Bone marrow transplantation versus chemotherapy in the treatment of very high–risk childhood acute lymphoblastic leukemia in first remission: results from Medical Research Council UKALL X and XI

Despite steady improvements in the management of acute lymphoblastic leukemia (ALL) in children, approximately 20% to 30% of patients relapse.1-4 While many of these relapses occur in so-called “standard risk” children, approximately 10% of these patients have clinical and biological features at diagnosis that identify them as at very high (VH) risk of relapse. There is no universally agreed definition of VH-risk patients, but such criteria have, in various studies, included a high initial white blood cell (WBC) count greater than 100 × 10⁹/L,5,6certain cytogenetic abnormalities (eg, Ph⁺ALL),7,8 and a slow response to induction chemotherapy2,9,10 or to pretreatment with steroids.2 

The treatment of this VH-risk group of patients has been intensified over the past decade either by chemotherapy alone or by high-dose chemotherapy and/or radiotherapy with allogeneic bone marrow transplantation (BMT). The benefit of transplantation in the first remission of ALL remains unclear, and there are few large studies addressing this important question. Until recently the impact of transplantation has been restricted by the limited availability of sibling donors, but as the availability of unrelated donors has increased,11 and the procedure of matched unrelated donor (MUD) transplantation has become safer,12 this form of treatment has become accessible for more children.

In the last 2 United Kingdom (UK) national protocols for children with ALL, Medical Research Council (MRC) UKALLX and UKALL XI, we attempted to define possible indications for transplantation in first remission and to compare the outcome of children who receive transplantations to those treated with chemotherapy alone. We prospectively collected information about human leukocyte antigen (HLA) typing in the patients and compared the outcome by treatment given and by the availability of an HLA-matched sibling donor, a “biologically randomized” control group.

The VH-risk patients in this study came from 2 consecutive UK MRC trials conducted during an 11-year period: MRC UKALL X (January 1985 to September 1990) and UKALL XI (October 1990 to March 1997). The trials were open to all children aged 1 to 15 years with ALL except those with surface membrane immunoglobulin-positive (Ig⁺) ALL (FAB L3 subtype), for whom there was a specific protocol. A number of infants were entered into UKALL X but not into UKALL XI, and they were all excluded from this analysis because they have a biologically different disease.13 

The children designated as VH risk were those with a WBC count greater than 100 × 10⁹/L at diagnosis in UKALL X and during the first 14 months of UKALL XI. The selection criteria for VH-risk patients was then refined in light of the analysis of prognostic factors in UKALL X, which led to the definition of a hazard ratio score incorporating gender and age as well as the WBC count (Figure 1).6 For the remaining period in UKALL XI, the VH-risk patients were identified by this hazard ratio score (a selection that resulted in relatively fewer girls and more older children being identified as high risk) if they had Ph⁺ or near-haploid ALL or if they had not achieved complete remission within 4 weeks of treatment.

The basic treatment was very similar in both trials, and both protocols have been described elsewhere.3,14 Induction treatment for all patients comprised the following: 1.5 mg/m² intravenous (IV) vincristine weekly; 6000 U/m² subcutaneous (SC)Erwinia L-asparaginase for 9 doses; 40 mg/m² oral (PO) prednisolone for 4 weeks; intrathecal methotrexate (MTX) for 2 doses, the dose derived according to age; and from 1985 to 1992, 45 mg/m² IV daunorubicin for 2 doses. After March 1992 in MRC UKALL XI, daunorubicin was omitted from induction because of concerns regarding possible long-term anthracycline toxicity.15Intensification blocks were given at weeks 5 and 20 to all VH-risk patients in both trials. These blocks consisted of a 5-day course of 200 mg/m² IV cytarabine, 100 mg/m² IV etoposide, 80 mg/m² PO thioguanine, and 40 mg/m² PO prednisolone with 45 mg/m² IV daunorubicin for 2 doses, 1.5 mg/m² vincristine for 1 dose, and intrathecal MTX for 1 dose.

From March 1992 all patients in UKALL XI had a randomization to receive, or not receive, a third intensification block, which was given over 8 weeks, starting at week 35. This block consisted of 10 mg/m² PO dexamethasone daily for 10 days, then reduced to zero over the next 4 days; 1.5 mg/m² IV vincristine on day 1 of weeks 35 to 38; 6000 U/m² SC L-asparaginase for 9 doses, starting on day 4 of the block; 600 mg/m² IV cyclophosphamide on day 1 of weeks 39 and 41; 75 mg/m² IV of SC cytarabine on days 1 to 4 of weeks 39 to 42; 60 mg/m²PO thioguanine daily during weeks 39 to 42 inclusive; and intrathecal MTX for 1 dose given at weeks 35 and 39.

In addition, in UKALL X, central nervous system (CNS)–directed treatment for VH-risk patients included 24 Gy cranial irradiation in 12 fractions with intrathecal MTX for 3 weekly dosages. In UKALL XI, children with a WBC count of more than 50 × 10⁹/L at diagnosis were randomized between 24 Gy of cranial irradiation in 15 fractions versus high-dose MTX with continuing intrathecal MTX every 12 weeks for up to 2 years of treatment. Cranial irradiation was deferred until the age of 2 years in younger children. High-dose MTX was given at 8 gm/m² IV for children aged 1 to 4 years and 6 gm/m² IV for those aged 4 years during weeks 9, 11, and 13. All patients received intrathecal MTX injections for 3 doses between induction and maintenance. Continuing or maintenance therapy for patients in both trials comprised 75 mg/m² PO 6 mercaptopurine daily; 20 mg/m² PO MTX weekly; 1.5 mg/m² IV vincristine monthly; and 40 mg/m² PO prednisolone for 5 days monthly, with an additional intrathecal MTX every 12 weeks for those children who had not received cranial irradiation. The total length of treatment was 2 years.

Patients identified as VH risk using the criteria described above and who had an HLA-compatible sibling were eligible to receive a transplantation in first remission. In these patients total body irradiation (TBI) and cyclophosphamide replaced cranial radiotherapy or high-dose IV MTX. The decision to proceed with HLA tissue typing and transplantation was taken by individual clinicians. Some patients received autologous BMT (ABMT) or MUD transplantation, although this procedure was not specified in the protocol. Information on the decisions regarding tissue typing and whether or not to receive a transplantation was obtained on all patients in a questionnaire sent to their clinicians. Both protocols received approval from institutional research ethics committees, and informed consent for all patients was obtained according to the guidelines of each center. The UKALL X patients have been followed to October 1998 and the UKALL XI patients to October 1999; the median follow-up was 8 years (range, 2.5 to 14 years).

Groups of patients were compared for differences in initial characteristics by means of chi-square tests. Differences in outcome by the prognostic factors, diagnosis WBC count, Ph chromosome status, ploidy, as well as by patient groups were analyzed by the log-rank method,16 using stratification to adjust for other factors when necessary. Comparisons of patients who received transplantations versus those who received chemotherapy were also adjusted for the time to transplantation by the Mantel-Byar method.17Using this method, patients who were designated to receive transplantations were categorized in the chemotherapy group in the numbers at risk until the time of transplantation, when they were then deemed to enter the transplantation group. To give a graphical display of the relative outcome of the chemotherapy and transplantation groups, descriptive survival curves were plotted after adjustment for the time to transplantation and for prognostic factors by using the adjusted log-rank observed minus the expected and its variance in each separate year.18 The prognostic factors used were those found to be independently predictive (diagnosis WBC count; Ph status of positive, negative, or unknown; and ploidy groups). These curves provide a pictorial representation of the differences between patient groups, but the survivals for one group cannot be compared with standard Kaplan-Meier survivals because they depend both on the results for the comparison group and on the distribution of prognostic factors in the 2 groups.

Six patients were treated by ABMT in first remission, and they were included with the patients receiving chemotherapy. Exclusion of these patients did not alter the overall results of the study. Only 25 patients received a MUD transplantation, and this number was too few to analyze as a separate group. In the comparisons, treatment received by these MUD transplant patients were analyzed in the same group as those receiving transplantations from a related donor.

Because of the inherent biases in comparing groups of patients by treatment received, comparisons were also made among those patients who were HLA-typed and also between patients with a matched sibling donor and those with no match. This process can be thought of as “genetic randomization” with an intention-to-treat analysis.19 Because some of the “no match” group did not receive chemotherapy alone, but received a transplantation from an unrelated donor, these analyses were repeated with the data for these patients censored at the time of transplantation. The primary end-points were event-free survival (time to any event) and overall survival. Time to relapse, censoring at death in remission, and time to death in remission were also analyzed.

For this study, 1586 patients, aged 1 to 15 years, were recruited by MRC UKALL X. Of these patients, 198 were classified as VH risk and were eligible for transplantation in first remission because they had a WBC count of 100 × 10⁹/L or above at diagnosis. Of the patients eligible for transplantation, 10 were excluded from these analyses: 8 failed to achieve first remission, 1 relapsed, and 1 died within 15 weeks of diagnosis. A further 2090 patients were recruited by UKALL XI, and 275 were classified as VH risk. We excluded 11 of these patients: 5 failed to achieve remission, 2 received a transplantation before achieving remission, 3 relapsed, and one died within 15 weeks. Of the remaining 264 patients, 31 entered UKALL XI before March 1992 and were classified as VH risk because of an initial WBC count greater than 100 × 10⁹/L. The other 233 patients entered the study after March 1992, when the high-risk selection criteria changed. Of these patients, 177 patients had a high hazard score (Figure 1), 22 had Ph⁺ ALL, 4 had near-haploid ALL, and 54 were late remitters (after 4 weeks), with 18 patients having 2 adverse features and 3 patients having 3 adverse features.

Thus, there were a total of 3676 patients aged 1 to 15 years registered into the UKALL X and XI MRC ALL trials. Of these patients, 473 patients (13%) were in the VH-risk category (as defined in each protocol), which made them eligible for transplantation in the first remission, and 452 patients (12% of the original ALL cohort entered into these 2 studies) achieved the stable remission needed to make transplantation feasible.

Of the 452 patients in stable remission, 92 patients had no siblings; 62 patients were not HLA-typed for other reasons; and in 12 patients, it was not known if tissue typing was done or not. A compatible sibling was found for 99 children (35%) of the 286 who were typed. Only 76 children (77%) with a donor received an MSD transplantation; 19 patients chose not to undergo transplantation; and 3 patients relapsed while waiting for their transplantation. Of these latter 22 patients, 8 received transplantations in second remission. In one case it was decided that the donor was too young, and this patient received a MUD transplant (Figure 2).

Comparison of the patients not HLA-typed by choice with those who were typed (Table 1) shows that typing was more likely to be performed in older children (P = .01), in those with a higher leukocyte count (P = .003), and in those with T-cell ALL (P = .003). Amongst those who were HLA-typed, there were no significant differences between patients with and without a donor except for a slightly higher chance of donor availability for younger children (P = .02).

A comparison of patients by treatment received (Table2) shows no significant differences in presenting features between those children who proceeded to transplantation and those who had chemotherapy or an ABMT, but more children with Ph⁺ ALL (P < .000 05) or with pseudodiploidy (P = .04) received a MUD transplant.

All 452 patients received standard induction therapy, although 224 patients (50%) had no daunorubicin. The first intensification block (IB) was received by 441 patients (98%), and 101 patients proceeded to the transplantation program. Of the 351 patients who proceeded with chemotherapy, 343 patients (98%) had a second IB (week 20), and 78 patients (22%) had a third IB (week 35). Of the 8 patients who had CNS disease at diagnosis, 4 patients received craniospinal radiotherapy with continuing chemotherapy and no second IB; the other 4 patients received a transplantation. The CNS-directed treatment was cranial radiotherapy at 24 Gy in 203 patients (58%), high-dose MTX in 88 patients (25%), and craniospinal radiotherapy in 4 patients with CNS disease at diagnosis.

Of those patients receiving a transplantation, 76 patients received a MSD transplantation, and 25 received a MUD transplantation. The transplantations were performed in 18 different centers, but MUD transplantations were undertaken in only 7 centers. The median time to MSD transplantation was 4 months, with all but 1 transplantation being done between 2 months and 1 year following diagnosis. The median time to MUD transplant was 7 months, with all but 1 transplantation done between 3 and 14 months. However, 1 MSD transplantation and 1 MUD transplantation were each done at 19 months, but the reason for the lateness of these transplantations was not recorded. The median follow-up after transplantation was 5 years.

Most patients (72 of 76 MSD patients and 23 of 25 MUD patients) received TBI, usually with cyclophosphamide. Graft-versus-host disease (GVHD) prevention measures for patients receiving MSD transplantations included cyclosporin (32%), cyclosporin plus MTX (39%), or T-cell depletion (18%). For patients receiving UDBMTs, the prevention measures included cyclosporin plus T-cell depletion (48%), cyclosporin plus MTX (24%), or cyclosporin alone (12%). Patients whose first transplantation failed did not undergo a second transplantation.

Table 3 shows the outcome of the chemotherapy group (including the patients who received ABMT) versus the transplantation group (MSD or MUD). Table4 shows the outcome by HLA-related donor availability comparing matched sibling donor versus no matched sibling donor.

There were significantly more deaths in remission in the transplantation group (transplantation, 18%, vs chemotherapy, 3%;P < .001) and in the matched sibling group (match, 16%, vs no match, 3%; P < .001). However, whereas the reduction in the odds of relapse seen in the comparison of the transplantation versus chemotherapy groups was 37% (log-rank,P < .01 after adjustment), the reduction in the odds of relapse in the match versus no match comparison was only 3% (not significant).

There was no significant difference in EFS or in overall survival in the comparisons of transplantation versus chemotherapy groups or in the patients with a matched sibling versus no matched sibling. The unadjusted estimated effect of transplantation on the risk of an event (relapse or death) was a reduction in the odds of 16%, while the effect in the matched sibling group was a reduction of only 3%. Adjustment for time to transplantation and prognostic factors decreased the estimated benefit of transplant from 16% to 9% in the odds of an event, and adjustment for prognostic factors suggested an adverse outcome in the matched sibling group, with a 37% increase in the odds of an event. Censoring at transplantation for the 18 patients in the no-match sibling group who had MUD transplantations made no material difference to the results, giving an increase in the odds of an event, after adjustment, of 30% (nonsignificant) in the matched sibling group.

Figures 3 to5 depict adjusted and unadjusted descriptive curves comparing EFS in the groups of patients. At 10 years after transplantation, the transplantation group had an 8.5% (95% confidence interval [CI], −3.5% to 20.5%) greater EFS than the chemotherapy-treated group (transplantation, 46.8%, vs chemotherapy, 38.3%) (Figure 3). Adjustment for time to transplantation and prognostic factors resulted in a smaller EFS difference of 6.0% (95% CI, −10.5% to 22.5%; transplantation, 45.3%, vs chemotherapy, 39.3%) (Figure 4). However, in the donor versus no donor comparison, adjusted for prognostic factors, the difference in EFS was 10.7% (95% CI, −2.6% to 24.0%) in favor of the group without a donor (donor, 39.7%, vs no donor, 50.4%) (Figure5).

Comparisons for survival show similar, but smaller, differences. The adjusted comparisons of survival give an increase of 3.5% (95% CI, −12.2% to 19.2%) at 10 years for the transplantation group (56.9%) versus the chemotherapy group (53.4%). The adjusted comparisons also give an increase of 7.2% (95% CI, −5.9% to 20.3%) at 10 years for those patients without a donor (61.2%) versus those patients with a donor (54.0%).

The overall numbers are inadequate for reliable subgroup analyses, but those that were done did not indicate any particular group who might benefit from transplantation. Patients with a higher WBC count did not show a trend for a greater effect from transplantation. This was not affected by the inclusion or not of patients with Ph⁺ disease or hypodiploidy, or if the analyses were confined to T or B lineage only. The following odds ratios (ORs) for EFS (with the Ph⁺ patients excluded) compares donor versus no donor patients within the following WBC groups: WBC count of less than 100×10⁹/L, 1.93 (95% CI, 0.88 to 4.20); WBC count of 100 to 149×10⁹/L, 1.18 (95% CI, 0.55 to 2.57); WBC count of 150 to 249×10⁹/L, 0.80 (95% CI, 0.39 to 1.64); WBC count of at least 250×10⁹/L, 0.85 (95% CI, 0.47 to 1.52) (test for trend, P = .09).

There were 15 deaths (20%) during remission in the patients who received a MSD transplantation: 4 deaths were due to infection; 2, acute GVHD plus infection; 2, chronic GVHD; 2, cardiotoxicity plus infection; and 1 each, cerebral hemorrhage plus infection, graft failure plus infection, graft failure plus hemorrhage, brain tumor, and a train accident. In the MUD transplantation group, there were 3 deaths (12%) in remission: 1 each from infection, acute GVHD plus infection, and chronic GVHD plus infection. Of the patients receiving chemotherapy, 9 patients (3%) died in remission: 7 died from infection and 2 from second malignancies (astrocytoma and lymphoma). Of the 6 patients receiving ABMT, 2 died: one from cytomegalovirus (CMV) pneumonitis and the other from chronic respiratory failure after developing restrictive lung disease.

In the study, 26 patients were identified as having Ph⁺ ALL. Of these patients, 12 (46%) were treated with chemotherapy; 2 patients remain alive in remission from their disease, but one of these patients had developed acute myeloid leukemia and had been treated with a MUD transplantation. Of the 14 patients (54%) who were treated by transplantation (3 MSD and 11 MUD), 8 patients remain alive and in remission.

Analysis of factors influencing prognosis showed that patients with a higher WBC count had a worse EFS (P = .01), as did those with Ph⁺ disease (P = .08) and boys compared with girls (P = .02). Age did not have a significant affect in this high-risk group. There was also significant heterogeneity between ploidy groups (P = .005), with the low hyperdiploid and hypodiploid groups having the worst prognosis, and the high hyperdiploid group having the best. These factors influencing prognosis were statistically significant in the chemotherapy-treated group. In the smaller transplantation group, these variables affected prognosis in the same direction, but because of the smaller numbers, they were not statistically significant, with the exception of the WBC count, which showed no clear trend. Sex was no longer significant after stratification by the WBC count, but the cytogenetic factors were independently significant.

From the chemotherapy group, 91 patients received a transplantation after relapse (17 autologous, 25 MSD, and 49 MUD transplantations). There was no significant difference in survival after relapse by first-remission treatment. Survival at 5 years after relapse was 24.7% (95% CI, 17.8% to 31.6%) for the chemotherapy group and 30.8% (95% CI, 14.1% to 47.5%) for the transplantation group.

In this large prospective study we compared the outcome of VH-risk ALL in children (13% of a cohort) both by treatment received (transplantation or chemotherapy) and by the availability of an HLA-compatible sibling donor. We extended the observations made in our previous study20 with longer follow-up and with the addition of patients treated in a subsequent but similar trial. The results showed a marginal possible benefit of 4.6% in adjusted EFS at 10 years for the patients who received transplantations in first remission compared to those treated with chemotherapy. BMT was associated with significantly fewer relapses (31%) than chemotherapy (55%), but there were many more deaths in remission, 18% compared to 3%, respectively. When the outcome between patients with and without an HLA MSD was compared, patients with an HLA donor had an adjusted 10-year EFS that was 10.7% lower than those with no donor. This comparison of outcome between the 2 biologically randomized and therefore unbiased groups suggests that there are problems of selection bias regarding the allocations to transplantation and chemotherapy, some of which can be allowed for, but some of which cannot. This study, using the best available comparative groups, those with and without an HLA-matched donor, suggests that for the majority of VH-risk childhood ALL patients, transplantation in first remission does not improve EFS.

We are aware of no other published studies that have attempted to control selection biases with an intention to treat analysis comparing patients with and without a histocompatible sibling or using adjusted EFS analysis. Most reports of transplantation in pediatric ALL patients in first remission describe the outcome of children receiving transplantation without any comparisons with children receiving chemotherapy.21-23 Others studies compared their results with matched chemotherapy-treated controls.24-26 The Italian cooperative group26 recently compared the outcome of 30 children with high-risk ALL patients who received transplantations within the first year after diagnosis to 130 matched controls. The patients were selected on the basis of leukocyte count, cytogenetics, and treatment response. The results suggested a benefit of transplantation, with a 4-year DFS of 58.5% in the transplantation group compared with 47.7% in the chemotherapy group. The high risk of early failure in the transplantation group was outweighed by the lower risk of relapse one year after transplantation. A recent Nordic study compared the outcome of 22 VH-risk ALL patients treated with first-remission transplantation with 44 matched control patients treated with chemotherapy25; the transplantation group had a DFS of 73% at 10 years compared with 50% in the control group. Both of these studies were relatively small and used selected controls rather than “biologically randomized” patients.

The critical evaluation of the place of transplantation in first-remission ALL is complicated by the lack of consensus about the identification of VH-risk patients and by the improving results with intensive chemotherapy. The Rome27 criteria and a recent re-evaluation by Smith et al28 both define higher risk patients as representing 30% to 40% of all cases. These high-risk patients are increasingly successfully treated with intensified chemotherapy.29-32 The Children's Cancer Group recently reported a 5-year EFS of 75% in a group of high-risk ALL patients who had initially responded slowly to treatment, when they were treated with intensified chemotherapy. This compared to a 5-year EFS of 55% in the patients receiving standard treatment.33 Transplantation is not justifiable for most of this large group of patients.

Our study was restricted to VH-risk patients. Using a WBC count in excess of 100 × 10⁹/L, as in the first 6 years of our study, 12% of patients were VH risk, while the hazard score using age, WBC count, and gender identified 11% of patients (mainly boys) who were VH risk. This compares with the 5% to 8% of patients now selected to be at VH risk by the Italian Group26 and the 10% of patients in the Berlin-Frankfurt-Munster (BFM) studies identified by a poor response to steroids and intrathecal MTX.9 It is likely that these various groups of VH-risk patients have similar although not identical clinical features. A more logical approach for future studies would be to select patients according to biological features, for example, a study of Ph⁺ ALL patients. This is a group at high risk of treatment failure, whether treated with chemotherapy or transplantation,34-39 although a combined study by the BFM and the Italian Group suggests that an early response to steroids in Ph⁺ patients may indicate a subgroup with a more favorable outcome.40 

We found that the transplantation group was associated with a treatment-related mortality of 18% compared with 3% for the chemotherapy group. This transplantation-related mortality (TRM) is in excess of the 10% to 12% in most recently published reports.21-23,25,26 The allogeneic transplantations were performed at 18 different centers compared with the fewer and perhaps more experienced 7 centers that performed MUD transplantations, which were associated with a lower TRM of only 12%. There is some evidence that centers performing more transplantations have a lower TRM,41 and it is not clear whether this decentralization was a factor in our study. A reduction in TRM could improve the outcome for VH-risk patients receiving transplantations in first remission.

The late effects associated with chemotherapy or transplantation treatment are predictable, particularly those that follow TBI given as conditioning prior to transplantation. We have not shown late effects data in this paper because they have been well described elsewhere.42 43 The objectives of this study did not include routine collection of this data and is therefore incomplete. The majority of children who have had TBI can be anticipated to be at risk of multiple endocrine deficiencies, infertility, bony dysplasias, cataracts, and second malignancies. Unless there is a clear advantage of transplantation over chemotherapy, this predictable TRM means that chemotherapy should be the preferred treatment option.

In conclusion we have not been able to show a significant advantage of first-remission transplantation in VH-risk ALL patients. There is a possibility that some benefit might become apparent with a reduction in TRM. In view of the heterogeneity of VH-risk ALL, future studies should probably concentrate on well-defined subsets of patients, such as those with Ph⁺ ALL. Such studies would require large-scale international collaboration.

Members of the MRC Working Party on Childhood Leukaemia during UKALL X and UKALL XI: O. B. Eden (chairman), C. C. Bailey, P. R. H. Barbor, A. Barrett, C. Barton, V. Broadbent, J. M. Chessells, S. I. Dempsey, P. J. Darbyshire, J. Durrant, P. Emerson, D. I. K. Evans, J. J. Fennelly, D. A. G. Galton, B. Gibson, R. Gray, I. M. Hann, R. M. Hardisty, M. Hewitt, F. G. H. Hill, J. Kernahan, D. J. King, J. Kohler, I. J. Lewis, J. S. Lilleyman, M. Madden, J. R. Mann, J. Martin, T. J. McElwain, S. T. Mellor, P. H. Morris Jones, A. Oakhill, J. Peto, M. Radford, J. K. H. Rees, S. M. Richards, R. F. Stevens, G. P. Summerfield, E. N. Thompson, H. Wallace, D. Webb, K. Wheatley, and A. Will.