Management of Relapsed Acute Myeloid Leukaemia

Results of treatment for acute myeloid leukaemia (AML) have steadily improved over the last decade, with remission rates of 60–80%, and a 5-year survival of > 40% reported by leading collaborative groups ( Appelbaum & Kopecky, 1997; Burnett et al, 1998 ; Mayer et al, 1994 ; Hann et al, 1998 ; Stevens et al, 1998 ). This improved outcome has enabled the identification of groups of patients with different prognoses, and the development of risk-directed treatment strategies ( Keating et al, 1988 , 1996; Buchner & Heinecke, 1996; Martinez-Climent et al, 1995 ; Priesler, 1991; Creutzig et al, 1990 ; Bloomfield et al, 1997 ; Buchner et al, 1998 ; Wheatley et al, 1999 ; Mitus et al, 1995 ). These improved response and survival rates have largely resulted from the intensification of chemotherapy combined with good supportive care, but have been associated with a treatment-related mortality rate of up to 15%. This mortality rate has been increased further by the widespread use of autologous and allogeneic bone marrow transplantation (BMT), which continue to be recommended at most centres for standard and poor risk patients fit enough to tolerate the procedures. However, relapse still remains the main cause of treatment failure, and occurs in around 50% of those who achieve a complete remission. Over 90% of relapses involve the bone marrow, whereas central nervous system (CNS) relapse is very uncommon following regimens with limited or even no intrathecal chemotherapy, and there is no evidence that cranial radiotherapy is required to prevent CNS disease. Most relapses are early, with 60% occurring in the first year from complete remission and few if any occurring after 4 years. The prognosis following relapse is poor, with 2-year overall survival rates in retreated patients around 20% ( Vignetti et al, 1996 ; Steuber et al, 1996 ; Vey et al, 1998 ; Stahnke et al, 1998 ; Webb et al, 1999 ; Keating et al, 1988 ; Thalhammer et al, 1996 ; Angelov et al, 1991 ), and therefore the most important strategy remains improved front-line therapy to reduce the relapse rate. The causes of relapse remain poorly defined, although a number of contributory mechanisms have been described. Multidrug resistance (MDR) due to expression of the drug efflux pump P glycoprotein (Pgp) encoded by the mdr 1 gene, expression of lung resistance protein (LRP), low levels of the enzyme topoisomerase II, increased levels of the enzyme glutathione synthetase, and inhibition of apoptosis due to increased expression of bcl-2 have been described and implicated ( Wood et al, 1995 ; Van Den Heuvel-Eibrink et al, 1997 ; List et al, 1998 ; Xu et al, 1996 ; Willman, 1997). Most emphasis has been placed on the role of Pgp; expression of Pgp is more common in secondary AML and after relapse, and patients demonstrating Pgp have lower remission and survival rates. Despite the availability of agents which reduce or block the action of Pgp such as cyclosporin, verapamil and PSC 833, there is no evidence at present that routine use of these MDR reversal agents is beneficial, although doses of cyclosporin and verapamil are limited by potential and actual toxicities, and work continues to identify agents with a more favourable therapeutic ratio ( Yin et al, 1998 ; List, 1996). A variety of different clinical and laboratory features have been proposed as a basis for the assignment of patients to risk groups with statistically significant differences in survival and relapse rates. There is widespread agreement that marrow cytogenetics determined at diagnosis are especially important with a better prognosis associated with the (8;21) and (15;17) translocations, and inversion of chromosome 16 ( Keating et al, 1988 ; Martinez-Climent et al, 1995 ; Wheatley et al, 1999 ). Conversely, monosomy of chromosome 7 or 5, deletion of the long arm of chromosome 5, abnormalities of the long arm of chromosome 3, and complex abnormalities are associated with a poor prognosis. Other reported risk factors include French–American–British (FAB) morphological subtype with a poorer prognosis for M0, M5, M6 and M7 not associated with Down's syndrome (these children have an especially low relapse rate), M1 without the presence of Auer rods, M2 with white cell count > 20 × 109/l, M4 with eosinophils < 3%, higher presenting white cell count (this represents a continuous variable with relapse risk proportional to presenting white cell count), secondary leukaemias, serum lactate dehydrogenase, and age ( Creutzig et al, 1990 ; Keating et al, 1996 ; Buchner & Helnecke, 1996; Bloomfield et al, 1997 ; Stahnke et al, 1998 ). Response to chemotherapy is important, with poorer outcome for patients with a relatively slow response rate assessed on bone marrows performed at several time points following chemotherapy ( Wheatley et al, 1999 ; Priesler, 1991). For example, in the MRC AML 10 trial, overall survival rates of 54%, 44% and 21% were respectively described in patients with < 5%, 5–20% or > 20% blasts in the bone marrow after recovery from the first course of chemotherapy. These factors have been combined in a variety of ways by the major collaborative study groups as the basis for identifying separate risk groups, and to allow risk-directed therapy; as an example, analysis of outcomes in the Medical Research Council (MRC) AML10 trial was used to discriminate between three risk groups with 5-year overall survival from the start of course 2 of chemotherapy of 70%, 48% and 15%, and relapse rates of 33%, 50% and 77% respectively using cytogenetics and response to therapy. It is now clear that these risk assessments maintain prognostic significance irrespective of whether treatment comprises chemotherapy alone or combined with BMT, although several of these risk assessment tools can only be applied after a time interval from diagnosis, and full information may not be available for every patient due to incomplete data, for example due to a failure to obtain cytogenetics. Despite these advances, the prediction of relapse in individual patients remains difficult or in most cases impossible, and most relapses occur amongst standard-risk patients who form the numerically largest group. There has been great interest in the assessment of methodologies for monitoring minimal residual disease by the detection of leukaemia-specific translocations using fluorescence in-situ hybridization (FISH), polymerase chain reaction (PCR) and standard cytogenetics, in order to assess whether these tests are predictive of relapse, and can be used as a guide to modify treatment. These techniques are applicable only to a minority of patients, and the interpretation of the data requires care ( Jurlander et al, 1996 ; Perot et al, 1993 ; Freireich et al, 1992 ; Campana & Pui, 1995; Nucifora et al, 1993 ; Radich, 1995; Miura et al, 1993 ). It is apparent that persistence of the (8;21) translocation and inversion 16 at low levels occurs in patients who remain in long-term remission. Recently, detection of persistent leukaemia-related aberrant patterns of surface antigen expression has been reported as carrying prognostic significance, with relapse in 67% of patients with > 5 × 10−3 residual cells in the first remission marrow compared with relapse in 20% of patients with values below this threshold ( San-Miguel et al, 1997 ). Successful reinduction of complete remission (CR) is possible in a substantial proportion of patients who relapse, with second CR rates of 40–60% achieved with combination chemotherapy or BMT in many studies ( Brown et al, 1996 ; Petersen et al, 1993 ), although, as in general these patients were selected and were allocated to non-randomized therapies, these series are subject to selection biases. Similar results have been achieved with a variety of intensive regimens, most commonly combining an anthracycline with standard-dose cytarabine with or without other agents, high-dose cytarabine alone or combined with an anthracycline, asparaginase or etoposide or using the purine analogues fludarabine or 2 chloro-deoxyadenosine combined with high-dose cytarabine, with or without an anthracycline and/or granulocyte-colony stimulating factor ( Estey et al, 1994 , 1998; Kornblau et al, 1996 ; Fleischhack et al, 1998 ; Wells et al, 1994 ; Archimbaud et al, 1995 ; Capizzi et al, 1984 ; Kern et al, 1998 ; Whitlock et al, 1997 ; Santana et al, 1992 ; Vahdat et al, 1994 ; Huhmann et al, 1996 ). It is possible that some differences exist between the efficacies of these combinations, that there may be advantages for one anthracycline over another ( Berman et al, 1991 ; Vogler et al, 1992 ; Mandelli et al, 1991 ; Hansen et al, 1991 ; Arlin et al, 1990 ; The AML Collaborative Group, 1998), for high-dose compared to standard-dose cytarabine ( Capizzi et al, 1984 ; Herzig et al, 1983 ) and for timed sequential administration of agents ( Archimbaud et al, 1995 ; Woods et al, 1996 ), but these issues can only be addressed by multicentre randomized trials. The role of granulocyte-colony stimulating factor (G-CSF) or granulocyte-macrophage colony stimulating factor (GM-CSF) is also unclear. These growth factors undoubtedly reduce the period of neutropenia following intensive chemotherapy by several days, but this effect has not been accompanied by improved survival, and other benefits such as reduced episodes of sepsis, antibiotic usage, or duration of hospitalization have not been consistently demonstrated ( Ohno et al, 1990 ; Buchner et al, 1991 ; Heil et al, 1995 ; Estey et al, 1994 ; Uyl de Groot et al, 1998 ). There is no evidence of increased relapse with their use and G-CSF has been widely employed in combination with chemotherapy in an attempt to increase cytotoxicity by the recruitment of leukaemic blast cells into S phase of the cell cycle, by alteration of cytarabine pharmacodynamics, or reduction in the ability of blasts to repair cytarabine-induced DNA damage, although it remains unclear whether there are benefits by this approach. An equally important consideration when selecting salvage chemotherapy is the cumulative toxicity of these regimens in heavily pretreated patients, and concerns regarding toxicity are reasonable grounds for discrimination between similarly effective treatments. For example, the risk of anthracycline-related cardiomyopathy ( Lipshultz et al, 1991 ; Liesner et al, 1994 ) may favour selection of a non-anthracycline containing regimen for reinduction therapy. This is especially important in subgroups of patients with a more favourable prognosis, mainly those with a long first remission (see below). Second CR rates have been related to length of first remission, initial risk group, age, and serum LDH ( Keating et al, 1989 ; Estey et al, 1996 ; Thalhammer et al, 1996 ; Stahnke et al, 1998 ; Webb et al, 1999 ), but on multivariate analysis length of first remission appears to be particularly important ( Table I). Second CR rates for patients with a first remission of < 1 year have been only 10–30%, with values of 40–70% achieved for individuals with first remission durations of 1–2 years or > 2 years respectively ( Thalhammer et al, 1996 ; Estey et al, 1996 ; Stahnke et al, 1998 ; Webb et al, 1999 ). Indeed, response rates in patients with first CR < 1 year have been so poor that it has been suggested that these individuals should be enrolled in phase I and II studies rather than be given standard chemotherapy ( Estey et al, 1996 ). It must also be remembered that, unfortunately, 60% of all relapsed patients fall into this group. In multivariate analysis, length of first remission has proved more important that initial risk group, although the latter maintains some independent effect on both second remission rate and survival ( Webb et al, 1999 ). Risk group and length of first remission are strongly correlated with longer remissions in good-risk patients, and conversely shorter remissions in the poor-risk group. As an example, in the paediatric arm of AML 10 poor-risk patients had a uniformly bad outcome following retreatment with intensive regimens, irrespective of the length of first remission, whereas intermediate and good risk patients have shown improving survival with longer durations of first remission. Specific mention is required of patients with acute promyelocytic leukaemia (APML) who may be successfully reinduced with all-trans retinoic acid (ATRA) as a single agent, although some relapsed cases manifest ATRA resistance following prior therapy with the drug ( Tallman et al, 1997 ; Mandelli et al, 1997 ; Thomas et al, 1998 ). Although there is evidence for high remission rates in de novo patients treated with ATRA plus an anthracycline ( Mandelli et al, 1997 ), it seems appropriate that relapsed APML patients are treated with ATRA and standard institutional chemotherapy for relapsed disease. The use of alternative retinoids is under investigation to offer an alternative in patients who develop resistance to ATRA, and arsenic trioxide may also be useful in this setting ( Chen et al, 1997 ), but requires adequate in vitro and in vivo study. One particular area of difficulty is the approach for patients who relapse following allogeneic BMT in first remission. A high proportion of these individuals may be considered unfit for further intensive therapy, and morbidity and mortality are high following a second BMT with 100 d mortality rates of 25–50% ( Sanders et al, 1988 ; Mrsic et al, 1992 ; Barrett et al, 1991 ; Radich et al, 1993 ). The relapse rate following second transplant is also high and disease-free survival rates are 10–40%. Features associated with poor outcome after second transplant include an interval between procedures of < 1 year, the presence of graft-versus-host disease (GvHD) following the first transplant, resistance to reinduction chemotherapy, older age and poor performance score ( Bosi et al, 1997 ; Cullis et al, 1992 ; Barrett et al, 1991 ). The outcome has been especially poor amongst individuals given a second transplant in relapse ( Mehta et al, 1997 ; Mrsic et al, 1992 ). Accordingly it is reasonable policy that such patients receive reinduction chemotherapy initially, with re-transplantation reserved for those who achieve remission. As the intensity of the initial preparative regimen has failed to erradicate disease, and given the increased toxicity of second procedures, modifications that allow the use of less intensive conditioning ( Slavin et al, 1998 ; Giralt et al, 1997 ), followed by immunomodulatory approaches to maximize the graft-versus-leukaemia effect, seem especially attractive (see below). In most studies, patients have received two or three courses of consolidation therapy following successful reinduction, with an autologous or allogeneic stem cell transplant as the final course in selected cases. Length of first remission has a major effect on survival irrespective of whether consolidation therapy is by chemotherapy or BMT; for those with first remissions of < 1 year survival has ranged from 0 to 15% at 3 years, whereas survival rates of 30–60% have been reported amongst patients with first remissions > 1 year ( Thalhammer et al, 1996 ; Keating et al, 1988 ; Estey et al, 1996 ; Stahnke et al, 1998 ; Webb et al, 1999 ) ( Table II). These data do not yet provide a clear-cut answer as to what proportion of these individuals may have been cured, although some patients have experienced second remissions of greater duration than the first (inversion of remission). The number of courses administered is often determined by the age and performance score of the patient. The toxic mortality rate for retreatment with intensive therapies has been between 7% and 50%, with the highest figures in transplanted patients. Conversely, there is evidence that the relapse rate is lower amongst transplanted patients ( Gale et al, 1996 ) and the advantages of BMT appear greatest in patients < 30 years who had first remissions > 1 year, but caution is required in the interpretation of these data as treatments following relapses have been non-randomized. There have been of survival after chemotherapy with outcomes for BMT from and the data are to selection at present there are no data which that overall survival is improved in transplanted patients compared to those given chemotherapy alone and the increased toxicity of BMT must be any potential due to reduced in transplant have the to include alternative of stem cells from and and ( et al, 1998 ). However, allogeneic particularly using are to relatively and fit patient the of individuals with There are data to the use of stem cells compared with marrow for allogeneic although it is clear that and count recovery is more with stem cells ( et al, 1998 ). The mortality rate is and further relapse remains a major cause of treatment failure following bone marrow occurring in of cases. There is evidence for a higher relapse rate following autologous compared with allogeneic but this is by a lower treatment-related mortality for autologous procedures, especially in children the mortality rate of is low ( et al, 1991 ; Petersen et al, 1993 ; Hann et al, 1998 ). patients, relapse is less common following BMT, due to the graft-versus-leukaemia of the high risk of further relapse and the treatment-related mortality modifications to the transplant have been studies of alternative conditioning to the standard approaches with of combined with or have demonstrated the of alternative agents, but there are no data to in survival and increased intensity of therapy has been associated with high rates of morbidity and treatment-related mortality ( Brown et al, 1995 ). cell have been for a of preparative regimens in ( & 1991). the development of preparative regimens is of great interest and may be the basis of an immunomodulatory approach to transplantation that is widely There is evidence for effect in relapse of AML is less common in patients with relapse rates are higher in remission has been following the of acute there is a lower relapse risk after BMT, a higher relapse risk with combined with than with higher relapse with the more than cyclosporin, and higher relapse in ( et al, 1990 ; et al, 1990 ; et al, 1990 ; et al, 1989 ; et al, ). 2 has immunomodulatory and has been shown to in some patients with relapsed AML ( et al, 1991 ; et al, 1994 , 1996). has been shown to increase in vivo of and levels of with of in vitro cytotoxicity when employed in and regimens and is the subject of several clinical ( et al, 1997 ; et al, ) following or salvage chemotherapy and with these regimens have been with and as the most occurring in of cases. a of and cell has been proposed as an alternative agent, but there are as yet no data to useful ( et al, 1996 ; et al, 1993 ). to and a useful effect after stem cell using cyclosporin or have not been shown to be ( et al, 1992 ). with in the treatment of relapsed AML has been the subject of several ( et al, 1997 ; Van & 1995; et al, 1997 ; et al, 1997 ) with CR in around 20% of selected patients with may be more effective if used in the setting of CR and minimal residual disease, that reinduction therapy should be administered prior to in patients with has in 60% of patients given but the has been applied using both and The of employed in these cases has and there is a time of several an effect is achieved for any given a further for prior retreatment with chemotherapy. administration may the of but this remains to be ( et al, 1993 ). of this approach is very important and should be in the setting of multicentre clinical trials. As 50% of patients who achieve first CR relapse, and most of these relapses occur early, disease, the development of agents, or to is of agents may the therapeutic for example is under in phase I and II ( et al, 1998 ). ATRA is in the treatment of but there is interest in possible in other of by apoptosis of bcl-2 ( et al, 1998 ). groups are randomized studies of ATRA in combination with intensive chemotherapy in de novo and relapsed patients ( Estey et al, 1998 ). to using or to a have been described ( et al, ; et al, 1998 ) and this approach is the subject of several clinical in relapsed patients. trioxide may be similarly effective therapy to ATRA in de novo and is effective in relapsed patients who ATRA resistance ( Chen et al, 1997 ; et al, 1998 ). The of action is increased This therapy appears and include and and further study in both relapsed and de novo cases. Over 90% of relapses occur in bone marrow alone or combined with other ( Table Accordingly the of data to the of bone marrow relapse, and of outcome for disease are ( et al, 1995 ), or by in series by bone marrow of relapse include central nervous system and to the of for are not but it is clear that survival is possible following chemotherapy and to such as the and CNS without to Despite in therapy, a high proportion of patients with AML still For the who relapse 1 year of first the is very poor, with only the long-term following further intensive therapy. with longer first CR may to further intensive therapy, BMT, and the of the approaches to the intensity of the preparative and the development of immunomodulatory all for the Despite a higher rate of second relapse than is following autologous BMT is also useful in this group. of may offer for in the especially difficult setting of relapse following BMT in first remission. However, it is unclear whether further are possible in to chemotherapy, given the high intensity of