The purpose of the present investigation was to use baseline clinical and laboratory parameters to distinguish subjects refractory to conventional periodontal therapy. Baseline clinical, microbial and host parameters were compared in 61 successfully-treated and 27 refractory subjects. Refractory subjects showed mean full-mouth attachment level (AL) loss and/or >3 sites with new AL loss >2.5 mm within 1 year after both scaling and root planing and surgery with systemic tetracycline. Successfully-treated subjects showed mean AL gain and no sites with new AL loss >2.5 mm after either regimen. Gingival redness, bleeding on probing, suppuration, supragingival plaque accumulation, pocket depth and AL were measured at 6 sites per tooth in each subject. The levels of 40 subgingival taxa were determined in subgingival plaque samples from up to 28 sites in each subject using checkerboard DNA-DNA hybridization. Serum antibody (Ab) to 85 subgingival species was determined using checkerboard immunoblotting. Levels of serum IgG2 and Gm23 allotype were measured using radial immunodiffusion; FcgammaRIIa and FcgammaRIIIb receptor haplotypes were determined using PCR and allele specific oligonucleotide probes. Odds ratios of a subject being refractory were determined by comparing measured parameters in the 2 subject groups using univariate and multivariate techniques. 17 of 151 clinical, microbial and immunological variables were significant using chi2 analysis after adjusting for multiple comparisons. For example, the odds ratios of a subject being refractory were 12.2, 5.4 and 6.9 if the subject had Ab >50 microg/ml to >9 species; S. constellatus counts >2.4% of the total DNA probe count or >2.1% of sites with AL >6 mm. The 17 significant predictor variables were used in logistic regression and discriminant analyses. Similar variables were selected using both analyses including the number of serum Ab to subgingival species >50 microg/ml, % S. constellatus in plaque samples and % sites with attachment loss >6 mm. In the logistic regression analysis model, the odds ratios associated with >9 species exhibiting >Ab 50 microg/ml, >2.1% of sites with AL >6 mm and >2.4% S. constellatus in plaque were 8.7, 6.8 and 2.4, respectively, after adjusting for other variables in the model. Discriminant analysis using these variables provided sensitivity, specificity, positive and negative predictive values of 0.66, 0.92, 0.80 and 0.85 respectively. Refractory periodontitis subjects could be distinguished using a subset of clinical, microbiological and immunological parameters.