Analysis of caries clinical trial data in the presence of study group imbalance

Imbalance of any important factor in caries clinical trials can bias group means, and make decisions based on such means of questionable value. Six hypothetical trials are presented which have varying degrees of imbalance in baseline DMFS. The crude group means for these trials vary radically, from one extreme in which the treated group has significantly less decay than the control group, to the opposite extreme in which the control group has significantly less decay than the treated group. Two methods for adjustment to correct for the imbalances, the analysis of covariance and blocking analysis, are discussed and applied to each trial. For these data sets the blocking analysis appears superior to the analysis of covariance in terms of maintaining the essential character of the data. An appendix explaining in detail how the blocking analysis is performed can be obtained on request to the author.