Background We previously reported that major depression in patients in the hospital after a myocardial infarction (MI) substantially increases the risk of mortality during the first 6 months. We examined the impact of depression over 18 months and present additional evidence concerning potential mechanisms linking depression and mortality.
Methods and Results Two-hundred twenty-two patients responded to a modified version of the National Institute of Mental Health Diagnostic Interview Schedule (DIS) for a major depressive episode at approximately 7 days after MI. The Beck Depression Inventory (BDI), which measures depressive symptomatology, was also completed by 218 of the patients. All patients and/or families were contacted at 18 months to determine survival status. Thirty-five patients met the modified DIS criteria for major in-hospital depression after the MI. Sixty-eight had BDI scores ≥10, indicative of mild to moderate symptoms of depression. There were 21 deaths during the follow-up period, including 19 from cardiac causes. Seven of these deaths occurred among patients who met DIS criteria for depression, and 12 occurred among patients with elevated BDI scores. Multiple logistic regression analyses showed that both the DIS (odds ratio, 3.64; 95% confidence interval [CI], 1.32 to 10.05; P=.012) and elevated BDI scores (odds ratio, 7.82; 95% CI, 2.42 to 25.26; P=.0002) were significantly related to 18-month cardiac mortality. After we controlled for the other significant multivariate predictors of mortality in the data set (previous MI, Killip class, premature ventricular contractions [PVCs] of ≥10 per hour), the impact of the BDI score remained significant (adjusted odds ratio, 6.64; 95% CI, 1.76 to 25.09; P=.0026). In addition, the interaction of PVCs and BDI score marginally improved the model (P=.094). The interaction showed that deaths were concentrated among depressed patients with PVCs of ≥10 per hour (odds ratio, 29.1; 95% CI, 6.97 to 122.07; P<.00001).
Conclusions Depression while in the hospital after an MI is a significant predictor of 18-month post-MI cardiac mortality. Depression also significantly improves a risk-stratification model based on traditional post-MI risks, including previous MI, Killip class, and PVCs. Furthermore, the risk associated with depression is greatest among patients with ≥10 PVCs per hour. This result is compatible with the literature suggesting an arrhythmic mechanism as the link between psychological factors and sudden cardiac death and underscores the importance of developing screening and treatment programs for post-MI depression.
We previously reported that patients with major depression while in the hospital after a myocardial infarction (MI) experienced a significantly elevated risk of mortality during the first 6 months after the MI.1 Depressive symptomatology has also been associated with elevated risks in various subgroups of patients with MI, including those with significant ventricular arrhythmias2 and males younger than 65 years.3 Although none of these studies have been able to explain the connection between depression and prognosis using measures of cardiac disease severity, some physiological mechanism must link the psychological symptoms of depression with mortality. This report extends the follow-up period for our original cohort from 6 to 18 months after the MI and further explores the mechanisms linking depression and post-MI prognosis.
Study procedures were approved by the Research Ethics Board of the Montreal Heart Institute on December 13, 1990. Patients were initially identified during the screening phase of a larger multicenter post-MI trial evaluating the prognostic significance of the signal-averaged ECG.4 Because recruitment procedures have been described in detail elsewhere,1 characteristics of the population sample are summarized briefly. Patients admitted to the Montreal Heart Institute between August 1991 and July 1992 for a documented MI were eligible to participate if they met study criteria. To be eligible, patients had to meet at least two of three criteria for an acute MI (typical chest pain lasting ≥20 minutes, presence of new pathological Q waves in two contiguous leads on the ECG, a peak creatinine phosphokinase [CPK] level of >1.5 times normal, or a CPK-MB value of >5% of a simultaneous CPK value exceeding the normal limit). Patients with MIs secondary to coronary artery bypass graft surgery or angioplasty, who had another condition likely to influence survival, who did not have adequate fluency in English or French to complete the interviews, who had cognitive difficulties, or who were not medically stable enough to complete a baseline interview between 5 and 15 days after the MI were ineligible. Of 337 eligible patients, 227 gave informed consent and were interviewed in the hospital. Five patients died before discharge. Thus, the final study sample includes 222 patients. As described previously,1 limited baseline data were available for the 105 eligible patients who did not take part in the study. The eligible nonparticipants were more likely to be women (P=.01) and were significantly older than the participants (P=.005) but did not differ in available measures of cardiac disease severity, including previous MI, smoking behavior, thrombolytic treatment, and left ventricular ejection fraction (n=96). Furthermore, survival status at 18 months was available for 101 of the 105 patients who did not take part. There were 15 deaths among the nonparticipants in comparison to 21 deaths among study subjects (P=.15).
Structured baseline interviews were carried out by a research assistant between 5 and 15 days after admission. These interviews included modified versions of the National Institute of Mental Health Diagnostic Interview Schedule (DIS),5 the 21-item self-administered Beck Depression Inventory (BDI),6 and questions about social and demographic characteristics. The DIS allows trained interviewers to collect data that are treated by computer to provide reliable psychiatric diagnoses based on the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition (DSM-III-R).7 The BDI is a self-report inventory designed to measure severity of depressive mood rather than to determine diagnosis of major depression. It was included to help describe the characteristics of post-MI depression. Scores on the BDI range from 0 to 63. Patients with BDI scores of ≥10 are considered to show at least moderate symptoms of depression.8
Because of time limitations in conducting interviews in hospital, the DIS was modified based on the work of Schleifer et al.9 The DSM-III-R duration criterion, which requires that symptoms of depression must have lasted at least 2 weeks, was not applied. Instead, patients who met symptom criteria from admission to the time of the interview were classified as depressed. Furthermore, because patients were in a healthcare setting, the criteria of seeking medical help for the depression and of experiencing impairment were not used for establishing the diagnosis of major depression in hospital.
The following baseline clinical variables were abstracted from patients’ hospital charts: history of previous MI, left ventricular ejection fraction measured by nuclear ventriculography, Killip class, frequency of premature ventricular contractions (PVCs) based on 18 to 24 hours of Holter monitoring, thrombolytic therapy, and prescription of β-blockade and angiotensin-converting enzyme (ACE) inhibitors at discharge. Holter monitoring data were available for 197 patients. There were no differences on any baseline variables between the patients with Holter and those without Holter monitoring. Left ventricular ejection fraction was available for 220 of the 222 patients in the final sample. Thus, there is little evidence of bias toward having more complete data on more severely ill patients.
All patients or their family members were contacted 18 months after the MI to determine patients’ survival status. A committee of three cardiologists independently classified all deaths as cardiac or noncardiac using information from death certificates, ambulance and hospital records, and interviews with family members or other individuals who had witnessed the death. The committee further classified cardiac deaths as due to arrhythmias, recurrent MI, or severe congestive heart failure or subsequent to cardiac procedures using the criteria from the Cardiac Arrhythmia Suppression Trial (CAST).10
Statistical analyses were carried out using spss for windows (version 6.0).11 All tests were two-tailed. Multiple logistic regression analysis and the χ2 statistic were used to assess the odds ratio for mortality in depressed versus nondepressed patients as assessed by each separate measure of depression (the DIS and the BDI dichotomized at 10). To select other variables related to prognosis, we followed procedures outlined by Hosmer and Lemeshow.12 Baseline measures were dichotomized at points suggested in the literature, and the odds ratio for each pair of groups was assessed using logistic regression and the χ2 statistic. Continuous variables were assessed twice, once as dichotomized variables and once as continuous measures. Baseline variables other than depression that were significantly (P≤.05) related to mortality in the bivariate analyses were entered into a backward stepwise logistic regression analysis to predict 18-month mortality. The resulting model was confirmed with forward analysis. Criteria for entry and removal were based on the likelihood ratio test with enter and remove limits set at P≤.05 and P≥.05.
To evaluate the prognostic importance of each measure of depression in relation to established measures of risk, we assessed the degree to which each measure of depression improved a predictive model based on other significant multivariate clinical predictors of 18-month cardiac mortality in the data set. The established predictors were entered into the model first, followed by the measure of depression being evaluated. The likelihood ratio test was used to compare the models, allowing us to examine the impact of depression after controlling for more traditional measures of post-MI risk.
All analyses were carried out twice: once with the two noncardiac deaths eliminated from the sample, and once with these patients included and classified as survivors. No differences emerged.
Baseline Variables Associated With In-Hospital Depression
The characteristics of patients meeting the modified DIS criteria in hospital have been described in detail previously.1 Patients depressed according to this measure were more likely to say they had no close friends (P=.031) and marginally more likely to be female (P=.057). They did not differ from nondepressed patients in age, education, whether they lived alone, smoking status before the MI, previous MI, thrombolysis, left ventricular ejection fraction, Killip class, PVCs, and prescription of β-blockade or ACE inhibitors.
The relations between baseline variables and BDI scores dichotomized at 10 (the point usually used to distinguish symptoms of mild to moderate depression from normal)8 are shown in Table 1⇓. Elevated BDI scores were significantly more common among women, people living alone, and those with a Killip class ≥2. Although education was only marginally related to elevated BDI scores when education was dichotomized at grade 8, the mean total of years of education among patients with elevated BDI scores was significantly less than the mean for people with low BDI scores. Patients prescribed ACE inhibitors at discharge were marginally more likely to have BDI scores ≥10, but those prescribed β-blockers were significantly less likely to have elevated scores. Finally, patients who did not receive thrombolysis were significantly more likely to have high BDI scores. There were no significant relations between BDI scores and age, having close friends, smoking status, previous MI, left ventricular ejection fraction, or PVCs.
We also examined baseline information about the prescription of antidepressants. However, the numbers were too small for statistical analysis. As reported previously,1 review of consultation records during the index admission revealed that only 3 patients were prescribed antidepressant medication (none received tricyclic antidepressants), and 1 of them died. However, postdischarge medication information was not available for patients who died, so we cannot examine the possible importance of later prescription of antidepressants.
During the first 18 months after the MI, a total of 21 patients died (9.5%). Nineteen of these patients died from cardiac causes (5 recurrent MIs, 11 arrhythmic deaths, 2 deaths from congestive heart failure, 1 death from cardiogenic shock after coronary artery bypass graft surgery), and 2 died from cancer.
Table 2⇓ shows the baseline characteristics of the patients who died of cardiac causes during the follow-up period in comparison to those who did not. Patients who died were more likely to have had a previous MI, to have had a left ventricular ejection fraction of <35%, to have had a Killip class of ≥2, and to have had ≥10 PVCs per hour during Holter monitoring. They were less likely to have been prescribed β-blockers and more likely to have been prescribed ACE inhibitors. Both major depression in hospital (as measured by the modified DIS) and depressive symptomatology (BDI scores) were significantly related to 18-month cardiac mortality.
Although there were nine deaths between 6 months and 18 months after the MI, only two of these deaths occurred among patients meeting DSM-III-R symptom criteria for major depression according to the modified DIS administered in the hospital. One of these patients died of lung cancer. In contrast, five of the nine deaths that occurred between 6 and 18 months took place in patients who had elevated BDI scores in the hospital. Thus, although the impact of major depression on cardiac mortality remained significant during the 18-month follow-up period, its impact occurred primarily in the first 6 months (odds ratio, 6.24; 95% confidence interval [CI], 1.88 to 20.67; P=.041). Among patients who survived to 6 months, major depression assessed by the DIS in the hospital no longer had an impact on survival (odds ratio, 1.04; CI, 0.12 to 8.98; P=.97). In contrast, the impact of BDI scores in the hospital remained over time. The odds ratio over 6 months was 5.62 (CI, 1.41 to 22.47). Among those who survived to 6 months, the odds ratio associated with the in-hospital BDI score was 13.0 (CI, 1.49 to 113.61; P=.0052).
Stepwise multiple logistic regression analysis was used to identify the most parsimonious group of predictors of 18-month cardiac mortality for use as covariates in the assessment of the independent prognostic importance of depression. Backward and forward procedures, based on the significant bivariate predictors (other than depression), yielded the same results. The final model (χ2; 3 df, 32.16; P≤.00001) included previous MI (odds ratio adjusted for other variables in model, 5.20; 95% CI, 1.50 to 17.97), PVCs dichotomized at 10 per hour (adjusted odds ratio, 6.06; 95% CI, 1.90 to 19.39), and Killip class (adjusted odds ratio, 6.16; 95% CI, 2.01 to 18.83).
To evaluate the impact of the in-hospital DIS-based diagnosis of depression and scores on the BDI (dichotomized at 10) on 18-month cardiac mortality after controlling for the standard measures of risk stratification that were significant in the data set, we forced previous MI, PVCs (dichotomized at 10), and Killip class (dichotomized at 2) into the logistic regression model, followed by each approach to assessing depression. The DIS-based diagnosis of depression did not significantly improve the predictive ability of the standard risk factors (likelihood ratio, 2.31; P=.13; adjusted odds ratio for major depression, 2.68; 95% CI, 0.77 to 9.31). However, the dichotomized BDI scores produced a significant improvement over the model based on previous MI, PVCs, and Killip class (likelihood ratio, 9.07; P=.0026; adjusted odds ratio, 6.64; 95% CI, 1.76 to 25.09). Thus, elevated BDI scores had a significant long-term prognostic impact independent of traditional post-MI risk factors. The continuous scores based on the BDI also improved the traditional risk model (likelihood ratio, 5.98; P=.015; adjusted odds ratio for a 1-point increase in BDI scores, 1.08; 95% CI, 1.02 to 1.15).
In an attempt to further explore the relation between the BDI and the other variables in the model, we evaluated two-way interactions between pairs of clinical variables in the multivariate model. The interaction of PVCs and the dichotomized BDI scores marginally improved the logistic model based on previous MI, Killip class, PVCs, and the dichotomized BDI scores (likelihood ratio, 2.80; P=.094).
Although not meeting the usual standards for significance, the potential importance of the interaction between PVCs and depression as a mechanism linking depression with outcome led us to explore the interaction. As the Figure⇓ shows, patients with PVCs who were depressed according to the BDI were at a substantially increased risk compared with other patients. The 18-month odds ratio for depressed patients with PVCs compared with all other patients was 29.17 (95% CI, 6.97 to 122.07; P≤.00001). The causes of death are also of interest (Table 3⇓). Among patients with elevated BDI scores and PVCs, five of the six deaths were arrhythmic, whereas in the other three groups no particular patterns emerged.
Because of evidence that the risk associated with PVCs may vary according to left ventricular function and other measures of disease severity,13 we examined the background characteristics of the four groups of patients shown in the Figure⇑. The high-risk group, ie, depressed patients with PVCs, were marginally more likely to have impaired left ventricular ejection fractions (P=.069). In an attempt to assess whether the risk associated with PVCs and depression was a surrogate for the interaction of PVCs and ventricular function, we examined outcomes by PVCs and depression separately in patients with impaired and normal ejection fractions. Although the cell sizes are too small to evaluate the interaction statistically, we found that the risk associated with the combination of PVCs and depression appears to be dependent on ejection fraction. Six of the 10 depressed patients with PVCs had low ejection fractions, and all of them died. There were no deaths among the 4 depressed patients with PVCs who had normal ejection fractions. In addition, much of the risk associated with low ejection fraction was concentrated in the patients with PVCs and elevated BDI scores. In fact, 6 of the 11 deaths among patients with low ejection fractions occurred in this group.
These results indicate that depressive symptomatology in hospitalized patients after MI has an important impact on prognosis during at least the first 18 months. Furthermore, the measurement of depression significantly improves risk stratification based on traditional post-MI risk factors.
Although patients fulfilling DSM-III-R symptom criteria for major depression were at significantly greater risk during the first 6 months after discharge than other patients, those with elevated levels of depressive symptomatology (as measured by the BDI) were also at increased risk. Although the DIS measured major depression in the hospital and predicted early post-MI deaths, probably related to this episode of depression, the BDI identified a number of patients who did not meet DSM-III-R symptom criteria in the hospital but who died later in the year. It is possible that elevated BDI scores predict later, postdischarge depressions in patients who do not meet standard criteria for major depression in hospital and that many of the later deaths may have occurred subsequent to major depressions that began after discharge. Additional evidence is needed to assess this hypothesis, but these results clearly suggest that depressive symptomatology should be included among the core group of potentially modifiable post-MI risk factors. Furthermore, both the DIS and the BDI may be valuable for in-hospital screening of patients at high risk of mortality because of depression.
These data also indicate that the prognostic impact of post-MI depression is related to PVCs. Depressed patients are not more likely to have arrhythmias, but the risk associated with depression is largely confined to those with PVCs and probably those who also have impaired left ventricular function. Furthermore, the prognostic impact of PVCs may be related to depression. Patients who were not depressed experienced little increase in risk associated with PVCs, even if they had low ejection fractions. This finding fits in well with animal-based models pointing to proarrhythmic factors as the link between psychological factors and sudden cardiac death.14 Previous clinical research, although limited, agrees with the present results. For example, Kennedy et al15 studied 88 patients undergoing programmed electrophysiological studies for severe arrhythmias. During 18 months, 3 of the 4 patients who were depressed at discharge died compared with 12 of the 84 patients who were not depressed (P=.015). More similar to the present data are the results of the Cardiac Arrhythmia Pilot Study (CAPS), which showed that BDI scores were not correlated with the frequency of PVCs among patients with ≥10 PVCs per hour at baseline.16 17 However, the lack of a relation between depression and PVCs does not mean that PVCs are not involved in the relation between depression and post-MI mortality. The CAPS investigators also found that BDI scores predicted 1-year mortality or cardiac arrest even after control for the number of consecutive PVCs at baseline.2 Our findings confirm the independent risk associated with elevated BDI scores and demonstrate that the impact of depression is most marked in patients with ≥10 PVCs per hour.
It should be noted that our population sample was not unusual in terms of levels of PVCs. Of patients who were Holter monitored between 7 and 15 days after the MI, 13.6% had ≥10 PVCs per hour. This is similar to levels reported in other nonselected post-MI populations.13 18 The observed 1-year mortality of 29.6% for patients with this level of PVCs also agrees with the literature.18 19
It is also of interest that the final results of CAST10 showed that although many studies had demonstrated that PVCs were associated with an increase in post-MI mortality, suppression of PVCs did not reduce mortality. In this context, the present data suggest that rather than trying to suppress arrhythmias, what may be necessary to improve survival in patients with PVCs is treatment of depression. However, it is by no means clear that reducing depression will improve survival. A well-designed clinical trial is obviously needed. Because the potential benefits and risks associated with the use of tricyclic antidepressants in post-MI patients remain controversial,20 additional research is also needed to determine the most appropriate treatment for post-MI depression.
Finally, the fact that BDI scores were significantly related to several background demographic, clinical, and treatment variables deserves some comment. Sex was significantly related to BDI scores. This agrees with other post-MI literature8 as well as the literature on depression in the general population.21 Women are more likely to become depressed than men, and the reasons remain unclear.
We found that patients with elevated BDI scores had less education than patients with scores in the normal range. Although the present sample was perhaps too small to show an impact of education on mortality, low education level has been linked to poor outcomes in many diseases, including established coronary artery disease and MI.22 23 24 25 The observed links between depression and low education and between depression and prognosis suggest that one of the pathways linking education and prognosis may involve depression.
Recent work by Case et al25 found that patients who lived alone after an MI were at increased risk of cardiac death at 1 year and cardiac recurrence at 1 year and that this risk was not explained by standard cardiac risk factors. However, depression was not measured. Our results linking elevated BDI scores with living alone and linking depression with post-MI prognosis support the idea that depression may have intervened between living alone and poor prognosis and could have accounted for the finding of Case et al.25 The fact that living alone was not related to prognosis in the present study may, like the lack of impact of education, be a sample-size phenomenon.
The observed link between BDI scores and Killip class indicates that in the present sample, depressed patients were somewhat sicker than nondepressed patients. However, statistical control for Killip class did not change the prognostic importance of BDI scores.
Finally, although it has been suggested that β-blockers may lead to depression in some patients,26 we found that patients who were not prescribed β-blockers had higher BDI scores. β-Blockade was not randomly assigned, and this result suggests that it was somewhat sicker patients—those for whom β-blockade was not seen as appropriate—who had higher BDI scores. Although there is no evidence in the present data to suggest that β-blockers cause depression in post-MI patients, it is possible that the short period of time that patients had been treated with β-blockers when interviewed in the hospital may not have been sufficient for the side effects of depression to become evident.
Because of the present study’s limitations in sample size and the number of cardiac deaths, it is important to consider the degree to which results agree with previous research. The bulk of the animal and human evidence linking cardiac events with psychological factors suggests that an arrhythmic mechanism is important in sudden cardiac death.27 These models are based on the idea that the combination of vulnerable myocardium after MI, acute ischemia, and negative emotional arousal could easily trigger fatal ventricular arrhythmias.14 In fact, recent research has shown that depressed patients with cardiac disease have reduced heart rate variability,28 suggesting that depression may be associated with changes in sympathetic–parasympathetic balance. Therefore, if sympathetic drive is increased or vagal tone is decreased, depression could be strongly proarrhythmogenic in patients at risk for ventricular arrhythmias. These links among depression, proischemic mechanisms, ventricular arrhythmias, and autonomic balance need additional study.
In conclusion, our results strongly support the key prognostic role of depression in post-MI patients and underscore the importance of further exploring links between the brain and the heart.
This study was supported by grants from the Joint Research Grant Program in Mental Health of the Fonds de la recherche en santé du Québec (FRSQ) and the Quebec Council of Social Research (CQRS), the Medical Research Council (MRC) of Canada, and the Montreal Heart Institute Research Fund. The contributions of Drs Duncan Stewart and Jean-Lucien Rouleau (who served on the committee to judge causes of death) and of Nicole Gélinas; Annick Girard; Danielle Beaudoin, RN; Louise Girard; Denise Pagé, RN; and Doris Morissette, RN (who participated in data collection and coding); are gratefully acknowledged. The authors also thank Dr Stanley Nattel for his helpful comments on an earlier version of the manuscript.
- Received July 12, 1994.
- Revision received September 19, 1994.
- Accepted September 28, 1994.
- Copyright © 1995 by American Heart Association
Ladwig KH, Kieser M, König M, Breithardt G, Borggrefe M. Affective disorders and survival after acute myocardial infarction: results from the post-infarction late potential study. Eur Heart J. 1991;12:959-964.
Rouleau JL, Talajic M, Sussex B, Warnica W, Davies R, Potvin L, et al. Changing patterns of patients having an acute myocardial infarction (AMI), their risk factors, risk stratification, and survival. Circulation. 1994;90(suppl I):I-500. Abstract.
Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. An inventory for measuring depression. Arch Gen Psychiatr. 1961;4:561-571.
American Psychiatric Association, Committee on Nomenclature and Statistics. Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition. Washington, DC: American Psychiatric Association; 1987.
Beck AT, Steer RA. Beck Depression Inventory Manual. San Antonio, Tex: The Psychological Corporation, Harcourt-Brace-Jovanovich; 1987:1-25.
Norusis MJ, SPSS Inc. SPSS for Windows, Release 6.0. Chicago, Ill: SPSS, Inc; 1993.
Hosmer D, Lemeshow S. Applied Logistic Regression. Toronto/New York: John Wiley & Sons; 1989.
Bigger JT Jr, Fleiss JL, Kleiger R, Miller JP, Rolnitzky LM. Multicenter Post-Infarction Research Group: the relationships among ventricular arrhythmias, left ventricular dysfunction, and mortality in the 2 years after myocardial infarction. Circulation. 1984;69:250-258.
Verrier RL. Behavioral stress, myocardial ischemia, and arrhythmias. In: Zipes DP, Jalife J, eds. Cardiac Electrophysiology From Cell to Bedside. Toronto: WB Saunders; 1990:343-352.
Kennedy GJ, Hofer MA, Cohen D, Shindledecker MA, Fisher JD. Significance of depression and cognitive impairment in patients undergoing programed stimulation of cardiac arrhythmias. Psychosom Med. 1987;49:410-421.
Bigger JT Jr, Weld FM, Rolnitzky LM. Prevalence, characteristics and significance of ventricular tachycardia (three or more complexes) detected with ambulatory electrocardiographic recording in the late hospital phase of acute myocardial infarction. Am J Cardiol. 1981;48:815-823.
Glassman AH, Roose SP, Bigger JT Jr. The safety of tricyclic antidepressants in cardiac patients: risk-benefit reconsidered. JAMA. 1993;269:2674-2675.