OBJECTIVE: The purpose of this study was to investigate the extent to which verification bias affects the sensitivity and specificity of MRI in the diagnosis of cruciate ligament tears. MATERIALS AND METHODS: Consecutively registered outpatients who underwent MRI evaluation of the knee were included in the study. The sensitivity and specificity of MRI were calculated for patients whose diagnosis was verified with arthroscopy. For patients who did not undergo arthroscopy, the effect of verification bias was estimated with global sensitivity analysis, a technique of graphic representation of whether a particular combination of sensitivity and specificity estimates is compatible with the observed data. RESULTS: Among the 356 patients included in the study, 82 patients (23%) had the MRI findings verified at arthroscopy. The sensitivity and specificity of MRI among patients who underwent arthroscopy were 38% and 90%. For patients whose disease status was not verified with arthroscopy, the influence of verification bias was estimated with global sensitivity analysis. The sensitivity of MRI ranged from 3% to 73%, and the specificity from 63% to 98%. The region comprising all possible combinations of sensitivity and specificity had a butterfly shape. The sensitivity and specificity pair estimated from cases verified with arthroscopy was included in this region. CONCLUSION: Verification bias did not greatly affect assessment of the diagnostic utility of MRI in the evaluation of cruciate ligament tears. The high specificity previously reported for MRI can be considered valid, but the sensitivity may not be as reliable.
OBJECTIVE: The purpose of this study was to investigate the extent to which verification bias affects the sensitivity and specificity of MRI in the diagnosis of cruciate ligament tears. MATERIALS AND METHODS: Consecutively registered outpatients who underwent MRI evaluation of the knee were included in the study. The sensitivity and specificity of MRI were calculated for patients whose diagnosis was verified with arthroscopy. For patients who did not undergo arthroscopy, the effect of verification bias was estimated with global sensitivity analysis, a technique of graphic representation of whether a particular combination of sensitivity and specificity estimates is compatible with the observed data. RESULTS: Among the 356 patients included in the study, 82 patients (23%) had the MRI findings verified at arthroscopy. The sensitivity and specificity of MRI among patients who underwent arthroscopy were 38% and 90%. For patients whose disease status was not verified with arthroscopy, the influence of verification bias was estimated with global sensitivity analysis. The sensitivity of MRI ranged from 3% to 73%, and the specificity from 63% to 98%. The region comprising all possible combinations of sensitivity and specificity had a butterfly shape. The sensitivity and specificity pair estimated from cases verified with arthroscopy was included in this region. CONCLUSION: Verification bias did not greatly affect assessment of the diagnostic utility of MRI in the evaluation of cruciate ligament tears. The high specificity previously reported for MRI can be considered valid, but the sensitivity may not be as reliable.
Authors: G Oldrini; P Gondim Teixeira; A Chanson; M L Erpelding; B Osemont; M Louis; A Blum Journal: Skeletal Radiol Date: 2012-09 Impact factor: 2.199