To expand on this: the heterogeneity is so high that they don't (as meta-analysis guidelines say you should) report the figures. This is why in the figures, you see tight confidence intervals around zero, but also equally-tight confidence intervals many effect-size units away. That's an alarm bell for trying to draw any sort of meaningful conclusion: it means there are so many variables at play that you have no idea whether the net should be zero or big. (Also a concern given the observation that many countries have seen dental problems fall considerably as they developed, which is a big confound.)