Commentary

This well-conducted Cochrane review has tested the efficacy of EMD (Emdogain) in comparison with open-flap debridement, GTR and various bone-grafting procedures, over a follow-up period of at least 1 year. The authors conclude that EMD is able to significantly improve PAL levels and PPD reduction compared with flap surgery, but note that the results may not have great clinical impact.

The review amply demonstrates the heterogeneity of clinical-trial methodology, which is so characteristic of periodontal RCT. Although the rigour of the Cochrane review methodology leaves little room for criticism, a few points may be noteworthy. The review identifies 10 RCT that fulfil the inclusion criteria. A quality assessment of these trials place only three trials in the high-quality category whereas five and two trials, respectively, are placed in the medium- and low-quality categories as regards the risk of bias. Nevertheless, all 10 RCT are included in the analyses.

The authors perform a metaregression analysis to assess the effect of various study characteristics — including trial quality — on the results and demonstrate that the mean effect of EMD on PAL change is 0.42 mm lower in the high-quality studies than in the lower-quality studies. Even so, the authors use absence of statistical significance to conclude that they were unable to explain the heterogeneity found between the studies. This, we submit, is a good illustration of the tyranny of the P value. We suggest that the combination of trial quality (0.42 mm), place of conduct (0.89 mm) and use of antibiotics (0.86 mm) may explain the heterogeneity between studies.

The review shows that a statistical analysis was performed in all 10 trials to test the hypothesis of equality of the baseline characteristics of test and to control patients/defects. If baseline differences are observed in RCT, it is a matter of definition that they have arisen by chance. It does not take a statistical test to prove this point. The purpose of randomisation is to obtain a balanced distribution of known, as well as unknown, confounders. Such a balance may be difficult to achieve in small studies, however, such as most of the RCT included in this review. Rather than performing statistical tests, the distribution of known confounders across treatment arms should be described (obviously this is not possible for the unknown confounders). If important imbalances are noted (not a statistical hypothesis testing issue) some measure of confounder-control in the analysis of the data would be appropriate.

We agree with the authors of the review when they state that the CAL improvements attributable to EMD therapy may not have a great clinical impact. We question, however, whether the quality of the RCT available is sufficiently high to allow for a more definite estimation of the clinical efficacy of this intervention.

Practice point

  • Currently, the evidence for a possible benefit of EMD in the treatment of intrabony defects is rather weak.