could identify the same outcome in a different study population. Outcomes are measured either by objective means or by self-report. For self-reported outcomes, however, there are explicit criteria for how the data were collected and evidence that the collection method would permit another investigator to repeat the study in another population.
The exposure measures are well defined. Self-report of exposure is acceptable so long as the method of collecting self-reports was well specified and there was evidence that the self-reports were reliable reflections of exposures. Job titles as surrogates for exposure were acceptable when the exposure of interest was inherent in the job (e.g., vibration exposure for those operating pneumatic chipping hammers).
The article was published in English.
The article was peer reviewed.
The study was done within the last 20 years (preferably).
No specific limitations were placed on study designs acceptable for consideration. The advantages of prospective studies, however, were recognized. For example, there were sufficient prospective studies of low back pain to examine these separately among the studies of physical factors and exclusively among the studies of psychosocial factors.
The literature reviews were conducted using computer-based bibliographic databases, with MEDLINE (National Library of Medicine, United States of America) a component of all searches. Additional databases included: NIOSHTIC (National Institute for Occupational Safety and Health, United States of America), HSELINE (Health and Safety Executive, United Kingdom), CISDOC (International Labour Organization, Switzerland), Ergoweb (Internet site of the University of Utah), Psychinfo, Oshrom, Ergonomics Abstracts, and ArbLine (National Institute for Working Life, Sweden).
The bibliographies of articles (particularly review articles) and the NIOSH comprehensive review (Bernard, 1997b) were examined to identify additional relevant articles.
Using these sources, a candidate list of articles was established and then systematically screened to determine which ones met the strict criteria, described above, for inclusion in the review. Each process reduced the list substantially. For physical work factors studied in association with back disorders, 255 studies were initially identified as relevant and 41 met the selection criteria and were reviewed. For psychophysical factors and back disorders, the search resulted in 975 references, which were then reduced to 21 work-related risk factor studies and 29 individual risk factor studies. For work-related physical factors and upper extremity disor-