18-Fluorine Fluorodeoxyglucose Positron Emission Tomography for the Diagnosis of Infection in the Postoperative Spine

Publisher Website
Google Scholar
Abstract
Background. Using conventional imaging methods, including magnetic resonance imaging and labeled leukocyte scanning, the diagnosis of infection in the postoperative spine remains a diagnostic challenge. Recently, promising results have been reported using 18F-fluorodeoxyglucose positron emission tomography for various infectious problems. This study aimed to investigate the value of 18F-fluorodeoxyglucose positron emission tomography in patients suspected of having spinal infection after previous surgery of the spine. Methods. Fifty-seven consecutive patients with a history of previous spinal surgery were prospectively included between February 1999 and June 2001. 18F-fluorodeoxyglucose positron emission tomography was performed 60 to 90 minutes after injection of 370 MBq 18F-fluorodeoxyglucose. Images were scored visually and semiquantitatively by two blinded, independent, certified nuclear medicine physicians, experienced with positron emission tomography. Differences were assessed by consensus. Results were correlated with final diagnosis allowing calculation of sensitivity, specificity and accuracy. Receiver operating characteristic analysis was performed to find optimal cut-off values. Results. Fifteen patients had spinal infection. Using the most sensitive cut-off values sensitivity, specificity and accuracy were 100%, 81%, and 86%, respectively, for both visual and semiquantitative scoring. In the group without metallic implants (n = 27), false positives (n = 2) only occurred in the first 6 months after surgery. In the group with metallic implants (n = 30), false positives (n = 6) were not confined to recently operated patients. Conclusions. Overall accuracy was excellent (86%) with a negative predictive value of 100%. 18F-fluorodeoxyglucose positron emission tomography holds promise to become the standard imaging technique in this difficult patient population, as it is straightforward, provides a rapid result (2 hours) and because accurate alternatives are lacking.

This publication has 25 references indexed in Scilit: