LETTER TO THE EDITOR
Composite confusion
David Massel, MD
Assistant Professor of Medicine, University of Western Ontario, London Health Sciences Centre, Victoria Campus, 375 South Street, Room 205 Colborne Building, London, Ontario, Canada, N6K 4J1
dmassel{at}lhsc.on.ca
In their retrospective post hoc analysis of the CADILLAC study, Dr. Cox et al. (1) state that the one-year composite adverse event rate (death, reinfarction, disabling stroke, or target vessel revascularization [TVR]) was greater after optimal percutaneous transluminal coronary angioplasty (PTCA) than routine stenting (21.9% vs. 13.8%, p = 0.001) and that this was driven largely by increased rates of ischemic TVR (19.1% vs. 9.1%, p = 0.001). As a result, they conclude that early and late outcomes can be further improved with routine stent implantation.
The conclusion seems excessive given results glossed over as simply nonsignificant. Perhaps what the investigators might have stated was that the difference in the composite outcome was driven entirely by the least clinically relevant and softest outcome event, namely TVR. Indeed, as shown in Table 3 of their study, the risks of death, recurrent myocardial infarction (MI), and disabling stroke were all higher (but not statistically significantly so) with stenting at 30 days. Moreover, at one year the relative increase in the risk of death with stenting was 31.3% and the absolute increase was 1.0%; this is about the same as the benefit of tissue plasminogen activator over streptokinase and far larger than any putative benefit of glycoprotein IIb/IIIa inhibitors in acute coronary syndromes. I cannot help but believe that had the estimates for the individual outcome events been reversed (that is, favoring stenting) readers would have seen phrases such as "strong trends" or "lack of statistical power" as an explanation. This worrisome increase in clinically important outcome events, also seen in other studies of stents in MI, should not be dismissed as just play of chance or as not significant. I would have thought it might merit a line or two in the discussion.
Could the investigators please provide readers with the differences between stent and optimal PTCA groups for: 1) death and disabling stroke, and 2) death, disabling stroke, and recurrent MI at 30 days and at 1 year?
Finally, should we not discard this silly notion of composite end points that equate a death and a disabling stroke as equivalent to a recurrent revascularization procedure? Is that how our patients view these events? Interpretive difficulties are sure to arise in problematic reports, such as the study by Cox et al. (1), where elements of the composite go in opposite directions (the treatment reduces TVR, but death, disabling stroke, and recurrent MI may be increased) (2). Careful wordsmithing often gives the illusion that all elements of the composite end point are favorably affected (3).
In the absence of a consensus-weighting scheme for elements of a composite, perhaps we need a hierarchical nomenclature for composites that make the results more transparent, particularly in the published abstract. For example, one could state that the composite was significantly lower among patients randomized to stenting (21.9% vs. 13.8%, p = 0.001, death [inc-ns], disabling stroke [inc-ns], re-MI [nd-ns], TVR [dec-sig]). All elements of the composite are reported, and they are ranked starting with death and followed by those of lesser clinical importance. Those elements deemed to be hard and objective are capitalized, whereas those that are more subjective or clinician-driven are given in lowercase. It is specified whether the point estimate is in keeping with an increase (inc), decrease (dec), or no difference (nd); and whether there is conventional statistical significance (sig) or not (ns).
 |
References
|
|---|
- Cox DA, Stone GW, Grines CL, et al. Outcomes of optimal or "stent-like" balloon angioplasty in acute myocardial infarction: The CADILLAC trial. J Am Coll Cardiol. 2003;42:971–977[Abstract/Free Full Text]
- DeMets DL, Califf RM. Lessons learned from recent cardiovascular clinical trials: Part I. Circulation. 2002;106:746–751[Free Full Text]
- Freemantle N, Calvert M, Wood J, Eastaugh J, Griffin C. Composite outcomes in randomized trials. Greater precision but with greater uncertainty? JAMA. 2003;289:2554–2559[Abstract/Free Full Text]
This article has been cited by other articles:
|
|
|
|
 
E. Lim, A. Brown, A. Helmy, S. Mussa, and D. G. Altman
Composite Outcomes in Cardiovascular Research: A Survey of Randomized Trials
Ann Intern Med,
November 4, 2008;
149(9):
612 - 617.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
Top
References