Raised cerebral artery blood flow velocity.

Need to avoid bias in controlled trials EDITOR,-It is accepted practice that in all quantitative research statistical tests should be used to estimate the likelihood that findings arose by chance and that authors should avoid overinterpreting data where these tests suggest that the likelihood is high. It is also accepted that to avoid bias in the analysis of randomised controlled trials clinically important endpoints must be specified at the design stage. A recent report of a clinical trial of two different surfactant preparations illustrates the dangers of neglecting these guidelines.' Although infants in the group treated with one of the preparations appeared to fair better on the predetermined clinical endpoints, none of the differences approached conventional levels of significance. The paper also reports data on a number of intermediate measures of oxygenation and ventilatory requirements, outcomes which the authors specifically state were not established at the design stage, where on some occasions, although not on others, differences do reach conventional levels of significance. In the discussion the authors begin by noting the need for caution in the interpretation of the results from this small study but, unfortunately, do not heed their own warning. They interpret the data on intermediate endpoints as demonstrating improved lung function in babies treated with Curosurf and devote much of the discussion to trying to rule out explanations other than that Curosurf is the more effective agent. They conclude the paper by stating their belief that Curosurf is more effective at improving oxygenation and reducing ventilatory requirements than Survanta in the first 24 hours, remind readers of the 'trend' towards lower incidence of adverse outcomes in the Curosurf group and suggest the need for these results 'to be confirmed in a larger randomised controlled trial'. The clear implication of their comments is that they believe that they have demonstrated that Curosurf is more effective than Survanta. The correct interpretation of the data presented is that the results would be compatible with Curosurf being either considerably better or considerably worse than Survanta at preventing the development of adverse clinical outcomes. The authors specifically mention that patients treated with Curosurf had a lower incidence of pneumothorax, intracerebral haemorrhage grades 3-4, and mortality: in fact, the adjusted 95% confidence limits of the odds ratios for each of these outcomes associated with being treated with Curosurf rather than Survanta are 0-08-2-88, 0-02-1-78, and 0-02-2-54, respectively, which suggests a high probability …