Assessment of the diagnostic accuracy of 18F-FDG PET/CT in prosthetic infective endocarditis and cardiac implantable electronic device infection: comparison of different interpretation criteria

Purpose The diagnosis of prosthetic valve (PV) infective endocarditis (IE) and infection of cardiac implantable electronic devices (CIEDs) remains challenging. The aim of this study was to assess the usefulness of 18 F-FDG PET/CT in these patients and analyse the interpretation criteria. Methods We included 41 patients suspected of having IE by the Duke criteria who underwent 18 F-FDG PET/CT. The criteria applied for classifying the findings as positive/negative for IE were: (a) visual analysis of only PET images with attenuation-correction (AC PET images); (b) visual analysis of both AC PET images and PET images without AC (NAC PET images); (c) qualitative analysis of NAC PET images; and (d) semiquantitative analysis of AC PET images. 18 F-FDG PET/CT was considered positive for IE independently of the intensity and distribution of FDG uptake. The gold standard was the Duke pathological criteria (if tissue was available) or the decision of an endocarditis expert team after a minimum 4 months follow-up. Results We studied 62 areas with suspicion of IE, 28 areas (45 %) showing definite IE and 34 (55 %) showing possible IE. Visual analysis of only AC PET images showed poor diagnostic accuracy (sensitivity 20 %, specificity 57 %). Visual analysis of both AC PET and NAC PET images showed excellent sensitivity (100 %) and intermediate specificity (73 %), focal uptake being more frequently associated with IE. The accuracy of qualitative analysis of NAC PET images depended on the threshold: the maximum sensitivity, specificity and accuracy achieved were 88 %, 80 %, 84 %, respectively. In the semiquantitative analysis of AC PET images, SUVmax was higher in areas of confirmed IE than in those without IE (∆SUVmax 2.2, p