Although AML is usually fatal, ~40% of younger patients can be cured.1, 2 Each successive year of first CR (CR1) correlates with prolonged survival and cure.3 To this end, we recently analyzed Eastern Cooperative Oncology Group (ECOG) data and observed that there were very few relapses in patients who maintained CR1 for at least 3 years (14/1229, 1.1%).4 Relapses occurring after 3 years from CR1 were considered late (LR), and usually occurred between 3 and 5 years after CR1 (12/14, 86%); two very LRs occurred after 5 years. Interestingly, the majority of LRs (82%) occurred in patients with normal cytogenetics, and no patients had an unfavorable karyotype. Of the 183 patients that underwent allogeneic hematopoietic cell transplantation (allo-HCT) on protocol (in CR1), there was only one LR (0.6%). Here, we describe three additional very LRs after allo-HCT (where post-transplant CR lasted >5 years) that occurred subsequent to our initial report. All four cases (Table 1) featured extramedullary relapses, suggesting a common post-transplant mechanism.

Table 1 Characteristics of four AML patients with late or very LR after allogeneic HCT
Full size table

In our first case, a 37-year-old white female presented with AML without cellular maturation. Her karyotype was normal. She achieved CR1 with standard therapy followed by four cycles of consolidation. She relapsed ~16 months after CR1 with dysplastic changes and 6% blasts. She proceeded directly to allo-HCT using PBSC from her HLA-matched brother. Conditioning consisted of cyclophosphamide and TBI. Cyclosporine and methotrexate were used as GvHD prophylaxis. She remained disease-free post transplant for 11 years, until she noted a small red nodule on her right lower leg. Lesional biopsy revealed AML with monocytic features (involving recipient cells as indicated by karyotyping). Blood counts were normal and a marrow biopsy revealed no evidence of AML (cytogenetics: 46, XY comprising donor cells). Positron emission tomography (PET) and CSF studies confirmed an isolated relapse. The patient was treated with involved field radiation therapy (IFRT) followed by high-dose cytarabine (HiDAC) and was in continuous remission for 2 years. She then had another cutaneous relapse on the right leg, again treated with IFRT, this time followed by maintenance lenalidomide and she has remained in third remission post transplant for 1 year to date.

In another case, a 33-year-old white female developed inv(16) AML with monocytic differentiation. She received standard induction and achieved CR1. Following one cycle of consolidation, she developed an isolated central nervous system relapse. Treatment with central nervous system-directed therapy and HiDAC led to a brief remission (<1 year). At progression (24% marrow blasts), re-induction chemotherapy was deferred and she received a myeloablative peripheral blood allo-HCT from her HLA-matched sister. She received cyclophosphamide and TBI conditioning with cyclosporine and methotrexate for GvHD prophylaxis. She remained in remission for >10 years post transplant before developing an extramedullary relapse (presenting as a right pelvic side wall mass). Additional staging (bone marrow biopsy and PET/CT) confirmed an isolated relapse. She underwent resection of the mass and partial omentectomy. Postoperatively she received IFRT to the pelvis, but declined additional chemotherapy. Less than 1 year later, she once again developed an isolated pelvic extramedullary recurrence. She was started on palliative therapy with decitabine, but developed progressive disease with right ureteral and small bowel obstruction. Ultimately, she succumbed to her disease.

In a third case, a 25-year-old white male presented with AML featuring monocytic differentiation and normal cytogenetics. He received standard induction chemotherapy as part of a non-ECOG clinical trial. He required two inductions attempts to achieve CR1, which was consolidated with two additional cycles of idarubicin and cytarabine as per protocol. Relapse was confirmed 3.5 years later with 30% marrow blasts. Re-induction was deferred and the patient underwent conditioning chemotherapy (cyclophosphamide and TBI) followed by infusion of his HLA-matched brother’s stem cells. GvHD prophylaxis included cyclosporine and methotrexate. Post transplant, he obtained a disease-free interval of 15 years before developing gastrointestinal symptoms and a white count of 150 × 109/L with 69% blasts. Marrow biopsy confirmed relapsed AML with complex cytogenetics. Salvage chemotherapy led to a cytogenetic remission that was consolidated with HiDAC and an infusion of lymphocytes from the original donor. Within a year, he developed left sided proptosis with diplopia on upward gaze. An orbital mass was discovered on imaging. Biopsy of the lacrimal gland confirmed involvement with AML. PET–CT confirmed widespread areas of increased metabolic activity in addition to the left orbit, (cervical/gastric lymph nodes and left testicle). Despite an aggressive attempt to salvage the patient (with chemotherapy and involved field radiation to the left orbit), the patient ultimately progressed with CSF involvement and died from his disease.

We describe four unique cases of LR AML after allo-HCT—one previously reported ECOG case4 and three new cases from patients treated at the Mayo Clinic. LR after allo-HCT is rare, and, to our knowledge, this is the largest case series of this kind to date. Also, although extramedullary relapse is known to occur relatively common after allo-HCT,5 late extramedullary relapse has been only rarely reported.6 Two findings stand out when examining our data: (1) all of the late recurrences featured extramedullary disease, suggesting a unifying mechanism underlying these relapses; and (2) three of the four LRs occurred very late (⩾10 years after transplant).

The differences in LR location (bone marrow vs extramedullary) and remission duration (late vs very late) in patients who were treated with chemotherapy induction and consolidation vs allo-HCT suggest different pathogenic mechanisms underlying late recurrence. LR after conventional chemotherapy, which typically occurs 3–5 years after CR1, is isolated to the bone marrow, and is more common in patients with a normal karyotype. It may represent relapse of a dormant subclone that has acquired new mutations or epigenetic alterations.7, 8 On the other hand, LR after allo-HCT, which appears to occur after much longer CR durations, is more likely related to defective or senescent immune surveillance.5, 9 Of course, in either case, late recurrence may also be therapy related, that is, a new therapy-related AML (t-AML).10 Although none of the ECOG LRs had cytogenetic evidence of t-AML, the patient in case 3 did present with clonal evolution to a complex karyotype (had a normal karyotype at initial diagnosis). Also intriguing is the possibility that the late recurrences presented in case 2 or 3 could have been donor cell derived, as these were sex-matched transplants and we do not have molecular chimerism data. However, donor-derived leukemias are rare, and the extramedullary nature of the relapses and other unique clinical features suggest that the LRs were from recipient cells.11

Interestingly, all four LRs after allo-HCT featured extramedullary involvement—and three were isolated extramedullary relapses—which suggests that impaired peripheral immune surveillance could be a very important risk factor for LR. Three of the LRs after allo-HCT (two of the very LR cases from the Mayo Clinic and the one ECOG LR case) had monocytic differentiation at diagnosis, which could partially explain the extramedullary recurrence. In addition, two of the three new cases had a normal karyotype and the other was inv(16) at initial diagnosis, consistent with our previous report demonstrating that most LRs have normal or favorable cytogenetics at diagnosis and none have poor-risk features.4 Importantly, the three Mayo Clinic LR patients were transplanted at first relapse with active disease (6–30% marrow blasts). These patients had no evidence of extramedullary disease at the time of transplant, but this raises the possibility that residual marrow leukemia is an additional risk factor for late extramedullary recurrence.

In conclusion, LR can occur very late and close attention should be paid to extramedullary sites in allo-HCT recipients. In surveillance and survivorship clinics, any abnormal skin or radiologic finding should be biopsied, as identifying a late extramedullary relapse early in its evolution may lead to more effective salvage therapy. For example, case 1 received HiDAC and IFRT to an isolated leg lesion and remained in continuous remission for 2 years. This was followed by another cutaneous relapse in the same leg, and she again received IFRT followed by maintenance lenalidomide, and has remained in remission for 1 year to date. The exact cause of late transplant-failure at extramedullary sites is unclear, but may result from the development of immune tolerance at peripheral sites (which are also shielded to a degree from chemotherapy and conditioning regimens).5, 6, 12 This reinforces the importance of an intact GvL effect for controlling dormant leukemia, particularly at peripheral or sanctuary sites, even years after transplant.13 In the future, therapies to augment the GvL effect, such as modified cellular-based therapies or checkpoint inhibitors, in addition to the standard approaches employed in our cases, should be considered for extramedullary relapse after allo-HCT, particularly if it occurs after a long period of immune control.