Management of data used in clinical trials which is unevenly distributed at baseline

In this study, 63 healthy adult males were selected for participation in a clinical trial. After initial screening to determine their suitability for participation in the trial, an appointment was scheduled for a professional prophylaxis to derive baseline data and to instruct volunteers in the use of a standard oral hygiene regimen for the trial. Volunteers were then randomly assigned to test and control groups. Follow-up appointments were scheduled two and four weeks thereafter for collection of data. The presence of dental plaque and calculus was assessed at each appointment. Data was to be analysed by using a Student's 't'-test for paired data, but significant differences in the plaque scores were found between the test and control groups at baseline. As a result, the data could not be analysed in this way to obtain a meaningful result. An initial scatter plot of the data revealed six subjects whose plaque scores varied considerably from the group. The aim of this analysis was to evaluate how two different approaches to statistical analysis of data in which significant differences between test and control groups at baseline might affect the inferences drawn from the data. Reasons for significant differences in baseline data in randomly selected groups of volunteers were considered, as well as methods of avoiding this situation. Suggestions on the management of situations in which databases significantly differ include the elimination of 'outliers' and the use of techniques like the Standard Normal Deviate (SND). The data for this study were analysed using these techniques. Although no statistical differences were identified between the two methods of analysis, numerical differences were evident which altered with the type of analysis used. It was concluded that techniques for transformation of the data, like those used in this study, made no difference to the results of subsequent statistical analysis by comparison with the original data set. The influence of sample selection and analysis of data on the results of clinical trials are discussed. The difficulties in interpretation of the results of statistical analyses were highlighted. The influence of imbalances in the data set on the interpretation of the results of statistical analysis was shown to be minimal in this clinical trial.