Depression following myocardial infarction affects about 20% of patients with myocardial infarction and is associated with a 2–2.5-fold increased risk for all-cause mortality, cardiovascular mortality, and cardiovascular events (1). Whether these effects can be prevented by antidepressant treatment remains unclear. From the observational studies on the cardiovascular effects of antidepressant medication, inconsistent results emerge, with studies suggesting cardioprotective effects (2, 3), no effects (4, 5), or even cardiotoxic effects of modern antidepressant drugs (6).

The experimental evidence for a protective cardiovascular effect of antidepressant treatment is equally inconclusive. In the Enhancing Recovery in Coronary Heart Disease Patients study (ENRICHD), it was found that cognitive behavior treatment (for post-myocardial infarction depression and social isolation) did not have an effect on cardiac prognosis compared to care as usual (7). Similarly, in the Myocardial Infarction and Depression—Intervention Trial (MIND-IT), no effects of antidepressant pharmacological treatment on cardiovascular prognosis were found (8, 9). However, because in these studies no strong effects of treatment on depression itself were reported, the question remains whether effective antidepressant treatment affects cardiovascular prognosis. In ENRICHD compared to usual care, cognitive behavior therapy resulted in a modest relative decrease in Hamilton Depression Rating Scale (HAM-D) score at 6 months (1 to 2.5 points on the HAM-D) that was no longer present at the 30-month follow-up (7). In MIND-IT, antidepressant medication (mirtazapine) was significantly better than placebo after 8 weeks of treatment (9), but no differences in depression status were found between patients in the intervention arm of the study and the care-as-usual arm at 18 months post-myocardial infarction (8). In two other studies, the Sertraline Antidepressant Heart Attack Randomized Trial (SADHART) (10) and a study by Strik et al. (11), a nonsignificant reduction in depressive symptoms as scored with the HAM-D was found in depressed patients with acute coronary syndromes that were treated with a selective serotonin reuptake inhibitor (i.e., sertraline and fluoxetine, respectively) compared to patients treated with placebo. Of interest, in the SADHART, significant improvements in depressive symptoms were observed in patients with severe depression, in depression with an onset before the acute coronary syndrome, and in patients with a history of depression (10, 12), a finding that needs to be further explored.

An exploratory analysis using ENRICHD data of depressed patients with a myocardial infarction was performed by Carney et al. (13). They reported that patients whose depression is refractory to cognitive behavior therapy and sertraline are at high risk for late mortality after myocardial infarction (occurring after 6 months). These results may be explained by the beneficial effects of antidepressant treatment but also by the possibility that a subtype of depression is associated with both a poor response to antidepressant treatment and a high risk for mortality. In the present study, we will evaluate the cardiovascular prognosis of post-myocardial infarction depressed patients who were enrolled in a randomized, controlled trial comparing placebo-controlled medication treatment for post-myocardial infarction depression (the MIND-IT). We hypothesized that responders after 24 weeks of antidepressive treatment would experience fewer new cardiovascular events than nonresponders. To evaluate if differences were due to the protective effects of antidepressant medication itself or to the presence of treatment-resistant depression with a poor cardiac prognosis, we added a comparison with untreated control subjects. These analyses, like the ones by Carney et al. (13), should be seen as preliminary because they were not based on random allocation and, as a result, one cannot rule out the possibility that the findings were produced by the presence of (unmeasured) confounders. Still, these analyses may help to clarify the importance of short-term response to antidepressant medication in preventing cardiovascular events.

The data were derived from MIND-IT, a multicenter randomized, controlled study on the effects of antidepressant therapy for post-myocardial depression on cardiovascular prognosis (14). For the present analyses, data from patients included in the “nested study,” the placebo-controlled efficacy study of mirtazapine (9), were combined with data from the overall study (8).

Inclusion and exclusion criteria have been described previously (14). In brief, we recruited consecutive patients (September 1999 through November 2002) who were hospitalized for acute myocardial infarction in 10 hospitals in the Netherlands. Patients were enrolled if they met World Health Organization (WHO) Monitoring of Trends and Determinants in Cardiovascular Disease (MONICA) criteria for definite myocardial infarction (15). Exclusion criteria were the occurrence of myocardial infarction while the patient was hospitalized for another reason, being unable to participate in the study procedures, having a disease likely to influence short-term survival, already receiving psychiatric treatment for depression, and participation in another clinical trial.

The institutional review board at each clinical center approved the study protocol, and all patients provided written informed consent before enrollment. In the study information it was emphasized that although all participating patients were screened for depression, antidepressive treatment was offered only to a random sample of the depressed patients and that all patients were free to seek help for mood problems.

Procedure

The patients were screened for depressive symptoms during hospitalization and at 3, 6, 9, and 12 months post-myocardial infarction with the Beck Depression Inventory (16). Those with depressive symptoms (i.e., a Beck Depression Inventory score ≥10) underwent a psychiatric evaluation using the WHO Composite International Diagnostic Interview (CIDI), auto version 2.1 (17, 18). The first CIDI interviews were performed not earlier than 3 months post-myocardial infarction to allow for natural recovery of depressive symptoms. Patients with a diagnosis of “depressive episode” according to the ICD-10 were randomly assigned to intervention or care as usual except for patients with a significant risk of suicide. We used a Zelen design (19); i.e., the patients allocated to the care-as-usual arm were not informed about their research diagnosis of depression in order to make sure that patients in the care-as-usual arm were truly representative of patients in usual care. The patients in the intervention arm were offered antidepressant treatment if the study psychiatrist confirmed a DSM-IV research diagnosis of major or minor depression.

Subjects

In MIND-IT, 4,780 subjects were assessed for eligibility, of which 2,177 (46%) of the patients met the inclusion criteria and agreed to participate. Of these, 375 patients developed a post-myocardial infarction depression, of whom 331 patients were randomly assigned to the intervention and the care-as-usual arm. At the start of the trial, we used a 1:1 ratio, but this was changed into a 2:1 ratio because the number of patients actually treated with antidepressants in the intervention arm was lower than expected. This resulted in 209 patients in the intervention arm and 122 in the care-as-usual arm.

Of the 209 patients in the intervention arm, 37 refused to visit a psychiatrist, nine patients were excluded because of the start of antidepressant treatment by the general practitioner, and in 28 patients, the psychiatrist did not confirm the research diagnosis of a depressive disorder. A total of 41 patients did not agree to receive double-blind, placebo-controlled treatment and were given open-label treatment with an antidepressant (N=8) and a referral to counseling or psychotherapy (N=15), whereas 18 patients refused any treatment. These 115 subjects were excluded from the present data analyses because no follow-up assessment of depression status at 24 weeks post-random assignment was conducted in these groups. The remaining 94 were offered double-blind mirtazapine/placebo treatment (i.e., the “nested study”) and used for the present analyses.

Of the 122 patients in the usual-care arm, we selected the post-myocardial infarction depressed patients who did not receive antidepressive treatment based on patient self-reports at 18 months post-myocardial infarction and reports from the patients’ general practitioners. In Figure 1, the flowchart of the study is shown.

Intervention

Double-blind mirtazapine/placebo was prescribed for a maximum period of 24 weeks, divided into an acute treatment period of 8 weeks and a continuation treatment period of 16 weeks. Mirtazapine is a nontricyclic, presynaptic α₂-antagonist that enhances both noradrenergic and serotonergic neurotransmission and is well tolerated by patients with coronary artery disease (20, 21). The starting dose was 30 mg/day, which could be increased to 45 mg/day and in case of severe side effects was lowered to 15 mg/day. After 8, 16, and 24 weeks, response to treatment was evaluated with the Hamilton Depression Rating Scale (HAM-D) (22). If after 8 weeks the HAM-D score was reduced by at least 50% or the HAM-D score was less than or equal to 9, treatment was continued. Otherwise, a switch to open treatment with citalopram (20 mg/day–40 mg/day) was made (23). For those for whom HAM-D scores were not available at 8 weeks (N=16), the effects were estimated based on the last-observation-carried-forward technique. After 24 weeks of treatment, the subjects were classified as responders (a reduction of at least 50% in HAM-D score or a HAM-D score less than or equal to 9) or nonresponders.

Baseline Variables

We collected demographic and cardiovascular data during hospitalization for the index myocardial infarction. The cumulative burden of medical comorbidity was assessed with a modified version of the Charlson Comorbidity Index (24). Further cardiovascular data included the left ventricular ejection fraction (assessed by either echocardiography or radionuclide ventriculography), the Killip class, and a series of cardiovascular risk factors (e.g., the presence of diabetes mellitus, cerebrovascular disease, hypertension). Depression characteristics were based on CIDI interviews with ICD-10 criteria and self-report data (the Beck Depression Inventory). With the CIDI, we assessed whether the onset of the post-myocardial infarction depression was within 3 months post-myocardial infarction or after and whether the post-myocardial infarction depression was a first-ever episode or a recurrent one.

Cardiac Events

The main outcome was a combined time-related variable consisting of cardiac death or hospital admission with an initial evaluation by a cardiologist (e.g., for nonfatal myocardial infarction, myocardial ischemia, revascularization, heart failure, or ventricular arrhythmia). Potential cardiac events were recorded at 12 months post-myocardial infarction (cardiac outpatient clinic, as part of usual cardiac care) and 18 months post-myocardial infarction (during the “outcome” CIDI interview session or through contact with the patient’s general practitioner). Potential cardiac events were reviewed and classified according to prespecified, established (24) criteria by an independent endpoint committee whose members were unaware of the patients’ treatment assignments. Discrepancies were discussed and decisions were taken by unanimity. All cardiac events occurring after random assignment were collected. However, the event had to occur more than 24 weeks after random assignment and within 18 months post-myocardial infarction in order to prevent bias (i.e., the event resulting in nonresponse to antidepressant treatment).

Depression Status at 18 Months Post-Myocardial Infarction

At approximately 18 months post-myocardial infarction, the CIDI interview was administered to determine whether the patients had a diagnosis of “depressive episode” according to the ICD-10. The CIDI interviews were conducted by trained research assistants, who were kept unaware of the patients’ random assignment status.

Statistical Analyses

Responders, nonresponders, and untreated control subjects were compared on baseline characteristics, including sociodemographic data, cardiac risk factors, cardiac status, and depression characteristics by means of Fisher’s exact test, the Kruskal-Wallis test, and the Mann-Whitney U Test where appropriate. Differences between the groups on cardiac events were analyzed with Kaplan-Meier curves, with outcome data censored at 18 months post-myocardial infarction, the time of last contact, withdrawal from the study, or the time of an event. Univariate and multivariate comparisons between responders, nonresponders, and untreated comparison subjects were conducted with Cox regression analysis, with untreated control subjects used as a reference group. All p values were two-tailed.

Of the 94 patients referred to the study psychiatrists in the nested study, 91 received actual double-blind mirtazapine/placebo treatment (47 mirtazapine and 44 placebo) (96.8%) (Figure 1). Of the 47 patients who received mirtazapine, 15 (32%) were classified as nonresponders at 8 weeks and given an open treatment with citalopram during weeks 8 through 16. Of the 44 placebo patients, 23 (52%) were nonresponders and received open treatment with citalopram during weeks 8 through 16. Therefore, of the 94 patients included in the nested study, 70 (47 mirtazapine patients and 23 placebo patients who were switched to citalopram) received actual antidepressant medication. At 24 weeks after random assignment, of these 70 patients, 27 (38.6%) were classified as responders and 43 as nonresponders (61.4%) (see Figure 1). Of the 122 patients randomly assigned to usual care, 98 received no formal care for their depression and were used as untreated comparison subjects.

Table 1 shows a comparison between responders, nonresponders, and untreated control subjects on baseline sociodemographic data, cardiovascular risk factors, cardiac status, and depression characteristics. The groups did not differ significantly on any of the baseline data. The mean Beck Depression Inventory scores were relatively low for all groups (i.e., between 12.0 and 14.4) compared to, for example, depressed patients in the ENRICHD (i.e., between 17.7 and 18.0) (7). However, mean baseline HAM-D scores in our group (17.0 and 18.4) were comparable to those in the ENRICHD study (17.7 and 17.8). This may be because of differences in study design; in the ENRICHD study, cognitive behavior therapy was started soon after the myocardial infarction, whereas in the MIND-IT, treatment was started after only 3 months post-myocardial infarction, which possibly resulted in fewer somatic depressive symptoms, such as fatigue (which are prominent in the Beck Depression Inventory). Of interest and in line with findings from the SADHART (12), the prevalence of recurrent depression seemed considerably higher in the responders group. However, a pairwise comparison between the responders and the nonresponders was not significant (p<0.10). We used sex, age, recurrence of depression, severity of depressive symptoms (the Beck Depression Inventory), myocardial infarction history, current smoking, and occurrence of early cardiac events (i.e., occurring within 24 weeks after random assignment) as potential confounders in the multivariate analyses.

Forty-eight cardiac events occurred between random assignment and 18 months after myocardial infarction, of which 24 were after the end of the 24-week antidepressant treatment. A total of 25.6% (95% confidence interval [CI]=14.8–40.3) of the nonresponders (11 of 43) experienced a cardiac event after 24 weeks; this percentage was 7.4% (95% CI=1.7–20.8) for the responders (two of 27) and 11.2% (95% CI=6.2–19.2) for the untreated control subjects (11 of 98). These percentages represented a significant overall difference among the groups (log rank=8.58, p=0.01) and specifically between the responders and the nonresponders (log rank=5.20, p=0.02). In Figure 2, the event-free survival for responders, nonresponders, and untreated control subjects is plotted.

A comparison of the event rates between untreated control subjects, responders, and nonresponders in a Cox regression analysis (Table 2) resulted in the following findings. First, an increased risk of cardiac events for nonresponders compared to untreated control subjects was found (hazard ratio=2.66, 95% CI=1.15–6.16, p=0.02), which remained after we controlled for age, sex, previous myocardial infarction, baseline Beck Depression Inventory score, history of depression, smoking, history of coronary artery bypass graft and percutaneous transluminal coronary angioplasty, and presence of peripheral vascular disease and after we controlled additionally for early cardiac events (hazard ratio=2.92, 95% CI=1.08–7.87, p=0.03). Second, no significant difference between responders and untreated control subjects was found (hazard ratio=0.52, 95% CI=0.12–2.37, p=0.41), which was unaltered after we controlled for confounders. Third, when directly comparing nonresponders to responders, we found an increased risk for nonresponders (hazard ratio=4.89, 95% CI=1.08–22.10, p=0.04), which remained similar after we controlled for confounders but was no longer statistically significant (hazard ratio=4.47, 95% CI=0.51–39.77, p=0.18).

Of the 168 (70 treated and 98 untreated) patients, a CIDI interview at 18 months was performed in 124 (73.8%). Of the responders at 24 weeks of treatment, 9.5% had an ICD-10 depressive episode at 18 months post-myocardial infarction. These percentages were 58.8% for the nonresponders and 33.3% for the untreated control subjects and were significantly different among the groups (Fisher’s exact test=14.2, p=0.001).

The primary finding of this study was that incomplete response to antidepressant treatment for post-myocardial infarction depression was associated with more prospective cardiac events even when compared to untreated control subjects. This suggests that by providing a standard antidepressant treatment, a subtype of treatment-resistant post-myocardial infarction depression with an impaired cardiovascular prognosis may be identified (13). This finding corroborates with several lines of research on the association between depression and cardiac mortality.

First, the impact of depression persistence on prognosis is stressed by most of the mechanisms supposed to be responsible for the association (e.g., reference 25). Whether because of increased sympathetic activation of the autonomous nervous system, elevated platelet reactivity, or increased immune system reactions, all mechanisms require a persistent depressive state in order to produce cardiovascular events. In line with this, we have recently shown elsewhere that among post-myocardial infarction depressed patients, a subgroup can be identified with persistent post-myocardial infarction depressive symptoms that, in addition, had the highest risk for new cardiac events in a period of 2.5 years, which was not explained by baseline cardiac impairments or the presence of additional risk factors (26). Second, despite large efforts in this field, there is no consistent evidence of pleiomorphic effects of antidepressant treatment, i.e., effects on cardiac prognosis irrespective of their effects on depression itself; in the ENRICHD trial, no evidence of cardiovascular effects attributed to antidepressant treatment was found in the absence of relevant antidepressant efficacy (7). Because the 95% CI of the cardiovascular event rates of the untreated control subjects (6.2%–19.2%) was clearly in between the rates found for the responders (7.4%) and the nonresponders (25.6%), no support was found either for a pleiomorphic effect or for a cardiotoxic effect of antidepressant treatment.

Post-myocardial infarction depressed patients whose depression is refractory to treatment thus seem to be at increased risk for new cardiac events. Of interest, of the early responders, only 9.5% were depressed at 18 months post-myocardial infarction, compared to 58.8% of the nonresponders. This finding suggests that it is persistent depression that is the key variable and that resistance to acute treatment may be a marker for that. Our findings thus stress the importance of early response to post-myocardial infarction antidepressant treatment and the need to develop more effective treatments for depression in the context of a myocardial infarction, such as stepped care in case of early nonresponse. We did not find support for the presence of a subgroup of patients with a treatment-resistant “vascular depression” (27, 28) because responders and nonresponders did not differ on any of the cardiac risk factors at baseline.

Among the strengths of our study was the random allocation to usual care versus active treatment. In contrast to earlier studies (2), we prevented bias due to baseline differences among these groups. Second, we randomly assigned patients only if their post-myocardial infarction depression was still present in the period 3–12 months post-myocardial infarction. This inclusion criterion has reduced the number of adjustment disorders with a depressed mood occurring relatively early after the event with a high rate of spontaneous recovery. Because it may be argued that in adjustment disorders recovery is less a function of treatment response, the results from our study more closely resemble a true treatment effect. Third, by counting cardiac events only if they occurred more than 24 weeks after random assignment, we reduced the chance that the postulated relation between events and treatment response is bidirectional or even the reverse. Of interest, differences among the responders and nonresponders, in fact, emerged after only 3–4 months of treatment (Figure 2). Fourth, by comparing the cardiac prognosis of responders and nonresponders to an untreated control group, we were able to distinguish whether the “pleiomorphic hypothesis” or the “persistent depression hypothesis” could explain our findings. A comparison of responders versus nonresponders alone could not have demonstrated or refuted the pleiomorphic hypothesis. This hypothesis can only be confirmed if exposure to antidepressive medication per se is associated with fewer cardiac events. On the other hand, comparing subjects treated with antidepressive medication to untreated patients alone without consideration of response to treatment would not have allowed us to study the possibility that the persistence of depression is of interest.

The present findings should be considered with caution because of the following study limitations as well. First and most important, the numbers of patients in the different groups were small, and because the data analysis was a secondary analysis of a complex trial, the patients had reached the relevant groups (responders and nonresponders) by different routes and no prior power calculation had been possible. Second, the number of cardiac events was small partly because we counted events only when they occurred after termination of the antidepressant treatment. However, because similar results have been reported before (13), we do not expect this to be an accidental finding. Third, because we do not know why some patients responded to antidepressant treatment and others did not, we cannot conclude that depression is a causal risk factor for new cardiovascular events. Fourth, the fact that we used a stepped protocol for the provision of antidepressive medication—mirtazapine as the first treatment of choice and citalopram in case of insufficient treatment response after 8 weeks—hindered us from further breaking down nonresponse to a specific antidepressant. Fifth, for the untreated control patients we did not have data on early (spontaneous) recovery, and as a consequence, these patients could not be further divided as responders or nonresponders as we did for the treated patients.

By monitoring the outcomes of post-myocardial infarction depression treatment, we may identify patients with refractory depression and an impaired cardiovascular prognosis. The challenge is thus to develop more effective treatments for post-myocardial infarction depression to further reduce insufficient treatment response. Second, we need to identify risk factors for nonresponse to antidepressant treatment. Recent work has shown that important predictors of response to sertraline are depression related and include onset, recurrence, and severity of post-myocardial depression (12). In our study, we found that depression recurrence but not severity was related to nonresponse. Our present findings leave the possibility open that post-myocardial depression may be causally involved in cardiovascular prognosis. However, the actual test for causality has to wait until the long-term course of post-myocardial infarction depression can be altered by more effective treatment.

The following investigators and institutions in the Netherlands participated in MIND-IT:

Steering committee: J. Ormel, Ph.D. (principal investigator), A.H. Schene, M.D., Ph.D., A. Honig, M.D., Ph.D., H.J.G.M. Crijns, M.D., Ph.D.

Study coordination: P. de Jonge, Ph.D., J.P. van Melle, M.D.

Data management: the Trial Coordination Center, Groningen, the Netherlands.

Clinical centers: Flevo Hospital, Almere: A.S.J.M. Sadee, M.D., L.M. Konijnenberg, M.D.; Academic Medical Center, Amsterdam: G. Casteelen, M.D., A.M.G. Kuyper, M.D., R.J.G. Peters, M.D., Ph.D.; Slotervaart Hospital, Amsterdam: M. Bax, M.D. Nij Smellinghe Hospital, Drachten: M. van der Linde, M.D., Ph.D., H. Teunenbroek, M.D.; Medical Spectrum Twente, Enschede: D.G. Buiten, M.D., G.P. Molhoek, M.D., Ph.D.; University Hospital Groningen: J.A. den Boer, M.D., Ph.D., J.F. May, M.D., Ph.D.; Tjongerschans Hospital, Heerenveen: H.P. den Daas, M.D., D.G. Jochemsen, M.D.; Atrium Medical Center, Heerlen: L.H.B. Baur, M.D., Ph.D., C.J.M. van den Berg, M.D., Ph.D.; Medical Center Leeuwarden: D.M. Tulner, M.D., C.J. de Vries, M.D.; University Hospital Maastricht: A. Honig, M.D., Ph.D., P.M.J.C. Kuijpers, M.D., A. Schins, M.D., Ph.D.

Endpoint committee: P.M.J.C. Kuijpers, M.D., Ph.D., J.F. May, M.D., Ph.D., R.J.G. Peters, M.D., Ph.D.