Background: The evidence stemming from trials on restorative materials is shaped not only by trial findings, but also trial design and validity. We aimed to evaluate both aspects in randomized controlled dental restorative trials published from 2005–2015. Methods: Using systematic review methodology, we retrieved trials comparing restorative or adhesive dental materials. Two authors independently assessed design, risk of bias, registration status, and findings of trials. Descriptive and regression analyses were performed. Results: 114 studies on 15,321 restorations placed mainly in permanent teeth of 5232 patients were included. Per trial, the median number of patients was 37 (25th/75th percentiles: 30/51). Follow-up was 24 (20/48) months. Seventeen percent of trials reported on sample size calculations, 2% had been registered. Most trials (90%) used US Public Health Service (USPHS) criteria, and had a high risk of bias. More recent trials were more likely to have been registered, to have reported on sample size calculations, to be of low risk of bias, and to use other than USPHS-criteria. Twenty-three percent of trials yielded significant differences between groups. The likelihood of such differences was significantly increased in older studies, studies with potential reporting bias, published in journals with high impact factor (>2), longer follow-up periods, and not using USPHS-criteria. Conclusions: The majority of dental restorative trials published from 2005–2015 had limited validity. Risk of bias decreased in more recent trials. Future trials should aim for high validity, be registered, and use defined and appropriate sample sizes, follow-up periods, and outcome measures.

Bonding of resin composite fillings, for example following root-canal treatment, is a challenge because remaining gaps grow and lead to failure. Here, phase-contrast-enhanced micro-computed tomography (PCE-CT) is used to explore methods of non-destructive quantification of the problem, so that countermeasures can be devised. Five human central incisors with damaged crowns were root-filled followed by restoration with a dental post. Thereafter, the crowns were rebuilt with a resin composite that was bonded conventionally to the tooth with a dental adhesive system (Futurabond U). Each sample was imaged by PCE-CT in a synchrotron facility (ID19, European Synchrotron Radiation Facility) with a pixel size of 650 nm. The reconstructed datasets from each sample were segmented and analysed in a semi-automated manner usingImageJ. PCE-CT at sub-micrometre resolution provided images with an impressive increased contrast and detail when compared with laboratory micro-computed tomography. The interface between the dental adhesive and the tooth was often strongly disrupted by the presence of large debonded gaps (on average 34% ± 15% on all surfaces). The thickness of the gaps spanned 2 µm to 16 µm. There was a large variability in the distribution of gaps within the bonding area in each sample, with some regions around the canal exhibiting up to 100% discontinuity. Although only several micrometres thick, the extensive wide gaps may serve as gateways to biofilm leakage, leading to failure of the restorations. They can also act as stress-raising `cracks' that are likely to expand over time in response to cyclic mechanical loading as a consequence of mastication. The observations here show how PCE-CT can be used as a non-destructive quantitative tool for understanding and improving the performance of clinically used bonded dental restorations.