Page 139 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

PET improves DLBCL response predictors
I-PET after 2 cycles), who analyzed different cutoffs for DS after 2 cycles, reported PPVs for DS4-5 and ∆SUVmax that were remarkably identical to our study (39% vs. 38% and 52% vs. 53%, respectively) [22]. A DLBCL subgroup analysis of the PETAL study also reports a more favorable PPV for ∆SUVmax I-PET assessment than for Deauville assessment [23].
Baseline clinical characteristics and chemoimmunotherapy sensitivity are both relevant factors in outcome prediction. This relevancy was demonstrated in our multivariable analysis, in which aaIPI and ∆SUVmax (reflecting chemosensitivity) were both independent predictors of 2-y PFS. Again, the subgroup with both high-intermediate or high aaIPI and a ∆SUVmax of 70% or less had a PFS clearly below 50% but was relatively small (6% of all patients). Selection of a poor-risk group of only 6% is justified both from a cost awareness perspective and because it is the group most likely not be cured by standard treatment. These patients can be treated within clinical trials investigating the efficacy of new drugs.
Several relatively small retrospective studies reported inconsistent results regarding associations of clinical characteristics and I-PET results (DS or ∆SUVmax) with survival in multivariable Cox models [7,22,24].Two prospective studies concluded that only I-PET and not IPI was independently associated with event-free survival [25,26]. The randomized phase III trials PETAL (I-PET after 2 cycles of R-CHOP21) and CALGB-50303 (I-PET after 2 cycles R-CHOP21 or DA- EPOCH-R [dose-adjusted etoposide, prednisone, vincristine, cyclophosphamide, doxorubicin, and rituximab]) also concluded that I-PET with ∆SUVmax (cutoff 66%) and IPI were independent predictors for event-free survival and PFS [11,27], respectively.
Baseline MTV assessment was not a strong predictor of 2-y PFS in our study (Table 2; Supplemental Table 1,3,5, and 7). We used a segmentation method applying a fixed SUV of at least 4.0, on the basis of a recent study showing that this method performed best and had a discriminative power similar to that of other segmentation methods [16]. Addition of dichotomized baseline MTV (345cm3 cutoff ) to ∆SUVmax did not improve the predictive value, but log-transformed continuous MTV added some independent predictive value when combined with ∆SUVmax. In a secondary analysis of the PETAL randomized clinical trial (DLBCL subset, I-PET after 2 cycles, same MTV software and methodology as
137
6a