宫内节育器临床研究设计与统计分析专家共识

Chinese experts’consensus on study design and statistical analysis for intrauterine device clinical research

我国现有宫内节育器(IUD)使用者超过1亿,约占世界使用总数的三分之二,但我国IUD临床研究总体质量不高,不利于保护广大育龄妇女的生殖健康。为指导国内医务工作者合理选择和使用IUD临床研究设计和统计分析方法,中华医学会计划生育学分会专家经充分讨论并结合我国实际制定本共识。共识指出,IUD临床研究全过程都应从流行病学与统计学角度考虑问题,包括但不局限以下内容:①明确研究问题,将其转换为可用数据回答的统计问题;②根据研究问题和实施条件匹配合理的研究设计,并考虑研究设计控制混杂和偏倚的能力;③在考虑混杂因素的基础上制定数据采集的内容和工作计划,按计划采集和整理数据;④合理选择分析数据集,并使用恰当的统计方法。当受试者特征分布不均衡时,应采用分层、倾向性评分、多元统计等方法进行数据分析;⑤分析结果的解释应回应原始研究问题,研究结论外推要考虑研究对象的代表性。

To date,the number of intrauterine device(IUD)users in China is more than 100 million,accounting for about two thirds of the world IUD users.Previous studies showed that the overall quality of IUD clinical researches in China was low,which potentially threatened women’s reproductive health.After a full discussion,experts from Chinese Society of Family Planning China Medical Association initiated and developed this consensus,which aimed to instruct Chinese clinicians to correctly use study designs and statistical analysis methods to carry out IUD clinical studies.This consensus points out that all kinds of IUD clinical studies should follow principals of epidemiology and medical statistics,including but not limited 1)define study questions and transform them into statistical items that can be answered by figures;2)select correct types of clinical study designs,which based on study questions,practical conditions and the ability of confounding control of the adopted study design;3)develop case report forms and data collect plan,considering the known confounding;collect and clean up study data as planned;4)select a proper dataset that minimizes potential risks of biases and confounding and employ correct statistical methods for data analysis.If the distribution of participants’characteristics is imbalanced,stratifying analysis,propensity score and/or multivariable regression analysis should be used to controll potential confounding;5)finding interpretations should answer the original study questions.Extrapolation of results from IUD clinical studies should consider the representativeness of study participants.