Statistical methods are presented for analyzing longitudinal periodontal loss of attachment data using subject based summary measures. The methods are illustrated by sing data from a 2-year clinical study in which a conservative periodontal therapy was evaluated. The 2-year study period was divided into the first 6-month period (treatment period) and the 2nd 18-months (maintenance period). Individual sites within patients were classified by their initial probing packet depth values: shallow, moderate or deep. Treatment and maintenance effects were assessed by using multivariate statistical methods (Hotelling T- square tests) jointly, and for each class of sites, separately. For this data set, it was shown that this conservative therapy produce significant improvement for deep sites, minor improvement for moderate sites, and significant deterioration for shallow sites.