Lies, damn lies, and random control trials in stuttering?

Tom Weidig

Random control trials (RCTs) are succesfully applied in many areas. Recently, they have been used to evaluate treatment sucess: see Lidcombe and Pagoclone study. I explain why a standard RCT in stuttering faces several conceptual difficulties, and might lead to an overly positive outcome. Specifically, I look at the statistical analysis of the outcome data of early childhood intervention. The natural recovery rate of dysfluent children significantly complicates the statistical study of the outcome data. I argue that the standard random control trial setup needs to be modified, because children are randomly assigned to the treatment or control group and by chance one group will have a higher natural recovery rate. I also point out other methodological difficulties like the possibility of a false positive due to a faster recovery of children that would have recovered anyway.