How to determine the best treatment: a mixed-treatment-comparisons meta-analysis (MTM) of trials of topical fluoride therapies for the prevention of dental caries

Tags: Oral
Salanti G, Higgins J, Marinho V

Background: A series of Cochrane reviews provides evidence on the effectiveness of topical fluoride therapies in various forms (toothpaste, gel, varnish and mouth-rinse) for the prevention of dental caries. Whereas evidence is based on a large number of trials for some forms of fluoride, there is more uncertainty about the effectiveness of other forms. Tailored placebo groups and among-study variation in trial settings contribute to heterogeneity and complicate the analysis. Mixed-treatment-comparisons meta-analysis (MTM) offers a technique to synthesise the entirety of the evidence in a single analysis taking into account sources of heterogeneity. Moreover, MTM can be used to rank treatments in effectiveness despite a lack of trials that compare them directly.

Objectives: We aim to determine the most effective treatment among fluoride toothpastes, gels, varnishes and rinses using evidence from all relevant randomised trials, to increase precision in the estimates of the least studied interventions, and to evaluate heterogeneity and inconsistency across trials and across comparisons.

Method: We compiled meta-analyses of placebo-controlled trials and head-to-head comparisons of the aforementioned fluoride interventions previously published in The Cochrane Library. We undertook a random-effects MTM synthesis within a Bayesian framework and meta-regression was used to account for variable study-level characteristics. We explore the consistency of information from different sources, and compare several assumptions regarding the compatibility of the various placebo implementations.

Results: The meta-analysis includes 150 trials corresponding to 13 different comparisons. The precision increases, compared with simple pair-wise comparisons, for the less studied treatments when MTM is applied. Fluoride varnish appears as the most effective treatment, and the other fluoride forms seem similarly effective. Accounting for the different forms of placebo does not change the core conclusions of the meta-analysis and we did not identify important disagreements among the direct and indirect sources of information. However, examples are given where information is inconsistent, yielding misleading results.

Conclusions: The analysis of treatment networks is undertaken with greater precision and offers more clinically relevant conclusions when MTM is applied. However, the use of such models should be accompanied with justification of their plausibility, and applied after evaluation of the underlying assumptions.