This Article Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Barrett-Connor, E. Search for Related Content PubMed PubMed Citation Articles by Barrett-Connor, E.

(Circulation. 1997;95:252-264.)
© 1997 American Heart Association, Inc.

Articles

Sex Differences in Coronary Heart Disease

Why Are Women So Superior? The 1995 Ancel Keys Lecture

Elizabeth Barrett-Connor, MD

the Department of Family and Preventive Medicine, University of California, San Diego.

Key Words: sex • coronary disease • women • men • survival

	Introduction

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
It was a great privilege to be asked to give the 1995 Ancel Keys lecture, following five very distinguished colleagues [Geoffrey Rose, Henry Blackburn, Jeremiah Stamler, Fred Epstein, and William Kannel] and honoring Ancel Keys. It is difficult to imagine how little we would know about CHD without their contributions.

The subject of my lecture, "Why are women so superior with regard to coronary heart disease?," is surely one of the most interesting of all epidemiological questions. My intention is to illustrate methodological problems and review old and new provocative results, making inferences from a variety of disciplines. My focus is on the origin of the gender gap, not postmenopausal estrogen therapy. There will be no conclusions, only reconstruction. My lecture is outlined in Fig 1.

View larger version (29K): [in this window] [in a new window]   Figure 1. Lecture outline.

	Sex Roles

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
Unhealthy Behaviors
For years it was thought that the excess male mortality was explained by unhealthy behaviors that were more socially acceptable for men than women. These behaviors might include cigarette smoking, heavy alcohol use, eating more red meat and fewer fruits and vegetables, and exposure to physical hazards. The usual view was that differences in behavior were more important determinants of the higher male mortality than inherent sex differences in physiology.¹ More recently, studies adjusting for unhealthy behaviors show that they contribute to but do not fully explain the increased risk of CHD in men.²

Work Outside the Home
It was also a popular premise that men were harmed by the stresses of the workplace; women were apparently protected by being at home relaxing with the children and the washing up. As Bernard notes, ". . . the husband went forth each day to grapple with the cruel cold world and the wife remained in the `heart' of the home, generating sweetness and light to bind the wounds inflicted on the husband and to serve in general as a hovering angel."³ Subsequently, several studies, such as those from Framingham,^{4 5} found that only low-level work (eg, clerical jobs with subordinate status) and limited education were associated with an increased risk of heart attack in women. Success outside the home was not dangerous; in fact, women in executive or professional positions have been found to have more favorable heart disease risk factors than women who stayed home.^{6 7} The expected epidemic of CHD when women entered the workplace did not occur.

Social Supports and Communication Style
Socialization and communication styles in women are different from those in men, an observation recently popularized in the best-selling book Men Are From Mars, Women Are From Venus.⁸ Less propensity to share feelings and the resultant inward anger may have biological consequences via increased levels of stress hormones or other mechanisms. Studies of neuroendocrine and cardiovascular reactivity generally show lower reactivity in women than men.⁹

Different socialization needs or skills may also be a reason why men tend to have fewer social supports than women. Social supports appear to be cardioprotective,¹⁰ although there may be alternative explanations for these associations.¹¹ Perhaps because men have fewer social supports and different communication skills, they are more likely than women to name their spouse as their primary confidant, to derive greater advantages from being married, and to die if bereaved in old age.^{12 13 14}

Coronary-Prone Behavior
In 1897, Osler highlighted an association between male behavior and CHD when he described the typical heart disease patient as a ". . . keen and ambitious man, the indicator of whose engine is always at full speed ahead."¹⁵ The concept of coronary-prone behavior was tested in the classic studies by Friedman and Rosenman,^{16 17} who showed that a behavior pattern (type A) characterized by aggressiveness, competitiveness, hostility, and time urgency was associated with CHD in both men and women (and noted that women who stayed home were more likely to be without these traits, ie, type B). This association was confirmed in prospective studies from Framingham, which showed that coronary-prone behavior, particularly anger and anxiety, was associated with an increased risk of heart attack in women.^{4 18} The importance of the type A behavior pattern has been difficult to demonstrate in more recent studies, and the focus has shifted from the aggressive and competitive components to hostility and time urgency.¹⁹

Masculine and Feminine Traits
Nevertheless, it may be premature to discard a role for the competitive and aggressive characteristics of type A behavior, which are central components of the "masculine" personality traits described by social psychologists.^{20 21} For example, Helgeson²² notes that traditional masculinity "seems a prescription for the most dangerous components of coronary-prone behavior." Anthropological studies of hunter-gatherer societies support an evolutionary advantage for masculine and feminine sex roles. In primitive societies, the hunters who provided the meat were usually men, whose success and survival depended on aggression and competitiveness, whereas the gatherers, who tended the children and collected the plant foods necessary for day-to-day survival, were usually women.

If masculinity and femininity are relatively stable traits and not merely reactive states, then it should be possible to study them as risk factors for CHD. Nevertheless, there has been considerable reluctance to study masculinity and femininity as traits, ie, constant, possibly heritable characteristics, perhaps in part reflecting feminist concerns that such labeling both limits women's options and ignores the effect of societal norms on behavior.

The BSRI²⁰ is based on the nonjudgmental premise that there are two distinct categories of behavior: masculine, with attributes that are particularly desirable or acceptable for men (eg, aggressive and competitive), and feminine, with attributes that are particularly desirable or acceptable for women (eg, nurturing, supportive). It also supposes that these behavior traits are stable over time. Bem further proposed that individuals who scored high for both masculine and feminine attributes (androgynous) might be the most flexible and healthiest.

To test this hypothesis, in 1988 we obtained BSRI scores from 664 men and 822 women from the Rancho Bernardo cohort, whose mean age was 69 years. They have now been followed for 6 years to determine whether masculinity, femininity, or androgynicity as defined by BEM would predict fatal CHD. Rancho Bernardo men were twice as likely to be classified as masculine than feminine, and women were twice as likely to be classified as feminine than masculine. An equal proportion of men and women were androgynous, and 23% of men and women could not be classified. As shown in Fig 2, masculine, feminine, and androgynous traits were unrelated to fatal CHD, but the number of CHD deaths (14 in men and 10 in women) was still very small. We plan to continue the follow-up and to repeat the BSRI to test for stability over time.

View larger version (20K): [in this window] [in a new window]   Figure 2. Sex-specific age-adjusted odds ratios for ischemic heart disease mortality by BSRI classification, Rancho Bernardo, Calif, 1988. Courtesy of Deborah Morton.

	Unisex Analyses

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
For many years, nearly all studies of CHD causes and treatment were conducted in men. This focus on men probably reflects their earlier onset of CHD and may have been more ageist than sexist. The current "catch-up" program features studies that include only women. Although some women-only studies, such as studies of the menopause transition and clinical trials of sex hormones, are needed and useful, they will not tell us much about sex differences, and they may be misleading about women as well.

Failure to study both sexes limits our understanding of each sex. I have chosen four examples to illustrate this point: physical activity, DHEAS, coronary artery reactivity, and parity.

Physical Activity
There is general agreement that physical inactivity is associated with an increased risk of heart disease in men.²³ Some of the strongest evidence of an inverse association between physical activity and CHD comes from the long-term follow-up of male university students.^{24 25} In contrast, studies of women have produced mixed results. For example, a large prospective study reported by Blair and colleagues²⁶ found increasing levels of exercise to be associated with a decreasing risk of heart disease in men but not in women. Possible explanations for this difference include the following. (1) Most women's physical activity is too low to show a benefit; (2) women are incorrectly classified as sedentary, because even the sedentary group does housework and gardening (which are not usually queried); (3) questionnaires focus on sports rather than female activities (vacuuming a two-story house requires the same energy expenditure as nine holes of golf); or (4) exercise has no benefit for women. (This last possibility is plausible if the exercise benefit is mediated by improving levels of something women usually already have, such as high levels of HDL cholesterol.)

A prospective analysis of fitness levels in >3000 women who were followed for 8 years showed that age-adjusted death rates from all causes (the majority from heart disease) decreased with increasing levels of physical fitness.²⁷ Similar results were observed in men. Since physical inactivity results in low levels of fitness, this study suggests that the previous failure to find an exercise benefit was due to poor measurement of the exposure (exercise). In other words, one likely explanation for the failure to find an inactivity-CHD association was the use of physical activity questionnaires that were designed for men, focused on sports, and neglected energy expended around the home.

Dehydroepiandrosterone Sulfate
DHEAS, an adrenal androgen, is present in higher levels in men than women (whose levels are further reduced by the administration of exogenous estrogen).²⁸ Although DHEAS is a weak androgen, circulating levels are very high (compared with other sex steroids) in both sexes, suggesting that it has an important if still undefined biological role.

Several years ago we published a 12-year prospective study showing that middle-aged men with low levels of DHEAS had a strikingly increased risk of fatal heart disease.²⁹ We did not look at the women initially, because there were so few CHD deaths. When we did look, we found that the reverse was true: high DHEAS levels in women were associated with an increased risk of CHD.³⁰ We have since measured DHEAS in a larger number of this cohort and followed everyone for 19 years.³¹ The lower risk in men with high DHEAS levels persists, although it is now less dramatic; the increased risk in women is gone, but DHEAS still offers no protection. Similar sex differences were reported by Herrington and colleagues,³² who found lower DHEAS levels in men but not women who had coronary atherosclerosis at angiography.

We do not understand these sex differences, but a recent report from Caracas provides some insights. Jakubowicz et al³³ treated 18 obese men and 29 obese women with a low-calorie diet for 2 months. Both men and women lost weight and dramatically reduced their insulin levels. In men, weight loss and insulin reduction were associated with a striking (presumably favorable) increase in DHEAS, but DHEAS levels did not change in women. These sex-specific results make it unlikely that any cardioprotective effect of DHEAS is mediated by a sex-neutral metabolic effect, such as the potent noncompetitive inhibition by DHEAS of glucose-6-phosphate dehydrogenase, the rate-limiting enzyme of the pentose pathway. We doubt that the cardioprotective effect of DHEAS is mediated by its conversion to estrogen, because plasma estrogen levels are not associated with CHD risk (see below). We have proposed that the apparent "men-only" benefit of high DHEAS levels is compatible with a DHEAS-androgen effect. This thesis postulates that androgens are good for men and estrogens are good for women, a teleologically attractive argument we have made elsewhere.³⁴

Coronary Artery Blood Flow
Pharmacological doses of estrogen have been shown to improve exercise-induced myocardial ischemia in postmenopausal women with angiographically proven CHD.³⁵ The importance of studying the effect of estrogen on the coronary arteries of both sexes is beautifully illustrated by a recent report from London. Collins and colleagues³⁶ used quantitative coronary angiography to study the effect of intracoronary administration of pharmacological doses of 17ß-estradiol in seven men and nine women with coronary artery disease. In both men and women, acetylcholine caused the expected vasoconstriction of atherosclerotic coronary arteries. This was rapidly reversed by estrogen in women but not in men, as shown in Fig 3. Coronary artery blood flow was also significantly improved by estrogen in women but was unchanged in men.

View larger version (30K): [in this window] [in a new window]   Figure 3. Change in diameter of coronary arteries produced by an intracoronary infusion of acetylcholine before and after intracoronary ß-estradiol in men and women. Reproduced with permission.36

The acuteness of the vasodilation argues against an endothelial effect mediated by lipid or lipoprotein changes. The sex-specific results suggest sex-specific receptors in the coronary artery, although the same group has found no evidence for classic estrogen receptors in the coronary arteries of women.³⁷ Whatever the mechanism, if these results are confirmed they point to the inappropriateness of assuming equivalent benefits of hormone treatment to prevent CHD in both sexes. These investigators are now using the same experimental design to study the effect of testosterone on coronary arteries in men and women.

Parity
As reviewed elsewhere,³⁸ some but not all studies have found that parity is associated with an increased risk of CHD. The largest study (of >100 000 death certificates in England and Wales) found a significant 20% increase in death from cardiovascular disease comparing multiparous with nulliparous women.³⁹

Possible reasons for this association are (1) confounding due to the association of multiparity with lower social class and less education (two powerful CHD risk factors in both sexes); (2) that parity is a marker for high fertility, which is related to some other CHD risk factor (this has been proposed for diabetes, for example); (3) an indirect association with the weight gain and central obesity that often persists after pregnancy; (4) the decrease in HDL and increase in insulin levels that tend to persist after pregnancy; (5) repeated hormone changes (such as the increase in cortisol observed during normal pregnancy⁴⁰ ; or (6) the stresses, anxiety, or habits of parenthood, which may include a less healthy diet.

To determine whether the parity-CHD association is independent of fertility, social class, and the metabolic changes of pregnancy, we studied parenthood in both men and women who were old enough to have completed their families before oral contraceptives were used, who were homogeneous with regard to education and social class, and who had also provided data on nonbiological children. In the Rancho Bernardo cohort, body mass index (shown in Fig 4) and waist-to-hip ratio were greater in both men and women who had more children. Several other CHD risk factors were less favorable in men who had more than four biological or nonbiological children.⁴¹

View larger version (12K): [in this window] [in a new window]   Figure 4. Mean age-adjusted body mass index by number of biological children. Rancho Bernardo men and women 55 to 84 years old, 1984 to 1987. Trends: men, P=.001; women, P=.004. Courtesy of Nancy J. Friedlander.

This rather simple study shows how research restricted to women may lead to some unnecessarily complex and erroneous biological hypotheses.

	Population Studies

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
Ecological studies of CHD have fallen out of favor, but they still have much to teach us.⁴² The remarkable universality of the sex difference in CHD death rates in countries that have very different CHD mortality experiences is a case in point, particularly when deaths occurring in very elderly men and women (whose actual cause of death is often uncertain) are excluded. The now familiar Kalin and Zumoff⁴³ graph (Fig 5) shows the rather consistent male-to-female ratio of 2.5 to 4.5 for fatal CHD in countries with very different rates. Only in China are CHD death rates said not to vary by sex⁴⁴ ; this observation requires confirmation.

View larger version (19K): [in this window] [in a new window]   Figure 5. Age-standardized coronary disease death rates in 1987 for men and women from 52 countries. Reproduced with permission.43

The virtually universal male excess of CHD implies a universal inherent female advantage that operates across populations with divergent rates of heart disease and lifestyles. The obvious explanation is that estrogen is good or that testosterone is bad. But the wide differences in CHD death rates between countries, coupled with the rather consistent sex ratios shown in Fig 5, suggest that this sex effect is mediated by heart disease risk factors that exert an effect on both men and women. In fact, the differences between countries are greater than the differences between sexes, suggesting that sex is not destiny with regard to CHD.

We might learn which risk factors are most apt to have a sex-specific effect by pursuing another well-known observation: Until recently there has been no consistent gender gap for stroke death, even when analyses are restricted to non-Asian countries that have low hemorrhagic stroke rates, as shown in Fig 6. In recent years the sex ratio for stroke has been increasing in many countries, reflecting a more rapid decrease in stroke rates in women than men.⁴⁵ Among the classic risk factors for CHD (hypertension, diabetes, cigarette smoking, and cholesterol), only cholesterol is not a strong independent risk factor for thrombotic stroke. By inference, total or LDL cholesterol is likely to play a central role in the gender gap for CHD risk.

View larger version (41K): [in this window] [in a new window]   Figure 6. International age-standardized stroke rates by sex, 1991. EW indicates England and Wales. Courtesy of Kay-Tee Khaw.

Similar clues to the CHD gender gap can probably be found by examining sex differences in different ethnic groups between and within countries. One might profitably consider the low CHD rates in men and women in Japan, despite the frequency of hypertension and cigarette smoking, rising rates of diabetes, and relatively low levels of endogenous estrogen. Is this because they have low plasma cholesterol levels or a phytoestrogen-rich diet? Kesteloot and Sasaki⁴⁶ used international lipid data to suggest that women's cardiovascular advantage exists because women can raise their HDL cholesterol levels in response to a diet high in saturated fat, whereas men cannot.

It is clear that we have a great deal more to learn from population studies of the sort pioneered by Ancel Keys.⁴⁷ In ecological comparisons, studies of exceptions or lack of exceptions to expected sex differences should be particularly useful.

	Endogenous Estrogen

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
The universal excess risk of CHD in men noted above, coupled with the apparent loss of the female advantage in women who had an early menopause, led to the hypothesis that endogenous estrogen is cardioprotective. Proof of this hypothesis has been surprisingly elusive.

Surrogate Studies: Menopause
For most women, the menopause marks the beginning of years of uninterrupted "estrogen deficiency," which might theoretically provide more information about the consequences of low estrogen levels than single assays of circulating hormones. The average age at menopause, 50 to 51 years, has been consistent for centuries.⁴⁸ Therefore, we might expect an effect of age on heart disease risk to be observed in vital statistics data.

However, as Furman noted in 1968,⁴⁹ the menopause has no discernible effect on the exponential increment in CHD death with age (Fig 7). It can be argued that the inability to see a change in risk at the age of menopause reflects a broad range for age at menopause or a weak correlation between the timing of the last menstrual period and the fall in estrogen. This explanation is unattractive, because, as shown in Fig 8, it is so very easy to see an abrupt midlife change in the age-related slope for breast cancer, a condition that is almost certainly related to endogenous estrogen levels.⁵⁰

View larger version (13K): [in this window] [in a new window]   Figure 7. Male and female mortality rates, US white population, 1955, as a function of age. Reproduced with permission.49

View larger version (15K): [in this window] [in a new window]   Figure 8. US women, semilogarithmic plots of age-specific heart disease and breast cancer death rates vs age (Gompertz plots), 1962. Reproduced with permission.50

As shown by Furman in 1968⁴⁹ (Fig 7), women do not catch up with men after the menopause. The decline in the ratio of male to female deaths that occurs in old age does not reflect an acceleration in women who have been postmenopausal longer; it does reflect a slowing of the acceleration rate that occurs in middle-aged men. This pattern has traditionally been ascribed to survival bias, the exhaustion of the pool of highest-risk men. Note, however, that the male midlife rate change shown in Fig 7 is more compatible with a change in male hormones than with an effect of the female menopause. This pattern has been replicated many times, most recently with death certificates from the United Kingdom.⁵¹

Premature Menopause
Studies of CHD risk in women who had an early "natural menopause," usually defined as menopause before age 40 years, have provided contradictory results. Most studies failed to control for important confounders such as social class and cigarette smoking. In addition, little is known about the factors that lead to early menopause or the psychological stresses that follow it.

The effect of early bilateral oophorectomy is easier to study only in the sense that a discrete event marks the onset of estrogen deficiency. Surgical menopause studies are confounded by the same factors that confound studies of early natural menopause plus the abrupt fall in testosterone levels that follows bilateral oophorectomy. The importance of the resulting elevations of pituitary hormones (eg, luteinizing hormone, follicle-stimulating hormone, and inhibin) is unknown. Thus, the consequences of bilateral oophorectomy in the premenopausal woman do not reflect estrogen deficiency alone.

Several early autopsy studies showed an increase in coronary artery disease in young women who had an oophorectomy,³⁸ but as the authors noted, it was not possible to separate the reason for the surgery from the outcome. In 1963, Ritterband and colleagues⁵² reported that surgically intact women had less heart disease than women who had bilateral oophorectomy or a hysterectomy only, but there was no difference in CHD rates in oophorectomized versus hysterectomized women. This early publication includes a remarkably complete discussion of various biases and confounders that complicate the understanding of oophorectomy as a risk factor. The authors conclude that some factor other than ovarian function is responsible for women's relative freedom from heart disease. Although these investigators were unaware of the (still unproven) hypothesis that hysterectomy causes premature ovarian failure (perhaps by accidentally damaging the ovarian arteries), this paper highlights the problems involved in attributing CHD to estrogen deficiency on the basis of oophorectomy.

Menopause Transition
The relatively abrupt fall in estrogen (and progesterone) during the menopause transition has led to a debate about whether the postmenopausal state is a deficiency disease or a natural phenomenon. It is well known that longevity after the menopause is a recent phenomenon; the average life expectancy was less than the age of usual menopause until the beginning of the 20th century.⁴⁸ Relatively few women survived to age 75 years, when clinical evidence of CHD (and osteoporosis) usually begins.

It is less well appreciated that women's premenopausal exposure to sex hormones is much different today than it was for her hunter-gatherer ancestors. As reviewed elsewhere,⁵³ women previously had a later menarche and then spent about 19 years of their lives in pregnancy and lactation and about 4 years in menstrual cycling. (In contrast, the typical modern woman spends about 35 years in menstrual cycling.) It is also likely that for much of human history, women experienced menopause in the hormonal milieu of lactation.⁵⁴ These anthropological observations suggest that women may be biologically adapted to a very different lifetime exposure to estrogen than they currently experience. What are the consequences?

Studies of women passing from premenopausal to postmenopausal status, modeled on the pioneer work by Kuller and colleagues^{55 56} in Pittsburgh, are under way at several sites in the United States and other countries. The prospective Pittsburgh study has found that the "modern menopause" is associated with changes in several heart disease risk factors, such as increasing LDL cholesterol levels, decreasing HDL cholesterol levels, and unfavorable changes in some hemostasis factors.⁵⁵ Natural menopause did not affect blood pressure, plasma glucose or insulin levels, or body weight.⁵⁷

Perhaps the most dramatic evidence suggesting that loss of endogenous estrogen increases cardiac risk is the sharp increase in LDL cholesterol that begins in the perimenopausal period and continues to at least age 60 years, with these higher levels sustained thereafter.⁵⁸ Interestingly, both cross-sectional and prospective studies show only a small decrease in HDL cholesterol levels at the time of menopause^{59 60 61} ; on average, HDL levels in women remain higher than HDL levels in men for at least another 30 years after menopause, raising the possibility that levels of this lipoprotein, often credited for some of the female advantage, are more related to obesity and lifestyle than to sex hormones.

	Endogenous Estrogen and CHD Risk Factors or Events

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
There are no data from clinical trials on the effects of physiological doses of estrogen on cardiovascular disease risk factors or risk in men or women. The oral route of administration, with its "first-pass" liver effect, appears to be responsible for one of the most favorable lipoprotein changes attributed to exogenous estrogen, the increase in HDL cholesterol seen with oral but not transdermal estradiol.⁶²

Therefore, the relation of physiological levels of estrogen and CHD is based on observational studies. As reviewed elsewhere, nearly all observational studies of postmenopausal women and middle-aged men show no positive association between endogenous estrone or estradiol levels and CHD risk factors, most notably HDL cholesterol.^{63 64} Although one prospective study that followed women through the menopause showed the expected striking decrease in estradiol levels and decrease in HDL cholesterol, the hormone-lipoprotein association was neither linear nor independent of covariates.⁶¹ Similarly, two prospective studies in men found no association between estrone or estradiol levels and the risk of future CHD.^{65 66} In women, one cross-sectional study of coronary atherosclerosis⁶⁷ and one prospective study⁶⁸ also found no association of CHD with sex hormones, including estrone, estradiol, and testosterone.

Obviously, these studies cannot completely exclude a cardioprotective effect of endogenous estrogen. A single hormone assay may be inadequate to correctly classify individuals with regard to their usual endocrine status. In addition, it is possible that blood levels do not parallel high turnover, which could reflect more biological activity. Perhaps studies of bioavailable estrogen are necessary, rather than the total estradiol or estrogen/sex hormone binding ratios usually used. Perhaps there is a threshold effect, such that postmenopausal estrogen levels are below the level of benefit. (This seems unlikely, because the CHD risk of men exceeds that of women throughout life, but men have higher estrogen levels than postmenopausal women of the same age.) Perhaps it is the estrogen/androgen ratio that is important, although several of the studies noted above did not find the hormone ratios to predict CHD risk either. Perhaps estrogen deficiency has only a transient effect on arteriosclerosis, such that early changes allow the process to begin.

	Risk Factors

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
Classic Heart Disease Risk Factors
In general, men have less favorable heart disease risk factors than women. For example, men in the Edinburgh Heart Study⁶⁹ had significantly (P<.001) less favorable levels of cigarette smoking, dietary fiber, vitamin C, blood viscosity, uric acid, HDL cholesterol, and triglycerides than women. Only three presumably cardioprotective factors were significantly more favorable in men than women: men had more reported physical activity and alcohol intake and had lower levels of fibrinogen.

Levels of common CHD risk factors do not explain the gender gap, however. The sex difference in incidence rates persists in studies adjusted for risk factor differences² or stratified by level of risk factor.⁷⁰ As shown in Fig 9, from the Renfrew-Paisley Study in Scotland,⁷⁰ women and men tend to have a similar dose response for each CHD risk factor and a similar relative risk, but women's absolute risk of CHD is much lower for any given level of risk factor. In this study, women in the highest female quintile of cholesterol (>7.2 mmol/L) had lower CHD death rates than men in the lowest male quintile (<5.0 mmol/L).⁷⁰ This was true even after the menopause.

View larger version (21K): [in this window] [in a new window]   Figure 9. Adjusted CHD mortality rates (deaths per 1000 patient-years) for both sexes by cigarette smoking, diastolic blood pressure, social class, and body mass index. Reproduced with permission.70

Metabolic Syndrome
The risk factors considered above contribute approximately the same relative risk to CHD risk in men and women. Clues to the gender gap might lie in a study of risk factors that contribute more to women's risk than men's (or vice versa). The relatively short list of interrelated variables that exert a stronger effect on female than male risk includes diabetes, low HDL cholesterol, and high triglycerides. These factors are part of a metabolic syndrome that has been associated with insulin resistance and small, dense LDL (phenotype B) dyslipidemia.⁷¹

Among these factors, the one with the most striking sex difference is diabetes. In most studies, diabetes is associated with a relative risk of CHD that is approximately twice as high in women as in men; this twofold female-to-male sex ratio occurs in both young adults with insulin-dependent diabetes and older adults with (primarily) non–insulin-dependent diabetes.⁷² In the Rancho Bernardo Study, after 12 years of follow-up, women with diabetes had a relative odds of fatal CHD that was similar to that of men with or without diabetes.⁷³ No other common risk factor so nearly erases the female advantage. I believe that a better understanding of how diabetes increases cardiovascular risk is central to understanding the causes of the gender gap.

Low HDL and high triglyceride levels, which occur more often in persons with non–insulin-dependent diabetes than in those without,⁷⁴ also have a stronger effect on CHD risk in women than men, although the differences are not as dramatic as for diabetes. And the excess CHD risk associated with diabetes appears to be independent of the increased risk associated with diabetic dyslipidemia.⁷⁵

Gordon and colleagues⁷⁶ used data from four prospective studies and regression analysis to show that the association between HDL and CHD was steeper in women than men: for each mg/dL increase in HDL there was an estimated 2% decrease in CHD risk in men and a 3% decrease in women. In stratified analyses, women have lower CHD rates than men at any given level of HDL cholesterol.^{77 78}

Hypertriglyceridemia is a feature, and probably an antecedent, of diabetes. The association of triglycerides with CHD risk also seems to be more prominent in women. In Framingham, for example, the regression coefficient for serum triglycerides and CHD was more than five times higher in women than men, although their mean triglyceride levels were lower.⁷⁹

The prevalence of small, dense LDL differs by both age and sex. In general, it is more common in men than in women and increases with age. It appears that the largest difference is around age 20 years in men and age 50 years (menopause) in women.^{80 81 82 83} Women appear to have a larger and less atherogenic LDL particle size than men.⁸⁴ There are no data yet that sufficiently describe whether or not there is a sex difference in the CHD risk associated with small, dense LDL. It has been clearly shown to be a risk factor in middle-aged men,^{85 86 87} but the sample sizes of women have not been large enough to make a rigorous statistical comparison.^{85 86}

Hyperinsulinemia or insulin resistance has been proposed to be the underlying cause of dyslipidemia and hypertension⁸⁸ and CHD.^{89 90} Although two of three early studies in men found that high fasting or postchallenge insulin levels predicted an increased risk of future heart disease in men without diabetes, six more recent prospective studies have been unable to find a positive linear association in men.⁹¹ Neither of the two prospective studies that reported women separately found any association between insulin levels and CHD in women.^{92 93}

If there is an insulin-CHD association in men but not in women, understanding this sex difference could help to solve the gender gap puzzle.⁹⁴ Or is this just a methodological problem stemming from incorrect assumptions about similarities of metabolism in men and women? This possibility is suggested by preliminary observations that serum insulin levels are more poorly correlated with insulin resistance in women than in men.⁹⁵

	Inheritance/Genes

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
Sex-Linked Inheritance
As shown in Fig 5, CHD mortality rates show a remarkably consistent sex ratio in different countries with very different CHD rates. This consistent sex ratio despite wide variations in lifestyle points to intrinsic genetic factors underlying the gender gap. Yet, the effect of sex-linked inheritance on chronic diseases manifest in adulthood has received relatively little attention from geneticists or epidemiologists. Perhaps it is just not "sexy" enough to study because the human phenotype nearly always defines the genotype; we rarely need a genetic research laboratory to determine which individual has XX or XY chromosomes. Or perhaps it is because the male-determining function of the human Y chromosome was not established until 1959. In any event, little attention has been paid by cardiologists or epidemiologists to the following question: What do the sex chromosomes do to make women less (or men more) susceptible to heart disease and when do they do it?

Human sex chromosomes determine our sexual phenotype during fetal life. In the presence of the XY chromosome, testosterone secreted by the fetal testes after the first trimester masculinizes (via dehydrotestosterone) and defeminizes (via estrogen) the cellular structures throughout the reproductive tract. It is perhaps less well appreciated that testosterone also masculinizes the brain during fetal life. The critical importance of the prenatal exposure period has been demonstrated in animal experiments⁹⁶ and by observations of humans who were exposed to chromosomally inappropriate hormones (eg, girls born with adrenal genital syndrome). These studies strongly suggest that several characteristically male or female behaviors are determined by hormone exposure in utero. (Women seem to have been particularly reluctant to accept the possibility of genetically determined, hormone-mediated male and female behaviors. Perhaps they fear that such data can be used to limit women's appropriate roles and capabilities.)

In addition, new technologies for anatomic and functional studies show sex differences in the brains of both animals and humans.^{97 98 99} Knowledge about (and acceptance of) brain sex differences have been enhanced by studies using new scanning methods that display brain activity during specific tasks; these studies show that men and women process and react to information differently. The role of these differences in defining human sex roles and behaviors (see above) and their direct or indirect relation to CHD risk await investigation.

Sex chromosomes may also play other relatively unexplored roles in CHD risk. For example, estrogen (but not testosterone) stimulates phagocytosis by macrophages,¹⁰⁰ and women's greater resistance to infection from infancy to old age is related in part to quantitative genes, although there is still some question as to whether they are carried on the X chromosome.^{101 102} These observations may be relevant to the gender gap, given the indirect evidence (based on associations with fibrinogen and C-reactive protein, for example) that atherosclerosis begins or progresses in part as an infectious or inflammatory disease.¹⁰³

In addition, autosomal genes may indirectly impact the female advantage. It has been recognized for >40 years that, even with familial hypercholesterolemia, affected women have premature heart disease about 10 years later than affected men.¹⁰⁴ The effects of apolipoprotein E polymorphism are not yet fully unraveled, but the observation that there may be fewer old men than old women with apo E4 alleles suggests a survival benefit for affected women.¹⁰⁵ Since both familial hypercholesterolemia and apo E4 genotype are associated with higher total cholesterol levels, some of this relative protection may reflect women's greater capacity to produce HDL cholesterol.

There are other possible mechanisms for female protection against genetic risk factors for CHD. For example, hyperhomocysteinemia, which can be both genetically and environmentally induced, is a recently rediscovered risk factor for heart disease.¹⁰⁶ Levels can be reduced by dietary folic acid; in many western cultures, fruits and vegetables are consumed in greater quantity by women than by men, who would therefore be more protected from the cardiovascular consequences of hyperhomocysteinemia.

Hemochromatosis is another common genetic condition in which a female attribute is protective. Approximately 10% of whites of northern European ancestry carry one abnormal gene for hemochromatosis and have a modest increase in the efficiency of iron absorption.¹⁰⁷ Whereas iron stores build up in modern men with no regular source of blood loss (no hookworms, no bloody battles), women have reproductive and menstrual bleeding to prevent iron overload. If high iron levels cause heart disease, as postulated by Sullivan,¹⁰⁸ some of the gender gap could be mediated by the blood loss of premenopausal women.

Most individuals with a family history of CHD probably have no major gene defect. It will be difficult to ascertain the role of "minor" susceptibility genes by use of classic pedigree studies if, as seems likely, gene effects are manifest only in specific environmental or behavioral contexts. For example, we have reported that family history of premature heart disease is associated with an increased risk of CHD only in women who smoke cigarettes.¹⁰⁹ Instead of the usual pedigree studies, candidate genes for common diseases of multifactorial etiology should also be studied by classic epidemiological methods in which the gene is included in the analysis as a risk factor. Certainly the geographical differences shown in Fig 5, as well as classic migration studies such as the Ni-Hon-San study,¹¹⁰ point to the strong environmental-genetic interplay that determines the cardiovascular risk for the majority.

Finally, Barker¹¹¹ has provided evidence that small birth weight and poor infant growth are associated with an increased risk of several common chronic diseases of adult life, including cardiovascular disease. These important studies highlight the nongenetic hazards of fetal life and the dangers of assuming that familial or even twin similarities are necessarily caused by our genetic makeup. The relation of fetal and infant growth to the gender gap is unstudied.

	Obesity/Male Fat Pattern

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
Epidemiological interest in generalized obesity as a CHD risk factor has been largely supplanted by interest in male-pattern (central or upper body) obesity. The association of "android" obesity with diabetes and atherosclerosis was clearly formulated by Vague¹¹² >40 years ago and was given new life by the Gothenburg studies, which showed that central obesity (estimated by waist-to-hip ratio) predicted future CHD in both men and women independent of body mass index.^{113 114} It is now fairly certain that the waist-to-hip ratio (or waist girth alone) is a marker for visceral adiposity, which in turn covaries with insulin resistance, dyslipidemia, diabetes, and CHD in both men and women.

Risk factors for central obesity (in both sexes) include cigarette smoking, alcohol intake, and stress.¹¹⁵ Male sex is by far the strongest risk factor: in nonobese Caucasian populations, there is almost no overlap in the distribution of the waist-to-hip ratio between the sexes (Fig 10).¹¹⁶ Male-pattern obesity is probably a genetically determined secondary sex characteristic. A strong heritable component to the acquisition of central obesity has been demonstrated in feeding studies of identical male twins.¹¹⁷

View larger version (15K): [in this window] [in a new window]   Figure 10. Percentage distribution of waist-to-hip ratio by sex in randomly selected men and women 54 years old, Gothenburg, Sweden. Reproduced with permission.116

Given the sex distribution of male-pattern obesity, it is plausible that when this fat pattern occurs in women, it and its associated metabolic abnormalities are caused by testosterone in excess of normal concentrations.¹¹⁸ Several studies in premenopausal women support this thesis. Thus, cross-sectional studies in premenopausal women suggest that higher levels of endogenous androgens are associated with increased activity of adipose tissue lipoprotein lipase, leading to enlarged abdominal adipocytes and possibly reflecting the number or affinity of testosterone receptors in visceral fat.^{119 120} Studies of women with polycystic ovary syndrome are compatible with an androgen-excess cause of visceral adiposity, as are results from an uncontrolled study of female-to-male transsexuals, who had a significant increase in visceral fat after 3 years of intramuscular testosterone.¹²¹

Although the prevalence of central adiposity increases after the menopause, evidence for an androgen-excess cause weakens. SHBG, a putative marker for hyperandrogenicity, has been rather consistently found to be positively associated with body fat distribution in both premenopausal and postmenopausal women,^{119 122 123 124 125 126} but three studies of bioavailable (non–SHBG-bound) testosterone in postmenopausal women found no association with current or future central obesity.¹²⁷ Some data do suggest that SHBG is not a good marker for androgenicity in women: for example, the absent correlation between SHBG and androgen levels in normal premenopausal or postmenopausal women or in women with moderate hirsutism.^{128 129} The absent association of central obesity with bioavailable testosterone in older women can be interpreted to mean that (1) a single assay does not correctly characterize the androgenic status of women, (2) poor stability of bioavailable testosterone in frozen samples precludes demonstrating an association, (3) testosterone is not associated with central obesity in older women, or (4) central girth is poorly correlated with visceral adiposity in older women. At present, evidence for a causal role for hyperandrogenism in the pathogenesis of postmenopausal central obesity remains weak.

In men the situation is clearer, if more counterintuitive, because three lines of evidence suggest that low testosterone levels cause male-pattern obesity. First, central obesity is usually first manifest between the ages of 35 and 40 years, when male testosterone levels are falling.^{130 131} Second, in cross-sectional and prospective studies, low, not high, levels of endogenous testosterone are associated with increased girth and visceral adiposity in men.^{132 133} Third, a small clinical trial has found that treatment with low doses of testosterone reduces central obesity in older men.¹³⁴

It is not difficult to postulate an evolutionary advantage for men to be able to store visceral or "android" fat and for women to store "gynoid" fat on the hips and buttocks. Men as hunters would need a rapid source of energy for the chase; visceral fat is readily mobilized and goes directly to the liver for gluconeogenesis. The ß-adrenergic lipolytic effect is higher in visceral adipocytes from men than from premenopausal or postmenopausal women.¹³⁵ For preservation of the species, women require a source of fat that is mobilized efficiently only when needed during pregnancy and lactation (which, as every woman knows, are the only ways short of starvation to move fat from the hips and thighs!). What was once a survival advantage is now a disadvantage: The opportunity to store more fat than is needed for the next "hunt" or pregnancy exists in sedentary affluent cultures, which are also marked by epidemics of diabetes and CHD.

Central obesity and its relation to abnormal glucose tolerance are reminiscent of Cushing's disease. Stress, which can raise cortisol levels, is a postulated cause of central obesity and CHD. Studies of how stress modulates the hypothalamic-pituitary-adrenal axis suggest interesting sex differences. For example, monkeys subjected to the stress of missing a single meal show elevated cortisol levels and suppressed luteinizing hormone secretion, unless the monkey is in the follicular (estrogen-dominant) phase of the menstrual cycle.¹³⁶ In young women, cortisol levels in response to a cold pressor stress test were higher during the luteal than the follicular phase of the menstrual cycle; men responded with higher cortisol levels than follicular-phase women.¹³⁷ Lindheim and colleagues¹³⁸ reported that postmenopausal women treated with transdermal estrogen had lower blood pressure, adrenocorticotropic hormone, cortisol, and norepinephrine response to psychological stress tests than placebo-treated women. Overall, these studies suggest an interaction whereby higher endogenous estrogen levels reduce cortisol response to stress (and bring us full circle to the sex-role psychosocial factors considered at the beginning of this article).

Thus, it is plausible that sex differences in visceral fat distribution hold the key to the gender gap in CHD. This thesis is compatible with the peculiarly noxious effect of diabetes in women, since diabetes is closely associated with central obesity. Unfortunately, analyses that pool sexes and adjust for waist-to-hip ratio to test this thesis are inappropriate when there is no overlap in the distribution of waist-to-hip ratio. (Such an analysis is analogous to a study of breast cancer in men and women that concludes that women are at greater risk because they have bigger breasts.) Studies of Indian Asians or other populations that show considerable overlap in waist girth between the sexes could be more useful in determining the degree to which visceral adiposity explains the gender gap in CHD.

	Reconstruction

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 
The dictionary defines reconstruction as "a theoretical view of something unknown," which seems an appropriate way to summarize the state of our knowledge with regard to the gender gap in CHD. Instead of a summary or a list of research priorities (almost impossible to provide unbiased by one's own funding anxieties), I offer seven statements about gender-gap research that I believe merit further discussion.

1. Sex role behaviors and related responses to stressors are inherited CHD risk factors determined by sex hormone exposure, particularly in utero (brain), plus environment.

2. Unisex analyses do not elucidate sex differences and may limit understanding of CHD in men and women.

3. Population studies, including ecological studies, are undervalued and underused.

4. Plasma estrogen levels do not explain CHD risk in either sex; new study methods and hypotheses are needed.

5. Risk factors, including lifestyle, lipids, blood pressure, and the metabolic syndrome, do not explain the sex difference in CHD.

6. Genes that impact the effect of environment or common heritable risk factors for heart disease differentially in men versus women are located on the sex chromosomes and probably have both hormone-mediated and non–hormone-mediated effects.

7. Male pattern obesity is not caused by high testosterone levels in men or postmenopausal women, but some stress hormone–androgen interaction may be important.

	Selected Abbreviations and Acronyms

 

BSRI = Bem Sex Role Inventory CHD = coronary heart disease DHEAS = dehydroepiandrosterone sulfate SHBG = sex hormone–binding globulin

	Acknowledgments

 
This research was partially supported by grants from the National Institute of Diabetes and Digestive and Kidney Diseases (grant DK-31801) and the National Institute on Aging (grant AG-07181).

	Footnotes

 
Reprint requests to Elizabeth Barrett-Connor, MD, Department of Family and Preventive Medicine, University of California, San Diego, 9500 Gilman Dr, La Jolla, CA 92093-0607. E-mail ebarrettconnor@ucsd.edu.

	References

Top Introduction Sex Roles Unisex Analyses Population Studies Endogenous Estrogen Endogenous Estrogen and CHD... Risk Factors Inheritance/Genes Obesity/Male Fat Pattern Reconstruction References

 

1. Waldron I. Why do women live longer than men? Soc Sci Med. 1976;10:349-362.
   
2. Wingard DL, Suarez L, Barrett-Connor E. The sex differential in mortality from all causes and ischemic heart disease. Am J Epidemiol. 1983;117:165-172.[Abstract/Free Full Text]
   
3. Bernard J. Ground rules for marriage: perspectives on the pattern of an era. In: Notman MT, Nadelson CC, eds. Women and Men: New Perspective on Gender Differences. Washington, DC: American Psychiatric Press, Inc; 1991:89-113.
   
4. Haynes SG, Feinleib M. Women, work and coronary heart disease: prospective findings from the Framingham Heart Study. Am J Public Health. 1980;70:133-141.[Abstract/Free Full Text]
   
5. Eaker ED, Pinsky J, Castelli WP. Myocardial infarction and coronary death among women: psychosocial predictors from a 20-year follow-up of women in the Framingham Study. Am J Epidemiol. 1992;135:854-864.[Abstract/Free Full Text]
   
6. Kritz-Silverstein D, Wingard DL, Barrett-Connor E. Employment status and heart disease risk factors in middle-aged women: the Rancho Bernardo Study. Am J Public Health. 1992;82:215-219.[Abstract/Free Full Text]
   
7. Haertel U, Heiss G, Filipiak B, Doering A. Cross-sectional and longitudinal associations between high density lipoprotein cholesterol and women's employment. Am J Epidemiol. 1992;135:68-78.[Abstract/Free Full Text]
   
8. Gray J. Men Are From Mars, Women Are From Venus. New York, NY: Harper Collins; 1992.
   
9. Manuck SB, Polefrone JM. Psychophysiologic activity in women. In: Eaker ED, Packard B, Wenger NK, Clarkson TB, Tyroler HA, eds. Coronary Heart Disease in Women. New York, NY: Haymarker Doyma; 1987:164-171.
   
10. Berkman LF, Syme SL. Social networks, host resistance, and mortality: a nine-year follow-up study of Alameda County residents. Am J Epidemiol. 1979;109:186-204.[Abstract/Free Full Text]
    
11. Shumaker SA, Hill DR. Gender differences in social support and physical health. Health Psychol. 1991;10:102-111.[Medline] [Order article via Infotrieve]
    
12. Gove WR. Sex, marital status, and mortality. Am J Sociol. 1973;79:45-67.
    
13. Russo NF. Reconstructing the psychology of women: an overview. In: Notman MT, Nadelson CC, eds. Women and Men: New Perspective on Gender Differences. Washington, DC: American Psychiatric Press, Inc; 1991:43-61.
    
14. Hu YR, Goldman N. Mortality differentials by marital status: an international comparison. Demography. 1990;27:233-250.[Medline] [Order article via Infotrieve]
    
15. Osler W. Lectures on Angina Pectoris and Allied States. New York, NY: D. Appleton & Co; 1897.
    
16. Friedman M, Rosenman RH. Association of specific overt behavior pattern with blood and cardiovascular findings: blood cholesterol level, blood clotting time, incidence of arcus senilis, and clinical coronary artery disease. JAMA. 1959;169:1286-1296.
    
17. Rosenman RH, Friedman M. Association of specific blood pattern in women with blood and cardiovascular findings. Circulation. 1961;24:1173-1184.[Medline] [Order article via Infotrieve]
    
18. Haynes SG, Levine S, Scotch N, Feinleib M, Kannel WB. The relationship of psychosocial factors to coronary heart disease in the Framingham study, 1: methods and risk factors. Am J Epidemiol. 1978;107:362-383.[Abstract/Free Full Text]
    
19. Matthews KA, Haynes SG. Type A behavior pattern and coronary disease risk: update and critical evaluation. Am J Epidemiol. 1986;123:923-960.[Free Full Text]
    
20. Bem SL. The measurement of psychological androgyny. J Consult Clin Psychol. 1974;42:155-162.[Medline] [Order article via Infotrieve]
    
21. Spence JT, Helmreich RL. Masculinity and Femininity: Their Psychological Dimensions, Correlates, and Antecedents. Austin, Tex: University of Texas Press; 1978.
    
22. Helgeson VS. The role of masculinity in a prognostic predictor of heart attack severity. Sex Roles. 1990;22:755-774.
    
23. Powell KE, Thompson PD, Caspersen CJ, Kendrick JS. Physical activity and the incidence of coronary heart disease. Annu Rev Public Health. 1987;8:253-287.[Medline] [Order article via Infotrieve]
    
24. Paffenbarger RS Jr, Wing AL, Hyde RT. Physical activity as an index of heart attack risk in college alumni. Am J Epidemiol. 1978;108:161-175.[Abstract/Free Full Text]
    
25. Paffenbarger RS Jr, Hyde RT, Wing AL, Hsieh CC. Physical activity, all-cause mortality, and longevity of college alumni. N Engl J Med. 1986;314:605-613.[Abstract]
    
26. Blair SN, Kohl HW III, Paffenbarger RS Jr, Clark DG, Cooper KH, Gibbons LW. Physical fitness and all-cause mortality: a prospective study of healthy men and women. JAMA. 1989;262:2395-2401.[Abstract]
    
27. Blair SN. Evidence for success of exercise in weight loss and control. Ann Intern Med. 1993;119:702-706.[Abstract/Free Full Text]
    
28. Barrett-Connor E, Khaw KT. The epidemiology of DHEAS with particular reference to cardiovascular disease: the Rancho Bernardo Study. In: Kalimi M, Regelson W, eds. The Biologic Role of Dehydroepiandrosterone (DHEA). Berlin, Germany: Walter DeGruyter and Co; 1990:281-298.
    
29. Barrett-Connor E, Khaw KT, Yen SS. A prospective study of dehydroepiandrosterone sulfate, mortality, and cardiovascular disease. N Engl J Med. 1986;315:1519-1524.[Abstract]
    
30. Barrett-Connor E, Khaw KT. Absence of an inverse relation of dehydroepiandrosterone sulfate with cardiovascular mortality in postmenopausal women. N Engl J Med. 1987;317:711. Letter.[Medline] [Order article via Infotrieve]
    
31. Barrett-Connor E, Goodman-Gr