Earlier work showed how to perform fixed-effects meta-analysis of studies or trials when each provides results on more than one outcome per patient and these multiple outcomes are correlated. That fixed-effects generalized-least-squares approach analyzes the multiple outcomes jointly within a single model, and it can include covariates, such as duration of therapy or quality of trial, that may explain observed heterogeneity of results among the trials. Sometimes the covariates explain all the heterogeneity, and the fixed-effects regression model is appropriate. However, unexplained heterogeneity may often remain, even after taking into account known or suspected covariates. Because fixed-effects models do not make allowance for this remaining unexplained heterogeneity, the potential exists for bias in estimated coefficients, standard errors and p-values. We propose two random-effects approaches for the regression meta-analysis of multiple correlated outcomes. We compare their use with fixed-effects models and with separate-outcomes models in a meta-analysis of periodontal clinical trials. A simulation study shows the advantages of the random-effects approach. These methods also facilitate meta-analysis of trials that compare more than two treatments.

Cost-effectiveness analysis was used to evaluate alternative methods of periodontal disease control. The alternatives considered included non-surgical and surgical procedures as well as the use of antimicrobial agents. Data on costs were obtained from American Dental Association publications of average charges for periodontal services. The concept of quality-adjusted tooth-years (QATYs) was developed to provide an outcome measure which could be compared across treatments. The conclusions of this analysis are as follows: (1) Conservative non-surgical treatments for periodontal disease control not only have costs lower than surgical alternatives, as would be expected, but also maximize expected quality-adjusted tooth-years over a wide range of estimates; (2) antimicrobial therapy used as an adjunct to non-surgical treatment is likely to be both effective and cost-effective; and (3) quality of tooth-years is a critical consideration in the determination of outcome of periodontal treatment. For example, when tooth-years are not adjusted for quality, differences between treatments are diminished, and surgical treatment becomes as good as or better than more conservative treatments for some levels of disease severity.

A meta-analysis was performed on 5 randomized controlled trials comparing surgical with non-surgical treatment for periodontal disease. The specific procedures considered were the modified Widman flap compared with scaling and root planning or curettage with anesthesia. We chose the most consistently reported outcomes, pocket depth and attachment level, for analysis. At 1 year of follow-up, surgical treatment reduced pocket depth more than non-surgical for all initial levels of disease, but by 5 years, only the deepest initial pockets (> 7 mm) showed significant improvement over non-surgically treated teeth (0.51 mm reduction, p < 0.01). Attachment level showed significantly better early results for non-surgical treatment for less diseased teeth, but by 5 years, all significant differences had disappeared. We computed quality scores following a method described by Chalmers. The mean quality score for study data analysis and presentation was 0.37 +/- 0.009 and for the study protocol, the mean quality score was 0.19 +/- 0.002. We find that this meta-analysis supports findings relating response to therapy with initial level of disease severity. We also find that the choice of outcome measure influences the choice of therapy, with surgical therapy providing greater benefit for probing depth and non-surgical therapy providing greater benefit for attachment level. These results must be viewed, however, in light of the low quality scores of the evaluated studies and the potential for bias due to lack of binding, the small mean treatment differences, and the observer measurement variability.

The presence of paired or multiple organs (arches, quadrants, teeth) and the chronic nature of many dental diseases suggest the use of split-mouth (trials in which each subject receives greater than or equal to 2 treatments, each to a separate section of the mouth) and cross-over research designs (trials in which each subject receives greater than or equal to 2 treatments in sequence). While these designs offer potential savings in resources, their usefulness can be negated if several strict scientific and statistical assumptions are not met. The primary prerequisites for the use of split-mouth and cross-over designs are that: (1) the disease to be investigated is relatively stable and uniformly distributed; (2) the effects of the treatments to be evaluated are short-lived or reversible for cross-over studies, or are localized for split-mouth designs. Other important factors that influence the appropriate use of these designs include: the method of treatment sequencing and assignment, and the cross-over rules used; blinding of patient assignment, patients and observers; assessment of order effects including period, carry-over or spill-over effects; the choice of statistical analysis, the sample size utilized, and the special importance of patients lost to study or of faulty data points. The objective of this study was to review 3 journals for studies using split-mouth or cross-over designs to determine how the assumptions underlying these research designs are considered and applied in dental research. The majority of studies used adequate methods for treatment allocation and sequencing; however, many studies failed to take advantage of the research designs in the statistical analysis of data. In addition, very few studies considered the possibility of order effects or reduced bias through blinding procedures.

The results of periodontal therapy vary by disease severity, outcome measure, and method of data analysis. Several clinical trials and a subsequent meta-analysis have demonstrated that, for teeth with severe disease, surgery decreases probing depth (PD) and increases attachment level (AL) more than non-surgical treatment. For other disease levels, the choice of therapy depends on the outcome measure. When clinical trials use two or more outcome measures (such as PD and AL), investigators ordinarily analyze each outcome separately. When the correlations are incorporated among the outcomes, a meta-analysis can use generalized-least-squares (GLS) regression to analyze multiple outcomes jointly. We applied the GLS multiple-outcomes model in a meta-analysis of 5 trials comparing surgical and non-surgical periodontal treatments, each assessing the outcomes PD and AL one year after treatment. The clinical conclusions are similar to those reported earlier, but our estimates of the relative benefits of surgical and non-surgical treatment should be more accurate, because the GLS method takes into account correlation between AL and PD. When correlations between the two outcomes rise, as they do with increasing severity of disease, the GLS estimates depart from those derived from separate analyses of PD and AL.