Some statistical aspects of a clinical study on dental caries in children

Abstract

A significant reduction in caries increment was obtained from dentifrices containing 0.4 per cent stannous fluoride and 0.76 per cent sodium monofluorophosphate, compared to a placebo. The trial was conducted on children aged 12–14 yr over a test period of 2 yr and whole mouth recording was used (150 surfaces/subject). Subsequent analyses showed that partial recording of subjects may be utilized in such trials. Nine methods of partial recording were compared; the most promising involved the examination of 66 surfaces/subject on teeth 4, 5, 6 and 7 in each quadrant and the use of two radiographs/subject compared with four or five required by other methods. The efficiency of stratification and covariance was compared in this trial where some loss of subjects occurred. When the loss of complete replicates exceeded 5 per cent, the single classification analysis of covariance with initial DMFS as a covariate was more efficient than the randomized block analysis of complete replicates where the replicates were formed by stratifying for sex, school, examiner and initial DMFS. No evidence was obtained to suggest that carious lesions develop more slowly or remain stationary longer at a certain stage in the presence of fluorine. The proportion of lesions classified as “smooth decalcified” or “rough decalcified” which reverted to intact was similar in the three treatment groups. The number of errors and wrong diagnoses varied considerably in different parts of the mouth. The error rate, that is the number of errors, reversals, and wrong diagnoses/DMF surface at the final examination, varied from 5 per cent in teeth 4, 5, 6 and 7, to 10 per cent in teeth 1, 2 and 3. The distribution of the residuals of DMFS increment was not normal. However, with the large sample size used in this experiment, the effect of the operation of the Central Limit Theorem was sufficient to make transformation unnecessary.