Gender differences in advanced heart failure: insights from the BEST study

Statistical analysis

Gender and treatment group comparisons were conducted. For continuous variables, the ttest was used except when the data were non-normal. This was the case with plasma norepinephrine (PNE), and Wilcoxon rank-sum test was used. For categorical variables, the chi-square test was used. The log-rank test was used to compare survival distributions (mortality rates) by gender and treatment group. Kaplan-Meier methods were used to construct survival curves. Cox regression models were used to examine the effect of covariates of interest on overall survival and in gender subgroups, and to estimate hazard ratios and 95% confidence intervals. Univariate analysis of associations with mortality were conducted within gender for each hypothesized predictor, followed by multivariate analyses within gender, with reduction of variables by backwards elimination of those found not significant at the p = 0.05 level (all models were adjusted for randomized treatment group assignment). Cox regression analysis was also used to examine potential interactions between gender, presence of CAD, and treatment group. The p values are reported from the paired ttest comparing treatment response from baseline, at 3 and 12 months, and overall mean estimates, respectively.

A type I error of 0.05 was used to denote statistical significance, and p values reported are unadjusted.

Baseline characteristics

Baseline characteristics are presented in Table 1. Women were younger than men, were more likely to be black, had a higher prevalence of nonischemic etiology, higher left and right ventricular ejection fraction (RVEF), higher heart rate, greater cardiothoracic ratio (CTR), higher blood urea nitrogen/creatinine ratio, higher prevalence of left bundle branch block (LBBB), lower prevalence of atrial fibrillation (Afib), current and past smoking, lesser use of anticoagulants and aspirin, and lower PNE compared with men.

Response to treatment

In women receiving bucindolol, a clinically significant slowing of the heart rate was noted at three months from 83 (beats/min) to 75 (beats/min) that was sustained at 12 months (p < 0.05). The slowing of the heart rate was significantly greater with bucindolol compared with placebo (p < 0.0001). Both RVEF and LVEF improved at three and 12 months in the study participants. In the bucindolol group, LVEF increased from 25 ± 7% at baseline, to 33 ± 13% at 12 months (p < 0.05), which was significantly higher than LVEF in the placebo group measured at 12 months, 28 ± 12% (p = 0.0004). Right ventricular ejection fraction was measured at 38 ± 14% at baseline, 42 ± 15% at 12 months in the placebo group (p ≤ 0.05). In the bucindolol group, the corresponding figures were 37 ± 14% and 47 ± 13% (p = 0.05), respectively. Right ventricular ejection fraction was significantly higher at 12 months in the bucindolol group (p = 0.0006).

Survival

No improvement in survival was noted in the bucindolol group compared with placebo (mortality 26% vs. 29%, respectively, p = 0.44). This finding was consistent in all subgroups regardless of etiology, ethnicity, or the presence of diabetes.

Analyzing crude mortality by gender demonstrated lower mortality in women compared with men (27% vs. 33%, respectively, p = 0.02) due to the significant difference in the nonischemic compared with ischemic group (19% vs. 27%, respectively, p = 0.009) (Table 2). New York Heart Association class III women had lower overall mortality than men (25% vs. 32%, respectively, p = 0.004), and a lower mortality in nondiabetic women compared with nondiabetic men (25% vs. 31%, respectively, p = 0.03) was noted as illustrated in Figure 1, which also displays the hazard ratios by gender for the prespecified variables.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Hazard ratios for mortality comparing genders for prespecified patient subgroups adjusted for treatment group. CAD = coronary artery disease; LVEF = left ventricular ejection fraction; NYHA = New York Heart Association.

Variables related to prognosis

The following preselected clinical characteristics and laboratory values were related to prognosis in a Cox proportional hazards regression model: NYHA functional class (class IV vs. III), CAD (presence vs. absence of CAD), ethnicity (black vs. non-black), age, body mass index (BMI), diabetes history, systolic blood pressure, LVEF, CTR, QRS duration, Afib (presence vs. absence), blood urea nitrogen/creatinine ratio, and treatment. Regression analyses were conducted separately for women and men.

In women, all the above-mentioned variables were found to be univariate predictors of survival with the exception of ethnicity, QRS duration, and treatment (Table 3). A parallel analysis was carried out in men (Table 4) with similar results. QRS duration, but not Afib, was found to be a univariate predictor of survival in men.

Ischemic etiology and measures of the severity of HF were found to be predictors of survival in multivariate regression models in both women and men (Tables 5 and 6). ⇓⇓These include CAD, LVEF, systolic blood pressure, CTR, and blood urea nitrogen/creatinine ratio. Coronary artery disease and LVEF appeared to be stronger predictors of prognosis in women (Table 5). Atrial fibrillation and NYHA were predictors in women, but not in men. QRS duration, BMI, age, heart rate, diabetes history, and treatment (bucindolol vs. placebo) were found to be predictors of survival in men only (Table 6).

A multivariate model (Table 7) comparing survival of females to males and adjusting for major risk factors indicates that gender is a significant predictor of outcome (p = 0.0046). There was a significant CAD by gender interaction (p = 0.011). The effect of gender on survival is modified by etiology (ischemic vs. nonischemic). In the nonischemic group, women have a better survival than men (p = 0.0093); however, in the ischemic group, there was a trend for a better survival in men (p = 0.1516) (Fig. 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Survival curves by gender within ischemia status. CAD = coronary artery disease.

Baseline characteristics and prognostic variables

Several baseline clinical and laboratory features were found to differ significantly between women and men. Some characteristics are expected to confer better prognosis, namely having a higher prevalence of nonischemic etiology, higher LVEF (11), lower occurrence of Afib (12), and lower PNE (11). Some other features, however, have been related to worse outcome including higher heart rate (13), CTR (11), and higher prevalence of LBBB (14).

A review of prior studies addressing gender differences in patients with advanced HF reveals certain common features in women consistent with our findings including a lower percentage of ischemic etiology (15–18)a higher percentage of African Americans (15,16), a higher heart rate (16,18), LVEF (15,18), CTR (19), higher prevalence of LBBB (17,19), and lower prevalence of Afib (15,18).

A more favorable outcome among women may theoretically be explained on the basis of the effects of sex hormones. Women with late menopause were found to have larger end-systolic volumes, lower LVEF, and lower filling rates as compared with early menopausal women (20). Similarly, premenopausal women have lower blood pressure compared with men of similar age, a pattern that is no longer seen after menopause (21). Estrogen decreases endothelin levels in postmenopausal women (22), and estrogen replacement in perimenopausal women results in reduction of systolic and diastolic blood pressure, and in total body norepinephrine spillover in response to mental stress (23). The potential vasodilator effect of estrogen may be mediated through the renin-angiotensin system, bradykinin, or nitric oxide (24–26). Moreover, sex hormones exhibit a favorable effect on hemostasis and thrombolysis. Significantly lower levels of plasminogen activator inhibitor were found in premenopausal women as compared with postmenopausal women or to men of similar age (27).

Gender-related differences in geometric remodeling and earlier onset of impaired LV systolic function in males was noted in animal models (28)as well as in humans with aortic stenosis (29)and hypertension (30). Myocyte cell loss pattern favoring females (31)and gender differences in gene expression have also been reported (32), and, very recently, reduced hypertrophy in women in postinfarction remodeling was speculated to reflect fundamental differences in cellular remodeling (33).

However, a more likely explanation is that these variables were derived from studies that have enrolled predominantly men. Our data indicate a differential prognostic value for some characteristics. For example, Afib was found to be a predictor in women but not in men. On the other hand, QRS duration and BMI were predictors only in men. Because of the relatively limited sample size, one should be cautious not to over-interpret these differences. However, two interesting findings should be mentioned. For every 1% increase in LVEF, there was a corresponding 4% decrease in mortality in women compared with a 1% decrease in men. Similarly, CAD conferred a 2.5-fold increase in the risk of mortality in women compared with a 1.5-fold increase in men. Thus, LVEF and CAD appear to be stronger predictors of prognosis in women.

Comparison with previous survival studies

A study in a young population of 65 women and 238 men with idiopathic dilated cardiomyopathy who were enrolled consecutively found no difference in survival rates between men and women (19). Adams et al. (15)reported their experience with 177 women and 380 men with HF and impaired left ventricular systolic function consecutively enrolled in an outpatient HF clinic. They found a significantly better survival rate for women with nonischemic etiology. Similar findings were reported from the Flolan International Randomized Survival trial, in 112 women and 359 men (16). Recently, Simon et al. (17)reported the survival rates in 515 female and 2,132 men in the Cardiac Insufficiency Bisoprolol study. Although they confirmed better survival for women compared with men, this difference was predominantly noted in the undefined etiology, and no difference in survival was seen between women and men in the nonischemic group. Likewise, in Metoprolol CR/XL Randomized Intervention Trial In Heart Failure, women had a 37% lower risk of dying than men after adjusting for baseline differences including ischemic etiology (18). Findings from BEST, however, indicate that the survival advantage of women was confined to the nonischemic patients. Therefore, one has to conclude that the survival advantage of women cannot be explained entirely on the basis of higher prevalence of nonischemic etiology. It is likely that other confounders not identified by measured baseline clinical and laboratory characteristics play a role.

Study limitations

Despite the attempt to maximize enrollment of women and to collect all pertinent information relevant to gender differences, this effort is by no means complete. The sample size is limited, and there are several factors that have not been measured. For example, gender-related differences in pharmacokinetics have been identified for a number of drugs including beta-blockers (34), and socioeconomic status has not been addressed. We consider our findings exploratory in nature and hypothesis-generating.

Conclusions

Major baseline differences in clinical and laboratory characteristics exist between men and women. The prognostic role of various predictors may vary in magnitude between genders, and the survival advantage of women in our study was confined to the nonischemic etiology. These data indicate that information collected on men with HF cannot be assumed to apply similarly or with the same magnitude to women, emphasize the importance of accounting for these differences in designing clinical trials, and highlight the need for both stratifying entry by gender and enrolling more women in HF trials.