Metabolic volume performs better than SUVmax in the detection of left ventricular assist device driveline infection.

Abstract

A continuous-flow left ventricular assist device (LVAD) is a new and highly promising therapy in supporting end-stage heart failure patients, either bridging them to heart transplantation or as a destination therapy. Infection is one of the major complications associated with LVAD implants. 18F-FDG PET/CT has already been shown to be useful in the detection of LVAD infection. The goal of this study was to compare the diagnostic accuracy of different PET analysis techniques (visual grading versus SUVmax and metabolic volume). We retrospectively analyzed 48 patients with implanted LVAD who underwent an 18F-FDG PET/CT that were either suspected to have a driveline or device infection or inflammation of unknown origin. PET/CT was analyzed qualitatively (visual grading) and quantitatively (SUVmax and metabolic volume) and matched to the final clinical diagnosis concerning driveline infection. The final diagnosis (standard of reference) was made at the end of clinically recorded follow-up or transplantation and included microbiological cultures of the driveline exit site and/or surgical samples, and clinical signs of infection despite negative cultures as well as recurrence of symptoms. Sensitivity, specificity, positive and negative predictive value were 87.5%, 79%, 81% and 86% for visual score, 87.5%, 87.5%, 87.5% and 87.5% for SUVmax and 96%, 87.5%, 88.5%, 95.5% for metabolic volume, respectively. ROC analysis revealed an AUC of .929 for SUVmax and .969 for metabolic volume. Both SUVmax and metabolic volume had a high detection rate of patients with driveline infection (21/24=91.5% true positive vs. 23/26=88.5% true positive, respectively). However, metabolic volume detected more patients without any infection correctly (1/22=4.5% false negative vs. 3/24=12.5% false negative). 18F-FDG PET/CT is a valuable tool for the diagnosis of LVAD driveline infection with high diagnostic accuracy. Particularly the use of the metabolic volume yields very high accuracy and performs slightly better than SUVmax.