The interim FDG-PET scans of the treatment intensification PETAL trial were re-analyzed in a comprehensive PET analysis to segment all lymphoma manifestations. The following principal findings arise from this analysis: (1) A fully automated analysis of interim FDG-PET/CTs from lymphoma patients is feasible. (2) The biomarkers derived from the comprehensive PET analysis are statistically significant prognosticators of TTP. (3) The mean-SUVAI parameter identified patients that benefitted from additional application of rituximab as treatment intensification, which could not be achieved using conventional PET metrics.
(R-)CHOP is the standard first-line treatment for patients with aggressive lymphoma, with cure rates of 6070%. In patients with (multiply) relapsed disease, several treatment options exist, such as high-dose chemotherapy with autologous hematopoietic stem cell transplantation, allogeneic transplantation, CAR-T cell therapy, immunomodulation, and others [15,16,17]. Current methods for early prediction of treatment failure, including Deauville-based iPET assessment, appear insufficient.
FDG-PET has a long track record of monitoring initial treatment response to systemic anti-cancer therapy [2, 3, 18]. In principle, early detection of treatment failure could trigger a change in therapy, aiming at improved outcome. However, often only a single target lesion is used to assess treatment failure and guide subsequent therapies. A single lesion, however, cannot accurately capture disease extent and severity. To overcome this limitation, a recent approach tries to employ ctDNA levels as comprehensive biomarker to assess the total lymphoma burden [19]. However, in FDG-PET is comprehensively analyzed, it can also quantify the total lymphoma burden and assess the metabolic heterogeneity of all manifestations. As the delineation of all disease manifestations is too time-consuming for clinical routine, AI-based PET analysis software, like the PARS prototype and others, have been proposed [12, 20].
For the conventional metric max-SUVmanual, which takes account of a single lymphoma manifestation, no statistically significant interaction of treatment intensification by additional rituximab was found in the present analysis. In contrast, for the mean-SUVAI metric, which averages the FDG uptake of all lymphoma manifestations, a statistically significant interaction with treatment intensification was observed. This indicates that the benefit of treatment intensification through additional rituximab is growing with increasing mean-SUVAI. This was corroborated by looking at patients with high mean-SUVAI who had statistically significantly longer survival when treated with two additional rituximab doses than with 6xR-CHOP alone. Interestingly, patients with high mean-SUVAI had higher baseline SUVmax compared to patients with low mean-SUVAI (Supplementary Table5). This indicates that patients with high mean-SUVAI might erroneously be read as iPET-negative due to their high baseline SUVmax, which could lead to a more pronounced relative reduction, despite metabolically active residual lymphoma at the interim timepoint. The finding is in line with recent studies indicating that a more complex PET analysis of lymphoma patients is superior to the IPI index [21].
In patients randomized to 8x(R-)CHOP versus 2x(R-)CHOP followed by the Burkitt protocol, no statistically significant interaction of a PET parameter and treatment intensification was found. However, patients with high mean-SUVAI had statistically significantly longer TTP when they were not treated with the Burkitt protocol. This seems paradoxical as especially patients with very high residual tumor activity seemed to have a disadvantage from therapy intensification. Also, conventional PET metrics such as highest uptake or change in highest uptake between baseline and interim could identify patients who were disadvantaged by the Burkitt protocol; highlighting the need for comprehensive PET analysis. The data, however, need to be interpreted with caution because of imbalances in baseline characteristics (Supplementary Table1).
Our study has several limitations. First, it was a retrospective re-analysis of the prospective PETAL trial. The present analysis was not pre-planned, which might cause an observational bias. Additionally, all patients receiving 6xR-CHOP and 6xR-CHOP+2R were included, but only a subfraction was truly randomized (178 of 397 patients). However, non-randomized patients receiving 6xR-CHOP or 6xR-CHOP+2R were recruited using the same inclusion criteria in the beginning and at the end of the study period, respectively, which should minimize potential biases. Finally, our primary endpoint was TTP which best reflects the impact of therapy on outcome [7,8,9]. In contrast, the PETAL trial employed event-free survival (EFS), which also included death unrelated to lymphoma and events such as treatment-related toxicity.