Abstract
Literature data on the diagnostic performance of phlebography, myelography, and CT
scan applied to patients with suspected lumbar disk herniation (LDH) are analyzed
to extract maximal information about their relative discriminatory power. Seventeen
papers meeting the selection criteria contain 13 reports on myelography, 6 on phlebography,
and 5 on CT. Sensitivity and specificity are considered simultaneously in logistic
ROC space. The reports of each procedure are effectively summarized by a linear regression
in logistic ROC space. Taking into account the individual confidence regions of sensitivity
and specificity obtained from each report, the slope of the regression line is estimated
by Generalized Least Squares (ML). This approach also allows to test the assumption
of a common odds ratio (i.e., of a unit slope). The simply to determine common odds
ratio as well as the perpendicular distance between the origin and the regression
line (as a good approximation to the area under the ROC curve) are used as a measure
for the discriminatory power of the procedures. For CT, homogeneity of sensitivity
turns out to be much more likely than a common odds ratio. Based on the available,
retrospective data, phlebography appears to have the highest performance in visualizing an LDH, followed by myelography and CT.
Key-Words
Lumbar Disk Herniation - Myelography - Phlebography - CT - ROC analysis - Diagnostic
Test Evaluation
