Current Controversies: Which Patients With Acute Myeloid Leukaemia Should Receive A Bone Marrow Transplantation? – An Adult Treater's View
As the majority of patients with acute myeloid leukaemia (AML) under the age of 60 years enter complete remission (CR) (Zittoun et al, 1995; Hann et al, 1997; Buchner et al, 1999), a top priority is to convert this initial response into cure. In this respect allogeneic transplantation from a matched sibling donor has been a standard approach for the last 20 years. This was followed by a similar enthusiasm for autologous transplantation after pioneering work demonstrated that cryopreserved marrow collected in remission could restore haemopoiesis after total body irradiation (TBI)-based myeloablation (Dicke et al, 1979). Several phase 2 and much registry data perpetuated the standing of transplantation as the desired approach to prevent relapse (Burnett et al, 1984; Stewart et al, 1985; Goldstone et al, 1986; Gorin et al, 1990; Lowenberg et al, 1990; Gale et al, 1996). Early comparative studies were rare, but did support the survival benefit of allogeneic bone marrow transplantation (BMT) when compared with contemporarily available chemotherapy (Appelbaum et al, 1988). The pioneers of transplant technology will have recognized that an important component of the curative mechanism was a graft-versus-leukaemia effect (GVL), and perhaps had little to do with the myeloablative treatment. The fact that the TBI dose has been shown to have an impact on relapse risk, if not on survival, suggests that both play a role (Clift et al, 1990). Early data demonstrating the relationship between relapse risk and the occurrence and severity of acute and/or chronic graft-versus-host disease (GVHD) supported the concept that the GVL effect played a role (Weiden et al, 1981). This tended to be supported by the clinical experience of T-cell depletion, which was highly effective in eliminating GVHD but was associated with an increased relapse risk. In none of these circumstances was the GVL effect as evident in AML as, for example, in chronic myeloid leukaemia (Marmont et al, 1989; Ringden et al, 1996). Further supportive information came from the limited data in identical twin transplants which demonstrated a high relapse rate and was taken to mean that the subclinical GVL effect – unrelated to GVHD – was making a contribution to allogeneic transplant cure. All these observations conditioned the development of autologous transplantation in which it was assumed that, due to the lack of a GVL mechanism, if it was to work then it was most likely to do so as a treatment of minimal disease, i.e. after consolidation had been optimized, rather than to achieve minimal residual disease. This was also partly influenced by the observation from registry data that autografts done early in CR were less effective than if done after 4–6 months, which presumably corresponded with the addition of consolidation chemotherapy (Gorin et al, 1991). As enthusiasm developed for treating all patients with either allogeneic or autologous BMT, there was also the realization that the patients who received the transplant were, because of the time delay, at a reduced risk of relapse. This ‘time-censoring’ effect called in to question whether there was indeed such a major advantage for transplantation (Gray & Wheatley, 1991; Gale et al, 1999). It was also not appreciated how small a proportion of newly diagnosed patients actually survive to transplantation (Berman et al, 1992). For all these reasons prospective trials were initiated principally to evaluate the role of autologous BMT in first remission against, or in addition to, chemotherapy. Patients were recruited at diagnosis so that the proportion of patients reaching transplant could be defined. This also enabled patients who had a matched sibling donor available to be tracked. They could be compared within the trial by ‘genetic randomization’ which compared those who had a donor with those who did not, as a surrogate for an ‘intent-to-treat’ analysis. Four major prospective trials were conducted in adults (Zittoun et al, 1995; Harousseau et al, 1997; Burnett et al, 1998, 2002; Cassileth et al, 1998) (Fig 1). They recruited 3737 patients between the late 1980s and mid-1990s. Of the 2687 eligible patients who entered remission, 711 were found to have a matched sibling donor and were intended to receive an allograft. Of the remainder, all of whom could have been randomized to autograft or chemotherapy only options, only 933 (47%) of available patients were actually randomized. This problem of compliance with the trial design is compounded by the finding that a substantial proportion of patients who were intended to (i.e. with a donor) or were randomized to an autograft did not receive it (Table I). Design of major trials assessing Allo or Auto transplantation. R, randomization; D, daunorubicin; A-Ara-C, Ara-C standard dose; Int-Ara-C, Ara-C intermediate dose; HD-Ara-C, Ara-C high dose; Ams, Amsacrine; T, Thioguanine; E, Etoposide; MACE, Amsacrine/Etoposide; Mid, Mitoxantrone; RBZ, Rubidazone; Ida, Idarubicin. This experience clearly delineated the problem in conducting trials of this nature and has some implications for what can be learned. First, for perfectly understandable clinical reasons several potential transplant candidates did not receive it. Second, the clinical results provided by these trials may not be generally applicable to the AML population as a whole but rather only to those who remain in CR for 4–6 months and are considered fit for transplantation. Third, there must be concerns about the use of intent-to-treat analyses if a high proportion of patients do not receive the transplant (Frassoni, 2000). All four trials have individual features that deserve comment. The European Organization for Research and Treatment of Cancer – Gruppo Italiano Malattie Ematologiche Maligne dell'Adulto (EORTC-GIMEMA) trial compared autograft as course 3 with further chemotherapy (Zittoun et al, 1995). A reduced relapse risk and superior disease-free survival was seen in the autograft arm but this did not translate into a survival benefit. Of note was that 56% of patients who could have been randomized were, and, of those, 74% actually received the autograft. A drawback of this trial was that the outcome for the chemotherapy arm was poorer than was reported by other groups. The major reason for the discrepancy between the superior disease-free survival but equivalent overall survival was that, of those patients who relapsed, sixty-five per cent were autografted and have had durable subsequent second remissions. In this trial allografted patients enjoyed a superior overall survival than either the autograft or chemotherapy arms. But this was not confirmed when the allografts were re-analysed on a donor versus no donor basis (Keating et al, 1998). The Groupe Ouest d'etudes de leucemies Aigues Myeloblastiques (GOELAM) from Western France compared autograft with high-dose chemotherapy, which was high-dose cytosine arabinoside (Ara-C) (Harousseau et al, 1997). No advantage was seen for the autograft arm in terms of disease-free or overall survival, but there was a trend in favour of reduced relapse risk. Good features of this trial were that a high proportion of patients that were randomized received the treatment and the autograft mortality was low. The allografted patients had a modestly better outcome when analysed on a donor versus no donor basis. The Medical Research Council (MRC) trial had a different trial design (Burnett et al, 1998). Here patients were to have received four intensive chemotherapy courses prior to transplant. This schedule turned out to have a long-term overall survival from diagnosis of 40% for the chemotherapy alone. The question was whether the addition of a transplant (allo or auto) was of benefit. Although both modalities made a major impact on relapse risk and were able to improve disease-free survival, no overall survival benefit was seen except for patients who survived beyond 2 years when the major risk period for relapse- and procedural-related deaths had passed (Burnett et al, 1998, 2002). In this large trial only 34% of patients who entered remission and could have been randomized were. In addition only 60% of patients who were allocated to autograft or had a donor available were transplanted. Finally, the US Intergroup Study followed a design which was similar to the two continental European trials in that transplant was compared with chemotherapy, in this case a single course of high-dose Ara-C (Cassileth et al, 1998). Overall, as in other trials, more patients allocated to chemotherapy received it (90%) and the compliance with the autograft allocation was quite low (54%). This was partly explained by the fact that there were delays in delivering the autograft so more patients relapsed during this period. The transplants reduced the relapse risk but the chemotherapy delivered a modestly superior overall survival. On the basis of these four trials a number of conclusions can be made. Both allogeneic and autologous transplant significantly reduced the risk of relapse, even in the context of the most efficient contemporary chemotherapy. Overall survival has not been improved either because the reduction in relapse risk did not compensate for the procedural-related mortality and the greater ability to salvage patients who relapse from chemotherapy (Table II). Of particular importance is that there was considerable compliance failure in all the trials, so that only a minority of patients in remission actually received the allocated treatment. This is an important issue in trial design. As the comparisons summarized were on an intent-to-treat basis, this defect could have under-represented the beneficial effects of transplant, i.e. reduction in relapse risk, and the disadvantages, e.g. procedural mortality and morbidity, since, as will be discussed below, patients are now recognized to be at substantially different risk of relapse once they enter remission, this balance of benefit versus risk must clearly be evaluated in the context of the risk of relapse. This assumes that the factors which predict relapse are not those which predict procedural risk. Given that transplantation has an apparently unavoidable associated mortality and morbidity, it is of importance, particularly as the results of the four large prospective trials suggest uncertainty of benefit, to establish which patients need a transplant. In order to do this several groups have attempted to define factors which predict relapse. These include cytogenetics, speed of disease response (Buchner et al, 1999; Wheatley et al, 1999), age, sex, initial white cell count and lactate dehydrogenase (LDH) levels (Buchner et al, 1999). Several studies in recent years have defined cytogenetics as the major determinant of relapse risk following chemotherapy or transplant (Ferrant et al, 1995; Gale et al, 1995; Slovak et al, 2000; Grimwade et al, 1998). The dramatic impact of the introduction of all-trans retinoic acid (ATRA) combined with chemotherapy has removed the requirement to consider transplant as first-line treatment for acute promyelocytic leukaemia (Fenaux et al, 1993; Mandelli et al, 1997; Tallman et al, 1997; Burnett et al, 1999a; Sanz et al, 1999). Similarly, but less universally accepted, is a similar understanding about t(8;21) and inv(16) disease. These patients have a high remission rate, a lower risk of relapse and a better response to reinduction therapy if relapse does occur (Bloomfield et al, 1998; Burnett et al, 1999a). It is of interest that in the MRC trial, when a donor versus no donor comparison was done in the good-risk patients with these three lesions, the relapse risk was reduced, but when the impact was looked at within the three cytogenetic entities, the impact was accounted for entirely by the t(15;17) subgroup (Burnett et al, 2002). This further suggests that allograft confers no gain for patients with t(8:21) and inv(16) disease. This has been endorsed by the experience in the current MRC AML12 trial in which no transplant has been used in this patient subgroup and equivalent survival is being achieved. There has been no consistent evidence to date to suggest there may be subgroups of these good-risk patients who do less well and may benefit from transplantation. Factors such as additional cytogenetic lesions, height of initial white cell count and slow entry to CR have not been shown consistently to have a negative impact on the favourable risk of these patients, to the extent of recommending transplantation, although more data needs to be collected. Data from the four trials, in which differing risk definitions of good risk were used, does not support the use of allo or autografts in good-risk disease (Table III). The exception is the experience from the US Intergroup study in which the outcome for transplant approaches was superior. However, it should be pointed out that the number of patients was small and, most notably, the outcome for the chemotherapy (control) arm was inferior to what should be achieved for good-risk disease. There is similarly a general consensus of which cytogenetic groups will respond poorly to existing chemotherapy. These comprise monosomy or deletion of chromosome 5, −7, 3q– complex changes and Philadelphia-positive disease. There is less agreement about whether or not to include 11q23 as poor risk. The MRC definition does not do so but it appears that a t(9;11) may respond better than patients with a t(10;11)(Grimwade et al, 1998). The MRC definition also included patients who were refractory to course 1 of therapy (i.e. > 15% blasts in the bone marrow), who subsequently entered complete remission. This group of patients have a lower chance of achieving remission and a short duration of remission with a 5-year survival of around 20% (Wheatley et al, 1999). Clearly, existing chemotherapy fails. The evidence that chemotherapy failures can be salvaged by transplantation is, however, limited. Based on information from the four trials, it was only in the US Intergroup trial that a significant benefit was shown for poor-risk patients (Table IV). As the rate of relapse in these patients is rapid there exists the possibility that several eligible patients relapsed before the transplant could take place and that only those who stayed in CR were transplanted. This particular issue was examined in the MRC 10 trial but did not seem to be a major factor (Burnett et al, 2002). In spite of the lack of consistent benefit in the trial experience and, given the lack of viable alternatives, it seems appropriate to deliver a transplant to this group of patients as soon as they are identified, thereby minimizing patients lost to relapse. The relapse rate post transplant is high and this sets the scene for developing enhanced GVL strategies, such as prophylactic donor lymphocyte infusion (DLI). Similarly, the improved reliability of matched unrelated transplantation makes this approach viable in this patient group and worthy of prospective evaluation. Patients not included in these two groupings have been regarded as standard risk with a relapse risk of about 60%. They comprise about 60% of the patients < 60 years. In this group the experience from the trials once again is inconsistent, although relapse is reduced and, in some cases, disease-free survival improved (Table V). The EORTC-GIMEMA and MRC trials showed some benefit for allograft but the admittedly smaller US Intergroup and GOELAM trials showed benefit for chemotherapy. Similar trends are apparent for autograft. Whether these differences can be explained by differences in protocol or protocol compliance is unclear, but it is worth noting that the GOELAM and US Intergroup studies had high dose Ara-C as the control arm, which appears to have achieved a particularly good results (survival was 51% and 55% respectively). The balance of data available from the trials therefore suggests that, although there is still uncertainty, this group of patients may benefit from a transplant approach, particularly if high dose Ara-C is not used. It is probable that important new prognostic factors will become recognized which may identify new patient subgroups. Of particular current interest is the demonstration of the presence of FLT3 receptor mutations in 20–25% of patients with AML (Nakao et al, 1996; Kiyoi et al, 1999; Rombouts et al, 2000; Kottaridis et al, 2001). This conveys a higher risk of relapse and may be particularly valuable in splitting the intermediate-risk group as defined by cytogenetics. A retrospective analysis of the MRC dataset in this respect showed that the relapse risk of the intermediate groups was substantially greater if a mutation was present (74%vs 48%) (Kottaridis et al, 2001). While defining new risk groups, particularly poor-risk groups, may be attractive as an exercise in suggesting groups of patients in need of improved treatment, the evidence that transplant can achieve that is, at the moment, scanty and still needs to be prospectively collected. As already discussed, those well-designed trials have in fact experienced significant compliance problems. This is highlighted in Table I. In part this was due to the belief that patients should receive consolidation chemotherapy pretransplant. This is likely to be the case for autograft because of the perceived need to undergo an in vivo purge before harvest. However, there is little evidence to sustain this belief in the context of allograft, as illustrated by a recent report from the International Bone Marrow Transplant Registry (Tallman et al, 2000). Of all the 2687 patients who entered these large trials and achieved remission, 1644 (61%) were intended for transplantation, but only 815 (50%) received it. The use therefore of an intent-to-treat analysis with a low compliance causes concern that it under-represents the benefit (antileukaemic effect) of transplant, which is even demonstrable with the low compliance achieved (Frassoni, 2000). The adverse effects (procedural mortality) will also be under represented. The assumption behind the ‘time censoring’ effect is that poorer prognosis patients will fail to receive a transplant. However, this particular point was examined in the donor-available patient group in the MRC trial (Burnett et al, 2002). It emerged that the survival of patients who had a donor, but who did not receive a transplant, was not significantly different from those with a donor who were transplanted. At least in the context of that protocol design, this sustains the argument that the role of allograft is uncertain. The other relevant issue is one of trial design. The prospect of transplantation would increase if it was intended to deliver it earlier in the treatment plan. The antileukaemic benefit may then fulfil its potential. The efficacy comparison would be between early transplant (as course 3 or even 2) with four or five courses of intensive treatment. This design, however, poses problems for including an autograft, when it is still assumed that in vivo purging is needed before stem cell collection. The comparative trial experience has involved only bone marrow as the source of stem cells. Peripheral blood is now widely adopted in the autograft setting. This has attractions for AML in which haemopoietic recovery is slow and is associated with some of the procedural-related mortality. Such data as are available at the moment suggest that, while peripheral blood harvests will enable more rapid reconstitution, there may be a higher risk of relapse (Vellenga et al, 1999; Powles et al, 2000; Reiffers et al, 2000). Data from the recent EORTC-GIMEMA trial suggest that the relapse risk is greatest when the CD34 harvest is highest (R. Willemze, personal communication). There is as yet insufficient firm data on which to advocate peripheral blood as the better source of allo-donations with respect to survival. In both these situations it is likely that more robust evidence will emerge in the near future. Improved molecular matching has improved the reliability and therefore the applicability of using unrelated donors (Busca et al, 1994). This approach might be useful for high-risk patients. However, as discussed earlier, the evidence for benefit of a matched sibling donor in this subgroup is not universal. It will be of interest to see whether the putative additional GVL effect of an unrelated donor can translate to clinical benefit. The antileukaemic effect could also be enhanced by using prophylactic DLI in patients expected to be at high risk after standard allograft. A reliable safe schedule for this approach has yet to be devised and validated. There is much current interest in non-ablative transplant approaches (Barrett, 2000; Kottaridis et al, 2000). Obtaining full donor chimaerism using a variety of less intensive preparative approaches is clearly feasible and the associated problems of graft-versus-host disease prevention are more reliably controllable. Given that the biology of the technique suggests that the graft-versus-tumour effect would take several weeks to act, most clinical activity has taken place with ‘slow’ diseases with indolent behaviour such as low-grade lymphoma. However, more recent data suggest that efficacy may be available for AML and myelodysplasia. This could potentially open a realistic prospect for older patients with AML. The MRC data suggests that standard allograft does not benefit patients over 35 years. However, the antileukaemic effect in older patients is still strong (relapse risk 43%vs 59%). This is more than compensated for by the increased mortality (26%vs 16%) of a standard allograft. Non-ablative transplantation should result in a reduced mortality but whether the antileukaemia effect is equivalent is not known and is worthy of prospective evaluation. Considerable efforts have been made to evaluate prospectively the role of autologous and allogeneic bone marrow transplantation in adults. These trials have had compliance problems but, when subjected to an intent-to-treat analysis and taking into account the risk of relapse, no consistent trend has emerged. The exception is that in all risk and age groups transplantation reduces relapse risk. This has not been translated to consistent survival advantage either because, as in good-risk cases, the associated mortality and ability to salvage chemotherapy failures outweighs any benefit. There appears to be no need to transplant cytogenetically good-risk patients if intensive chemotherapy can be delivered. High-dose Ara-C consolidation may be at least equivalent to transplant for standard-risk patients. There are, however, potential options for improving transplant effectiveness. More data is needed to evaluate any advantage that peripheral blood may confer. To this end, the recently completed EORTC-GIMEMA AML10 trial which randomly compared blood and bone marrow autografts will be pivotal. The early delivery of treatment may increase overall effectiveness – particularly for patients at highest risk of relapse, in whom the addition of antileukaemic GVL effect perhaps using prophylactic DLI justifies evaluation, as does the potential contribution of non-ablative transplantation. As there is no subgroup of patients who clearly benefit from a transplant in first remission and bearing in mind the evidence that in this age group there is gradual improvement with chemotherapy, it remains important that transplantation is continuously evaluated in the context of a clinical trial.
Read more
70