Search for a command to run...
Large B-cell lymphomas (LBCL) display variable outcomes to treatment related to the underlying heterogeneity of the disease. While transformative survival outcomes are observed in relapsed or refractory (r/r) LBCL with therapies such as anti-CD19 Chimeric Antigen Receptor (CAR) T therapy,1 disease outcomes still vary. In ZUMA-7, a phase-3 randomized trial (NCT03391466) evaluating treatment of r/r LBLC with the anti-CD19 CAR T-cell axicabtagene ciloleucel (axicel) compared to historical standard of care (SOC) rituximab-based chemotherapy followed by transplant in responders, most patients responded profoundly well to axicel, but many eventually experienced disease progression.2 Imaging Positron Emission Tomography/Computed Tomography (PET/CT) is the gold standard to assess response post-treatment following the Lugano 2014 criteria; however, a large proportion of patients experience disease progression after a complete remission (PET-CR) PET/CT assessment.3 Genomic profiling of circulating tumour DNA (ctDNA) to detect minimal residual disease (MRD) is increasingly utilized for response assessment and risk stratification in lymphoma indications4, 5 and importantly has been incorporated into LBCL National Comprehensive Cancer Network (NCCN) clinical care guidelines.6 The clonoSEQ assay7 assesses MRD by determining dominant B-cell clones within the tumour and subsequently tracking these clones in the periphery to monitor disease levels over time. Here, we used clonoSEQ to retrospectively analyse 65 axicel and 48 historical SOC patients (characteristics Table S1, schema Figure S1) from ZUMA-7 with sufficient pretreatment tumour biopsies and at least one post-treatment plasma sample (see details of sample collections and data analyses in Methods S1). This subset had higher event-free survival (EFS) compared to the overall study,2 with median EFS of 30.2 months for axicel and 5.2 months for historical SOC patients (Figure 1A). Overall, we successfully established a malignant clone ID in 80% (117/147) of tumour samples for tracking and 79% (412/521) of plasma samples calibrated with a matched ID for longitudinal assessment. Notably, 89% of patients receiving axicel achieved MRD negativity within 12 months post-randomization, compared to 69% in the historical SOC group (Figure 1B). Across all patients, EFS was significantly improved in those who achieved MRD negativity at any point, herein referred to as the ‘best MRD response’ analogous to the best overall response derived from PET/CT imaging, compared to those who never achieved MRD negativity (Figure 1B). Patients treated with axicel showed a higher rate of ever achieving MRD negativity than the historical SOC group and a best MRD response of MRD negative was strongly associated with improved responses. The rate of MRD negativity increased over time and stabilized by day 150 post-randomization for the axicel arm, indicative of continued disease clearance prior to this time point (Figure S2A). MRD negativity had a high rate of concordance with PET-CR for both treatments overall, but the concordance was lower for axicel than for historical SOC at day 50 post-randomization (Figure 1C). These findings suggest potential challenges in assessing early efficacy by MRD in lymphoma within the first month following CAR T treatment if prolonged CAR-mediated tumour cell lysis results in ctDNA release and detection despite radiological disease clearance. We next assessed the performance of the last available MRD result up to month 12 in stratifying long-term patient responses using the last PET/CT result as a benchmark comparison to calibrate as a definitive disease measure. Both detection methods show an improved durable benefit of axicel over historical SOC, although PET/CT was more robust in stratifying EFS (Figure S2B). Landmark assessments such as PET-CR by end of treatment (EOT) can be predictive of response duration, but up to 20% of patients with LBCL still experience disease relapse after a first CR.8 MRD may provide suitable sensitivity and specificity to complement imaging9 and MRD has been shown to better stratify at EOT than PET-CR while further identifying relapses in those achieving a PET-CR.3 As EOT is not defined for CAR T therapy, we explored the association of the day 150 post-randomization MRD time point as a predictor of treatment outcomes, as this was a scheduled collection capturing approximately 1 month after EOT for historical SOC and approximately 2–3 months after axicel infusion. We hypothesized that day 150 post-randomization could be a suitable landmark as earlier time points might capture active molecular disease clearance seen in our above concordance analysis, and day 56 and day 90 post-infusion have previously been shown to have the lowest rates of MRD positivity in patients with durable responses to axicel.10 Furthermore, later time points might be subsequent to many progression events. Accordingly, we determined that MRD-positive patients at day 150 have lower EFS than MRD-negative patients, albeit not statistically significant given the low number of MRD-positive patients (Figure 1D). Similarly, patients achieving a PET-CR on day 150 did not have significantly increased EFS compared to non-CR patients (Figure 1E). While neither MRD nor PET/CT showed a significant ability to stratify patient response at day 150 post-randomization, both measures showed significant patient stratification at each time point up to month 12, although PET-CR showed a higher degree of significance compared to MRD across all time points (Figure S3A,B). Of patients achieving PET-CR on day 150, those who were concurrently MRD-positive had worse EFS than MRD-negative patients (Figure 1F), indicating that combined imaging and MRD could provide the greatest response prediction. Furthermore, of the three MRD-positive patients in CR, one patient remained in CR at 1 year, indicating active disease clearance at day 150, whereas two patients relapsed shortly after. Patients who received axicel and achieved both PET-CR and MRD negativity trended towards improved EFS compared to respective historical SOC patients (Figure 1G). The sensitivity of the assay to detect MRD positivity in non-CR patients was low at each time point, and the predictive values fluctuated, indicating increased sensitivity might be required to detect MRD positivity in non-CR patients (Figure S2C). While clonoSEQ was previously shown to have predictive value at pretreatment and 1-month post-axicel treatment in third-line or later LBCL,10 others have shown that some patients may not have MRD detected prior to treatment4, 5 or around the time of clinical relapse.11, 12 Here, 80% of tumour samples passed QC with malignant clones detected, and of those, clonoSEQ MRD did not outperform PET/CT imaging. One potential limitation here is the relatively low number of subjects available for analysis, as MRD was a post hoc assessment in ZUMA-7 using available material with expected missingness across time points.13 Meta-analysis across future studies may thus better support these findings. In our analysis, clonoSEQ did not perform better than PET/CT imaging and was inferior in stratifying patient responses at any time point tested. One explanation for this may be attributed to the overall lower frequencies of trackable clones observed in LBCL, in contrast to other haem malignancies such as multiple myeloma or leukaemias with more circulating disease.14 Due to the single-target approach of clonoSEQ, the sensitivity in indications with lower ctDNA may be reduced compared to assays that profile a diverse set of mutations and show improved detection rates at low ctDNA fractions.15 Thus, given the low detection rates and lack of improvement over imaging in this study, an assay with improved sensitivity might be required to monitor MRD in LBCL. Overall, the utility of MRD in lymphoma is rapidly progressing towards standardization and improved sensitivity with the anticipation for adoption in routine clinical monitoring. All authors conceived and designed the study; D.M. and S.V. performed the statistical analysis; B.M., D.M., S.V. and R.S. designed, performed and interpreted the data analysis; and all authors participated in the manuscript writing and review and provided final approval of the manuscript. We thank the patients who participated in this trial and their families, caregivers and friends; the trial investigators; coordinators; and healthcare staff at each site. This study was funded by Kite, a Gilead Company. All authors report employment and stock ownership in Kite Pharma, a Gilead Company. The ethical committees of participating institutions approved this study. All patients gave written informed consent according to the Helsinki declaration. NCT03391466. Kite is committed to sharing clinical trial data with external medical experts and scientific researchers in the interest of advancing public health. Kite shares anonymized individual patient data for studies of Kite or Gilead compounds approved in the United States and the EU with a marketing authorization date on or after 1 January 2014 upon request or as required by law or regulation with qualified external researchers based on submitted curriculum vitae and reflecting non-conflict of interest. The request proposal must also include a statistician. Approval of such requests is at Kite's discretion and is dependent on the nature of the request, the merit of the research proposed, the availability of the data and the intended use of the data. Data requests should be sent to [email protected] and will be addressed within 60 days. Consideration for data requests will commence at the time of manuscript publication. Data S1: Figure S1. Figure S2. Figure S3. Table S1. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.