Comparator choice in cariology trials limits conclusions on the comparative effectiveness of caries interventions

Comparator choice has been found as one major factor impacting on the overall evidence supporting clinical interventions. We performed social network analysis (SNA) on trials on the prevention/management of caries/carious lesions, hypothesizing that certain comparators are proportionally overinvestigated, and others underinvestigated, and that comparisons within comparator classes are preferred over comparisons between classes.

A systematic review of randomized controlled trials on the prevention/management of caries/existing carious lesions was carried out. All comparators were classified at each of three levels of granularity, becoming more detailed with each level: (1) degree of invasiveness (noninvasive, microinvasive, or invasive), (2) the specific noninvasive, microinvasive, or invasive approach, and (3) the actual material or technique used. SNA was used to evaluate trial networks.

Searching electronic databases found 4,774 articles; of which, 764 were relevant and 605 were included. The networks for all levels were polygonal. There was a high degree of separation of comparisons in prevention vs. management trials. Invasive comparators were tested most frequently (number of comparators: 592), mainly in management trials. Noninvasive comparators were tested next often (464), mainly in caries prevention. Microinvasive strategies were tested next often (234), in both prevention and management trials. On more granular levels, few interventions dominated the networks. Regardless of the level, most trials compared within and not between classes. Prevention trials were mainly conducted in children (number of trials in adults/children/both: 37/240/11), whereas those on managing lesions were conducted in both children and adults (117/176/21).

Comparator choice in cariology trials is driven by indication and limits conclusions on the true comparative effectiveness of all strategies. There are a variety of comparators that have not been, but should be, compared with one another, which should be addressed by future trials. Factors underlying trialists' comparator choice need to be identified.