Clinical characteristics and post-transplant clinical outcomes

A total of 257 adults with IDH2-mutated AML undergoing alloHCT in CR1 at a CIBMTR reporting site between 2013 and 2019 were included in this study (Table 1). The median age was 60.1 (range 19.8–79.3), 84% were Caucasian, 50% were female, 57% had a Karnofsky performance score of 90 or above, and 48% had an HCT-CI comorbidity index of 3 or greater. In addition to IDH2 mutations in all patients, NPM1 and/or FLT3-ITD mutations were reported at initial diagnosis in 41%. Transplants were from peripheral blood 75% of the time, 49%  had myeloablative conditioning, and 60% were from matched unrelated donors.

Table 1 Patient clinical characteristics.

Overall, the 1-year and 3-year cumulative incidence of relapse in this cohort was estimated as 16% and 24%, respectively (Fig. 1a, Supplementary Fig. 1). Of 57 relapses recorded, 41 (72%) occurred within 12 months following transplantation. For all baseline characteristics studied (Table 1), using univariable competing risk regression considering NRM as the competing risk event, only cord blood compared to peripheral blood as a graft type (HR: 2.3, 95% CI 1.2–4.5; Supplementary Fig. 2A) was related with higher relapse risk.

Fig. 1: Clinical outcomes of AML patients with IDH2 mutations at baseline and the association with MRD after allogeneic hematopoietic cell transplant.

Cumulative incidence of relapse (left) and overall survival (right) are shown for IDH2 mutated patients (a) for the entire cohort (b) based on the presence (Flow MRDpos) or absence (Flow MRDneg) of reported clinical flow cytometry measurable residual disease (MRD), and (c) based on the presence (IDH2 NGS MRDpos) or absence (IDH2 NGS MRD neg) of residual IDH2 variants by next generation sequencing (NGS) MRD assay.

Overall, the 1-year and 3-year OS in this cohort were estimated as 81% and 71%, respectively (Fig. 1a). Median follow-up time among censored patients was 25 months. Two baseline characteristics identified by univariable Cox regression analysis as associated with differences in OS were HCT-CI (3+ vs. 0, HR: 2.8, 95% CI: 1.3–5.9) and AML group (transformed from myelodysplastic syndrome/myeloproliferative neoplasm vs. de novo HR: 2.1, 95% CI: 1.1–4.3; Supplementary Fig. 2B).

Pre-transplant flow cytometry and residual IDH2m detection

Flow cytometry is commonly performed for patients with AML in remission prior to transplant as a test of MRD to estimate post-transplant relapse risk [12], although the clinical utility of such testing as currently performed has been questioned [17, 40]. 247 of the 257 (96%) patients in this cohort had pre-transplant remission flow cytometry results reported to the CIBMTR registry, of which 25 (10%) tested positive. Testing positive by flow cytometry pre-transplant did not predict any differences in relapse or OS compared with those testing negative (Fig. 1a).

Testing for IDH2m persistence for patients with AML in remission prior to transplant is not routinely performed clinically. A custom error-corrected NGS assay was validated to detect IDH2 mutations down to a variant allele fraction (VAF) of at least 0.1% (Supplementary Fig. 3A). NGS analysis detected residual IDH2m in CR1 pre-transplant blood samples from 130 patients (51%) with VAFs ranging from 0.05 to 56% (median: 3%; Supplementary Fig. 3, Supplementary Table 1). Those with IDH2m detected (IDH2 NGS MRDpos) had increased relapse (3 yrs: 29% vs. 18%, +11%, 95% CI: 0.2% to 22%; overall p = 0.03), decreased OS (3 yrs: 58% vs. 83%, −25%, 95% CI: −13% to −38%; overall p < 0.001), and decreased relapse-free survival (RFS 3 yrs: 53% vs. 70%, −17%, 95% CI: −4% to −30%; overall p = 0.002) compared to those testing negative (IDH2 NGS MRDneg, Fig. 1c). The OS for IDH2 NGS MRDneg patients was estimated to be 83% (95% CI: 75% to 89%, p < 0.001) at 3 years and only 6 of the 21 deaths in this group were relapse related.

Association of age, residual IDH2m burden and type with clinical outcome

To study the association of IDH2m detection in pre-transplant CR1 and clinical outcomes in younger and older patients, subgroups were created by using 60 years of age as a cutoff (≥60 vs. <60). For both age groups, decreased OS was observed for those testing positive for IDH2m compared to those testing negative (<60 yrs, 3 yrs: 62% vs. 90%, −28%, 95% CI: −12% to −43%, overall p < 0.001; ≥60 yrs, 3 yrs: 54% vs. 73%, −19%, 95% CI: −1% to −40%, overall p = 0.04; Fig. 2a, b; Supplementary Fig. 4). OS for the younger group testing negative for persistent IDH2m was 90% (95% CI: 81% to 95%, p < 0.001); among 9 patients who died in this group only 3 experienced relapse. Older patients had an increased rate of relapse when testing positive for persistent IDH2m compared to those testing negative (3 yrs: 33% vs. 12%, +21%, 95% CI: 6% to 36%, overall p = 0.007). Interestingly, the level of IDH2m persistence was significantly higher in older versus younger patients (median VAFs: 5.4% vs. 1.4%, ≥60 yrs vs. <60 yrs, p = 0.01). To examine a potential dose effect of IDH2m persistence, patients were divided into two VAF subgroups with 2.5% as the cutoff, but the risk of relapse was not significantly different between the VAF groups (Fig. 2c). Among patients with residual IDH2m detected, 13% (n = 17) had R172 variants and the remaining had R140 variants. The type of residual IDH2m did not further stratify the risks of clinical outcomes (Fig. 2d).

Fig. 2: Baseline characteristics for IDH2 mutated AML patients and the association with clinical outcomes after allogeneic hematopoietic cell transplant.

Cumulative incidence of relapse (left) and overall survival (right) are shown for IDH2 mutated patients by the presence (IDH2 NGS MRDpos) or absence (IDH2 NGS MRDneg) of residual IDH2 variants and by (a) age group below 60 years old, (b) age group 60 years or above, (c) different variant allele fraction (VAF) groups, and (d) by IDH2 mutation type.

Co-occurrence of mutated IDH2 with mutated NPM1 and FLT3-ITD

As IDH2 mutated AML often also has mutations in NPM1 and/or FLT3-ITD [2, 4], the cohort was separated into those with (n = 106, 41%) or without (n = 151, 59%) either of these two mutations reported at initial diagnosis. In the IDH2 mutated patients who did not also have mutated NPM1 and/or FLT3-ITD at baseline, IDH2m were commonly detected in pre-transplant remission blood (n = 80, 53%) and was associated with increased relapse and decreased OS compared with testing negative (relapse 3 yrs: 31% vs. 12%, +19%, 95% CI: 6% to 32%, overall p = 0.01; OS 3 yrs: 56% vs. 79%, −23%, 95% CI: −6% to −40%, overall p = 0.005; Fig. 3, Supplementary Fig. 5).

Fig. 3: Clinical outcomes for IDH2-mutated AML patients without baseline mutations in NPM1 or FLT3-ITD and the association with residual IDH2 variants.

Cumulative incidence of relapse on the left and overall survival on the right for patients based on the presence (IDH2 NGS MRDpos) or absence (IDH2 NGS MRDneg) of residual IDH2 variants.

In contrast, for patients known to also have mutations in NPM1 and/or FLT3-ITD at initial diagnosis (n = 106), detection of persistent IDH2m pre-transplant was also common (N = 50, 47%) but was prognostic only for OS and not relapse (Fig. 4a, Supplementary Fig. 6). Since the persistence of NPM1 and/or FLT3-ITD variants in pre-transplant CR1 blood has already been shown to be strongly associated with worse post-transplant clinical outcomes [17], the utility of these markers was validated in this specific setting. As anticipated the detection of persistent NPM1 and/or FLT3-ITD variants (n = 20, 19%) was strongly associated with increased relapse, while those with only residual IDH2m detected (n = 35, 33%) had similar relapse risk as those testing negative (3 yrs: 65% vs. 12% vs. 16%; overall p < 0.001; p < 0.001 for NPM1/FLT3-ITD vs. negative, p = 0.8 for IDH2m only vs. negative; Fig. 4b). Both residual mutation positive groups had inferior OS compared to the negative group while patients with residual NPM1/FLT3-ITD had the worst estimated OS (3 yrs: 41% vs. 71% vs. 94%; overall p < 0.001). None of the patients testing negative (n = 51, 48%) experienced relapse-related mortality among the 4 reported death events.

Fig. 4: Clinical outcomes of IDH2-mutated AML patients with co-mutated NPM1 or FLT3-ITD at baseline and the association with residual variants.

Cumulative incidence of relapse on the left and overall survival on the right for patients based on (a) the presence (IDH2 NGS MRDpos) or absence (IDH2 NGS MRDneg) of residual IDH2 variants or (b) the presence of residual NPM1 and/or FLT3-ITD variants (NPM1/FLT3-ITD NGS MRDpos), the presence of only residual IDH2 variants (IDH2 NGS MRDpos), or the absence of residual variants (NGS MRDneg).

Modifying effects of other clinical characteristics on residual IDH2m detection

Given evidence that the intensity of the conditioning regimen used prior to transplant can impact clinical outcomes in MRD positive patients [22], we next examined the modifying effect of conditioning intensity and IDH2m persistence. Patients were grouped into low intensity (reduced intensity conditioning (RIC) and nonmyeloablative (NMA)) and high intensity (RIC with melphalan and myeloablative conditioning (MAC)) regimens. When considering all patients, those with IDH2m persistence who received RIC/NMA had the highest rate of relapse (3 yrs: 35%, 95% CI: 20–51%) and lowest OS (3 yrs: 51%, 95% CI: 34–66%; Fig. 5a, Supplementary Fig. 7), but there was no statistically significant difference among the residual IDH2m patients based on treatment regimen. Conditioning intensity was not randomized in this retrospective observational study but was likely determined, at least in part, by judgements regarding patient performance status and comorbidity. Age is an important factor, and as expected RIC/NMA was more frequently given to older patients (<60 yrs vs. ≥60 yrs: 11% vs. 42%, Chi-squared test p < 0.001). While there were few cases of patients <60 years old with persistent IDH2m who received RIC/NMA, when looking at those ≥60 years, higher rates of relapse were observed in patients testing IDH2m positive when given RIC/NMA compared to those testing negative (1 yr: 29% vs. 0, +29%, 95% CI: +13.3% to +43.8%, p < 0.001; Fig. 5b).

Fig. 5: Evaluation of conditioning intensity modifying the association of residual IDH2 mutations and clinical outcomes.

Cumulative incidence of relapse on the left and overall survival on the right for (a) the entire cohort of IDH2-mutated AML patients, and (b) those ≥60 years of age based on the presence (IDH2 NGS MRDpos) or absence (IDH2 NGS MRDneg) of residual IDH2 variants and conditioning intensity received. MAC myeloablative conditioning, Mel melphalan, RIC reduced intensity conditioning, NMA nonmyeloablative conditioning.

For the full cohort, multivariable Cox regression analysis for OS indicated that patients with persistent IDH2m had a higher risk of relapse compared to those testing negative (HR: 2.5, 95% CI: 1.5–4.2, p < 0.001) after adjusting for other variables of interest, which also had significant differences in OS risk including hematopoietic cell transplantation-comorbidity index and age group (Table 2). Residual IDH2m remained prognostic in relapse using competing risk regression (IDH2 NGS MRDpos vs. IDH2 NGS MRDneg, HR: 2.1, 95% CI: 1.4–5.7, p = 0.003) when considering graft type and IDH1 baseline status.

Table 2 Estimated hazard ratios of clinical outcomes in multivariable analyses.

The prognostic significance for persistent IDH2m remained when patients did not have NPM1 and/or FLT3-ITD mutations at baseline (HR for OS: 2.5, 95% CI: 1.3–4.7; HR for relapse: 3.3, 95% CI: 1.5–7.2; Table 2) when including other clinically important characteristics in the regression model. For patients with NPM1 and/or FLT3-ITD mutations at baseline, residual NPM1 and/or FLT3-ITD variants was the most significant variable in predicting both OS and relapse outcomes (OS, HR: 10, 95% CI: 3.2–31.2, p < 0.001; relapse, HR: 21, 95% CI: 7–63, p < 0.001), while residual IDH2m was only significant for relapse (HR: 4.36, 95% CI: 1.4–13.7, p = 0.01), when adjusting for other covariates of interest (Table 2).

Since the IDH2 inhibitor enasidenib was approved by the FDA on August 1st of 2017, patients were divided into two subgroups with alloHCT date before (n = 44, 17%) or after (n = 213, 83%) the date of drug approval to examine the potential therapeutic impact. Although the patient numbers were not balanced between the two groups, the difference in rates of relapse between the residual IDH2m positive and negative groups was greater before the approval date versus after (before: 45% vs. 23%, after: 26% vs. 15%, overall p = 0.03; Fig. 6, Supplementary Fig. 8A). There was also a difference in NRM comparing those testing positive versus negative for residual IDH2m (before: 0 vs. 18%, after: 22% vs. 10%, overall p = 0.005), with 9/10 death events relapse related for patients in the IDH2m positive group transplanted before August 2017. The same patterns in clinical outcomes were observed when removing patients with baseline FLT3-ITD mutation (Supplementary Fig. 8B).

Fig. 6: The association of enasidenib drug approval date and residual IDH2 mutations on clinical outcomes.

Cumulative incidence of relapse on the left and non-relapse related mortality (NRM) on the right for IDH2-mutated AML patients based on the date of transplant occurring before or after the approval date of enasidenib (August 2017) and the presence (IDH2 NGS MRDpos) or absence (IDH2 NGS MRDneg) of residual IDH2 variants.