I am writing in reference to the paper, Longevity of materials for pit and fissure sealing—Results from a meta-analysis, by Künisch et al., Dental Materials 2012;28:298–303. The authors suggest two conclusions based on their meta-analysis: firstly that resin-based sealants are to be recommended (no problem with that); secondly that light-polymerizing materials are preferred over the auto-polymerizing types. This latter conclusion is based on the assumption that light-polymerizing materials are “faster and less error-prone”. Both the preference and the reasoning are incorrect, in my opinion.

I would argue that most light-polymerizing materials take longer to finish if one takes into account the need to fully irradiate each surface for the required minimum time for full cure at full depth (unless, possibly, only one surface of one tooth is addressed). Similarly, the assumption that auto-polymerizing materials are less error-prone is false. The example concerned the incorporation of bubbles in the mixing process. Apart from the fact that careful mixing precludes the incorporation of most bubbles (and that more advanced methods would preclude it entirely), the far more important issue of stress-relaxation during polymerization being much more favorable for the auto-polymerizing materials was ignored. In addition, the disadvantage of a costly light that requires regular testing and maintenance for full output was ignored, as was the difficulty of knowing that any particular aliquant of sealer is fully polymerized, especially for opaque sealants, and in sites of limited access. Nevertheless, this preference and reasoning emerges out of the blue: neither was supported by data; neither was the subject of the study. Gratuitous opinions cannot be conclusions.

Secondly, it seems that the authors relied on wildly implausible data for light-polymerizing sealants in their rush to anoint these materials as the best—one of the graphs in Fig. 1 showed that light-polymerizing sealants possess the remarkable aptitude to improve in terms of retention over time! This, of course, is patently ridiculous, yet the authors do not address this glaring impossibility (their model gave 2-, 3-, and 5-year fitted retention rates of 77.8%, 80.4% and 83.8% respectively [Authors’ Table 1]). There must either be a miscalculation (Bayes notwithstanding), or their adopted model invalidated itself. No sealant can improve in retention over time unless defects are regularly rectified, but this of course would render a study of longevity of retention null and void. In any event, the ‘ground truth’ of the data has not been recognized. To report fitted-model estimates as facts is unhelpful, at best.

There are other concerns as well (such as the non-independence of repeated observations), but I find it disappointing that neither our colleagues nor the reviewers could see the incomprehensibility of this last point and, at the very least, prompt some discussion of it in the text, if not the abandonment of their meta-analysis in favor of some other more realistic model.

Using an unquestioned impossible result to justify unsubstantiated opinion cloaked as a definitive conclusion for a supposedly authoritative clinical recommendation is not what I would expect in such an important analysis, from such an honorable team of authors, in such a respected scientific journal.