Considerations in the statistical analysis of clinical trials in periodontitis

On the analysis of life tables for dependent observations

Applying methods assuming independence when observations are positively correlated means that confidence intervals become too short and significance levels of statistical tests are less extreme. This paper discusses elements of life-table analysis based on the standard product-limit estimator and a modified Greenwood formula for its variance to be used for dependent observations. An application from oral surgery is given. Erroneously assuming independence in the analysis of a life table could have serious consequences. It is demonstrated in a simulation study that the confidence levels can be much too low. The proposed modification of the Greenwood formula for the variance of the estimated survival function most often results in confidence levels not too much below the required level. Using the upper bound for the variance will give conservative confidence intervals but also larger standard errors. Averaging individual group survival curves should only be considered for situations with large groups.

Subgingival minocycline hydrochloride ointment in moderate to severe chronic adult periodontitis: a randomized, double-blind, vehicle-controlled, multicenter study

The safety and efficacy of subgingivally-applied 2% minocycline ointment was evaluated in a randomized, double-blind study of 103 adults with moderate to severe periodontitis. Two groups were compared; one received the test minocycline ointment and the other a vehicle control. Both groups had scaling and root planing at baseline, after which the test or control ointments were applied with an applicator into the periodontal pockets at baseline, and at 2, 4, and 6 weeks. Assessment of clinical response was made by measuring probing depth and probing attachment level and gingival bleeding. These measurements were made at baseline prior to scaling and root planing, and at weeks 4 and 12. Microbiological assessment of the subgingival flora was carried out with DNA probes at baseline, and at weeks 2, 4, 6, and 12 to identify and quantify Porphyromonas gingivalis, Prevotella intermedia, and Actinobacillus actinomycetemcomitans. Subgingival minocycline ointment resulted in statistically significantly greater reduction of P. gingivalis at weeks 2, 4, 6, and 12; P. intermedia at weeks 2, 4, 6, and 12; and A. actinomycetemcomitans at weeks 6 and 12. Probing depth reductions were seen for both groups at weeks 4 and 12; however, this reduction was statistically significantly greater in subjects treated with minocycline ointment. Reduction in gingival index and probing attachment gain were seen in both groups, however, the differences between the groups were not statistically significant.(ABSTRACT TRUNCATED AT 250 WORDS)

Change detection on longitudinal data in periodontal research

Longitudinal data of attachment level (AL) or the alveolar bone level are often used to assess the progression of periodontal disease. This paper tries to identify the most efficient method to detect the changes of AL in a general periodontal research environment; that is, a sequential decision based on multiple sites. Several existing methods suggested in the periodontal research literature such as the tolerance, running median, cusum, and regression methods as well as change-point detection methods in the statistical literature are examined. It is found that the regression method is most convenient among the several methods that are equally effective in change detection. Formulae, tables and their usage are discussed in detail.

Design and analytic concepts for periodontal clinical trials

Aspects of the design and analysis of periodontal clinical trials are surveyed from a biostatistical perspective. Design issues discussed include protocol preparation, subject selection and its documentation, randomization, problems associated with the sample versus population paradigm in sampling of microbes and gingival fluid constituents, quality control, cross-over and split-mouth versus parallel-arm designs, blinding, and multicenter trials. Analytic discussion deals with the definition and choice of analytic unit, appropriate methods for the analysis of data from multiple sites within the same subject, the nature and application of randomization tests, interim analyses, subgroup analyses, and multiple comparison issues. Examples are provided to illustrate the feasibility of analyzing site-specific data while accounting for intra-subject correlation, which represents the increased similarity of sites chosen from the same mouth as compared to sites from different patients.

Conduct of multicenter trials to test agents for treatment of periodontitis

The conduct of multicenter trials to test agents for treatment of periodontitis is conceptually similar to that of single center trials, but the requirement that centers ultimately be combined into a single result places much more stringent requirements on uniformity and control. Multicenter trials should be considered only if numbers of subjects or the ability to generalize to other treatment settings becomes a driving theme. General guidelines for the conduct of studies of this sort are suggested and problems that occur are discussed. Despite formidable difficulties that may occur, multicenter studies provide a convincing demonstration of an agent's therapeutic efficacy for the treatment of periodontitis.

The power of tests to detect differences between periodontal treatments in published studies

10 studies comparing periodontal treatment modalities were re-examined to see if they had adequate power to detect true differences. Attachment level (AL) and pocket depth (PD) were the 2 variables assessed. A statistical test's power refers to its probability of detecting a significant sample difference in treatment means, given a predetermined value for alpha (level of significance), delta (a clinically meaningful underlying difference), and the sample size. Studies were included that stratified their data by initial pocket depths, reported sample size, and lasted at least 6 months. Power calculations were done for 173 treatment comparisons, using delta = 0.5 mm and alpha = 0.05. For shallow pockets (1-3 mm), most studies had a strong chance of detecting true differences (median power = 83%). For moderate pockets (4-6 mm), median power dropped to 38%. However, median power dropped to 14% for deep pockets (> 6 mm), with 75% of the tests having less than a 20% chance of detecting a 0.5 mm difference. Many of the modalities reported as "not significantly different" from each other have not had a fair trial, especially for deep pockets. In order to improve a study's power, 4 factors are discussed: the number of compared treatments, the expected noise or random error, the patient sample size, and the average number of sites per patient for each pocket depth category.

Validity issues in split-mouth trials

The main purpose of the split-mouth design is to remove all components related to differences between subjects from the treatment comparisons. By making within-patient comparisons, rather than between-patient comparisons, the error variance (noise) of the experiment can be reduced, thereby obtaining a more powerful statistical test. Unfortunately, comparisons made on a within-patient basis have potential disadvantages. Treatments may have effects on experimental units other than those which they were assigned to (carry-across effects). Such effects cannot be estimated from split-mouth data. Neither can treatment effects be estimated. The estimable parameter in a split-mouth design is the treatment effect plus the sum of all carry-across effects. Unless a priori knowledge indicates that no carry-across effects exist, reported estimates of treatment efficacy are potentially biased. In the design of split-mouth clinical trials, potential gain in precision should be carefully weighed against a potential decrease in validity.

Reproducibility of an electronic probe in relative attachment level measurements

4 probing designs have been employed to investigate the reproducibility of the Florida Probe. 3 groups (each composed of 10 subjects) were selected for the study: healthy adults, gingivitis subjects, and periodontitis subjects. The 4 probing designs were as follows: (a) the probe tip was left in the sulcus between successive probings; (b) the probe tip was removed from the gingival margin between probings but the next probing followed immediately; (c) successive whole-mouth probings were interrupted by a 5-min interval and a mouthrinse; (d) there was a 4-week interval between each probing. 3 measurements were taken for each design. The main purpose of this study was to identify variance components in the attachment level variation. The maximum probing error standard deviation was found to be around 0.3 mm, which is considerably smaller than that found in most previous studies. The errors associated with the periodontal condition and probing effect were also estimated. The variance components obtained here can be used for determining the sample size in controlled clinical studies.

Multi-level statistical models in studies of periodontal diseases

Periodontal data typically have a hierarchical structure, with sites grouped within individuals, and individuals grouped within communities. Also, the occasion may be regarded as another level since the acquired knowledge indicates that periodontal disease activity may vary over time. Conventional statistical tests are based on unilevel analysis of data. However, this approach to statistical analysis is often inconvenient in periodontal research because of the variation in the outcome variables between the various levels in the hierarchy. Lately there have been important developments in the statistical theory which have made available powerful statistical techniques for analyzing multilevel or hierarchical data. This report describes a new approach for analyzing periodontal data and uses an illustrative example to build a model which explains part of the variability in the response variable. The results from this analysis are then compared to results from an earlier report which uses unilevel methods and the findings discussed. The present multilevel approach has several advantages over unilevel methods, mainly due to its statistical validity and efficiency. Further, it permits the incorporation of explanatory variables measured at the site and the subject levels, and those which vary across the time points. Multilevel analyses have a promising potential and are expected to have a significant impact on periodontal research.

Logical and analytic issues in dental/oral product comparison research

Possible meanings of "superiority" and "equivalence" in comparative dental studies are considered. A proposed refinement of vocabulary is introduced to precisely distinguish between these meanings. Implications for data analysis of different interpretations of "superiority" and "equivalence" are explored, as are implications for study design where design and analytic considerations are inseparable. Conceptual problems of equivalence studies are stressed. Choice of analytic unit(s), scaling of dependent variables, use of surrogate variables, and validity checking are discussed as they pertain to comparative studies. Analytic methods for correlated responses and categorized responses are briefly surveyed, particularly with reference to periodontal research.

Toward strategies for the analysis of periodontal disease clinical trials

We consider design, analysis and regulatory issues relating to clinical trials in periodontal disease and identify complications commonly associated with such studies. Alternative statistical procedures that can be used for the analyses of data from periodontal research are reviewed and a case study of the analysis of a Phase II periodontal disease clinical trial is provided to illustrate the use of one of these procedures.

Microbiology of healthy and diseased periodontal sites in poorly controlled insulin dependent diabetics

A group of poorly-controlled insulin dependent diabetes mellitus (IDDM) patients were examined in a cross-sectional design for total microbial levels, microbial incidence, and the percent levels of selected periodontal microorganisms. These organisms were selected on the basis of prior reports that associated them with either periodontal disease or health. One periodontally-healthy and one periodontally-diseased site were examined in each IDDM patient. Increased levels of the periodontal pathogens Prevotella intermedia, P. melaninogenica spp., Bacteroides gracilis, Eikenella corrodens, Fusobacterium nucleatum and Campylobacter rectus (formerly Wolinella recta) were found at the periodontal diseased sites. Increased prevalence of the organisms P. intermedia, P. melaninogenica spp., and C. rectus were found at the diseased sites. A significantly higher percentage of P. intermedia was found at the sites exhibiting deep pockets and attachment loss.

Predictive power of various models for longitudinal attachment level change

Several statistical models that have been suggested in the periodontal literature for describing longitudinal attachment level changes, such as the gradual loss, single-burst, multiple-burst, and random walk models as well as other models introduced in this paper are compared by their power to predict future attachment loss. The data used in this analysis is from 1061 sites of 8 subjects, with moderate to severe periodontal disease, monitored monthly for about a year. This study found that none of the suggested models could significantly outperform the naïve mean predictor, which predicts the future attachment level from the past mean. It was also found that no single model, such as the burst, gradual, or random walk, together with measurement error can fully explain the variation in the data. These results indicate that in the course of one year, the attachment level change may not follow the same model. Consequently, a model that fits well to past data cannot be accurately extended to the future.

Application of matched pair procedures to site-specific data in periodontal research

Application of standard statistical procedures to site-specific data in periodontal research is invalid unless site-to-site dependencies are accounted for. In this paper, we present the adjustments required for valid application of matched pair procedures, including the paired t-test and McNemar's chi 2 test for correlated proportions. Examples are given involving data arising from: (1) the comparison of pre- and post-treatment clinical measurements; (ii) split-mouth protocols.

Multicenter evaluation of tetracycline fiber therapy: I. Experimental design, methods, and baseline data

The study design and baseline characteristics of a multicenter trial to test the effectiveness and safety of locally delivered tetracycline for treatment of adult periodontitis are described. Local delivery was provided by 0.5 mm diameter ethylene vinyl acetate copolymer fibers loaded 25% with tetracycline hydrochloride which were placed into periodontal pockets and maintained by an adhesive for 10 (+/- 2) days. A total of 113 subjects (56 male and 57 female; mean age 49.3 yr) at five centers participated in the study. Subjects were selected who had 4 nonadjacent teeth with 6-10 mm pockets that bled on probing. The selected sites in each subject were randomly assigned to 4 test groups: tetracycline fiber, control fiber, scaling with root planing, or untreated. A balanced experimental design was thereby established in which each subject contributed equally by providing 4 clinically comparable sites for evaluation. To provide a more specific model for testing periodontitis therapy, gingivitis was treated prior to the initiation of the study by prophylaxis with supragingival calculus removal and home care instruction. Clinical response variables measured were pocket depth reduction, attachment level gain and bleeding on controlled-force probing measured at baseline, 30 d, and 60 d. Levels of 6 bacterial species selected as probable periodontal pathogens were measured by DNA probe analysis of plaque samples. The design of this study provided several unique analytical opportunities. Controls included a comparison with conventional treatment, analysis of vehicle effects, and effects at untreated sites. Comparison of the test group with controls permitted evaluation of the principal variables that could effect interpretation of results.(ABSTRACT TRUNCATED AT 250 WORDS)

Destructive forms of periodontal disease in adolescents. A 3-year longitudinal study

We studied the periodontal disease progression and the relationship between some forms of destructive periodontitis in a group of Brazilian adolescents with high prevalence of periodontal destruction, and evaluated the adequacy of reducing data by limiting the measurement of disease to first molars and by aggregating and pooling site-scores into subject-level scores. Over a period of 3 years 222 adolescents were examined annually by bite-wing radiographs. Individuals displaying arc-shaped bone lesions adjacent to greater than or equal to 2 first molars were diagnosed as juvenile periodontitis (JP) patients, while those with greater than or equal to 1 first molars showing vertical lesions were regarded as periodontal risk subjects. Teenagers with greater than or equal to 2 first molars exhibiting longitudinal bone loss were defined as high-risk patients. At the ages of 13 and 16 years, 3 (1.3%) and 4 (1.8%) subjects had JP, while 12 (5.4%) and 28 (12.6%) were regarded as periodontal risks; 8 (3.6%) subjects were assigned to the high-risk group; 4 (1.8%) 13-year olds had greater than or equal to 1 first molars missing; hence, no loss of posterior teeth occurred over a period of 3 years. Analyzing the data at the site-level revealed progressive loss and little fluctuation in the alveolar bone height in the high-risk group, and a more pronounced bone loss at the mesial than at the distal surfaces. Simpler data sets were constructed by aggregating some of the site-scores or by pooling these into subject-level scores.(ABSTRACT TRUNCATED AT 250 WORDS)

A study of inter- and intra-examiner reliability of pocket depth and attachment level

Factors affecting inter- and intra-examiner variability for probing pocket depth and periodontal attachment level were studied on a sample of 20 patients, each given six full-month examinations, two by each of 3 examiners. The major determinant of both dimensions of variability was the degree of periodontal destruction, measured as either pocket depth or level of attachment: the greater the degree of destruction, the greater the disagreements within and between examiners. Differences in variability were generally minor between sites on anterior and sites on posterior teeth, between midsites and proximal sites, and between sites on the facial and sites on the lingual surface. When differences were found between the average variances for different kinds of sites, they may have been due to corresponding differences between the average depths, and not to any inherently greater difficulty in measuring one kind of site than another.

The relationship between probing attachment loss and computerized radiographic analysis in monitoring progression of periodontitis

This study attempted to evaluate quantitative changes in radiographic density as an indicator of progression of periodontitis. Twenty-one subjects with a history of periodontitis were monitored at baseline, 3, 6, and 9 months using duplicate probing attachment level (PAL) measurements from stents and computer assisted densitometric image analysis (CADIA) of standardized radiographs. Results indicate that the majority of sites exhibited no PAL change during the 9-month period; however, the percentage of sites with loss increased with time. A mean of 6.1% of the sites/patient exhibited probing attachment loss during the study, as compared to a mean of 38.3% of the sites/patient that exhibited a loss of radiographic density. Due to the two dimensional nature of radiographs, density analysis was calculated in terms of radiographic "complexes" of multiple probing sites. There was significantly more density loss at complexes with greater than or equal to 2 mm of attachment loss than at sites with no change in PAL at 9 months; there was no such difference noted at 3 and 6 months. Also, density loss tended to increase as more sites within each complex experienced PAL. Although there was a significant correlation between mean density and PAL changes during the same time interval, there were wide variations at individual sites. This study suggests that there is a complex relationship between density change on radiographs and PAL change. The difficulties inherent in comparing highly sensitive new technologies to relatively imprecise clinical measurements of the attachment level are discussed.(ABSTRACT TRUNCATED AT 250 WORDS)

Cross-sectional and longitudinal reliability for clinical measurement of attachment loss

Progression of marginal periodontitis in humans is characterized by acute exacerbations during short periods of time followed by periods of remission. The analytical procedures used to distinguish actively progressing sites from non-progressing sites within an individual's dentition are controversial. The purpose of the present investigation was to use measurements of probing depth and attachment level from a prospective longitudinal study of episodic periodontal disease progression to examine measurement and diagnostic reliability. In 10 systemically healthy adult human subjects with untreated advanced periodontitis, probing depth and attachment levels were measured at baseline and every 30 days for 10 to 12 months. Measurements were made at 6 sites of each tooth using an acrylic onlay as a reference point and a pressure sensitive probe. Replicate measurements were made on sites showing apparent changes in attachment loss of 2 mm or more in a month, and their contralateral counterparts. Cross-sectional reliability was described by means and standard deviations for the absolute differences between replicate measurements: 0.63 +/- 0.87 mm for anterior teeth, 0.58 +/- 0.80 mm for premolars and 0.69 +/- 0.91 mm for molars. The mean differences tended to decrease with increasing time on study and tended to be greater in deeper pockets. Longitudinal reliability was described by analyzing site-specific serial attachment level measurements across the study time period. Maximum likelihood methods were used to estimate false positive and false negative diagnostic rates associated with the diagnosis of attachment loss based on minimum threshold levels of 1 and 2 mm. The use of a 1 mm threshold resulted in estimated false positive rates of 0.08 to 0.11 and false negative rates of 0.11 to 0.15.(ABSTRACT TRUNCATED AT 250 WORDS)

The pH of gingival crevices and periodontal pockets in children, teenagers and adults

Gingival crevice and periodontal pocket pH, measured directly with glass micro-electrodes, was near neutral at most sites in most individuals (mean pH 6.92 +/- 0.03 SEM, 69 subjects). Periodontal state ranged from healthy to periodontitis but neither clinical evidence of gingivitis at a site nor pocket depth were associated with crevicular pH different from that at healthy sites. This finding contradicts earlier reports that gingivitis is associated with a crevicular pH as alkaline as pH 9.06. Metallic antimony electrodes as used by earlier investigators were found to give pH readings that were too high by as much as 1.5 pH units in the presence of organic reducing agents of the type produced by oral bacteria within gingival crevices. In contrast, glass micro-electrodes respond only to hydrogen ions and thereby provided accurate measurements of pH even in the presence of organic reducing agents. Loss of CO2 to the atmosphere from biological fluids that are bicarbonate buffered resulted in a shift to alkaline pH by as much as 1 pH unit. As a result, only measurements taken within gingival crevices or periodontal pockets can provide accurate measurements of crevice or pocket pH.

Efficiency of split-mouth designs

The purpose of this paper is (1) to investigate the similarity of the amount, distribution, and severity of periodontal disease of the within-patient experimental units, (2) to estimate the relative efficiencies of split-mouth designs when compared to whole-mouth designs, and (3) to discuss how stratification on initial pocket depth can result in large differences in the power of the test-statistics in the different disease categories. Periodontal disease characteristics are not always homogeneously distributed over the within-patient experimental units and this heterogeneity can reduce the efficiency of split-mouth designs. In particular, if analyses are stratified on initial pocket depth, sites with an initial probing depth deeper than 6 mm may be small in number and asymmetrically distributed when compared to sites with an initial probing depth less than 6 mm. This may result in large differences of the power of the test statistics among the different disease categories and should lead to a careful interpretation of the statistical significance tests. When disease characteristics are symmetrically distributed over the within-patient experimental units and a sufficient number of sites is present per experimental unit, the split-mouth design can provide moderate to large gains in relative efficiency. In the absence of a symmetric disease distribution, whole-mouth clinical trials may be preferable.

A 6-year study on the pattern of periodontal disease progression

The present longitudinal radiographic investigation was designed to study the change in the alveolar bone height over 6 years in relation to tooth type, age and the presence of previous bone loss in a group of 142 subjects (age 18-67 years) who were not under systematic periodontal treatment. The subjects were examined radiographically at baseline, and after 2 and 6 years, and sites showing significant (greater than or equal to 1 mm) change in the alveolar bone height during 2 consecutive examinations were identified. 90.2% of all examined sites showed no bone loss during the 6 years, while 8.6% and 1.2% of the sites exhibited bone loss during one or both examination periods, respectively. 3 groups of subjects within the studied population were identified. Approximately 5% of the subjects had high rates of periodontal disease progression whilst approximately 70% demonstrated very few or no sites with bone loss, and approximately 25% had a moderate level of disease progression during the study period. The results also showed that the number of sites exhibiting alveolar bone loss during the study period varied with respect to tooth type, age of the patient and presence of bone loss at baseline. In addition, significantly different rates of bone loss were disclosed with relation to tooth type (P less than 0.01), age (P less than 0.0001) and initial bone loss (P less than 0.01). It was concluded that periodontal disease progression occurs infrequently and may take the form of discrete or recurrent episodes of loss of periodontal support which can be influenced by the location of the site, age of the subject and previous periodontal disease experience.

A new approach to investigating associations in periodontal disease data

This paper presents a method for the analysis of data originating from studies of destructive periodontal disease. The proposed method is an extension of the Mantel-Haenszel technique for the analysis of case-control studies and allows for expression of the site-specificity of destructive periodontal disease while maintaining the individual as the unit of analysis. Using data originating from a cross-sectional study of periodontal breakdown and oral hygiene parameters in a random sample of adult rural Kenyans, the proposed method is illustrated and the results compared with results obtained when two alternative analytical methods are used. The results demonstrate that the choice of analytical strategy may have profound implications for the conclusions to be drawn. Depending on the strategy chosen, one may draw conclusions which are qualitatively different and the present study indicates that the direction of these differences is not predictable.

Statistical models for data from periodontal research

Many factors have been hypothesised either to characterise groups and individuals at risk for periodontal disease or to be markers of periodontal breakdown. In order to identify these as associated either with disease status or progression, a statistical association between the factor and a measure of disease will have to be demonstrated. The statistical modelling of data arising from periodontal research presents special problems. These include the large number of measurements made in each subject, the large magnitude of measurement error compared to the changes in attachment level, the analysis of longitudinal studies, the lack of a measure of instantaneous rate of attachment loss and controversies over the nature of the progression of the disease. We consider statistical methods currently available in the light of these difficulties and identify areas in which further research is necessary.

Adjustment of frequently used chi-square procedures for the effect of site-to-site dependencies in the analysis of dental data

Donner and Banting (1988) presented a method of adjusting the standard chi-square test for the comparison of proportions when this procedure is applied to multiple-site dental data. The adjusted procedure allows for a comparison of two or more groups of subjects with respect to the prevalence over all sites of a specified characteristic, while taking site-to-site dependencies into account. In this paper, similar adjustments are presented for two other frequently used chi-square procedures, the chi-square test for linear trend and the Mantel-Haenszel test. Examples are given.

A 10-year retrospective study of periodontal disease progression

The purpose of this study was to record the rate of periodontal disease progression over a 10-year period with respect to individual subject, age and tooth type. 283 subjects, who had undergone a full-mouth radiographic examination in 1974-76 and at that time were 25-70 years old, were randomly selected from a larger patient sample for a new radiographic examination in 1985-86. 201 subjects (71%) agreed to participate. For each respondent, tooth loss over the 10-year period was calculated. The radiographic bone height at the mesial and distal aspect of all teeth was assessed by measuring the distance between the cemento-enamel junction and the bone crest. Difference in periodontal bone height between the 2 examinations was calculated for each tooth site. The results revealed that the mean number of teeth lost over the 10-year period was 3.8 (SD 4.6). Tooth mortality, expressed as a % of the no. of teeth present at the initial examination, increased with age and varied between 2.9%-28.5%. In all age groups, molars had the highest and canines the lowest rate of tooth mortality. The frequency of subjects with loss of greater than 10 teeth was highest for the initially 45-year old individuals (20%). 7 individuals (3%) had become edentulous. The mean annual reduction of alveolar bone height varied between 0.07-0.14 mm in the age groups 25-65 years. The initially 70-year old subjects showed a statistically significant higher annual rate of bone loss (0.28 mm).(ABSTRACT TRUNCATED AT 250 WORDS)

The detection of in vitro produced periodontal bone lesions by conventional radiography and photographic subtraction radiography using observers and quantitative digital subtraction radiography

Changes in the periodontium produced by removal of bone cylinders at one interdental site of a dry human mandible, were recorded radiographically. These artificial lesions had diameters of 0.3 mm increasing to 1.4 mm in steps of 0.1 mm. Radiographs were obtained using 3 different X-ray tube potentials and 3 different amounts of radiation. These 9 exposure variables resulted in 9 series of radiographs from the artificial lesions. Observers were asked to determine the presence or absence of these periodontal bone lesions on conventional radiographs and on photographically subtracted images. The images were also evaluated by a quantitative digital subtraction technique. This study showed that the smallest periodontal bone changes were detected with the quantitative digital subtraction technique compared to the other methods using observers. On photographically subtracted images, smaller bone changes were detected by the observers than on conventional radiographs. Only the detection threshold of the quantitative digital subtraction technique was influenced by the 2 exposure factors: kVp and mAs.

The impact of regression towards the mean on probing changes in studies on the effect of periodontal therapy

The purpose of this report was to evaluate the magnitude of the regression towards the mean (RTM) relative to the observed changes in probing measurements following periodontal therapy in 2 groups of patients. Regression lines were calculated for observed changes in probing depths and probing attachment levels related to the initial probing depths of the individual sites. Regression lines, adjusted for the RTM effect, were also calculated. For the extreme subgroups of recorded shallow and deep sites, the RTM effect (mm) was found to amount to a maximum of 0.6 mm. It appears unlikely that erroneous conclusions have been made in previous clinical studies evaluating probing depth changes in deep sites, since the RTM accounts for a limited portion of the observed changes. For probing depth changes in shallow sites and for probing attachment level changes in both shallow and deep sites, the RTM accounts for a larger proportion of the observed mm changes. It therefore seems prudent to be careful about interpretation of these changes in clinical studies, where the RTM effect has not been adjusted for. Whenever possible, duplicate initial recordings should be taken to allow determination of and subsequent adjustment for the RTM effect.

Public health implications of recent research in periodontal diseases

Knowledge of the epidemiology, natural history, and bacterial etiology of the periodontal diseases has advanced considerably as a result of research conducted through the 1980s. Prevention and control of these conditions, however, remains mechanical, cumbersome, and often impractical, based as it is on bacterially nonspecific plaque removal for an indeterminate period. This research has not yet changed the content of public health programs, but it does affect the way the programs are applied. Because sever, generalized disease seems to be less prevalent than previously thought, the need of regular, routine professional care for everybody is questioned. Professional care in a public health context is likely to be more efficient when targeted toward those with severe disease. Dental health education for personal oral hygiene is still supported by scientific studies, though a targeted approach and careful assessment of educational content is needed. Until predictive screening methods for identifying susceptible individuals are developed, selection of priority groups for education and treatment should be guided by epidemiologic data.

Analysis of site-specific data in dental studies

Much attention has been given recently to issues in the statistical analysis of site data arising from clinical trials in periodontics. As pointed out by Imrey (1986) and Fleiss et al. (1987), it is unreasonable to expect sites within the same patient to respond independently, and thus conventional statistical methods are not directly applicable. In this paper we develop a procedure for comparing the percentage of sites showing a given characteristic across two or more groups of patients that accounts for this lack of independence. The procedure is a straightforward extension of the standard Pearson chi-square test. Two examples are given.

Accuracy of repeated single versus averages of repeated duplicates of probing depth measurements

The present study aims to determine to which extent averages or deepest recordings of repeated duplicates of probing depth measurements may increase the accuracy as compared to repeated single measurements. In total, 1247 sites were recorded in 8 patients with advanced periodontitis. The patients had received initial treatment. Each site was probed 4 times with time intervals of 100 min. The standard deviation of differences between repeated single measurements of 0.97 mm decreased with a factor square root of 2 to 0.69 mm for differences between averages of repeated duplicates of measurements. Differences between the averages and the deepest recordings of repeated duplicates of measurements showed a non-Gaussian distribution. This implies that the type-I error can not be computed on the basis of the standard deviation of the error with the use of parametric statistical analysis. The best estimate for the type-I error is the observed frequency of differences. The type-I error for differences of 3 mm or more between repeated single measurements decreased from 1.5% to 0.9% and 0.2%, when deepest recordings and averages of repeated duplicates of measurements, respectively, were compared.

Microbiologic assessments to enhance periodontal diagnosis

Associations between microbes and progressive periodontal lesions provide the biologic rationale for using specific organisms as bacteriologic markers for disease activity. Rapid, specific and relatively inexpensive laboratory tests to identify bacteria have facilitated increased testing by clinicians. However, assays must be interpreted with regard to current concepts of etiology and pathogenesis of periodontal diseases. This paper addresses commonly used microbiologic assays and their shortcomings.

A re-examination of within-mouth correlations of attachment level and of change in attachment level

Using a data set from the Forsyth Dental Center consisting of repeated measurements of periodontal attachment level on 22 patients over 1 year, we performed correlation analyses at 4 time points (baseline, 1 month, 6 months and 9 months) for 3 units of measurement (sites, teeth and quadrants). Values of the intraclass correlation coefficient were obtained for the cross-sectional measurements and for changes from baseline. The values were positive and often of appreciable magnitude, indicating that measurements of attachment level and of change in attachment level within the same mouth are positively correlated. Taking sites rather than patients as the units of statistical analysis in comparative clinical trials in periodontology will therefore tend to produce underestimated standard errors and overstated statistical significance.

Change in attachment level

Attachment level or pocket depth measurements are used to detect change in periodontal attachment. However, measurement error limits their usefulness for this purpose. The aims of this study are twofold. The first aim was to test the assumption, used in previous reports of measurement error, that attachment level measurements are normally distributed. The second aim was to estimate error rates encountered when assessing periodontal attachment level change with attachment level or pocket depth measurements. Two simulation methods are compared in their ability to reproduce the distribution of differences in replicated measurements. A simulation method based on a normal distribution was not able to reproduce the actual distribution of differences between replicate measurements. In contrast simulations based on resampling recreated the distribution of differences in replicated attachment level measurements. Due to the inability of simulations based on the normal distribution to reproduce the distribution of differences in replicated measurements, simulations based on the resampling procedure were used to estimate error rates. Determining change in attachment level or pocket depth by either single or paired measurements resulted in low Type I error rates for both single and paired measurements. For single measurements of attachment level, the Type I error rate was 0.0074 for a 3-mm change. Paired measurements of attachment level had a Type I error rate of 0.0014 for a 3-mm change. However, false-positive rates were found to be much higher. For single measurements of attachment level, the false-positive rate was 0.32 for a 3-mm change.(ABSTRACT TRUNCATED AT 250 WORDS)