Acute myeloid leukemia (AML) relapse after allogeneic hematopoietic cell transplant (allo-HCT) is challenging. Data on extramedullary relapse (EMR) after allo-HCT are limited. We analyzed 215 patients with AML who underwent allo-HCT in our institution between January 2005 and December 2015. We limited this retrospective review to patients who received a MA conditioning, were in complete remission (CR) at the time of transplant and who received a matched sibling transplant, all other patients were excluded to avoid heterogeneity. Seventy-seven (35.8%) patients experienced disease relapse, 45 had BMR, and 32 had EMR. The only variable that was statistically associated with EMR post allo-HCT was male sex (OR = 3.2 (1.2, 8.2), p-value = 0.01); there was a trend for association between transplant in >CR2 and EMR (OR = 0.38 (0.14, 1.06), p-value = 0.06). The median overall survival (OS) after relapse for all relapses was 10 months (95% CI 4.839–15.161). The median OS for BMR group was 8 months (95% CI 2.850–13.150) and 14 months for the EMR group (95% CI 5.776–22.224); however, this was not statistically significant, p-value = 0.4. Multivariate analysis revealed that gender, treatment modality, and time from allo-HCT to relapse (≥12 vs. <12 months) have significant association with the post-relapse death. Male gender was the only significant factor associated with EMR.
Acute myeloid leukemia (AML) relapse after allogeneic hematopoietic cell transplantation (allo-HCT) remains a significant cause of mortality and morbidity. Bone marrow relapse (BMR) is the most frequent presentation. Extramedullary relapse (EMR), in isolation, or along with BMR has been reported in the literature; however, limited information to guide practitioner on how to manage EMR is available. The standard therapy for EMR typically includes a combination of cytotoxic chemotherapy along with radiation and, if feasible cellular therapies or a second transplantation. The mechanism of EMR after allo-HCT for AML remains poorly understood. Tissues with limited drug penetration (central nervous system (CNS) and gonads) are well known sites of EMR for both transplant and non-transplant patients. In the post-transplant setting, EMR probably depends on transplant related variables and disease related factors. The cytotoxic CD8 positive T cells are responsible for the graft-versus-leukemia effect after transplantation and some tissues may have reduced levels or reduced penetration by these cells, and act as a reservoir or niche that later on leads to florid relapse. It remains unclear why these tissues have reduced immunologic surveillance, although a number of mechanisms have been reported including clonal evolution with partial loss of HLA genes and preferential T-cell homing [1,2,3,4]. Disease-related factors like inv(16), t(8;21), CD56 expression, and tetraploidy have been reported as well [5,6,7]. This manuscript aimed to estimate the frequency of EMR and BMR and to explore any differences between relapse in the bone marrow (BM) and extramedullary sites in terms of risk factors and clinical course.
Patients and methods
Between 1 January 2005 and 31 December 2015, 268 AML patients underwent allo-HCT at our institution. Two hundred and fifteen patients were included in this retrospective analysis after excluding patients not in CR at the time of transplant, patients who received less than fully myeloablative (MA) regimen, and patients who received any non-matched sibling transplant (matched unrelated, mismatched, or alternate donor transplant). Of these 215 patients, 77 relapsed. Thirty-two of them experienced EMR, while 45 experienced BMR. All the patients received MA conditioning and matched sibling donor transplant.
EMR was defined as extramedullary involvement regardless of BM involvement; BMR was defined as relapse with no evidence of extramedullary involvement. EMR was diagnosed by physical exam findings or imaging (computed tomography, positron emission tomography or magnetic resonance imaging). When possible, histological confirmation was performed. CNS relapse was diagnosed when leukemic cells were identified in the cerebrospinal fluid. CR was defined as <5% blasts in the bone marrow (BM) with normal complete blood count. Time to post-relapse survival has been calculated from the date of relapse to date of last follow-up or death. Acute graft-versus-host disease (aGVHD) and chronic GVHD (cGVHD) were defined according to the standard criteria [8, 9]. GVHD prophylaxis included cyclosporine A with short-term methotrexate. This study was approved by our institution review board.
Patient characteristics were summarized using frequencies for categorical variables and medians with ranges for continuous variables. Further, categorical variables and continuous variables were compared using Chi-squared test and Wilcoxon rank-sum test, respectively. Probabilities of post-relapse survival have been calculated using Kaplan–Meier estimator with variance estimated using Greenwood’s formula. Survival curves were compared using log-rank test. Multivariate analysis of post-relapse survival was applied using Cox proportional hazard model. Patients, disease, and transplant related variables were tested using step-wise backward procedure. These variables included age, sex, cytogenetics, disease status at allo-HCT, extramedullary disease at presentation, graft source, time from diagnosis to transplant (≤6 vs. >6 months), and time from allo-HCT to relapse (≤12 vs. >12 months). The site of relapse (EMR vs. BMR) will be considered as a main effect and will be included in the model in all steps of model building regardless of its significance. Results will be summarized as hazard ratio (HR) with 95% confidence level. Interactions between main effect and significant variables will be tested. All models will be tested for proportional hazard assumptions. If violated, time-dependent covariate will be constructed. Covariates with p-value <0.05 were considered significant in the final model. Association between site of relapse (EMR vs. BMR) and proposed risk factors will be evaluated using logistic regression. Variables with p-value <0.1 were retained into the final multivariate model using step-wise backward procedure. The effect of cGvHD on relapse was evaluated using the Cox regression model with cGvHD included in the model as a time-dependent covariate. the multivariate model will be construed with all significant factors using step-wise backward procedure.
A total of 268 patients with AML underwent allo-HCT in our institution from 2005 until 2015. Patients who received less than fully MA conditioning, cord blood or non-matched sibling allo-HCT, and patients not in CR at transplant were excluded from this analysis. Two hundred and fifteen patients were included in this retrospective review. Of these patients, 77 have relapsed (35.8%). Thirty-two (42% of all relapses) of them experienced EMR, while 45 (58% of all relapses) experienced BMR. Table 1 summarizes the patients characteristics of all the relapsed patients. The only statistically different characteristic between the BMR and the EMR group was the sex distribution with more females in the BMR group. Age, cytogenetics, graft source, conditioning (total body irradiation (TBI) vs. non-TBI), treatment modality, extramedullary disease at presentation, disease status at HCT, time from allo-HCT to relapse, and time from diagnosis to allo-HCT were similar between the BMR and the EMR group.
The median OS after relapse for all relapses (BMR and EMR) was 10 months (95% CI 4.839–15.161). The median OS for the BMR group was 8 months (95% CI 2.850–13.150), and 14 months for the EMR group (95% CI 5.776–22.224); however this was not statistically significant, p-value = 0.4 (Fig. 1a, b). Univariate analysis revealed that gender, treatment modality, and time from allo-HCT to relapse (≥12 vs. <12 months) have significant association with the post-relapse death. Extramedullary at presentation showed a trend of association with post-relapse death, HR = 2.1 (0.9, 4.4), p-value = 0.05 (Table 2). These factors remained significant in multivariate analysis after forcing the relapse site variable into the final model. However, relapse site did not show significance, HR = 0.8 (0.42, 1.46), p-value = 0.4. Males were at lower risk of post-relapse death as compared to females, HR = 0.5 (0.3, 0.9), p-value = 0.04. Longer time interval from HCT to relapse was also associated with lower risk of post-relapse death, HR = 0.3 (0.16, 0.66), p-value = 0.002. Patients with EM disease at presentation were at higher risk of post-relapse death, HR = 2.4 (1.03, 5.7), p-value = 0.04. Patients who received palliative treatment were at higher risk of death compared to patients who received 2nd allo-HCT or DLI, HR = 3.3 (1.6, 6.4), p-value = 0.001 (Table 3).
Factors associated with EMR
In univariate analysis, gender showed a significant association with relapse site. Males were more likely to develop EMR as compared to females, OR = 3.2 (1.2, 8.2), p-value = 0.01. However, other risk factors did not show any significant association. Gender was the only significant factor associated with EMR. There was no association between EMR and cGvHD, odd ratio = 1.4 (0.5, 3.7), p-value = 0.4. There was a trend to develop EMR for patients transplanted in >CR2; however, it did not reach statistical significance OR = 0.38 (0.14, 1.06), p-value = 0.06 (Table 4).
cGvHD as a time-dependent covariate showed insignificant association with relapse. HR = 0.9 (0.9, 1), p-value = 0.1.
Often, EM relapses of leukemia patients post allo-HCT are studied together regardless of the primary disease and age group (AML, ALL, MDS, adult, pediatric, etc.) [10,11,12,13,14,15,16,17,18]; however, the primary disease and the age have major impact on the physiology and outcomes of the relapse and this makes interpreting the results and drawing firm conclusions from these studies difficult. To minimize these confounders, we focused our review only on relapsed AML in adult patients. Our study suffers from the usual retrospective-reviews limitations; however, the fact it comes from a single institution with uniform treatment protocols and guidelines minimizes the heterogeneity in the treatment received and the patient’s characteristics which help to compare other variables between groups and to draw important conclusions. Out of 268 patients transplanted for AML, 215 patients fulfilled the inclusion criteria and were retrospectively analyzed. Seventy-seven patients relapsed post allo-HCT (35.8%), thirty-two of them experienced EMR (42%), while 45 experienced BMR (58%). The incidence of EMR in the literature ranges from 0.65% up to 41% [12, 16, 17, 19,20,21] and this variability is due to multiple factors including underdiagnoses because imaging is not an integral part of AML follow-up post-transplant, reporting all leukemia EMR together (AML, ALL, CML, etc.), heterogeneity in the definition of EM disease and underreporting the type of relapse in the registries. Sites of EM relapse were soft tissues predominantly (25/32 = 78% of cases), followed by CNS (5/32 = 15.6% of cases) and visceral (2/32 = 6.25% of cases). Diverse sites have been reported in the literature and almost any site and any organ can be affected, but consistently soft tissues and CNS are the most frequently involved sites [10,11,12, 19, 22]. In this study, the time to relapse post allo-HCT was not different between the two groups (EMR vs. BMR), which does not correlate with previous reports, that have reported longer time to relapse in EMR cases [10,11,12, 14, 17,18,19, 23]. We analyzed cGVHD, cytogenetics, disease status at HCT, sex, graft source, conditioning (TBI vs. non-TBI), extramedullary disease at presentation, time from HCT to relapse, and time from diagnosis to transplant for association with relapse site. In univariate analysis, gender was the only statistically significant factor associated with relapse site. Males were more likely to develop EMR as compared to females, OR = 3.2 (1.2, 8.2), p-value = 0.01. However, other risk factors did not show any significant association. There was no association between EMR and cGvHD. Male gender was never reported as a predisposing factor for EMR in previous reports. EM presentation and leukemia subtypes (FAB) M4/M5 were consistently found to be predictive of EM relapse in the reported literature; however, other risk factors like cytogenetics, disease status at transplantation, time to relapse, conditioning regimen (TBI vs. non-TBI), and age were not consistently predictive and some reports even showed contradicting results [10,11,12,13,14, 16, 18, 23]. The median OS after relapse for all relapses (BMR and EMR) was 10 months (95% CI 4.839–15.161). The median OS for BMR group was 8 months (95% CI 2.850–13.150) and 14 months for the EMR group (95% CI 5.776–22.224); however, this difference was not statistically significant, p-value = 0.4. This OS rate compares better to previously reported survival rates after AML relapse post allo-HCT (4–7 months) [19, 20], and this is probably due to the younger population in our study and therefore the propensity to use more aggressive and curatively intended therapy, and because patients not in CR at the time of transplant were excluded from this analysis. Univariate analysis revealed that gender, treatment modality, and time from allo-HCT to relapse (≥12 vs. <12 months) have significant association with the post-relapse death. Extramedullary at presentation showed a trend of association with post-relapse death, HR = 2.1 (0.9, 4.4), p-value = 0.05. These factors remained significance in multivariate analysis; however, relapse site did not show significance, HR = 0.8 (0.42, 1.46), p-value = 0.4.
In summary, in this homogeneous population we showed that EMR represent a large proportion of AML relapses post allo-HCT (42%), and we confirmed that soft tissues and CNS are the most frequently affected sites (as in previous reports). Our study showed that the time to relapse and the survival post-relapse are similar between EMR and BMR. Interestingly, this is the first study to report male gender as a risk factor for EMR, whether this is linked to the menopausal status or the hormonal differences between males and females vs. other disease and patient variables needs to be investigated in future trials.
The characteristics, risk factors, and outcomes of EMR post allo-HCT for AML are still not well defined, probably due to the retrospective nature of the reported studies addressing this issue and the loose inclusion criteria in these studies that allow heterogeneous groups to be lumped and studied together. Ideally we should establish adequate diagnostic algorithms to accurately define the incidence and prevalence of this entity. Also studying isolated EMR separately from combined BMR and EMR and from isolated BMR will be of value to better understand and appreciate the differences between each relapse type. Our study is one of the largest retrospective studies focusing on EMR restricted to AML patients, rather than studying EMR in general with inclusion of different leukemia subtypes. Our study design, in addition to the fact that it is a single institution study addressing a homogeneous population treated similarly, minimized the confounders that most of the other studies suffered from. The study population is rather a young population and probably not representative of all AML patients; results of this study should not be generalized to all AML patients.
StÃlzel F, Hackmann K, Kuithan F, Mohr B, Fussel M, Oelschlãgel U, et al. Clonal evolution including partial loss of human leukocyte antigen genes favoring extramedullary acute myeloid leukemia relapse after matched related allogeneic hematopoietic stem cell transplantation. Transplantation. 2012;93:744–9.
Sackstein R. A revision of Billinghamâ€™s tenets: the central role of lymphocyte migration in acute graft-versus-host disease. Biol Blood Marrow Transplant. 2006;12:2–8.
Jiang YZ, Cullis JO, Kanfer EJ, Goldman JM, Barrett AJ. T cell and NK cell mediated graft-versus-leukaemia reactivity following donor buffy coat transfusion to treat relapse after marrow transplantation for chronic myeloid leukaemia. Bone Marrow Transplant. 1993;11:133–8.
Berthou C, Leglise MC, Herry A, Balcon D, Hardy E, Lessard M, et al. Extramedullary relapse after favorable molecular response to donor leukocyte infusions for recurring acute leukemia. Leukemia. 1998;12:676–81.
Au WY, Kwong YL, Lie AK, Ma SK, Liang R. Extra-medullary relapse of leukemia following allogeneic bone marrow transplantation. Hematol Oncol. 1999;17:45–52.
Byrd JC, Weiss RB. Recurrent granulocytic sarcoma. An unusual variation of acute myelogenous leukemia associated with 8; 21 chromosomal translocation and blast expression of the neural cell adhesion molecule. Cancer. 1994;73:2107–12.
Kuwabara H, Nagai M, Yamaoka G, Ohnishi H, Kawakami K. Specific skin manifestations in CD56 positive acute myeloid leukemia. J Cutan Pathol. 1999;26:1–5.
Przepiorka D, Weisdorf D, Martin P, Klingemann HG, Beatty P, Hows J, et al. 1994 Consensus conference on acute GVHD grading. Bone Marrow Transplant. 1995;15:825–8.
Sullivan KM. Acute and chronic graft-versus-host disease in man. Int J Cell Cloning. 1986;4(S1):42–93.
Harris AC, Kitko CL, Couriel DR, Braun TM, Choi SW, Magenau J, et al. Extramedullary relapse of acute myeloid leukemia following allogeneic hematopoietic stem cell transplantation: incidence, risk factors and outcomes. Haematologica. 2013;98:179–84.
Chong G, Byrnes G, Szer J, Grigg A. Extramedullary relapse after allogeneic bone marrow transplantation for haematological malignancy. Bone Marrow Transplant. 2000;26:1011–5.
Lee KH, Lee JH, Choi SJ, Kim S, Seol M, Lee YS, et al. Bone marrow vs extramedullary relapse of acute leukemia after allogeneic hematopoietic cell transplantation: risk factors and clinical course. Bone Marrow Transplant. 2003;32:835–42.
Ge L, Ye F, Mao X, Chen J, Sun A, Zhu X, et al. Extramedullary relapse of acute leukemia after allogeneic hematopoietic stem cell transplantation: different characteristics between acute myelogenous leukemia and acute lymphoblastic leukemia. Biol Blood Marrow Transplant. 2014;20:1040–7.
Cunningham I. Extramedullary sites of leukemia relapse after transplant. Leuk Lymphoma. 2006;47:1754–67.
Clark WB, Strickland SA, Barrett AJ, Savani BN. Extramedullary relapses after allogeneic stem cell transplantation for acute myeloid leukemia and myelodysplastic syndrome. Haematologica. 2010;95:860–3.
Goker H, Gunes G, Demiroglu H, Malkan UY, Haznedaroglu I, Sayinalp N, et al. Extramedullary relapses after allogeneic stem cell transplantation for leukemia: clinical characteristics, cumulative incidence and risk factors. In: ASCO Annual Meeting Proceedings. p. e18005, Chicago, 2015.
Szomor A, Passweg JR, Tichelli A, Hoffmann T, Speck B, Gratwohl A. Myeloid leukemia and myelodysplastic syndrome relapsing as granulocytic sarcoma (chloroma) after allogeneic bone marrow transplantation. Ann Hematol. 1997;75:239–41.
Shimoni A, Rand A, Shem-Tov N, Yerushalmi R, Hardan I, Nagler A. Isolated extra-medullary relapse of acute leukemia after allogeneic stem-cell transplantation (SCT); different kinetics and better prognosis than systemic relapse. Biol Blood Marrow Transplant. 2009;15:59.
Bekassy AN, Hermans J, Gorin NC, Gratwohl A. Granulocytic sarcoma after allogeneic bone marrow transplantation: a retrospective European multicenter survey. Acute and Chronic Leukemia Working Parties of the European Group for Blood and Marrow Transplantation. Bone Marrow Transplant. 1996;17:801–8.
Mortimer J, Blinder MA, Schulman S, Appelbaum FR, Buckner CD, Clift RA, et al. Relapse of acute leukemia after marrow transplantation: natural history and results of subsequent therapy. J Clin Oncol. 1989;7:50–57.
Simpson DR, Nevill TJ, Shepherd JD, Fung HC, Horsman DE, Nantel SH, et al. High incidence of extramedullary relapse of AML after busulfan/cyclophosphamide conditioning and allogeneic stem cell transplantation. Bone Marrow Transplant. 1998;22:259–64.
Au WY, Ma SK, Kwong YL, Lie AKW, Shek WH, Chow WC, et al. Acute myeloid leukemia relapsing as gynecomastia. Leuk Lymphoma. 1999;36:191–4.
Harris AC, Mageneau J, Braun T, Kitko CL, Choi SW, Ferrara JLM, et al. Extramedullary relapse in acute leukemia following allogeneic hematopoietic stem cell transplantation: incidence, risk factors and outcomes. Biol Blood Marrow Transplant. 2013;16:S177–8.
Conflict of interest
The authors declare that they have no conflict of interest.
About this article
Cite this article
Alhashim, N., Aljurf, M., Hassanein, M. et al. Extramedullary relapses after allogeneic stem cell transplantation for acute myeloid leukemia: clinical characteristics, incidence, risk factors and outcomes. Bone Marrow Transplant 53, 838–843 (2018). https://doi.org/10.1038/s41409-018-0093-9
Isolated cerebrospinal fluid relapse of acute myeloid leukemia status post stem cell transplant: A neuroleukemiosis manifesting as polyneuropathy
American Journal of Hematology (2020)
Extramedullary Relapse of Acute Myelogenous Leukemia after Allogeneic Hematopoietic Stem Cell Transplantation
Biology of Blood and Marrow Transplantation (2019)