Prediction of Relapse after Allogeneic Stem Cell Transplantation Using Individualized Minimal Residual Markers; The Prospective Nordic Study NMDSG14B

Introduction

One third of patients with myelodysplastic syndrome (MDS) will relapse after allogeneic stem cell transplantation (SCT), with a dismal prognosis. Early detection of relapse enables pre-emptive treatment and may potentially reduce relapse risk, but is limited by the lack of sensitive markers for minimal residual disease (MRD). We developed a pipeline where patient-specific mutations, as determined by a myeloid next generation sequencing (NGS) panel are tracked using sensitive digital droplet PCR (ddPCR).

Method

We designed a prospective Nordic study (NMDSG14B; NCT02872662) enrolling all patients with MDS, mixed MDS / MPN or AML with myelodysplasia related disease and < 30% marrow blasts undergoing SCT in the Nordic region. We hypothesized that personalized MRD detection by ddPCR can predict clinical relapse earlier than conventional methods. Patients were included before SCT and serial bone marrow samples were collected before, and 1 and 3 months post SCT, and thereafter every third month for 2 years or until relapse or death. Blood samples were collected monthly. The MRD results were not available for the treating physicians. MRD positivity was defined based on the background noise of the specific ddPCR-assays and varied between 0.05-0.1% VAF.

Results

Three-hundred and sixteen patients were screened between 2016 and 2020, of which 19 were excluded due to lack of mutation or disease progression preventing SCT. We here present data of 254 patients followed ≥ 6 months after SCT. Median age was 64 (18-78) years and 59% were male. Most WHO subgroups of MDS (n=166), MDS/MPN (n=39), AML (n=8) and therapy-related disease (n=41), were represented. Risk profile according to IPSS-R was very low (n=13), low (n=32), intermediate (n=46), high (n=60) and very high (n=32). The majority of patients received pre-SCT treatment consisting of HMA (n=159) and / or intensive chemotherapy (n=59) while 60 patients did not receive disease-modifying treatment prior to SCT. The most common mutations were ASXL1 (n=69), TET2 (n=58), SRSF2 (n=57) and TP53 (n=44). No mutation was identified in 10 pts, and NGS data is still pending for 11 patients.

After a median follow-up of 436 days, estimated 2 years overall survival and relapse free survival were 72% and 63%, respectively. Cumulative incidence of NRM and relapse at 2 years was 16% and 20%, respectively. Forty-six patients relapsed after a median of 170 (53-733) days, and the estimated median survival following relapse was 197 days. The most common pre-SCT mutations in the relapsed cohort were TP53 (n=19), DNMT3A (n=11) and RUNX1 (n=9). Thirty-seven patients died due to non-relapse mortality (NRM) after a median of 83 (4-754) days.

To date, MRD results are available for 64 patients. Relapse was preceded by positive MRD in 14 out of 15 patients a median of 79 (21-173) days before clinical relapse. The 15^th patient had an extra-medullary relapse only. Borderline positive MRD samples < 0.2% VAF within 100 days after SCT followed by negative samples were seen in 11 non-relapse patients. Twenty-four of 37 patients in continuous complete remission (CCR) were consistently MRD neg. Six CCR patients had positive MRD after 100 days; two with transient borderline peaks (<0.1%) at 6 months; two with transient peaks > 0.1%, which turned negative when the patients developed GVHD; one patient with slowly decreasing MRD which turned negative first after 1 y, and finally one patient with prevailing KIT mutation (>700 days post-SCT) despite negative BCOR and STAG2. Two MRD+ patients died from NRM without showing signs of clinical relapse.

Discussion

In summary, we show that our pipeline of personalized MRD-assessment, based on patient-specific mutations is feasible with a high sensitivity to predict relapse. An update of study progression will be presented at the meeting.

Figure 1

View largeDownload slide

Figure 1

View largeDownload slide

Close modal