Department of Orthodontics and Dentofacial Orthopedics, Hellenic Air Force General, Hospital, Athens, Greece.
Department of Orthodontics and Dentofacial Orthopedics, School of Dentistry, University of Bern, Bern, Switzerland.
J Periodontol. 2023 Aug;94(8):967-975. doi: 10.1002/JPER.22-0653. Epub 2023 Mar 8.
Clustering effects can be encountered in periodontology and implant dentistry research. The aim of this study was to identify studies with clustering effects published in periodontology and oral implantology specialty journals and to assess the frequency by which clustered designs are correctly accounted for in the statistical analysis.
Ten periodontology and oral implantology journals were searched to identify studies with clustering effects published between January 1, 2019 and July 31, 2021. Descriptive statistics and frequency distributions were calculated. Associations between the correct statistical handling of clustering effects and study characteristics were explored.
A total of 695 studies were included of which 45.0% correctly accounted for clustering effects in the statistical analysis. Certain journals (p < 0.01) and animal studies (p < 0.01) had lower odds of correctly accounting for clustering effects in the statistical analysis, whereas per unit increase in impact factor (p < 0.001), involvement of statistician (p < 0.001) and when the study design included either repeated measures only (p < 0.01) or both clustering and repeated measures (p < 0.001) had higher odds. The most commonly used tests were the mixed models or generalized estimating equations (64.2%).
Greater awareness of the importance of accounting for clustering effects is required to prevent incorrect inferences being drawn. Incorrect inferences are related to lack of data independence and the artificial inflation of the sample size which can result in statistically significant results which are not genuine. This issue can be further exaggerated in combination with publication bias.
在牙周病学和种植牙科研究中可能会遇到聚类效应。本研究的目的是确定在牙周病学和口腔种植学专业期刊上发表的具有聚类效应的研究，并评估在统计分析中正确考虑聚类设计的频率。
搜索了 10 种牙周病学和口腔种植学期刊，以确定 2019 年 1 月 1 日至 2021 年 7 月 31 日期间发表的具有聚类效应的研究。计算了描述性统计和频率分布。探讨了正确处理聚类效应与研究特征之间的关联。
共纳入 695 项研究，其中 45.0%的研究在统计分析中正确考虑了聚类效应。某些期刊（p<0.01）和动物研究（p<0.01）在统计分析中正确考虑聚类效应的可能性较低，而影响因素的单位增加（p<0.001）、统计学家的参与（p<0.001）以及研究设计包括仅重复测量（p<0.01）或同时包括聚类和重复测量（p<0.001）时，正确考虑聚类效应的可能性较高。最常用的检验是混合模型或广义估计方程（64.2%）。
需要提高对正确考虑聚类效应重要性的认识，以防止得出不正确的结论。不正确的结论与数据独立性的缺乏以及样本量的人为膨胀有关，这可能导致统计学上显著的结果，但这些结果并不是真实的。在与发表偏倚相结合的情况下，这个问题可能会进一步加剧。
