The use of repeated measures analysis of variance for plaque and gingival indices

Clinical trials for anti-gingivitis and anti-plaque agents commonly use the mean of Silness and Löe plaque indices and Löe and Silness gingival indices as response variables. The aim of this report is to determine if data from anti-plaque and anti-gingivitis clinical trials using Silness and Löe plaque indices and Löe and Silness gingival indices satisfy conditions necessary for the use of the univariate or multivariate approach to repeated measures. These conditions are multivariate normality, homogeneity of variance-covariance matrices, and for the univariate approach, a type-H variance-covariance matrix. Data from 5 separate clinical trials representing a wide range in sample size, pretreatment mean gingival and plaque indices and treatment effects were used to test these conditions. Either the univariate or multivariate approach to repeated measures was found to be appropriate for both responses of the 5 clinical trials. Thus, means of Silness & Löe and Löe and Silness gingival indices meet the necessary conditions for use of either the univariate and/or multivariate approach to repeated measures. However, significant time-treatment interactions are a common occurrence in these types of clinical trials and must be evaluated carefully. The analyses in this study were carried out using SAS. Other mainframe statistical software packages and many micro-computer statistical software packages have routines to analyze repeated measures experiments with analysis of variance methods. However, some of the packages may omit the multivariate approach to repeated measures or may not include interactions between within-subject and between-subject effects. These packages should be used with caution.