J Thorac Cardiovasc Surg 2004;127:891-892
© 2004 The American Association for Thoracic Surgery
Radial versus right internal thoracic artery for myocardial revascularization
Massimo Caputo, MDa,
Barnaby C. Reeves, DPhilb,
Gianni D. Angelini, FRCSa
a Bristol Heart Institute, University of Bristol, Bristol , United Kingdom
b Health Services Research Unit, London School of Hygiene & Tropical Medicine, London, United Kingdom
To the Editor:
Buxton and colleagues1 recently reported interim results of a randomized controlled trial (RCT) of alternative conduits for the second graft in patients having coronary artery bypass grafting. They concluded that "the 5-year interim results do not support the hypothesis that the radial artery (RA) has superior patency to or is associated with fewer clinical events than free right internal thoracic (RITA) or saphenous vein (SV) grafts." We argue below that this conclusion is misleading.
First, the paper does not present 5-year results. The last patient was recruited 2 months before the paper was received. The median duration of follow-up is not stated but from the Kaplan Meier graphs appears to be about 2.5 years. Plotting survival graphs when few patients remain at risk disguises the imprecision of the estimates at these time points.2
Second, the paper represents an interim analysis but does not state (1) whether the analysis was specified in the protocol, (2) the criteria for acting on the results of the interim analysis, or (3) whether any action has subsequently been taken. It is not clear whether the trial continues to recruit.
Interim analyses of treatment effects aim to prevent participants and other patients receiving a treatment known to be inferior.3-5 They should be specified in the protocol, together with proposed actions (ie, stopping rules).3,4 Interim results should be disclosed only to the Data and Safety Monitoring Board (DSMB), not to the investigators.3,5 If the DSMB recommends specific actions (eg, stopping the trial early), the investigators have responsibility for definitive analyses. Statistical criteria for actions are usually set at a more stringent level than for final analyses (eg, P < .001).3,5 If the DSMB concludes no action is required, disclosing interim results can prejudice successful completion of the trial.3
Third, the authors report a target sample size based on higher risks of events than were observed. For patency, the sample size is also much smaller than the target. Consequently, the analyses of patency have little power to detect the target difference. For example, with about 35 patients in each group (ie, for the comparison of RA versus RITA), the analysis had only about 28% power to detect the stated target difference. The analyses of survival free from cardiac-related events also have less power than suggested because only half the participants had reached about 2.5 years of follow-up.
We believe the target sample size is optimistic, even for the planned 10-year analysis. A relative risk smaller than the 0.33 implied by the predicted event risks is likely to be clinically worthwhile. Table 1 shows the power and sample sizes for different baseline event risks (bracketing those chosen by the authors), assuming a constant relative risk of 0.67 (our estimate of the minimum effect likely to change practice). For the tabulated baseline event rates, the rates for the comparison group would be 8% and 11%, respectively (absolute differences of 4% and 5%).
Fourth, the results may also be seriously biased. The authors excluded 40 randomized patients from the graft patency study because of the "poor quality of the randomly assigned conduit (eg, damaged during removal)" or because "the appropriate coronary artery was not grafted." The investigators should have kept these patients in the study as damage to the conduit or inappropriate grafting are likely to affect the overall outcome for a patient and may be related to the type of graft. The decision to exclude a patient could not have been blinded to the randomly assigned conduit. The distribution of excluded patients by group is not reported.
The conclusion is stated as a refutation of the hypothesis that RA grafts have superior patency than RITA or SV grafts. This conclusion is misleading because the analyses had very little chance of detecting the target differences even if they truly exist. The publication is likely to influence the attitudes of cardiac surgeons and may jeopardize future better-designed and better-executed RCTs of an important question.
 |
References
|
|---|
- Buxton BF, Raman JS, Ruengsakulrach P, Gordon I, Rosalion A, Bellomo R, et al. Radial artery patency and clinical outcomes: five year interim results of a randomized trial. J Thorac Cardiovasc Surg. 2003;125:1363–1371[Abstract/Free Full Text]
- Pocock SJ, Clayton TC, Altman DG. Survival plots of time-to-event outcomes in clinical trials: good practice and pitfalls. Lancet. 2002;359:1686–1689[Medline]
- Pocock SJ. Clinical trials: a practical approach. Chapter 10: Monitoring trial progress. Chichester: Wiley; 1983. p. 142-59
- Meinert CL, Tonascia S. Clinical trials: design, conduct and analysis. Monographs in epidemiology and biostatistics, Vol 8. Oxford: Oxford University Press; 1986. p. 208-31
- Pocock SJ. Statistical and ethical issues in monitoring clinical trials. Stat Med. 1993;12:1459–1469[Medline]
Top
References