ABSTRACT
A great deal of controversy has existed in the periodontal literature as to whether the site or the subject should be the unit of analysis. Using the site as the unit of analysis assumes that observations of sites within the same subject are independent and ignores between subject variation. The purpose of this report is to evaluate the influence that the unit of analysis has on estimating the number of necessary site specific bacterial samples from each subject. The number of bacterial samples per subject was defined as the number of samples that would insure a clinician at a 95% confidence level that, if the bacteria were present in a subject, it would be discovered. From two data sets in which 20 to 30 bacterial samples were taken from each subject and data generated from a simulation, appropriate within-subject sample size was determined. In one data set the presence of Actinobacillus actinomycetemcomitans, Porphyromonas gingivalis, Bacteroides forsythus, and Prevotella intermedia was determined by indirect immunofluorescence. In the other data set the presence of A. actinomycetemcomitans, P. gingivalis, and P. intermedia was determined using DNA probes. Results of this study demonstrate that there is a large between subject variation in site specific bacterial prevalence, as indicated by an elevated intraclass correlation. Simulated data in this report demonstrated that the number of necessary bacterial samples per subject increased with increasing values of intraclass correlation. The number of necessary within subject samples also increased with decreasing prevalence rate. For A. actinomycetemcomitans, which had a low prevalence rate (0.11 to 0.18), the number of necessary samples per subject was very high (31 to 35).(ABSTRACT TRUNCATED AT 250 WORDS)