Cochrane系统评价中如何处理分配方案隐藏(英文)

How Allocation Concealment Is Handled in Cochrane Reviews

目的 ①记录和分析纳入Cochrane系统评价研究的分配方案隐藏描述和分级情况 ;②向各评价小组反馈存在的问题或错误的分级及分级方式 ;③建议怎样正确描述和分级分配方案隐藏。方法 提取Cochrane图书馆 2 0 0 3年第 1期的 15 96篇系统评价有关分配方案隐藏分级和描述情况。共提取了来自 10个Cochrane系统评价小组 (CRGs)的 10 8% ( 173/ 15 96 )系统评价 (共 184 4个研究 )的资料 ;记录和分析了分配方案隐藏描述和分级与Cochrane系统评价手册的方法有差异或矛盾的研究。结果 目前 ,纳入Cochrane系统评价研究的分配方案隐藏的分级方式有可能不太准确。其原因在于对分配方案隐藏方法的描述有误 (占 184 4个研究中 38 6 % ) ,而其中分级错误至少占 9 2 %。分配方案隐藏为A的研究中 ,最常见的是各种信封法 ( 133/ 6 75 ,占19 7% )。而在分配方案隐藏为B的研究中 ,最常见的是分配方案隐藏方法不清楚或未提及 ( 2 6 / 6 6 5 ,占 6 4 % )。结论 因正确的分配方案隐藏对防止随机对照试验出现偏倚至关重要 ,所以必须对分配方案隐藏进行准确的分级和描述。我们需要改进Cochrane系统评价纳入研究中的分配方案隐藏的描述和分级。因半数以上分配方案隐藏为D的研究是系统评价员疏漏了分级造成缺省 。

Objective ① To document the way in which allocation concealment is described and coded for studies included in Cochrane Reviews. ② To feed back any gaps or miscodings to individual review groups. ③ To suggest changes and expansions to advice on how to code and describe allocation concealment methods.Methods The coding and description of methods of allocation concealment for studies included in all 1 596 reviews on issue 1, 2003 of The Cochrane Library are being extracted.So far results are available for 10.8% (173/1 596) of reviews containing 1 844 studies, from 10 Collaborative Review Groups (CRGs).Discrepancies, and inconsistencies with the Cochrane Reviewers' Handbook, are being documented and analysed.ST5HT5SSResults The current coding of the adequacy of allocation concealment in studies included in Cochrane reviews is not likely to be very accurate.This is due to failure to describe methods of allocation concealment (38.6% of the sample of 1 844 studies) as well as miscoding (at least an additional 9.2%).The most common method for studies coded A was some variation of envelope use (133/675-19.7% of all A codes). The most common 'method' for studies coded B was method unclear or not described in the report of the study (426/665, 64% of all B codes). Conclusions Since adequate allocation concealment is so important in protecting against bias in randomised controlled trials, it needs to be accurately coded and described.We need to improve how this is done for studies included in Cochrane Reviews.Since over half the studies coded as D were likely to have been where reviewers omitted to enter a code, the default should be changed from D to 'code not supplied'.Structural changes to RevMan are suggested-ideally the addition of a separate new study quality assessment table with fixed headings as well as the facility to enter free text.Suggestions for improving coding in particular reviews will be fed back to CRGs in the next stages of this project.Suggestions for additions to the Cochrane Reviewers' Handbook are also made.