The basic design for human clinical trials which examine prevention of gingivitis recurrence is well known. However, the anti-gingivitis efficacy of triclosan-containing dentifrices has not been routinely demonstrated using this design, regardless of formulation. In an effort to better understand the subtleties of design which show triclosan efficacy, a study was undertaken to prospectively test how subjects with a range of baseline disease levels (which encompassed the range of gingivitis severity documented in the triclosan literature) benefit from a triclosan dentifrice. This trial was a randomized, double-blind, parallel group, six-month study where subjects brushed twice daily with either triclosan/pyrophosphate, triclosan/copolymer, triclosan placebo or sodium fluoride control (NaF) dentifrice following a prophylaxis. Both gingivitis and bleeding were measured using the Löe-Silness Gingival Index (GI) and plaque was measured using the Turesky Plaque Index. Comparisons between treatments were performed using an analysis of covariance. At month 3, the triclosan/pyrophosphate group had a relative mean plaque benefit of 14.4% versus the triclosan placebo group (with an associated p-value of 0.004), and subjects using triclosan/copolymer had a mean reduction in GI bleeding sites of 16.2% relative to those brushing with the NaF control (with an associated p-value of 0.031). The overall results did not demonstrate a treatment effect for gingivitis or plaque for either triclosan-containing dentifrice at month 6. There was evidence of modest anti-gingivitis efficacy in the triclosan placebo. In an effort to better understand how baseline disease severity may have impacted the trial outcome, additional analyses were performed to investigate whether the magnitude of a triclosan effect was related to baseline gingivitis levels. An analysis of covariance model incorporating a baseline group interaction effect indicated that the magnitude of the treatment differences depended on the baseline scores The effect was most pronounced for GI bleeding sites. Further analysis showed that differences between triclosan dentifrices and the NaF control dentifrice were only present for subjects with more than 33 to 63 bleeding sites, depending on the specific comparison. These findings were illustrated through a series of subset analyses on subjects with > or = 20 (entire study population), > or = 31, > or = 40 and > or = 45 baseline GI bleeding sites. The magnitude of the treatment effect for both GI and GI bleeding sites increased with each succeeding subset. Unlike that required for other antimicrobial agents used in oral care products, these findings suggest that a study design which includes subjects with more severe gingivitis at baseline has the required sensitivity to demonstrate treatment benefits for triclosan dentifrices.