Clinical effectiveness and cost-effectiveness of stem cell transplantation in the management of acute leukaemia: a systematic review

Types of SCT

Treatment pathway

Treatment of acute leukaemia is divided into several phases. The first phase of treatment after a patient is diagnosed with acute leukaemia is induction therapy. The aim is to rapidly kill most of the tumour cells and get the patient into a state of complete remission, which is defined as < 5% blasts in the bone marrow, normal peripheral blood counts and no other symptoms or signs of the disease. 3 Once complete remission has been achieved, a phase of consolidation/intensification follows to eradicate remaining leukaemia cells as much as possible and to prevent the return of the disease. SCT may be considered during first complete remission if the patient is judged to be at high risk of relapse. Following consolidation, patients with ALL are usually given central nervous system (CNS) prophylaxis involving cranial irradiation and intrathecal chemotherapy and further maintenance chemotherapy. Autologous SCT is sometimes used as further consolidation following long-term chemotherapy. CNS prophylaxis and maintenance chemotherapy are usually not needed for patients with AML.

If relapse occurs, reinduction therapy can be given and SCT may be considered during subsequent complete remission. For patients who fail to achieve complete remission following first or subsequent induction therapy, further salvage chemotherapy may be given and allogeneic SCT may also be considered.

Additional supportive treatments that deal with infections, haemorrhage and other symptoms that are associated with either the disease itself or the aforementioned treatments may also be needed throughout disease management.

Definition of the intervention and indications

Place of the intervention in the treatment pathway

Patients who are diagnosed with acute leukaemia are usually given an induction treatment which aims to bring about complete remission of the disease. Once complete remission has been achieved, a phase of consolidation/intensification therapy is followed to prevent the relapse of the disease. Further maintenance therapy may also be needed. The effectiveness of these chemotherapies (and sometimes radiotherapy) in inducing remission and preventing relapse varies between types of leukaemia (AML vs ALL), age, and other prognostic factors. For example, complete remission following induction therapy has been achieved in over 90% of children with ALL3 but only in < 65% of older adults (≥ 56 years) with AML. 6 SCT provides alternative treatment options that have the potential to reduce the risk of relapse or even cure the disease when chemotherapy alone fails to eradicate the disease and/or when the patient’s prognosis remains poor owing to a high risk of relapse. Nevertheless, SCT is not always successful and does not always prevent relapses, and the procedures (in particular allogeneic SCT) involve significant risk of treatment-related mortality and morbidity. Whether SCT would be considered as an alternative/additional treatment to chemotherapy and/or no further therapy in the management of acute leukaemia therefore depends very much on balancing the potential benefit that SCT may confer (taking into account the potential harm that SCT may cause) against the potential risk of relapse or poor prognosis if SCT is not carried out.

SCT is usually considered during the first complete remission following induction therapy, or during subsequent remissions after disease relapse has occurred. Allogeneic SCT may also be considered for patients who are unable to achieve complete remission with chemotherapy (see Figure 3). Once SCT is judged to be potentially beneficial, further issues need to be considered to inform the choice of the type of SCT. For patients who have an HLA-matched sibling donor, allogeneic SCT is usually preferred over autologous SCT as the former is thought to have the advantage of a graft-versus-leukaemia effect whereas the latter has the possibility of reintroducing leukaemic cells to patients. For patients without an HLA-matched sibling donor, allogeneic SCT from an HLA-matched unrelated donor may also be considered. Autologous SCT may be considered for patients with no suitable donor, for patients who may not be able to tolerate the intensive conditioning regimen required for allogeneic SCT, such as older patients and patients with comorbidity, and for patients in whom the higher risk of mortality associated with allogeneic SCT is considered unacceptable.

FIGURE 3.

Potential use of stem cell transplantation in the treatment pathway for the management of acute leukaemia.

For patients who do not have a matched donor, possible treatment options when remission has been achieved include autologous SCT, further chemotherapy or no further therapy. For patients who fail to achieve remission, possible treatment options include allogeneic SCT, further chemotherapy or no further therapy.

The use of SCT in the treatment pathway for acute leukaemia is partly reflected in the conduct of controlled clinical trials, the structure of which has been summarised by Johnson et al. ,7 as shown in Figure 4. Not all possible comparisons have been tested in the trials owing to practical difficulties and ethical issues.

FIGURE 4.

Typical simplified schema for trials comparing sibling allogeneic or autologous transplantation with conventional chemotherapy in acute leukaemia. Adapted from Johnson et al. 7

Because any intention to perform allogeneic SCT is restricted by the availability of a matched donor, randomisation to receive either allogeneic SCT or alternative treatment (e.g. autologous SCT or chemotherapy) is possible only among patients with a matched sibling or unrelated donor. Even within this patient population, recruitment to randomised trials comparing allogeneic SCT with chemotherapy is difficult, either because of the significant treatment-related mortality and serious side effects associated with allogeneic SCT, which are considered unacceptable to patients and clinicians, or, conversely, because of their belief that allogeneic SCT (particularly when a matched sibling is available) is the treatment of choice and hence unwillingness to accept the possibility of not having SCT.

Despite the practical difficulties in carrying out a randomised study comparing allogeneic SCT with alternative treatments, it has been pointed out that whether or not a sibling is an HLA-matched donor depends on the random assortment of genes at fertilisation, and thus a comparison of patients with a sibling donor with those without a donor is effectively a randomised evaluation. 8 This ‘donor versus no donor’ comparison has therefore been used to evaluate the effectiveness of allogeneic SCT from sibling donor compared with alternative treatment options. The ‘intention to treat’ (ITT) (i.e. to carry out allogeneic SCT from a sibling donor) here is solely dictated by the availability of a matched sibling donor. The term ‘donor versus no donor (DvND) studies’ will be used in this report to describe this type of study.

Relevant comparators

Potentially relevant comparisons for this review included different forms and techniques of SCT compared with each other, with further chemotherapy or with no further therapy. Possible treatment options were identified on the basis of the characteristics of the specific patient group in question (e.g. adults with AML with high risk of relapse) and its place in the treatment pathway (e.g. in first complete remission). Therefore, appropriate comparators were determined for AML and ALL for a given age/risk group and disease stage. Not all the potential comparators were considered relevant for each patient group/disease stage.

Specific decision problems

The following specific decision problems were identified on the basis of the scope stated in the commissioning brief of this National Institute for Health Research (NIHR) Health Technology Assessment (HTA) report, taking into account the treatment pathway described above and additional advice from our clinical and methodological experts. As the aim of the review was to provide a critical assessment and summary of best available evidence, these decision problems (DPs) were listed in view of their clinical and economic relevance and not according to whether good-quality evidence related to these decision problems existed.

Search strategy

Full search strategies are detailed in Appendix 1. A scoping search was undertaken to identify existing reviews and other background material. Searches for systematic reviews were carried out using the Aggressive Research Intelligence Facility (ARIF) search protocol and included searches of MEDLINE (Ovid) 1950 to April week 3, 2008, EMBASE (Ovid) 1980 to week 16, 2008, MEDLINE In-Process & Other Non-Indexed Citations (Ovid) as at 24 April 2008 and the Cochrane Library (Wiley) 2008 Issue 3 [Cochrane Database of Systematic Reviews (CDSR), Database of Abstacts of Reviews of Effects (DARE) and HTA databases]. Searches of the same databases, together with the Cochrane Library (Wiley) 2008 Issue 3 [Cochrane Central Register of Controlled Trials (CENTRAL) and NHS Economic Evaluation Database (EED) databases], were undertaken to estimate the volume and nature of primary studies for the topic.

Study selection criteria

Data extraction and quality assessment strategy

Data from included systematic reviews and meta-analyses were extracted by one reviewer on to a data extraction form (spreadsheet) adopted from a proforma that had been designed for critical appraisal of systematic reviews. The data extraction was checked by another reviewer to ensure accuracy.

Methods of synthesis and structure of the report

A narrative overview of published systematic reviews and meta-analyses is provided in Chapter 3. Findings from these studies, as well as new evidence from primary studies and information on relevant ongoing trials, is presented in Chapter 4 according to the structure of the decision problems outlined in Chapter 1. Data have been tabulated and summarised in a narrative manner for each decision problem. The strength and weakness of evidence from existing systematic reviews/meta-analyses, and new evidence not covered by these studies and its potential impact, are discussed. An inventory identifying areas of priority for future primary research and evidence synthesis is provided in Chapter 7. No de novo quantitative synthesis of data was planned or carried out.

Explanation of comparisons presented in DPs 1–8

Search strategies

The following sources were searched to identify information on cost-effectiveness:

• MEDLINE, EMBASE, DARE and NHS EED via the Cochrane Library

• internet sites of national economic units.

There were no date or language restrictions. The reference lists of included systematic reviews were also checked. Full details of the search and databases searched are provided in Appendix 1. The end date for the searches was January 2009.

Study selection criteria

The titles and abstracts of records retrieved from the searches for cost-effectiveness, as well as any other potentially relevant articles that were identified through additional searches, were examined for inclusion by a single reviewer (CH). It was originally hoped that two reviewers would be available to screen, but this was not possible owing to resource constraints. One reviewer was thought to be acceptable, given the experienced nature of the reviewer. The criteria used are listed as follows.

Data extraction and quality assessment strategy

Data from the included economic evaluations were extracted and critically appraised according to a published checklist by Philips et al. 9 The key items of information to be assessed included:

• details of the study characteristics, such as form of economic analysis, comparators, perspective, time horizon and modelling used

• details of the effectiveness and cost parameters, such as effectiveness data, health state valuations, resource use data, unit cost data, price year, discounting assumptions and productivity costs

• details of the results and sensitivity analysis.

Details and data of costs studies, which we did not originally intend to review, were extracted using a similar framework to that used in an NIHR HTA programme review by Johnson et al. 7 on SCT. No detailed critical appraisal was carried out, although major challenges to validity were recorded as notes.

The data extraction and quality assessment was checked by another reviewer (LA) to ensure accuracy. Disagreement between reviewers was resolved by discussion with involvement of a third reviewer when necessary.

Methods of analysis/synthesis

Results of the review of cost-effectiveness literature were tabulated and summarised in a narrative manner for each decision problem. Again the framework employed by the HTA review by Johnson et al. 7 on SCT was employed where still appropriate. The strengths and weaknesses of existing economic evaluations were discussed and potential evidence gaps that might impede a reliable economic evaluation or that might be associated with major uncertainty were identified.

Johnson 1998

This HTA report7 evaluated the effectiveness and cost-effectiveness of SCT compared with conventional chemotherapy for the treatment of a number of haematological and other solid malignancies. It appears to be the first systematic review to assess the efficacy of SCT in acute leukaemia. Studies were identified by electronic search in CancerLit, EMBASE, MEDLINE and NHS EED, by hand searching of conference proceedings of major societies, and by searching within the UK cancer registries/databases.

Randomised controlled trials, DvND studies and cohort studies comparing SCT with chemotherapy were included in the review; the authors did not include case series. For each disease entity, a decision, based on the number of trials, patients and events, but without knowledge of the results, was made as to which level of evidence should be considered. Thus, if sufficient evidence was available from the RCTs, no other studies were reviewed. Similarly, if sufficient evidence was available from the RCTs plus DvND studies, no further studies were reviewed. Otherwise, all categories of study were included but quantitative analyses were performed only for RCTs and DvND studies. No data synthesis was carried out for cohort studies and no conclusions were drawn from them. End points considered were survival and progression-free survival (PFS); odds ratios (ORs) were calculated for each end point.

The analysis of identified DvND studies revealed a survival advantage for allogeneic transplantation over chemotherapy among adults with AML in CR1. For RCTs comparing autologous transplantation with chemotherapy, there was no good evidence of a survival difference between the two treatment approaches but there was some suggestion that autologous SCT might improve PFS. For adult AML in CR2, no RCTs or DvND studies were identified that compared transplantation with chemotherapy; two retrospective cohort studies comparing the two treatment approaches were found but no conclusion was drawn from them.

In the case of children with AML, the results of identified trials that compared allogeneic transplantation with chemotherapy once again suggested a benefit in favour of allogeneic transplantation, for both survival and PFS. However, unlike the case of adults with AML in CR1, all RCTs that compared autologous transplantation with chemotherapy in children in CR1 tended to favour conventional therapy.

The authors also concluded that it was not possible to determine whether transplantation – allogeneic or autologous – offered any benefit over conventional chemotherapy in the consolidation of CR1 in adult ALL, mainly because the trials identified at the time were generally relatively small and inconsistently reported.

For children with ALL in first complete remission, one CCT and two cohort studies of allogeneic SCT versus conventional chemotherapy were identified but these did not yield any data on which to draw conclusions. In CR2, despite SCT being perceived as the treatment of choice in the one relevant CCT comparing allogeneic SCT with conventional chemotherapy, there was again insufficient data on which to draw any conclusion.

Bleakley 2002

The systematic review and meta-analysis by Bleakley et al. 23 on BMT for children with AML in CR1 included DvND studies comparing the outcome of patients with and without a histocompatible family donor, as well as RCTs comparing autologous BMT with other treatments. For both types of study, only trials enrolling patients from 1985 were included, as BMT was not widely recommended before then.

Relevant studies were identified by electronic searching of several databases (including the Cochrane Controlled Trials Register, MEDLINE, EMBASE, CancerLit) and registers of current cancer trials, hand searching of specialist journals and conference proceedings, descendent searching of identified trials and relevant reviews, as well as contact with researchers. Full texts of trials were assessed for methodological quality using a simple quality assessment system. Outcome measures were relapse, overall survival (OS), disease-free survival (DFS) and treatment-related mortality (TRM).

Relative risk (RR) and absolute risk difference were calculated for each outcome measure and results were pooled where appropriate. The authors’ conclusion was that allogeneic transplantation from a histocompatible family donor improved outcomes among children with AML in CR1. They were, however, unable to comment on the value of autologous BMT compared with chemotherapy. This was because of the presence of substantial heterogeneity between the included RCTs, which made it inappropriate to present summary scores.

Levi 2004

Levi et al. 21 carried out a meta-analysis comparing autologous BMT transplantation and intensive chemotherapy in adults with AML in CR1. A computerised search of MEDLINE was conducted from January 1966 to March 2003 for RCTs of autologous BMT in this group of patients. Identified studies were appraised and scored using the standard developed by Chalmers et al. 27 Risk of dying (death rate) and the risk of relapse or death (event rate) were calculated for each arm; rate ratios of death and events rates for each study were then calculated and aggregated.

The overall estimates of rate ratio indicated that autologous BMT had no advantage over chemotherapy or no further treatment concerning death rate but was superior to chemotherapy with regards to event rate. The authors therefore concluded that autologous BMT improved event-free survival (EFS) but not OS, when compared with chemotherapy or no further treatment in patients with AML in CR1.

Nathan 2004

This meta-analysis20 was quite similar to the study conducted by Levi et al. 21 Two types of trial designs were included: prospective cohort studies that offered allogeneic BMT to all eligible patients in CR1 with an available donor and randomly assigned all remaining patients to either autologous BMT or chemotherapy/no further treatment; RCTs that compared autologous BMT with chemotherapy or no further treatment in all patients in CR1.

Studies were identified by an electronic search of databases (MEDLINE/PREMEDLINE, EMBASE, the Cochrane Controlled Trials Registry and CancerLit) and by manual searching of the reference lists of articles and relevant reviews that were retrieved in full. Publications finally included were assessed for validity with the use of an eight-item checklist. Primary outcomes were OS and DFS and secondary outcomes were TRM and survival of relapsed patients.

A point estimate of the ratio of survival probabilities comparing the autologous BMT group with the chemotherapy group derived from the Kaplan–Meier estimates of survival (as reported in the studies) served as the effect size statistic. As in the meta-analysis by Levi et al. ,21 the aggregated results of this study revealed that the OS between the two treatment strategies was similar, and that compared with patients who received chemotherapy or no further therapy, patients who received autologous BMT had a better DFS. This led the authors to conclude that the results did not support the routine use of autologous BMT in adults with AML in CR1.

Schlenk 2004

Schlenk et al. 25 performed an IPD meta-analysis on adults with core binding factor (CBF) AML [defined by the presence of t(8;21) or inv(16)] treated between 1993 and 2002 in prospective German AML treatment trials. Their aim was to evaluate prognostic factors for recurrence-free survival (RFS) and OS and to assess the impact of different post-remission therapies in this group of patients.

Survival analysis showed similar OS and RFS for the two groups and ITT analysis for post-remission therapy revealed no difference in RFS and OS between intensive chemotherapy and autologous SCT in the t(8;21) group and between chemotherapy, autologous SCT and allogeneic SCT in the inv(16) group.

In the t(8;21) group, significant prognostic variables for longer RFS and OS were lower WBC and higher platelet counts; loss of the Y chromosome in male patients was prognostic for shorter OS. In the inv(16) group, trisomy 22 was a significant prognostic variable for longer RFS. For patients who experienced relapse, CR2 was significantly lower in patients with t(8;21), resulting in a significantly inferior survival duration after relapse compared with patients with inv(16).

Hahn 2005

The study by Hahn et al. 13 on the role of cytotoxic chemotherapy with SCT in the therapy of ALL in children is the third publication to result from the initiative of the ASBMT to sponsor evidence-based reviews of the scientific and medical literature for the use of blood and marrow SCT in the therapy of selected diseases.

Literature searching was carried out within MEDLINE, spanning the period 1 January 1980 to 3 January 2005, and evidence was gathered from all published articles related to SCT for ALL in children. Meeting abstracts and data from non-peer-reviewed journals were excluded, as were case reports (≤ 10 patients), reviews, consensus reports, practice guidelines or laboratory studies with no clinical correlates. All included studies were graded based on the quality of their design and the strength of their evidence as per the system devised by Harbour and Miller. 28

There was no statistical pooling of results; details were conveyed in text format and in summary tables. Among its recommendations, the review stated that when compared with chemotherapy among children with ALL in CR1, benefit was demonstrable only for matched related allogeneic SCT in very high-risk [Philadelphia chromosome-positive (Ph+ve) only] cases, and that SCT was not recommended for standard or other high-risk (e.g. induction failure, hypodiploidy, etc.) patients in CR1, except in the context of clinical trials.

In the case of patients in CR2, there was a cautious recommendation of matched related allogeneic transplantation over chemotherapy, because part of the evidence came from one prospective trial that did not demonstrate a benefit for transplantation when analysed by the presence versus absence of a related donor in an ITT analysis; no recommendations were made with regard to unrelated allogeneic transplantation versus chemotherapy on the grounds of insufficient evidence.

The review also concluded that regimens including total body irradiation (TBI) had better outcomes than regimens without TBI. No recommendations were made between autologous and allogeneic SCT as the comparison between these two had not been adequately studied at the time.

Yanada 2005

The purpose of this meta-analysis18 was twofold: to identify the overall efficacy of allogeneic SCT for patients with AML in CR1 and to assess the efficacy for patients stratified into favourable, intermediate and poor cytogenetic risk groups. Literature searching was conducted in MEDLINE for DvND studies published between 1995 and 2003 that, among others, used an ITT analysis and assessed outcomes in terms of OS. The authors used only articles published after 1995 in consideration of the changes in procedures for allogeneic SCT.

Five studies were eventually used for the meta-analysis. Hazard ratios (HRs) were used to assess the survival advantage of allogeneic SCT compared with non-allogeneic SCT, and cytogenetic risk categories were examined using meta-regression analysis. Results showed that there was a statistically significant advantage from allogeneic SCT in terms of OS in this group of patients.

In addition, meta-regression analysis indicated that the benefit of allogeneic SCT was further increased for the poor cytogenetic risk group, mildly increased (but not significant) for the intermediate-risk group and lost for the favourable cytogenetic risk group. The authors’ conclusion was that this suggested that the efficacy of allogeneic SCT for patients with AML in CR1 depended on cytogenetic risk.

Hahn 2006

This is the fourth review14 to result from the ASBMT initiative mentioned above, and it investigated the roles of cytotoxic therapy and haematopoietic SCT in the treatment of ALL in adults. Literature searching, selection and grading criteria were the same as for the review of SCT in ALL in children. And, just as in the previous reviews, there was no statistical pooling of results; details were conveyed in text format and in summary tables.

The study revealed that in adults with ALL in CR1, SCT yielded outcomes similar to chemotherapy and was thus not recommended as first choice therapy in standard-risk patients. For high-risk patients, there were no direct comparisons, but some data suggested an advantage for SCT. In CR2, SCT was recommended over chemotherapy, as a sizable fraction of patients achieved extended leukaemia-free survival compared with chemotherapy alone; however, there are no direct comparative data.

In the comparison between the transplantation techniques, there was a preponderance of evidence favouring allogeneic over autologous SCT, and in the comparison of conditioning regimens, the authors concluded that there were not enough data to make a recommendation of the superiority of any one regimen. However, there appeared to be a benefit for TBI-containing regimens compared with non-TBI-containing regimens.

Visani 2006

The aim of this systematic review15 was to define, according to the rules of evidence-based medicine, the role of allogeneic SCT compared with autologous SCT and intensive chemotherapy after achievement of CR1 in adults with AML. The authors searched MEDLINE for reports of cohort studies, RCTs and meta-analyses from 1985 through to January 2005 and also scrutinised the references of the identified articles. To be included, studies had to satisfy strict methodological criteria and had to consider global mortality (OS) and/or DFS as primary outcomes. Aggregation of results was not deemed necessary by the authors because of the inclusion of a reliable recent meta-analysis. Details of each included study were presented in structured tables and in written text.

In their analysis with regard to the comparison between autologous SCT and intensive chemotherapy/no further therapy, the authors opined that there were still contradictory opinions about the role of autologous transplantation in patients with AML in complete remission, as a possible advantage of autologous SCT over intensive chemotherapy was not clearly supported by data from the clinical trials. There was no evidence that autologous SCT was superior in terms of OS to chemotherapy; a differentiated approach in patients with high-, standard- or low-risk AML could also not be suggested because of insufficient data.

The second end point of this review was to establish whether allogeneic SCT was superior to other therapeutic options in improving DFS and/or OS in adults with AML. As no studies effectively compared allogeneic SCT and intensive chemotherapy, the authors mainly focused their analysis on a DvND comparison between allogeneic and autologous SCT. No overall benefit of allografting on survival was demonstrated by any trial. The review concluded by suggesting that, in view of the inconclusive evidence and the emergence of genetic subgroups, there is a need for randomised trials directly focusing on the single entities. Hence, the cure for AML could eventually become the cure for each specific AML subset with its peculiar biological, molecular and prognostic features.

Yanada 2006

This meta-analysis19 was carried out to provide precise estimates of the clinical efficacy of allogeneic SCT as post-remission therapy for adult patients with ALL, in the presence of inconsistent results being reported by a number of studies. Literature searching was conducted in MEDLINE to include original articles of DvND studies based on an ITT analysis which assessed outcomes in terms of OS.

Quantitative data synthesis involved the calculation of HRs and the estimation of summary HRs. Results demonstrated that patients in the donor groups had significantly better survival than patients in the no-donor groups, but the overall result revealed significant heterogeneity, which was attributable to the high-risk group on subgroup analysis. Furthermore, when only high-risk patients were included in the analysis, the superiority of the survival advantage was even greater.

In addition, meta-regression analysis was employed to further explore the source of heterogeneity in survival analyses and it revealed that compliance with allogeneic SCT had a significant and positive correlation with survival, i.e. the greater the proportion of patients who actually received allogeneic SCT, the better the survival of the donor group. No beneficial effects of autologous SCT were observed.

The authors concluded that the findings demonstrated that allogeneic SCT improved the outcome of adult patients with high-risk ALL and, as such, allogeneic SCT should be considered for such patients if a suitable donor was available.

Cornelissen 2007

Cornelissen et al. 22 presented the results of the HOVON-SAKK collaborative study group, which evaluated outcomes for patients with AML in CR1 that were entered in three consecutive studies according to a DvND comparison. In addition, as reports of three previous major trials failed to demonstrate a significant benefit in terms of OS and also showed disconcordant results in terms of DFS, the authors aggregated their results and those of the three previous studies in a meta-analysis so as to enhance statistical power.

Disease-free survival and OS were analysed by donor availability and broken down for cytogenetic risk category as well as age category. The findings from all four studies were highly consistent and the confidence intervals of the HR estimates for the separate studies all contained the pooled estimate. Pooled results revealed both an enhanced DFS and a statistically significant OS benefit in all AML patients in CR1. However, subgroup analysis revealed that this benefit was not observed in patients with favourable cytogenetics, with benefit being restricted to patients without favourable cytogenetics. The analysis also indicated that in patients younger than 40 years of age, the advantage in the donor group was present but less pronounced for DFS, but absent for OS.

Oliansky 2007

This systematic review,16 assessing the role of cytotoxic chemotherapy with SCT in the treatment of AML in children, was the fifth of the evidence-based review series arising from the ASBMT initiative. Literature searching for published reports was carried out within MEDLINE, spanning the period 1990–2006. Selection and grading criteria were the same as for the review of SCT; study details were conveyed in text format and in summary tables.

The authors concluded that for children with AML in CR1, autologous SCT and chemotherapy had equivalent survival outcomes, whereas allogeneic SCT was superior to chemotherapy in terms of both OS and leukaemia-free survival. There was also a consensus recommendation for matched related donor allogeneic SCT over chemotherapy in CR2, although there was a lack of evidence comparing these two therapeutic options.

With regard to transplantation techniques, the conclusion was that matched related donor allogeneic SCT had superior survival outcomes compared with autologous SCT in CR1. For CR2, the recommendation was to use allogeneic over autologous SCT. However, this was based on consensus rather than on evidence from trials, as this comparison yielded no evidence that one had better outcomes than the other. In addition, the comparison of allogeneic SCT myeloablative conditioning regimens demonstrated no advantage of one regimen over another with respect to survival data or late effects.

Orsi 2007

Orsi et al. 26 conducted a survival meta-analysis from individual patient information, to compare allogeneic transplantation versus chemotherapy/autologous transplantation using an ITT approach, in patients with ALL in CR1. CCTs, wherein allocation to allogeneic transplantation was based on donor availability, to be included in the study were identified by electronic searching of MEDLINE and the Iowa Drug Information System (IDIS) compact disks and subjected to stringent inclusion criteria. Study quality was rated as excellent, good, sufficient or poor, based on subjective assessment.

Event-free survival was chosen as the outcome index for the review as it was the more frequently reported outcome. Data of individual survivals were derived either from the original data provided by the trial’s authors or from the information contained in the original survival graphs and the event-free individual survival data were reconstructed on the basis of this. The meta-analytic EFS curves were then generated for the two treatments; the mean survival gain per patient was estimated and a simplified cost-effectiveness assessment was also carried out.

Results showed a statistically significant EFS difference in the experimental group versus the controls, with a mean survival gain of 1 year per patient. The authors concluded that allogeneic transplantation inpatients with ALL in CR1 improves EFS compared with chemotherapy or autologous transplantation.

Schaich 2007

This IPD meta-analysis24 was carried out on adult patients with AML and trisomy 8 (+8) (as a sole or an additional aberration) that were part of the same German trials as those in the study by Schlenk et al. 25 Trisomy 8 is among the commonest genetic aberrations seen in AML, but the prognostic significance of this and the best consolidation strategy for patients with this numerical aberration remain unclear. The aim of the meta-analysis was, therefore, to compare different consolidation strategies and to reveal new prognostic factors for survival.

Survival analysis showed no significant difference between patients with +8 as a sole aberration and those with +8 and one additional cytogenetic aberration. Results also showed no significant difference on OS among high-dose cytarabine, allogeneic SCT or autologous SCT. A positive impact of allogeneic SCT on RFS compared with other post-remission strategies was however demonstrated.

In addition, multivariate analysis including clinical, laboratory, cytogenetic and therapeutic variables identified age, extramedullary disease and the percentage of +8 positive metaphases at diagnosis as independent prognostic factors for OS. A hierarchical model was built by combining these three variables, whereby patients with +8 could be classified into low-, intermediate- and high-risk groups.

The authors therefore concluded that patients with AML with +8 do not form a homogeneous group and that allogeneic SCT may be the superior post-remission strategy for improving RFS.

Oliansky 2008

This was the sixth review17 to result from the ASBMT’s initiative and it evaluated the role of haematopoietic SCT in the therapy of adult patients with AML. Articles were identified by a MEDLINE search for literature published between 1990 and 2007. Selection and grading criteria were the same as for the previous reviews; study details were conveyed in text format and in summary tables.

Based on the survival data present within the studies, it was the authors’ conclusion that there was no significant advantage of autologous SCT over chemotherapy in CR1. For allogeneic SCT versus chemotherapy in CR1, a survival advantage was demonstrable for patients < 55 years old with high-risk cytogenetics but not for patients with intermediate or low risk.

Among other conclusions, it was stated that allogeneic SCT be preferred over autologous SCT if a matched related donor was available and that PBSCT was recommended over BMT. However, there was no evidence of a survival advantage in the following instances: unpurged versus purged SCT; T-cell replete versus T-cell depleted; comparison of two or more high-dose therapy conditioning regimens; and comparison of two or more myeloablative conditioning regimens.

IQWiG 2007

Among others, the aim of this German review,12 entitled Stem cell transplantation in adults with acute lymphocytic (ALL) or acute myeloid leukaemia (AML), full text not available in English, was to investigate in detail four aspects of SCT: allogeneic SCT with unrelated donors in ALL/AML; autologous SCT in ALL; non-myeloablative allogeneic SCT in ALL/AML; and SCT with in vitro manipulation of the graft in ALL/AML.

All study types that included a control group were considered, including retrospective studies. Case series were included for the group of patients with refractory disease who could not achieve remission with standard therapy. The end points selected were outcomes that enabled an assessment of patient-relevant therapy goals such as OS, DFS, therapy-related complications and health-related quality of life.

Literature searching was performed in MEDLINE, EMBASE and Cochrane CENTRAL in August 2005 and updated in December 2006. Several other sources were screened to identify further published studies. These included institutions that published evidence reports or were specifically involved in stem cell therapy or research, corresponding study groups, reference lists of relevant publications and reviews and congress proceedings.

No robust evidence showing a clear advantage for the interventions was available for any research question. Interestingly, indications of a superiority of SCT over chemotherapy could only be inferred for non-myeloablative therapy with a related donor in patients with AML. Likewise, the authors posited that the use of allogeneic SCT with dose-reduced conditioning might show an advantage in patients with refractory ALL or AML.

No evidence of an additional benefit was shown in patients with ALL or AML or their subgroups for the following subtypes or modifications of SCT: allogeneic SCT with myeloablative conditioning (compared with non-myeloablative conditioning) as well as in vitro manipulation of the graft in allogeneic or autologous SCT (compared with transplantation without manipulation of the graft). Likewise, no additional benefit of non-myeloablative therapy or autologous SCT (both vs chemotherapy) could be inferred from the data.

In patients with ALL, AML and their subgroups, no evidence of a benefit of allogeneic SCT with an unrelated donor versus chemotherapy could be inferred from direct comparative studies. However, the review states that evaluation of the available literature pointed to the possibility of a benefit as well harm in this regard.

Trials comparing allogeneic SCT with other therapeutic treatment strategies

Trials comparing autologous SCT with other post-remission treatment strategies

Evidence from existing reviews

Two existing systematic reviews (Johnson et al. 7 and Oliansky et al. 17) covered this decision problem but neither of them identified any CCT or RCT of adults with AML in CR2. The 1998 HTA report by Johnson et al. 7 found two retrospective cohort studies comparing allogeneic transplantation with chemotherapy for the consolidation of second remission in AML (Gale et al. 199170 and Gale et al. 199670), although the authors suspected that it was likely that some patients were included in both studies. However, in keeping with the decision made by the authors of the HTA report with regard to cohort studies, no conclusions were drawn from these studies.

The sole study retrieved by Oliansky et al. 17 was the 1996 one by Gale et al. ,70 which presented a retrospective analysis of 501 adult patients (≤ 50 years of age) with AML in CR2. Patients were treated with either chemotherapy (n = 244) or HLA-matched sibling donor allogeneic SCT (n = 257) at over 80 centres between 1980 and 1989. The two groups were similar with respect to sex and WBC count at diagnosis but transplanted patients were younger (median age 26 years vs 35 years), had briefer first remissions, and more had FAB M2 subtype than their chemotherapy counterparts.

The 3-year probabilities of TRM were 56% for transplant and 7% for chemotherapy and, although the probability for a 3-year leukaemia-free survival was higher in the transplant group compared with the chemotherapy group (26%, 95% CI 20% to 32% vs 17%, 95% CI 12% to 23%), this difference was not statistically significant. Outcomes adjusted for time to treatment, age, and duration of CR1 were also presented.

Evidence from primary studies not included in existing reviews

Our search for primary studies yielded one full paper and a meeting abstract. The paper by Thomas et al. 202 assessed the safety and efficacy of SCT, mainly autologous SCT, as consolidation therapy in patients with acute promyelocytic leukaemia (APL: AML FAB subtype M3) who relapsed and achieved CR2. Fifty adult patients with a first relapsed APL, of whom 39 had been previously treated with all-trans-retinoic acid, entered a multicentre trial of oral all-trans-retinoic acid until complete remission was achieved, followed by timed sequential chemotherapy. Forty-five patients were in complete remission after induction therapy and five patients died from infection during aplasia following chemotherapy. Of those who achieved complete remission, 11 had a familial HLA-identical donor and were allografted; the other 34 patients were scheduled for autologous PBSCT (ITT group) but transplantation was actually carried out in only 22 patients. Nine of the patients who did not undergo transplantation received maintenance chemotherapy.

The 3-year DFS rates of the ITT group and the group that received only chemotherapy were 63% and 51% respectively, while the 3-year DFS rate of actually autografted patients was 77% compared with 11% for patients who underwent allogeneic transplantation. Among the 17 autografted patients that remained in CR2, nine had already reached a longer CR2 than CR1. In addition, results of detection of PML/retinoic acid receptor-alpha by reverse transcription-polymerase chain reaction (RT-PCR) after autologous transplantation showed negative findings in eight of the nine patients tested. The authors concluded that, while allogeneic SCT may be too toxic for patients with relapsed APL, the clinical/molecular outcomes of autografted patients were encouraging.

The abstract by Vignetti et al. 203 presented the findings of 243 patients with AML in relapse enrolled into a GIMEMA-AIEOP (Associazione Italiana Ematologia Oncologia Pediatrica) prospective study that investigated BMT and immunotherapy with interleukin-2 (IL-2) in this group of patients. Of the 101 patients in complete remission after consolidation, 41 were transplanted (10 received allogeneic and 31 autologous BMT), 32 were randomised to receive or not receive IL-2 and 28 were ineligible either for transplant or for randomisation (and did not receive any other treatment until relapse and/or death). The criteria for transplantation or randomisation were not stated in the protocol.

Probabilities of DFS, projected at 3.5 years were 33% and 37% for autologous and allogeneic transplant respectively, 19% for the randomised group and 34% for the group of patients not eligible to receive any other treatment after consolidation. No definitive conclusions were reached but the authors pointed out that DFS was similar in the transplanted group (either allogeneic or autologous) and in the group of patients in complete remission not eligible for post-remission therapy.

Ongoing trials

No ongoing trial among this group of patients was identified.

Overall findings and discussion

No studies from higher up the evidence hierarchy were identified in the two systematic reviews covering this decision problem: the only evidence came from two retrospective cohort studies, one of which appeared in both reviews. Thus, while Johnson et al. 7 did not draw any conclusion from these studies (in keeping with their concerns about less robust studies), the recommendation by Oliansky et al. 17 that patients in CR2 receive an allogeneic SCT if a donor is available, otherwise autologous SCT, was based on expert opinion and clinical practice.

The results of the study by Thomas et al. 202 tend to favour autologous SCT over allogeneic SCT and chemotherapy but the study recruited only a small number of patients who make up a specific subtype of relapsed AML. The GIMEMA-AIEOP study reported by Vignetti et al. 203 found a higher probability of DFS at 3.5 years for allogeneic SCT compared with autologous SCT (p-values not stated) and seems to be in keeping with the recommendation of Oliansky et al. 17

Conclusion

Even though the review by Oliansky et al. 17 was fairly recent (2008), it identified only one retrospective study. Our search of primary studies identified a full text and a meeting abstract on the use of stem cells in relapsed AML, with the full paper describing the use of stem cells in the treatment of relapsed APL. Furthermore, no ongoing trials were identified. This may indicate a genuine lack of evidence for this decision problem. More primary research is therefore needed.

Trials comparing allogeneic SCT with other therapeutic options

RCTs comparing autologous SCT with other post-remission treatment strategies

Evidence from existing reviews

Only the 2007 review by Oliansky et al. 16 covered this decision problem and identified one study by Wells et al. 90 that was classified as level 2 evidence. That study presented outcomes of allogeneic BMT versus chemotherapy only, for 101 children with refractory or first relapse AML enrolled in the CCG 2951 trial. However, the Oliansky review16 concluded that there was a lack of evidence comparing allogeneic SCT with chemotherapy in children with AML in CR2.

Evidence from primary studies not included in existing reviews

Our search of primary studies identified only one RCT that addressed the comparison between autologous SCT and other post-remission treatment strategies. Webb et al. 210 described the results from MRC AML 10 trial, in which patients who achieved CR1 received autologous BMT, a matched unrelated donor allograft, a family donor allograft or chemotherapy. However the study did not present the results of comparisons between the treatment options but rather between stages of remission (CR1 and CR2).

Ongoing trials

No ongoing trial was identified.

Overall findings and discussion

The authors of the only review addressing this decision problem stated that there is a lack of evidence comparing matched related donor allogeneic SCT with chemotherapy in CR2, and therefore based their recommendations on the views of the expert panel, which favours the use of matched related donor allogeneic SCT, if available, over both chemotherapy and autologous SCT. Furthermore, the results of the identified primary study do not improve our existing knowledge, which is likely to remain the same as no ongoing trials were identified.

Conclusion

There is no good evidence from systematic reviews or primary studies with regard to this decision problem, nor is there a prospect for much information to emerge in the near future as no ongoing trials were identified. Given the limited evidence available, further primary studies may be required to inform an evidence-based decision on the role of SCT for AML in children in CR2 or beyond.

Trials comparing allogeneic SCT with other therapeutic options

RCTs comparing autologous SCT with other post-remission treatment strategies

Evidence from existing reviews

No DvND studies or RCTs of adults with ALL in CR2 were found in any of the identified reviews. Hahn et al. ,14 however, recommended SCT over chemotherapy, based on non-analytical studies with no direct comparative data, in which a sizeable fraction of patients achieved extended leukaemia-free survival compared with chemotherapy alone.

Evidence from primary studies

No RCTs or DvND studies addressing this decision problem were identified in our search of primary studies. One cohort study by Tavernier et al. 218 was identified. It examined the outcome of 421 adult patients who entered the LALA-94 trial and subsequently experienced relapse. 181 patients achieved CR2 and genoidentical allogeneic SCT was performed in 55 patients and three patients received donor lymphocyte infusions. Forty-four transplantations were performed from an unrelated donor (of which four were cord blood). The authors concluded that most adult patients with recurring ALL could not be rescued using currently available therapies, although allogeneic SCT remains the best therapeutic option.

Ongoing or recently closed trials

No ongoing/recently completed trials were identified.

Overall findings and discussion

No RCTs or CCTs were identified from the search of existing reviews. Although studies comparing allogeneic and autologous SCT were presented in one review, they were of very small sample sizes, from lower down the evidence hierarchy and arrived at conflicting conclusions. Therefore, the recommendation favouring SCT over chemotherapy by the authors of that review was based on non-analytical studies with no direct comparative data, in which a sizeable fraction of patients achieved extended leukaemia-free survival compared with chemotherapy alone. The conclusion of the primary study identified by us is in consonance with their stance.

Conclusion

It is apparent that robust evidence about the role of SCT in the management of adults with ALL in CR2 or relapse is lacking. In this regard, the importance of well-designed and adequately powered trials that will help to inform decision-making cannot be overemphasised.

Trials comparing allogeneic SCT with other therapeutic options

RCTs comparing autologous SCT with other post-remission treatment strategies

No RCTs comparing autologous SCT with other post-remission treatments were identified in either of the reviews. Our search for primary studies identified one RCT (the PETHEMA ALL-93 trial, described earlier), in which very high-risk patients without a matched sibling donor were randomly allocated to autologous SCT or chemotherapy. No significant difference was found between autologous SCT and chemotherapy groups in 5-year DFS [44% (95% CI,29% to 60%) vs 46% (95% CI 32% to 62%)] and OS [45% (95% CI 31% to 612%) vs 57% (95% CI 43% to 73%)]. Given the very limited volume of evidence from RCTs, no systematic review is currently needed. Evidence from the only RCT identified does not favour the use of autologous SCT, hence the scope for further primary studies may be limited.

Trials comparing allogeneic SCT with other therapeutic options

RCTs comparing autologous SCT with other post-remission treatment strategies

No RCTs comparing autologous SCT with other post-remission treatments in children with ALL in CR2 or relapse were identified in any of the reviews. Our search for primary studies and ongoing trials identified one study (MRC UKALL R1, mentioned earlier) in which patients with ALL in CR2 without a matched sibling donor were to be randomised to either autologous SCT or chemotherapy. However, the planned randomisation failed (only 9% of those eligible were actually randomised), highlighting the difficulties of running RCTs with patients who have relapsed from a previous trial. 226

PBSCT vs BMT

CBSCT vs BMT

Comparison of two or more myeloablative regimens

RIC regimen vs myeloablative conditioning regimen

Evidence from systematic reviews and meta-analyses

Two existing systematic reviews (Hahn et al. 14 and Oliansky et al. 17) covered this decision problem but neither of these identified any CCT or RCT that addressed the decision problem’s subquestions

However, the review by Hahn et al. 15 identified a retrospective study reported both by Granena et al. 182 and Garcia et al. 183 that addressed the comparison between autologous SCT with purged and unpurged stem cells. The study presented the result of 52 patients on purged or 23 patients on unpurged autologous BMT in CR1, CR2 and CR3 between 1987 and 1993. There was no statistically significant difference between the two groups with a 3-year DFS of 46.8% and 25.6% in the purged and unpurged groups, respectively (p = 0.13). There was an improvement in DFS in the purged groups for patients over 15 years old following multivariate analysis.

The review by Oliansky et al. 18 also reported on three retrospective studies (Miller et al. ,179 Gorin et al. 180 and Chao et al. 181) that compared autologous SCT with purging with autologous SCT without purging. The studies were carried out between 1982 and 1993 and included patients ranging from 50 (Chao et al. 181) to 294 (Miller et al. 179). The age limits for inclusion were 47 years (Gorin et al. 180) and 60 years (Miller et al. 179 and Chao et al. 181).

The OS for all the included studies was not stated except for the study by Miller et al. ,179 in which the 3-year OS was 63% in the purged BMT compared with 40% in the unpurged BMT, though the p-value was not stated. Similarly, the leukaemia-free survival was longer in the purged BMT than the unpurged BMT in the same study, with the p-value not stated either. However, there was no difference between the two intervention arms for leukaemia-free survival in the studies by Gorin et al. 180 and Chao et al. 181

Evidence from studies not included in previous reviews

Our search of primary studies did not yield any study that addressed this question.

Ongoing trials

No ongoing trial was identified among this group of patients.

Overall findings and discussion

Only retrospective studies identified from the two reviews provided evidence on the comparison between autologous purged and unpurged SCT; these are evidence of a lower level. 7 Neither Oliansky et al. 17 nor Hahn et al. 14 made any recommendations due to the poor quality of the available data.

Conclusion

There is a lack of evidence for this decision problem. The number of autologous SCTs carried out in patients with acute leukaemia has decreased substantially in recent years (see Figure 1 in Chapter 1). The lack of a current role for autologous SCTs concords with the findings of previous sections (Chapter I, Decision problems), which indicate that in general autologous SCTs offer no advantage over other treatment options in the management of acute leukaemia. Further studies of purging methods in autologous SCT is therefore not a priority.

Evidence from systematic reviews and meta-analyses

Only one review (Oliansky et al. 17) focused on this decision problem. Three studies were identified and one of them was an RCT. Wagner et al. 184 presented the outcomes of 101 adult patients with AML randomised to undergo an unrelated allogeneic BMT with either T-cell depleted marrow and ciclosporin or a T-cell replete BMT with methotrexate and ciclosporin. The follow-up period was 4.2 years. There was no difference between the two groups in terms of DFS; OS was not stated.

Evidence from primary studies

Our search for primary studies did not yield any additional papers.

Ongoing/recently closed trials

No ongoing trials comparing T-cell depleted with T-cell replete allogeneic SCT were identified.

Overall findings and discussion

Using results from the only RCT identified, Oliansky et al. 17 concluded that there was no evidence of a survival advantage with T-cell depleted grafts.

No additional publications were retrieved from our search of primary studies and no ongoing/recently completed trials were identified.

Conclusion

There is a palpable lack of evidence for this decision problem. More primary research in the form of well-designed and adequately powered RCTs is required.

Nature of decision problems

Number of included studies

There were 15 studies contributing information on costs, two of which were reviews7,248 and 13 primary studies. 249–261 Six studies contributed information on cost-effectiveness, two of which were reviews7,248 and four primary studies. 250,255–257 Finally, there were no economic models.

Costs – validity of included studies

Costs – results of included studies

The results of included studies are summarised in Table 18. Concerning costs it is very difficult to disentangle the findings from the two previous systematic reviews beyond the obvious expense of SCT in acute leukaemias.

This review confirms that but offers some more specific information. First, the high cost applies irrespective of the type of transplantation. However, there is considerable variability from study to study, seen particularly for the costs of allogeneic transplantation. This is likely to be driven by many factors such as chance variation, country, year, costing method, precise procedures employed, aspects of transplantation included in the costs, and variation in the nature of the populations. For allogeneic transplantation, costs of 50,000–100,000 are typical, irrespective of whether they are expressed in euros or US dollars. For autologous transplantation, the available costs are generally < US$50,000. In both cases it is notable that there are few estimates based on recent data, offering the possibility of considerable inaccuracy in the published estimates, either through the influence of inflation or improved efficiency through greater proficiency with the SCT techniques.

Given the variability, it is possible that studies offering internal comparisons may provide more robust cost estimates. On this basis it seems clear that both allogeneic and autologous SCT costs considerably more than conventional care, even when that conventional care involves high-dose chemotherapy. Comparing the costs of allogeneic with autologous transplantation in adult AML offers little certainty with mixed results across the four studies addressing this question. This is in marked contrast to the general level of costs for allogeneic and autologous transplants in individual studies, which suggests a very clear difference with autologous transplantation being less expensive. The discrepancy between the two sources of estimate on the relative cost of autologous and allogeneic transplantation suggests that this issue deserves further and closer investigation.

Finally, one study indicates greater costs associated with allogeneic transplantation using matched unrelated donors from those using matched siblings as donors.

Cost-effectiveness – validity of included studies

Cost-effectiveness – results of included studies

The results of included studies are summarised in Table 20.

Previous reviews have offered few specific conclusions about the cost-effectiveness of SCT. However even in this they are contradictory in that Johnson et al. 7 feels cost-effectiveness is driven by cost alone, in which case the high cost of transplantation mitigates against cost-effectiveness. In contrast Redaelli et al. 248 indicates that cost-effectiveness of allogeneic transplantation has been demonstrated.

In this review, although there is only one additional primary study not considered in the two previous reviews, more detailed assessment of the studies does reveal some additional insights. First, the quality of the assessments is noteworthy. All are based on extremely small numbers of patients. Further, all are extremely simple in approach, consistent with the standard of economic evaluation prevailing in health care at the time the studies were reported, but it is debatable whether such approaches could be considered adequate estimates of current cost-effectiveness. The date of the costings, mentioned in the previous section, alone indicates that this is likely to be a problem. It should also be noted that there is a methodological flaw in the methods used by Yu et al. 250 to estimate the incremental cost-effectiveness of the assessed interventions. While cost-effectiveness analysis should determine the additional cost for an additional unit of outcome (e.g. life-year), Yu et al. 250 divided the total mean cost associated with the treatment by the median years of survival resulting from this treatment. Correcting this error reveals that high-dose chemotherapy (HDC) appears to be dominant (both cheaper and more effective) against allogeneic transplantation. In the study by Barr et al. 255 when the ICER is expressed as cost per life-year saved (LYS) the cost-effectiveness looks acceptable, but when expressed as cost per quality-adjusted life-year (QALY) (with the arithmetical error corrected) SCT is unlikely to be considered cost-effective using current commonly used thresholds. Given the above, the unequivocal statement by Redaelli et al. 248 that the cost-effectiveness of SCT has been demonstrated is difficult to sustain in the face of the available data, even if there is widespread belief in its validity.

Finally, both previous reviews commented only on the cost-effectiveness of allogeneic transplantation. It should be noted that Barr et al. 255 also provide some information on the cost-effectiveness of autologous transplantation. With the same provisos indicated above, this suggests that the cost-effectiveness of autologous transplantation may be inferior to that of allogeneic transplantation. This may also need further investigation given the uncertainty noted about the relative costs of allogeneic and autologous transplantation for AML noted in the section on costs.

Economic models – validity and results

There were no economic models relating to DPs 1 and 2.

Summary for DPs 1 and 2

Ostensibly there appears to be a wealth of information on the costs and some information on cost-effectiveness of allogeneic transplantation in adults with AML. Unfortunately the age and quality of the studies mean that there is considerably more uncertainty than there might at first appear. The expense of allogeneic transplantation does seem to be confirmed, but what drives the observed marked variation and whether the figures fairly represent current costs is debatable. Previous statements about the proven cost-effectiveness of allogeneic transplantation appear to deserve careful re-examination using up-to-date methods and more credible sample sizes. Much less is known about the costs and cost-effectiveness of autologous transplantation and, again, past statements about the relative cost of allogeneic and autologous transplantation deserve re-examination.

Nature of decision problems

Number of included studies

There were two studies claiming to contribute information on costs, both of which were reviews. 7,248 The same two studies7,249 also claimed to contribute information on cost-effectiveness. There were no economic models.

Relevance, validity and results of included studies

Closer examination of the scope of the included studies in each of the reviews by Johnson et al. 7 and Redaelli et al. 248 (see Tables 15 and 16 in previous section) makes it clear that any general conclusions made in the reviews about costs and cost-effectiveness in acute leukaemias or AML are not generalisable to children as many included studies are clearly restricted to adults who would have been treated with different protocols and required different doses of drugs compared with children. Although a few of the included studies have mixed populations, children are always in the minority. We identified no new studies addressing costs or cost-effectiveness of transplantation in AML in children.

Summary for DPs 3 and 4

There are currently no published estimates of costs or cost-effectiveness relevant to this decision problem. Given the low frequency of AML in children the absence of specific information on costs and cost-effectiveness of transplantation may not be problematic. However, it does seem likely that both costs and cost-effectiveness may differ in important ways from adults.

Nature of decision problems

Number of included studies

There were two studies contributing information on costs, one of which was a review7 and one a primary study. 255 Three studies contributed information on cost-effectiveness, one of which was a review7 and two primary studies. 26,255 There were no economic models.

Costs – validity of included studies

Costs – results of included studies

The results of included studies are summarised in Table 22.

Based on the very limited information provided by this study, it appears that allogeneic transplantation in ALL in adults is as expensive as transplantation in AML, but control treatments may be more expensive. This leads to the cost saving associated with transplantation observed in the study by Barr et al. 255 The very small scale of this study must however be re-emphasised.

Cost-effectiveness – validity of included studies

For reasons analogous to cost, none of the general conclusions made by Johnson et al. 7 on cost-effectiveness in acute leukaemia were felt to be generalisable to DPs 5 and 6. The review made no specific conclusions about cost-effectiveness in ALL.

The relevance of the two included primary studies is confirmed in Table 23. The study by Orsi et al. 26 considered much greater numbers of participants than other cost-effectiveness studies by meta-analysing data from existing trials. In this respect it is more akin to an economic model than a primary evaluation of cost-effectiveness. It did not, however, use a computer-based model to integrate the data and so is considered in the cost-effectiveness section. A problem associated with the approach used by Orsi et al. ,26 is that the chemotherapy and autologous transplantation participants in the control arms of the meta-analysed trials could not be separated, so the comparator is distinct from those encountered in other economic evaluations in this review.

Cost-effectiveness – results of included studies

The results of the two primary studies are tabulated in Table 24.

The study by Barr et al. 255 indicates that, in adult ALL, allogeneic transplantation was considerably more cost-effective than the control treatment, which is not further specified but presumed to be chemotherapy. Specifically, allogeneic transplantation dominated control, which means that it was more effective at reduced cost. No indication is given about the uncertainty around results owing to small numbers of observations. However, the sensitivity analyses, which explore the effect of variation in several inputs such as use of discounting, extending the time horizon, increasing costs and changing the estimates of effect on survival, suggest that the results are sensitive to changes contrary to claims in the paper.

In the study by Orsi et al. 26 allogeneic transplantation appeared to improve survival but at increased cost. The ICER of €45,000 per LYS is broadly comparable to the cost-effectiveness thresholds of cost per additional QALY as they appear in the NICE Guide to the methods of technology appraisal. 279 This estimate indicates that allogeneic transplantation is less cost-effective than in the study by Barr et al. ,255 but that may be accounted for by the fact that the comparison group in Barr et al. 255 is chemotherapy and in Orsi et al. 26 control or autologous transplantation. Unfortunately, Orsi’s estimate of cost-effectiveness is undermined by uncertainty, the key sources of which are that the cost studies used to calculate the cost-effectiveness did not consider patients with ALL and that the results are highly sensitive to the variation in a single variable, the effect of which the investigators examined. An advantage of the study was a much greater number of participants on which estimates of effect on survival were based, afforded by meta-analysis of existing trials.

Economic models – validity and results

With the exception of the study by Orsi et al. ,26 discussed in the previous section, which might be considered by some as more akin to an economic model than a primary evaluation of cost-effectiveness, there were no economic models addressing DPs 5 and 6.

Summary for DPs 5 and 6

There is very limited evidence on the costs and cost-effectiveness of transplantation in ALL. This supports the unsurprising observation that, like transplantation in AML, allogeneic transplantation in ALL comes at a high cost. There is, however, also a suggestion that the high cost is justified by the level of additional clinical benefit in terms of improved survival, particularly relative to chemotherapy. However, the limited number of studies, the small size of one and the observed limitations of the other strongly suggest that this conclusion deserves further investigation. There is no information on the cost and cost-effectiveness of autologous transplantation.

Nature of decision problems

Number of included studies

There were two studies claiming to contribute information on costs, one of which was a review7 one a primary study. 262 One review7 also claimed to contribute information on cost-effectiveness. There were no economic models.

Costs – validity of included studies

Costs – results of included studies

The results of the single included cost study by Madero et al. 262 are summarised in Table 26. The most notable feature is that the absolute level of costs (US$14,000–20,000) is considerably lower than the costs encountered for the other decision problems. Whether this is because of the patients being children, the condition being ALL, the country in which health care was provided or the costing method employed is unknown.

Cost-effectiveness and economic models – validity and results of included studies

The only study that appears to contribute information on cost-effectiveness is the review by Johnson et al. 7 However, as before on closer inspection there are actually no studies included that are relevant to the cost-effectiveness of transplantation in children with ALL. There are thus effectively no studies contributing to the assessment of cost-effectiveness for DPs 7 and 8.

Summary for DPs 7 and 8

There is little information on the costs and cost-effectiveness of transplantation for ALL in children. The single small study available suggests that allogeneic transplantation costs are much lower. This observation, however, requires verification.

Nature of decision problem

What is the cost-effectiveness of BMT versus PBSCT versus CBSCT in the management of acute leukaemia. The need to separate evidence by given types of leukaemia (AML/ALL), populations (age/risk) and disease stage stipulated for clinical effectiveness was relaxed for cost-effectiveness.

Number of included studies

There were seven studies contributing information on costs, two of which were reviews7,248 and five primary studies. 252,253,262–264 Three studies claimed to contribute information on cost-effectiveness, two of which were reviews7,248 and one a primary study. 263 Finally, there was one full economic model. 263

Costs – validity of included studies

Costs – results of included studies

The table of comparative costs (Table 29) suggests reasonably consistently that using bone marrow as the source of stem cells increases the costs of transplantation by approximately US$5000–20,000 relative to peripheral blood. van Agthoven et al. 252 is the exception to this, showing no difference between bone marrow and peripheral blood. It is not clear what this difference might be due to, but inclusion or not of follow-up costs does not seem to be an explanation as all estimates included this to some degree. Despite the general consistency, the small size of the studies does require a note of caution to be added to the interpretation too. There is no information to gauge whether the observed differences could have been explained by chance alone.

There was only one study providing comparative costs on cord blood, which showed that cord blood also adds to the cost of transplantation relative to bone marrow or peripheral blood as sources of stem cells for transplantation. The limited amount of data on which the study is based again needs to be remembered.

Cost-effectiveness and economic models – validity and results of included studies

For reasons analogous to costs, the suggestion that the reviews by Johnson et al. 7 and Redaelli et al. 248 might contribute useful evidence on cost-effectiveness for DP9 was rejected. This left just one piece of original research by Costa et al. 263 to be considered. This was also an economic model, the only one to be encountered during the course of the review. The applicability of this study to DP9 can be seen from information in Table 28. However, it needs to be clearly noted that Costa et al. 263 provides estimates of cost-effectiveness for cord blood as source of stem cells for transplantation relative to bone marrow or peripheral blood. It does not deal with the more commonly encountered question of the incremental cost-effectiveness of bone marrow relative to peripheral blood as a source of stem cells in transplantation.

The only model-based study identified in this review was carried out by Costa et al. 263 The study was conducted in Canada from the perspective of the regional health-care system. The methodological validity of the study is appraised below.

Model structure

For the purposes of the study, Costa et al. 263 constructed a Markov model with five distinctive health states. To be specific, at the end of each 1-year cycle, patients in the model could survive without complications, survive with complications (chronic GvHD, leukaemia relapse, infection) or die. The model follows patients for 20 years.

The choice of a Markov type model is justified as such a model appears to be suitable for representing the underlying biological and clinical process of the considered disease (Barton et al. ). 280 Similarly, the choice of a 20-year time horizon is considered adequate for the population and disease in question. However, it is unclear whether the cycle length of 1 year is short enough to capture expected changes in disease progression. A shorter cycle might have been more appropriate for the specific decision problem.

Data

Data on treatment effectiveness were obtained from a review of the available literature. Relevant data were survival at different time intervals and complication rates. Supplementary data necessary for the analysis, such as estimates of life expectancy, was obtained from regional (Quebec, Canada) life-expectancy tables.

The systematic review appears to have been conducted in a systematic manner. Costa et al. 263 searched a number of major electronic bibliographic databases and followed widely accepted criteria to ensure compliance with ‘good practice’ in meta-analysis of observational studies. 281

Costa et al. 263 also made specific assumptions (GvHD and disease relapse were assumed to occur in year 1 if the timing was not specified in the original studies; no mortality was attributable to disease or transplantation procedure after year 5) which were justified on the grounds of available evidence from the literature. In the absence of data, they further assumed that the long-term survival of the patients followed that of the general population but was only 50% of the population estimate. The assumptions appear reasonable, but alternatives were explored by sensitivity analysis. Additionally, it is not clear whether benefits (life-years) were undiscounted to reflect their present value.

Cost input for the model was obtained from the host hospital’s financial department and from existing Canadian studies on BMT/ PBSCT. Relevant costs included direct medical costs of hospitalisation, inpatient and outpatient medications and costs related to nursing and physicians’ input. All costs estimates were discounted annually at 3%. Costa et al. 263 considered a comprehensive list of cost categories related to the disease and treatments in question. However, it is not clear whether the cost of identifying the unrelated donor is included.

Cost-effectiveness results and sensitivity analysis

The cost-effectiveness results of the analysis are reported in terms of ICERs (cord blood transplantation vs no transplantation; BMT/ PBSCT vs no transplantation). ICER estimates were accompanied by 95% confidence intervals. In addition, the Costa et al. 263 presented graphs depicting alternative point estimate ICERs resulting from univariate sensitivity analysis and cost-effectiveness acceptability curves (CEACs) from probabilistic sensitivity analysis (PSA).

To assess the impact of uncertainty, Costa et al. 263 carried out univariate sensitivity analysis on survival (10 years survival for transplantation group, 3 years for no transplantation) and cost (250% increase), as well as PSA on survival, complication rates and costs.

The methods employed for dealing with uncertainty in this study appear sound. Carrying out both deterministic (univariate) analysis and PSA allows a better representation of the uncertainty surrounding the model parameters and the effect of these parameters on the study results. 282 In addition, the probability distributions assigned to parameters subject to PSA (survival, complication rates, beta distribution) is consistent with current recommendations, although the use of the triangular distribution for costs may not represent the most appropriate choice. 283

However, sensitivity analysis in this study appears to concentrate on parameter uncertainty, although structural assumptions, such as allowing for only two states for patients without transplantation, and methodological assumptions, e.g. different discount rates for benefits, could have a significant impact on the study results and, thus, should have been explored in further sensitivity analyses.

At this point it is worth mentioning that Costa et al. 263 did not attempt to estimate outcomes that combine life expectancy with health-related quality of life, e.g. QALYs. According to the authors, this was owing to the absence of quality of life estimates in the literature.

As a general conclusion, the Costa et al. 263 study appears to be of sound internal validity. The methods used in the study are in broad agreement with suggested ‘good practice’ and with the NICE Guide to the methods for technology appraisal. 279 In addition, the review of the Costa et al. 263 model reveals that the main structure of the presented model can be used to assess different decision questions related to SCT. In brief, this would require: (1) modifying (or possibly adding) health states to facilitate relevant treatment comparisons; (2) reducing the cycle length to ensure that all the transitions between health states are captured; and (3) identifying relevant evidence in terms of treatment effectiveness, baseline patient progression, costs, etc. and populating the model with appropriate distributions representing the available evidence and the uncertainty surrounding that evidence. Once such evidence becomes available, adapting the Costa model, or constructing a similar structure, will be relatively straightforward. The results of the study by Costa et al. 263 are summarised in Table 30.

Notwithstanding the limitations of the model identified, many of which reside in the limitations of the data used for estimates of effectiveness and cost, it appears that both cord blood and bone marrow or peripheral blood are cost-effective as sources of stem cells in allogeneic transplantation of acute leukaemias, relative to no transplantation. The costs used in this model are, however, low relative to the costs of allogeneic transplantation identified earlier in the chapter, and it is debatable whether the ‘no transplantation’ option really represents the costs and outcomes that would be expected in the absence of SCT. Different assumptions about longevity in the no transplantation arm are considered in a univariate sensitivity analysis and small increases from the base case of 0.75 in the life-years assumed with no transplantation bring about a marked deterioration in the cost-effectiveness ratios.

The study by Costa et al. 263 relates only to allogeneic transplantation, so the cost-effectiveness of different stem cell sources in autologous transplantation does not appear to have been addressed.

Summary for DP9

There is limited information about the costs and cost-effectiveness of different sources of stem cells in transplantation for acute leukaemias.

For arguably the most important comparison between bone marrow and peripheral blood, the costs of bone marrow appear to be greater than (or at worst the same as) peripheral blood. There are no assessments of cost-effectiveness.

Comparing cord blood with bone marrow or peripheral blood suggests in one study that the cost of cord blood is greater and that all options appear to be cost-effective. There is, however, considerable uncertainty about this finding and the study relates only to allogeneic transplantation.

Nature of decision problem

What is the cost-effectiveness of various conditioning regimens, including standard myeloablative regimens and RIC regimens (mini-SCT)? The need to separate evidence by given types of leukaemia (AML/ALL), populations (age/risk) and disease stage stipulated for clinical effectiveness was relaxed for cost-effectiveness.

Number of included studies

There were two studies claiming to contribute information on costs, one of which was a review7 and one a primary study. 251 One review7 also claimed to contribute information on cost-effectiveness. There were no economic models.

However, there are no included studies in the review by Johnson et al. ,7 in which the costs and cost-effectiveness of myeloablative transplantation are compared with non-myeloablative (mini-SCT), so effectively the only published information to inform this decision problem is the primary study on costs by Cordonnier et al. 251

Costs – validity of included studies

Costs – results of included studies

As indicated in Table 32, the study by Cordonnier et al. 251 indicates that costs for myeloablative and non-myeloablative transplantation are similar. Early excess costs for myeloablative therapy were offset by excess costs for non-myeloablative regimens in the second 6 months owing to late complications and re-admissions. It needs to be noted that groups compared differ not only by whether the regimen was myeloablative but also by source of stem cells. The probable reduced cost of transplants using peripheral blood as a stem cell source, discussed in the previous section, suggests that this alone should have led to reduced costs in the group receiving the non-myeloblative regimen. Further caution is required in interpretation, given that this is a single study with a small number of participants.

Cost-effectiveness and economic models – validity and results of included studies

As already indicated, because the review by Johnson et al. 7 does not actually include any cost-effectiveness studies relevant to this decision problem, there is effectively no information on cost-effectiveness.

Summary for DP10

There is very limited information.

Tentatively, there is some evidence that myeloablative and non-myeloablative regimens have similarly high costs in allogeneic transplantation in AML. However, given the causes of uncertainty noted, this finding requires replication.

There is no evidence on the cost of myeloablative regimen in ALL or autologous transplantation, and a complete absence of any evidence at all on cost-effectiveness.

Nature of decision problem

What is the cost-effectiveness of autologous SCT with purging compared with autologous SCT without purging? The need to separate evidence by given types of leukaemia (AML/ALL), populations (age/risk) and disease stage stipulated for clinical effectiveness was relaxed for cost-effectiveness.

Number of included studies

There was one study, a review,7 claiming to contribute information on costs. The same review7 also claimed to contribute information on cost-effectiveness. There were no economic models.

However, there are no included studies in the review by Johnson et al. 7 in which the costs and cost-effectiveness of purging are compared with no purging, so effectively there is no published information to inform this decision problem.

Nature of decision problem

What is the clinical effectiveness and cost-effectiveness of T-cell-depleted allogeneic SCT compared with T-cell-replete allogeneic SCT? The need to separate evidence by given types of leukaemia (AML/ALL), populations (age/risk) and disease stage stipulated for clinical effectiveness was relaxed for cost-effectiveness.

Number of included studies

There was one study, a review,7 claiming to contribute information on costs. The same review7 also claimed to contribute information on cost-effectiveness. There were no economic models.

However, there are no included studies in the review by Johnson et al. 7 in which the costs and cost-effectiveness of T-cell depletion are compared with no T-cell depletion, so effectively there is no published information to inform this decision problem.

Main findings

This review revealed a significant paucity of evidence on most of the considered decision problems. While there exists a wealth of information regarding the costs and some information on cost-effectiveness of allogeneic transplantation in adults with AML (DPs 1 and 2), there is very limited evidence on the relative costs and cost-effectiveness of different techniques of SCT against further chemotherapy for AML in children (DPs 3 and 4) and for ALL in adults and children (DPs 5 and 6 and DPs 7 and 8 respectively).

Similarly, there is little evidence on the costs and cost-effectiveness of BMT versus PBSCT versus CBSCT in the management of leukaemia in general (DP9), with indications that the costs of BMT may be greater than those for PBSCT.

There is also very limited information relating to the cost and cost-effectiveness of various conditioning regimens, including standard myeloablative regimens and RIC regimens (DP10). Results from the only relevant study (Cordonnier et al. 251) suggest that myeloablative and non-myeloablative regimens have similarly high costs in allogeneic transplantation in AML; however, such a conclusion is subject to considerable uncertainty and requires replication.

Finally, no studies providing evidence on the cost-effectiveness of autologous SCT with purging compared with autologous SCT without purging (DP11) or the cost-effectiveness of T-cell-depleted allogeneic SCT compared with T-cell-replete allogeneic SCT (DP12) were identified, thus there is virtually no published information to inform these decision problems.

At this point, it must be pointed out that, when evidence exists, it is surrounded by considerable uncertainty. This is mostly owing to (1) the age of the included studies; (2) methods used in studies not necessarily according with widely accepted methods for conducting economic evaluations in the UK;279 and (3) likely differences in patterns of care and use of health-care services across different countries (UK, Netherlands, Spain, Canada, France, Taiwan).

Limitations of primary data

With regards to the primary data used in the reviewed cost and cost-effectiveness papers, it must be noted that, in the majority of the studies, evidence on costs and treatment effectiveness came from clinical studies with a small number of participants and, in most cases, a limited time horizon. Such limitations have a negative impact on the robustness of the reported results and contribute to uncertainty in the reported cost-effectiveness estimates.

It is also worth mentioning that there is a paucity of evidence on health-related quality of life associated with relevant health states. Only one of the reviewed studies attempted to express treatment effectiveness in terms of outcomes that take into account quality of life (QALYs). This is an important limitation as the use of QALYs facilitates decision-making on the use of available resources across different clinical areas and is suggested as the outcome of choice by existing guidelines (NICE 2008). 279

Conclusions in light of limitations

Bearing in mind the limitations in the available evidence, safe conclusions on what treatment is or is not cost-effective cannot be made. While allogeneic SCT appears to be effective compared with chemotherapy in adults and children with AML and adults and children with ALL with high and very high risk of relapse, whether this strategy is also cost-effective is debatable.

Suggestions for practice

The evidence in this review is not sufficient to allow any recommendations for practice to be made without further research.

Recommendations for research

Primary research is needed to provide more precise evidence on treatment effectiveness and impact on health-related quality of life, as well as up-to-date information on health service use and costs associated with transplantation from the perspective of the UK NHS. Decision-analytical models should be then constructed, possibly in the form of Markov models, to synthesise such information and provide evidence to inform the relevant decision problems.

Potential biases

The probability of finding a matched sibling donor is related to the patient’s age (availability of siblings for donation) and family size (number of siblings). There may be differences in these aspects between donor and no-donor groups, and results may be confounded if these factors are not adjusted for.

Another potential bias may occur during enrolment of patients into trials if patients and/or physicians already have knowledge of the availability of a matched sibling donor prior to enrolment. For example, patients who are at high risk of relapse may be more likely to participate in a trial if they have a donor and less likely to be enrolled if they do not because of the perceived benefit of allogeneic SCT and perceived lack of efficacy of alternative treatments. Conversely, patients with a good prognosis may be less likely to participate in a trial if they do have a matched sibling donor owing to the risk associated with allogeneic SCT. To reduce such enrolment bias, it is important that patients are enrolled into trials before the results of HLA typing are known to the patient and physician.

During the preparation of this report, it became apparent that determining whether an analysis reported in a research paper is based on a DvND comparison can be difficult owing to inadequate description of methods. Many studies that could have contributed data for DvND comparisons were (justifiably) excluded from existing reviews because reported analyses were based on actual treatments received. It is not clear whether the lack of reporting of results based on DvND comparisons in these studies was because of incomplete typing, lack of awareness of this method or selective reporting of more ‘statistically significant’ results. The last raises concern of potential publication bias, which does not appear to be a problem in a recent review published during the preparation of this report197 but should be carefully examined in future systematic reviews.

We also found several cases in the literature in which DvND comparison was clearly intended but the analyses actually performed did not strictly follow the principle of genetic randomisation. Examples were the inclusion of allogeneic SCT from matched unrelated donors in the donor group,113,115 inclusion of all patients without a sibling in the no-donor group, and classifying patients with a matched sibling who refused or was unable to donate stem cells in the no-donor group. 22,211 The actual impact of the potential bias that could be introduced is unclear. Practical issues relating to DvND comparison have been discussed elsewhere. 284

Limitations

Not all patients with a matched sibling donor actually receive allogeneic SCT. The statistical power for DvND comparison to detect a difference in the effectiveness between allogeneic SCT and other treatment options decreases as the proportion of patients who have a matched sibling donor but do not actually receive allogeneic SCT increases. DvND comparison therefore almost always underestimates the difference in effectiveness (if it exists) between allogeneic SCT from a matched sibling donor and other treatment options. In addition, treatments received in the ‘no-donor’ group varied between studies depending on the study protocols. Therefore the DvND comparisons may not be directly comparable between studies. The clinical application of such evidence may also be limited because of the varied and sometimes not well defined treatment option received by the ‘no-donor’ group.