General principles

When reporting the results of a randomized trial, one should keep two goals in mind. Firstly, the onus is on the authors to show that the trial has been designed, conducted and analysed to appropriate standards - or if it has not, to explain why and discuss the possible impact of any limitations on the interpretation of the results. As other chapters demonstrate, the fact that a trial is referred to as randomized is not sufficient to confirm true randomization with adequate concealment; further, the use of proper randomization does not imply that all other aspects of the trial design will render it free from bias; and even the conduct of a well designed, properly randomized trial does not preclude the possibility of bias being introduced in the analysis. The first goal should therefore be to report the results in a structured way, which includes all the information necessary for a reader to judge whether these standards have been reached. For randomized trials, many aspects of structure have been formalized in the 'CONSORT' statement, which is discussed in detail later.

The second goal is to present all the data necessary to make a decision on the overall merits of one treatment compared with another. It is important to recognize that such a decision may often involve a complex weighing up of several advantages and disadvantages, and that decision makers - for example, patients or clinicians - will weight the individual components differently. A trial report should therefore not attempt to dictate - though of course as the people closest to the trial, the authors' opinions will be of interest - but to inform. Key to this, as discussed in Chapter 9, is the need to provide estimates of treatment effect on all relevant outcome measures and not simply the p-value resulting from a hypothesis test.

0 0

Post a comment