Combined Visual and Semiquantitative Evaluation Improves Outcome Prediction by Early Midtreatment 18F-FDG PET in Diffuse Large B-Cell Lymphoma.

ABSTRACT

The purpose of this study was to assess the predictive and prognostic value of interim FDG PET (iPET) in evaluating early response to immunochemotherapy after 2 cycles (PET-2) in diffuse large B-cell lymphoma (DLBCL) by applying 2 different methods of interpretation: the Deauville visual 5-point scale (5-PS) and a change in SUV (ΔSUV) by semiquantitative evaluation. Methods: In total, 145 patients with newly diagnosed DLBCL underwent pretreatment PET and PET-2 assessment. PET-2 was classified according to both 5-PS and percentage ΔSUV. Receiver-operating-characteristic analysis was performed to compare the accuracy of the 2 methods for predicting progression-free survival. Survival estimates, based on each method separately and combined, were calculated for iPET-positive (iPET+) and iPET-negative (iPET-) groups and compared. Results: Both with 5-PS and with ΔSUV-based evaluations, significant differences were found between the progression-free survival of iPET- and iPET+ patient groups (P < 0.001). Visually, the best negative predictive value (NPV) and positive predictive value (PPV) occurred when iPET was defined as positive if the Deauville score was 4-5 (89% and 59%, respectively). Using the 66% ΔSUV cutoff reported previously, NPV and PPV were 80% and 76%, respectively. ΔSUV at the 48.9% cutoff, reported for the first time here, produced 100% specificity along with the highest sensitivity (24%). The 5-PS and a semiquantitative ΔSUV of less than 48.9% for each PET-2 gave the same PET-2 classification (positive or negative) in 70% (102/145) of all patients. This combined classification delivered NPV and PPV of 89% and 100%, respectively, and all iPET+ patients failed to achieve or remain in remission. Conclusion: In this large consistently treated and assessed series of DLBCL patients, iPET had good prognostic value interpreted either visually or semiquantitatively. We determined that the most effective ΔSUV cutoff was 48.9% and that when combined with 5-PS assessment, a positive PET-2 result was highly predictive of treatment failure.