Over the past two decades the outcome of therapy for patients with AML has improved considerably and steadily. Between 65 and 85% of adult patients under 60 years of age enter complete remission (CR) with current chemotherapy.1,2 Thereafter, there are three major therapeutic options: further courses of chemotherapy, autografting and allografting. The results of recent studies tend to confirm that results of allografting have consistently improved3 and that they are never inferior4,5 to other options. However, any comparison of allograft results with chemotherapy or autograft is complicated by selection processes5 which inevitably take place in the absence of true randomisation and evaluation on an intention-to-treat basis. As a consequence, it is still debated whether it is appropriate or not to offer BMT automatically to all eligible patients.
In a number of large studies autografting was never inferior to chemotherapy4,5,6,7 when patients were analysed on an intention-to-treat basis. However, in spite of a continuous exchange of data (meetings, publications etc) there is no general consensus as to the best post-remission therapy.4,5,6,7,8,9,10,11,12 This situation is exemplified by the MRC AML 10 study (560 non-randomised out of 1131)6 where there was an appreciable loss of patients from the study due to patient or physician decision.
There is no doubt that randomised studies generate relevant information and are the cornerstones of major progress. The EORTC-GIMEMA AML 84 and MRC AML 10 trials5,6 have both provided very important information. First, the 40% overall survival in the MRC AML 10 trial6 represents a major achievement in the treatment of AML. Unlike the EORTC-GIMEMA AML 8 study,4 the MRC study included four courses of chemotherapy before autograft and this may explain the brilliant results in patients who stopped therapy. However, should the chemotherapy schedules and doses change, the same results, particularly in the good risk category, might not be duplicated.1,4,5,6,7,8,9,10,11,12 Second, the MRC study5 has shown that patients who have an HLA-identical sibling donor fare slightly better than other patients. However, patients who have an HLA-identical sibling and do not proceed to allo-transplant do worse than patients lacking a donor: this shows that patients who are allografted are a selected cohort.5 Third, in both studies4,6 autograft was superior to chemotherapy for LFS but not for overall survival (MRC (57 vs 40% P = 0.2)6 and EORTC-GIMEMA (56 vs 46% P = 0.43)).4 Nevertheless, in spite of the statistics the perception of the superiority of autografting was reflected in the fact that in the subsequent EORCT-GIMEMA study (AML-10), all patients were assigned to receive an autograft with only two exceptions: (1) if an HLA-identical sibling was available, or (2) in the case of acute promyelocytic leukaemia. In the MRC study, on an intention-to-treat analysis the number of relapses was substantially lower in the autologous group than in the group assigned to no further treatment (64/190 (37%) vs 101/191 (58%), P = 0.0007), resulting in superior disease-free survival at 7 years (53 vs 40%, P = 0.04). This was in spite of the fact that only two thirds (126/190) of patients assigned to the autograft arm did in fact receive the planned treatment.
In contrast to the two previous studies,4,6 two other studies do not show superiority of transplant over chemotherapy. Indeed, the GOELAM study8 in which there was a good concordance between assignment and number of patients actually receiving the treatment, did not show any superiority of autograft over chemotherapy. The US Intergroup results9 showed an advantage of chemotherapy vs autograft. However, in the latter study, only 54% of patients received the planned autograft and 21 patients out of 116 randomised, who had refused the assigned treatment, were included in the analysis.
It should be recognised that most of the reported studies face major problems. First, the low accrual of patients to transplant (autograft). This is due primarily to failure to reach randomisation4,6 due to various reasons preventing patients from undergoing the transplant.10,11 This helps to explain how the value of transplant could be missed in the structured studies. Second, it is impossible to know what happened between the time of randomisation and treatment delivery to cause bias. Some of it relates to design of the trials. For instance, the fact that patients were planned to have four courses of chemotherapy6 provides a further opportunity for patients to become unfit for the autograft.
The main reason for publishing the results of clinical trials in major scientific journals is the hope that physicians use such information to guide their decisions. Therefore, if the emerging conclusion of the above studies is that there is no major difference among treatments, why choose the most demanding option? However, I would like to argue that interpreting the results only on an intention-to-treat basis is not necessarily appropriate for the decision-making process. No doubt this method identifies correctly the true impact of a given treatment on the disease and avoids biases related to the selection process. At the same time, the risk that the conclusions of these studies may convey misleading messages is high. If one has to advise a patient with AML in first complete remission who is eligible for an allogeneic transplant, then only the figures for patients who actually underwent allogeneic transplants should be considered, not the intention-to-treat figures. The same holds true for a patient with AML in first complete remission whose marrow or mobilized peripheral blood cells have been successfully harvested (why should we consider patients in whom it was not possible to obtain a suitable harvest?). Interestingly, results of patients who actually underwent transplantation were not reported in the results of the trials4,6,8,9 to comply with the conventional wisdom arising from a statistics-oriented orthodoxy. The patient in front of you asking advice about her/his post-remission therapy does not represent the whole population. She/he is selected by the fact that she/he is in remission, is in good general condition and meets the criteria for allografting or autografting. Consequently, we should only discuss the figures for patients who actually underwent the procedure. These are the only data providing the correct information on the probability that this particular patient will become a long-term disease-free survivor. I realise that this viewpoint embodies the potential risk of denying the value of prospective randomised trials. I do not intend to challenge the concept that a new treatment modality should be evaluated through a prospective randomised trial. However, for an individual patient, not included in a prospective randomised study, actual data may be a more appropriate reference. Finally, since major differences in the outcome for the same procedure do exist among various centres,13 when advising a patient with AML in first remission about the therapeutic options, the results of chemotherapy, allografting or autografting in the particular centre where the post-remission treatment is planned, should also be carefully evaluated before making the final choice.
|
