Copyright © 2004 by the European Society of Cardiology.
Letter to the Editor
Bias in the evaluation of evidence linking depression to cardiovascular mortality: reply
Department of Cardiology, Auckland City Hospital, Private Bag 92024, Auckland 1030, New Zealand
Department of Preventive and Social Medicine, University of Otago, Dunedin, New Zealand
* Corresponding author
E-mail address: rstewart{at}adhb.govt.nz
We wish to respond to questions raised by Frasure-Smith and colleagues regarding our study, and to document differences in our interpretation of the evidence linking depression to increased mortality after myocardial infarction. In Table 5, we summarised previous studies selected using an objective criterion (
30 fatal events).1 For all established cardiovascular risk factors, an association with adverse outcomes has been observed across diverse populations. We therefore did not exclude studies because their subjects were participants in a clinical trial or cardiac rehabilitation programme, or because of usage of various medical treatments. We acknowledge that individual studies have limitations, but we are concerned that exclusion of studies from review on subjective grounds would bias evaluation of the evidence. The emphasis given to small positive studies, multiple reports from a single patient population, and studies reporting data only from selected subgroups, has also been misleading.
In our study we used the General Health Questionnaire (GHQ), which was originally designed to detect general psychiatric morbidity in the community. There is, however, a close correlation between scores on the GHQ and scores on questionnaires designed to detect depression alone. In a systematic review of case-finding instruments for depression, the GHQ's sensitivity and specificity were similar to those of other self-administered "depression" questionnaires including the Beck Depression Inventory.2 In a previous analysis we assessed changes in mood over time among patients randomised to pravastatin versus placebo.3 The "Likert" method was used to score the GHQ for this purpose because it has a more normal distribution of responses. In the current study we wished to identify "cases" of depression at baseline, and therefore used the standard method to score the GHQ and prespecified the widely used "case" threshold of 5. Our results and conclusions were, however, similar irrespective of the method used to score the GHQ.1
With respect to timing, mood assessed during hospitalisation may be strongly influenced by the severity and consequences of the cardiac event. On the other hand, "usual mood" (even when assessed before the myocardial infarction) might be important if "depressive symptoms" predict an increase in mortality months or years later.
Our previous study of patients in a randomised clinical trial provided evidence that psychological symptoms are not influenced by treatment with pravastatin or by cholesterol reduction.3 Furthermore, there was no evidence of a treatment interaction in the association between depression and outcomes.1 In the current study, therefore, we presented our findings for all participants combined rather than by subgroup. We are sceptical about the use of subgroup analysis to evaluate possible complex interactions, particularly when such analysis is not prespecified and when the biological mechanism is unclear.
To assess whether an association between depression and increased cardiovascular mortality is causal, it is important to adjust for confounding. Depressive symptoms are more likely with poor health and fatigue due to physical illness such as heart failure, but the associated increase in cardiovascular mortality may be explained by mechanisms independent of depression. Similarly, lack of social support may increase cardiovascular mortality by other mechanisms such as poor access to medical care.
Large randomised clinical trials provide more reliable evidence than observational studies. In the only such study completed to date (ENRICHD),4 the intervention had a modest impact on depression, but did not reduce mortality or reinfarction (300 versus 299 events). Our interpretation of current evidence concerning the hypothesis that depression increases cardiovascular mortality after myocardial infarction evidently differs from that of Frasure-Smith and colleagues.
References
- Stewart RAH, North FM, West TM et al. Depression and cardiovascular morbidity and mortality: cause or consequence? Eur. Heart J. 2003;24:2027–2037.
[Abstract/Free Full Text] - Mulrow CD, Williams JW, Gerety MB et al. Case-finding instruments for depression in primary care settings. Ann. Intern. Med. 1995;122:913–921.
[Abstract/Free Full Text] - Stewart RA, Sharples KJ, North FM et al. Long-term assessment of psychological well-being in a randomized placebo-controlled trial of cholesterol reduction with pravastatin. Arch. Intern. Med. 2000;160:3144–3152.
[Abstract/Free Full Text] - Berkman LF, Blumenthal J, Burg M et al. Effects of treating depression and low perceived social support on clinical events after myocardial infarction: the Enhancing Recovery in Coronary Heart Disease Patients (ENRICHD) randomized trial. JAMA. 2003;289:3106–3116.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||