牙科随机对照试验中的发表偏倚:哪些因素影响结局的统计学显著性
该研究纳入2017年1月1日至2021年12月31日发表于牙科6个专科领域12本高影响因子期刊的474项随机对照试验,其中321/474(67.7%)报告了统计学显著的主要结局。多变量模型显示专科领域(p=0.01)、洲别(p=0.003)和注册(p=0.004)与显著性相关;牙髓病学和正畸学的随机对照试验较牙周病学更不易报告显著效应,注册试验报告显著结局的比值更低(OR=0.52)。作者认为各牙科领域均存在偏好发表“成功”且统计学显著结局的现象。
要点
- 474项牙科随机对照试验中321项(67.7%)报告了统计学显著的主要结局。
- 牙髓病学(OR=0.40)和正畸学(OR=0.41)较牙周病学更不易报告显著效应。
- 注册试验报告显著结局的比值低于未注册试验(OR=0.52;95%CI:0.34,0.81)。
核心结论
- 各牙科领域均存在偏好发表被认为“成功”且统计学显著结局的现象,正畸学和牙髓病学相对更平衡。
- 试验未注册仍较普遍,且与报告统计学显著效应相关。
研究设计与样本
- 研究纳入2017年1月1日至2021年12月31日发表于牙周病学、牙髓病学、修复牙学/口腔修复学、正畸学、儿童牙科、口腔颌面外科6个领域各2本最高影响因子期刊(Clarivate Analytics,2020)的全部随机对照试验。
- 共识别并纳入474项随机对照试验,主要结局为报告显著/不显著主要结局的比例,并检验期刊、发表年份、影响因子、资助、注册等特征与显著性的关联。
主要发现
- 474项随机对照试验中321项(67.7%)报告了统计学显著的主要结局。
- 多变量模型显示专科领域(p=0.01)、洲别(p=0.003)和注册(p=0.004)对显著性有显著影响。
- 与牙周病学相比,牙髓病学随机对照试验报告统计学显著效应的比值更低(OR=0.40;95%CI:0.22,0.76),正畸学亦更低(OR=0.41;95%CI:0.23,0.74)。
- 有强证据表明注册试验报告统计学显著结局的比值更低(OR=0.52;95%CI:0.34,0.81)。
关键数值与推荐
- 主要结局为随机对照试验报告显著/不显著主要结局的比例。
- 多变量模型检验的预测因素包括专科领域、洲别、注册等特征。
临床怎么用
- 作者指出,该实证报告的结果提示应关注系统评价结论的解读,因其很大程度上取决于某一主题下随机对照试验结局的可获得性和性质,可能影响合并效应的合成估计及其方向。
局限性
- 原文未报告局限性
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
OBJECTIVES: To record the proportion of Randomized Controlled Trials (RCTs) reporting significant (versus non- significant) primary outcomes, published across 12 high impact journals in Dentistry, covering 6 specialty domains. Associations with certain journal, publication and outcome characteristics were examined. METHODS: We identified and included all RCTs published from January 1st, 2017 to December 31st, 2021 in the two journals with the highest impact factors (Clarivate Analytics, 2020) from each of the following domains: Periodontology, Endodontics, Restorative Dentistry/ Prosthodontics, Orthodontics, Paediatric Dentistry, Oral and Maxillofacial Surgery. The primary outcome was the proportion of significant/ non- significant findings reported for the primary outcomes under study, while a range of characteristics such as: journal, year of publication, impact factor, funding, registration and others, were tested for associations. RESULTS: A total of 474 RCTs were identified and included, with the majority reporting statistically significant outcomes (321/474; 67.7%). The multivariable model revealed significant effects of predictors related to specialty domain (p = 0.01), continent (p = 0.003) and registration (p = 0.004). Compared to Periodontology, RCTs published in Endodontics (OR= 0.40; 95%CIs: 0.22, 0.76) and Orthodontics (OR= 0.41; 95%CIs: 0.23, 0.74) were less likely to present statistically significant effects. There was strong evidence that registered trials presented lower odds of reporting statistically significant findings (OR= 0.52; 95%CIs: 0.34, 0.81). CONCLUSIONS: The entirety of dentistry domains demonstrated preferential publication practices of outcomes considered as "successful" and statistically significant, with domains such as Orthodontics and Endodontics being more balanced. Trial non- registration is still prevalent and associated with reporting of statistically significant effects. CLINICAL SIGNIFICANCE: The findings of this empirical report bring attention to the interpretation of Systematic Reviews (SRs) conclusions. These largely depend on the availability and nature of outcomes of randomized controlled trials (RCTs) on a topic, which may impact on the synthesized estimate of a pooled effect and its direction.
引用
- Mikelis F, Tzanetakis GN, Eliades T, et al. Publication bias in randomized controlled trials in dentistry. What factors affect statistical significance of outcomes? J Dent. 2022. PMID: 35690226. doi:10.1016/j.jdent.2022.104183
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。