Comparaison de différents indices métaboliques et de la méthode SULTAN dans l’évaluation de la réponse thérapeutique par TEP au 18 FDG dans le cancer du sein métastatique

Objective. - 18F-FDG PET is a valuable tool in the evaluation of therapeutic response in breast cancer. This retrospective study was designed to compare the performance-of six metabolic indices and to define their optimal thresholds, in patients treated with chemotherapy or hormone therapy for metastatic breast cancer. The performances of a parametric analysis by SULTAN method were also evaluated.Methods. - Twenty patients, who underwent from 2 to 7 PET during the follow-up were analyzed. For each target, six indices were measured: SUVmax (maximum Standardized Uptake Value), SUVpeak, SAM (Standardized Added Metabolic activity), metabolic volume (MV), SUVmean using an adaptive threshold, and TLG (total lesion glycolysis). The percentage change of each target between each PET was calculated. A method based on parametric imaging (SULTAN) was also applied to each patient. The results were compared to the gold standard, defined by clinical evaluation, biological and morphological imaging RECIST 1.1 criteria. A per-lesion and per-patient analysis were performed and the optimal thresholds for each indices were calculated.Results. - For the per-lesion analysis, 61 targets and 111 evolutions with 67 responders (R) and 44 non-responders (NR) were studied. Using ROC curve analysis and intercomparison, SUVmax, SUVpeak and SUVmean were significantly better than SAM, TLG and VM (P < 0.05). Using the optimal thresholds of -21%, -21%, -34%, -48% and -23% for SUVmax, SUVpeak, SUVmean, SAM and TLG respectively, these five indices were significantly correlated with the gold standard. SUVmax, SUVpeak and SUVmean showed the best performances of sensitivity (88%, 87% and 78% respectively), specificity (93%, 93% and 98% respectively) and negative predictive value (NPV) (84%, 69% and 74% respectively). For the per-patient analysis, 42 pairs of PET with 22 R and 20 NR were studied. Only SUVmax and SUVpeak were correlated to gold standard with the 30%-PERCIST-threshold and with optimal thresholds with performances of sensitivity of 73% and 77%, specificity of 95% and NPV of 76% and 79%. Parametric analysis with SULTAN showed excellent performances in the per-lesion and per-patient analysis (sensitivity 84% and 82%, specificity 98% and 90%, NPV 80% and 82%, respectively).Conclusion. - SUVmax and SUVpeak appeared the best indices to evaluate metabolic response in metastatic breast cancer. The SULTAN method was a reliable method to assist interpretation. (C) 2014 Elsevier Masson SAS. All rights reserved.