Therapy-related acute myeloid leukemia and myelodysplasia after high-dose chemotherapy and autologous stem cell transplantation

During the last decade the issue of therapy-related myelodysplasia (t-MDS) and acute myeloid leukemia (t-AML) following high-dose chemotherapy (HD-CT) and autologous stem cell transplantation (ASCT) for malignant diseases has become increasingly important, as discussed in previous reviews.1,2 The number of patients with hematologic malignancies and chemosensitive solid tumors undergoing this new type of treatment has expanded dramatically, and because HD-CT and ASCT are followed by a limited morbidity and low mortality, much attention has been paid to the long-term outcome, including secondary malignancies. t-MDS and t-AML are the predominant types of secondary malignancies observed after HD-CT and ASCT. The actuarial risk of leukemic complications has varied considerably, but has often been very high compared to the results in studies of patients treated for the same diseases by conventional chemotherapy and radiotherapy. Several risk factors, the clinical and cytologic characteristics, the chromosome abnormalities, therapy, and survival of this interesting subgroup of patients with t-MDS and t-AML have been discussed in many studies. An important question has been the extent to which the high risk of leukemic complications relates to prior therapy with reinfusion of preleukemic hematopoietic precursor cells at ASCT, the extent to which the high risk could result directly from the HD-CT and ASCT procedures per se, or finally in a few cases could result from chemotherapy administered after ASCT for late relapse of the primary tumor.

Risk and risk factors

Following a few early reports briefly discussing single cases of t-MDS and t-AML,3-7 at least 16 studies,8-23 3 of which were a follow-up of previous reports,9,11,14 have evaluated the actuarial risk of t-MDS and t-AML following HD-CT and ASCT (Table1). In these studies the cumulative risk of t-MDS and t-AML has varied widely from 1.1% at 20 months16and up to 24.3% 43 months after ASCT.17 Much of this difference possibly relates to the wide confidence intervals, because the calculations have often been based on only a few cases of leukemia. However, major variations in risk factors from study to study may also explain the surprising differences observed in these well-conducted and well-controlled studies.

In several previous studies of t-MDS and t-AML following conventional chemotherapy, the risk has been shown to increase with patient age and with the exposure to alkylating agents.24-29 More precisely, in patients with Hodgkin disease the risk has been shown to increase by the square of age and is almost directly proportional to the cumulative dose of alkylating agents.30 In patients treated with HD-CT and ASCT, the risk of t-MDS and t-AML seems also to increase with age and with the burden of alkylating agents administered before the transplantation procedure (Table 1). In the study from Nebraska12 and in the follow-up study from Minnesota9 of patients transplanted for Hodgkin disease and non-Hodgkin lymphomas, ages over 40 and 35 years, respectively, were risk factors (Table 1). In the EBMT cooperative study20 of the same type of patient, age at transplantation was an independent and significant risk factor. In the French cooperative study of patients transplanted for Hodgkin disease,18 age over 40 years was a significant risk factor for the development of all types of secondary malignancy.

Increased duration of previous chemotherapy and treatment with alkylating agents were significant risk factors in 1 study from the Dana Farber Institute of patients with non-Hodgkin lymphomas13 (Table 1), and in a later study from the same institution,15 the number of prior relapses, also reflecting the burden of previous chemotherapy, was a significant risk factor. In the EBMT cooperative study,20 the interval from diagnosis of lymphoma to ASCT, likewise reflecting the burden of previous therapy, was a significant risk factor. In the British cooperative study of patients transplanted for Hodgkin disease,19 the quantity of previous therapy and chemotherapy with MOPP and with lomustine were all significant and independent risk factors for t-MDS and t-AML. In the study from Arkansas21 of patients transplanted for multiple myeloma, all patients who developed t-MDS belonged to a subgroup of 117 patients initially treated with more than 1 cycle of conventional chemotherapy including an alkylating agent, whereas 71 patients who underwent ASCT after only 1 cycle of chemotherapy did not develop leukemic complications (P = .02). In a study from Newcastle16 of patients transplanted for Hodgkin disease and non-Hodgkin lymphomas, in most cases early in the course of their disease, the cumulative risk of t-MDS and t-AML was only 1.1% at 20 months, whereas in our simultaneously published study from Copenhagen17 of multitreated patients of the same type, a very high risk of 24.3% at 43 months was observed. However, this high risk was at the same level as in our previous studies of patients multitreated for the same diseases with conventional chemotherapy.30 These observations all indicate a marked dose-response effect of prior chemotherapy with alkylating agents on the risk of t-MDS and t-AML following ASCT.

The risk of t-MDS and t-AML was low in the 2 studies of ASCT for solid tumors22,23 (Table 1). In both studies patients were apparently less pretreated than patients in the lymphoma studies, and patients with germ cell tumors were in general young. In addition, in the study of germ cell tumors, 2 cases of t-MDS were excluded from the risk estimate because the authors considered them as unrelated to therapy.23 Thus, different factors may have contributed to the low risk of t-MDS and t-AML observed in the 2 studies of HD-CT and ASCT for breast cancer and germ cell tumors.

In previous studies of patients treated traditionally with chemotherapy, including alkylating agents, for lymphomas and solid tumors, the leukemogenic potential of additional high-voltage radiotherapy has been controversial, because no increase in risk was observed in several studies.25,26,28-30 The risk of t-MDS or t-AML after radiotherapy alone, observed in 1 study, was approximately 10-fold lower than the risk following treatment with alkylating agents.27 In 2 independent studies from the Dana Farber Institute,13,15 the first a study of patients treated with ASCT for non-Hodgkin lymphomas, the second of patients treated with ASCT for Hodgkin disease and non-Hodgkin lymphomas, previous radiotherapy, in particular irradiation toward the pelvic region, was a significant risk factor for development of t-MDS and t-AML (Table 1). Perhaps even more important, in the study from Nebraska12 total body irradiation (TBI), used in the preparative regimen for ASCT, reached almost significance as a risk factor. A comparison of the 2 studies from the Dana Farber Institute supports that TBI could be a risk factor. In patients transplanted for non-Hodgkin lymphoma, the use of TBI in the preparative regimen for ASCT resulted in a very high actuarial risk of 18% at 6 years and 19.8% at 10 years,13,14 whereas in the parallel study from the same institution of patients transplanted for Hodgkin disease and non-Hodgkin lymphomas conditioned solely by chemotherapy,15the risk was only 4.2% at 5 years. Finally, in the recently published EBMT cooperative study,20 radiotherapy used in the conditioning regimen for ASCT was a significant and independent risk factor. Low-dose whole-body or hemibody irradiation was used many years ago as therapy for indolent lymphoid malignancies but was abandoned because it resulted in many cases of AML,31 and an actuarial risk of 17% after 15 years was observed.32 As a result of this experience, TBI in the preparative regimen for ASCT is now in many centers used more selectively in patients with malignancies that are highly sensitive to radiotherapy.

Among other risk factors under discussion for development of t-MDS and t-AML was prior splenectomy. In the French cooperative study18 of patients with Hodgkin disease, splenectomy was a risk factor with borderline significance. However, in the British cooperative study19 of the same type of patient, splenectomy was not a significant and independent risk factor. In the follow-up study from the Dana Farber Institute14 a lower number of CD34⁺ cells reinfused at ASCT was a risk factor. In the study of patients with multiple myeloma from Arkansas21 a high risk of 12% at 4 years was observed in the subgroup of 117 patients transplanted late in the course of their disease and 73% of these had received tandem transplantation. A specific effect on the risk of leukemia of repeated transplantations was confirmed by the EBMT cooperative study,20 in which the number of transplants received by each patient was a significant risk factor.

In previous studies of patients with malignant diseases treated conventionally with chemotherapy, often combined with radiotherapy, major variations in the risk estimates of t-MDS and t-AML have been observed. In patients treated with HD-CT and ASCT, similar variations are observed, likewise with an often much higher risk of t-MDS and t-AML in many smaller single-institute studies8-11,13,14,17,21 in which cytogenetic data are provided for each patient, as compared to a lower risk in many of the larger cooperative studies.18-20,23 These obviously benefit from more statistical power, and in the smaller studies a selection for publication of series presenting a very high risk, and the delimitation of the cohorts, could both contribute to a relatively high risk. However, a difference in case ascertainment should also be considered. The systematic use of bone marrow cytogenetics as an important diagnostic tool in most single-institute studies, especially in patients with refractory cytopenia but without an excess of blasts in the blood or the bone marrow, and a more careful clinical follow-up, even of patients with relapse of their primary tumor after ASCT, could also contribute to the difference in risks observed between the 2 types of study. Only rather few undiagnosed cases of t-MDS may lead to a marked underestimate of the actuarial risk of t-MDS and t-AML.

Some years ago, hematopoietic stem cell support for ASCT was always obtained as direct aspirates from the iliac crest without chemotherapy priming. Subsequently, because of a number of potential advantages, the harvesting procedures were changed, and CD34-enriched cells were obtained from peripheral blood after chemotherapy priming and administration of hematopoetic growth factors. The effects of this change on the outcome of the primary malignancy and on the risk of t-MDS and t-AML after ASCT have never been studied prospectively. A retrospective, matched-pair comparison of patients transplanted for Hodgkin disease treated with stem cells from the 2 sources has, as expected, demonstrated an equivalent survival outcome.33However, 3 independent retrospective studies have so far suggested,18 or even significantly demonstrated,8,9,11 a higher risk of t-MDS and t-AML following ASCT with CD34-enriched cells isolated from peripheral blood after chemotherapy priming and growth factors, as compared to ASCT using CD34⁺ cells from the bone marrow without pretreatment. There could be at least 3 reasons for this important difference. First, hematopoietic precursor cells critically damaged by chemotherapy priming, harvested before DNA repair is completed, reinfused and thereby forced to a marked proliferation and self-renewal, could explain the excess of t-MDS and t-AML with the use of cells isolated from peripheral blood. Second, CD34-enriched cells from peripheral blood could, for some reason, be more contaminated with preleukemic, clonogenic cells originating from the prior treatment, than cells isolated from the bone marrow. Third, an increased awareness of the risk of t-MDS and t-AML after the switch from harvesting cells from bone marrow to harvesting cells from peripheral blood, is also a consideration.

Biologically, CD34-enriched cells isolated from bone marrow without priming, and cells obtained from peripheral blood after chemotherapy priming, differ in several respects. Most importantly, they possess a different capacity to restore bone marrow function after myeloablative therapy. The more effective recovery of hematopoiesis, particularly of thrombopoiesis, by using cells isolated from peripheral blood, possibly relates to a higher content of clonogenic, normal hematopoietic precursor cells. This source of cells could likewise preferentially contain clonogenic preleukemic cells.

The CD34-enriched cells from the 2 sources show other important differences. As an example, in leukemia, cells isolated from peripheral blood after HD-CT may contain fewer or no cytogenetically abnormal cells, although such cells are still present in the bone marrow.34,35 Although this finding does not unambiguously indicate reasons for a higher risk of t-MDS and t-AML after ASCT with CD34-enriched cells from peripheral blood, it emphasizes the major biologic differences between the cells obtained by the 2 methods.

The possibility that the higher risk of t-MDS and t-AML observed following the shift of stem cell harvest from bone marrow to peripheral blood could actually be due to an increasing awareness of the problem has recently been supported by data from the EBMT cooperative study.20 In this study, the risk of leukemic complications increased significantly in time from 1978, whereas no difference in the risk was observed in relation to the procedure for stem cell harvest.

The cellular origin of t-MDS and t-AML following HD-CT and ASCT has been subject to much debate. The significant impact of primary chemotherapy with alkylating agents on the risk of t-MDS and t-AML after transplantation, outlined in different ways in several studies,13,15,17,19-21 with an actual risk similar to that of patients treated intensively with traditional types of chemotherapy,30 points toward a disease origin in events before transplantation, but perhaps triggered by a reinfusion of preleukemic hematopoietic precursors. Such an origin is supported by the very short latent period from HD-CT and ASCT to t-MDS or t-AML observed in many patients (see below), and by a recent study36 in which specific cytogenetic abnormalities observed in t-MDS or t-AML after transplantation were shown to be present already in the stem cell harvest in 9 of 12 patients.

Other findings, however, support the view that t-MDS and t-AML could be a direct result of the transplant procedure. The facts that the use of TBI in the preparative regimen for ASCT, and the use of CD34-enriched cells from peripheral blood after chemotherapy priming, have both been reported to increase the risk of t-MDS and t-AML, indicate that at least some cases of t-MDS and t-AML could be directly initiated or triggered by the transplantation procedure. A leukemogenic effect of TBI could seem intriguing because its purpose is a complete bone marrow ablation. However, in allogeneic bone marrow transplantation it has been demonstrated that residual recipient stem cells may often survive a preparative regimen of high-dose cyclophosphamide and TBI, as mixed chimerism was observed in 51% of such patients.37 Analogously, surviving hematopoietic precursors could be the origin of t-MDS and t-AML after ASCT. Also, the significantly increasing risk of t-MDS and t-AML with the number of transplants observed in the EBMT-conducted study,20 and, in cases of AML de novo, a shift in cytogenetic characteristics from balanced translocations to deletions or loss of chromosome arms 5q and 7q at relapse after conditioning with busulfan and cyclophosphamide,38 could be taken into account as a leukemogenic effect of the HD-CT and ASCT procedure itself.

The extent to which the origin of t-MDS and t-AML predominantly relates to prior chemotherapy or to the ASCT procedure has recently been addressed by the major British cooperative study19 of 4576 patients treated for Hodgkin disease. In this study, the general risk of t-MDS and t-AML was closely related to prior therapy with alkylating agents, and in a multivariate analysis, when taking the extent of this therapy into account, HD-CT and ASCT treatment in a subgroup of 595 patients was not associated with a significantly increased risk of leukemic complications (RR 1.83, CI 0.66-5.11). Thus, most cases of t-MDS and t-AML following HD-CT and ASCT are probably initiated by therapy prior to transplantation, whereas a minor part possibly is the result of the transplantation procedure.

The latent period from HD-CT and ASCT to development of t-MDS and t-AML has been specified for at total of 100 patients (Table2). Four cases were observed within 4 months, an additional 6 cases were observed within 8 months, and as many as 26 of 102 cases of t-MDS and t-AML were observed up to only 12 months after ASCT. Because previous studies have shown that a latent period of 12 months or less is rather rare in patients treated with conventional chemotherapy including alkylating agents,24-27,30these data likewise, as discussed, support an origin of the disease as being mainly prior to ASCT. Five cases of t-MDS were observed more than 5 years after ASCT. All 5 patients had previously received chemotherapy with alkylating agents, all 5 had characteristic defects of chromosome 5 or 7 or both, and there was no information on chemotherapy for relapse of their primary malignancy after ASCT. These cases emphasize that, although the general risk of t-MDS and t-AML decreases markedly 5 to 7 years after cessation of therapy,30 cases of late occurrence may occasionally be observed.

In MDS and AML de novo, most cases can be diagnosed and subgrouped morphologically according to the French-American-British (FAB) classification.39,40 In MDS de novo, this classification is an important and independent prognostic factor that can contribute substantially to a prognostic score.41 Whereas many cases of overt t-AML and t-MDS with a high percentage of blasts can be classified according to the FAB criteria, cases of t-MDS with a low percentage of blasts after ASCT often do not meet the FAB criteria for classification.42-44 The bone marrow is often hypoplastic and fibrotic in such patients, and dysplastic changes are observed regularly, also in cases without evidence of t-MDS. For this reason neither the diagnosis nor a prediction of the outcome can be based on bone marrow pathology in such patients, but must rely on the presence of refractory cytopenia as well as clonal chromosome abnormalities in the bone marrow. The course of the disease may differ even in cases with chromosome abnormalities, and an aggressive type as well as a more indolent type of t-MDS have been described.42 

The disease presented as t-MDS in 132 patients and as overt t-AML in only 36 patients in well-documented reports. This large excess of t-MDS is characteristic for leukemic complications following alkylating agents45-47 and can be taken into account for a generally high diagnostic accuracy in these studies.

Detailed cytogenetic characteristics have been published for at least 75 patients with t-MDS or t-AML after HD-CT and ASCT.5,7,8,10,12,13,16,17,21-23,36,38,44Only 4 of these patients presented a normal karyotype, whereas 17 patients (23%) had a complex karyotype with more than 5 different cytogenetic abnormalities (Table 3). As in patients with t-MDS and t-AML in general,45-47 deletions or loss of the long arms or loss of the whole chromosomes 5 or 7 were most common, confirming that the majority of these cases are a result of prior therapy with alkylating agents. Interestingly, 11 patients presented balanced translocations to chromosome bands 11q23 or 21q22 and 2 patients a t(9;22). Twelve of these 13 patients had previously received chemotherapy with DNA topoisomerase II inhibitors, confirming the association between this type of chemotherapy and t-AML with balanced chromosome aberrations.48-51 Likewise, in the follow-up study from City of Hope,11 exposure to DNA topoisomerase II inhibitors before ASCT and priming with high-dose etoposide were significantly associated with the development of t-MDS and t-AML with balanced translocations to chromosome bands 11q23 and 21q22.

The presence of a cytogenetically abnormal clone in the bone marrow after HD-CT and ASCT is considered by most investigators as diagnostic for a leukemic complication, also in cases without an excess of blasts or other cytologic abnormalities characteristic of t-MDS or t-AML. A few cases of this type with uncharacteristic cytogenetic abnormalities have been shown to relate to bone marrow involvement of the primary tumor, detected due to the presence of cytogenetically unrelated clones.52 In other cases, cytogenetically abnormal clones, with uncharacteristic chromosome aberrations, have been observed in some patients without other evidence of MDS.10,11,13,42,53The origin and the significance of these findings have been discussed. In some cases the abnormalities have persisted for prolonged periods of time; in other cases refractory cytopenia and t-MDS have gradually developed. In our experience of t-MDS and t-AML54 as well as in other series, a few patients turn out to be long-term survivors. These are often patients with t-MDS and a single cytogenetic abnormality, mainly monosomy 7, and they may survive for many years before the disease, in all cases, shows progression.

The survival of patients with t-MDS and t-AML after HD-CT and ASCT was in general poor, as demonstrated in Table4. Median survival was 6 months and thereby identical to the survival of patients with t-MDS and t-AML in general.45,54 However, 11 patients were still alive more than 18 months after diagnosis, without further treatment. Five of these 11 long-term survivors, whose disease was discussed in more detail, presented defects of chromosome number 7, in 3 cases as the only cytogenetic abnormality present. Another 3 patients presented a normal karyotype.

Because of the generally poor response to conventional, intensive, antileukemic chemotherapy of patients with t-AML in general and after ASCT, allogeneic bone marrow transplantation has been attempted. In one study of patients with t-AML in general, the actuarial disease-free survival was 24% at 3 years.55 In this study, however, only 5 of 18 patients presented characteristic abnormalities of chromosomes 5 and 7, and only 2 a complex karyotype. In a study from Seattle,56 the 5-year disease-free survival of patients transplanted for t-AML in general was only 7.8% with no difference between untreated patients and patients brought into remission with intensive antileukemic chemotherapy before transplantation. Finally, in the study from the Dana Farber Institute all 13 patients who underwent allogeneic bone marrow transplantation for t-AML after ASCT have died.14 For this reason intensive antileukemic chemotherapy with subsequent harvest of karyotypically normal progenitor cells followed by HD-CT and ASCT has been attempted, and so far demonstrated a clinical effect in a few cases.35 To what extent this type of therapy or other regimens will be a future option, remains to be determined.

As demonstrated in an important joint study from Stanford and City of Hope, cytogenetic screening of the bone marrow before harvesting cells for ASCT is of major importance, particularly in heavily pretreated patients, to avoid transplantation in patients with an already established t-MDS.57 

Because of the poor prognosis, attempts have been made to predict the development of t-MDS and t-AML following ASCT. This has so far been possible to some extent in women by studying clonality, as revealed by the X-chromosome inactivation pattern of the human androgen receptor gene.58,59 To what extent this type of analysis will find a place in clinical practice remains to be determined.

Although t-MDS and t-AML after HD-CT and ASCT do not seem to differ from t-MDS and t-AML after conventional chemotherapy and radiotherapy as far as clinical characteristics, phenotype, and chromosome abnormalities are concerned, the leukemic complications following ASCT present specific problems. Because of the high risk observed in many studies and the dramatically increasing number of patients undergoing this type of therapy, efforts to decrease the risk of t-MDS and t-AML should have a high priority together with a search for more effective types of therapy in patients who have acquired a leukemic complication. Among the possibilities of decreasing the risk of t-MDS and t-AML in patients suffering from a primary tumor with poor prognostic indices is transplantation early in first complete remission. This may improve the outcome with regard to the primary tumor as well as reduce the risk of t-MDS and t-AML. Such measures are already undertaken in many centers, primarily in patients with malignant lymphomas. Likewise, a restrictive use of TBI in the preparative regimens for ASCT must be considered if the risk of leukemic complications should be shown to outweigh the potential benefit of radiotherapy.

Because of its many advantages, harvesting of CD34⁺ cells from peripheral blood after chemotherapy priming has become the dominant standard procedure in obtaining stem cells for transplantation. In consequence, it is now impossible prospectively to solve the question as to what extent this procedure is particularly leukemogenic. Instead, a modification of the priming procedure should be considered. Most rational would be to substitute alkylating agents and DNA topoisomerase II inhibitors in the preparative regimens for ASCT by cytostatic drugs without leukemogenic potential, such as the antimetabolites. Also, the methods for an early detection of clonality before and after transplantation will possibly improve, and if clonality is observed, purging of the CD34⁺ cells harvested for ASCT could be a possibility.

The generally poor results of intensive antileukemic chemotherapy in patients with advanced t-MDS and t-AML after HD-CT and ASCT is analogous to the experience in t-MDS and t-AML following conventional chemotherapy.47-49 They relate primarily to an initial phase of t-MDS observed in most patients and to the cytogenetic characteristics with deletion or loss of the long arms of chromosomes 5 and 7 and a complex karyotype. In addition, in t-MDS and t-AML after HD-CT and ASCT a reduced pool of hematopoietic stem cells may explain the many treatment-related deaths due to persistent bone marrow hypoplasia after intensive antileukemic therapy. Although allogeneic bone marrow transplantation in some studies offers rather poor results, it may at present represent the best choice of therapy in younger patients with advanced t-MDS or t-AML following HD-CT and ASCT, but also new therapeutic approaches should be considered, such as autologous tandem transplantations.

Note added in proof. Since submission of this manuscript, 3 important studies have been published discussing t-MDS and t-AML after HD-CT and ASCT.60-62 

Micallef et al reported an actuarial risk of t-MDS and t-AML of 14.2% and 36.5% at 5 and at 10 years after HD-CT and ASCT in 230 patients transplanted for non-Hodgkin lymphomas.60 Prior therapy with fludarabine and older age were significant risk factors. Twenty-two out of 24 patients with cytogenetic data available showed monosomy of chromosome arms 5q or 7q. The study confirms the high risk of t-MDS and t-AML observed in other single institute studies using cytogenetics as a diagnostic tool.

Yakoub-Agha et al demonstrated a 2-year overall survival of 30% in 70 patients with t-MDS or t-AML after HD-CT and ASCT following allogeneic bone marrow transplantation.61 Age 37 years or younger, female sex, negative CMV serology, and the use of an intensive conditioning regimen were predictive for a more favorable outcome, and patients with abnormalities of chromosome 7 and a complex karyotype had an inferior survival rate. The study confirms that allogeneic bone marrow transplantation must be considered in younger patients with t-MDS or t-AML after HD-CT and ASCT.

Krishnan et al reported a follow-up study of patients transplanted for Hodgkin disease and non-Hodgkin lymphomas.62 The actuarial risk of t-MDS and t-AML was 8.6% at 6 years. An increased risk was observed in 61 out of 612 patients primed with VP-16 for stem cell harvest (P = .002), which likewise was significantly associated with development of t-AML with 11q23/21q22 chromosome abnormalities (P = .006). The study for the first time demonstrates that the type of priming may be of importance.

The authors are indebted to Professor Flemming Skovby, MD, to Mr Harry Cowan, BSc, and to Hanne Nielsen for their help in preparing the manuscript.