Original Research

Sex-Based Differences in Causes of Hospitalization for Coronary Heart Disease.
By Arthur L Klatsky, MD; Mary Anne Armstrong, MA

Introduction
Possible sex-based disparities with respect to atherosclerotic coronary heart disease (CHD) have become a subject of intense research and debate. Areas of interest include epidemiologic or clinical aspects,^1-4 differences in risk factors,^2,3,5-9 problems of diagnostic evaluation,^10-12 disparities in either treatment or prognosis,^2,3,9,13-19 and prevalence of nonatherosclerotic coronary syndromes or of noncardiac syndromes that mimic CHD problems.^10,20,21 Several of these areas of interest may have implications pertinent to various forms of bias and raise the possibility of biological, sex-based differences in manifestations of this most prevalent cardiovascular disease.

The Framingham Heart Study² presented prospective population data showing possible sex-based differences in initial clinical manifestation of CHD. These data showed that the initial CHD manifestation in men was more likely to be acute myocardial infarction (MI) or sudden cardiac death, whereas CHD in women was more likely to initially manifest as stable or unstable angina. In the Framingham cohort, clinical CHD developed in women a decade later than in men, and the prognosis for CHD was worse in women than in men. Similar findings have been reported in several clinical studies^1,3,5 that have speculated that these differences in CHD manifestations and prognosis may be explained by sex differences in risk factors.

We here report prospective Kaiser Permanente (KP) data that show substantial, sex-based differences in clinical CHD hospitalization diagnoses. We also report data about risk-related traits and later mortality among patients hospitalized for CHD.

Methods
Subjects and Data
Baseline data were obtained from results of voluntary health examinations given to 128,934 KP members in San Francisco and Oakland, California from 1978 through 1985. The examination included queries about sociodemographic status, habits, medical history and symptoms, and health measurements.²²

Ascertainment of Hospitalization for CHD
Subjects were monitored until December 31, 1991; until death; until termination of health plan membership; or until first CHD-related hospitalization at a KP facility in Northern California. Duration of patient follow-up totaled 889,611 person-years for the 128,934 persons in the study population. Primary diagnosis for 3931 persons (63% men) hospitalized for CHD included acute MI (ICD-9 code 410), "other acute" CHD syndromes (mainly "unstable angina," ICD-9 code 411), "angina pectoris" (ICD-9 code 413), and "chronic ischemic heart disease" (ICD-9 codes 412 or 414) (Table 1).

Analytic Methods
The Cox proportional hazards model was used for age-adjusted and multivariate analyses. Most multivariate models included age, sex, race, education, marital status, body mass index, cigarette smoking, and a composite "CHD risk" variable consisting of an affirmative response to any of 12 baseline items pertaining to CHD risk or symptoms.^23,24 Some models also included systolic blood pressure, blood glucose level, and total blood cholesterol level. We also studied subsequent mortality in relation to first CHD hospitalization diagnosis by using an automated matching system²⁵ to screen for deaths in California through 1994. This result was validated by manual review conducted by two investigators (MAA, ALK).

Results
Table 2 shows demographic characteristics and selected traits of persons hospitalized for any CHD-related diagnosis. On average, women were older than the men by about three years. There was a substantially greater proportion of black women than black men. In general, women were more likely to have CHD-related risk traits, including smoking, lifelong alcohol abstinence, the CHD risk variable, and history of diabetes (Table 2). This higher prevalence of CHD-related traits among women was generally similar for the four subsets of CHD-related clinical diagnoses. These traits were all included among covariates used in the multivariate analyses.

Among patients who had any CHD-related diagnosis, male/female relative risks for age-adjusted and multivariate analyses were identical (RR = 2.0; 95% CI = 1.9-2.1; p < 0.001). Table 3 shows age-adjusted and multivariate relative risk for the four subsets of CHD-related clinical diagnoses. Each subset shows a preponderance of men at p < 0.001. However, the data showed a greater male-female disparity (p < 0.001) between 1) risk of hospitalization for either acute MI (RR = 2.7) or chronic ischemic heart disease (RR = 3.3) and 2) risk of hospitalization for other acute syndromes (RR = 1.5) or angina (RR = 1.6). Very small differences were observed between age-adjusted and multivariate odds ratios. Independence of these disparities from race was confirmed by observation that the male-female difference in risk was similar for each racial group represented--white, black, and Asian (data not shown).

Although the established CHD predictors were generally more strongly related to CHD risk in women, the strength of this relation differed little across the subsets of CHD-related clinical diagnoses (Table 4). Age, body mass index, and the composite CHD risk variable were strong predictors in both sexes; for each clinical diagnostic subset, the composite CHD risk variable was a stronger predictor in women than in men. The stronger relation between smoking and CHD diagnoses in women was present in all diagnostic subsets except angina; and for all diagnostic subsets, the inverse relation between alcohol use and CHD diagnoses was slightly stronger in women than in men. Total blood cholesterol level was strongly related to all diagnostic subsets in both sexes. Systolic blood pressure and blood glucose level showed the expected relations to diagnostic subsets in both sexes with no major disparities except that black men were at substantially lower risk than white men for acute MI and chronic ischemic heart disease, whereas black women were at lower risk for acute MI and other acute syndromes. Men were the only group in which the data showed a substantial inverse relation between higher educational attainment and acute MI or angina.

Data obtained as late as ten years after hospitalization showed identical mortality rates for men and women: 29% from all causes and 16% from CHD. Women initially diagnosed with acute MI or chronic ischemic heart disease were 15% more likely than men to die of any cause or of CHD, whereas women who were initially hospitalized for stable or unstable angina were 20% less likely than men to die of any cause or of CHD. However, after adjustment for age, this sex-based difference in mortality was significant at p < 0.05 for comparison of any two diagnostic subsets in any combination. For each diagnostic subset, multiple hospitalizations for CHD occurred slightly more in men (46%) than in women (39%).

Eight women and two men were hospitalized for Prinzmetal's variant angina (ICD-9 code 413.1) or Syndrome X (ICD-9 code 413.9), and 2620 persons (47% of whom were women) were hospitalized for nonspecific chest pain syndromes (ICD-9 code 786.5)

Comments
These prospective hospitalization data are similar to the Framingham Study CHD incidence data² in suggesting that CHD in men is more likely to manifest initially as acute MI, whereas CHD in women is more likely to manifest initially as angina. Relatively high rates of CHD manifesting in women initially as acute MI and nonacute MI syndromes were also reported in several clinical series.^1,3,5,21 Although this finding has been accepted as a true difference,²⁶ possible selection bias should be examined. Selective gender-related admission would seem unlikely to influence hospitalization for acute MI, but hospitalization for diagnostic evaluation of chest pain could be influenced by sex-related bias about perceived likelihood of CHD. That 53% of patients hospitalized for nonspecific chest pain were men is indirect evidence against the possibility that hospital admissions for stable or unstable angina were substantially biased with respect to sex. The preponderance of men diagnosed with chronic ischemic heart disease is more problematic and difficult to evaluate, because inclusion of some patients in this mixed group represented surgery or other procedures, whereas inclusion of other patients represented sequelae of CHD, such as heart failure and arrhythmia. The relatively high preponderance of men diagnosed with chronic ischemic heart failure is therefore likely to represent a combination of higher male risk for sequelae of acute MI (eg, heart failure) and a greater likelihood that men with any CHD diagnosis will undergo interventional procedures (eg, surgery).

Our finding that standard coronary risk factors were more prevalent in women hospitalized for CHD than in men hospitalized for CHD is generally compatible with other published findings.^2,3,5-9 However, our findings showed no consistent sex-related disparity for the subsets of clinical diagnoses we studied. In addition, the small differences between age-adjusted and multivariate data shown in our study (Table 3) suggest independence from the covariates as a group. Thus, our data do not support the hypothesis that risk factor disparities explain sex differences in initial clinical manifestation of CHD.

Could the explanation be that a larger proportion of women admitted for a non-MI, CHD-related diagnosis have nonatherosclerotic syndromes and that these syndromes are erroneously coded as atherosclerotic CHD? Women are more likely to have Prinzmetal variant angina²⁷ and Syndrome X²⁸; and eight of the ten patients so diagnosed in our study were women. If such dilution of the group of women as acute coronary syndromes and angina were substantial, these diagnoses could reasonably be expected to affect follow-up mortality data, because nonatherosclerotic syndromes have a better prognosis than atherosclerotic disease. As already stated, survival in women was not statistically significantly different than for men in any subset of CHD-related clinical diagnoses.

Our age-adjusted data for mortality rates differ from those in some other studies.^2,13,19,26 In those studies, women diagnosed with acute MI were substantially older than men compared with our subjects, and a greater proportion of these women than men were diagnosed with Q-wave infarctions, heart failure, and cardiogenic shock. Death among men, in contrast, was more likely to occur outside the hospital setting.

Could this disparity in CHD-related diagnoses for men and women indicate a biological sex difference in coronary atherosclerosis? Data relevant to this speculation are sparse. In one report,²⁹ CHD plaques in eight young women with fatal CHD were compared with CHD plaques in older men and women with fatal CHD. Compared with the older men and women, these young women had a substantially higher percentage of cellular fibrous tissue and lipid-rich foam cells and lesser amounts of dense, fibrous calcified tissue; this result was interpreted as suggesting greater potential reversibility of plaque formation. However, a study using intravascular ultrasound analysis³⁰ failed to show any quantitative or qualitative difference in coronary atherosclerosis between men and women. Other reports^31,32 suggest different effects of male and female hormones on the interaction of lipid-laden macrophages and endothelial damage. The pertinence of these reports to a sex-based difference in clinical manifestation of CHD is unclear. The possibility of a biological, sex-related difference in atherosclerotic plaque composition remains hypothetical.

Effects of estrogenic hormones are the most obvious explanation for a biological, sex-based difference in CHD manifestations. Although the role of these hormones in prevention of CHD has become highly controversial, estrogenic substances clearly could affect atherosclerotic mechanisms--not only via lipid effects but also via effects on endothelial function and vascular injury response.^26,33 In addition, a sex-based difference (possibly hormonally mediated) may also be inherent in thrombophilic or spontaneous antithrombotic tendencies,^34-37 and this difference might affect the relative likelihood of acute MI vs other CHD syndromes. We did not ascertain which patients in our study were receiving hormone replacement therapy at first hospitalization for CHD.

With respect to first CHD hospitalization, we conclude: 1) among persons hospitalized for CHD-related diagnoses, men are more likely to be diagnosed with acute MI, whereas women hospitalized for CHD-related diagnoses are more likely to have unstable or stable angina, 2) this biological, sex-based disparity is independent of established risk traits for CHD, 3) no explanation for this independence is clear, and 4) a biological, sex-based difference in atherosclerotic plaque composition, endothelial function, or thrombotic tendency is likely.

Previous Presentation: Portions of this material were presented in Abstract form at the 46th Annual Session of the American College of Cardiology in Anaheim, CA, March 1997.

Acknowledgments: This research was supported by grants from the National Institute on Alcohol Abuse and Alcoholism (NIAAA) (1 R01 AA 10830-01); the Alcoholic Beverage Medical Research Foundation; and the Direct Community Investment Benefit Program.

We thank Cynthia Landy for data collection, Harald Kipp for programming assistance, and Sally McBride Allen for technical assistance.

To Original Research Index >>