A statistical approach to the ecology of Actinobacillus actinomycetemcomitans in subgingival plaque

Structured complexity may be a key feature of the microbiota residing in periodontal ecosystems. Recently, several attempts have been made to analyse the mutual interdependence of certain organisms in subgingival plaque and/or to identify clinical conditions which might influence presence or absence of these bacteria in subgingival plaque. The aim of the present analysis was to compare the results of different models derived from logistic regression of several clinical factors on the presence of Actinobacillus actinomycetemcomitans in subgingival plaque of adult periodontitis patients all harbouring this organism. Models assuming independence of observations or assuming independent observations with the level of response depending on the individual subject were compared to models considering the correlated structure of the acquired data by employing Generalized Estimating Equation (GEE) methods. Significant associations found when neglecting the correlated observations within a given subject as, e.g., between A. actinomycetemcomitans and amount of supragingival plaque, bleeding on probing, or attachment loss, became generally spurious in the GEE approach. Thus, presence of A. actinomycetemcomitans only depended on periodontal probing depth. It was concluded that the correlated structure of observations in an oral cavity should be carefully considered when associations between clinical factors are calculated.