Discussion

AML in older adults has higher rates of induction failure and relapse than AML in younger adults.^[1,4,5,12] In this multi-institutional study of AML patients aged ≥60 undergoing allogeneic transplantation in first CR, we found that baseline clinical and genetic variables, but not remission MRD status, were the predominant determinants of posttransplant risks. High baseline molecular risk was driven independently by FLT3-ITD and TP53 mutations, which contributed to relapse risk, and by JAK2 mutations, which contributed to NRM, possibly as a consequence of potentiated inflammatory signaling that persisted after transplant.

We found that nearly 80% of patients had persistent mutations before transplant despite having achieved a morphologic CR. Whereas some patients had clonal remissions with only DNMT3A or TET2 mutations that did not impact outcomes, most had persistent MDS- and AML-associated mutations that were linked to an elevated risk of relapse. Recent results have indicated that intensification of transplant-preparative regimens could mitigate the risk of MRD-associated relapse.^[13] However, in this cohort of real-world older patients with AML, we observed a higher rate of NRM than was reported in clinical trials of older patients with AML^[8,38] or in younger patients who received RIC.^[13] These results raise the concern that intensified conditioning would not be tolerated in older patients with elevated vulnerability to treatment-related toxicity^[39] and highlights the need for novel approaches to mitigating risk of relapse in this population.

We found that the likelihood of MRD positivity was closely related to other baseline features of AML, including the presence of high-risk gene mutations, such as TP53, high-risk karyotypes, and prior MDS. Moreover, among patients with MRD-positive remissions, those with secondary genetic ontogeny or TP53 mutations at diagnosis had higher levels of molecular MRD than those with de novo genetic ontogeny. Previous studies have not determined whether AML mutations that are present at remission are an independent risk factor for relapse or reflect high-risk AML biology.^[13,36] Our findings suggest that the high rate of molecular persistence in this cohort may be related in particular to the evolution of many of these leukemias from MDS, consistent with the association between MDS and advancing age.

The distinction between molecular persistence and relapse risk is further supported by the observation that secondary genetic ontogeny was associated with persistence but not relapse, whereas clinically defined secondary AML was associated with relapse but not persistence. A possible contributor to this distinction is the higher rate of hypomethylating agents administered to patients with sAML before their AML diagnosis (40 of 91 patients with sAML [44%] vs 28 of 127 [22%] with secondary ontogeny), which may have selected for more therapy-resistant subclones that drove relapse in the sAML group. Consistent with this, the 3-year cumulative incidence of relapse (CIR) for sAML vs non-sAML was 45.8% vs 31.3% (P = .02), whereas the 3-year CIR for molecular ontogeny (de novo vs secondary vs TP53) was 28.3% vs 30.8% vs 84.8% (P < .0001).

In this cohort, MRD positivity did not have an independent impact on LFS after consideration of other pretransplant risk factors. Consistent with prior studies, MRD negativity was significantly associated with reduced risk of relapse, but this benefit was offset by elevated NRM among higher risk groups. These results are particularly important in the context of a recent study from a randomized cohort showing that MAC mitigates the relapse risk associated with MRD positivity.^[13] Although those findings have prompted a push to employ myeloablative regimens for eligible AML patients with persistent molecular disease after induction,^[40] they may not be applicable to many older patients, given the younger age of that cohort (median 55 and capped at 65), the small number of genes sequenced with omission of many MDS-associated genes commonly mutated in older patients with AML (such as SRSF2 and U2AF1), and the requirement that all patients be eligible for MAC before randomization. Indeed, although our cohort did not include enough patients receiving MAC to directly compare outcomes to RIC, the high rate of NRM in our and other older cohorts suggests in general that older patients with AML with MRD may not derive the same benefit from conventional treatment intensification as younger, fitter patients. Instead, older AML patients with persistent molecular disease before transplantation may benefit from nonintensive approaches aimed at achieving molecular clearance and reducing relapse risk, such as targeted MRD erasers after transplant maintenance therapy^[41-47] and, pending the results of ongoing clinical trials, novel monoclonal antibodies^[48] or immunologic therapies.^[49]

Germline DDX41 mutations predispose to development of myeloid malignancies later in life,^[18] and were present in 5.4% of patients in this cohort. We found that DDX41 mutations were associated with overall favorable outcomes, including MRD-negative remission after initial therapy and prolonged LFS after HCT. The improved LFS in older patients with DDX41 mutations was driven entirely by a very low risk of relapse, and the risk of NRM was similar to that of other patients. This finding suggests that patients with germline DDX41 mutations may benefit from strategies to minimize transplant toxicity, including the selection of lower-intensity conditioning regimens associated with reduced risk of NRM^[50,51] or the use of posttransplant cyclophosphamide to minimize the risk of chronic graft-vs-host disease.^[52,53]

Given the evolving nature of AML therapy, conclusions from retrospective analyses are fundamentally limited in their application to current and future therapies. Further, the time of MRD sample collection and assessment relative to initial induction, subsequent consolidation or maintenance therapies, and transplant-preparative regimen may contribute heterogeneity to observed outcome associations. As such, use of this baseline model to guide clinical decision making for older patients with AML will require validation by subsequent prospective studies of cohorts with uniform prior treatment histories and sample collection time points. Nevertheless, our results indicate that most of the posttransplant relapse risk in older patients with AML is encoded at the time of diagnosis and that the likelihood of molecular MRD positivity in these patients is primarily a reflection of baseline genetic risk. Attempting to eradicate persistent AML mutations with dose intensification may be counterproductive in older patients, given their high risk of treatment-related toxicity. This observation highlights both the importance of optimizing initial therapy for older patients with AML who are transplant candidates through use of liposomal daunorubicin/cytarabine and FLT3 inhibitors,^[54,55] as well as the development of nonintensive strategies for mitigating relapse risk in older patients who remain MRD positive before transplantation. Taken together, our results provide a framework for developing risk-adapted strategies and interpreting the relative prognostic impact of baseline and remission genetics in older patients with AML who are transplant candidates.