Page 131 - 18F-FDG PET as biomarker in aggressive lymphoma; technical and clinical validation

PET improves DLBCL response predictors
MTV values had a nonnormal distribution and were log-transformed using the natural logarithm. We used both the continuous and the dichotomized MTV with a prespecified cutoff adopted from the PETAL study to identify a high- MTV (>345 cm3) and a low-MTV group (MTV ≤345 cm3) [8].
I-PET4 scans were centrally reviewed by 2 independent reviewers from a pool of 10 reviewers [13] according to DS criteria [9,17]. Discrepancies were resolved by adjudication. DS4-5 was categorized as no complete metabolic response (PET- positive), and DS1-3 was categorized as complete metabolic response (PET- negative) [9,17]. DS4 was assigned when tumor SUVmax exceeded hepatic SUVmax by fewer than 3 times, and DS5 was assigned when there were new lymphoma lesions or when tumor SUVmax was 3 or more times hepatic SUVmax [9]. The accuracy of other DS cutoffs (i.e. 1 vs. 2-5, 1-2 vs. 3-5, and 1-4 vs. 5) for I-PET4 were evaluated in sensitivity analyses.
In patients with a baseline PET scan and an I-PET4 scan with DS2-5, we measured the change in SUVmax between baseline and I-PET4 (∆SUVmax). For DS1, ∆SUVmax was set at 100% reduction [9]. We applied a prespecified ∆SUVmax cutoff of 70% reduction between baseline and I-PET4 to define a positive (≤70%) or negative (>70%) I-PET result [10].
Statistical analysis
The primary outcome measure was 2-y PFS, defined as time from randomization to disease progression, relapse, or death from any cause within 2 y [18]. Survival curves were obtained with Kaplan-Meier analyses for PFS stratified by dichotomized PET response criteria and compared with log-rank tests. We used univariate and multivariable Cox proportional hazards regression models to assess the effects of baseline clinical factors (aaIPI, age, B-symptoms, MTV, sex, treatment arm) and I-PET4 response criteria (DS, ∆SUVmax) on 2-y PFS. A backward Wald elimination procedure was used to test which prognostic factors were independently associated with 2-y PFS. In addition, 2x2 contingency tables were constructed to calculate diagnostic measures (i.e. sensitivity, specificity, positive predictive value [PPV], and negative predictive value [NPV]) to predict 2-y PFS. Sensitivity, specificity, predictive values, univariate hazard ratio (HR) and receiver operating-characteristic curve were used to define the optimal I-PET4 response criteria to predict 2-y PFS. We examined whether the addition of
129
6a