055 Use of Non-Randomised Studies In The Guideline Process: The GRADE Approach. (15th August 2013)

• Record Type: Journal Article
• Authors: Santesso, N; Mustafa, R; Schunemann, H
• Is Part Of: BMJ quality & safety. Volume 22(2013)Supplement 1
• Page Start: A29
• Page End: A30
• Publication Date: 2013-08-15
• Subjects: Medical care -- Quality control -- Periodicals; Health facilities -- Risk management -- Periodicals; Medical errors -- Prevention -- Periodicals; 362.106805
• DOI: 10.1136/bmjqs-2013-002293.86
• Languages: English
• ISSNs: 2044-5415

Abstract

Abstract : Background: The GRADE approach to guideline development requires a review of the best available evidence which includes randomised controlled trials (RCTs) and non-randomised studies (NRS). Objectives: Describe the use of NRS as a replacement, a sequence, or a complement for RCTs, in a World Health Organization guideline using the GRADE approach. Methods: We searched the literature using no study type limits for the effect of screening and treatment of precancerous lesions on patient or population important outcomes and for baseline risks. We assessed quality of the evidence using GRADE. Results: Depending on the outcomes, we found few to no RCTs. When there was low/very low overall quality evidence from RCTs, we used NRS studies with no independent control groups to compare proportions between groups and calculate a relative effect of treatment and this evidence replaced the RCT evidence with similar/higher quality evidence. We found no evidence in RCTs for long-term outcomes, such as spontaneous abortion. Therefore, we used data from NRS (cohort studies) for premature delivery (a surrogate) to provide sequential evidence. For evidence about baseline risk of precancerous lesions and other outcomes, we used NRS a complement to the RCT data. Discussion: Data from NRS provided evidence in three ways. One key criterion to consider when grading this evidence is indirectness due to indirect comparisons, surrogate outcomes or varying population risks. Implications for Guideline Developers/Users: Guideline developers can use GRADE and these methods when there is no evidence or low/very low quality evidence from RCTs.