High‐dose chemotherapy and autologous bone marrow or stem cell transplantation versus conventional chemotherapy for women with metastatic breast cancer

Abstract

Background

Although metastatic breast cancer is often responsive to conventional chemotherapy, it remains ultimately incurable. Autologous transplantation of bone marrow or peripheral stem cells (in which the patient is both donor and recipient) has been considered a promising technique because it allows much higher doses of chemotherapy to be used.

Objectives

To compare the effectiveness of high dose chemotherapy and autologous bone marrow or stem cell transplantation (autograft) with conventional chemotherapy for women with metastatic breast cancer. Outcomes were survival rates, treatment‐related toxicity and quality of life.

Search methods

We used the Cochrane Breast Cancer Group search strategy, adding these terms: bone marrow transplantation, stem cell transplantation, autologous stem cell support. The following databases were searched up to August 2010: MEDLINE, EMBASE, The Cochrane Library and ASCO (American Society of Clinical Oncology). We also searched the Cochrane Breast Cancer Group database and trial registries for unpublished trials, and checked the reference lists of articles found.

Selection criteria

Randomised controlled trials comparing the effectiveness of high dose chemotherapy and autograft with conventional chemotherapy for women with metastatic breast cancer.

Data collection and analysis

Six randomised controlled trials met the inclusion criteria. Two independent reviewers extracted data.

Main results

In total 437 eligible women were randomised to receive high dose chemotherapy with autograft and 413 were randomised to receive conventional treatment. There were fifteen treatment‐related deaths among the high dose group and two in the control (conventional dose) group (RR 4.08, 95% CI 1.40 to 11.93). There was no statistically significant difference in overall survival between the high dose and control groups at three years or five years. At five years of follow up, there was a statistically significant difference in event‐free survival, favouring the high dose group (RR 2.84, 95% CI 1.07 to 7.50). Toxicity was more severe in the high dose group. Only one of the trials has followed up all women for five years.

Authors' conclusions

Although there is evidence that high dose chemotherapy and autograft significantly improves event‐free survival compared to conventional chemotherapy in women with metastatic breast cancer there is no significant evidence of benefit in overall survival. High dose chemotherapy with bone marrow or stem cell transplantation should not be given to women with metastatic breast cancer outside of clinical trials.

Plain language summary

High dose chemotherapy and autologous bone marrow or stem cell transplantation versus conventional chemotherapy for women with metastatic breast cancer

Advanced or 'metastatic' breast cancer is cancer that has spread beyond the breast and underarm lymph nodes to other parts of the body. Although metastatic breast cancer is often responsive to conventional chemotherapy it does not provide a cure. The dose of chemotherapy that can be given to an individual is limited because it is unsafe in high doses and can seriously damage the bone marrow, creating high risk of serious infection. One treatment that was considered promising at the start of the 1990's was use of autograft, which involves transplantation of a woman's own bone marrow or peripheral stem cells to regenerate her bone marrow. Autografting allowed the administration of chemotherapy doses many times higher than could otherwise be used. This systematic review aimed to compare the evidence from randomised controlled trials comparing high dose chemotherapy with conventional chemotherapy.

This review identified six randomised trials including 437 women receiving high dose chemotherapy with autograft and 413 women receiving conventional chemotherapy treatment. In the group receiving the high dose chemotherapy, there were 15 treatment‐related deaths as opposed to two in the conventional chemotherapy arm. Although the high‐dose treatment did not increase overall survival at 5 years compared with conventional treatment, women on the high‐dose treatment survived significantly longer before experiencing recurrence of cancer. Treatment side‐effects were worse in the high‐dose group. On the basis of this review, the authors conclude that high dose chemotherapy with bone marrow or stem cell transplantation should not be given to women with metastatic breast cancer outside of clinical trials.

Background

Description of the condition

Breast cancer is the most common cancer occurring in women and is the primary cause of cancer death among women worldwide. The incidence of breast cancer is increasing in most countries and although survival rates have improved in Europe and North America this improvement relates mainly to disease diagnosed at an early stage (Bray 2004). Advanced or 'metastatic' breast cancer is cancer that has spread beyond the breast and underarm lymph nodes to other parts of the body. Although metastatic breast cancer is often responsive to conventional chemotherapy or endocrine (hormonal) treatments it remains ultimately incurable (Stockler 2000).

Description of the intervention

Researchers from the 1970s onwards described a dose‐response relationship in the action of chemotherapy drugs against cancer (Frei 1980); thus it has been observed that in the treatment of breast cancer the percentage of women responding to therapy is positively associated with the dose intensity of the drugs received, dose intensity being a function of both the dose and the timing of the chemotherapy regimen (Hryniuk 1984). The technique of bone marrow or peripheral stem cell transplantation using the patient's own tissue (a treatment known as autograft) was considered an exciting development because it addressed the problem of life‐threatening bone‐marrow toxicity which had previously limited the dose of chemotherapy drugs that could be safely given.

How the intervention might work

The use of autograft with chemotherapy permits the administration of chemotherapy doses many times higher than could otherwise be used. Results of animal studies were encouraging, as were non‐randomised patient trials. During the 1980s and 1990s several non‐randomised patient trials reported prolonged survival times using high dose chemotherapy and autograft for women with advanced breast cancer (Antman 1992Peters 1988Williams 1992).

Why it is important to do this review

Since clinical trials started, a large number of women with breast cancer have been treated with high dose chemotherapy and autograft and in the 1990s over 41,000 people in the United States underwent the procedure. However there is scant clinical evidence of its efficacy, with evidence from non‐randomised studies being criticised for participant selection bias as well as other design weaknesses (Eddy 1992; Mello 2001). The first randomised trials, sponsored by the U.S. National Cancer Institute, began in 1991. Several randomised controlled trials have now been completed and the aim of this review is to consider them.

Objectives

To compare the effectiveness and safety of high dose chemotherapy and autologous bone marrow or stem cell transplantation with conventional chemotherapy for women with metastatic breast cancer.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials that compare the effectiveness of high dose chemotherapy and autologous bone marrow or stem cell transplantation (autograft) with conventional chemotherapy for women with metastatic breast cancer.

Types of participants

Women with metastatic breast cancer either at initial diagnosis or as a recurrence.

Types of interventions

High dose chemotherapy and autologous bone marrow or stem cell transplantation with conventional chemotherapy, regardless of duration of therapy. Not limited to first line therapy. High dose chemotherapy was defined as chemotherapy at a dose sufficient to require bone marrow transplantation or stem cells to restore immune function.

Types of outcome measures

Studies included at least one of the following outcomes:

• Treatment related mortality;

• 1, 2, 3, 5 or 7 year overall survival;

• 1, 2, 3, 5 or 7 year event‐free survival (surviving with no evidence of recurrence of disease);

• Morbidity, including non‐haematological toxicities such as nausea and vomiting, white cell measures, new malignancies

• Quality of life

• Time to tumour progression and

• Overall survival time

Search methods for identification of studies

Electronic searches

1. The Cochrane Breast Cancer Group's Specialised Register was searched utilising their search strategy with the addition of the following terms: bone marrow transplantation, stem cell transplantation, stem cell support, autologous stem cell support.

2. The Cochrane Central Register of Controlled Trials was searched in August 2010.

3. The following electronic databases were searched for controlled clinical trials: MEDLINE (1966 to September 2010), EMBASE, (1980 to September 2010), PsycINFO (1984 to September 2010) and Cinahl (1982‐September 2010), using the search strings recommended in the Cochrane Reviewer's Handbook with the addition of specific terms (see Appendix 1 and Appendix 2 for EMBASE and MEDLINE search strategies).

4. The database of abstracts on the treatment of breast cancer from the annual meetings of ASCO (American Society of Clinical Oncology) was searched from 1995 to October 2010, using the following string: '"randomized*"' and '"breast*"' and '"stem*" OR "graft*" OR "transplantation*" OR "marrow*" OR "autograft*"OR "ABMT*" OR "PBSC*" OR "PBPC*" OR "progenitor*"' in abstract .

5. The following websites were searched in October 2010: Breast Cancer International Research Group, the National Cancer Institute and Current Controlled Trials.

6. The WHO International Clinical Trials Registry Platform (ICTRP) was searched in September 2010 (see Appendix 3 for WHO ICTRP search strategy).

Searching other resources

Reference lists of articles retrieved were hand searched.

Data collection and analysis

Selection of studies

One review author scanned the titles and abstracts of articles retrieved by the search and removed those that were very clearly irrelevant. The full text of all
potentially eligible studies was retrieved. Two review authors independently examined the full text articles for compliance with the inclusion criteria and
selected eligible studies. Review authors corresponded with study investigators if required, to clarify study eligibility. Disagreements were resolved by consensus or by discussion with a third review author.

Data extraction and management

Data were extracted from eligible studies using a data extraction form designed by the review authors. Where studies had multiple publications, the
main trial report was used as the reference and additional details supplemented from secondary papers. Review authors corresponded with study
investigators in order to resolve data queries. Two review authors independently extracted the data with any disagreements resolved by a third review author.

Assessment of risk of bias in included studies

The included studies were assessed for risk of bias using the Cochrane risk of bias assessment tool to assess random sequence generation; allocation concealment; completeness of outcome data, selective outcome reporting and other potential sources of bias. Two authors assessed these domains, with any disagreements resolved by consensus or by discussion with a third author. The conclusions were presented in the Risk of Bias table and incorporated into the interpretation of review findings by means of sensitivity analyses (see below).

The following methods of random sequence generation were considered adequate:

• random number table

• computerised random number generator

• coin tossing

• shuffling cards or envelopes

• throwing dice

• drawing of lots

The following methods of allocation concealment were considered adequate:

• central allocation, including telephone, web‐based, and pharmacy‐controlled randomisation

• sequentially numbered drug containers of identical appearance

• sequentially numbered, opaque, sealed envelopes

Outcome data were considered as complete if either of the following applied:

• all women randomised were analysed

• data were imputed for those missing

A study was assessed as being free of the risk of selective outcome reporting if both the following applied:

• the published report included all expected outcomes

• outcomes were reported systematically for all comparison groups, based on prospectively collected data

Measures of treatment effect

For dichotomous data, the numbers of events in the control and intervention groups of each study were used to calculate relative risk ratios (RR) using the Mantel‐Haenszel method for fixed effects, with 95% confidence intervals (CIs). TIme to event data were reported descriptively for overall and event‐free survival.

Dealing with missing data

The most complete data set feasible was assembled. The data were analysed on an intention‐to‐treat basis as far as possible and attempts were made to obtain missing data from the original investigators. Some studies had multiple publications and reported outcomes at differing follow‐up times. The data included in this review are the most mature results available for each study. Most of the available data were immature (e.g. five year outcomes were estimated when not all participants had been randomised for five years). Where triallists reported survival rates based on immature data, these rates were used in our tables of comparison. A note has been made in the Characteristics of included studies where applicable. Where results were available only as percentages or only presented in graphs, the numerators and denominators were back‐calculated from the data available and the tables of comparison completed accordingly. This has been noted in the Characteristics of included studies where applicable

Assessment of heterogeneity

Clinical heterogeneity between trials was assessed through evaluation of potential differences between participants, interventions and outcomes within each study.

Assessment of reporting biases

In view of the difficulty in detecting and correcting for publication bias and other reporting biases, the authors aimed to minimise their potential impact by
ensuring a comprehensive search for eligible studies and by being alert for duplication of data. It was planned that if there were ten or more studies in an analysis, a funnel plot would be used to explore the possibility of small study effects (a tendency for estimates of the intervention effect to be more beneficial in smaller studies).

Data synthesis

Where trials appeared to be clinically comparable, data were pooled to obtain a relative risk (RR) using the Mantel‐Haenszel method for fixed effects. Quality of life data were not combined because only descriptive data were available. No individual patient data were available to allow pooling of time‐to‐event data.

Subgroup analysis and investigation of heterogeneity

Statistical heterogeneity was assessed by inspecting the scatter in the data points on the graphs and the overlap in their confidence intervals and, more formally, by checking the I² quantity (Higgins 2003). This quantity describes the percentage of total variation across studies that is due to heterogeneity rather than chance. Interpretation of a given degree of heterogeneity will differ according to whether the estimates show the same direction of effect but it was planned to tentatively assign adjectives of low, moderate and high heterogeneity to I² values of 0% to 25%, 26% to 74% and 75% to100% respectively. A priori, it was planned to look at the possible contribution of differences in trial design to any meta‐analyses with an I² value of 50% or more.

No sub‐group analyses were conducted.

Sensitivity analysis

There were insufficient studies to conduct sensitivity analyses.

Results

Description of studies

Included studies

Six trials were identified that included high dose chemotherapy and autologous bone marrow or stem cell transplantation for the treatment of metastatic breast cancer. One of the trials was available as conference abstract and slide presentation only (IBDIS 2003). Additional information was requested from all authors and replies were received from the principal investigators of all studies although not all provided data for the specific time points for follow up nor for all questions relating to study quality. For details on trial characteristics see the Characteristics of included studies and Table 1; Table 2 and Table 3.

1. High dose chemo regimens for included studies (all doses per m2 unless otherwise).

Drug	ECOG 2000	IBDIS 2003	NCIC 2001	PEGASE 03 2002	PEGASE 04 1999	Schmid 2005
Induction phase	As for control arm	As for control arm but 3 cycles only	6 cycles of anthracycline or taxane based chemotherapy as for control	As for control arm	4‐6 courses as for control arm	No treatment in common with control arm
Cyclophosphamide	6000 mg/m2 by continuous infusion over 4 days	Phase 2: 6gm	6000 mg/m2 divided into 4 equal doses over 4 days	6000 mg/m2 by continuous infusion over 4 days	120 mg/kg	4.4 gm/m2 X 2 cycles 6 weeks apart
Mitoxantrone			70 mg/m2 divided into 4 equal doses over 4 days		45mg/m2	45mg/m2 X 2 cycles 6 weeks apart
Thiotepa	500 mg/m2 by continuous infusion over 4 days	Phase 2: 800mg		800 mg/m2 by continuous infusion over 4 days
Carboplatin	800 mg/m2 as continuous infusion over 4 days	Phase 1: AUC 18	1600 mg/m2 divided into 4 equal doses over 4 days
Melphalan					140 mg/m2
Etoposide		Phase 1: 1200mg				2.5 gm/m2 X2 cycles 6 weeks apart
Ifosphamide		Phase 1: 12000 mg

2. Control arm ‐chemotherapy doses.

Drug	ECOG 2000	IBDIS 2003	NCIC 2001	PEGASE 03 2002	PEGASE 04 1999	Schmid 2005
Description		Progressed to phase 2 if no disease progression during phase 1	6‐9 cycles anthracycline based chemotherapy, or taxane based for women previously exposed to anthracyclines	Given before randomisation:	3‐9 cycles of conventional dose chemotherapy, 90% anthracyline‐based. Median cycles 6
cyclophosphamide	Induction: 100 mg/m2 X orally for 14 days X 4‐6 cycles. Post randomisation: same regime for up to 24 cycles	Phase 2: 600mg X2 X4 cycles		500mg/m2 every 21 days X 4 cycles
methotrexate	Induction: 40mg/m2 (if prior >400mg/m2 doxorubicin) X 4‐6 cycles. Post randomisation: same regime for up to 24 cycles	Phase 2: 40mg X 2 X 4 cycles
fluorouracil	Induction: 500 mg/m2 X twice per cycle X 4‐6 cycles. Post randomisation: same regime for up to 24 cycles	600 mg X 2 X 4 cycles		500mg/m2 every 21 days X 4 cycles
epirubicin			If used, total dose 840 mg/m2	100 mg/m2 every 21 days X 4 cycles
doxorubicin	Induction: 30 mg/m2 X twice per cycle X 4‐6 cycles (if prior <400‐500 mg/m2)	Phase 1: 50mg/m2 X 4 cycles	If used, total dose 450 mg/m2			60 mg/m2 X 6‐9 cycles
paclitaxel						200 mg/m2 X 6‐9 cycles
mitoxantrone
etoposide
prednisone	Optional with induction chemotherapy
doxetaxel		75mg/m2 X 4 cycles

3. Study quality of the included studies.

Study	Concealed allocation	Randomisation method	Complete follow up	Prognosis balanced	Control gp contamina	Any other bias?	Median follow up
ECOG 2000	Yes	Computer randomised. Stratified for response to initial chemo, site of metastasis, age, oestrogen receptor status, & study location	184/199	Yes except more bone metastases in control group (p=0.04)	3/83 received high dose on relapse	Unequal numbers on arms due to numerous stratification factors; 20 (10%) refused treatment assignment	69.5 months
IBDIS 2003	Yes	Centrally randomised	Yes	Yes	None mentioned	More women in the high dose arm completed all cycles of chemotherapy (84% vs 74%)	47 months
NCIC 2001	Method not stated	Method not stated	Yes	Yes	2/111 received high dose off study	21/112 (18.7%) of high dose group did not receive assigned treatment.	48 months
PEGASE 03 2002	Method not stated	Method not stated	Yes	Yes	2/91 received high dose on relapse	9/89 (10%) of high dose group did not receive assigned treatment	48 months
PEGASE 04 1999	Method not stated	Method not stated	Yes	More women in the HDC group had pulmonary (48% versus 21%; p=0.009)and cerebral metastases (6% versus 0%) than the control group	Not stated	1. Only 12/29 (41%) of control arm received chemotherapy after randomisation. 2. Significantly more lung lesions in high dose arm: these are more chemosensitive than other visceral tumours.	Median time of follow‐up 92 months in high dose arm, arm, 87months in control arm.
Schmid 2005	Yes	Stratified for menopausal status and ER status, blocks of 4, computer generated	Yes	Yes	5/44 received high dose on relapse	7/48 (14%) of high dose group did not receive assigned treatment	52 months

Trial Design

All six trials were randomised and compared high dose treatment to conventionally‐dosed chemotherapy. One study (Schmid 2005) stated that cross‐over to high dose chemotherapy was pre‐planned for patients in the control group with disease progression. One (PEGASE 03 2002) stated that treatment at disease progression would be decided by the treating physician: this could include high dose chemotherapy and autograft for women on the control arm. The other trials did not state a plan for those in the control group with disease progression. One study (NCIC 2001) stated that maintenance chemotherapy could be used for the control arm at the discretion of the trial investigator.

All the trials were multi‐centred. Two were conducted in France (PEGASE 03 2002; PEGASE 04 1999), one in the United States (ECOG 2000) and one in Canada (NCIC 2001). Two trials were international, one with centres in Ireland, the UK, Switzerland, Spain and Greece (IBDIS 2003) and one with centres in Germany and Austria (Schmid 2005).

Five studies provided power calculations: ECOG 2000 was designed at baseline in 1991 to have a power of 90% to detect a doubling of median survival with high dose chemotherapy and autologous bone marrow transplant. This required 348 women to be randomised. As recruitment problems were encountered, the design was revised in 1996 and new power calculations were given using 85% power with 164 women required to be randomised. Eventually 188 women were randomised. NCIC 2001 was powered to have 80% power to detect an 18% difference in survival at 2 years. PEGASE 03 2002 was powered to detect a 20% improvement in event‐free survival at three years and 180 women were required to be randomised. PEGASE 04 1999 was powered to detect a 25% improvement in event‐free survival at two years, with 156 women required to be randomised. Only 61 women were eventually randomised because of recruitment problems. Schmid 2005 failed to meet accrual targets and was terminated early, thus the planned two‐part statistical design was abandoned and exploratory data analyses only were applied

Description of Participants

All the included participants had metastatic breast cancer. In four trials participants had to be under 60 years old, while the NCIC 2001 abstract specified only that women must be aged at least 16 years. All studies excluded women who had had prior chemotherapy for metastatic breast cancer. See Characteristics of included studies and Table 4 for more details.

4. Prognostic factors in the included studies.

Study ID	Median Age	Metastatic sites	Estrogen rec +ve	Adjuvant chemo	Induction response	Adjuvant hormones	Local Radiotherapy	No. metastatic sites
ECOG 2000	46	Soft tissue and nodes 51%, bone 44%, lung & pleura 34%, liver 28%	46%	56%	Complete response 24%, partial response 76%	32%
IBDIS 2003	45	Lung/pleura 35%, liver 39%, node only 12%, single site only 39%	30%	23%	Protocol had no chemotherapy before randomisation	47%	Not stated
NCIC 2001	47	Breast 25%, bone 58%, lung 18%, liver 27%	31%	61%	Complete response 12%, partial response/non‐measurable 88%.	32%
PEGASE 03 2002	46	Bone 50%, lung 31%, liver 53%, distant lymph nodes 37%, normal bone biopsy 89%	HDC 59%	46%	Complete response 6%, partial response 87%, other/unknown 5%	24% on tamoxifen	Providing that less than one third of the haemopoietic areas were treated within 6 weeks before induction
PEGASE 04 1999	44	Bone 46%, lung 31%, cerebral 3%, only one involved site 20%	similar between groups	54%	Complete response 21%, partial response 79%	20% (a further 7% uncertain)	62%
Schmid 2005	50	Soft tissue and nodes 13.6%, liver 60%, lung 42%, bone 48%	70.7%	54%	Not applicable	Not stated

Description of Interventions

Five trials administered initial conventionally‐dosed chemotherapy before starting high dose treatment (ECOG 2000, NCIC 2001; PEGASE 03 2002; PEGASE 03 2002). In four of these trials all participants had to demonstrate complete or partial response to chemotherapy before randomisation, while in the fifth (IBDIS 2003) participants were randomised before chemotherapy commenced but only participants responding to initial chemotherapy proceeded to the high dose treatment. The rate of complete response to initial chemotherapy varied between 6% and 24%. In four studies, both groups received the same initial chemotherapy and the control group had additional conventionally‐dosed chemotherapy, while in one trial (PEGASE 03 2002) the control group had no further chemotherapy after randomisation. One study (Schmid 2005) compared 'upfront' high dose chemotherapy (with no induction course) with conventional chemotherapy.

In all trials high dose chemotherapy was administered with blood progenitor cell support and all trials stated that GCSF (granulocyte colony‐stimulating factor) was used to stimulate mobilisation of stem cells from the bone marrow. Women on both arms of the trial received GCSF in one study (IBDIS 2003).

The regimen of high dose chemotherapy differed for each trial: see Characteristics of included studies, Table 1 and Table 2 for details.

Description of Outcomes

Five studies have published their results (ECOG 2000; NCIC 2001, PEGASE 03 2002; PEGASE 04 1999; Schmid 2005). As noted above, most of the outcomes reported were based on immature data: median follow up times and the range of follow up times, where this information was available, are presented in Table 3.

All studies presented data on the median length of time for which women survived from randomisation to disease recurrence or death, and all had details of treatment‐related deaths and adverse events.

Outcomes are presented at five years for three trials (ECOG 2000; IBDIS 2003; PEGASE 04 1999) and at three years for three trials (NCIC 2001, PEGASE 03 2002; Schmid 2005). No seven year data were available. Data were mature for only one trial (PEGASE 04 1999).

Quality of life has so far only been reported by ECOG 2000 and NCIC 2001.

Excluded studies

Three studies were excluded, one (Madan 2000) because it was designed to compare different time frames for high dose therapy rather than comparing high dose chemotherapy with conventional treatment, one (Vredenburgh 2006) because the study design did not include a non transplantation/conventional treatment arm and one (Bezwoda 1995) due to breach of scientific honesty. See Characteristics of excluded studies.

Risk of bias in included studies

Allocation

All trials stated that they were randomised controlled trials with women randomised to either high‐dose chemotherapy or to a control group that received conventional dose chemotherapy. Four trials used stratification. In ECOG 2000 trial stratification was by partial or complete response to induction chemotherapy, predominant site of distant metastasis, age, oestrogen receptor status and participating centre. The ECOG 2000 triallists commented that the numerous stratification factors in this trial resulted in skewed assignment to high‐dose and conventional‐dose chemotherapy groups, with unequal numbers in the two groups. In NCIC 2001, participants were stratified by response to and type of induction chemotherapy, disease site and hormone receptor status. In Schmid 2005 trial participants were stratified by menopausal status and hormone receptor status. In PEGASE 04 1999 stratification was by participating centre.

Three trials (ECOG 2000, IBDIS 2003, Schmid 2005) were given a score of A for allocation concealment as they used central telephone/fax allocation, indicating that treatment allocation was carried out in a central trial office remote from the clinical centres. The other trials received a score of B for unclear allocation concealment as no details were given on allocation in spite of requests for information from the authors.

Blinding

Blinding was only mentioned in one trial (ECOG 2000), which stated that at the time of the second interim analysis (conducted in November 1998 after 94 deaths had occurred) the data‐monitoring committee recommended that the study be unblinded because the likelihood that the study would show a significant difference in favour of high dose chemotherapy at the final analysis was very low. The study author noted that single blinding of the outcomes was conducted up to this point.

Incomplete outcome data

Although a high proportion of women did not complete their allocated treatment, particularly in the high dose arms, most studies used intention‐to‐treat analysis. Thus in most cases all women randomised were included in analysis in the groups to which they were randomised, regardless of the treatment they received. With the exception of one small trial (PEGASE 04 1999), all studies included some immature data for survival outcomes (i.e. not all women had completed follow‐up).

ECOG 2000
This trial randomised 199 women and analysed 184, including nine who were ineligible for minor reasons. A further 15 were found to be ineligible for major reasons and were excluded from analysis. Twenty women (10%) did not complete their allocated treatment, six in the high dose arm (of whom five received either no chemotherapy or conventional chemotherapy and one underwent an alternative high dose therapy) and 14 in the control arm (of whom ten underwent high dose treatment, three received no chemotherapy and there were no data for one). Three women in the control group had high‐dose chemotherapy when they relapsed: it is not clear whether this was part of the treatment protocol.

IBDIS 2003
The study intended to randomise 264 women, but accrual failed and ultimately only 110 were randomised, all of whom were included in analysis by intention‐to‐treat. 46/56 (82%) women randomised to high dose treatment completed all five cycles.

NCIC 2001
Twenty‐one of 112 women randomised to high dose chemotherapy did not proceed with the treatment: 11 had progressive disease, six refused, two experienced toxicity and two had inadequate stem cell collection. Two women in the control group received high dose chemotherapy with autograft off the study. Analysis was by intention to treat for 111/112 women. The single participant excluded from analysis was ineligible due to failure to respond to induction chemotherapy.

PEGASE 03 2002
179 women were randomised. All women randomised were included in analyses of survival rates. Nine women (10%) randomised to the high dose group did not receive high dose chemotherapy and three women randomised to the control group had high dose chemotherapy with autograft.

PEGASE 04 1999
This trial randomised 61 women and analysed results by intention‐to‐treat. Two women randomised to the high dose arm (7%) did not receive the prescribed treatment due to rapid disease progression. Only 12 out of 29 women randomised to the control arm had chemotherapy subsequent to randomisation although the treatment protocol proposed that they have two to four additional courses of treatment as maintenance therapy.

Schmid 2005
This trial randomised 93 women and analysed them all by intention‐to‐treat. Seven of the 48 (15%) women randomised to the high dose group did not proceed to high dose chemotherapy. All women who were randomised to standard chemotherapy received at least one course of chemotherapy; it is not stated what proportion completed all their allocated treatment. Five women randomised to the control group were eligible to have high‐dose chemotherapy with autograft when they relapsed and two of them did so.

Selective reporting

All studies reported all expected outcomes.

Other potential sources of bias

Observation bias: In NCIC 2001, regularly scheduled radiological assessment of disease status was not required after chemotherapy completion: hence assessment of progression‐free survival may have been subject to observer bias.

Prognostic imbalance: There was an imbalance between the groups in prognostic factors at baseline in the following studies:
ECOG 2000
The two groups were well balanced for prognostic factors apart from a statistically significant (p=0.04) imbalance with respect to bone metastasis: 37% of the high dose group and 53% of the control group had bone metastasis
PEGASE 03 2002
The two groups were well balanced apart from a higher number of women with hormone‐receptor positive disease in the high dose group (oestrogen receptor positive: high dose group 66%, control group 52%; progesterone receptor positive: high dose group 60%, control group 42%).The high‐dose group also had a higher proportion with metastases to liver (57% versus 50%) and distant lymph nodes (41% versus 34%), but lower proportion of lung metastases (30% versus 34%).
PEGASE 04 1999
There was imbalance between the two groups with respect to pulmonary metastases (high dose group: 46%, control group 14%). This may have favoured the high dose arm as pulmonary lesions are more sensitive to chemotherapy than other visceral lesions. The duration of chemotherapy received also favoured the high‐dose group since, as noted above, only 12 out of 29 women randomised to the control arm had chemotherapy subsequent to randomisation.

Effects of interventions

Four hundred and thirty eight women were randomised to receive high dose chemotherapy with autograft (experimental group) and four hundred and thirteen women were randomised to receive conventional treatment (control group). Results for each outcome are described below. For positive outcomes (overall and event‐free survival), a risk ratio above one favours the high‐dose chemotherapy group, while for negative outcomes (treatment related death) a risk ratio above one favours the conventional chemotherapy group.

Treatment‐related deaths

There were 15 deaths among 437 women who received high dose chemotherapy with autograft and 2 deaths reported among 413 women randomised to the control group. This was statistically significant, favouring the control group (RR 4.08, 95% CI 1.40 to 11.93).

Overall survival

Three year follow up

In a meta‐analysis of three trials (NCIC 2001; PEGASE 03 2002; Schmid 2005), there was no statistically significant difference between the groups in overall survival at three years (RR 1.10, 95% CI 0.86 to 1.42).

Five year follow up
In a meta‐analysis of three trials (ECOG 2000, IBDIS 2003; PEGASE 04 1999), there was no statistically significant difference between the groups in overall survival at five years (RR 1.50, 95% CI 0.96 to 2.34).

Event‐free Survival

Three year follow up

In a meta‐analysis of two trials (PEGASE 03 2002; Schmid 2005), there was no statistically significant difference between the groups in event‐free survival at three years (RR 2.35, 95% CI 0.86 to 6.41). There was moderate heterogeneity in this result (I² statistic 70%).

Five year follow up

In a meta‐analysis of three trials (ECOG 2000, IBDIS 2003, PEGASE 04 1999), there was a statistically significant difference in event‐free survival at five years, favouring the high dose group (RR 2.84, 95% CI 1.07 to 7.50).

Morbidity

Four studies provided information on morbidity. As expected, women who received high dose chemotherapy with autograft were far more likely than controls experience severe haematological toxicity (such as severe leucopenia, thrombocytopenia and anaemia) and also sepsis, gastrointestinal toxicity (such as nausea, vomiting and diarrhoea) and/or organ toxicity (such as cardiac, pulmonary, hepatic and central nervous system complications). In most cases these toxic effects were temporary and of short duration. See Table 5.

5. Morbidity ‐ descriptive data.

Study ID	Haemopoietic	Gastrointestinal	Pulmonary	Cardiac	Renal	Neuro	Infection	Other	Triallist comment
ECOG 2000	High dose group: leukopenia 96%, thrombocytopenia 95%, anaemia 69% CMF: leukopenia 52%, thrombocytopenia 5%, anaemia 6%	High dose: Nausea 30%, diarrhoea 25%, vomiting 8%, mucositis 5% CMF: nausea 1%, diarrhoea 1%, vomiting 1%, mucositis 2% (all Grade 3 or 4)	High dose: 7% CMF: 1%	High dose 8% CMF: 0%	High dose: 9% CMF: 1%	High dose: 6% CMF: 0%	High dose group: 31% CMF: 2% (Grade 3 or 4)		The incidence of nonfatal but serious adverse effects was greater in the group assigned to high dose chemotherapy,
IBDIS 2003
NCIC 2001				Cardiac dysfunction High dose group: 11% Standard therapy: 5% p=0.47			Grade 3 or 4 febrile neutropenia or infection High dose group: 60/91Standard therapy: 5/111 p=<0.0001	Secondary malignancy High dose group: 1/112 Standard therapy: 2/112	More grade 3‐4 toxicity (haematologic and non‐haematologic) in high dose arm
PEGASE 03 2002	High dose: neutropenia = median 10 days; thrombocytopenia < 20 = median 12 days; FEC: Neutropenia grade 3 or 4 = 38% of cycles	Grade 3 or 4 mucositis: High dose: 30% FEC: 2% Nausea and vomiting: High dose: 49%		No grade 3 or 4 toxicities observed	No grade 3 or 4 toxicities observed	High dose: 2.6%	High dose: Grade 3 or 4 = 30% FEC: Grade 3 or 4 = 1% of cycles		Toxicity after FEC manageable.
PEGASE 04 1999	High dose: neutropenia < 0.5 = median 13 days; thrombocytopenia = median 10 days	High dose: No grade 3 or 4 toxicity	High dose: pulmonary aspergillus infection requiring surgery (1 woman)	High dose: No grade 3/4 toxicity				Median hospitalisation in High dose arm: 25 days (range 20‐44)	HDC: 57% women who received HDC experienced grade 3‐4 extra‐haematological toxicities
Schmid 2005	Grade 3 or 4 toxicities: High dose group: white blood cell toxicity 100% Haemoglobin toxicity 9.3% Platelet toxicity 77.2% Control group: white blood cell toxicity 41.5% Haemoglobin toxicity 5.1% Platelet toxicity 8.1%	Grade 3 or 4 toxicities: High dose group: Nausea/vomiting 54.1% Stomatitis 20.3% Diarrhoea 9.5% Control group: Nausea/vomiting 7.0% Stomatitis 0.8% Diarrhea 0.0% Control group: Nausea/vomiting 54.1% Stomatitis 20.3% Diarrhoea 9.5%	Grade 3 or 4 toxicity: High dose group: 5.5% Control group: 0.8%	Grade 3 or 4 toxicity: High dose group: 1.4% Control group: 0.0%	Grade 3 or 4 toxicity: High dose group: 1.3% Control group: 0.0%		Grade 3 or 4 toxicity: High dose group: 39.2% Control group: 2.3%	Grade 3 or 4 hepatic toxicity High dose group: 0.0% Control group: 0.4% Late cardiotoxicity: High dose group: 1/48 women Control group: 2/45 women

Quality of life

Two studies reported on quality of life (ECOG 2000, NCIC 2001).

In ECOG 2000 three quality of life measures were collected: POMS (general measure of emotional distress), MOS (general health and well‐being) and SDS (symptoms specifically referable to chemotherapy). Not all women completed the surveys but failure to complete them was equal in both groups. There were no statistically significant differences at baseline, but at six months total mood disturbance was significantly worse in the high dose arm (21.0 versus 7.3, p=0.03). The groups did not differ with respect to the reported incidence of side effects specifically attributable to chemotherapy, nor did their levels of bodily pain or vitality (as measured by the MOS) differ significantly.

NCIC 2001 used two quality of life measures: EORTC C‐30 (quality of life of cancer patients), FACT‐BMT subscale (quality of life in bone marrow transplant patients), as well as five study‐specific questions. At baseline the high dose group had significantly worse scores for social function but otherwise there was no significant difference between the groups. Immediately following treatment, quality of life scores were significantly worse in the high dose group in the following areas: physical function, role function, social function, fatigue, dyspnoea, global quality of life and FACT‐BMT subscale. At six and nine months' follow‐up most scores had recovered but there was still a significantly increased prevalence of breathlessness, bruising and bleeding in the high dose group.

Time to tumour progression

All studies reported on time to tumour progression; however no individual patient data were available and therefore combining results was not possible. Four studies found a statistically significant difference between the two groups in time to tumour progression (IBDIS 2003, NCIC 2001; PEGASE 03 2002, PEGASE 04 1999). In all these studies the time to tumour progression was longer in the high dose group. Two studies (ECOG 2000; Schmid 2005) found no statistically significant difference between the groups in time to tumour progression. See Table 6 for details.

6. Median time to tumour progression.

Study Name	High dose chemo	Control group	p‐value	median follow up
ECOG 2002	9.6 months	9.1 months	0.31	69.5 months
IBDIS 2003	14.4 months	10.6 months	0.0033	60 months
NCIC 2001	11 months	9 months	0.006	48 months
PEGASE 03 2002	11 months	6.6 months	0.0001	48 months
PEGASE 04 1999	12 months	6 months	<0.0056	87‐92 months
Schmid 2005	11.1 months	10.6 months	0.67	52 months

Overall survival

All studies reported on median overall survival time; however no patient individual data were available and therefore combining results was not possible. One of the trials (PEGASE 04 1999) reported a significant difference (p=0.0294) between the two groups in median overall survival time, benefiting the high dose group. The other five trials found no statistically significant difference between the groups for this outcome. See Table 7 for details.

7. Median overall survival time.

Study ID	High dose	Control group	p value	Median follow up
ECOG 2000	26 months	26 months	N/S	69.5 months
IBDIS 2003	31.2 months	26.4 months	0.145	60 months
NCIC 2001	24 months	28 months	0.43	48 months
Pegase 03 2002	33.6 months	27.3 months	0.8	48 months
Pegase 04 1999	44.1 months	19.3 months	<0.0294	87‐92 months
Schmid 2005	26.9 months (95% CI 21.6‐32.1)	23.4 months (95% CI 12.4‐34.5)	0.60	52 months

Discussion

The results of this review show a statistically significant advantage in event‐free survival for women with metastatic breast cancer receiving high dose chemotherapy. There is also evidence of a longer median time to tumour progression in the high dose group. Moreover although no statistically significant benefit in overall survival has been demonstrated, there is a non‐significant trend towards improved overall survival in the high‐dose group (at five years: RR 1.50, 95% CI 0.96 to 2.34). This finding was not evident in the first version of this review, possibly because any benefits for high dose treatment were overshadowed by significantly higher treatment‐related mortality in the high dose arm.

There are several issues making it essential to interpret these results with extreme caution. It would take only one event in the high‐dose group at five years to negate the statistically significant effect for event‐free survival at this time point, and the effect is also nullified by the use of a random‐effect statistical model. Further data are required as none of the analyses for survival involve more than 355 women. At present the follow up of most of the studies is incomplete and one has not been published. Moreover four studies mention that some participants from the control group had treatment with high dose chemotherapy and autograft at relapse (ECOG 2000, NCIC 2001, PEGASE 03 2002, Schmid 2005), making it more difficult to compare the progress of the two groups. There is moderate heterogeneity in the results for event‐free survival at three years, which is probably attributable to there being only two trials in this comparison. Heterogeneity is low or absent for other outcomes.

The authors of PEGASE 04 1999, who reported positive results for high dose treatment, commented that it would be inappropriate to conclude from their findings that high dose chemotherapy has a real impact on overall survival. They pointed out a number of limitations in their study, notably a higher prevalence in the high dose group of pulmonary lesions (which are more chemosensitive than other visceral lesions) and a difference in the total duration of chemotherapy received due to the failure of 17/29 women in the control group to receive post‐randomisation maintenance chemotherapy as planned. They noted that there was probably selection bias in the trial, with oncologists referring only women who would be expected to be good candidates for high dose treatment and this may limit the generalisability of results. In addition they speculated whether the superior results in the high dose arm might relate to hormone deprivation, as all women in the high dose arm became postmenopausal whereas many remained premenopausal in the control arm.

Some researchers have suggested that while the studies have not demonstrated a statistically significant advantage in long term survival from high dose chemotherapy, a particular subgroup may be shown to benefit. This possibility was investigated by ECOG 2000 and NCIC 2001 investigators. NCIC 2001 investigators looked specifically at the following factors: type of induction chemotherapy, level of response to induction, whether adjuvant chemotherapy was used and the presence or absence of visceral disease. However no benefit was shown for any subgroup. ECOG 2000 investigators looked at level of response to induction, age, oestrogen receptor status and site of metastatic disease. The only statistically significant interaction observed between treatment and subgroup was for age. They reported a 28% lower hazard of death for women aged over 43 and assigned to conventional treatment, and a 33% higher hazard of dying for women aged 42 or less and assigned conventional chemotherapy. The number of women in these subgroups was small: there were a total of 61 women aged 42 years or less and 123 women aged 43 to 61 years.

High dose chemotherapy with autograft is associated with a significant risk of treatment‐related death, a high incidence of serious acute side effects and severely diminished quality of life following treatment. Although most of these effects are short term and quality of life gradually improves, women still report significant impairment six to nine months later. However in the 1990s high dose therapy was often viewed as worthwhile for women with high risk or advanced disease (Nieto 2000). As a result many women in the USA have been treated outside of a clinical trial: data from the Autologous Blood and Marrow Transplant Registry suggest that in the USA during the 1990s over 40,000 women with breast cancer received this treatment though fewer than 1000 were recruited to clinical trials (ABMTR 2002). This review underscores the importance of randomised controlled trials to provide reliable evidence about effective treatments for women with metastatic breast cancer.

Authors' conclusions

Implications for practice.

There is insufficient evidence to support the routine use of high dose chemotherapy with autograft for women with metastatic breast cancer. There is some evidence that this treatment prolongs disease‐free survival but no statistically significant evidence that it prolongs overall survival. Moreover high dose chemotherapy with autograft is associated with considerable morbidity and a significant risk of treatment‐related mortality.

Implications for research.

High dose chemotherapy with bone marrow or stem cell transplantation should not be given to women with metastatic breast cancer outside of clinical trials. It seems unlikely that further studies are warranted unless more compelling evidence emerges of a beneficial effect, or until there are further encouraging technical or scientific developments.

What's new

Date	Event	Description
6 February 2018	Review declared as stable	No new studies have been conducted on this topic since the previous update of this review. Therefore we do not expect to update this review.

History

Protocol first published: Issue 3, 2001
Review first published: Issue 3, 2005

Date	Event	Description
10 October 2010	New search has been performed	Performed search for new studies on the 10th of October 2010. Data checked against published results at median follow‐up 48 months for two studies: NCIC 2001 and PEGASE 03. Minimal changes as most data were previously available in unpublished form. Text reformatted with RevMan 5 headings and updated Risk of Bias tables.
5 August 2008	Amended	Converted to new review format.
12 May 2005	New citation required and conclusions have changed	Substantive amendment

Appendices

Appendix 1. EMBASE search strategy

1. Controlled study/ or randomized controlled trial/
2. double blind procedure/
3. single blind procedure.mp. [mp=title, abstract, cas registry/ec number word, mesh subject heading]
4. single blind procedure/
5. crossover procedure/
6. drug comparison/
7. placebo/
8. random$.ti,ab,hw,tn,mf.
9. latin square.ti,ab,hw,tn,mf.
10. crossover.ti,ab,hw,tn,mf.
11. cross‐over.ti,ab,hw,tn,mf.
12. placebo$.ti,ab,hw,tn,mf.
13. ((doubl$ or singl$ or tripl$ or treb$) adj5 (blind$ or mask$)).ti,ab,hw,tn,mf.
14. (clinical adj5 trial$).ti,ab,hw,tn,mf.
15. or/1‐14
16. nonhuman/
17. animal/ not (human/ and animal/)
18. or/16‐17
19. 15 not 18
20. exp breast neoplasms/
21. breast neoplasm$.tw.
22. (breast adj5 carcinoma).tw.
23. (breast adj5 cancer).tw.
24. (breast adj5 malignan$).tw.
25. or/20‐24
26. bone marrow transplantation/
27. ("bone marrow" adj5 transplant$).tw.
28. ("stem cell" adj5 transplant$).tw.
29. ("stem cell" adj5 transplant$).mp. [mp=title, abstract, cas registry/ec number word, mesh subject heading]
30. ("stem cell" adj5 transplant?).mp. [mp=title, abstract, cas registry/ec number word, mesh subject heading]
31. ("stem cell" adj5 transplant?).tw.
32. ("bone marrow" adj5 transplant?).tw.
33. "progenitor cell$".tw.
34. PBPC.mp. [mp=title, abstract, cas registry/ec number word, mesh subject heading]
35. ABMT.tw.
36. autograft.tw.
37. autotransplant.tw.
38. or/26‐37
39. 19 and 25 and 38

Appendix 2. MEDLINE search strategy

1. randomized controlled trial.pt.
2. controlled clinical trial.pt.
3. Randomized controlled trials/
4. random allocation/
5. double‐blind method/
6. single‐blind method/
7. or/1‐6
8. clinical trial.pt.
9. exp clinical trials/
10. (clin$ adj25 trial$).ti,ab,sh.
11. ((singl$ or doubl$ or tripl$ or trebl$) adj25 (blind$ or mask$)).ti,ab,sh.
12. placebos/
13. placebo$.ti,ab,sh.
14. random$.ti,ab,sh.
15. Research design/
16. or/8‐15
17. animal/ not (human/ and animal/)
18. 7 or 16
19. 18 not 17
20. breast neoplasms/
21. breast neoplasm$.tw.
22. (breast adj5 carcinoma).tw.
23. (breast adj5 cancer).tw.
24. (breast adj5 malignan$).tw.
25. or/20‐24
26. bone marrow transplantation/
27. ("bone marrow" adj5 transplant$).mp. [mp=title, abstract, cas registry/ec number word, mesh subject heading]
28. ("stem cell" adj5 transplant$).tw.
29. ("stem cell" adj5 support).tw.
30. "progenitor cell$".tw.
31. PBPC.mp. [mp=title, abstract, cas registry/ec number word, mesh subject heading]
32. PBPC.tw.
33. ABMT.tw.
34. autograft.tw.
35. autotransplant$.tw.
36. or/26‐35
37. 19 and 25 and 36

Appendix 3. WHO ICTRP search strategy

Basic Search:
1. advance breast cancer AND high dose chemotherapy AND autologous bone marrow
2. advance breast cancer AND high dose chemotherapy AND stem cell transplantation
3. metastatic breast cancer AND high dose chemotherapy AND autologous bone marrow
4. metastatic breast cancer AND high dose chemotherapy AND stem cell transplantation
5. advance breast cancer AND conventional chemotherapy
6. metastatic breast cancer AND conventional chemotherapy
7. advance breast cancer AND high dose chemotherapy AND bone marrow transplantation
8. advance breast cancer AND high dose chemotherapy AND stem cell support
9. metastatic breast cancer AND high dose chemotherapy AND bone marrow transplantation
10. metastatic breast cancer AND high dose chemotherapy AND stem cell support

Advanced Search:
1. Condition: advance* breast cancer* OR advance* breast carcinoma* OR advance* breast neoplasm* OR metastatic breast cancer* OR metastatic breast carcinoma * OR metastatic breast neoplasm * 
Intervention: high dose chemotherapy* OR autologous bone marrow transplant* OR bone marrow transplant* OR stem cell transplant* OR stem cell support OR conventional chemotherapy*

2. Condition: advance* breast cancer* OR advance* breast carcinoma* OR advance* breast neoplasm* OR metastatic breast cancer* OR metastatic breast carcinoma * OR metastatic breast neoplasm *
Intervention: high dose chemotherapy* AND autologous bone marrow transplant* OR conventional chemotherapy*

3. Condition: advance* breast cancer* OR advance* breast carcinoma* OR advance* breast neoplasm* OR metastatic breast cancer* OR metastatic breast carcinoma * OR metastatic breast neoplasm *
Intervention: high dose chemotherapy* AND bone marrow transplant* OR conventional chemotherapy*

4. Condition: advance* breast cancer* OR advance* breast carcinoma* OR advance* breast neoplasm* OR metastatic breast cancer* OR metastatic breast carcinoma * OR metastatic breast neoplasm *
Intervention: high dose chemotherapy* AND stem cell transplant* OR conventional chemotherapy*

5. Condition: advance* breast cancer* OR advance* breast carcinoma* OR advance* breast neoplasm* OR metastatic breast cancer* OR metastatic breast carcinoma * OR metastatic breast neoplasm *
Intervention: high dose chemotherapy* AND stem cell support OR conventional chemotherapy*

Data and analyses

Comparison 1. High‐dose chemotherapy versus conventional therapy.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Treatment‐related deaths	6	850	Risk Ratio (M‐H, Fixed, 95% CI)	4.08 [1.40, 11.93]
1.2 Overall survival	6		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
1.2.1 Three year follow‐up	3	495	Risk Ratio (M‐H, Fixed, 95% CI)	1.10 [0.86, 1.42]
1.2.2 Five year follow‐up	3	355	Risk Ratio (M‐H, Fixed, 95% CI)	1.50 [0.96, 2.34]
1.3 Event‐free survival	5		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
1.3.1 Three year follow‐up	2	272	Risk Ratio (M‐H, Fixed, 95% CI)	2.35 [0.86, 6.41]
1.3.2 Five year follow‐up	3	355	Risk Ratio (M‐H, Fixed, 95% CI)	2.84 [1.07, 7.50]

1.1. Analysis.

Comparison 1: High‐dose chemotherapy versus conventional therapy, Outcome 1: Treatment‐related deaths

1.2. Analysis.

Comparison 1: High‐dose chemotherapy versus conventional therapy, Outcome 2: Overall survival

1.3. Analysis.

Comparison 1: High‐dose chemotherapy versus conventional therapy, Outcome 3: Event‐free survival

Characteristics of studies

Characteristics of included studies [ordered by study ID]

ECOG 2000.

Study characteristics
Methods	Randomised controlled trial Method of randomisation: Central. Stratified according to type of response to induction chemotherapy (partial or complete), predominant site of distant metastasis (visceral or other), age (<= 42 years or > 42 years), oestrogen receptor status (positive, negative or unknown) and cooperative group Allocation to groups: Central Number enrolled for induction chemotherapy: 553 Number of patients randomised: 199 Number of patients analysed: 184 Number of withdrawals: 15 women (9 on high dose arm, 6 on control arm) found to be ineligible: of these, nine did not have a documented response to induction chemotherapy, three were oestrogen receptor positive and had not received hormone therapy, two had disease progression in the interval before randomisation and one had insufficient data Failure to receive protocol therapy: Six on high dose arm (5 refused, one had different high dose therapy), 14 on control arm (10 had high dose therapy, 3 had no therapy, one unknown) "Intention to treat" analysis: yes Number of centres: 9 US centers Source of funding: National Cancer Institute and US Healthcare. Years: 1990 to 1997
Participants	Initial eligibility criteria: Women with metastatic breast cancer (local or distant metastatic disease), pre or post menopausal, age 18 to 60 years, no prior chemotherapy for metastatic disease, at least 6 months since any prior adjuvant chemotherapy, at least one prior hormonal treatment if ER +ve (unless life threatening visceral disease present of hormone receptor status unknown without hormonal trial), ECOG performance status 0 or 1. Eligibility for randomisation: Inclusions: Documented complete or partial response to induction chemotherapy, normal renal, cardiac, pulmonary and hepatic function, adequate granulocyte and platelet levels, <8 weeks since induction chemotherapy. Exclusions: Bone marrow involvement with tumour, severe medical or psychiatric problems Country: USA
Interventions	All patients meeting initial eligibility criteria received induction chemotherapy of oral cyclophosphamide, doxorubicin and I/V fluorouracil (dosage depended on whether or not they had previously received doxorubicin). After receiving induction chemotherapy, patients were reevaluated and were eligible for randomisation if they had had a complete remission or partial remission. Patients were withdrawn if they had new lesions or progression. Randomisation had to occur within 8 weeks of the last dose of induction chemotherapy. Patients were randomised to either high dose chemotherapy (a single course of carboplatin ‐ 200mg per metre square, thiotepa ‐ 125 mg per metre square and cyclophosphamide ‐ 1500 mg per metre square all per day as continuous infusion for four days) followed by autologous stem‐cell transplantation or conventional chemotherapy (24 cycles of cyclophosphamide, methotrexate and fluorouracil)
Outcomes	Overall survival Progression free survival Adverse events
Notes	Power calculation: 85% power to detect a doubling of the median survival with two sided P value of 0.05 with sample of 164. [assuming median survival of 2.5 years (1year) for women with complete (partial) response to initial chemotherapy] Median time of follow up = 69.5 months. Not all patients have completed five year follow up (‐ five year survival data presented in the tables of comparison in this review have been calculated from percentage survivals cited by triallists).
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Centrally randomised by computer
Allocation concealment (selection bias)	Low risk	A ‐ Adequate, remotely allocated
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	184/199 of randomised participants included in analysis, reasons for withdrawal provided. Potential for bias unclear
Selective reporting (reporting bias)	Low risk	All expected outcomes reported systematically
Other bias	High risk	There was a statistically significant (p=0.04) imbalance between the groups at baseline: 37% of the high dose group and 53% of the control group had bone metastasis

IBDIS 2003.

Study characteristics
Methods	Randomised controlled trial Method of randomisation/allocation: Centrally randomised; allocated by sequential opaque envelopes Number of patients randomised: 110 Number of patients analysed: 110 Failure to complete all cycles of treatment: High dose arm 16% (9/56); control arm 26% (14/54) "Intention to treat" analysis: Yes Number of centres: 17 Source of funding: Pharmaceutical Years: 1997 to 2001
Participants	Women with metastatic breast cancer who had not had prior chemotherapy for metastatic disease. Prior adjuvant chemotherapy allowed > 1 year pre‐enrolment. No prior invasive malignancy. No prior platinum, mitomycin, <300mg/m2 doxorubicin or 450mg/m2 epirubicin. Normal echo. No or minimal pelvic radiotherapy, adequate blood counts, no marrow or CNS metastases, adequate hepatic function, no prior or concomitant malignancy (except for basal cell carcinoma of the skin or in‐situ neoplasia of the uterine cervix). Countries: Ireland, Switzerland, UK, Spain, Greece
Interventions	Initial chemotherapy: All women received 3 cycles of doxorubicin 50mg/m2 and docetaxel 75mg/m2. Women randomised to the high dose arm who had not progressed on the initial chemotherapy then received (within 9 weeks of randomisation) ifosfamide 12000mg/m2, carboplatin AUC 18 and etoposide 1200mg/m2, followed after 3‐6 weeks by cyclophosphamide 6gm/m2 and thiotepa 800mg/m2. Women randomised to the conventional dose arm who had not progressed on initial chemotherapy then received an additional cycle of the initial chemotherapy, followed by cyclophosphamide 600mg/m2, methotrexate 40mg/m2 and fluorouracil 600mg/m2 twice per cycle for 4 cycles. All cycles were administered with GCSF, with PBPC support for high dose arm
Outcomes	Overall survival Progression‐free survival Treatment‐related death Median overall survival Median progression‐free survival time
Notes	Projected accrual of 264 women: Accrual failed 1999; accrual terminated with only 110 women. Trial externally audited in 2004 with full chart review of 59/110 women Median follow‐up 5 years: five year survival data presented in the tables of comparison in this review have been calculated from percentage survivals cited by triallists.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Centrally generated random number sequence
Allocation concealment (selection bias)	Low risk	A ‐ Adequate; allocated by sequential opaque envelopes
Incomplete outcome data (attrition bias) All outcomes	Low risk	All randomised participants included in analysis
Selective reporting (reporting bias)	Low risk	All expected outcomes reported systematically
Other bias	High risk	"Regularly scheduled radiological reassessment of disease status was not required after chemotherapy completion; hence assessment of progression free survival may be subject to observer bias"

NCIC 2001.

Study characteristics
Methods	Randomised controlled trial Method of randomisation: not stated Stratified by response to induction, hormone receptor status, presence or absence of visceral disease, type of induction. Allocation to groups: not stated Number registered to receive induction chemotherapy: 386 Number of patients randomised: 224 Number of patients analysed for efficacy: 223 (excluding one woman in control arm found to be ineligible because she did not have complete or partial response to induction chemo) Failure to receive protocol therapy: 23 on high dose arm (11 because disease progressed, six refused, two had inadequate stem cell collection, two due to toxicity, two unexplained) "Intention to treat" analysis: For 223/224 women Number of centres: 27 Source of funding: NCIC Years: 1997‐2000
Participants	Women with metastatic breast cancer or locoregional recurrence (chest wall or axillary lymph node) after mastectomy, aged 16 years or more who had had no prior chemotherapy for metastatic breast cancer and who responded to induction chemotherapy. ECOG performance status 0‐2. No CNS metastases. Country: Canada
Interventions	All women received four cycles conventional dose chemotherapy, either anthracycline‐based, or (if previously exposed to anthracyclines) taxane‐based. Women responding to this treatment were then randomised. Women randomised to the high‐dose arm then received one or two more cycles of the induction chemotherapy, followed by cyclophosphamide 6gm/m2, mitoxantrone 70mg/m2 and carboplatin 1600mg/m2 given in divided doses over four consecutive days. Women on the control arm received two to four more cycles of the initial chemotherapy. All women with hormone receptor positive disease received Tamoxifen or another hormonal treatment. Women in the high dose arm with solitary bone or soft tissue metastases received local radiotherapy. Controls received radiotherapy at the physician's discretion.
Outcomes	Progression‐free survival Overall survival Quality of life
Notes	Median follow‐up 48 months. Survival data are immature. Survival data presented in this review have been calculated from survival graphs presented by triallists.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method not stated
Allocation concealment (selection bias)	Unclear risk	B ‐ Unclear; method not stated
Incomplete outcome data (attrition bias) All outcomes	Low risk	123/124 randomised participants included in analysis for efficacy. All eligible patients who received at least one cycle of chemotherapy post randomisation were included in safety analysis, based on treatment received
Selective reporting (reporting bias)	Low risk	All expected outcomes reported systematically
Other bias	Low risk

PEGASE 03 2002.

Study characteristics
Methods	Randomised controlled trial Method of randomisation: not stated Allocation to groups: not stated Number enrolled for induction chemotherapy: 308; of whom 186 eligible for randomisation, of whom 7 refused to be randomised. Number of patients randomised:179 Number of patients analysed: 179 Number of withdrawals: nil stated; 80 of 88 women completed HDC Failure to receive protocol therapy: Nine on high dose arm (in two cases because stem cell collection unsuccessful), two on control arm (both received high dose chemotherapy) "Intention to treat" analysis: yes Number of centres: 23 Source of funding: Federation Nationale des Centres de Lutte Contre le Cancer (FNCLCC) and the Societe Francaise de greffe de moelle (SFGM) Years:01/95‐06/01
Participants	Women with metastatic breast cancer with at least one measurable target requiring first line chemotherapy, who had an objective response to induction chemotherapy and showed signs of disease progression within one month of inclusion in the study. Less than 60 years old, performance status less than 2, life expectancy greater than 3 months. Adjuvant cumulative dose of epirubicin < 450 mg/m2 or of doxorubicin < 300mg/m2. Women with bone, ascites, pleural effusion or pulmonary or cutaneous lymphangitis as the only metastatic site were excluded. Country: France
Interventions	All patients received induction chemotherapy of four cycles of FEC 100 (fluorouracil 500 mg/m2; cyclophosphamide 500 mg/m2; epirubicin 100 mg/m2). Responding patients were randomised as soon as they responded but not later than 4 cycles to high dose thiotepa (800 mg/m2) and cyclophosphamide (6000 mg/m2) with PBPC reinfusion at 96 hours after completion of chemotherapy or no further treatment. Harvesting of PBPC performed under G‐CSF following FEC 100 course, minimum of 2 x 10 6 CD34+cells/kg collected. Treatment at relapse left to the choice of the clinician.
Outcomes	Progression free survival Overall survival Relapse or death Restrictions on quality of life
Notes	Only 80 of the 89 in the high dose treatment arm were transplanted. Power calculations:180 patients needed in order to detect a 20% improvement in relapse‐free survival at three years Only 80 of the 89 in the high dose treatment arm were transplanted. Median follow‐up: 48 months. All survival data are immature. Survival data presented in this review have been calculated from percentages quoted by the triallists.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method not stated
Allocation concealment (selection bias)	Unclear risk	B ‐ Unclear; method not stated
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants included in survival analyses (personal correspondence with author)
Selective reporting (reporting bias)	Low risk	All expected outcomes reported systematically
Other bias	High risk	Prognostic imbalances at baseline in hormone‐receptor status, proportion with liver and distant lymph node metastases and lung metastases.

PEGASE 04 1999.

Study characteristics
Methods	Randomised controlled trial Method of randomisation=stratified by centre Allocation to groups: not stated Number of patients enrolled for induction chemotherapy: 105 Number of patients randomised: 61 Number of patients analysed: 61 Number of withdrawals: none stated Failure to receive protocol therapy: Two women on high dose arm, due to rapid disease progression "Intention to treat" analysis: Yes Number of centres: >30 Source of funding: Federation Nationale des Centres de Lutte Contre le Cancer (FNCLCC) and the Societe Francaise de greffe de moelle (SFGM) Years: 09/92‐12/96
Participants	Women with metastatic breast cancer (at diagnosis or first relapse), responding to anthracycline‐based conventional chemotherapy (a maximum of 9 cycles), not more than 60 years old, life expectancy greater than 3 months, performance status less than 2. Any brain metastases non‐life threatening and chemosensitive. Adjuvant cumulative dose of epirubicin < 450 mg/m2 or of doxorubicin < 300mg/m2. Country: France
Interventions	All patients received induction chemotherapy (3‐9 cycles of anthracycline‐based conventional chemotherapy). Patients were then randomised to either high‐dose chemotherapy (cyclophosphamide 120 mg/kg, mitoxantrone 45 mg per metre square, melphalan 140 mg per square metre, + filgrastim) or to a proposed*(see Notes) 2‐4 additional cycles of conventional treatment, using the same regimen where possible. Collection of PBPC was done after randomisation alone of in combination with chemotherapy. A minimum of 2 x 10 6 CD34+ cells/kg was be harvested. PBPC were infused at least 48 hours after Melphalan and 6 days after mitoxantrone All women received Tamoxifen 20 mg daily for five years if receptor positive. Radiotherapy "if necessary" Treatment of relapses determined in consultation with individual oncologist.
Outcomes	Primary end‐point was overall survival Relapse‐free survival, progression‐free survival, relapse rates. Ancillary studies put forward on cost‐efficiency ratio, quality of life, residual disease in PBPC products, biological markers
Notes	Power calculations found that 156 subjects were required to detect a 25% improvement in relapse free survival rates at 2 years. Accrual failed ‐ only 61 women recruited. Data are mature. Median time of follow‐up 87 months in CDC arm, 92 months in HDC arm. Survival data presented in this review have been calculated from percentages reported by triallists * Only 12/29 women in the standard dose arm received any chemotherapy after randomisation
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method not stated
Allocation concealment (selection bias)	Unclear risk	B ‐ Unclear; methods not stated
Incomplete outcome data (attrition bias) All outcomes	Low risk	All randomised participants included in analysis
Selective reporting (reporting bias)	Low risk	All expected outcomes reported systematically
Other bias	High risk	There was imbalance between the two groups with respect to pulmonary metastases (high dose group: 46%, control group 14%). This may have favoured the high dose arm as pulmonary lesions are more sensitive to chemotherapy than other visceral lesions. The duration of chemotherapy received also favoured the high‐dose group since, as noted above, only 12 out of 29 women randomised to the control arm had chemotherapy subsequent to randomisation.

Schmid 2005.

Study characteristics
Methods	Randomised controlled trial Method of randomisation. Centrally randomised. Stratified by menopausal status and hormone receptor status Allocation to group: By fax from central trial office Number enrolled for induction chemotherapy: not stated Number randomised: 93 (see note) Number of patients analysed: 93 Number of withdrawals: nil stated Failure to receive protocol therapy: seven on high dose arm (one had hepatitis, three had rapid progression or early death, two had inadequate stem cell collection, one declined), nil on control arm "Intention to treat" analysis: yes Number of centres: 19 Source of funding: not stated Years: 04/98 to 11/01
Participants	Women aged 18‐60 with metastatic breast cancer without prior cytotoxic chemotherapy for metastatic disease At least one bi dimensionally measurable lesion that had not been irradiated, Excluded: women who had had any adjuvant chemo within 6 months of the study or previous high dose chemotherapy, brain metastases, bone metastases as the only site of measurable disease, serious disease which might impair ability to receive protocol treatment Countries: Germany and Austria
Interventions	Women randomised to double high dose chemotherapy (cyclophosphamide 4.4 gm/m2; mitoxantrone 45 mg/m2 and etoposide 2.5 gm/m2 repeated once after 6 weeks) with PBSCT or to 6‐9 courses of standard chemotherapy (doxorubicin 60 mg/m2; paclitaxel 200 mg /m2 q3 weeks, max 6 courses additional 3 courses of paclitaxel if a complete response seemed achievable). G‐CSF mobilized stem cells were collected before and after the first course of high dose chemotherapy. All Hormone Receptor Positive Status patients received tamoxifen. Cross‐over High Dose Chemotherapy was planned at relapse for patients showing a complete response to doxorubicin and paclitaxel.
Outcomes	Primary end point was complete response rates Secondary objectives progression‐free survival, overall survival, toxicity
Notes	Planned with a two‐step statistical design with 220 women and 140 women per arm for the two phases, respectively. However, due to accrual failure, trial terminated prematurely without completion of the first phase and only exploratory data analyses applied. Data are immature. Survival data presented in this review have been calculated from percentages reported by triallists.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Patients were randomized by study center in a 1:1 ratio to one of the two treatment groups, using a stratified random permuted block design (block size 4)."
Allocation concealment (selection bias)	Low risk	A ‐ Adequate; centrally allocated. "Randomization was performed centrally at the coordinating study center, and was based on the order in which information on potential patients was received by fax."
Incomplete outcome data (attrition bias) All outcomes	Low risk	All randomised women included in analysis
Selective reporting (reporting bias)	Low risk	All expected outcomes reported systematically
Other bias	Low risk

AUC: Area Under Curve; chemotherapy dose titrated according to individual kidney function
ER +ve: Oestrogen receptor positive
GCSF: Granulocyte colony‐stimulating factor
PBPC: Peripheral blood progenitor cells

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Bezwoda 1995	Subsequent to the presentation of the Bezwoda trial results at the 1999 ASCO Annual Meeting there was an investigation by the University of Witwatersrand into possible breaches of scientific honesty and integrity concerning the research underlying the presentation. The data and information presented in the materials was incorrect.
Madan 2000	Women in the control group crossed over to the high dose arm if disease progressed ‐ a comparison of different time frames for high dose therapy rather than comparing high dose with conventional.
Vredenburgh 2006	The study design did not include a non transplantation or "conventional" treatment arm.

Characteristics of ongoing studies [ordered by study ID]

GEBDIS (Kanz).

Study name	GEBDIS
Methods
Participants	Women with chemo‐naive metastatic disease
Interventions	HDC: ICE then Cy/TT versus observation ICE then Cy/TT (at relapse)
Outcomes
Starting date
Contact information
Notes	Accrual target 350

GITMO (Rosti).

Study name	GITMO
Methods
Participants	Metastatic breast cancer
Interventions	HDC: ICE then ICE then TT/mel versus standard‐dose ICE
Outcomes
Starting date	Feb 2000
Contact information
Notes	Accrual target 240

ICE: ifosfamide, carboplatin and etoposide
A‐Txtx: adriamycin and Taxotere
TT: Tiotepa
Cy: cyclophosphamide

Contributions of authors

CF drafted the protocol, searched for and selected the studies, extracted the data and wrote the text of the review. SH selected the studies, extracted the data, entered the data and completed the included and excluded studies table. RB edited the protocol, selected the studies, extracted the data and commented on the draft on several occasions, providing content advice. AL commented on the draft of the protocol, provided statistical advice and commented on the draft on several occasions. JMB checked the data extraction and entry and commented on the draft. JMB and CF have updated the review twice (in 2007 and 2010).

Sources of support

Internal sources

• University of Auckland, New Zealand

External sources

• RAND Corporation, Santa Monica, California (Supported by the Robert Wood Johnson Foundation Grant #044128), USA

Declarations of interest

Russell Basser is the principal investigator of a trial of high dose chemotherapy with autologous bone marrow transplantation in women with non‐metastatic breast cancer, (IBCSG Trial 15). He has previously served as a consultant for AstraZeneca, Pharmacia, Wyeth, Amrad, Amgen, Novartis, Flauding and BMS.

Stable (no update expected for reasons given in 'What's new')