopausal women without prior MI. Computerized pharmacy data and medical records were used to measure use of estrogens. Cardiovascular risk factors recorded from medical records, computerized pharmacy and laboratory data, and telephone interviews were used as adjustment variables. RESULTS In this study, 8.5% of case and 13.9% of control subjects were current users of estrogens. The relative risk (RR) of MI for current estrogen users was 0.51 (95% CI ) relative to never users, adjusted for age, study year, weight, angina, and duration of treated diabetes. Among current estrogen users, risk of MI tended to decline with each additional year of estrogen use (adjusted RR = 0.78, 95% CI ). Of those studied, 45.3% of case and 37.7% of control subjects were past users of estrogens (adjusted RR = 1.22,95% CI ). CONCLUSIONS This study suggests that use of postmenopausal estrogens does not increase risk of MI in diabetic women and that sustained use may be of benefit. Evidence from observational studies suggests that estrogen replacement therapy diminishes the increased risk of cardiovascular disease associated with menopause, and a smaller body of research suggests that the addition of progestin does not attenuate this benefit (1-3). Some commentators consider hormone replacement therapy to be of particular value for women at high risk for heart disease (4-7), but for many risk groups, this claim has not been well studied. In women of postmenopausal age, diabetes has a prevalence of 4-7% (8) and is associated with a threefold or higher increased risk of coronary heart disease (9-12). Yet, there have been few studies of the risks and benefits of hormone replacement therapy in diabetic women. Estrogens and progestins have known and hypothesized effects that may be of particular importance in diabetic patients. In a randomized trial conducted among postmenopausal diabetic women, glycated hemoglobin levels decreased and suppres- From the Cardiovascular Health Research Unit (R.C.K., S.R.H., N.L.S., B.M.P); the Departments of Epidemiology (R.C.K., S.R.H., N.S.W, N.L.S., K.M.N., B.M.R), Biostatistics (PWW), Medicine (B.M.E), and Health Services (B.M.P), University of Washington; the Fred Hutchinson Cancer Research Center (N.S.W); and the Center for Health Studies (K.M.N.), Group Health Cooperative of Puget Sound, Seattle, Washington. Address correspondence and reprint requests to Robert C. Kaplan, MS, Cardiovascular Health Research Unit, Suite 1360, 1730 Minor Ave., Seattle, WA rkaplan@u.washington.edu. Received for publication 31 December 1997 and accepted in revised form 1 April K.M.N. has received an honorarium for presenting a paper at an event sponsored by the American Heart Association and supported by Wyeth-Ayerst Laboratories. Abbreviations: CEE, conjugated equine estrogen; GHC, Group Health Cooperative of Puget Sound; HMO, health maintenance organization; Ml, myocardial infarction; RR, relative risk. sion of hepatic glucose production by insulin increased relative to placebo after 6 weeks of 17-(3 estradiol therapy (13). In the Postmenopausal Estrogen/Progestin Interventions Trial, conjugated equine estrogen (CEE) or combined CEE-progestin therapy given over 3 years to healthy women had no effect on 2-h or fasting insulin levels compared with placebo, while 2-h glucose increased significantly and fasting glucose decreased significantly in active treatment groups relative to placebo (14). Estrogen use causes an increase in serum triglycerides that may exacerbate diabetic hypertriglyceridemia (14-16). The beneficial increase in HDL cholesterol associated with hormone replacement therapy appears to be smaller in diabetic women than in nondiabetic women (15). Despite previous findings that hormone replacement therapy is not associated with weight gain in healthy women (17), a recent trial of 17- estradiol in postmenopausal diabetic women showed an increase in BMI, but no change in waistto-hip ratio, relative to placebo (13). We present results from a case-control study of hormone replacement therapy and risk of incident myocardial infarction (MI) in pharmacologically treated diabetic women. This study uses a subset of subjects from previously published studies (2,18). In addition, case and control subjects were identified during an additional year for the analyses presented here. RESEARCH DESIGN AND METHODS The setting for this study was Group Health Cooperative of Puget Sound (GHC), a 500,000-member health maintenance organization (HMO) in western Washington State. Subjects Case subjects were postmenopausal GHC enrollees with diabetes who sustained an incident fatal or nonfatal MI between July 1986 and December We identified case subjects from computerized hospital discharge data maintained by GHC and from Washington State death files. Case subjects were eligible only if their medical records indicated that they were postmenopausal at the time of their event. We presumed that women over age 55 were DIABETES CARE, VOLUME 21, NUMBER 7, JULY

2 Estrogens and MI in diabetic women postmenopausal if there was no indication of menopause status in the medical records. Case subjects were eligible only if they were between the ages of 30 and 79, if they had at least four visits to a GHC provider, if they did not have a history of prior MI, and if their medical record showed evidence of pharmacologically treated diabetes. Control subjects were diabetic women drawn from a stratified random sample of postmenopausal women enrolled at GHC. Like the case subjects, control subjects were eligible only if they were between the ages 30 and 79, if they had at least four visits to a GHC provider, and if they did not have a history of MI. We assessed whether control subjects were postmenopausal and diabetic in the same manner as for case subjects. The control group was a subgroup of control subjects from an ongoing study of hormone replacement therapy and risk of MI in the general population of female GHC enrollees (2,18). In the parent study, control subjects were randomly selected from among women enrolled at GHC during the same calendar year as MI events of case subjects, and they were frequency-matched to case subjects at a ratio of 2:1 on age by 10-year strata. There were more case subjects than control subjects in the present analysis of diabetic women because of the higher prevalence of diabetes in MI case subjects relative to the general population. Case validation and eligibility All MI cases identified in the computerized hospital discharge and death data were validated by a review of the outpatient medical records maintained by GHC, which include summary information from hospitalizations and care provided outside the GHC system. Events were excluded if they did not meet standard criteria for definite or probable MI. Difficult cases were referred to and reviewed by study physicians (S.R.H., B.M.E). Case subjects were excluded if their event was a complication of surgery or a medical procedure. The methods of case identification and validation used in this study are described more fully elsewhere (2). Reference dates All subjects were assigned a reference date. For case subjects, the reference date was the date of hospitalization, or the date of death for fatal out-of-hospital events. For control subjects, the reference date was a randomly selected date from the year of the study period during which they were selected as control subjects. Collection of data on cardiovascular risk factors Information on risk factors for cardiovascular disease was collected from GHC outpatient medical records. Risk factors included duration of treated diabetes, history of cardiovascular conditions, blood pressure, menstrual and reproductive history, cholesterol and other laboratory values, social history (including education level and occupation), and behavioral history (including tobacco and alcohol use). Data abstractors were not blinded to casecontrol status as they conducted medical record review, although they were unaware of the study hypothesis. Glycosylated hemoglobin values were collected from computerized laboratory data rather than from medical records. Information on education level was collected from telephone interviews as well as from medical records. Use of insulin and oral hypoglycemic medications was assessed using computerized pharmacy data. We collected data on cardiovascular risk factors at the time immediately preceding the reference date and used the most recent value available before the reference date for laboratory measures and vital signs. Collection of data on hormone use GHC's computerized pharmacy database was the primary source of information about use of estrogens and progestins. Since 1976, GHC has coordinated its inhouse pharmacy services through an automated system, which contains a patient identifier, drug type and dose, dosing instructions, and dispensing date for every prescription filled at a GHC pharmacy. In a previous study, a telephone survey of GHC members indicated that 94.5% filled all prescriptions and 96.8% filled "all or almost all" prescriptions at a GHC pharmacy (2). We assessed current use of estrogens by searching the pharmacy data for the most recent prescription before the reference date and calculating its duration based on quantity of pills dispensed, dosing instructions, and an assumption that patients complied with instructed regimens 80% of the time. Subjects were defined as current users if their most recent estrogen prescription provided enough pills to last through the reference date. For current users, we determined the cumulative duration of all estrogen prescriptions in the pharmacy data. When a new prescription was filled before the previous one had elapsed, we assumed that pills remaining from the earlier one were not taken. We used the same method to determine whether progestins were also used at the reference date. We defined subjects as past users if they were not current users but had two or more prescriptions for estrogens in the computerized pharmacy data, or if the medical records indicated a history of estrogen use. For past users identified in pharmacy data, we estimated the time since stopping estrogen use from the expected runout date of the last estrogen prescription before the reference date. For past users identified only by medical records, we could not estimate time since stopping estrogen use. Statistical analysis We estimated the relative risk (RR) of MI associated with current use of estrogens and with cumulative duration of use in current users relative to never users. We also estimated the RR for past use, and for time since stopping use, relative to never use. In the analyses of cumulative duration of use and time since stopping use, we divided estrogen users into three categories by tertiles, and we tested for trend among subjects with non-zero estrogen use using a continuous linear variable. RRs and 95% CIs were approximated by odds ratio estimates from logistic regression models (19). Logistic regression models included the matching variables age and study year as adjustment variables. We adjusted for other variables if they changed the RR estimates for estrogen use, including time enrolled at GHC, number of physician visits in the year before the reference date, height, weight, BMI, systolic and diastolic blood pressure, history of congestive heart failure, history of stroke, current treated hypertension, current smoking, current angina, surgical menopause, hysterectomy, duration of treated diabetes, age at first treatment for diabetes, glycosylated hemoglobin level, creatinine level, use of insulin and sulfonylureas, race, education level, usual occupation, and marital status. These covariates were also used as stratification variables to examine whether RRs for estrogen use varied in subgroups of the population. Duration of treated diabetes was not recorded in the medical records for 14 subjects, and in these cases, we estimated the time offirsttreatment as the day of the first prescription for insulin or oral diabetes medications in the pharmacy data. We used 1118 DIABETES CARE, VOLUME 21, NUMBER 7, JULY 1998

3 Kaplan and Associates Table 1 Characteristics of MI case and control subjects among postmenopausal diabetic women at GHC, July 1986 through December 1994 n Age (years) Time enrolled at GHC (years) Physician visits in last year Time since last menstrual period (years) Surgical menopause (%) BMI (kg/m 2 ) Systolic blood pressure (mmhg) Diastolic blood pressure (mmhg) Glycosylated hemoglobin (%) Random glucose (mmol/1) Blood urea nitrogen (mmol/1) Creatinine (umol/1) Total cholesterol (mmol/1) HDL cholesterol (mmol/1) Current smoking (%) Current angina (%) Current treated hypertension (%) History of stroke (%) History of congestive heart failure (%) White race (%) Married (%) Duration of treated diabetes (years) Age first treated for diabetes (years) First treated before age 40 (%) Insulin use, ever (%) Insulin use, last 6 months (%) Glyburide use, ever (%) Glyburide use, last 6 months (%) Other sulfonylurea use, ever (%) Other sulfonylurea use, last 6 months (%) Case subjects Control subjects Data are means or %. Data represent characteristics at the time of reference date, or the most recent data available before reference date for laboratory values and vital signs. education level reported in the telephone interview if it was available and relied on information from the medical records for 74 subjects who did not report education level. Values were relatively complete for most covariates and were missing for >5% of subjects only for education level (n = 137), glycosylated hemoglobin level (n = 55), occupation (n = 40), time since menopause (n = 40), race (n = 37), total cholesterol level (n = 35), HDL cholesterol level (n = 106), marital status (n = 28), hysterectomy (n = 28), and surgical menopause (n = 27). We omitted subjects with missing values of adjustment or stratification variables from regression models that included these variables. As a sensitivity analysis, we repeated RR estimation using three alternative approaches: 1) using an assumption of 100% compliance with dosing instructions, rather than 80% compliance, for defining estrogen use; 2) excluding subjects in the group of never users who had been enrolled at GHC for fewer than 10 years (n = 42 case and n = 23 control subjects); 3) excluding subjects first treated for diabetes before age 40 (n = 19 case subjects, no control subjects). None of these alternative approaches changed the results appreciably, and so we present only the results of the main analyses. RESULTS We identified 212 cases of incident MI in postmenopausal treated diabetic women who met selection criteria. Of these, 12 (5.7%) were out-of-hospital deaths, 23 (10.8%) were in-hospital deaths, and 177 (83.5%) were nonfatal Mis. The 212 cases eligible for this study represent 21.1% of the 1,004 female GHC enrollees who sustained an incident MI between 1986 and We identified 122 treated diabetic control subjects who met eligibility criteria from among 2,242 control subjects included in the parent study. On average, case subjects had a longer duration of treated diabetes and an earlier age of first treatment than did control subjects, although almost all subjects were first treated for diabetes after age 40 (Table 1). Case subjects were more likely than control subjects to be treated with insulin and less likely to be treated with glyburide. Case subjects also had more medical visits in the past year; had a higher BMI; had higher values of glycosylated hemoglobin, creatinine, blood urea nitrogen, and total cholesterol; were more likely to be current smokers; and were more likely to have current angina and a history of stroke or heart failure. Forty-seven subjects (31 case subjects, 16 control subjects) experienced menopause before age 40, due in most cases (n = 34) to bilateral oophorectomy. According to computerized pharmacy data, 18 case (8.5%) and 17 control (13.9%) subjects were current users of estrogens at the reference date (Table 2). Among case subjects, five subjects used >0.9 mg of estrogens per day, and the rest used doses of <0.625 mg. Among control subjects, all 17 current users were on daily doses of ^0.625 mg. Among current estrogen users, seven case and four control subjects were also current users of a progestin (medroxyprogesterone acetate) at the reference date. Current use of estrogen among control subjects was more common during the period (10/63,15.9%) than during the period (7/59,11.9%). Ninety-six case subjects (45.3%) were past users of estrogens (Table 2). Of these, 32 were identified by pharmacy data, and an additional 64 by medical records. Fortysix control subjects (37.7%) were past users of estrogens. Sixteen were identified by pharmacy data and an additional 30 by medical records. The category of never users contained 98 case subjects (46.2%) and 59 control subjects (48.4%). Current use of estrogens was associated with an RR of MI of 0.51 (95% CI ) relative to never use, after adjustment for age, study year, current angina, weight, and duration of diabetes (Table 2). Past use of estrogens was associated with a multivariate-adjusted RR of 1.22 (95% CI ) relative to never use. Further adjustment for other covari- DIABETES CARE, VOLUME 21, NUMBER 7, JULY

4 Estrogens and MI in diabetic women Table 2 Adjusted RR of MI according to use of postmenopausal estrogens, among diabetic women at GHC, July 1986 through December 1994 Case subjects Use of estrogens (n = 212) Never Current Past Control subjects Age- and year- Mukivariate- (w = 122) adjusted RR (95% CI) adjusted RR (95% CI) * 0.57( ) 1.30( ) 1.0* 0.51 ( ) 1.22 ( ) The multivariate-adjusted RR was adjusted for age, study year, weight, current angina, and duration of treated diabetes. "'Reference category. ates did not affect the RR estimates, and RRs did not differ appreciably in subgroups defined by age, study year, BMI, current hypertension, current angina, surgical menopause, hysterectomy, duration of treated diabetes, or race. Relative to never users, the risk of MI was lower for current estrogen users with longer cumulative duration of use in multivariate-adjusted analyses (Table 3). Among current users, the multivariateadjusted RR for each additional year of estrogen use was 0.78 (95% CI ). Among past users of estrogens, there was little variation in the risk of MI in relation to how recently estrogens had been stopped. Women who stopped estrogen less than 3 years in the past (n = 14) had an RR of 0.90 (95% CI ) relative to never users, after adjustment for age, study year, angina, and duration of treated diabetes. Multivariate-adjusted RRs were 0.78 (95% CI ) for women who stopped using estrogens 3-5 years in the past (n = 13) and 1.50 (95% CI ) for women who stopped 5-16 years in the past (n = 21), relative to never users. CONCLUSIONS Among postmenopausal women being treated for diabetes, current use of estrogens was associated with a reduced risk of MI relative to never use, although the 95% CI for the RR was wide. Among current estrogen users, there was a trend toward decreasing risk of MI with longer cumulative duration of use. Past users of estrogens had a similar or slightly higher risk of MI than did never users. A strength of this study includes the use of computerized files to measure drug use, which capture nearly all prescription medications for GHC enrollees. Nonetheless, it was necessary to use the medical records to determine drug use during periods of time not included in the computerized data. We also used the medical records to validate MI cases, assess diabetes and menopause status, and measure adjustment variables. A limitation of this study was the small number of subjects and the low prevalence of hormone use, which decreased statistical power. The small sample did not include enough hormone users to examine separately various estrogen doses or combined estrogen-progestin therapy. Also, the observed associations could have been biased to the extent that factors influencing the starting or stopping of estrogen use were separately related to the likelihood of developing a MI (20,21). We used information from the medical records, telephone interviews, and computerized pharmacy and laboratory data to adjust for cardiovascular disease risk factors, but there may have been other sources of confounding that we did not measure. In this study of treated diabetic women, 13.9% of control subjects were current users of postmenopausal estrogens. The prevalence of current estrogen use among control subjects was 20.8% in the parent study, which sampled postmenopausal women from the same HMO without regard to diabetes status (18). Thus, in this HMO setting, use of postmenopausal hormones was less common among diabetic women than among the general population of women. The diabetic women were more likely than women in the general HMO population to have treated hypertension (63.1 vs. 35.1% among control subjects) and angina (11.5 vs. 6.3% among control subjects) (18). Some authors have suggested that hormone replacement therapy is more likely to benefit women with known cardiovascular conditions such as high blood pressure and angina (4,7). Although the RR for current estrogen use in this diabetic population was not significantly different from the null, the point estimate of 0.51 was consistent with studies of general populations (1,2,22). In an analysis of data from the Nurses' Health Study, Grodstein et al. (1) reported an inverse relationship between estrogen use and coronary disease among the subgroup of their population with diabetes, although they did not report an RR. There was a trend toward lower risk of MI with longer duration of estrogen use in this study. Studies of general populations have not produced consistent findings regarding the relationship between duration of estrogen use and risk of MI. Henderson et al. (22) reported an association between lifetime duration of estrogen use and decreased Table 3 Adjusted RR of MI according to cumulative duration of postmenopausal estrogen use in current users, among diabetic women at GHC, July 1986 through December 1994 Use and cumulative duration of use of estrogens Case subjects Control subjects Never Current <2 years 2-6 years >6 years Linear trend (per year use)t Age- and yearadjusted RR (95% CI) Multivariateadjusted RR (95% CI) * 1.0* ( ) 0.52( ) 0.25 ( ) 0.85( ) 1.18( ) 0.69( ) 0.18( ) 0.78( ) The multivariate-adjusted RR was adjusted for age, study year, weight, current angina, and duration of treated diabetes. *Reference category; tamong current esrogen users only DIABETES CARE, VOLUME 21, NUMBER 7, JULY 1998

5 Kaplan and Associates mortality from MI. Rosenberg et al. (23) did not report an overall association between risk of MI and duration of use, although duration was associated with a decreased risk among recent estrogen users. However, other studies have not found an association between duration of estrogen use and coronary heart disease (1,24). This study suggests that use of postmenopausal estrogens does not increase risk of MI in diabetic women and that sustained use may be of benefit. Because diabetic women are at high absolute risk of MI, a clinical intervention that decreases risk even to a small relative degree is important. More research on the association between postmenopausal estrogens and coronary disease risk is needed to quantify the risks and benefits of hormone replacement therapy in diabetic women. Acknowledgments The research reported in this article was supported by grants HL40628, HL43201, and HL53375 from the National Heart, Lung, and Blood Institute, Bethesda, MD. B.M.E is a Merck/SER Clinical Epidemiology Fellow (sponsored by the Merck Foundation, Rahway, NJ, and the Society for Epidemiologic Research, Baltimore, MD). References 1. Grodstein F, Stampfer MJ, Manson JE, Colditz GA, Willett WC, Rosner B, Speizer FE, Hennekens CH: Postmenopausal estrogen and progestin use and the risk of cardiovascular disease. N Engl J Med 335: , Psaty BM, Heckbert SE, Atkins D, Lemaitre RN, Koepsell TD, Wahl PW, Siscovick DS, Wagner EH: The risk of myocardial infarction associated with the combined use of estrogens and progestins in postmenopausal women. Arch Intern Med 154: , Falkeborn M, Persson I, Adami H, Bergstrom R, Eaker E, Mohsen HLR, Naessen T: The risk of acute myocardial infarction after oestrogen and oestrogenprogestogen replacement. Br J Obstet Gynaecol 99: , American College of Physicians: Guidelines for counseling postmenopausal women about preventive hormone therapy. Ann InternMed 117: , Coope JKM, Coope JR: HRT, a general practice approach: how to reach the most vulnerable. BrJ CM Pract 50:38-43, National Cholesterol Education Program: Second report of the Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults. Circulation 89: , Barrett-Connor E: Risks and benefits of replacement estrogen. Annu Rev Med 43: , Harris MI, Hadden WC, Knowler WC, Bennett PH: Prevalence of diabetes and impaired glucose tolerance and plasma glucose levels in US population aged years. Diabetes 36: , Butler WJ, Ostrander LD, Carman WJ, Lamphiear DE: Mortality from coronary heart disease in the Tecumseh Study. Am] Epidemiol 121: , Barrett-Connor E, Wingard DL: Sex differential in ischemic heart disease mortality in diabetics: a prospective population-based study. Am J Epidemiol 118: , Pan W-H, Cedres LB, Liu K, Dyer A, Schoenberger JA, Shekelle RB, Stamler R, Smith D, Collette R Stamler J: Relationship of clinical diabetes and asymptomatic hyperglycemia to risk of coronary heart disease mortality in men and women. Am] Epidemiol 123: , Kleinman JC, Donohue RP, Harris MI, Finucane FF, Madans JH, Brock DB: Mortality among diabetics in a national sample. Am] Epidemiol 128: , Brussaard HE, Gevers Leuven JA, Frolich M, Kluft C, Krans HMJ: Short-term oestrogen replacement therapy improves insulin resistance, lipids and fibrinolysis in postmenopausal women with NIDDM. Diabetologia 40: , PEPI Trial Writing Group: Effects of estrogen or estrogen/progestin regimens on heart diseaseriskfactors in postmenopausal women. JAMA 273: , RobinsonJG, Folsom AR, Nabulsi AA, Watson R, Brancati FL, Cai J: Can postmenopausal hormone replacement improve plasma lipids in women with diabetes? Diabetes Care 19: , Walsh BW, Schiff I, Rosner B, Greenberg L, Ravnikar V, Sacks FM: Effects of postmenopausal estrogen replacement on the concentrations and metabolism of plasma lipoproteins. N Engl] Med 325: , Kritz-Silverstein D, Barrett-Connor E: Long-term postmenopausal hormone use, obesity, and fat distribution in older women. JAMA 275:46-49, Heckbert SR, Weiss NS, Koepsell TD, Lemaitre RN, Smith NL, Siscovick DS, Lin D, Psaty BM: Duration of estrogen replacement therapy in relation to the risk of incident myocardial infarction in postmenopausal women. Arch Intern Med 157: , Stata Corporation: Stata Reference Manual: Release 5.0. College Station, TX, Stata Corporation, Barrett-Connor E: Postmenopausal estrogen and prevention bias. Ann Intern Med 115: , Matthews KA, Kuller LH, Wing RR, Meilahn EN, Plantinga P: Prior to use of estrogen replacement therapy, are users healthier than nonusers? Am ] Epidemiol 143: , Henderson BE, Paganini-Hill A, Ross RK: Decreased mortality in users of estrogen replacement therapy. Arch Intern Med 151:75-78, Rosenberg L, Palmer JR, Shapiro S: A casecontrol study of myocardial infarction in relation to use of estrogen supplements. Am J Epidemiol 137:54-63, Folsom AR, Mink PJ, Sellers TA, Hong C-P, Zheng W, Potter JD: Hormonal replacement therapy and morbidity and mortality in a prospective study of postmenopausal women. Am] Public Health 85: , 1995 DIABETES CARE, VOLUME 21, NUMBER 7, JULY