2.1

Evidence from laboratory trials

A systematic search of the dental literature up to September 2014 identified eight laboratory and nine clinical controlled trials that compared HVGICs against dental amalgam. Analysis of the trial data showed that the effect direction identified by the two types of trials were not the same: Laboratory results indicate statistically significant (p < 0.05) better results for amalgam than for HVGIC, while results from clinical controlled trials indicate no statistically significant difference (p > 0.05). The effect magnitude, measured as the median with 25% and 75% percentile range of the Standard mean difference (SMD) between HVGIC and amalgam per trial, for laboratory trials was SMD −3.04 (−5.34; −0.24) and that for clinical controlled trials was SMD −0.08 (-0.38; 0.12). The difference in effect magnitudes between laboratory and clinical controlled trial results was highly statistically significant (p = 0.0014) .

This indicates that evidence from laboratory trials, particularly for head-to-head comparisons of HVGIV versus amalgam, does neither reflect the effect direction nor the effect magnitude of clinical controlled trial results and thus cannot be considered suitable for clinical guidance.

2.2

Evidence from uncontrolled clinical (longitudinal) trials

A systematic literature search for trials published between 2002 and 2012 identified ten and two prospective, uncontrolled clinical trials for HVGICs and amalgam, respectively. Within the same time period three randomised control trials (RCT), comparing HVGIC versus amalgam restorations were found. Similar to the results from laboratory trials , naïve-indirect comparison between HVGIC and amalgam restorations of the uncontrolled trial results yielded a different effect direction and magnitude than direct HVGIC versus amalgam comparisons in RCTs. The established effect direction from uncontrolled trials showed in favour of amalgam restorations and a more than six times higher Odds of HVGIC restoration to fail than that of comparable amalgam restorations as effect magnitude (median Odds ratio = 6.29 with 25% and 75% percentile of 1.34 and 19.27, respectively). In contrast, the established effect direction from RCTs indicated no superiority/inferiority of both restoration type and a statistically non-significant Odds ratio in the failure rate of both types of restoration as effect magnitude (median Odds ratio = 1.00 with 25% and 75% percentile of 0.81 and 1.20, respectively). The difference in effect magnitudes between naïve-indirect comparisons of data from uncontrolled trials and direct comparisons from RCTs was highly statistically significant (p = 0.0013) .

This indicates that, like evidence from laboratory trials, evidence from uncontrolled clinical trials does neither reflect the effect direction nor the magnitude of clinical controlled trial (RCT) results and thus should not be considered suitable for clinical guidance.

2.3

Evidence from controlled clinical trials without ART

The number of prospective clinical controlled trials that directly compared HVGIC restorations placed after conventional cavity preparation against amalgam or composite restorations as potential gold standard for load-bearing posterior tooth restoration in permanent teeth, is limited. A search in PubMed up to the 29th of April 2016, using a number of different strings of search terms (see Additional file 1), revealed only seven published trial reports of which four [10,11 and 12,13] reported results from the same trials at different follow-up periods. The quality of all trials needs to be rated as poor, due to lack of reported adequate randomisation method, lack of reported allocation concealment and lack of reported sample size calculation. In addition, all trials included only low sample sizes, ranging total at baseline between 61 and 222 tooth restorations per trial. Consequently, the non-statistically significant (p > 0.05) differences between the results of the intervention groups shown in all trials may thus be ascribed to low statistical power, due to too low samples sizes and not to clinical merits of the placed restoration types. Statistical pooling of all trial results by meta -analysis may further prove difficult, due to large in-between-trial differences in, e.g. type of control intervention, follow-up period, and patient characteristics.

2.4

Controlled clinical ART trials

The bulk of clinical evidence regarding HVGIC restorations in head-to-head comparison against amalgam or composite restorations originates from controlled clinical trials with HVGIC restorations placed using ART. While the evidence for HVGIC versus composite resin restorations is still scant , more evidence is available for HVGIC versus amalgam. A systematic review in 17 English and non-English databases accepted 38 trials that included more than 10 000 placed tooth restorations . The results indicated no statistically significant differences (p > 0.05) in the failure rates between HVGIC and amalgam restorations in single- and multiple surface tooth cavities after up to six years. While many of these trials also suffered from too low sample sizes, their results could be sufficiently pooled in at least three meta -analyses, combining their total sample sizes to between 942 and 3762 tooth restorations at point of follow-up. Pooling of results was permissible, because of only moderate in-between-trial differences (I ² = 40.5–62.7%). All trials suffered from high bias risk, due to insufficient randomisation and allocation concealment, high risk of performance- and detection bias, as operator, patient and evaluator blinding was impossible, due to the obvious differences in appearances between HVGIC and amalgam, as well as high risk of attrition bias, due to loss-to-follow up. In contrast, the risk of publication and reporting bias appeared to be low .

The inclusion of ART trials as evidence for HVGIC restorations has been repeatedly questioned . The main reasons were the apparent differences in evaluation criteria for ART restorations and how ART restoration failures were defined as compared to prevailing USPHS criteria, commonly applied to other restoration types . Another reason for rejecting ART evidence was concerns regarding the clinical conditions under which ART restorations were placed . However, all objections are relevant to uncontrolled ART studies, only, and not to controlled trials where patient allocation to either HVGIC or conventional (e.g. amalgam or composite resin) restorations is conducted randomly, thus effectively distributing treatment conditions evenly and where the same evaluation criteria and definitions are applied to all restoration groups and not just to HVGIC/ART restorations.

A subsequent systematic literature search, conducted in October 12, 2016, indicated no change in the systematic review conclusions on which this short communication is based (Additional file 2).