Eighty-two patients were treated in a split mouth design with coronal scaling (CS), root planing (RP), modified Widman surgery (MW), and flap with osseous surgery (FO) which were randomly assigned to the various quadrants in the dentition. Following phase I and phase II therapy, the patients received supportive periodontal treatment (SPT) at 3-month intervals for up to 7 years. Clinical attachment level (CAL) was determined initially, post-phase I, post-phase II and prior to each SPT appointment. If a site lost > or = 3 mm of CAL from its baseline, it was classified as a breakdown site. Baselines were the initial exam for sites treated by CS and 10 weeks post-phase II for sites treated by RP, MW, and FO. Data were grouped by probing depth (PD) severity at the initial exam and at post-phase II. The breakdown for CS sites was assessed separately from RP, MW, and FO sites because of different baselines and retreatment protocols. Sites treated by CS had a higher incidence of breakdown than the other therapies through year 1 of SPT. The breakdown incidences/year for RP and MW sites were similar and greater than for FO sites in 1 to 4 mm and 5 to 6 mm PD categories. Breakdown incidence of RP sites was greater than MW sites which was greater than FO sites initially > or = 7 mm. Differences in incidence of breakdown between therapies after recategorizing data by post-phase II PD were the same as above, except no difference was present between RP and MW sites > or = 7 mm. Breakdown incidences were greater in increasing PD severities regardless of when they were categorized. There was no further loss of CAL one year after retreatment in 88% of sites. Patients with higher breakdown incidences tended to be smokers at the initial exam.