Article type
Year
Abstract
Introduction:
Objectives: To discuss methodological aspects of conducting systematic reviews on occupational risk factors for musculoskeletal pain.
Methods: We have recently completed three systematic reviews evaluating occupational risk factors for low back, neck, and shoulder pain. In contrast to reviews of clinical trials, methodological guidelines for conducting systematic reviews of (occupational) observational studies are not yet available. We encountered several problems related to the search strategy, quality appraisal, data extraction, and analysis. These problems and some potential solutions will be presented at the Cochrane Colloquium.
Results: Optimal search strategies for aetiological studies have not yet been designed. The harvest of our computerized searches was rather poor: 28 out of 1363, 51 out of 1026, and 21 out of 560 titles, respectively, met the selection criteria (2-5%). Separate, but similar checklists were designed for the quality appraisal of cohort, case-control, and cross-sectional studies. Only 44%, 10% and 65% of the studies on back, neck, and shoulder pain scored at least 50% of the maximum attainable scores. Disagreements between two independent reviewers ranged between 16 and 20% of all scored items. Some items that did not discriminate well between studies may have to be excluded from the checklists (e.g. clearly stated objective). Other items caused many disagreements (e.g. data collection on history of pain, use of an appropriate statistical model) and may need revision. Data extraction and analysis was severely hampered by the wide range in exposures and methods used for assessment of exposures and outcomes. The often inadequate data presentation impeded statistical pooling. Quantitative measures of association were frequently missing, and adjustments for covariates were often absent or insufficiently reported. The use of two different methods of combining the results in a best-evidence synthesis were considered: an evaluation of criteria of causality, and the design of levels of evidence
Discussion: Despite the methodological hurdles that still need to be taken concerning the identification of relevant studies, quality appraisal, data extraction, and synthesis of available evidence, the application of a systematic approach for reviews on occupational risk factors for musculoskeletal pain is worthwhile.
Objectives: To discuss methodological aspects of conducting systematic reviews on occupational risk factors for musculoskeletal pain.
Methods: We have recently completed three systematic reviews evaluating occupational risk factors for low back, neck, and shoulder pain. In contrast to reviews of clinical trials, methodological guidelines for conducting systematic reviews of (occupational) observational studies are not yet available. We encountered several problems related to the search strategy, quality appraisal, data extraction, and analysis. These problems and some potential solutions will be presented at the Cochrane Colloquium.
Results: Optimal search strategies for aetiological studies have not yet been designed. The harvest of our computerized searches was rather poor: 28 out of 1363, 51 out of 1026, and 21 out of 560 titles, respectively, met the selection criteria (2-5%). Separate, but similar checklists were designed for the quality appraisal of cohort, case-control, and cross-sectional studies. Only 44%, 10% and 65% of the studies on back, neck, and shoulder pain scored at least 50% of the maximum attainable scores. Disagreements between two independent reviewers ranged between 16 and 20% of all scored items. Some items that did not discriminate well between studies may have to be excluded from the checklists (e.g. clearly stated objective). Other items caused many disagreements (e.g. data collection on history of pain, use of an appropriate statistical model) and may need revision. Data extraction and analysis was severely hampered by the wide range in exposures and methods used for assessment of exposures and outcomes. The often inadequate data presentation impeded statistical pooling. Quantitative measures of association were frequently missing, and adjustments for covariates were often absent or insufficiently reported. The use of two different methods of combining the results in a best-evidence synthesis were considered: an evaluation of criteria of causality, and the design of levels of evidence
Discussion: Despite the methodological hurdles that still need to be taken concerning the identification of relevant studies, quality appraisal, data extraction, and synthesis of available evidence, the application of a systematic approach for reviews on occupational risk factors for musculoskeletal pain is worthwhile.