We established written guidelines to help the clinician determine the surgical needs for each tooth in the dentition, and showed that two clinicians independently examining a patient agreed on his or her surgical needs, with = 0.85, which is regarded as excellent.⁸ As these studies were double-blinded, this eliminated any bias on the part of the clinical examiner in terms of surgical needs. Also, any method errors associated with clinical judgment were distributed randomly to both groups and, if large, prejudiced against the detection of significant differences. Despite these potential confounders, we found that the use of metronidazole plus scaling and root planing for one week significantly reduced the surgical needs of these patients compared with the needs of patients who received placebo plus scaling and root planing.^5,6

Subsequently, we⁷ initiated a study in 1989 with a goal of determining how much periodontal surgery and extraction of periodontally involved teeth could be eliminated by a treatment protocol that included several rounds of systemic treatment, local antimicrobial treatment or both. In this study,⁷ a combination of débridement of the tooth surfaces—that is, scaling and root planing—and short-term use of systemic metronidazole or doxycycline followed, if necessary, by the local delivery of either metronidazole or chlorhexidine to individual teeth resulted in a 93 percent reduction in the need for periodontal surgery around individual teeth, and a 66 percent reduction in tooth extraction over that recommended in the initial treatment plan. In fact, 73 of the 90 patients who would have been scheduled for periodontal surgery or for extraction of four or more teeth did not require any surgery or extraction. However, this reduction in surgical needs would be of little value to most patients if the antimicrobial treatment merely postponed the surgical needs to the following year or some point thereafter. In this report, we describe the subsequent surgical needs of the participants whom we followed for 5.1 years (median value) after completion of the initial antimicrobial treatment.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
Subjects. We recruited from an inner-city dental clinic in Detroit (Detroit Receiving Hospital) subjects with advanced periodontal disease who normally would require surgical intervention. We screened them for an anaerobic infection in sub-gingival plaque samples taken from the tooth that had the greatest pocket depth or mobility in each quadrant of the dentition. A subject who had three or more plaque samples with spirochetes greater than 20 percent of the microscopic count, was capable of hydrolyzing benzoyl-DL-arginine naphthylamide, or BANA (in other words, was BANA-positive),¹ or both qualified for participation in our study, in which both systemic and locally delivered antimicrobial agents would be used in lieu of surgical procedures to treat the anaerobic periodontal infection. The BANA test measures the presence in dental plaque of an arginine hydrolase that is possessed by Porphyromonas gingivalis, Treponema denticola and Bacteroides forsythus, which are three anaerobic species consistently associated with periodontal infections.^1–3,8 The bacteriologic results, which were consistent with a reduction in BANA-positive species in the plaque, will be reported in the future.

Patients signed consent forms approved by the institutional review boards of both the University of Detroit Mercy and the University of Michigan schools of dentistry. Then the subjects were given a detailed baseline examination by two calibrated external examiners, both of whom were periodontists not involved with the treatment of these patients. For each subject, the examiners had access to probing measurements of relative attachment levels and pocket depths (about six sites per tooth) obtained using a constant-pressure probe (Florida Probe, Florida Probe Corp., Gainesville, Fla.); two bitewing and four periapical radiographs showing bone levels surrounding existing teeth, obtained under standardized conditions using a cephalostat; and furcation and tooth mobility scores. (Furcation was scored as follows: 0, no furcation; 1, slight indentation; 2, pronounced indentation; 3, through-and-through penetration. Mobility was scored as follows: 0, no mobility; 1, slight mobility to touch; 2, mobility of 1 to 2 mm; 3, obvious looseness with mobility > 2 mm.^5,6)

The examiners documented each subject’s surgical needs (the number of teeth needing surgical procedures or extraction) using criteria previously described,^5,6 and those with four or more teeth needing surgery, extraction or both were entered into the experimental treatment protocols.

The examiners’ decisions as to whether or not a tooth needed surgery had an interexaminer correlation coefficient of r = .94 and an intraexaminer correlation coefficient of r = .98. The tooth-by-tooth percentage agreement was 90 percent and the was 0.85.^7,8 Molars that needed surgery had an average 6.9-mm probing depth and average relative attachment levels ranging from 11.4 mm to 13.0 mm (Florida probe readings from occlusal surface to bottom of pocket) (Table 1). Single-rooted teeth that needed surgery had an average 7.2- to 7.4-mm probing depth and average relative attachment levels of 12.2 to 13.1 mm. Most teeth had some degree of mobility, and those recommended for extraction had significantly higher mobility scores—that is, 1.7 for molars and 2.0 for single-rooted teeth (Table 1).

Experimental treatment protocols. Treatment phase. The sequences of treatments and the types of treatments are described in detail elsewhere,⁷ and we will briefly summarize them here. All teeth in all patients, including the 143 teeth initially recommended for extraction, were débrided by a single periodontist/hygienist team during three to five visits over a period of two to six months. No patient received more than five hours of treatment, and most received about three and one-half to four hours. After débridement, the patients were randomly assigned in a double-blind clinical design to receive, for two weeks, either metronidazole (500 milligrams twice per day), doxycycline (100 mg daily) or placebo tablets or capsules. Four to six weeks later, the patients were re-examined independently by the same two clinicians who initially had examined the patients. If they had no teeth needing surgery, the subjects went directly to the maintenance phase of treatment; otherwise, they were re-treated with either systemic or locally delivered antimicrobial agents. No patient received more than two systemic treatments and three local treatments, and most received fewer.

Maintenance phase. We considered as a single group the 90 patients who entered the maintenance phase of treatment and who had been treated by nonsurgical procedures. The study hygienist then arranged to débride these patients’ teeth at three-month intervals. However, many patients were not reliable in returning for these débridement visits, so that in the statistical analysis we recorded the number of débridements actually given for each patient. The patients also were scheduled to have a detailed clinical examination by a single external examiner (S.S.) at periodic, approximately annual intervals for five or more years. This examiner was blinded as to which antimicrobial treatments the patient had received previously. He had access to recent full-mouth probing depths and relative attachment level measurements that had been obtained by the clinical periodontist (J.R.G.) using a constant-pressure probe, and to full-mouth radiographs taken under standardized conditions with a cephalostat.⁹ He determined the number of teeth that needed periodontal surgery or extraction based on several factors⁶:

– the extent of probing depth and relative attachment levels;

– the presence or absence of bleeding or exudate on probing;

– the root topography and nature of bony defects as evidenced by radiographs;

– the nature and extent of any furcation involvement;

– the magnitude of tooth mobility;

– whether access to the root surface would be adequate for thorough root débridement.

If this examiner recommended that teeth undergo surgery or extraction, such treatments were immediately offered to the patient. Any teeth that became symptomatic were promptly treated, usually by extraction.

Prophylactic antimicrobial treatment in maintenance phase. We had expected that there would be some erosion of the initial treatment effect during the maintenance period because of the return of the anaerobic infection in the pockets. For this reason, we incorporated a nested clinical trial of the prophylactic usage of metronidazole into the study design. After the first and the second periodic examinations, the patients were given either metronidazole (14 tablets containing 500 mg metronidazole) or placebo (14 tablets) to be taken unsupervised for one week using a double-blind design, or were given no tablets. Thereafter, the patients continued to be scheduled for débridements by the hygienist every three months, but there was no additional treatment with systemic metronidazole or placebo after any subsequent periodic examinations.

Statistical analysis. At each periodic examination, the clinical examiner recorded the status of each tooth in each patient to achieve a total number of teeth that needed surgery or extraction or that had been extracted. If teeth were extracted because of nonperiodontal causes, they were not included in this count. The results are presented in the tables as median values, owing to the bimodal nature of the data; in other words, patients either needed or did not need some type of surgical procedure. We compared the values at each periodic examination with the values at the entry to the maintenance phase (after completion of the initial antimicrobial treatments). We reported the recidivism rate separately for single-rooted and multirooted teeth, and also for teeth that had or had not been recommended for surgery or extraction at the initial baseline examination. The latter analysis provided information on the relapse rate of the initially severely diseased teeth, as well as the appearance of new disease in the initially less diseased teeth.

We determined the differences between the number of teeth per patient that needed surgery and extraction at entry into the maintenance phase and at each periodic examination for each patient, and we used the average values of the differences as the dependent outcome in a series of linear regression models run for each time interval. We entered in the models any predictor variables—such as age, sex, race (which we classified as white, black or other), smoking habits (never smoked, had quit smoking or currently smoked), diagnosis (chronic or adult periodontitis vs. aggressive or early-onset periodontitis), systemic medication taken in the first round of treatment (placebo, doxycycline or metronidazole), and the number of rounds of systemic (one or two) and local antimicrobial treatments (one, two or three)—that might affect the reduction in surgical needs during the treatment phase of the protocol. We also entered in the models other factors that would be operating only during the maintenance phase, such as the number of débridements given by the hygienist and the type of antimicrobial treatment given at the first two periodic examinations (metronidazole, placebo or no treatment). Based on an analysis of various full models, we entered candidate predictor values (P < .25 in the full model) and values that have been reported to be risk factors (such as age and smoking^10,11) into reduced models until we obtained the most parsimonious model that combined the highest r² with the fewest number of predictor variables.

We analyzed the data using linear mixed models implemented through Proc Mixed in SAS Version 6.12 (SAS Institute Inc., Cary, N.C.) because of the longitudinal nature of the clinical design. These linear models account for within-subject autocorrection over time using subject-specific random effects, which would adjust for missing data at the various examinations in the maintenance phase. Thus, this model would address some of the concerns that arose owing to patient attrition. We also used a generalizing estimating equation, or GEE, model because the measurements were obtained for the same subjects over time. GEEs account for correlation among the observations over time obtained from the same subject by constructing a robust covariance matrix in estimation of the standard errors of the regression coefficients.¹² In the GEE model, we used the same predictor variables that we used in the linear regression models. This model would address the dependent nature of the data related to repeated measurements in the same patient. We performed all procedures using SAS Version 6.12.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
Ninety subjects entered the maintenance phase, and 46 were lost to the study during the following four- to seven-year period. Some "active" patients were unreliable in making scheduled visits for both the scaling and root planing sessions and the periodic examinations, despite repeated efforts to schedule them. The data reported reflect those obtained from subjects who had four periodic clinical examinations, although it took from 48 to 89 months (median 5.1 years) to obtain these measurements. The number of delivered scaling and root planing sessions varied from 11 to 29, and we accounted for this variability in the statistical models.

The patients who remained active tended to have had slightly higher surgical needs at the baseline examination than did the patients who were lost from the study. For example, the 44 patients who remained in the study for 5.1 years averaged 9.6 teeth per patient that needed surgery or extraction at the baseline examination, whereas the 46 patients who had dropped out of the study had 7.7 teeth per patient in need of surgery or extraction at the baseline examination.

The 81 patients who were seen at least once in the maintenance phase had 8.7 teeth per patient that needed periodontal surgery or extraction at the baseline examination, and the initial antimicrobial treatment protocol reduced these surgical needs to 1.1 teeth per patient (Table 2). This treatment had reduced the surgical needs of single-rooted teeth by 93 percent and of multirooted teeth by 83 percent from baseline values. During the next 1.1 to 5.1 years, there was a 3 to 7 percent relapse in surgical needs in the singlerooted teeth and a 1 to 25 percent relapse in the multirooted teeth (Table 2). After 5.1 years, the number of multirooted teeth per patient that needed surgery or extractions or had been extracted was 1.68, compared with 0.68 at entry to the maintenance period and 3.9 teeth at baseline (Table 2). The majority of these teeth—in other words, 1.1 teeth per patient—had been or needed to be extracted.

We subtracted the number of teeth per patient that needed surgery or extraction or had been extracted at each periodic examination in the maintenance period from the values that existed for each patient at entry into the maintenance period. The increase in teeth with surgical needs was 0.06 teeth per patient after 1.1 years, 0.22 teeth per patient after 2.3 years, 0.51 teeth per patient after 3.6 years and 0.86 teeth per patient after 5.1 years (Table 3). We developed linear regression models that examined the contributions of a number of factors—age, smoking status, the use of systemic antimicrobials during the initial treatment, the subject’s initial surgical needs, the number of scaling and root planing visits during the maintenance period, and the use of metronidazole or placebo at the periodic recall visits—to the increase in the number of teeth with new surgical needs at each periodic examination (Table 3). We found no effect of the initial systemic use of antimicrobials on the results in the maintenance phase. Only a few of these parameters were significantly related to new surgical needs in the recall period. For example, in the 63 subjects seen after 2.3 years, the unsupervised use of metronidazole after the 1.1-year examination was significantly associated (P = .05) with a reduced need for surgery or extraction, compared with subjects in the no-treatment group. There was a tendency for this beneficial effect still to be evident in the 51 patients seen after 3.6 years (P = .06). The use of the placebo tablets also reduced the surgical needs relative to those of the no-treatment group, but the effect was not significant. At the 5.1-year examination, there was a significant increase in surgical needs in the subjects who were smokers (P = .04) (Table 3).

The results of the mixed model analysis are shown in Table 4. Subjects who were smokers at baseline were significantly more likely (P = .013) to have new surgical needs compared with subjects who never had smoked. The number of sessions of scaling and root planing given in the maintenance phase varied from 11 to 29 per patient, but there was no effect of the number of these sessions on the number of teeth with new surgical needs. Subjects who had received metronidazole treatment in the maintenance phase were significantly less likely to have teeth needing surgery or extraction (P = .008) than were those who did not receive any tablets (Table 4). We observed a similar tendency in the placebo group (P = .07).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

The unsupervised usage of systemic metronidazole by certain patients during the maintenance period may have resulted in additional reductions in surgical needs.

This sustained benefit of initial metronidazole treatment was observed by Soder and colleagues,¹⁴ whose subjects, after receiving scaling and root planing, were assigned randomly to either metronidazole or placebo for one week and then were given scaling and root planing every six months for five years. The patients with complete healing, defined as the absence of inflamed sites with a pocket depth of 5 mm or greater, occurred only in the metronidazole group.

The success of the initial antimicrobial treatment protocols⁷ indicated that the suppression of the targeted BANA-positive, anaerobic species—such as P. gingivalis, T. denticola and B. forsythus—as well as other anaerobes, had been of a magnitude that allowed the host’s defense mechanisms to establish periodontal health. In previous double-blind studies,^4–6,15 metronidazole significantly reduced the plaque levels of spirochetes and black-pigmented species for several weeks following treatment. In the present study, there was a reduction in the number of BANA-positive plaques in the period immediately after treatment (data not shown). Others have shown that metronidazole plus débridement reduced plaque levels of P. gingivalis, T. denticola and B. forsythus, as monitored with DNA probes, to almost undetectable levels during, and immediately after, metronidazole administration, and that this reduction persisted for up to 90 and 180 days after cessation of treatment.¹⁶

Doxycycline appears to specifically reduce plaque levels of T. denticola for periods of at least one year after administration.¹⁷ It should be noted that the bacteriologic species currently associated with periodontal disease are implicated mainly from cross-sectional studies, which cannot discern whether the increase in these species is the cause or the result of periodontal disease. Thus it is possible that the presence of spirochetes and BANA-positive enzymes at high levels in the plaques may not be causative of the periodontal infection. Even so, they served as markers that enabled us to select antimicrobial agents that were active against anaerobes and that appeared to be clinically successful.

We obtained data suggesting that the unsupervised usage of systemic metronidazole by certain patients during the maintenance period resulted in additional reductions in surgical needs. The statistical models showed that subjects who had received prophylactic metronidazole at the first two periodic examinations, relative to those who had received no treatment, had significantly reduced surgical needs (as shown by the negative estimate) at the 2.3-year examination (P = .05), and a tendency for reduction at the 3.6-year examination (P = .06) (Table 3). This surgery-sparing role of metronidazole was more apparent in the mixed-model analysis (P = .008), in which the longitudinal nature of the data increased the sample size (Table 4). We chose metronidazole over doxycycline because we had considerable data on its efficacy,² and because resistant organisms rarely are seen with metronidazole.² We terminated the prophylactic usage after the third periodic examination, owing to the 43 percent attrition rate we encountered. The unsupervised use of placebo medication tended to reduce surgical needs compared with the no-medication group (Table 3 and Table 4), so that some unmeasured patient response to the use of medications could be contributing to the positive findings noted.

We observed a cumulative attrition rate of about 50 percent during the 5.1 years of follow-up. This rate is consistent with a 47 to 70 percent attrition rate that was observed in patients treated in private practices.¹⁸ Cumulative attrition rates for longitudinal studies of five years’ duration or more that evaluated various débridement and surgical procedures in an institutionalized setting ranged from 6 percent to 42 percent.¹⁹ A 35 percent attrition rate was observed over a five-year period in the Swedish metronidazole study by Soder and colleagues.¹⁴ Our first-year attrition rate of 10 percent is comparable with the 5 to 18 percent rate reported in longitudinal studies involving antimicrobial agents that followed subjects for 12 weeks to one year after completion of treatment.^20–23

It is possible that the subjects who returned did so because they could perceive a treatment benefit, whereas the dropouts did not perceive any such benefits. The active subjects stated that they perceived a benefit, but as the dropouts did not respond to follow-up phone calls or letters, we were unable to address this issue. Another possibility could be that the subjects who remained in the study might have had less periodontal disease initially than the dropouts and therefore were more likely to benefit from treatment. This type of bias was not evident, as the subjects who remained active in the study had more teeth recommended for surgery or extraction at the baseline examination than did the patients who subsequently were lost (Table 3). Because of the missing data, we used the mixed model that adjusts for missing values^12,24 for the statistical analysis of longitudinal data (Table 4).

It would be of interest to compare the results of the present antimicrobial treatment paradigm with the traditional surgical treatment paradigm that enables about 80 to 85 percent of the patients receiving treatment to retain most of their teeth. But this outcome is based on reports obtained from clinicians treating patients in their private practices,^25–27 and from clinical trials¹⁹ in which "hopeless teeth" were extracted before treatment and in which the clinical examiner was not always blinded to the treatments given. The extraction of hopeless teeth confounds the interpretation of the efficacy of these studies, because only teeth that were expected to respond to treatment were treated. These studies cannot be compared with our present study, which used a blinded examiner to evaluate surgical needs and in which hopeless teeth were not extracted before treatment.

The present study indicated that the extraction of some hopeless teeth might be premature, as 66 percent of the teeth recommended for surgery at the baseline examination responded to antimicrobial therapy to the extent that even periodontal surgery was not needed.⁷ In fact, few teeth were extracted during this study, and then mostly in a few subjects. Among the 81 subjects who were seen at least once in the maintenance phase, 17 had one tooth extracted, three had two teeth extracted, one had three teeth extracted and one had six teeth extracted. Four of the 51 subjects seen at the 3.6-year examination had lost from four to six teeth; and four of the 44 subjects seen at the 5.1-year examination had lost from four to eight teeth. No subject became edentulous during this study.

	CONCLUSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
These results would indicate that patients have a choice in treatment options: either the traditional approach of surgery or extraction of hopeless teeth, or an approach based on an antimicrobial strategy.² Both options bring the patient to the maintenance phase, where routine scaling and root planing can maintain the beneficial results obtained by either initial therapy. The potential additional advantage of annual doses of metronidazole given prophylactically would need to be substantiated in additional studies. The antimicrobial approach, by reducing the need for labor-intensive surgical procedures, would seem to offer the patient a less expensive treatment protocol. This advantage alone should be sufficient reason for exploring treatments based on the specific plaque paradigm.²

If dental disease, and especially periodontal disease, can be demonstrated to be a risk factor for cardiovascular disease,^28,29 it will be one that is modifiable, and both patients and cardiologists will be seeking cost-effective, user-friendly means of treating it. Treatments based on the control of specific anaerobic periodontal infections would offer such possibilities.