Adjusted survival curves showing the probability of surviving free of heart failure (nonfatal or fatal) according to baseline score on the Center for Epidemiological Studies Depression Scale (CES-D). Curves are adjusted for age; sex; race; history of myocardial infarction, diabetes, or angina; blood pressure; cholesterol levels; electrocardiographic abnormalities; disability; smoking; and Systolic Hypertension and the Elderly Program treatment status.
Abramson J, Berger A, Krumholz HM, Vaccarino V. Depression and Risk of Heart Failure Among Older Persons With Isolated Systolic Hypertension. Arch Intern Med. 2001;161(14):1725-1730. doi:10.1001/archinte.161.14.1725
Investigators have shown that depression is associated with an increased risk of coronary heart disease in general and myocardial infarction in particular. However, it is unknown whether depression, independent of its association with myocardial infarction, is a risk factor for heart failure.
This study examined whether depression was a predictor of incident heart failure among 4538 persons aged 60 years and older with isolated systolic hypertension who were enrolled in the Systolic Hypertension in the Elderly Program (SHEP). Depression was defined as a score of 16 or more at baseline on the Center for Epidemiological Studies Depression Scale (CES-D). The relationship between depression and heart failure was assessed using Cox proportional hazards regression.
The average follow-up was 4.5 years. Heart failure developed in 138 (3.2%) of 4317 nondepressed persons and in 18 (8.1%) of 221 depressed persons. After controlling for age; sex; race; history of myocardial infarction, diabetes, or angina; blood pressure; cholesterol levels; electrocardiographic abnormalities; smoking; disability; and SHEP treatment group, depressed persons had more than a 2-fold higher risk of developing heart failure compared with nondepressed persons (hazard ratio, 2.59; 95% confidence interval, 1.57-4.27; P<.001). After additional adjustment for the occurrence of myocardial infarction during follow-up, depressed persons remained at elevated risk of heart failure (hazard ratio, 2.82; 95% confidence interval, 1.71-4.67; P<.001).
Depression is independently associated with a substantial increase in the risk of heart failure among older persons with isolated systolic hypertension. This association does not appear to be mediated by myocardial infarction.
INVESTIGATORS have shown that depression is associated with an increased risk of developing coronary heart disease in general1- 3 and myocardial infarction (MI) in particular.3,4 Studies5- 8 have also shown that depression is a negative prognostic factor among persons with established coronary heart disease. Less is known about the relationship between depression and heart failure. One study9 reported an association between depression and higher death rates among persons with existing heart failure. However, prior investigations have not addressed whether depression predicts the onset of heart failure. Because depression is associated with increased MI risk and because MI is a major risk factor for heart failure, one would expect depression to be associated with an increased risk of developing heart failure. In addition, however, depression has been associated with excessive activation of the sympathetic nervous system,10,11 and heightened activation of this system is thought to be one of the processes involved in the pathogenesis of heart failure.12 As such, depression-induced sympathetic nervous system activation represents a separate pathway, not necessarily related to MI risk, by which depression could increase the risk of incident heart failure. If a link were established between depression and increased risk of heart failure, it would add to the growing evidence of depression's negative impact on cardiovascular health. In the present study, we examined whether depression was predictive of incident heart failure among older persons with isolated systolic hypertension enrolled in the Systolic Hypertension in the Elderly Program (SHEP).
SHEP was a randomized, double-blind, placebo-controlled, multicenter clinical trial designed to test the efficacy of antihypertensive therapy in reducing stroke incidence among older persons with isolated systolic hypertension.13 Major inclusion criteria were an age of 60 years or older, a baseline systolic blood pressure (SBP) of 160 mm Hg or higher, and a diastolic blood pressure (DBP) of less than 90 mm Hg. Major exclusion criteria were a baseline SBP of 220 mm Hg or higher and a recent episode (6 months before baseline) of a major cardiovascular event such as MI or stroke. In addition, persons with uncontrolled heart failure at baseline were excluded from SHEP. Investigators randomized 4736 participants who met these criteria to either stepped-care treatment or placebo. The drugs used for the active treatment were chlorthalidone, 12.5 mg increased to 25 mg (step 1), and, if necessary, atenolol, 25 or 50 mg (step 2). For persons whose baseline SBP was more than 180 mm Hg, the goal was to reduce the SBP to below 160 mm Hg; for persons whose baseline SBP was between 160 and 179 mm Hg, the goal was a reduction of at least 20 mm Hg in SBP. Participants were followed up for an average of 4.5 years. SHEP demonstrated that the active treatment resulted in a significant reduction in the incidence of stroke.14 Additional details about the design and conduct of SHEP can be found elsewhere.13,14
Uncontrolled heart failure was one of the SHEP exclusion criteria. However, compensated heart failure was not one of the exclusion criteria. According to information gathered from extensive physical examinations made at baseline, physicians determined that 16 SHEP participants had had compensated heart failure within 1 year before the start of the study. These 16 persons were excluded from our analyses. We additionally excluded 182 persons with missing data on depression or any of the other variables used in our analyses. After all exclusions, the sample size for the present study was 4538 persons.
At baseline, depression was assessed by the Center for Epidemiological Studies Depression Scale (CES-D).15 The CES-D is a 20-item self-report scale designed to measure depressive symptoms in the general population. Each item assesses how frequently a participant experiences a particular depressive symptom. A given item is scored on a 4-point scale ranging from 0 to 3. Scores from the individual items are then summed to arrive at a total CES-D score that can range from 0 to 60, with higher scores representing higher levels of depressive symptoms. For those with 1 to 3 missing CES-D items (n = 103), we imputed total CES-D scores by calculating the average score on the nonmissing items and then multiplying this average score by 20. Participants with more than 3 missing items on the CES-D were excluded from our analyses. For purposes of this study, depression was defined as a score of 16 or more on the CES-D at baseline. This score has been shown to provide a reasonable approximation to a clinical diagnosis of depression.15,16
The outcome in this study was fatal or nonfatal heart failure. Some SHEP participants experienced multiple heart failure events during follow-up, but we only considered the first event in our analyses. When a suspected cardiovascular or cerebrovascular event occurred in SHEP, staff at participating centers forwarded relevant clinical information (including electrocardiographic [ECG] data, cardiac enzyme levels, chest x-ray reports, and death certificates or autopsy reports) to a coordinating center so that a coding panel could adjudicate diagnoses. The coding panel included 3 physicians, at least 1 of whom was a cardiologist. For a diagnosis of heart failure, the coding panel required 1 of the following to be present: paroxysmal nocturnal dyspnea, dyspnea at rest, orthopnea, or New York Heart Association classification III cardiac disease. In addition, 1 of the following criteria also needed to be present: rales, ankle edema (2+ or greater), tachycardia (heart rate of ≥120/min after 5 minutes of rest), S3 gallop, jugular venous distension, or radiographic evidence of heart failure.14 Diagnoses of heart failure were not made if the symptoms or signs listed herein were due to pulmonary conditions such as pneumonia or chronic obstructive pulmonary disease.
A number of baseline factors were included as control variables in our analyses. These factors included age, sex, race (white vs nonwhite), history of MI or angina as determined by a physician, self-reported history of diabetes, SBP and DBP, total cholesterol and high-density lipoprotein cholesterol (HDL-C) levels, and presence of ECG abnormalities. We also controlled for self-reported smoking (current vs former or never), presence of any disability in activities of daily living,17 and SHEP treatment status (active treatment vs placebo).
Another control variable that we considered was the occurrence of MI during follow-up. Of the MIs that occurred during follow-up, we only controlled for those that happened before a heart failure event. As with heart failure, a coding panel of 3 SHEP physicians diagnosed MIs during follow-up on the basis of clinical information gathered by SHEP staff. A diagnosis of MI during follow-up required the presence of typical symptoms associated with acute MI and either typical ECG changes (including new Q waves) or enzyme levels at least 1.25 times greater than normal.
The first step in the analysis was to examine bivariate associations between the dichotomous depression variable under study and the baseline control variables. The t test was used for continuous control variables and the χ2 test was used for categorical control variables. We used a series of Cox proportional hazards regression models to examine the association between baseline depression and subsequent risk of heart failure after adjusting for baseline control variables. Time to first heart failure event was the outcome of interest in these Cox models. The proportional hazards assumption was checked by including a time-dependent covariate representing the interaction between depression and follow-up time. A nonsignificant P value for this covariate (P>.20) was taken as evidence that the proportional hazards assumption had been satisfied. Age, SBP, DBP, total cholesterol level, and HDL-C level were entered as continuous variables in the Cox models. Other control variables were entered according to the categories noted herein.
We also performed a number of additional analyses. First, we examined the association of depression with heart failure after adjusting for MI events that occurred subsequent to baseline. This analysis was accomplished by running a Cox model that included a time-dependent covariate for the occurrence of MI during follow-up. Second, we considered the possibility that the relationship between depression and heart failure varied according to sex and SHEP treatment group, respectively. To assess this possibility, we ran models that included appropriate interaction terms. Third, because SHEP participants completed the CES-D scale every 6 months during follow-up, we were able to assess how updated CES-D values were related to heart failure risk. To do this, we ran a model that included a time-dependent CES-D variable. Fourth, some investigators have noted that depressive symptoms exert a graded effect on cardiovascular outcomes.4,18 To evaluate this possibility, we divided baseline CES-D scores into 3 levels: low (0-7), medium (8-15), and high (≥16). We then included these levels in a Cox model with the 0 to 7 group as the referent group. These levels were chosen because they were thought to represent clinically meaningful differences in CES-D scores, with the highest group being consistent with clinically diagnosed depression.
Of the 4538 persons included in this study, 4317 were free of depression at baseline, and 221 were depressed according to the CES-D. Persons who were depressed were significantly less likely to be male (P<.001) and more likely to be nonwhite (P<.001) (Table 1). Depressed persons had significantly higher SBP readings (P = .01) and higher HDL-C levels (P<.01), but there were no significant differences in history of MI or diabetes.
During follow-up, heart failure events occurred in 138 (3.2%) of the 4317 nondepressed persons and in 18 (8.1%) of the 221 depressed persons. The crude relative risk indicated that depressed persons were significantly more likely than nondepressed persons to develop heart failure during follow-up (relative risk, 2.55; 95% confidence interval [CI], 1.59-4.08; P<.001). Table 2 shows that after adjustment for demographic factors (model 1), the association between depression and heart failure incidence was still strong and significant (hazard ratio [HR], 2.68; 95% CI, 1.63-4.42; P<.001). In model 2, which additionally adjusted for history of MI, diabetes, and angina; SBP and DBP; lipid levels; ECG abnormalities; current smoking; and treatment status, depression continued to be a predictor of increased heart failure risk (HR, 2.59; 95% CI, 1.57-4.27; P<.001). In both of these models, the time-dependent interaction between depression and follow-up time was not significant (P>.60), suggesting that the proportional hazards assumption was satisfied. Thus, the effect of depression on heart failure risk was fairly constant during the study and was not confined to a particular period, such as the early part of the follow-up. To further verify that depression was not simply associated with early heart failure events (therefore suggesting that depression assessed at baseline might be secondary to subclinical heart failure), we refitted model 2 after excluding heart failure events that occurred within the first year of follow-up. In this refitted model, depression was still strongly associated with an increased risk of heart failure (HR, 2.54; 95% CI, 1.39-4.66; P = .002). Figure 1 shows the adjusted probability of surviving free of heart failure, according to baseline depression status.
In additional analyses, we found no evidence that the association between depression and heart failure was mediated by incident MI. We ran a fully adjusted Cox model (adjusting for all of the variables in model 2 of Table 2) that included a time-dependent covariate for incident MI. In this model, incident MI was an overwhelmingly strong risk factor for heart failure (HR, 22.68; 95% CI, 15.92-34.49; P<.001). Nevertheless, the model showed that depression was still associated with a significant increase in the risk of heart failure; the association actually appeared slightly stronger than in previous models (HR, 2.82; 95% CI, 1.71-4.67; P<.001).
The association between depression and heart failure did not vary significantly according to sex or treatment status. A fully adjusted Cox model that included a term for the interaction between depression and sex showed that the effect of depression was somewhat lower in women compared with men, but the interaction was not significant (for women: HR, 2.05; 95% CI, 1.05-4.00; for men: HR, 3.68; 95% CI, 1.76-7.70; P = .25 for the interaction between depression and sex). A fully adjusted model that included a term for the interaction between depression and treatment group indicated that the effect of depression in the placebo and treatment groups was similar (for the placebo group: HR, 2.67; 95% CI, 1.45-4.92; for the treatment group: HR, 2.45; 95% CI, 1.04-5.76; P = .92 for the interaction between depression and treatment).
To determine the role of depression over time on heart failure risk, a dichotomous CES-D variable (<16 vs ≥16) was modeled as a time-dependent variable, which took into account longitudinal assessments of the CES-D throughout the follow-up. This variable proved to be a significant predictor of heart failure, even after full adjustment for all of the control variables (HR, 2.00; 95% CI, 1.21-3.31; P = .007). However, the effect was somewhat smaller than it was for the baseline dichotomous CES-D variable. We also used the follow-up CES-D scores to calculate change in CES-D scores from baseline. When modeled as a time-dependent covariate, we found that a change (increase) in CES-D scores from baseline was a risk factor for heart failure. However, baseline depression remained a significant predictor of heart failure (HR, 3.12; 95% CI, 1.76-5.52; P<.001), even when the increase in CES-D score from baseline and other variables were accounted for.
Finally, we examined whether CES-D scores exerted a graded effect on the risk of heart failure. As noted earlier, we trichotomized baseline CES-D scores as 0 to 7 (n = 3697), 8 to 15 (n = 620), and 16 or higher (n = 221). After adjustment for all of the control variables, we found that those in the 8 to 15 group did not have a higher risk of heart failure compared with those in the 0 to 7 group (HR, 0.75; 95% CI, 0.46-1.22; P = .24). However, those in the 16 or higher group had a significantly higher risk of heart failure compared with the subjects in the 0 to 7 group (HR, 2.46; 95% CI, 1.48-4.07; P<.001).
The main finding of this study was that depression, as indicated by a score of 16 or more on the CES-D, was associated with a substantial and significant increase in the risk of developing heart failure during follow-up among older persons with isolated systolic hypertension enrolled in SHEP. This association was observed after controlling for a number of factors, including age; sex; race; history of MI, diabetes, or angina; SBP and DBP; total cholesterol and HDL-C levels; ECG abnormalities; smoking; disability; and SHEP treatment status. Others have reported that the prevalence of depression is higher in persons with heart failure19 and that depression is associated with an increased risk of death among persons with existing heart failure.9 To our knowledge, our investigation is the first study to establish an independent association between depression and the subsequent development of heart failure.
In addition to the main result noted, we conducted a number of additional analyses. First, we found that baseline depression was associated with a higher risk of heart failure, even when the occurrence of MI during follow-up was controlled for. Second, we failed to find evidence that the association between depression and heart failure significantly varied according to sex. This is in contrast to some prior studies of depression and cardiovascular outcomes, which have reported significant interactions between depression and sex.18,20 Third, we failed to observe a graded association between depressive symptoms and heart failure. Instead, our findings suggested that depressive symptoms increase heart failure risk only when those symptoms reach a sufficiently high level (ie, a level considered an approximation of a clinical diagnosis of depression). Other studies, however, have reported graded associations between depression and cardiovascular outcomes such as coronary heart disease.4,18 Fourth, we analyzed how depression levels during follow-up were related to heart failure risk. An earlier study of the SHEP data reported that a change (increase) in the CES-D score from baseline, modeled as a time-dependent covariate, was associated with a higher risk of MI and stroke.21 We found that this time-dependent covariate was also associated with increased heart failure risk, although baseline depression continued to predict heart failure when this time-dependent covariate was accounted for. We also found that a time-dependent dichotomous (<16 vs ≥16) depression variable was associated with higher heart failure risk, but the association was somewhat weaker than it was for the baseline dichotomous depression variable. Thus, compared with updated measures of depression over time, a single measure of depression at baseline in SHEP was as good or better at predicting heart failure.
What could account for the association between depression and increased heart failure risk observed in this study? One possible explanation for our findings is that, before baseline, heart failure in a subclinical or compensated stage led to depression, and then the subclinical disease later manifested itself as clinical disease during follow-up. Such a sequence of events would make it seem as if depression led to heart failure when in fact the opposite would be true. If this were the case, one might expect that the relation between baseline depression and subsequent heart failure would primarily appear during the early part of the follow-up period. However, our results indicated that the effect of depression was not confined to the early part of follow-up but was, instead, relatively constant throughout the study. Thus, our results do not appear to be attributable to a scenario in which subclinical heart failure led to depression.
A second possible explanation is that high-level depressive symptoms led to MI events and that the latter subsequently led to heart failure. In theory, this explanation seems plausible because there is evidence that depression may increase the risk of MI,4 and MI is, in turn, a major risk factor for heart failure. However, we found that depression predicted increased heart failure risk after controlling for a history of MI and the occurrence of MI during follow-up. Thus, MI does not appear to mediate the association we observed between depression and increased heart failure risk.
Factors known to differ between depressed and nondepressed persons at baseline may provide a third possible explanation for our findings. For example, we found that depressed persons were significantly more likely to be female and nonwhite. We also found that depressed persons had higher SBP levels and greater limitations in activities of daily living at baseline. After controlling for these factors though, we still found a strong association between depression and a higher risk of heart failure. Thus, unless there was residual confounding due to imprecise control of these factors, such factors do not appear to account for our results.
Fourth, our results may be due to factors that we did not have any information about. We did not, for example, have any information on the duration of systolic hypertension or the duration of medical therapy before baseline. Compliance with medication regimens is another important factor that was not included in our analyses. Evidence indicates that depressed persons may be less likely to comply with medical advice,22 and such noncompliance could be a potential explanation for our results. Finally, we did not have information on sympathetic nervous system activity in our study. Depressed persons tend to show heightened activation of the sympathetic nervous system.10,11 In addition, heightened activation of this system is believed to play an important role in the pathogenesis of heart failure.12,23- 25 Thus, depression-induced sympathetic nervous system activation is a plausible physiological pathway that could account for our findings.
The present study had a number of strengths. Heart failure events were ascertained prospectively according to rigorous criteria, and all diagnoses were confirmed by a panel of 3 physicians. In addition, all subjects underwent an extensive medical examination at baseline, which allowed accurate assessment of medical history and health status. Thus, we were able to control for a number of other potential risk factors for heart failure, such as MI, blood pressure, cholesterol levels, and smoking.
However, this study also had some weaknesses. Our measure of depression was based on a CES-D score of 16 or more. The CES-D is a scale that assesses depressive symptoms, and a score of 16 or more on the CES-D is not necessarily equivalent to a clinical diagnosis of depression. Clinical diagnoses may have provided a more precise method of measuring our exposure of interest. However, previous studies15,16 have indicated that a score of 16 or more on the CES-D is a reasonable approximation of a clinical diagnosis of depression. The CES-D is one of the most commonly used depression scales in epidemiologic studies because it is shorter and easier to administer in population studies than are structured clinical interviews for mental disorders used in clinical settings. In this respect, our study follows the approach of previous epidemiologic studies2,18,20 that have linked depression to an increased risk of coronary heart disease. A second weakness of our study was that our study population was a select group of people who all had isolated systolic hypertension and had agreed to participate in a clinical trial. Isolated systolic hypertension is fairly common among elderly persons though,26 so our results may have fairly broad applicability. Nevertheless, future studies should confirm our findings in more general, community-based settings. A third weakness was that our results were based on a relatively small number of events (n = 18) in the depressed group. Although this did not prevent us from detecting a significant association for the main effect of depression, it may have precluded detection of significant interactions between depression and other variables, such as sex. A fourth weakness is that we did not have information on whether the heart failure events were due to systolic or diastolic ventricular dysfunction, and, therefore, we were unable to determine the association between depression and these different subtypes of heart failure.
In summary, our results indicate that depression is associated with an increased risk of heart failure among older persons with isolated systolic hypertension. This association is not explained by demographic characteristics, baseline health status and medical history, or MI risk. This study adds to the growing evidence that depression increases the risk of adverse cardiovascular events.
Accepted for publication January 11, 2001.
Corresponding author and reprints: Jerome Abramson, PhD, Emory Center for Outcomes Research, Emory West, 1256 Briarcliff Rd NE, Suite 1 N, Atlanta, GA 30306 (e-mail: email@example.com).