BACKGROUND: The Ottawa knee rule is a clinical decision aid that helps rule out fractures and avoid unnecessary radiography. PURPOSE: To summarize evidence about the accuracy of the Ottawa knee rule. DATA SOURCES: Relevant English- and non-English-language articles were identified from PreMEDLINE and MEDLINE (1966-2003), EMBASE (1980-2003), CINAHL (1982-2003), BIOSIS (1990-2003), the Cochrane Library (2002, Issue 3), the Science Citation Index database, reference lists of included studies, and experts. STUDY SELECTION: Articles were included if they reported enough information to determine the sensitivity and specificity of the Ottawa knee rule for detecting fractures confirmed either radiologically or in combination with follow-up. DATA EXTRACTION: Two reviewers independently extracted data on study samples, the ways that the Ottawa knee rule was used, and methodologic characteristics of studies. DATA SYNTHESIS: Of 11 identified studies, 6 involving 4249 adult patients were considered appropriate for pooled analysis. The pooled negative likelihood ratio was 0.05 (95% CI, 0.02 to 0.23), the pooled sensitivity was 98.5% (CI, 93.2% to 100%), and the pooled specificity was 48.6% (CI, 43.4% to 51.0%). CONCLUSION: A negative result on an Ottawa knee rule test accurately excluded knee fractures after acute knee injury. However, because the rule is calibrated toward 100% sensitivity and actual fracture prevalences are usually low, large-scale, multicentered studies are still needed to establish the cost-effectiveness of routinely implementing the rule.