Exercise treatment for work-relevant neck pain: does the body-of-literature used for a qualitative synthesis of trial results differ according to methodological quality assessment approach?

Gabrielle van der Velde, Maurits van Tulder, Pierre Côté, Sheilah Hogg-Johnson, Peter Aker, J. David Cassi

Background: Two systematic review methods are commonly used in the spine-pain literature: the Cochrane Collaboration Back Review Group (CCBRG) and best-evidence synthesis (BES). They differ by their approach to assessing the quality of clinical trials and qualitatively synthesising trial results, suggesting a potential for discrepant conclusions regarding effectiveness.
Objective: To determine whether the body-of-literature used for a qualitative synthesis of results of trials examining the effectiveness of exercise for work-relevant neck pain (WRNP) differs according to review approach: the CCBRG and BES methods.
Methods: MEDLINE (1966-2003), EMBASE (1980-2003), CINAHL (1982-2003), and Cochrane Controlled Trials Register (2003) were searched without language restrictions. Randomised or controlled clinical trials of exercise for WRNP were included. Two approaches were used to assess methodological quality of identified trials and qualitatively synthesise results. The CCBRG method assessed quality using a scale from which was calculated a summary score. High quality trials were defined as those scoring a minimal value (6 or greater) out of a maximal score (11). Results were to be synthesised by stratification into levels of evidence based on quality scores. The BES approach critically appraised methodology. Trials were accepted on strength of their scientific merit or rejected due to methodological shortcomings. Results were to be synthesised according to consistency of the evidence across accepted studies.
Results: The search yielded 3920 non-duplicate citations; 40 were identified as WRNP studies; 13 fulfilled the inclusion criteria. CCBRG quality scores for these 13 trials ranged from two to seven out of 11; two trials were defined as high-quality. All 13 trials were accepted for inclusion into a CCBRG qualitative synthesis. Of 13 trials appraised by BES, four were accepted for inclusion into its qualitative synthesis (including the two high-quality studies accepted by CCBRG approach).
Conclusion: The body of literature which would be used for the CCBRG and BES qualitative syntheses is different. This may have important consequences on their respective conclusions on the effectiveness of exercise treatment for WRNP.

Ottawa 2004 O-059