RESUMO
AIM: To investigate the factors predicting non-surgical periodontal treatment responses using multilevel multiple regression. MATERIAL AND METHODS: Forty men (mean 45.6 years) were recruited; 20 were smokers. A 12-month reduction in probing pocket depth (PPD) and gain in probing attachment level (PAL) of 5814 sites were analysed, with 594 being initially diseased sites (initial PPD> or =5 mm). RESULTS: Variance Component models showed that site-level variations contributed about 70-90% of the total variance. About a 10% reduction of the total variations of PPD reduction in initially diseased sites was achieved with the inclusion of the 10 predictors in the multilevel multiple regression. Multilevel multiple regression showed that three predictors, subject level: non-smokers; tooth-level: anterior teeth; and site level: sites without plaque at baseline, were significantly associated with a greater reduction in PPD in initially diseased sites over the 12-month study period (p<0.05). No consistent predictor was found for PAL gain. CONCLUSION: Multilevel analysis was applied on periodontal treatment response data. Smokers showed less favourable PPD reduction at deep sites after non-surgical periodontal therapy.