TO THE EDITOR: In a series of 95 large-vessel vasculitis patients investigated with 18F-FDG PET imaging, Dashora et al. recently tested the performance of qualitative (PET vascular activity score [PETVAS]) and semiquantitative (SUV and tissue-to-background ratio [TBR] relative to liver and blood activity) scoring methods (1). Regarding the latter methods, 9 territories were created in each patient by segmenting the aorta and branch arteries. A territory score was calculated by averaging the SUVmax assessed in each axial region of interest that was manually drawn across the territory, and a global summary, SUVArtery, was then calculated by averaging all territory scores. Liver TBR (TBRLiver) and blood TBR (TBRBlood) were computed by dividing SUVArtery by a mean liver and blood SUV, respectively. The performance of each metric was assessed in association with reader interpretation of vascular PET activity and with physician assessment of clinical disease activity, including the area under the receiver-operating-characteristic curve. Tables 2 and 3 by Dashora reported the metrics performance against the 2 reference standards; this performance was poor–poor for SUVArtery (area under receiver-operating-characteristic curve, 0.67–0.59) and good–poor for TBRLiver and PETVAS (areas under receiver-operating-characteristic curve, 0.85–0.66 and 0.87–0.65, respectively) (1). TBRBlood had slightly lower performance than TBRLiver.
Since TBRLiver involves SUVArtery, which results from SUVmax averaging, we suggest that instead of using SUVArtery, we use an averaged SUVmax obtained from N hottest voxels (SUVmax-N) irrespective of their location within the 9 vascular territories (2). Both SUVArtery and SUVmax-N take into consideration the heterogeneity of the vessel-wall uptake, but N can actually be much greater than the total number of regions of interest used by Dashora et al. for calculating SUVArtery. Since the greater the N number, the lower the SUVmax-N variability, a more reliable TBRLiver can thus be provided than with SUVArtery (2,3). A previous assessment of treatment response in a Takayasu arteritis patient illustrates the possible magnitude of N, with SUVmax-N pooling N = 4,100 and 515 voxels, corresponding to a hottest volume V = 100 and 12.6 mL, respectively (4). SUVmax-V might be preferred to SUVmax-N, for the voxel volume depends on the PET system at a given center. For assessing response to treatment in a large-vessel vasculitis patient, it has been previously shown that V (or N) should be set in the scan showing the lowest total 18F-FDG–positive volume, which is expected to be posttreatment one (4). For assessing the severity of large-vessel vasculitis inflammation as in the study of Dashora et al., we suggest that standard SUVmax-V–based TBRLiver metrics might be relevant, using an arbitrary value of V defined by expert consensus (e.g., of 10 cm3). Additionally, we suggest that the hottest volume V corresponding to a standard value of SUVmax-V–based TBRLiver could also be investigated by Dashora et al. as a further metric. This TBRLiver standard value should be greater than 1, as is consistent with the qualitative territory score of 3 used in PETVAS (arterial uptake > liver uptake). The standard might be set at 1.33 according to TBRLiver data reported in Table 3 by Dashora et al. for physician assessment of clinical disease activity, that is, between the clinical-active range and the clinical-remission range (1.33 = 1.27 + 1.96 × 0.03 ≈ 1.46–1.96 × 0.06) (1). A similar line of argument provides a TBRBlood standard value of 2.43 (from Table 3 of Dashora et al. (1)).
To conclude, we fully agree with the authors that qualitative metrics for assessing large-vessel vasculitis inflammation severity with 18F-FDG PET, such as PETVAS, are attractive in clinical practice because of ease of implementation and ease of interpretation. However, we believe that SUVmax-V–based TBRLiver (or SUVmax-V–based TBRBlood) could also be used daily if manufacturers are encouraged to make SUVmax-V (or SUVmax-N) easier to assess than currently (2–4).
Footnotes
Published online Sep. 2, 2021.
- © 2022 by the Society of Nuclear Medicine and Molecular Imaging.
- Received for publication November 1, 2021.
- Revision received August 16, 2021.
- Accepted for publication September 2, 2021.