Covariance analysis (for comparison of caries increments) in clinical trials of caries-prophylactic agents

The validity of the assumptions of linearity, parallelism and homogeneity underlying the use of covariance analysis were tested using data from two 3-year clinical trials of caries-prophylactic agents in which over 900 children aged 11-12 years participated. Three covariables were used; initial DMFS (IDMFS), initial surfaces at risk (ISAR) and IDMFS X ISAR/IDMFS + ISAR. For all three covariables the assumptions of linearity and parallelism could not be rejected. However, wide departures from the assumption of homogeneity were found, and it was concluded that analysis of covariance was not appropriate for comparing the caries increments investigated.