Fourteen studies (19 articles) that provided outcome data for 8,600 participants were included. Six studies were RCTs and eight were CCTs. Four of the eight CCTs were considered fair quality, three were good quality and one was poor quality. One of the six RCTs received a Jadad score of 5, one a score of 3, one a score of 2 and three a score of 1.
Based on random-effects models, dental caries was significantly reduced with xylitol-containing chewing gum versus no chewing gum (prevented fraction 58.66%, 95% CI 35.42 to 81.90, p<0.00001; six studies), with xylitol-sorbitol-containing chewing gum (prevented fraction 52.82%, 95% CI 39.64 to 66.00, p<0.00001; five studies) and with sorbitol-containing chewing gum (prevented fraction 20.01%, 95% CI 12.74 to 27.27, p<0.00001; five studies), but not with sorbitol-mannitol-containing chewing gum (prevented fraction 10.71%, 95% CI -20.50 to 41.93, p=not significant; three studies).
There was significant heterogeneity in all meta-analyses. None of the sensitivity analyses for the xylitol or xylitol-sorbitol studies resulted in the intervention effect becoming non-significant. The effect became non-significant for the sorbitol studies when only RCTs were considered.
The regression analysis for xylitol suggested that larger doses were associated with a greater effect (R2=49%). For sorbitol the correlation was smaller (R2=34%).
Results for Rosenthal's fail-safe N suggested that a large number of unpublished trials with a null-effect (between 89 and 565) would be required to offset the effects seen.