Evaluation of the Quality of Prognosis Studies in Systematic Reviews

Abstract

To provide valid assessments of answers to prognostic questions, systematic reviews must appraise the quality of the available evidence. However, no standard quality assessment method is currently available. To appraise how authors assess the quality of individual studies in systematic reviews about prognosis and to propose recommendations for these quality assessments. English-language publications listed in MEDLINE from 1966 to October 2005 and review of cited references. 163 systematic reviews about prognosis that included assessments of the quality of studies. A total of 882 distinct quality items were extracted from the assessments that were reported in the various reviews. Using an iterative process, 2 independent reviewers grouped the items into 25 domains. The authors then specifically identified domains necessary to assess potential biases of studies and evaluated how often those domains had been addressed in each review. Fourteen of the domains addressed 6 sources of bias related to study participation, study attrition, measurement of prognostic factors, measurement of and controlling for confounding variables, measurement of outcomes, and analysis approaches. Reviews assessed a median of 2 of the 6 potential biases; only 2 (1%) included criteria aimed at appraising all potential sources of bias. Few reviews adequately assessed the impact of confounding (12%), although more than half (59%) appraised the methods used to measure the prognostic factors of interest. Reviews may have been missed by the search or misclassified because of incomplete reporting. Validity and reliability testing of the authors' recommendations are required. Quality appraisal, a necessary step in systematic reviews, is incomplete in most reviews of prognosis studies. Adequate quality assessment should include judgments about 6 areas of potential study biases. Authors should incorporate these quality assessments into their synthesis of evidence about prognosis.