To study the effects of a step-wise treatment regimen on Actinobacillus actinomycetemcomitans-(Aa) associated periodontitis, 4 clusters among 33 patients harboring the organism were followed during successive periods of systemic minocycline plus mechanical debridement and minocycline plus modified Widman flap treatment. Localized periodontitis was found in 2 clusters, one with 7 localized juvenile periodontitis patients and a 24-year old male with localized destruction and extremely low plaque levels (LJP), and the other consisting of 10 patients with plaque and gingivitis and a wider age range (16 to 54 years, LP). Generalized severe and moderate periodontitis was found in 2 clusters which were further discriminated by severe gingivitis and high levels of supragingival plaque (9 patients, GSP), and mild inflammation and low plaque levels (6 patients, GMP). Mean percentages of Aa, as determined by selective cultivation of microbiota from at least 2 periodontal pockets of 6 mm or more were 63, 16, 33, and 7.8% in the clusters (P < 0.01). Six months after active treatment, Aa was present in 6/9 patients and 50% of sites in GSP, and 3/6 patients and 46% of sites in GMP patients. In contrast, the organism was virtually eliminated by scaling and flap procedures in the localized periodontitis clusters, and did not reappear after 6 months (P < 0.05). Combined antibiotic, mechanical, and surgical therapy resulted in a persistence of 20% of sites with residual probing depth of > or = 4 mm in GMP patients after active therapy. At this point, 3 of the GMP patients and 1 GSP patient left the study. Multiple regression analysis showed a significant influence of log-transformed numbers in Aa in cheek and saliva samples at the end of the study, and cluster on the percent residual number of sites with periodontal probing depth of > or = 7 mm (P < 0.001). The present results suggest that the applied therapy would be appropriate in localized forms of Aa periodontitis, but inappropriate in more severe and generalized forms to predictably eliminate Aa. Controlled long-term studies with larger groups of patients will be needed to establish the difference in treatment response suggested by these studies.