| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J Thorac Cardiovasc Surg 2004;127:892-893
© 2004 The American Association for Thoracic Surgery
Letter to the editor |
Department of Intensive Care and Research, Austin Hospital, Melbourne, Victoria , Australia
Department of Cardiology, Austin Hospital, Melbourne, Victoria, Australia
Department of Cardiac Surgery, Austin Hospital, Melbourne, Victoria, Australia
We thank Professor Angelini and colleagues for raising several important issues in relation to our trial.1 However, we do not agree that the conclusion of our article is misleading, and we believe that the issues raised in their letter can be simply and easily addressed.
The article presents the results 5 years after trial inception. We make that perfectly clear in the text and display the mean duration of follow-up in a graphic format with the Kaplan-Meier graphs. Pocock's concerns about the imprecision of estimates as displayed by such graphs in the presence of few events relates to the risk of conveying a visual impression of a difference where none exists. This is of moot concern in a graphic display of actual lack of difference.2
This interim analysis was not planned; it represented a response of the group of cardiac surgeons, cardiologists, intensivists, and cardiac anesthesiologists involved in the trial to the aggressively promoted view that radial artery conduits are "best," a view also strongly espoused by Professor Angelini and colleagues3 on the basis of nonrandomized data. A decision to undertake an interim 5-year graft patency analysis was made to ensure that, with the best of the data available so far, the design of the trial did not expose our patients to "unsafe surgical practice." No criteria were formally set for "stopping" the trial. The data were presented to the physicians involved in the trial and, following discussion, a unanimous decision was made to continue the study. The enrollment of 556 patients is now complete.
We agree about the importance of interim analyses for the purpose of patient safety. However, the methodology suggested by the correspondents is typically that of large multicenter drug company-sponsored trials, where major financial conflicts of interest exist. In such studies, the need for an impartial Data and Safety Monitoring Board (DSMB) is obvious. The need for a DSMB in single-center studies, where the patients are known to all physicians involved and are regularly seen in outpatient clinics, is empirically unproven. Furthermore, the statistical criteria for action cannot, in our opinion, be sensibly taken as an absolute but must be seen within a Bayesian analysis of the pretest probability of a particular outcome being correct.4 We note that no gold standard exists to empirically validate the sensitivity and specificity of current statistical criteria for trial cessation. The probabilities of a difference between groups1 approximate unity, a far cry from the extreme statistical "stopping" values as suggested by O'Brien and Fleming and the Peto-Haybittle rule to avoid a type 1 error.5,6
Our study might be underpowered. However, the number of events seen in this interim group of patients may not be representative of the whole population. Furthermore, we anticipate that most of the outcome events will be seen in the latter half of the trial, and therefore, the estimation of sample size based on information from the interim analysis is inappropriate. We invite caution and consider that more interim information is needed before a recalculation of sample size is necessary. We agree that our report contains data with limited statistical power. As this was a safety analysis aimed at excluding only a major difference between the 2 groups, we believe our observations are important to the continued conduct of this and other studies, hence the publication.
All 438 patients1 were included in the clinical outcomes analyses, which were based on an "intention-to-treat." Forty-two patients were excluded from the graft patency study: 21 were radial artery trial grafts and 21 were controls (16 right internal thoracic artery and 5 saphenous vein grafts). Two patients of the 42 refused angiography and 23 were excluded because of graft disease. In 2 patients the wrong conduit was used and in the remaining 15 patients, the correct grafts were used, although the surgical technique was not as defined in the protocol. We understand the correspondents' concern about possible selection bias due to the exclusion of patients who did not receive the correct graft or in whom the surgeon did not target the correct artery. However, such decision is logical because it is not reasonable to assess the patency of a radial graft when no such graft exists. The same applies to assessing flow to a vascular bed that was actually not grafted. We consider that, in a study such as this, angiographic analysis on "per-treatment" principles is the correct approach. We further note that this decision was made a priori.
Finally, the conclusions of the article are correct given the data available. Furthermore, all the necessary information is available in detail for readers to understand the limitations that surround such conclusions. There is no attempt to mislead and whether our report will influence surgical attitudes remains uncertain. Furthermore, it is an "open verdict" and we cannot see how it could jeopardize future randomized controlled trials, given our conclusion that "the continued evaluation of the RA is justified." For our part, we are very committed to carrying out good randomized trials in cardiac surgery and will continue to do so in a branch of medicine where, relatively speaking, they are uncommon.
 |
References |
|---|
 Top References |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
