Article type
Year
Abstract
Background: The GRADE (Grading of Recommendations Assessment, Development and Evaluation) methodology has been proposed to grade quality of evidence and strength of recommendations in clinical guidelines.
Objectives: To assess if Cochrane systematic reviews (SRs) contain all the information to compile the GRADE profiles.
Methods: We compiled the GRADE profile using the GRADE software for each outcome included in a preliminary list identified by the panel involved in producing international World Health Organization guidelines for treatment of opiate addiction. We analysed: number of profiles and outcomes for which we had to go back to primary studies; missing information; number of clinical outcomes for which we had to compute again the effect measures (relative risk (RR), risk difference (RD), weighted mean difference (WMD)).
Results: We compiled seven profiles (28 outcomes). For 5/7 profiles and 15/28 outcomes we had to go back to primary studies. The most frequent missing information were: setting, country, number of subjects allocated to groups, length of follow up, raw data for SRs without meta-analysis. Information on quality of primary studies was not always exhaustive. We had to recalculate 22/28 effect sizes. The RD and the range of risk in the control group were never reported in Cochrane SRs. Interpretation of guidelines on how to downgrade the evidence is not always straightforward, particularly when only one study is included or only qualitative results are reported.
Conclusions: Cochrane SRs do not report all the required information to compile the GRADE profiles for clinical guideline development. There is a need for defining a detailed list of information to be reported in Cochrane SRs relevant to policy decisions to avoid going back to primary studies. RevMan should be adapted to compute the required quantitative information. Many systematic reviews do report qualitative evidence, but this cannot be properly taken into consideration in the present version of the GRADE software.
Objectives: To assess if Cochrane systematic reviews (SRs) contain all the information to compile the GRADE profiles.
Methods: We compiled the GRADE profile using the GRADE software for each outcome included in a preliminary list identified by the panel involved in producing international World Health Organization guidelines for treatment of opiate addiction. We analysed: number of profiles and outcomes for which we had to go back to primary studies; missing information; number of clinical outcomes for which we had to compute again the effect measures (relative risk (RR), risk difference (RD), weighted mean difference (WMD)).
Results: We compiled seven profiles (28 outcomes). For 5/7 profiles and 15/28 outcomes we had to go back to primary studies. The most frequent missing information were: setting, country, number of subjects allocated to groups, length of follow up, raw data for SRs without meta-analysis. Information on quality of primary studies was not always exhaustive. We had to recalculate 22/28 effect sizes. The RD and the range of risk in the control group were never reported in Cochrane SRs. Interpretation of guidelines on how to downgrade the evidence is not always straightforward, particularly when only one study is included or only qualitative results are reported.
Conclusions: Cochrane SRs do not report all the required information to compile the GRADE profiles for clinical guideline development. There is a need for defining a detailed list of information to be reported in Cochrane SRs relevant to policy decisions to avoid going back to primary studies. RevMan should be adapted to compute the required quantitative information. Many systematic reviews do report qualitative evidence, but this cannot be properly taken into consideration in the present version of the GRADE software.