Methodological issues in analyzing psychological test scores in pediatric clinical trials

Clinical trials to address drug dosing, safety, and efficacy issues in the pediatric population are becoming more common. In some studies, tests of mental ability are administered at regular intervals in drug trials for treatment of children with HIV, certain types of cancer, sickle cell anemia, and diabetes. The test scores are used to examine differences between treatments in efficacy and safety over time. In addition to the well-known problems of analyzing repeated measures with incomplete data profiles, the analyses of these data are complicated by a number of unique features, including that children can be so ill that their raw scores cannot be mapped to a normed scaled score, and that children may be in the studies long enough that they transition between the age-appropriate instruments. These issues are often ignored in data analyses and can potentially cause incorrect conclusions regarding treatment efficacy and safety. This article describes these issues and their possible consequences. A simple approach to determine their impact on the statistical analysis of a particular clinical trial is suggested. The approach is illustrated with data from a Phase III trial in HIV-infected children.