Clinical Relevance of Minimal Residual Disease Monitoring in NPM1 Mutated AML: A Study of the AML Study Group (AMLSG)

Background: Recent studies revealed that Nucleophosmin1 mutated (NPM1^mut) based minimal residual disease (MRD) monitoring allows identification of patients (pts) with acute myeloid leukemia (AML) at high risk of relapse. NPM1^mut MRD monitoring remains of clinical relevance even in the context of concurrent gene mutations, e.g . FLT3 internal tandem duplications (FLT3- ITD^pos) and DNA methyltransferase 3 (DNMT3A^mut). However, data are varying with regard to the type of material that was used (bone marrow, BM; peripheral blood, PB) and the clinically relevant time points of MRD assessment.

Aims: To confirm our previous results on the clinical relevance of NPM1^mut transcript levels (TL) in an extended cohort of 611 younger AML pts [18 to 60 years (yrs)] harboring NPM1^mut.

Methods: All 611 pts were enrolled in one of four AMLSG [AMLHD98A (n=46; NCT00146120); AMLSG 07-04 (n=199; NCT00151242); AMLSG 09-09 (n=256; NCT00893399); AMLSG 16-10 (n=110; NCT01477606)] treatment trials. Treatment comprised two cycles of induction therapy (idarubicin, cytarabine, etoposide) with or without ATRA or gemtuzumab ozogamicin, or 1 cycle of daunorubicin and cytarabine followed by 1 to 4 cycles of high-dose cytarabine (n=363, 59%), autologous (n=19, 3%) or allogeneic stem cell transplantation (n=162, 27%). 67 (11%) pts did not receive or complete post-remission therapy. NPM1^mut TL (ratio of NPM1^mut/ ABL1 transcripts x 10⁴) were determined by RQ-PCR; the sensitivity of the assays was 10^-5 to 10^-6. Median follow-up (FU) for all patients: 3.2 yrs.

Results: A total of 3527 BM and 2812 PB samples from 611 NPM1^mut pts were analyzed at diagnosis [n=532 (BM); n=385 (PB)], after each treatment cycle [n=1790 (BM); n=1264 (PB)] and during FU [n=1205 (BM); n=1163 (PB)]. Pretreatment NPM1^mut TL in BM and PB were not associated with clinical characteristics or AML-associated gene mutations and had no impact on relapse-free and overall survival (OS). Cox regression analysis using NPM1^mut TL in BM and PB as continuous variable at different time points during therapy was significantly associated with shorter remission duration (RD) and OS.

After twotreatment cycles a > 2log reduction of NPM1^mut TL in BM and PB was significantly associated with a superior OS [p<.0001 (BM); p<.0001 (PB)] and a lower cumulative incidence of relapse (CIR) [p=.0002 (BM); p=.02 (PB)]. At this time point 63/395 (16%) pts became RQ-PCR negative (RQ-PCR^neg) in the BM. CIR at 4 yrs was 10% for RQ-PCR^neg pts versus 40% for RQ-PCR positive (RQ-PCR^pos) pts (p<.0001) and 4-yrs OS was 82% vs 63%, respectively; (p=.01). With regard to PB 129/293 (44%) pts became RQ-PCR^neg which was also associated with a lower CIR (17% vs 49%, p<.0001) but not with a difference in OS (p=.07). Multivariable analysis revealed RQ-PCR^neg in the BM as an independent factor for better OS (p=.02) and longer RD (p=.001), whereas in PB RQ-PCR^neg was only significant for the endpoint RD (p<.0001). In explorative subgroup analyses achievement of RQ-PCR^neg in the BM was significantly associated with a lower CIR independent of the FLT3 -ITD (p=.006) and DNMT3A (p=.04) mutation status. Achievement of RQ-PCR^neg in BM was significantly superior in the FLT3 -ITD^wildtype / DNMT3A^wildtype (26%) and the FLT3 -ITD^pos/ DNMT3A^wildtype (25%) genotypes compared to the FLT3 -ITD^pos/ DNMT3A^mut (8%) and FLT3 -ITD^wildtype/ DNMT3A^mut (8%) genotypes (p <.001).

At the end of treatment, 129/284 pts (45%) were RQ-PCR^neg in the BM. CIR was 22% for RQ-PCR^neg pts versus 48% for RQ-PCR^pos pts (p<.0001). Again, this translated into a better OS (73% vs 61%, respectively; p=.01). Multivariable analysis revealed RQ-PCR^neg in the BM and PB as an independent factor for better OS [p=.05 (BM); p<.0001(PB)] and longer RD [p<.0001 (BM); p<.0001(PB)]. During FU, the CIR at 4 yrs for pts exceeding our previously defined cut-off value of >200 NPM1^mut TL was 88% in each BM and PB with a median time to relapse of only 1.7 and 1.6 months, respectively.

Conclusions: In our study NPM1^mut MRD assessment in BM after 2 treatment cycles was highly informative to identify pts at an increased risk of relapse. Attainment of RQ-PCR^neg was significantly associated with a lower CIR and a better OS, independent of concurrent FLT3 -ITD and DNMT3A mutation . However,achievement of RQ-PCR^neg strongly correlated with FLT3 -ITD^pos/ DNMT3A^mut and FLT3 -ITD^wildtype/ DNMT3A^mut genotypes. During FU, TL above >200 were highly predictive for relapse in both BM and PB samples.