Author + information
- Deirdre Lane, PhDa
- Douglas Carroll, PhDb
- Gregory Y.H. Lip, MD, FACCc ()
We welcome rejoinders to our editorial (1)by Carney et al. and Pickering et al. The precise nature of the association between depression and cardiac mortality in myocardial infarction (MI) patients and, in particular, whether it is a causal association, is an important public health matter; thus, we are grateful that our arguments have sparked debate. However, there appear to be a number of misconceptions in these responses as to our purpose in writing the editorial and, indeed, in what we were trying to say.
Pickering and colleagues accuse us of being “inappropriately pessimistic.” This they ascribe to our failure to consider prospective observational studies of depression in participants initially free of cardiac disease, focusing instead solely on studies of depression (and anxiety) in MI patients. The results from the latter, they readily concede, afford a “weak” case for a causal association between depression and cardiac mortality. Studies in initially disease-free populations were simply beyond the scope of our editorial and the brief given us. We would agree with Pickering et al. that the data from such studies are somewhat more persuasive (2). We would, however, make two points here. First, there are sufficient instances where risk factors for mortality vary between disease-free and diseased populations to suggest that caution is warranted in generalizing from one to the other. Second, confounding cannot be summarily dismissed as a possible explanation of the association between depression and cardiac mortality in studies of initially disease-free participants; confounding always has to be considered as a potential explanation in observational epidemiological research (3). We also appreciate that it can be difficult on occasion to distinguish between confounders and mediators. This is one of the reasons why experimental evidence is so critical (4). What we are arguing is that the data to date do not preclude the possibility that depression after MI may not be a cause, however mediated, of cardiac mortality, and for the reasons articulated above, it is not at all clear to us how prospective observational studies in initially disease-free populations will help resolve this issue.
Furthermore, we were not asserting, as Carney et al. seem to suggest, that depression is not an identified risk factor for cardiac mortality after MI. What we are questioning, as we had hoped that we had made plain, is whether depression is an independentrisk factor (i.e., a fundamental cause of cardiac mortality) so that successful intervention for depression would improve survival after MI. Based on the available evidence at this juncture, particularly the results of the ENRICHD study, the one substantial published randomized controlled trial (RCT) (5), it seems to us that the parsimonious conclusion is that the independence of depression still remains to be established. Moreover, we would submit these null results from ENRICHD should cause us to pause and consider alternative explanations for the pattern of results that have emerged from observational epidemiological studies.
Carney et al. argue that the equivalence of outcome in the ENRICHD trial between the intervention and usual care groups most likely reflected the modest, albeit statistically significant, between-group differences in depression post-intervention, and they chide us as to whether we would dismiss cholesterol as a risk factor on the basis of an ineffectual cholesterol-lowering drug. Clearly we would not: but this is because there is now strong experimental evidence from elsewhere implicating cholesterol (e.g., 6). However, there was a time earlier in the history of cholesterol research when the risk status of cholesterol was controversial and it was perfectly appropriate to be skeptical (7). Likewise, we would be among the first to shift our position on the nature of depression as a risk for mortality after MI were positive experimental data to become available. We have addressed the interpretation by Carney et al. of the results of the ENRICHD ancillary study elsewhere (8). Here we would simply make the point that correlational and experimental data are not evidentially equivalent.
Because both Carney et al. and Pickering et al. use cholesterol research as a metaphor, we should point out that other, possibly more apposite, analogies could be drawn. For example, consider the cautionary tale of hormone replacement therapy (HRT). That HRT was apparently protective against cardiac disease was demonstrated in numerous prospective observational studies. Indeed, the authors of a meta-analysis of these studies argued that the summary estimate protective effect, 0.5, was such that is was unlikely to be explained by confounding factors (9). However, when evidence from robust clinical trials eventually became available, it indicated no reduction in cardiac risk associated with HRT; if anything, a possible increased risk was suggested (10,11).
This example serves to emphasize the occasional fragility of conclusions of causality drawn from observational epidemiological evidence and the importance of experimental data. We trust that our editorial was never regarded as an argument against further RCTs; it is only through these that we shall resolve the true nature of the risk conferred by depression. However, the diagnosis and treatment of depression in cardiac disease patients should not depend on the outcome of such trials. Its prevalence, persistence, and impact on quality of life should provide sufficient imperative.
- American College of Cardiology Foundation
- Lane D.,
- Carroll D.,
- Lip G.Y.H.
- Wulsin L.R.,
- Singal B.M.
- Davey Smith G.,
- Ebrahim S.
- Macleod J.,
- Davey Smith G.
- ↵Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet 1994;344:1383–9
- Muldoon M.F.,
- Manuck S.B.,
- Matthews K.A.
- ↵Lane D, Lip GYH, Carroll D. Is depression following acute myocardial infarction (AMI) an independent risk for mortality? Am J Cardiol 2004;93:1333–4