Methods
Participants
The EULAR task force on GC therapy is a multidisciplinary committee consisting of 16 experts from 7 European countries (8 rheumatologists, 1 endocrinologist, 1 rheumatologist/epidemiologist, 4 rheumatic patients as patients' representatives and 2 research fellows). The objective was to formulate 10 recommendations on the management of medium/high-dose systemic GC therapy in rheumatic diseases by identifying and critically appraising evidence in the literature. The strength of each recommendation was evaluated.
Experts' Consensus and Delphi Rounds
As a first step, a general systematic literature search was performed aiming at identifying prospective follow-up studies in which medium/high-dose GC therapy was administered systemically. This search was not limited to rheumatic diseases (see online supplementary appendix 1 for details on this search). We used the databases PubMed, EMBASE and Cochrane Library; search results on adverse events (AEs) were expressed in events per patient year and odds ratios (ORs) (not corrected for disease activity or comorbidity) and summarised in tables (see online supplementary appendix 2), using the software Comprehensive Meta Analysis V2. The results were presented at the first group meeting to initiate group discussions identifying important topics. After the first meeting, each task force member independently formulated 10 propositions related to management of medium/high-dose GC use in rheumatic diseases. The Delphi technique was used to reach consensus on the propositions as follows. The initial propositions were listed and overlapping propositions were amalgamated. The list was returned to the members with the request to select the 10 most important propositions in this first round. A proposition was accepted if over three-quarters of the members selected it in the first round, two-thirds in the second round, and half in the third and fourth rounds. A proposition was removed if it was selected by one-quarter of the participants or less in the first round, one-third or less in the second round, and half or less in the third and fourth rounds. After 4 rounds, 10 propositions of which the text had been optimised by an English native speaker remained and were agreed upon by all participants.
Systematic Literature Search of the 10 Propositions
After agreement on 10 propositions, additional proposition-specific searches were performed using PubMed, EMBASE and Cochrane Library by two research fellows (see online supplementary appendix 3 for details on inclusion and exclusion per search). Results of the different databases were combined and duplicates were excluded; issues regarding inclusion or exclusion of articles were resolved by discussion and consensus. Articles evaluating the value of a recommendation were selected and in case of lack of evidence, circumstantial evidence was looked for. References of articles found were screened for additional evidence.
Categorising Evidence and Strength of Ecommendations
The quality of evidence based on study design was categorised according to the EULAR hierarchy (Table 1).
After the proposition-specific literature searches, evidence regarding each of the recommendations was subjected to group discussion; the final recommendations were approved by all members. For each proposition, the strength of recommendation (SOR) was graded using an A–E ordinal scale (A=fully recommended, B=strongly recommended, C=moderately recommended, D=weakly recommended and E=not recommended) and a visual analogue scale (VAS, 0–100 mm, 0=no agreement and 100=maximal agreement). The members were asked to consider both the quality of evidence presented and their own clinical experience while grading. For each proposition, the mean VAS and 95% CI, and the percentage of strongly to fully recommended (A–B) propositions were calculated. This grading method has not been fully evaluated, but is, in our view, valuable to give SOR for recommendations which cannot be or have not been assessed in randomised controlled trials (RCT); SOR has been used for other EULAR recommendations too.
During the meetings, members were asked to discuss items which should be the focus of future research. These items were combined into a research agenda.