Depression and cardiovascular disease: Selected findings, controversies, and clinical implications from 2009

Summary and clinical implications

The Nurses’ Health Study analysis by Whang et al suggested that antidepressant use triples the risk of sudden cardiac death in healthy women, and the authors suggested that the association between fatal ventricular arrhythmias and antidepressant use be examined further. ² The analysis of the Women’s Health Initiative by Smoller et al found that use of SSRIs and tricyclic antidepressants doubled the risk of fatal stroke in healthy women.⁴ The analysis of the Women’s Ischemia Syndrome Evaluation by Krantz et al revealed a doubling of the risk of CVD and death in women taking antidepressants who had been referred for coronary angiography.³ At the same time, May et al found no increase in the risk of heart failure conversion with antidepressant use in patients with CAD,⁵ and Kim et al found no increase in the risk of bleeding with use of SSRIs compared with non-SSRI antidepressants in patients with CAD undergoing CABG.⁶ Finally, in a population- and community-based case-control study, SSRI use was associated with an increased risk of sudden death, particularly in patients with CVD.⁷ So what are we to make of these findings?⁸

In all observational studies (including those reviewed above), unmeasured confounders pose a threat to the validity of any causal conclusions. A study recently tested some of the proposed confounders that might have existed in the above studies. Waldman et al examined racial differences in depressive symptoms and antidepressant treatment among a cohort of 864 consecutive patients with CHD undergoing diagnostic coronary angiography (727 white and 137 African American).⁹ While levels of depression were similar between the white and African American patients, the African Americans were less likely than their white counter-parts to receive antidepressant medications. Patients with only some high school, men, and patients with more severe depressive symptoms were significantly more likely to receive a prescription for antidepressants. Clearly, low education, male sex, and elevated depressive symptoms are related to poor prognosis for CHD, and the simple interpretation that antidepressant use is causing poorer outcomes is problematic.⁸

Two additional interpretations of the observational findings should be considered. First, confounding by indication (depressive symptom severity) might exist in these studies.¹⁰ In other words, patients who are prescribed antidepressant medication may be those with the most severe depressive illness, and it could be this severity, rather than the antidepressant use, that is causally implicated in the CVD incidence.¹¹ However, all of the studies reviewed above either directly controlled for depressive symptom severity (at least as obtained at baseline) or used propensity scores or stratified subjects based on depression severity. The results showed an increased risk among those who were taking antidepressants. However, none of the studies examined depressive symptom severity during or at the end of the study or depression diagnosis and severity before antidepressant use; these data are needed for a clearer understanding of whether the results were confounded by indication.

Second, these findings are also consistent with a treatment-resistant depression phenotype.¹² Krantz et al caution that it is not clear from their observational study³ whether medication use itself or depression refractory to treatment is implicated in the increased risk of CVD events and mortality. Depression that is refractory to treatment may be the type of depression that places patients at risk for sudden death, stroke, or CHD recurrence, so it may not be the antidepressant use per se that is associated with this risk. This phenomenon was documented in a secondary analysis¹³ of the largest-to-date randomized controlled trial of patients with MI undergoing treatment for depression (ENRICHD).¹⁴ It showed that those whose depressive symptoms did not respond to treatment had a higher risk of late mortality (ie, death ≥ 6 months after acute MI). This finding was replicated in 2009 in an important follow-up¹⁵ of the SADHART trial¹⁶; among patients with MI and major depression, treatment-resistant depression (ie, depression that failed to improve substantially during treatment with either sertraline or placebo) was strongly and independently associated with long-term mortality (HR = 2.39; 95% CI, 1.39–2.44; P < .001).¹⁵

What is needed next? Testing the alternative hypothesis— ie, that an unmeasured confounder may exist—is difficult, requiring new observational studies and measurement of the putative third common causes or previously unmeasured confounder. Other putative confounders would then be hypothesized and would need to be included in additional observational studies. To properly test the putative confounding by depressive symptom severity, future observational studies should examine initial depressive symptom severity prior to antidepressant use and then collect data on depressive symptom severity and antidepressant use as time-varying covariates to CVD outcomes. To test whether treatment-resistant depression is the phenotype driving the spurious observational association between antidepressant use and increased risk of CVD, the phenotype and its underlying causal mechanisms need to be better understood. Of course, rigorous and adequately powered randomized controlled trials of antidepressant use in patients with CVD would be a more straightforward way to test the observational association between antidepressant use and increased risk of CVD. We turn now to the recently published randomized controlled trials in this field.

NEW EVIDENCE FROM RANDOMIZED TRIALS IN PATIENTS WITH CVD AND DEPRESSION

Concerns have been voiced for some time about the ability to effectively treat depression and whether an effective depression treatment will affect the risk of CVD recurrence and mortality.⁸ Adding to these concerns is our limited knowledge of the causal pathways and behavioral and biologic mechanisms implicated in this risk association.¹ For these reasons, results from new randomized controlled trials, such as the four summarized below, are important.

Rollman et al compared the effectiveness of telephone-delivered collaborative care (treatment group) and usual physician care (control group) for improving mental health quality of life and reducing depressive symptoms in 302 patients with depression after CABG.¹⁷ Patients were observed for 8 months following randomization. Mental health quality of life and depressive symptoms were both significantly improved in the treatment group relative to the control group. Significantly more patients in the treatment group had a 50% or greater reduction in depressive symptoms (50.0% vs 29.6% in control group, P < .001; number needed to treat = 4.9 [95% CI, 3.2–10.4]). Men particularly benefited from the treatment. This trial suggests that collaborative care can be delivered effectively (and potentially cost-efficiently) over the phone.

Freedland et al also evaluated depression treatment in 123 patients with major or minor depression who underwent CABG.¹⁸ Their primary objective was to determine the efficacy of two behavioral treatments (cognitive behavioral therapy [CBT] or supportive stress management) compared with usual care. Significantly more patients in both the CBT group (71%) and the stress management group (57%) had a low score (indicating less severe depressive symptoms) on the clinician-based Hamilton Rating Scale for Depression compared with the usual care group (33%). These results were maintained 6 months after the end of the trial. Secondary measures of depressive symptoms, anxiety, and quality of life were also significantly improved in the depression treatment groups compared with the usual care group. This trial is important for the following reasons:

• The use of a second control group, the stress management group, represents a strict, high-quality design that controls for professional attention, generic or placebo therapy effect, and time or effort on the part of the patient.
• The second control group also provides treatment options for the patient, as both CBT and stress management were beneficial.
• Outcome assessors were blinded to treatment assignment, an important design feature in behavioral trials.

In a rigorously conducted randomized, double-blind, placebo-controlled trial, Carney et al tested whether 2 g/day of omega-3 acid ethyl esters (eicosapentaenoic acid [EPA] and docosahexaenoic acid [DHA]) improved depressive symptoms in 122 patients with major depression and CHD.¹⁹ Patients in both the omega-3 and placebo groups received sertraline (50 mg/day) during the 10-week trial. A deficiency of omega-3 fatty acids has been implicated in both depression and CHD and is a possible causal link between the two diseases. Also, there is some evidence that the efficacy of antidepressants is increased by the addition of omega-3 supplementation. Unfortunately, there were no differences in self-reported or clinician-assessed depressive symptoms or in predefined depression remission at the study’s end. The trial included a 2-week adherence run-in period, ensuring that medication adherence in the trial was excellent (97%), and concluded that omega-3 supplementation, at least at these dosage levels, does not improve depression outcomes.

In the Coronary Psychosocial Evaluation Studies (COPES) randomized controlled trial, Davidson et al compared 6 months of an enhanced care intervention with usual depression care among 157 patients with ACS and persistently elevated depressive symptoms.²⁰ Under the enhanced care intervention, patients received either problem-solving therapy or antidepressant medication, depending on their preference, and then had the option of later augmentation with the other treatment, intensification of the initial treatment, or switching of treatments, if indicated by depressive symptom severity (stepped care). The purpose of the trial was to determine the acceptability and efficacy of depression treatment among patients with ACS, who often neither agree with the diagnosis of depression nor had been seeking treatment for depression. Significantly more patients were satisfied with their depression care in the intervention arm, and depressive symptoms and major adverse cardiac events (nonfatal MI, hospitalization for unstable angina, or all-cause mortality) were significantly reduced in the intervention arm compared with the usual care arm. The absolute numbers were very small; at the end of the trial, 3 patients in the intervention group and 10 patients in the usual care group had major adverse cardiac events (4% and 13%, respectively; log-rank test, χ²₁ = 3.93; P = .047). The results suggested that involving patients in the type of depression care they receive (medication and/or psychotherapy) and stepping treatments aggressively may be methods to improve the treatment of depression in patients with CHD. In addition, persistently depressed patients may be an interesting patient group to select for future trials, as usual care has resulted in large reductions in depressive symptoms in some previous trials but not in this study of patients with persistent depression.