The upper and lower extreme percentiles of change in clinical variables were analyzed for a period of four years after completion of periodontal therapy in order to assess the patterns of disease recurrence after treatment. This analysis has demonstrated periods of healing and relapse of sites within individuals, among different individuals, and among different clinical assessment variables utilized. As not all of the sites within different individuals heal or relapse at the same rate, the random asynchronous burst model of pathogenesis seems appropriate to incorporate the different clinical assessment parameters and different sites in different individuals simultaneously. Long-term recession and pocketing, however, may demonstrate opposite trends, giving the erroneous interpretation that attachment levels remain constant with time, whereas, in fact, an active "dynamic equilibrium" of tissue remodeling is established. A statistical model based on the dynamic equilibrium concept would enable account to be taken of the multifactorial interactions that incorporate site and subject interrelationships between different clinical assessment parameters. The dynamic equilibrium model may be suitable therefore as the basis on which statistical analysis of the progression of periodontal disease may be considered. However, as the value of this or other models of the periodontal disease process have not been established, the analysis of data to derive meaningful conclusions from complex statistical techniques may be premature.
Key words: periodontitis, recurrence, intractable, relapse, random burst
The pathogenesis of periodontal diseases has been assumed to follow a linear progression with time as a result of studies that have been based on mean values and standard deviations for statistical analysis of epidemiological data. This type of analysis may be inadequate as large numbers of cases may be pooled, obtaining a mean value that may conceal significant individual variations occurring within the overall trend. It has been suggested that statistical analysis should include mean values supplemented by frequency distributions of those sites undergoing major changes. As a result of these analyses, other models of periodontal disease progression have been proposed. These models include the linear progression model, the random burst model, the episodic burst model, and the asynchronous multiple burst model of periodontal disease progression.
Goodson et al. have indicated that only a small minority of sites may be susceptible to destructive periodontal disease, while other studies have indicated that only a small number of individuals may be susceptible to the destructive process. Other workers have observed the occurrence of variable patterns of tissue attachment loss. The observed patterns of tissue attachment loss have included these:
Questions have been raised as to which of these hypotheses best reflect the true nature of the pathogenesis of periodontal diseases. Several studies have highlighted the inaccuracy of assessments based on the use of a periodontal probe. Some workers have observed that a high proportion of the perceived change in clinical parameters was a result of Type 1 statistical errors, while others have described soft tissue measurement errors with a periodontal probe as being as great as 1 mm, even in the hands of an experienced operator. Much has been written about site specificity of the destructive process. It has been shown that angular defects were most frequently observed radiographically on maxillary first premolars, and that the highest frequency of advanced alveolar bone loss and soft tissue destruction was observable in the incisor regions. Lindhe et al. have shown that the periodontal disease process is episodic, with the majority of the degenerating sites occurring in a small percentage of the subjects. The sites most susceptible to destruction were in the molar regions. Poor response to treatment has been shown to occur in the furcations of molar teeth, while the greatest resolution was seen to occur on single-rooted teeth with the deepest pre-treatment pockets. There is therefore some evidence that different sites in the mouth respond differently to periodontal therapy.
The assessment of periodontal disease progression is further complicated by the controversy that exists in respect to whether comparisons of sites within individuals are utilized as statistical analyses, or the individuals themselves are used as the units of measurement, since sites within individuals are subject to covariance, which could introduce bias into the assessment of the results. As no account of this covariance is made in the analysis of data relating to single individuals, intra-patient comparisons may give misleading results. A dilemma therefore exists as to whether individuals, or sites within individuals, should be used as the unit of analysis in statistical evaluation. It has been shown that using individuals rather than sites as the unit of measurement would tend to produce underestimated standard errors and overestimated statistical significance. Halzonetis et al. have described three types of destructive patterns: minor periodontitis, predominantly molar periodontitis, and generalized periodontitis, in an attempt to reconcile the problem of intra-patient and intersite comparisons. In this study, different patterns of attachment loss were observed in relation to teeth, sites, and surfaces in each group. There is also some evidence from other studies that some sites may demonstrate improvement while other sites concurrently relapse, and that these changes may occur over a protracted period after treatment. This demonstrates that a complex pattern of tissue remodeling occurs, in which some sites remain stable, others improve, and still others relapse with time.
The aim of this investigation was to observe the patterns of recurrent periodontal disease over a period of four years in patients who had received periodontal therapy.
Ten individuals ages 33 to 59 who had multiple residual pockets after completion of an initial phase of treatment that included oral hygiene instruction and scaling and root planing were selected for the study. The group consisted of seven females and three males. Criteria for inclusion in the study were the presence of residual pockets requiring surgery, an absence of relevant medical history, and a consistent ability to maintain good plaque control as monitored by the O'Leary Plaque Control Record is over an estimated period of time after completion of the initial phase of therapy. Prior to undertaking the surgery, the amount of recession and the pocket depths were assessed. Thereafter, these variables were assessed every six months for the four-year period of study. All measurements were undertaken by the same experienced investigator using a Hu-Friedy periodontal probe with a tip diameter of 0.62 mm. In a preliminary pilot study, reproducibility of measurements using a periodontal probe had been shown to be in excess of 86%. The attachment levels were derived by summation of the data for recession and pocketing in each case. Measurements were taken at multiple predetermined sites around the teeth without the use of stems, which had been shown to give less accurate results in a pilot study carried out prior to this study.
To minimize the inaccuracies that prevail in data derived from clinical probing measurements, it was decided to evaluate only the extremes of change. Thresholds were placed on the data so as to include in the analysis only those 5% of sites demonstrating the upper and lower extremes of change in each time period for each assessment parameter. The frequency distribution of the resultant data was then classified into groups that displayed the greatest change and the least change for each time period. The data then were subclassified as follows:
Categories: To indicate whether the defects had periodontal pocketing that was initially mild (0-3 mm), moderate (4-5 mm), or severe (>6 mm).
Surface: To indicate whether differences could be observed between different tooth surfaces.
Tooth: To see if anterior teeth, premolars, or molars demonstrated any observable differences.
Sites: To establish if differences in the frequency distribution could be seen in facial, interdental, and furcation areas.
The results of the analysis described above are illustrated in tables 1 to 6. From table 1 it can be seen that only a small number of sites showed minimal recession and that these increased in the first year postoperatively, after which progressively fewer sites showed minimal recession. This indicates that recurrent recession occurred after one year postoperatively (A). The sites that showed the least recession were predominantly in the mild and moderate categories. Initially minimal recession occurred palatally, but later more buccal sites fell into this category. The increase in sites falling within this category implies some creeping reattachment of the gingival tissues in the buccal sites occurred up to two years postoperatively, which regressed after two years. This creeping reattachment was predominantly in the anterior teeth (B). Minimal recession occurred initially in the posterior regions, although these sites were also susceptible to recession with time. The sites that showed the least recession were predominantly interdental (C).
Conversely, evaluation of those sites that showed the greatest (most) degree of recession demonstrated that this occurred incrementally with the defect depths (D) and in the predominantly interdental regions (E). The total number of sites with extreme recession decreased in the first year postoperatively, indicating creeping reattachment, but more sites progressively entered this group, an indication that additional sites become susceptible to recession in the long term. These notably increased between three and four years postoperatively and seemed to be associated with the interdental and palatal surfaces of the deepest defects (P).
When the amount of recession occurring in each individual is examined, as illustrated in table 4, it readily can be seen that the overall totals for active sites are misleading in that there appear to be different patterns of recession in different individuals. In some individuals (numbers 1, 4, 9) recession remained stable with time, while other individuals (5, 7, 10) demonstrated a peak in the number of sites with greatest recession by one year postoperatively, which reduced with time. Individuals 2, 3, 6, and 8 reached a peak in sites with greatest recession by two years postoperatively. In contrast, individual 4 demonstrated increasing numbers of sites entering the "least" recession category. This implies that this individual was not susceptible to recession and tended to resolve recession with time. In sites showing the least recession, individual 5 had 46 sites after six months, which reduced to 19 by year one; and resumed to baseline at year two, with a marked increase at year four. This may indicate a "burst" of recurrent recession at one year postoperatively and possibly a "burst" of regeneration of the soft tissue at four years postoperatively. By contrast, individual number 7 had five sites with significant recession until one year postoperatively when possibly a "burst" of regeneration occurred, leaving no sites demonstrating recession until four years postoperatively. At this point, seven sites were demonstrating recession, indicating a "burst" of destructive activity. A similar trend is seen in individual number 10, but in this individual it was clearly continuing at four years postoperatively. Individuals 2 and 9 show peaks in sites exhibiting the least recession at six months and at three years after surgery, while individual 10 had minimal sites with recession which then increased at three years after surgery.
Thus it can be seen that recession postoperatively is highly variable in different individuals, with peaks in the sites demonstrating postoperative recession and creeping reattachment at different rates and at different time periods in each individual.
Table 2 illustrates the reduction in pocket depths over the period of study. It can be seen that a small number of sites showed greater than 6 mm of closure (A). The best pocket reductions occurred in the deepest defects on the buccal surfaces (B) and the interdental sites (C). It can be seen that with the exception of one year postoperatively, when some sites showed recurrent pocketing, there is a progressive increase in the number of sites with the greatest healing with time, and this is predominantly in the posterior teeth (D), with some sites demonstrating continuing healing up to three or four years after treatment. In the case of the premolars (E), a number of sites showed greater inconsistency during the period of study, which may indicate less stability of the tissues associated with these teeth.
Greater numbers of pockets showed little healing, with a tendency for long-term healing to occur up to one year postoperatively when some relapse occurred. After this time, progressively fewer sites demonstrated minimal reduction, indicating that continued reduction in pocket depths may occur over a long period of time (i.e., four years in this study) once general periodontal health has been established (F). The minimal improvement sites were predominantly confined to the mild defects (A) and to a lesser extent the moderate defects (G), the buccal surfaces ~, and interdental regions )) in all of the teeth, demonstrating the greatest improvements in pocketing. When the data relating to individuals are evaluated (table 5), it can be seen that only a small number of sites showed maximum resolution. In some subjects (numbers 2, 6, 7, 9, and 10) very few sites demonstrated maximum healing, which then remained stable throughout the study period. In other subjects (1 and 8) the number of sites with greatest Stats 2 resolution peaked by the first year postoperative assessment and remained stable throughout the time period under investigation, while in others (3, 5) good resolution occurred initially, with some relapse followed by incremental improvements with time until the end of the study. By contrast, patient 3 had a progressive increase in sites of maximal healing up to three years postoperatively.
Similar variations in the patterns of healing of individuals can be seen in the assessment of those pockets that achieved the least healing. In individuals 2 to 10, it clearly can be seen that a substantial number of sites show minimal healing by six months after treatment, which reduced markedly by one year after treatment, indicating that maximal healing of pockets only occurs at up to one year after surgery. By two years postoperatively, notable recurrence of pocketing, indicated by the increase in the overall number of sites showing minimal healing, was evident. The number of sites with least resolution was again markedly reduced by the three-year assessment period, after which they remained stable in all except individuals 4, 5, and 8.
Thus it can be seen that periods of healing and recurrence of periodontal pockets occur at different intervals in different individuals. Both the rate and the stability of sites showing maximal and minimal resolution of pocketing also was variable in each individual.
Table 3 demonstrates how attachment levels changed after treatment. The greatest gain of attachment was in the deepest defects (A) and interdentally (B), with a tendency for them to be on the buccal surfaces rather than the palatal surfaces (C).
Sites showing minimal gain in attachment were fairly constant for the first three years after treatment, but a noticeable reduction in these sites occurred by the four-year assessment period. This observation complements the observation noted above in that a "burst" of healing apparently took place toward the end of the study at both extremes of the group under investigation. The sites showing the least improvement were predominantly in the mild group (D). There was a tendency for the number of sites in this group to peak at one and three years postoperatively, which indicates that a number of sites lost attachment to nearly baseline levels in "bursts" of destructive activity. The interdental sites showed the greatest tendency for long-term healing, as it was these teeth that showed the greatest reduction by four years postoperatively (E).
When individuals within the group are observed (table 6), similar differences as noted before are seen, albeit to a lesser degree. Similar peaks of attachment gain and loss can be seen in different individuals at different time periods as described for the variables of recession and pocketing. Those sites with the least gain in attachment show similar extensive variations between individuals over the four-year period of study. Individuals 2, 3, and 6 showed increasing numbers of sites with minimal healing until one year after surgery, implying continued attachment loss, after which the numbers continuously reduced throughout the study period. This implies that after an initial period of recurrent attachment loss, long-term continued healing took place. However, in some individuals (4, 10) after a period of initial healing, more sites entered the category of "least gain in attachment," indicating a tendency for recurrent attachment loss to occur two to three years after treatment. Other individuals (1, 7, 8) remained stable for the duration of the study, following number of sites with least change in attachment until three years after treatment, after which a "burst" of healing took place with a marked reduction in the sites of this group. Individual 9 showed two peaks in the number of sites with little gain in attachment at one year and three years after surgery.
Thus, when loss or gain of attachment is assessed, it can be seen that the rate of change of attachment levels is different in each individual. In addition, in some sites and individuals the attachment levels achieved after the initial phase of treatment did not remain stable. There was also some evidence of "burst of healing" and "bursts of destruction" as noted previously. This confirms the previous observations in respect to recession and pocketing.
Closer examination of tables 4-6 shows a trend in recession acting in reverse of the trend in pocketing, giving the appearance of constant levels of attachment. Thus it seems that periods of activity, either in terms of recurrence or resolution of the disease, may result in a "dynamic equilibrium" being observed because of cancellation effects resulting from increased recession with concurrent decreased pocketing in a number of sites or vice versa.
It has been shown that patient variability may occur within a group of individuals under study, and this variability is evident in all of the clinical variables used in this assessment. In addition, not all individuals or pockets within an individual heal at the same rate, or remain healthy after treatment. Sites undergoing "bursts" of increased pocket depth or "bursts" of reduction in pocket depth are evident throughout the study. However, when only the total number of sites for the whole group are evaluated, these trends are lost. Thus, both subject and site interrelationships must be utilized in order to more readily assess the interpretation of statistical analysis. Halazonetis et al. have observed that the heterogeneity of patient groups obscures potentially important clinical associations with different patterns of attachment loss. Although this paper has concurred with this observation, it has been shown additionally that the three assessment variables of recession, pocketing, and attachment loss show different rates and patterns of variation with time. These variations may give distorted results when only one variable is used for evaluation, or erroneous cancellation effects when the data are pooled. Some workers) 2 3 4 have theorized that only a small number of individuals and sites are susceptible to periodontal breakdown, but such observations may be erroneous due to the cancellation effects or an inability adequately to statistically model the minor changes taking place concurrently.
Therefore, in the vast majority of sites that appear clinically stable, it may be that a "dynamic equilibrium" has been established in which the actual alterations in the patients and site variables are lost. It has been shown that different patterns of progression of periodontal disease may arise purely from the assumptions made in mathematically modeling the analysis rather than from any true clinical effects. This study has highlighted the complexity of the varying responses of different sites as measured by different assessment variables at vary-in" time periods during the study. Sites may heal at various rates or, conversely, may be more susceptible to breakdown with time. In this study recurrent periodontal pocketing occurred with time, predominantly in the posterior teeth. This concurs with the study by Lindhe et al., but in addition variations the susceptibility of sites to recurrent recession and loss of attachment were noted. Furthermore, some sites demonstrated long-term improvements in the variables of recession, pocketing, and soft tissue attachment. As recurrent periodontal diseases did not recur in the anterior regions, this study agrees with others in which it has been observed that the best healing occurred in single-rooted teeth. This study, however, also has demonstrated that site variations are prevalent. More buccal and interdental site pockets, particularly on anterior teeth, achieved better results than in other sites. It also has been shown that additional factors to be considered include subject variability. Periods of activity have been demonstrated which may include not only "bursts" of destruction but also "bursts" of healing that may occur concurrently, resulting in further distortions of the analysis.
Imrey has highlighted the problems associated with statistical analysis of the data from periodontal disease observations. He has suggested that a number of transformations might result in better interpretation of the data. In addition, techniques such as cluster sampling, which enable a differential subject weighting to be applied to high-risk sites, may be helpful. Before complex, interactive multivariate analyses can be developed for use with periodontally derived data, however, more needs to be known about the patterns of microbiological infection, the pathological progression of the disease process itself, and the underlying model of the progression of disease, which would cause different interpretations of the data to be made and which would, in turn, give different results on statistical evaluation. Currently available multiple comparison techniques like Scheffe's multiple procedure are too nonspecific and may yield conflicting results with the simpler univariate tests of convariance like the F-test. This highlights the difficulties currently facing the statistician who is attempting to analyze data to derive meaningful conclusions. Other workers also have emphasized the complexities and difficulties associated with evaluation of periodontal disease by currently available statistical methods because of the multiplicity of special problems associated with the nature of the disease process and the measurement thereof. They have concluded that further research is necessary to establish more reliable methods for statistically analyzing the progression of periodontal disease.
It should also be borne in mind that multivariate statistical analyses require normally distributed parametric data. Data relating to the progression of periodontal disease are not necessarily normally distributed, and in any case, this type of analysis is not suitable for a disease model based on "random bursts" of activity and quiescence. With this in mind, no attempt has been made in this investigation to statistically analyze or develop a model for analysis of the long-term response of the tissues to periodontal treatment. The low number of individuals in this study also would preclude meaningful analysis of the data in this way.
Conclusions drawn from this study must be considered with great caution, because of the small number of individuals under investigation, and the lack of any hypothesis testing owing to the complexity of the data, which has precluded meaningful statistical analysis of the observations made. In this study the following has been suggested: