Interobserver agreement of qualitative analysis and tumor delineation of 18F-fluoromisonidazole and 3'-deoxy-3'-18F-fluorothymidine PET images in lung cancer.

As the preparation phase of a multicenter clinical trial using (18)F-fluoro-2-deoxy-d-glucose ((18)F-FDG), (18)F-fluoromisonidazole ((18)F-FMISO), and 3'-deoxy-3'-(18)F-fluorothymidine ((18)F-FLT) in non-small cell lung cancer (NSCLC) patients, we investigated whether 18 nuclear medicine centers would score tracer uptake intensity similarly and define hypoxic and proliferative volumes for 1 patient and we compared different segmentation methods. Ten (18)F-FDG, ten (18)F-FMISO, and ten (18)F-FLT PET/CT examinations were performed before and during curative-intent radiotherapy in 5 patients with NSCLC. The gold standards for uptake intensity and volume delineation were defined by experts. The between-center agreement (18 nuclear medicine departments connected with a dedicated network, SFMN-net [French Society of Nuclear Medicine]) in the scoring of uptake intensity (5-level scale, then divided into 2 levels: 0, normal; 1, abnormal) was quantified by κ-coefficients (κ). The volumes defined by different physicians were compared by overlap and κ. The uptake areas were delineated with 22 different methods of segmentation, based on fixed or adaptive thresholds of standardized uptake value (SUV). For uptake intensity, the κ values between centers were, respectively, 0.59 for (18)F-FDG, 0.43 for (18)F-FMISO, and 0.44 for (18)F-FLT using the 5-level scale; the values were 0.81 for (18)F-FDG and 0.77 for both (18)F-FMISO and (18)F-FLT using the 2-level scale. The mean overlap and mean κ between observers were 0.13 and 0.19, respectively, for (18)F-FMISO and 0.2 and 0.3, respectively, for (18)F-FLT. The segmentation methods yielded significantly different volumes for (18)F-FMISO and (18)F-FLT (P < 0.001). In comparison with physicians, the best method found was 1.5 × maximum SUV (SUVmax) of the aorta for (18)F-FMISO and 1.3 × SUVmax of the muscle for (18)F-FLT. The methods using the SUV of 1.4 and the method using 1.5 × the SUVmax of the aorta could be used for (18)F-FMISO and (18)F-FLT. Moreover, for (18)F-FLT, 2 other methods (adaptive threshold based on 1.5 or 1.6 × muscle SUVmax) could be used. The reproducibility of the visual analyses of (18)F-FMISO and (18)F-FLT PET/CT images was demonstrated using a 2-level scale across 18 centers, but the interobserver agreement was low for the (18)F-FMISO and (18)F-FLT volume measurements. Our data support the use of a fixed threshold (1.4) or an adaptive threshold using the aorta background to delineate the volume of increased (18)F-FMISO or (18)F-FLT uptake. With respect to the low tumor-on-background ratio of these tracers, we suggest the use of a fixed threshold (1.4).