TO THE EDITOR: In the preclinical arm of a coclinical trial, Savaikar et al. recently optimized 18F-FDG PET imaging biomarkers of response to a combined docetaxel and carboplatin therapy in patient-derived tumor xenografts involving triple-negative breast cancer (1). Twenty-one necrotic-core-phenotype tumors and 13 solid tumors were examined. Besides a preclinical PERCIST paradigm, 43 imaging metrics were evaluated, both in the whole tumor and in a single highest-intensity tumor slice. These metrics included SUVmean obtained from various fixed percentages of the SUVmax thresholds (SUVTh) and SUVmean obtained from the voxels involved in a sphere centered at the SUVmax voxel (SUVpeak). Spheric volumes of 4, 14, and 33 mm3 (radius of 1, 2, and 3 voxels, respectively) were considered, leading to SUVP4, SUVP14, and SUVP33, respectively. In particular, Bland–Altman plots of test–retest data allowed us to estimate an SUV25 (i.e., using a 25% of the SUVmax threshold) reproducibility percentage (R; 95% level of confidence) of about 20% and 25% for solid and necrotic tumors, respectively (from Figs. 3C and 3G, respectively, in Savaikar et al.). Finally, a coined quantitative response assessment score favored SUV25 followed by SUVP14 as optimal metrics of response to therapy in patient-derived tumor xenograft models.
We would like to stress the central role of R in assessing treatment response for any investigated SUV metrics, that is, the minimal relative change between 2 SUVs assessed from 2 successive examinations that is required to be considered a significant difference (2). In this connection, we suggest that a further SUV metric, that is, the SUVmax-V (defined as an average SUV computed from an arbitrary total hottest volume, regardless of the location of the hottest voxels included within the 18F-FDG–positive lesion), might be particularly suitable in the current context involving 21 tumors with a necrotic-core phenotype (and with varying tumor dimensions), thus exhibiting a low 18F-FDG uptake at the core and well-separated 18F-FDG–positive areas (Fig. 2 in Savaikar et al.). Indeed, it has been previously shown, in lung cancer patients, that the R of SUVmax-N, which is an average SUV computed from the N hottest voxels (N denotes the number of pooled voxels) regardless of their location within an 18F-FDG–positive lesion, was significantly lower for an N of 30 than is the R of SUVpeak obtained from SUVmax and its 26 neighboring voxels (3). In a subsequent study, SUVmax-40 was found to more likely represent the most metabolically active portions of tumors than was SUVpeak, which was obtained from the voxels involved in a 1-mL sphere centered at the SUVmax voxel, with close R performance (4). Finally, the SUVmax-N procedure for treatment-response assessment has been described in a Takayasu-arteritis patient, emphasizing that the greater the N value, the lower the SUVmax-N R and, hence, the more efficient the metrics (Table 1 in Caubet et al. (5)). Since the voxel volume may depend on the PET system, it is noteworthy that instead of SUVmax-N, one could alternatively use SUVmax-V. When comparing baseline scans with posttreatment scans, volume should be set in the scan showing the lowest total 18F-FDG–positive volume but at the greatest possible value, since the greater the volume value, the lower the SUVmax-V R.
To conclude, Savaikar et al. addressed the important issue of reaching a consensus on the reproducibility of imaging metrics for assessing response to therapy in oncology animal models (1). We suggest that the SUVmax-V metrics may have a place in this toolbox, with volume set at the greatest possible value in the scan showing the lowest tumor uptake (which is expected to be the posttreatment one). Finally, in the current series, whether R of SUVmax-V for V = 14 and 33 mm3 might be lower than R of SUV25, SUVP14, and SUVP33 remains to be assessed.
Footnotes
Published online Jan. 10, 2020.
- © 2020 by the Society of Nuclear Medicine and Molecular Imaging.