RESUMO
Evidence synthesized based on randomized controlled trials (RCT) results are recognized as the pinnacle of research excellence; however, the conduction of RCT in workplace environment is not always possible. This study comparatively reviewed evidence from RCT and non-RCT studies in which participants performed workplace exercise for musculoskeletal pain control. Up to February 2011, PubMed, MEDLINE, Embase, Cochrane, PEDro and Web of Science databases were searched. All trials that evaluated workplace exercise interventions for controlling musculoskeletal pain were included. The PEDro scale was used to rate the studies' quality, PRISMA and Cochrane recommendations were applied, and association between frequencies of effect size categories (small, moderate, large) from various outcomes by study type was tested (2x3 contingency table). The search yielded 10239 references in English, from which 21 RCT and 12 non-RCT were selected. Both groups of studies presented methodological flaws including descriptions of randomization, blinding of examiners and absence of intention-to-treat analysis for the RCT, and further absence of controls and blind assessor for the non-RCTs. RCTs had significantly more moderate and large effect size reported in their results compared to non-RCTs (p=0.04). Considering the difficulties in randomizing participants in occupational settings, all studies would benefit from better describing pertinent methodological information.