PT  - JOURNAL ARTICLE
AU  - Britto, J
AU  - Domecq, J
AU  - Murad, M
AU  - Guyatt, G
AU  - Montori, V
TI  - 081 The Endocrine Society Guidelines: Implications of Strong Recommendations with Low Quality Evidence
AID  - 10.1136/bmjqs-2013-002293.112
DP  - 2013 Aug 01
TA  - BMJ Quality & Safety
PG  - A38--A38
VI  - 22
IP  - Suppl 1
4099  - http://qualitysafety.bmj.com/content/22/Suppl_1/A38.2.short
4100  - http://qualitysafety.bmj.com/content/22/Suppl_1/A38.2.full
SO  - BMJ Qual Saf2013 Aug 01; 22
AB  - Background In 2005, the Endocrine Society (TES) adopted the GRADE system of developing clinical practice guidelines. This system facilitates the formulation of evidence-based recommendations by explicitly describing the confidence in estimates (quality of evidence) and strength of each recommendation. Objectives To describe and characterise the relationship between confidence in estimates and strength of recommendation in TES guidelines. Methods We included all published TES guidelines between 2005 (when TES started using GRADE) and 2011. Independently and in duplicate, reviewers extracted, for each recommendation: disease area, confidence in estimates and design of the studies considered, and strength of recommendation. In strong recommendations with low quality of we developed and applied a taxonomy of appropriate recommendations and identified those we considered inappropriate. Results Most of the 357 recommendations issued were supported by evidence warranting low or very low confidence in estimates (256, 72%). Evidence cited in support of these recommendations came exclusively from observational studies in 233 recommendations (65%). Most recommendations were strong (206, 58%); of these, 121 (59%) were supported by evidence warranting low or very low confidence in estimates. In 101/121 (83%), we identified a compelling rationale for the recommendations; in 20 (17%), we did not. Conclusions Most TES strong recommendation based on low quality evidence are justified and appropriate, but a substantial proportion are not. Implications for Guideline Developers Guideline developers should carefully justify any strong recommendations based on low confidence in effect estimates.