LAUSR

Differences in effect estimates between early primary trials included in a meta-analysis and the pooled estimate of meta-analysis might indicate potential novelty bias. The objective of this study was to assess the presence of novelty bias in a sample of studies published in periodontology and implant dentistry. On August 7, 2020, we searched the PubMed database for meta-analyses of clinical studies published between August 2015 and August 2020. Meta-analyses with at least 4 primary studies were selected for assessment. We fitted logistic regression models using trial characteristics as predictors to assess the association between these characteristics and 1) the odds of the first trial's estimate to be included in the meta-analysis confidence interval (CI) and 2) the odds of overlap between the first trial's CI and the meta-analysis prediction interval (PI). Ninety-two meta-analyses provided data for assessment. In absolute values, 70% of the meta-analyses have a pooled estimate smaller than the corresponding estimate of the first trial, although there was overlap of the CI of estimates from the first trial and the meta-analysis in 87% of the cases. This is probably due to the small number of trials in most meta-analyses and the subsequently large uncertainty associated with the pooled effect estimate. As the number of trials in the meta-analysis increased, the odds of the treatment effect estimate of the first trial to be included in the meta-analysis CI decreased by 15% for every additional trial (odds ratio, 0.85; 95% CI, 0.73 to 0.96). Meta-analytic effect estimates appear to be more conservative than those from the first trial in the meta-analysis. Our findings show evidence of novelty bias in periodontology and implant dentistry; therefore, clinicians should be aware of the risk of making decisions based on the information reported in new trials because of the risk of exaggerated estimates in these trials.