Acute myeloid leukaemia Bone marrow transplantation for ... - Nature

4 downloads 63 Views 128KB Size Report
RR and absolute risk reduction for treatment effect across all trials were calculated, using DerSimonian and Laird ran- dom effects model.20 Analyses were ...
Bone Marrow Transplantation (2002) 29, 843–852  2002 Nature Publishing Group All rights reserved 0268–3369/02 $25.00 www.nature.com/bmt

Acute myeloid leukaemia Bone marrow transplantation for paediatric AML in first remission: a systematic review and meta-analysis M Bleakley1, L Lau1, PJ Shaw1,2 and A Kaufman2 1

Oncology Unit, The Children’s Hospital at Westmead, Sydney, NSW, Australia; and 2Department of Paediatrics and Child Health, University of Sydney, Australia

Summary: For children with AML in CR1, the major consolidation therapies are BMT, ABMT and intensive chemotherapy. The relative effectiveness of these strategies is still debated. We conducted a systematic review and meta-analysis of trials to determine the effectiveness of BMT and ABMT in CR1 in paediatric AML. Eligible studies enrolled patients ⬍21 years from 1985 to 2000 with AML in CR1. Two groups of studies were identified: (1) Those comparing the outcome of patients with and without a histocompatible family donor; and (2) Randomised controlled trials (RCT) comparing ABMT with non-myeloablative chemotherapy. The relative risk statistic was calculated for outcomes of interest in each trial. If there was no excessive heterogeneity between trials the results were pooled, and an overall relative risk and risk difference for treatment effect across trials were calculated. Results of the analysis showed that allocation to BMT reduced risk of relapse and improved disease-free and overall survival. For ABMT, heterogeneity of effect between RCTs prevented pooling of results. In conclusion, BMT from a histocompatible family donor improves patient outcome. Data are insufficient to determine whether this is true for all subgroups of AML, and whether ABMT is superior to non-myeloablative chemotherapy. An individual patient data meta-analysis is required to further evaluate the available data. Bone Marrow Transplantation (2002) 29, 843–852. DOI: 10.1038/sj/bmt/1703528 Keywords: ABMT; AML; pediatric; systematic review; meta-analysis

The prognosis of paediatric AML has improved over the last four decades. Currently more than 80% of children diagnosed with AML achieve remission. In half of these patients remission is sustained and long-term cure achieved.1 For patients achieving remission, BMT has been Correspondence: M Bleakley, Oncology Unit, The Children’s Hospital at Westmead, Sydney, NSW 2145, Australia Received 21 May 2001; accepted 24 January 2002

considered the treatment of choice if a matched sibling donor (MSD) is available.2 The advantages of BMT in maintaining remission are balanced by mortality, related mainly to the complications of infection and GVHD, and the long-term toxicity of treatment. The overall survival advantage for BMT is not clear, particularly as chemotherapy becomes more intensive. It is unclear if all subgroups of AML benefit equally from BMT. BMT in CR1 may not be the best therapy for a good prognosis AML patient with a reduced risk of relapse and an improved salvage rate following relapse. There have been no randomised clinical trials (RCTs) comparing BMT with other forms of post-remission therapy in children or adults. BMT is widely believed to be the most effective anti-leukaemic therapy available in AML. Consequently RCTs have not been believed to be ethical. ‘Mendelian’ or ‘biologic’ randomisation has been used to compare BMT with other therapies. The outcome of all patients with a matched sibling donor (MSD) is compared to those without, in trials where BMT is recommended for all patients with a MSD. Such Mendelian randomisation in a ‘donor vs no donor’ study is regarded as an unbiased method of comparing intervention effects in the absence of true RCTs.3,4 The major treatment options for AML patients with no MSD, are ABMT, further intensive post-remission nonmyeloablative chemotherapy or no further therapy.1,2,5–8 A number of studies have randomised children to receive ABMT or chemotherapy,9–12 and others have compared ABMT with no further therapy.13–15 AML is a heterogenous disease with a number of morphologic and cytogenetic subtypes. These have been associated with better or worse prognosis within AML. In the United Kingdom Medical Research Council’s 10th AML trial (MRC AML 10), patients classified into good, standard and poor risk subgroups (by cytogenetics and responsiveness to initial chemotherapy) had a risk of relapse of 33%, 51% and 78%.16 The Berlin–Frankfurt–Munster (BFM) group have used morphological group and day 15 marrow to determine a standard risk and high risk group, with an event-free survival (EFS) of 68% for the standard risk patients and 33% for high risk patients in BFM 87.17,18 Stratification of patients into such risk groups and evaluation of response to different treatments (eg BMT vs ABMT vs chemotherapy) within these risk groups would

Bone marrow transplantation for paediatric AML M Bleakley et al

844

assist in identifying the appropriate treatment strategy for an individual patient. In this systematic review, the effect of BMT and ABMT on relapse and survival in paediatric AML patients in CR1 is described. The first aim of this review was to determine the effectiveness of BMT and ABMT in preventing relapse and improving survival beyond 3 years after first remission in paediatric AML. Secondly, we wished to determine the effectiveness of BMT and ABMT in preventing relapse and improving survival beyond 3 years after first remission in good risk groups, standard and poor risk groups of AML in children. Patients and methods

Eligible studies Two types of studies were eligible: prospective cohort studies on paediatric AML comparing the outcome of patients with and without a histocompatible family donor, and RCTs on paediatric AML comparing ABMT with other treatments (non-myeloablative consolidation chemotherapy or no further therapy). The prospective cohort studies included were those which (1) Had an explicit policy of tissue typing family members of patients and recommending BMT to all patients in CR1 who had a histocompatible family donor (HLA 5 or 6/6 match). For convenience, we refer to such donors as ‘matched sibling donors’ (MSD); (2) Allocated patients to BMT or comparative therapy groups exclusively on the basis of donor availability; and (3) Compared the outcome of patients in the donor/BMT group with patients in the no donor/comparative therapy group using intention to treat analysis. For both types of study, only trials enrolling patients from 1985 to 2000 were included, since BMT was not widely recommended for all patients with a MSD prior to 1985 and transplantation supportive care after 1985 would be sufficiently comparable to current practice. Eligible studies were paediatric trials reporting AML patients less than 21 years old, who achieved CR1 and were treated within a paediatric oncology unit and/or on a treatment protocol of a collaborative paediatric oncology group. Studies that treated children on adult protocols within adult units were not eligible. Outcome measures Outcomes of interest were relapse after remission, survival and disease-free survival (DFS) with a follow-up time of 3 or more years and treatment-related mortality (TRM, defined as all deaths in remission in all studies). As all participants in the donor vs no donor comparisons had survived induction and achieved remission, DFS and EFS are equivalent outcome measures. Search strategies Relevant studies were identified by electronic searching of the Cochrane Controlled Trials Register, Medline, Embase, Bone Marrow Transplantation

CancerLit and registers of current cancer trials (CancerNet and the UKCCCR Register of Cancer Trials), by handsearching of 17 specialist journals and conference proceedings from three societies. Citations from identified trials and relevant reviews were also followed. Researchers participating in comparative trials in paediatric AML were contacted to identify unpublished studies.

Data extraction Studies identified from the above sources were screened independently by three reviewers for the eligibility criteria specified above. If this could not be done satisfactorily from the title and abstract, a full text version of the published study was obtained for assessment. Two reviewers using a data extraction pro forma independently performed data extraction. Methodological quality of eligible trials was independently assessed unblinded by two reviewers using the full-text article and a simple quality assessment system. Kappa scores for agreement between investigators on three study quality assessment criteria were determined. If incomplete information was provided in the published report, we sought additional data from the authors.

Statistical analysis All analyses were performed according to original treatment allocation (intention-to-treat analysis). Relative risk (RR) and absolute risk difference were calculated for each outcome measure in each trial. Quantitative (Mantel– Haenszel heterogeneity test statistic19) and qualitative heterogeneity between trials was evaluated. If there was no evidence of excessive qualitative or quantitative heterogeneity between trials, the results were pooled and an overall RR and absolute risk reduction for treatment effect across all trials were calculated, using DerSimonian and Laird random effects model.20 Analyses were conducted with the latest version of the Cochrane statistical package, Revman 4.1. For analysis of time-to-event data, the hazard ratio was estimated using information reported in the included studies (the logrank P value, the number of patients allocated to each treatment group and number of events in each allocated treatment group). The log rank expected number of events and variance were used to calculate the log hazard ratio (HR) for individual trials. In the absence of excessive heterogeneity, the log HR for individual trials were then combined to give an overall log HR, calculated using methods described by Parmer et al21 (method 4.3 formula 13, 11 and 4 with an overall log hazard ratio calculated as a weighted average of the individual log hazard ratios according to formulas 1 and 2). To investigate treatment effect within risk groups, data were extracted on outcomes of patients stratified by prognostic risk as defined in the MRC AML 10 trial.16

Bone marrow transplantation for paediatric AML M Bleakley et al

Results Description of studies Six prospective cohort studies comparing the outcome of paediatric AML patients with and without an MSD fulfilled the inclusion criteria.9,11–13,22,23 The characteristics of these six studies are shown in Table 1. The proportion of patients with an MSD who received an MSD BMT ranged from 72 to 100% and was greater than 90% in four of the six studies. Four RCTs comparing ABMT with other treatment were identified.9–14 In three RCTs patients were randomized between ABMT and further chemotherapy,9–12 while one RCT compared ABMT with no further therapy.13,14 The characteristics of these four RCTs are shown in Table 2. All of these randomised comparisons were conducted within prospective multicentre trials involving inception cohorts of paediatric AML patients. Patients were eligible for randomisation if they achieved CR and did not have an MSD. Between 50% and 73% of eligible patients were random-

Table 1

ised and between 54% and 100% of randomised patients received their allocated treatment. Fourteen studies were excluded from the systematic review of BMT for the following reasons: no intention to treat analysis,10,25–30 insufficient data to determine quality of study,31–33 no published or unavailable unpublished data (CCG 2961, unpublished; POG 9421 unpublished), BMT for high risk AML only18 and outcome not from time of remission.34 All reported randomised controlled trials of ABMT were included in the meta-analysis of ABMT.

845

Outcomes: donor vs no donor studies of BMT Relapse: The proportion of patients suffering relapse was significantly lower in the donor than in the no donor group in the pooled result from five studies9,11–13,23,24 in which proportion of relapse was reported (RR 0.59, 95% CI 0.48– 0.72) (Figure 1). This corresponded to a reduced risk of relapse of 18% (Table 3). The crude mean risk of relapse in the no MSD group across trials was 44.2%, whilst the relapse rate amongst those with donors was 23.9%. There

Characteristics of included studies in systematic review of BMT for paediatric AML in first remission

Study ID Accrual period

Participants

Interventions

AEIOP LAM 879 3/87–3/90

Age ⬍15 years Complete remission after 1– 2 courses induction chemotherapy Excluded DS, sAML, MDS

BMT (TBI-Cy) vs other Other included ABMT (BAVC) or chemotherapy (SPC) alone

CCG 21322 1/86–2/89

Age ⬍21 years Complete remission after 1– 2 courses of induction chemotherapy Excluded infant M5 leukemia Age ⬍21 years Complete remission after 1– 2 courses of induction chemotherapy Excluded DS, sAML, MDS, GS

CCG-289111,12 10/89–5/93

MRC AML 1013 5/88–4/95

RAHC23 5/87–11/92

LAME 89/9124 12/88–11/93

Duration of follow-up

Proportion of Proportion of patients patients with MSD without MSD not receiving MSD BMT receiving BMT

28 months projected 5 years

92%

99%

BMT (TBI-Cy) vs chemotherapy

5 years

81%

⬎99%

BMT (BuCy) vs other Other included ABMT (BuCy) or chemo

3 years

91%

95%

Age ⬍15 years Complete remission after 1– 2 courses induction chemotherapy included DS, sAML, MDS Age ⬍15 years Complete remission after 1–4 courses induction chemotherapy Excluded MDS. Included DS, sAML

BMT (TBI-Cy/BuCy) vs other Other included ABMT (TBICy/BuCy) or no further therapy BMT (BuCy) vs ABMT (BuCy)

7 years

72%

98%

4 years

100%

100%

Age ⬍20 years Complete remission after 1– 2 courses of induction chemotherapy Excluded DS, sAML, MDS, M0, biphenotypic leukemia

BMT (BuCy/TBICy/TAM) vs chemotherapy

4 years

100%

99%

DS ⫽ Downs syndrome; sAML ⫽ secondary AML; MDS ⫽ myelodysplastic syndrome; GS ⫽ granulocytic sarcoma. Bone Marrow Transplantation

Bone marrow transplantation for paediatric AML M Bleakley et al

846

Table 2

Characteristics of included randomised controlled trials in systematic review of ABMT for paediatric AML in first remission

Study ID Accrual period

Participants

Interventions

Proportion of Proportion of eligible patients patients randomised randomised to ABMT receiving ABMT

Proportion of patients randomised to control receiving allocated management strategy

AEIOP LAM 879 3/87–3/90

Age ⬍15 years Complete remission after 1– 2 courses induction chemotherapy Excluded DS, sAML, MDS

ABMT (BAVC/non purged) vs chemotherapy (nonHDAC)

70%

66%

54%

CCG 289111,12

Age ⬍21 years Complete remission after 1– 2 courses of induction chemotherapy Excluded DS, sAML, MDS, GS Age ⬍15 years Complete remission after 1– 2 courses induction chemotherapy Included DS, sAML, MDS

ABMT (BuCy/purged) vs chemotherapy (HDAC)

73%

⬎95%

⬎95%

ABMT (TBICy/BuCy/non purged) vs no further therapy

50%

88%

100%

Age ⬍21 years Complete remission after 1– 2 courses induction chemotherapy Excluded sAML, MDS Included DS

ABMT (BuCy, purged) vs chemotherapy (HDAC)

68%

62%

⬎95%

MRC AML 1013,14

POG 882110

DS ⫽ Downs syndrome; sAML ⫽ secondary AML, MDS ⫽ myelodysplastic syndrome; GS ⫽ granulocytic sarcoma

was no statistically significant heterogeneity in RR and risk difference between trials. The HR, estimated from the log-rank statistic, was only available for the CCG 289111,12 and LAME 89/9124 trial. The pooled HR was 0.53 (95% CI 0.39–0.70), indicating patients in the MSD group had a reduced hazard of relapse compared to the no donor group. Disease-free survival: The proportion of patients who relapsed or died following treatment was significantly lower in the MSD group than in the no donor group in the pooled result from all six included studies (RR 0.71 95% CI 0.58–0.86) (Figure 1). There was no significant heterogeneity between trials in risk difference, however quantitative heterogeneity in relative risk was found (␹2 ⫽ 9.66, 5 df, P ⫽ 0.085). The pooled HR calculated for the three trials11,12,22,24 for which time-to-event data was available was 0.67 (95% CI 0.58–0.81) with patients with an available donor having a reduced hazard of relapse or death. Overall survival: In each of the four trials11,12,13,22,23 in which overall survival (OS) was reported, the RR of death was reduced in the MSD group, but this effect was only statistically significant in one of the trials.11,12 In the individual trials the risk of death for patients with MSDs was reduced by 6% to 25%. Whilst the studies showed the same direction of effect, there was heterogeneity in the magnitude of effect for RR (␹2 ⫽ 9.51, 3 df, P ⫽ 0.023) and risk difference (␹2 ⫽ 7.58, 3 df, P ⫽ 0.056). A pooled result for the overall risk of death was determined as the direction of effect was consistent across studies. The pooled Bone Marrow Transplantation

RR of death was 0.68 (95% CI 0.48–0.95) (Figure 1) with an absolute difference risk difference of death of 15% (95% CI 0.05–0.25) (Table 3). The pooled HR calculated for the three trials11–13,22 for which time-to-event data were available were 0.69 (95% CI 0.58–0.83), with the MSD group having a reduced hazard of death. Treatment-related mortality: The absolute number of deaths was small in each individual study. There was gross heterogeneity in direction of effect in TRM. In the MRC AML10 trial13 TRM was higher in the donor group. In contrast, three of the other trials9,11,12,24 reported a higher TRM amongst patients lacking an MSD who underwent alternative post-remission therapy (Figure 1). The apparent effect, however, was not statistically significant in any of the four trials. Randomised controlled trials of ABMT The results are summarised in Figure 2 and Table 4. For each outcome of interest there was qualitative heterogeneity, with or without statistically significant quantitative heterogeneity. For this reason it was inappropriate to present a pooled estimate of effect for RR or HR statistics. Relapse: Only the POG 8821 trial10 showed that the ABMT group had a significantly lower risk of relapse than the control group (RR 0.70, 95% CI 0.53–0.93). The risk of relapse in the ABMT and chemotherapy group was 41% and 58%, respectively, with a risk difference of 17% (95% CI 4–30%).

Bone marrow transplantation for paediatric AML M Bleakley et al

847

Comparison: 01 Relapse Outcome: 01 event = relapse

Treatment n/N

Control n/N

11 / 24 32 / 181 9 / 33 22 / 74 3 / 10

71 / 103 128 / 356 55 / 116 96 / 214 8 / 21

20.3 35.4 12.0 28.8 3.5

0.66(0.42,1.05) 0.49(0.35,0.69) 0.58(0.32,1.04) 0.66(0.45,0.97) 0.79(0.26,2.35)

358 / 810 77 / 322 Total (95% CI) Test for heterogeneity chi square = 2.02 df = 4 P = 0.73 Test for overall effect z = -5.06 P = 0.00001

100.0

0.59(0.48,0.72)

Study AEIOP LAM 87 (9) CCG 2891 (11,12) LAME 89/91 (24) MRC AML 10 (13) RAHC (23)

.1

Comparison: 02 Relapse or death Outcome: 01 event = death or relapse

Study AEIOP LAM 87 (9) CCG 213 (22) CCG 2891 (11,12) LAME 89/91 (24) MRC AML 10 (13) RAHC (23)

RR (95% CI random)

.2

Favours treatment

MSD n/N

No donor n/N

12 / 24 61 / 113 42 / 140 9 / 33 33 / 85 4 / 11

75 / 103 185 / 298 171 / 310 60 / 116 114 / 230 10 / 23

1

Weight %

5

10

Favours control

RR (95% CI random)

161 / 406 615 / 1080 Total (95% CI) Test for heterogeneity chi square = 9.66 df = 5 P = 0.085 Test for overall effect z = -3.40 P = 0.00007 .1

.2

Favours treatment

Comparison: 03 Death Outcome: 01 event = death

54 / 113 34 / 140 25 / 85 2 / 11

5

RR (95% CI random)

14.5 28.5 22.6 9.1 20.9 4.3

0.69(0.45,1.04) 0.87(0.72,1.05) 0.54(0.41,0.71) 0.53(0.29,0.95) 0.78(0.58,1.05) 0.84(0.34,2.08)

100.0

0.71(0.58,0.86)

10

Favours control

RR (95% CI random)

RR (95% CI random)

161 / 298 150 / 310 91 / 230 10 / 23

35.6 30.8 28.1 5.6

0.88(0.71,1.10) 0.50(0.37,0.69) 0.74(0.52,1.07) 0.42(0.11,1.59)

115 / 349 412 / 861 Total (95% CI) Test for heterogeneity chi square = 9.51 df = 5 P = 0.023 Test for overall effect z = -2.24 P = 0.02

100.0

0.71(0.58,0.95)

CCG 213 (22) CCG 2891 (11,12) MRC AML 10 (13) RAHC (23)

Donor n/N

1

Weight %

Weight %

Study

MSD n/N

RR (95% CI random)

.1

.2

Favours treatment

1

5

10

Favours control

Comparison: 04 Treatment-related mortality Outcome: 01 event = treatment-related death Study AEIOP LAM 87 (9) CCG 2891 (11,12) LAME 89/91 (24) MRC AML 10 (13) RAHC (23)

MSD n/N

No donor n/N

0 / 24 0 / 181 1 / 33 11 / 85 1 / 11

9 / 103 3 / 356 8 / 116 16 / 230 2 / 23

RR (95% CI random)

Weight %

RR (95% CI random)

9.0 8.2 15.6 54.3 12.9

0.22(0.01,3.64) 0.28(0.01,5.40) 0.44(0.06,3.39) 1.86(0.90,3.85) 1.05(0.11,10.33)

100.0

0.97(0.40,2.38)

13 / 334 38 / 828 Total (95% CI) Test for heterogeneity chi square = 5.05 df = 5 P = 0.28 Test for overall effect z = -0.06 P = 1 .1

.2

Favours treatment

1

5

10

Favours control

Figure 1 BMT for paediatric AML in CR1 outcome of donor (BMT) vs no donor (alternative intervention) studies.

Bone Marrow Transplantation

Bone marrow transplantation for paediatric AML M Bleakley et al

848

Disease-free survival: The MRC AML10 study13,14 was the only trial that found an improved DFS in the ABMT group. Unlike the other three studies, it compared ABMT with no further therapy and showed a significantly reduced risk of relapse or death in the ABMT group (RR 0.59, 95% CI 0.37–0.96 and HR 0.49, 95% CI 0.27–0.89).

Table 3 Pooled results of donor (BMT) vs no donor (alternative postinduction intervention) comparative studies in systematic review of BMT for paediatric AML in first remission: relative risk and risk difference

Outcome

Relapse Relapse or death Death

Relative risk (95% confidence interval)

Risk difference (95% confidence interval)

0.59 (0.48–0.72) 0.71 (0.58–0.86) 0.68 (0.48–0.95)

⫺0.18 (⫺0.24, ⫺0.12) ⫺0.17 (⫺0.24, ⫺0.09) ⫺0.15 (⫺0.25, ⫺0.05)

Overall survival: In the CCG 2891 trial,11,12 the OS was significantly lower in the patients randomised to ABMT (RR 1.34, 95% CI 1.06–1.70 and HR 1.43, 95% CI 1.04– 1.97). The risk difference was 0.14 (95% CI 0.03–0.25). There was no significant difference in the risk of death between ABMT and control groups in the POG8821 study10 and the UK MRC 10 trial.13,14

Relative risk indicates risk of adverse event in donor group relative to the no donor group. Risk difference indicates the percentage reduction in absolute risk of adverse outcome in the donor group compared to the no donor group. A random effect model was used to pool data.

Treatment-related mortality: TRM was low (less than 6%) in the AEIOP LAM 879 and MRC AML 1013,14 studies

Comparison:01 Relapse Outcome: 01 event = relapse Study AEIOP LAM 87(9) MRC AML 10(13,14) POG 8821(10) Total (95% CI)

ABMT n/N

Conrol n/N

27/35 16/50 43/106

26/37 26/50 66/114

37.0 26.1 36.9

86/191

118/201

100.0

RR (95% CI Random)

Weight %

Test for heterogeneity chi-square = 7.62 df = 2 P = 0.022 Test for overall effect z = -1.17 P = 0.2 .1

Study AEIOP LAM 87(9) CCG 2891(11,12) MRC AML 10(13,14) POG 8821(10) Total (95% CI)

.2

Favours treatment

Comparison:02 Relapse or death Outcome: 01 event = relapse or death

1

5

10

Favours control

ABMT n/N

Conrol n/N

28/35 90/150 16/50 71/115

27/37 80/160 27/50 75/117

25.9 30.2 13.2 30.7

205/350

209/364

100.0

RR (95% CI Random)

Weight %

Test for heterogeneity chi-square = 8.04 df = 3 P = 0.045 Test for overall effect z = -0.02 P = 1 .1

Comparison:03 Death Outcome: 01 event = death Study CCG 2891(11,12) MRC AML 10(13,14) POG 8821(10) Total (95% CI)

.2

Favours treatment

1

5

10

Favours control

ABMT n/N

Conrol n/N

83/150 15/50 69/115

66/160 21/50 66/117

40.2 17.9 41.8

167/315

153/327

100.0

RR (95% CI Random)

Weight %

Test for heterogeneity Chi-square = 5.18 df = 2 P = 0.075 Test for overall effect z = 0.60 P = 0.5 .1

Comparison:04 Treatment-related mortality Outcome: 01 Treatment-related death Study AEIOP LAM 87(9) MRC AML 10(13,14) POG 8821(10) Total (95% CI)

.2

Favours treatment

1

10

ABMT n/N

Conrol n/N

1/35 1/50 11/115

3/37 2/50 3/117

28.7 26.7 44.5

13/200

8/204

100

RR (95% CI Random)

Test for heterogeneity Chi-square = 4.48 df = 2 P = 0.011 Test for overall effect z = 0.12 P = 0.9 .1

.2

Favours treatment

Figure 2 ABMT for paediatric AML in CR1 outcome of randomised controlled trials. Bone Marrow Transplantation

5

Favours control

1

Weight %

5

10

Favours control

Bone marrow transplantation for paediatric AML M Bleakley et al

Table 4 Outcome of randomised controlled trials of ABMT vs chemotherapy and ABMT vs no further therapy for paediatric acute myeloid leukemia in first remission: relative risk and risk difference for individual trials Outcome

Relative riska (95% confidence interval)

849

Risk differenceb (95% confidence interval)

Relapse

POG 8821 MRC AML10 AEIOP LAM

0.70 (0.53–0.93) 0.62 (0.38–1.00) 1.10 (0.83–1.45)

Relapse or death

MRC AML 10 AEIOP LAM CCG 2891 POG 8821

0.59 1.10 1.20 0.96

Death

CCG 2891 MRC AML 10 POG 8821

1.34 (1.06–1.70) 0.71 (0.42–1.22) 1.06 (0.86–1.32)

(0.37–0.96) (0.85–1.42) (0.98–1.47) (0.79–1.17)

POG 8821

⫺0.17 (⫺0.30, ⫺0.04)

MRC AML 10

⫺0.22 (⫺0.41, ⫺0.03)

CCG 2891

0.14 (0.03, 0.25)

a Relative risk indicates risk of adverse event in donor group relative to the no donor group. Absolute risk difference indicates the percentage reduction in absolute risk of adverse outcome in the donor group compared to the no donor group b Risk reductions are presented only for trials in which there was a significantly reduced relative risk of the adverse event in one of the treatment groups

and did not differ significantly between ABMT and control groups. However in the POG8821 study10 where marrow was purged for autograft, TRM was significantly higher (10%) in the ABMT group compared to the chemotherapy controls (RR 3.73, 95% CI 1.07–13.03). TRM was not reported in CCG 2891.11,12 Treatment effect in different risk groups None of the six cohort studies or four RCTs presented stratified treatment outcomes according to morphological or cytogenetic risk groups of paediatric AML. The MRC AML 10 trial13,14 reported the outcome of paediatric patients in good, standard and poor risk cytogenetic–morphological risk groups, but this was not presented according to intervention groups. Although this analysis included data on 2200 children with AML achieving complete remission in 10 clinical trials, treatment effect in different risk groups could not be evaluated based on published data. Discussion This review has summarised the best available evidence on the effect of BMT and ABMT in paediatric AML. Ten trials of BMT or ABMT of adequate methodological quality were identified, representing over 2200 children with AML in CR.9,11–13,22–24 Amongst these studies, patients allocated to BMT on the basis of having an MSD available were found to have a reduced risk of relapse and an improved DFS and OS, compared to patients without donors. The risk of relapse in remission, if there is an MSD available for consolidation therapy with BMT, is reduced by 18% whilst the risk of death is reduced by around 15%. A reduced risk of relapse with BMT was shown in all trials and was statistically significant for the larger studies. The risk of death was also consistently reduced amongst patients allocated to BMT, although the magnitude of the effect varied between studies. Four RCTs of ABMT vs other post-remission treatment strategies in paediatric AML were identified.9–14 The sum-

marised results represent the best available evidence on the effect of ABMT for consolidation therapy in paediatric AML in CR1 amongst patients lacking an MSD. Substantial heterogeneity existed between the included RCTs in the outcomes of relapse, DFS, OS and TRM. It was therefore not appropriate to present summary scores for outcome effects. Therefore, there was insufficient evidence to determine the effect of ABMT compared to chemotherapy, or no further therapy, in preventing relapse or death in paediatric AML in CR1. It was not possible to comment on the value of BMT or ABMT for subgroups of AML with different risks of relapse. Our conclusions are only valid if our results are accurate. This assumes that: (1) donor vs no donor studies are a valid method of quasi-randomisation; (2) pooling of data from studies with different follow-up periods is appropriate; (3) the interventions used in the included studies were qualitatively similar; and (4) a high proportion of patients received the allocated treatment. It is generally accepted that donor vs no donor studies are a valid method of quasi-randomisation. Such studies have no deliberate treatment allocation for an individual. The observed outcome is, therefore, less likely to be caused by baseline differences between patient groups and potential confounding factors are more likely to be distributed evenly between groups. Furthermore, these studies avoid the survivor bias present in studies that simply compare the outcome of patients receiving BMT with other management strategies. The studies selected for inclusion in the systematic review all fulfilled the quality criteria of allocation to treatment group solely on the basis of MSD availability and implementation of intention-to-treat analysis. Although it is difficult to establish intention to transplant, we specifically required that the trials stated a requirement for patients in CR1 to be tissue typed and the outcome analysed on this basis. In five9,11,12,22–24 of the six trials, investigators reported that donor and no donor patient groups were comparable on the basis of known prognostic variables at baseline. In three9,22,24 of the six studies, data were available demonstrating these similarities. One can therefore be confident that the observed improved patient outcome in the Bone Marrow Transplantation

Bone marrow transplantation for paediatric AML M Bleakley et al

850

donor group reflects a true intervention effect. Our strict entry criteria resulted in the exclusion of several other studies in the analysis. One clear limitation of this systematic review and metaanalysis is the pooling of data from studies with different follow-up times (28 months to 7 years). This would also be addressed by updating individual patient data from the included studies and using a standard follow-up time across the pooled studies. Pooling of studies with different followup times is not ideal but, as most events occur before 3 years, combining data with follow-up times of 3 or more years was therefore considered reasonable.9–14,22–24 There are potentially important differences between the studies included in the interventions that were compared. Heterogeneity in outcome between the ABMT trials may relate to differences between the trials in the details of ABMT or control intervention. Such differences included comparison of ABMT vs no further therapy,13,14 rather than comparison with further non-myeloablative chemotherapy;9–12 comparison with chemotherapy control intervention with10–12 or without9,13,14 the use of high-dose cytarabine; and the use of purging and different conditioning regimens (TBI vs chemotherapy) in the ABMT group. There were not sufficient numbers of studies to perform meaningful subgroup analysis. In the included studies there was some variability in the proportion of patients who actually received the treatment to which they were allocated. Incomplete compliance with the intended post-remission therapy means that the difference in intervention effect is likely to be underestimated. The proportion of patients with donors who actually received BMT, ranged from 72% to 100%. This is an acceptable proportion of patients overall, comparing favourably with compliance with allocated interventions in many RCTs. If more patients with donors had received BMT, the intervention effect may have been even more apparent. The patients randomised to ABMT or control formed only a small proportion of the total patient numbers in the trials (50–73%). Compliance with allocation to chemotherapy or ABMT was poor in some trials (54– 100%). In many cases the results in individual trials found no significant difference between ABMT and control groups, which may partially reflect the fact that many patients did not receive the intended intervention. There is no doubt that a substantial proportion of children with AML can be cured by chemotherapy. There is also no doubt that BMT from an MSD still offers the best chance of DFS and OS overall. There is, however, an urgent need to identify which patients will benefit most, or least, from more intensive therapy and BMT. This systematic review was able to determine the average effect of BMT on relapse and survival for paediatric patients with AML. AML is a heterogeneous group of disorders and individual patients have quite different baseline risks of relapse and death following remission. In MRC AML 10, patients classified into good, standard and poor risk subgroups had a baseline risk of relapse after remission of around 33%, 51% and 78% respectively.16 It is unlikely that the intervention effect of BMT is of the same magnitude within each risk group. Although it is hypothesised that BMT will be of greatest advantage within the poor or standard risk groups, it may

Bone Marrow Transplantation

also offer benefit to those in the good risk group. None of the reports included here allowed evaluation of this hypothesis. Prospective RCTs remain the gold standard for evaluation of treatment effect in populations and subgroups. However, if the newly formed Children’s Oncology Group initiates a frontline protocol for AML in 2001/2002, the result of this study will not be known for around 10 years. It is therefore vital that the best use be made of the available data. Meta-analysis of the published data does not allow treatment effect within risk groups to be determined. Our conclusions are that these data should be re-analysed using individual patient data (IPD). IPD meta-analysis is a form of meta-analysis recognised as a gold standard, particularly in the context of clinical trials using time to event outcome data.35–38 IPD meta-analysis is based on individual patient data, in which the separate trial results used in the metaanalysis come from the central analysis of the raw data from each trial. Updated information on each patient entered into each study is collected centrally. The data for each trial are then analysed separately to obtain summary statistics. Patients are therefore compared directly only with other patients in the same trial. For the purposes of therapy for paediatric AML, the main benefit of IPD meta-analysis is the ability to undertake subgroup analyses for important hypotheses about difference in effect. This would allow us to determine if the morphological and cytogenetic risk groups identified from MRC AML 10 and BFM are generally applicable. An additional potential benefit of an IPD analysis is the possibility of including data from the many excluded studies and re-evaluating them using intentional to treat analysis. Individual patient meta-analyses are particularly well suited to summarising and synthesising the data from cancer trials, as has already been done for AML for induction therapy,39 as well as for other diagnoses.35,40,41 Allocation of patients to MSD BMT reduced the risk of relapse significantly and improved DFS and OS. The average effect of MSD BMT is to reduce risk of relapse by 18% and improve survival by 15%. From a clinician’s perspective, this means that in order to prevent one death amongst children with AML in first remission one needs to perform seven MSD BMTs. The survival advantage provided by BMT needs to be considered in the context of the risks of the procedure. These include short- and long-term morbidity, quality of life and financial costs. These were not addressed in the included studies. In the absence of conclusive evidence, both ABMT and non-myeloablative chemotherapy appear to be reasonable treatment options for paediatric AML patients lacking an MSD. As the relapse rate tended to be lower amongst patients randomised to ABMT in both the POG 8821 and MRC AML 10 trial, it could be argued that ABMT would be the preferred current treatment option. However, given the marginal benefit and the reduced rate of attainment of second remission for patients who relapse after ABMT, a better strategy may be to continue intensive chemotherapy for those patients without an MSD and look for an alternative allogeneic donor in the event of poor response or relapse with chemotherapy alone.14

Bone marrow transplantation for paediatric AML M Bleakley et al

Acknowledgements M Bleakley and L Lau were supported by the Leukaemia Research and Support Fund. The authors would like to thank Drs W Woods and T Alonzo for contributing additional data for analysis. We would also like to thank Drs D Henry, J Robertson and D O’Connell for their comments on the systematic review, and Drs M Stevens and E McCahon, for manuscript review.

References 1 Silverman LB, Weinstein HJ. Treatment of childhood leukemia. Curr Opin Oncol 1997; 9: 26–33. 2 Golub TR, Weinstein HJ, Grier HE. Acute myelogenous leukemia. In: Pizzo PA, Poplack DG (eds). Principles and Practice of Pediatric Oncology. Lippincott-Raven: Philadelphia, 1997, pp 463–482. 3 Gray R, Wheatley K. How to avoid bias when comparing bone marrow transplantation with chemotherapy. Bone Marrow Transplant 1991; 7 (Suppl. 3) 4–12. 4 Ljungman P, de Witte T, Verdonck L et al. Bone marrow transplantation for acute myeloblastic leukemia: an EBMT Leukemia Working Party prospective analysis from HLA typing. Br J Haematol 1993; 84: 61–66. 5 Ebb DH, Weinstein HJ. Diagnosis and treatment of childhood acute myelogenous leukemia. Ped Clin N Am 1997; 44: 847–859. 6 Burnett AK. Tailoring the treatment of acute myeloid leukemia. Curr Opin Oncol 1999; 11: 14–19. 7 Pui CH. Acute leukemia in children. Curr Opin Hematol 1996; 3: 249–258. 8 Lowenberg B, Downing JR, Burnett A. Acute myeloid leukemia. New Engl J Med 1999; 341: 1051–1062. 9 Amadori S, Testi AM, Arico M et al. Prospective comparative study of bone marrow transplantation and postremission chemotherapy for childhood acute myelogenous leukemia. J Clin Oncol 1993; 6: 1046–1054. 10 Ravindranath Y, Yeager AM, Chang MN et al. Autologous bone marrow transplantation versus intensive consolidation chemotherapy for acute myeloid leukemia in childhood. New Engl J Med 1996; 334: 1428–1434. 11 Woods WG, Kobrinsky N, Buckley JD et al. Timed-sequential induction therapy improves post-remission outcome in acute myeloid leukemia: a report from the Children’s Cancer Group. Blood 1996; 87: 4979–4989. 12 Woods WG, Neudorf S, Gold S et al. Aggressive postremission (REM) chemotherapy is better than autologous bone marrow transplantation (BMT) and allogeneic BMT is superior to both in children with acute myeloid leukemia (AML). Proc Annu Meet Am Soc Clin Oncol 1996; 15: 368. 13 Stevens RF, Hann I, Wheatley K et al. Marked improvements in outcome with chemotherapy alone in paediatric acute myeloid leukemia: results of the United Kingdom Medical Research Council’s 10th AML trial. Br J Haematol 1998; 101: 130–140. 14 Burnett AK, Goldstone AH, Stevens RF et al. Randomised comparison of addition of autologous bone-marrow transplantation to intensive chemotherapy for acute myeloid leukemia in first remission: results of MRC AML 10 trial. Lancet 1998; 351: 700–708. 15 Webb DK, Wheatley K, Harrison G et al. Outcome for children with relapsed acute myeloid leukemia following initial therapy in the Medical Research Council (MRC) AML 10 trial. Leukemia 1999; 13: 25–31. 16 Wheatley K, Burnett AK, Goldstone AH et al. A simple,

17

18 19

20

21 22

23

24

25

26 27

28

29

30

31

robust, validated and highly predictive index for the determination of risk-directed therapy in acute myeloid leukemia derived from the UK MRC 10 trial. Br J Haematol 1999; 107: 69–79. Cruetzig U, Ritter J, Schellong G. Identification of two risk groups in childhood acute myelogenous leukemia after therapy intensification in study AML-BFM-83 as compared with study AML-BFM-78. Blood 1990; 75: 1932–1940. Cruetzig U, Zimmerman M, Ritter J et al. Definition of a standard-risk group in children with AML. Br J Haematol 1999; 104: 630–639. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959; 22: 719–723, in The Cochrane Collaboration Handbook Version 3.0.2, 1997. www.cochrane.org/hbook.htm DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled clinical trials 1986; 7: 177–188, in The Cochrane Collaboration Handbook Version 3.0.2, 1997. www.cochrane.org/hbook.htm Parmer MK, Tori V, Stewart L. Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Stat Med 1998; 17: 2815–2834. Wells RJ, Woods WG, Buckley J et al. Treatment of newly diagnosed children and adolescents with acute myeloid leukemia; a Children’s Cancer Group study. J Clin Oncol 1994; 12: 2367–2377. Shaw PJ, Bergin ME, Burgess MA et al. Childhood acute myeloid leukemia; outcome in a single centre using chemotherapy and consolidation with busulphan/cyclophosphamide for bone marrow transplantation. J Clin Oncol 1994; 12: 2138–2145. Michel G, Leverger G, Leblanc T et al. Allogeneic bone marrow transplantation vs aggressive post-remission chemotherapy for children with acute myeloid leukemia in first complete remission. A prospective study from the French Society of Pediatric Hematology and Immunology (SHIP). Bone Marrow Transplant 1996; 17: 191–196. Woods WG, Kobrinsky N, Buckley J et al. Intensively timed induction therapy followed by autologous or allogeneic bone marrow transplantation for children with acute myeloid leukemia or myelodysplastic syndrome: a Children’s Cancer Group pilot study. J Clin Oncol 1993; 11: 1448–1457. Hermann J, Zintl F, Krause M et al. Therapy of acute myeloid leukemia in children-results of the AML II/87 multicenter study. Kinderarztliche Praxis 1991; 59: 321–327. Malik S, DeLaat C, Harris R et al. Allogeneic bone marrow transplantation vs autologous bone marrow transplantation vs chemotherapy in childhood acute myeloid leukemia. Proc ASCO 1993; 12: 323. Boulad F, Gillio A, Small T et al. Bone marrow transplantation for the treatment of childhood acute myelogenous leukemia – the MSKCC experience. American Society for Hematology 36th meeting conference proceedings. Blood 1994; 84: 205a. Lie S, Johnmundsson G, Mellander l et al. A population-based study of 272 children with acute myeloid leukemia treated on two consecutive protocols with different intensity; best outcome in girls, infants, and children with down’s syndrome. Br J Haematol 1996; 94: 82–86. Vowels MR, Lam-Po-Tang R, Ford D et al. Allogeneic and autologous bone marrow transplantation for acute non-lymphoblastic leukemia. J Paediatr Child Health 1994; 30: 319–323. Kigasawa H, Akiyama Y, Ohira M et al. Prospective comparison between postinduction chemotherapy and various types of stem cell transplants for childhood acute myeloid leukemia

851

Bone Marrow Transplantation

Bone marrow transplantation for paediatric AML M Bleakley et al

852

32

33

34 35 36 37

using Japan cooperative protocols ANLL91. SIOP XXVII Meet Abstr 1995; O–127. Matsuzaki A, Eguchi H, Miyazaki S et al. Treatment of childhood acute myelogenous leukemia with protocol ANLL93 containing allogeneic and autologous stem cell transplantation during first remission. SIOP Meet Abstr 1999; P72. Mimaya J, Horikoshi Y, Shimizu H et al. Myelogenous leukemia in children. ANLL 9205 study by Children’s Cancer and Leukemia Study Group (CCLSG). Rinsho Ketsueki Jpn J Clin Hematol 1997; 38: 100–107. Ortega J, Olive T, Diaz de Heredia et al. Autologous vs allogeneic BMT after post induction intensification therapy in AML in children. Bone Marrow Transplant 1995; 15: 234. Clarke M, Stewart L, Pignon JP et al. Individual patient data meta-analyses in cancer. Br J Cancer 1998; 77: 2036–2044. Stewart LA, Clarke MJ. Practical methodology of meta-analyses (overviews) using updated individual patient data. Stat Med 1995; 14: 2057–2079. Stewart LA, Parmer MK. Meta-analysis of the literature or of

Bone Marrow Transplantation

38

39

40 41

individual patient data: is there a difference? Lancet 1993; 341: 418–422. Steinberg KK, Smith SJ, Stroup DF et al. Comparison of effect estimates from a meta-analysis of summary data from published studies and from a meta-analysis using individual patient data for ovarian cancer studies. Am J Epidemiol 1997; 145: 917–925. The AML Collaborative Group. A systematic collaborative overview of randomised trials comparing idarubicin aith daunorubicin (or other anthracyclines) as induction therapy for acute myeloid leukaemia. Br J Haematol 1998; 103: 100–109. Tierney J for the Sarcoma Meta-Analysis Collaboration (SMAC). Adjuvant chemotherapy for localised resectable soft tissue sarcoma in adults. Lancet 1997; 350: 1647–1654. Tierney JF, Mosseri V, Stewart LA et al. Adjuvant chemotherapy for soft tissue sarcoma: review and meta-analysis of the published results of randomised clinical trials. Br J Cancer 1995; 72: 467–475.