Zimmerman S
University of Texas, M.D. Anderson Cancer Center, Houston.
J Periodontal Res. 1992 Jul;27(4 Pt 2):425-32. doi: 10.1111/j.1600-0765.1992.tb01709.x.
This paper reviews considerations involved in the interpretation of results of individual periodontal clinical trials and the synthesis of findings of multiple studies. The purpose of each study must be considered, particularly whether it is an epidemiological investigation, prevention trial or therapeutic clinical trial. The definition of clear end-points, whether they be hygiene end-points associated with calculus and plaque or disease end-points that measure extent of gingivitis and periodontitis, is essential. Other statistical considerations include the following: adequate sample size, balanced design, use of covariates, issues of the appropriate sampling unit, and proper statistical analysis. Statistical issues that must be considered include the types of tests of hypotheses, use of multiple comparison procedures when more than two groups are involved, corrections for multiple testing and precise estimation of parameters with corresponding confidence intervals. When comparing across clinical trials, three sources of heterogeneity that must be considered are: treatments, diseases, and target populations. The technique of meta-analysis is described for the statistical interpretation of data arising from several trials. A meta-analysis is conducted for six dental clinical trials that evaluated a triclosan/copolymer formulation for the reduction of plaque, calculus or gingivitis. Results indicate that the first five studies, which were reported simultaneously, showed a common effect, i.e., that the formulation successfully reduced either plaque or calculus. When the sixth study, reported at a different time, was added to the analysis, the collection of studies did not satisfy the homogeneity assumption necessary for estimating a common effect size. This illustrates the importance of testing for homogeneity before pooling possibly dissimilar studies.
本文回顾了个体牙周临床试验结果解读及多项研究结果综合分析中涉及的相关考量因素。必须考虑每项研究的目的，尤其是它是流行病学调查、预防试验还是治疗性临床试验。明确终点的定义至关重要，无论是与牙石和菌斑相关的卫生学终点，还是衡量牙龈炎和牙周炎程度的疾病终点。其他统计学考量因素包括：足够的样本量、均衡设计、协变量的使用、合适抽样单位的问题以及恰当的统计分析。必须考虑的统计学问题包括假设检验的类型、涉及两组以上时多重比较程序的使用、多重检验的校正以及参数的精确估计和相应的置信区间。在比较不同的临床试验时，必须考虑的三个异质性来源是：治疗方法、疾病和目标人群。文中描述了荟萃分析技术用于对多项试验产生的数据进行统计学解读。对六项评估三氯生/共聚物配方减少菌斑、牙石或牙龈炎效果的牙科临床试验进行了荟萃分析。结果表明，同时报告的前五项研究显示出共同效应，即该配方成功减少了菌斑或牙石。当加入另一时间报告的第六项研究进行分析时，这些研究的集合不满足估计共同效应大小所需的同质性假设。这说明了在汇总可能不同的研究之前进行同质性检验的重要性。
