|Home | About | Journals | Submit | Contact Us | Français|
Don't believe everything you read. That's the message from several articles in this week's BMJ.
Neill Adhikari and colleagues systematically reviewed the results of 12 clinical trials of nitric oxide in the treatment of acute respiratory distress syndrome (doi: 10.1136/bmj.39139.716794.55). A selective vasodilator, nitric oxide has been widely used in intensive care units because of its demonstrated beneficial effects on oxygenization in critically ill patients. But this meta-analysis of small trials totaling more than 1200 patients found no significant effect on hospital mortality and an increased risk of renal dysfunction. In an accompanying editorial, Niall Ferguson discusses the theoretical advantages of nitric oxide in catastrophic respiratory failure (doi: 10.1136/bmj.39168.568692.BE). Unfortunately this seems to be a case when physiological improvements do not lead to better clinical outcomes, which Ferguson calls a “recurring theme in critical care.”
A more general but less obvious point is made by Ignacio Ferreira-González et al in a systematic review of cardiovascular trials using composite end points (doi: 10.1136/bmj.39136.682083.AE). “Composite end points” is journal speak for the creation in a clinical trial of an outcome measure from several presumably related end points, in order to increase the trial's efficiency. Thus the investigators might combine death and myocardial infarction rates in a cardiovascular drug trial or several measures of cognitive impairment in an Alzheimer's disease study. Using composite end points allows a smaller sample size and can better capture the overall impact of a therapeutic intervention.
This works well when the combined end points are of similar importance and similar frequency and if they are not subject to important biases. But it can be misleading if, for example, death is a rare outcome and the other end points are subject to bias, such as revascularization rate or admissions to intensive care. This study found that in some of the trials many of the most important components of the composite end points had smaller treatment effects than the less important ones when they were reported separately. And many trials did not report the component outcomes separately.
In an editorial (doi: 10.1136/bmj.39176.461227.80), Nick Freemantle and Mel Calvert agree with this paper's findings and express concern that the misleading use of composite end points may lead to approval by regulators of inappropriate claims for drugs. They advocate clear language specifying which outcomes have been shown to be affected by the drug in question.