Page 138 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

Chapter 6a
elimination, yielding log-transformed MTV, an age of more than 60 y, B symptoms, and ∆SUVmax as factors independently associated factors with 2-y PFS (Supplemental Table 1)
Overall survival analyses
The results of the response criteria and uni- and multivariable analyses for 2-y overall survival are presented in Supplemental Tables 5-7 and Supplemental Figure 4.
Discussion
In this multicenter study, DLBCL I-PET after 4 cycles of R(R)-CHOP14 added predictive value to baseline clinical characteristics (aaIPI) for 2-y PFS, with high NPVs (82%-86%) independent of all I-PET response criteria. However, the PPV was still relatively low. Combining clinical and PET data showed that aaIPI and ∆SUVmax were independently associated with 2-y PFS, with HRs of 3.2 and 5.0, respectively. Adding log-transformed baseline MTV only slightly improved the predictive value combined with the ∆SUVmax response criteria. As a secondary objective, we compared the most commonly used visual and semiquantitative criteria and externally validated that ∆SUVmax criteria were the optimal I-PET4 criteria to predict 2-y PFS, with a HR of 4.8 and a PPV of 53%.
On the basis of the PPV and univariate HR in I-PET, the DS5 cutoff performed best, with a PFS clearly below 50% for the DS5 group. However, the percentage of DS5-positive patients was low (5%), but this group could be of interest for future new therapy strategies. The univariate HR for 2-y PFS with DS4-5 cutoff in I-PET4 was 3.1 (95%CI, 2.1-4.5), which is similar to the pooled HR of 3.1 (95%CI, 2.5-3.9) in a systematic review, even though in that review I-PET was performed after 1-4 cycles of treatment and less strict I-PET response criteria were applied [19]. The NPV for 2-y PFS in our study was 85%, which is in line with these previous studies generally reporting NPVs above 80% (range, 64%– 95%, [19]).
Two recent retrospective DLBCL studies analyzed the value of I-PET after 4 cycles [20,21], and both concluded that ∆SUVmax had a higher accuracy and PPV than DS in predicting PFS. The retrospective study from Itti et al. (n=114,
136