Acrobat interim results: Alloheme surveillance enables early detection of relapse in post-HCT AML and MDS patients

Abstract Introduction: Relapse is the leading cause of mortality after allogeneic hematopoietic cell transplantation (HCT) for acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS). Current peripheral blood-based standard-of-care surveillance methods, such as CD3 and CD33 lineage-specific chimerism by short tandem repeat–PCR (STR-PCR), are routinely used for engraftment monitoring. However, they have limited sensitivity and lack standardization, which hinders their effectiveness in detecting relapse. The ACROBAT study (NCT04635384) evaluates whether serial assessment of peripheral blood with AlloHeme can enable earlier and more precise detection of relapse. Methods: Adult patients with AML or MDS receiving their first allo-HCT across 11 US sites were prospectively enrolled in the ACROBAT study. All participants underwent periodic assessments using the AlloHeme NGS-based ultra-sensitive chimerism test (CareDx, Brisbane, CA) on CD33+ (myeloid) and CD3+ (T-cell) cell subtypes and whole blood at 14 post-HCT time-points over 2 years: weeks 4, 6, 8, 10, 12, and months 4, 5, 6, 9, 12, 15, 18, 21, and 24. An increase of ≥0.2% in recipient chimerism between two consecutive time points was defined as increasing mixed chimerism (iMC). Complete chimerism (CC) was defined as a donor chimerism of ≥99.8%. Relapse outcomes were determined per the Center for International Blood & Marrow Transplant Research (CIBMTR) criteria by the treating physicians at each site. Here we present the findings of the 18-month interim analysis. Results: A total of 227 patients were enrolled, including 162 with AML and 65 with MDS. At the data cut-off date, the median follow-up was 23 months (IQR: 10 – 24). During this period, 51 subjects (22.5%) experienced relapse, and 45 (19.8%) subjects died due to non-relapse mortality (NRM). A time-varying risk model (TVM) with CD33 iMC as the time-varying variable and NRM as the competing risk for relapse demonstrated a hazard ratio (HR) of 54.0 for relapse (95%CI = 22.4, 142.8; p < 0.001). CD33 iMC by AlloHeme demonstrated a high sensitivity (87.5%) and a high specificity (85.2%); with 22.5% prevalence of relapse, it resulted in a positive predictive value (PPV) of 59.6%, and a negative predictive value of 96.5% for relapse prediction. We identified some subjects who were transiently CD33 iMC positive but later achieved CC status. If we consider the transient positive subjects as negative, then the specificity and PPV increased to 95.8% and 84.8% respectively. Among subjects who were CD33 iMC positive before or at the time of relapse event, the test provided a median lead time of 36 days to clinical relapse (IQR: 18–72 days), with 47% relapses identified at least 45 days before relapse. Conclusion:Serial CD33-iMC monitoring post-allo HCT identified relapses in AML and MDS patients with high sensitivity and specificity and with a median lead time of 36 days. These interim results of the ACROBAT study suggest that incorporating AlloHeme surveillance in routine post-HCT monitoring of AML and MDS patients could facilitate the early detection of relapse.