Out of 154 results of our initial literature search, 8 studies were found to be eligible under the predetermined criteria.1,5-11 Four studies were performed on EARL-accredited systems.6,8,9,11 In total, 330 PHVs were assessed. Study characteristics are described in Table 1 and boxplot representations of the SUVmax findings for each study are shown in Figure 1.

In the seminal publication by Saby et al,1 72 patients suspected of PVE were prospectively included. Findings on FDG PET/CT were compared to the final diagnosis, defined according to the modified Duke criteria after a follow-up period of 3 months. Visual analysis was based on hypermetabolism in prosthetic and periprosthetic areas on both attenuation-corrected (AC) and non-AC (NAC) images. SUVmax was measured as the average of 3 measurements from 3 volumes of interest (VOI) of 5 mm³ at equal distances from each other. Additionally, a 35 mm³ VOI was placed in the right atrium blood pool in a location without significant metabolic activity from myocardium, and target-to-background ratio (TBR) was calculated as (SUVmax valve/SUVmax atrial blood pool). The reported sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and global accuracy for visual assessment were 73%, 80%, 85%, 67%, and 76%, respectively. By adding FDG PET/CT to the modified Duke criteria, sensitivity increased to 97%.

Bartoletti et al5 only included patients with proven PVE in their case series of 6 patients. They did not describe the measurement method for the reported values. They found a large variation in SUVmax, with several relatively low values (median 4.2, range 3.2-10.0). In patients with the lowest values, antibiotic therapy had been started before FDG PET/CT with resolution of fever and other symptoms of infection.

Rouzet et al6 included 39 patients with a total of 45 PHVs in their study on FDG PET/CT and radiolabelled leukocyte scintigraphy in PVE. Visual analysis and SUV measurement were performed based on the same methods as reported by Saby et al and final diagnosis was also based on modified Duke criteria after 3 months follow-up. The reported sensitivity, specificity, PPV, NPV, and global accuracy for visual assessment were 93%, 71%, 68%, 94%, and 80% respectively. SUVs for patients classified as ‘rejected’ by the modified Duke criteria were relatively high (median 4.9, range 3.3-6.2), but patients with no visually discernible uptake of FDG in the region of the prosthetic valve were excluded from the semi-quantitative analysis. Since it is reasonable to assume that these would have represented lower SUVmax values, the reported SUVmax values will therefore most likely be skewed to the higher end of the spectrum.

The study by Pizzi et al7 included both patients with PHVs and cardiac implantable electronic devices (CIEDs). The reported sensitivity, specificity, PPV, and NPV for the total group were 84%, 75%, 81%, and 78%, respectively when compared to the modified Duke criteria at 3 months. For the purpose of this review only the values of the patients with PHVs were included. Visual analysis was comparable to the methods outlined by Saby et al, but SUV analysis differed in that SUVmax was measured at any abnormal area, and blood pool measurement was based on the mean standardized uptake value (SUVmean) as obtained with a 30 mm³ VOI at the thoracic descending aorta. TBR was calculated as (SUVmax prosthesis/SUVmean blood pool). The region of interest in scans without visually detectable uptake near the prosthetic valve was placed around the metallic components of the valve alone, without inclusion of the adjacent tissues and blood pool, which most likely resulted in a lower SUVmax and may explain the exceptionally low values found in the ‘rejected’ category (median 0.5).

Fagman et al8 included 11 patients scanned for suspected PVE. Additionally, 19 normal controls were added in the form of patients with prosthetic heart valves scanned for malignancy. Visual analysis, comparable to the methods described by Saby et al, resulted in a sensitivity of 75% and specificity of 84%, based on patients with definite (N = 8) or without (N = 19) PVE. Semi-quantitative analysis was performed measuring SUVmax in or directly adjacent to the prosthetic aortic valve. Blood pool values were determined by measuring SUVmax in five circular ROIs on consecutive slices in the lumen of the descending aorta at the level of carina, avoiding inclusion of potential uptake in the wall of the aorta. TBR was calculated as (prosthetic valve SUVmax/SUVmax descending aorta).

In their study, Salomäki et al9 included both native valves and PHVs suspected of endocarditis, confirming that FDG PET/CT is less capable of diagnosing native valve endocarditis (only 1 out of 6 cases detected). For the purpose of this review only the data of the 16 patients with PHVs were considered. Visual analysis was performed as described by Saby et al, resulting in a reported sensitivity of 100% (6/6 cases) but a specificity of only 60%. SUVmax was measured in a VOI covering the valve or prosthesis area based on co-registered CT images. The mean blood pool values were measured in the ascending aorta excluding the vessel wall (mean radioactivity in a VOI of 6.8 cm³) to calculate TBR. Two noticeably high values (7.8 and 7.2) skewed the overall findings in the ’rejected’ category (N = 5) upwards (median 4.8, range 2.9-7.8), with one value reported as being this high due to a foreign body reaction and one due to imaging relatively early after implantation (6 weeks).

Jiménez-Ballvé et al10 compared different interpretation criteria in their cohort consisting of patients with PHVs and/or CIEDs compared to the Duke pathological criteria if tissue was available or the decision of an endocarditis expert team after a minimum of 4 months follow-up. Using criteria comparable to those used by Saby et al in the whole group, visual analysis resulted in reported sensitivity, specificity, PPV, NPV, and global accuracy of 100%, 73%, 80%, 100%, and 87%, respectively. Again, for this review only the data in PHV patients compared to the final diagnosis were included. SUVmax was measured in the area under suspicion and was compared to physiological uptake (also measured as SUVmax) in the mediastinal blood pool, calculated by measuring a VOI with 3 mm diameter in the ascending aorta, and in the liver, calculated by measuring a VOI with 3 cm diameter drawn in the right hepatic lobe excluding any areas of inhomogeneous or focally increased uptake. As in the study by Rouzet et al, patients with no visually discernible uptake of FDG in the region of the prosthetic valve were excluded from the semi-quantitative analysis, probably skewing reported SUVmax values to the higher end of the spectrum.

Mathieu et al11 recently studied 51 patients with 54 PHVs with no suspicion of PVE to define normal variants and values of FDG uptake. Indications for PET were oncology (N = 26), suspicion of prosthetic valve endocarditis subsequently excluded (N = 17), and history of vasculitis (N = 11). Visual analysis was descriptive, with FDG uptake described as absent, homogeneous, or focal in 13%, 80%, and 7% of AC images and 44%, 50%, and 6% of NAC images, respectively. SUV measurements were performed according to the protocol as described by Saby et al and Rouzet et al, resulting in a median SUVmax of 3.5 (range 2.1-8.0). In subgroup analysis, values were higher in patients referred with a history of vasculitis (median 4.7, range 3.0-8.0) than in the other patients referred for oncologic indications (median 3.3, range 2.1-5.7) or rejected PVE (median 3.5, range 2.1-4.7), even though metabolic activity in the wall of the ascending aorta did not differ significantly between groups.