Taxane‐containing regimens for metastatic breast cancer

Davina Ghersi; Melina L Willson; Matthew Ming Ki Chan; John Simes; Emma Donoghue; Nicholas Wilcken

doi:10.1002/14651858.CD003366.pub3

. 2015 Jun 10;2015(6):CD003366. doi: 10.1002/14651858.CD003366.pub3

Taxane‐containing regimens for metastatic breast cancer

Davina Ghersi ^1,^✉, Melina L Willson ², Matthew Ming Ki Chan ^3,⁴, John Simes ⁵, Emma Donoghue ⁶, Nicholas Wilcken ^7,⁸

Editor: Cochrane Breast Cancer Group

PMCID: PMC6464903 PMID: 26058962

Abstract

Background

It is generally accepted that taxanes are among the most active chemotherapy agents in the management of metastatic breast cancer. This is an update of a Cochrane review first published in 2003.

Objectives

The objective of this review was to compare taxane‐containing chemotherapy regimens with regimens not containing a taxane in the management of women with metastatic breast cancer.

Search methods

In this review update, we searched the Cochrane Breast Cancer Group Specialised Register, MEDLINE, EMBASE, the World Health Organization's International Clinical Trials Registry Platform (WHO ICTRP), and ClinicalTrials.gov on 14 February 2013 using keywords such as 'advanced breast cancer' and 'chemotherapy'. We searched reference lists of articles, contacted study authors, and did not apply any language restrictions.

Selection criteria

Randomised controlled trials comparing taxane‐containing chemotherapy regimens to regimens without taxanes in women with metastatic breast cancer. We included published and unpublished studies.

Data collection and analysis

Two review authors independently assessed trial quality and extracted data. We derived hazard ratios (HRs) for overall survival, time to progression, and time to treatment failure where possible, and used a fixed‐effect model for meta‐analysis. We represented objective tumour response rates and toxicity as risk ratios (RRs). We extracted quality of life data where present.

Main results

This review included 28 studies. The updated analysis included 6871 randomised women, while the original review had 3643 women. Of the 28 included studies, we considered 19 studies to be at low risk of bias overall; however, some studies failed to report details on allocation concealment and methods of outcome assessment for those outcomes that are more likely to be influenced by a lack of blinding (for example tumour response rate). Studies varied in the taxane‐containing chemotherapy backbone, and the comparator arms and were categorised into three groups: Regimen A plus taxane versus Regimen A (2 studies); Regimen A plus taxane versus Regimen B (14 studies); and single‐agent taxane versus Regimen C (13 studies). Thirteen studies used paclitaxel, 14 studies used docetaxel, and 1 study allowed the investigator to decide on the type of taxane; the majority of studies delivered a taxane every 3 weeks. Twenty studies administered taxanes as first‐line treatment, and 21 studies involved anthracycline naïve women in the metastatic setting. The combined HR for overall survival and time to progression favoured the taxane‐containing regimens (HR 0.93, 95% confidence interval (CI) 0.88 to 0.99, P = 0.002, deaths = 4477; and HR 0.92, 95% CI 0.87 to 0.97, P = 0.002, estimated 5122 events, respectively) with moderate to substantial heterogeneity across trials. If the analyses were restricted to studies of first‐line chemotherapy, this effect persisted for overall survival (HR 0.93, 95% CI 0.87 to 0.99, P = 0.03) but not for time to progression (HR 0.96, 95% CI 0.90 to 1.02, P = 0.22). Tumour response rates appeared to be better with taxane‐containing chemotherapy in assessable women (RR 1.20, 95% CI 1.14 to 1.27, P < 0.00001) with substantial heterogeneity across studies. Taxanes were associated with an increased risk of neurotoxicity (RR 4.84, 95% CI 3.18 to 7.35, P < 0.00001, 24 studies) and hair loss (RR 2.37, 95% CI 1.45 to 3.87, P = 0.0006, 11 studies) but less nausea/vomiting compared to non‐taxane‐containing regimens (RR 0.62, 95% CI 0.46 to 0.83, P = 0.001, 26 studies). Leukopaenia and treatment‐related death did not differ between the two groups (RR 1.07, 95% CI 0.97 to 1.17, P = 0.16, 28 studies; and RR 1.00, 95% CI 0.63 to 1.57, P = 0.99, 23 studies, respectively). For quality of life measures, none of the individual studies reported a difference in overall or any of quality of life subscales between taxane‐containing and non‐taxane chemotherapy regimens.

Authors' conclusions

Taxane‐containing regimens appear to improve overall survival, time to progression, and tumour response rate in women with metastatic breast cancer. Taxanes are also associated with an increased risk of neurotoxicity but less nausea and vomiting compared to non‐taxane‐containing regimens. The considerable heterogeneity encountered across studies probably reflects the varying efficacy of the comparator regimens used in these studies and indicates that taxane‐containing regimens are more effective than some, but not all, non‐taxane‐containing regimens.

Keywords: Female; Humans; Antineoplastic Agents, Hormonal; Antineoplastic Agents, Hormonal/therapeutic use; Antineoplastic Agents, Phytogenic; Antineoplastic Agents, Phytogenic/therapeutic use; Breast Neoplasms; Breast Neoplasms/drug therapy; Breast Neoplasms/mortality; Breast Neoplasms/pathology; Bridged‐Ring Compounds; Bridged‐Ring Compounds/therapeutic use; Disease Progression; Paclitaxel; Paclitaxel/therapeutic use; Randomized Controlled Trials as Topic; Tamoxifen; Tamoxifen/therapeutic use; Taxoids; Taxoids/therapeutic use

Plain language summary

Taxane‐containing regimens for metastatic breast cancer

Review question

We reviewed the evidence about the effect of taxane‐containing chemotherapy regimens in women with metastatic breast cancer. This is an update of a Cochrane review first published in 2003.

Background

Treatment for women with metastatic breast cancer (that is, cancer that has spread beyond the breast) usually involves chemotherapy to try to shrink or slow the growth of the cancer. Chemotherapy can involve a single drug or a combination of drugs. Paclitaxel and docetaxel are chemotherapy drugs known as taxanes. Taxanes can inhibit cancer cells from dividing and reproducing, and their adverse effects can include nausea, vomiting, and hair loss, as well as allergic reactions, which can be reduced by premedication. We wanted to examine whether or not taxane‐containing chemotherapy improves survival and extends time to disease progression in women with metastatic breast cancer.

Study characteristics

The evidence is current to February 2013. We included 28 studies that randomised 6871 women. Women were assigned to receive either a taxane‐containing chemotherapy regimen (single taxane or in combination with other chemotherapy drugs) or a non‐taxane chemotherapy regimen. There were variations in the taxane‐containing chemotherapy regimen and the non‐taxane treatments. Approximately half of the studies used paclitaxel and the other half used docetaxel, and in the majority of cases, taxanes were administered every three weeks. Of the 28 studies, 20 studies included women who received taxanes as their first treatment after their diagnosis of metastatic breast cancer, and 21 studies involved women who had not been previously treated with anthracyclines in the metastatic setting. From those studies reporting median follow‐up, this ranged from 9 months to 69 months.

Key results

This review showed that chemotherapy regimens including taxanes improved survival and decreased the progression of metastatic breast cancer. If the analyses were restricted to those studies where women received taxanes as their first treatment after their diagnosis of metastatic breast cancer, the survival benefit persisted. Taxanes also appeared to cause tumours to shrink more than chemotherapy regimens without taxanes. However, there were differences in side effects. The risk of experiencing neurotoxicity (tingling of hands and feet) with taxanes increased compared to non‐taxane chemotherapy. Hair loss also seemed to be more likely with taxane than with non‐taxane‐containing regimens. However, less nausea/vomiting was observed with taxanes. There was no difference in the rates of leukopaenia (low white blood cells) or treatment‐related deaths between taxane and non‐taxane chemotherapy. Of the studies that reported quality of life measures, there did not appear to be any differences (overall or on subscales) in quality of life between the two groups.

Quality of the evidence

We considered 19 out of the 28 studies to be at low risk of bias overall. However, some studies failed to report details on concealing drug treatments and methods of outcome assessment for those outcomes more likely to be at risk of bias (for example tumour response rate). The degree of variability seen across the included studies probably reflects the varying efficacy of the non‐taxane chemotherapy regimens used in these studies and indicates that taxane‐containing chemotherapies are more effective than some, but not all, non‐taxane‐containing regimens.

Background

Description of the condition

Breast cancer is the most common type of cancer in women, with more cases being diagnosed in less developed compared to more developed regions (Ferlay 2015). It is the most common cause of cancer death among women in less developed regions and the second most frequent cause of cancer death in more developed regions (Ferlay 2015). In 2012 there were an estimated 1.67 million new cases and approximately 522,000 deaths from breast cancer worldwide; an age standardised death rate of 12.9 (per 100,000) (Ferlay 2015).

The stage of breast cancer at the time of diagnosis is an important indicator of prognosis. Once breast cancer becomes metastatic it is rarely curable, with reported median survival of 18 to 24 months from the time of recurrence, although some women do experience long‐term survival (Hayes 1995; NCI 2003). Although there is no randomised evidence comparing chemotherapy with observation in women with metastatic breast cancer, it is widely accepted that women with metastatic disease should receive some form of systemic therapy at some time during the course of their disease.

Description of the intervention

Chemotherapy is considered by many to be the appropriate first treatment option for women with multiple sites of recurrence or where visceral disease is not easily treated by local modalities (Beslija 2009; Hayes 1995; NCI 2003). Chemotherapy is also considered to be useful in women whose cancer is hormone refractory, or expected to be hormone resistant (Hortobagyi 1996).

It is generally accepted that taxanes are among the most active chemotherapy agents in the management of metastatic breast cancer. The term 'taxanes' describes a group of drugs used in the treatment of cancer, specifically paclitaxel (Taxol® Bristol‐Myers Squibb) and docetaxel (Taxotere® Rhone‐Poulenc Rorer). The first taxane, paclitaxel, was identified in 1971 as part of a National Cancer Institute program that screened medicinal plants for potential anti‐cancer activity. It was originally isolated from the bark of the Pacific yew tree (Taxus brevifolia, native to western North America), but now a semisynthetic form is derived from the needles and twigs of the more common Himalayan or European yew (Taxus bacatta). Paclitaxel was first used in clinical trials in 1983 (BMS 1996). Docetaxel was first synthesised in 1986 and is similar, although not identical, to paclitaxel in its mechanism of action.

How the intervention might work

Taxanes are unique as they affect cell structures known as microtubules (or spindle fibres). In normal cell growth, microtubules are formed when a cell starts dividing and when the cell stops dividing, the microtubules are broken down or destroyed. Taxanes work by blocking the microtubules from breaking down. Cancer cells then become blocked with microtubules and stop dividing hence potentially slowing the growth of the cancer or killing the cells. The known side effects of paclitaxel include hypersensitivity reactions (such as shortness of breath or skin rash), myelosuppression (neutropaenia), peripheral neuropathy, cardiac rhythm disturbances, joint or muscle pain, diarrhoea, nausea and vomiting, or hair loss. Patients often receive premedication before receiving taxanes to prevent possible allergic reactions. The side‐effect profile of docetaxel is similar to that of paclitaxel, although docetaxel causes less neuropathy and more myelotoxicity (Vasey 2001).

Why it is important to do this review

In the previous version of this review, the primary aim was to assess taxane use in the first‐line setting, but the follow‐up data were insufficient for this to be adequately assessed. In this review update, data on an additional 3000 participants were available, with time‐to‐event data for 87% of the participants randomised, compared to 57% in the original review. Only one other systematic review and meta‐analysis appears to have been published, in 2008, but the focus was on taxanes alone or in combination with anthracyclines (refer to Piccart‐Gebhart 2008). The Piccart‐Gebhart 2008 review included first‐line therapy using individual participant data but did not conduct assessments of trial conduct and reporting or drug side effects. An update of the efficacy and safety of taxanes overall and in the first‐line setting in the form of an updated Cochrane review seemed warranted given the availability of mature follow‐up data and new trial data.

Objectives

The original review was conducted as part of a series of reviews comparing more intense (or more active) chemotherapy with less intense (or less active) chemotherapy in women with advanced (metastatic) breast cancer.

The objective of this review and review update was to compare taxane‐containing chemotherapy regimens with regimens not containing a taxane in the management of women with metastatic breast cancer. Subquestions within the review were:

subquestion A: regimen A plus taxane versus regimen A (e.g. doxorubicin plus docetaxel versus doxorubicin alone)
subquestion B: regimen A plus taxane versus regimen B (e.g. doxorubicin plus docetaxel versus doxorubicin plus cyclophosphamide)
subquestion C: single‐agent taxane versus regimen C (e.g. docetaxel versus doxorubicin plus cyclophosphamide)

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs).

Types of participants

Women with advanced (metastatic) breast cancer, either newly diagnosed or recurrent. Trials that included both women with metastatic disease and women with locoregionally recurrent disease were only eligible for inclusion if it was possible to distinguish between the two groups (that is the data were reported separately) or if women with isolated locoregional recurrence were less than 20% of the total group. We applied no age restrictions.

In the protocol for this review we proposed to include trials in which the women randomised to receive chemotherapy were receiving first‐line treatment (that is no previous chemotherapy given except as adjuvant therapy). In the original review, as very few completed trials involved first‐line treatment, all trials meeting the remaining eligibility criteria were included in the review. Results were presented separately for all trials (that is all lines) and first‐line only. In the 2013 review update, the number of first‐line trials increased, but for completeness, we still reported all lines and analysed as per the original review.

Types of interventions

Intervention group: Any chemotherapy regimen containing a taxane.
Comparator: Any chemotherapy regimen not containing a taxane.

Participants may also have received endocrine therapy if the study planned to give it to both treatment groups.

Trials may or may not have specified recommended treatment upon disease progression/initial treatment failure. This treatment may have included cross‐over to the alternative treatment arm of the trial. We did not include trials where the primary intention was to investigate sequencing of treatment regimens, including, for example:

trials where participants received a given number of cycles of one regimen, followed by a given number of cycles of another regimen (randomisation being to which regimen commenced first);
trials where regimens were alternated (e.g. one cycle of regimen A followed by one cycle of regimen B followed by a second cycle of regimen A, etc.).

Types of outcome measures

Primary outcomes

1. Overall survival  2. Time to progression

Secondary outcomes

3. Time to treatment failure  4. Objective tumour response rate  5. Toxicity  6. Health related quality of life

For the purpose of this review, the following outcome definitions apply:  1. Overall survival: time from date randomised to date of death (any cause).  2. Time to progression: time from date randomised to date of progression or death (any cause). May also be referred to as progression‐free survival.  3. Time to treatment failure: time from date randomised to date of progression, death (any cause), withdrawal due to adverse event, participant refusal, or further anticancer therapy for documented progression.  4. Objective tumour response rate: the proportion of participants with a complete or partial response.

This review also attempted to investigate treatment‐related death which, for the purpose of this review, was defined as death due to the toxicity of the drug and not to disease progression. If an individual trial did not include the definition used by that trial but used the terms "toxic death" or "lethal toxicity", then we included the information in the review.

Search methods for identification of studies

Electronic searches

For the review update, we searched the following databases or registries.

Cochrane Breast Cancer Group (CBCG) Specialised Register on 14 February 2013. Details of the search strategy applied by the Group to create the register, and the procedure used to code references, are described in the Group's module on the Cochrane Library. The register includes both published and unpublished (including ongoing) trials. The CBCG codes 'advanced' and 'chemotherapy' were applied to the specialised register and combined with the keywords (imported with the references from MEDLINE) 'Taxol', 'docetaxel', or 'paclitaxel', and a search of all non‐indexed fields for the following text words: taxane, taxanes, taxol, taxotere, paclitaxel, paxene, nsc‐125973, docetaxel, or anzatax.
MEDLINE (via OvidSP) from 2008 to February 2013, see Appendix 1
EMBASE (via Embase.com) from 2008 to 14 February 2013, see Appendix 2
World Health Organization International Clinical Trials Registry Platform search portal (http://apps.who.int/trialsearch/) for all prospectively registered and ongoing trials on 14 February 2013, see Appendix 3
ClinicalTrials.gov register (http://clinicaltrials.gov/ct2/search) on 14 February 2013 for additional unpublished and ongoing studies, see Appendix 4

Searching other resources

We also searched the reference lists of other related literature reviews. In the original Cochrane Review, the systematic reviews searched included Fossati 1998 and Stockler 2000 as well as review articles identified by the search strategy. In the 2013 review update, we screened the references in the systematic review by Piccart‐Gebhart 2008.

We obtained a copy of the full article for each reference reporting a potentially eligible trial. In the 2013 review update, the review authors contacted the trial authors to provide additional information if data were available in abstract form only.

Data collection and analysis

Selection of studies

In the original review and 2013 review update, two review authors (original review: DG, ED; 2013 review update: MC, MW) applied the selection criteria to each reference identified by the search strategy. A third review author (NW) resolved any discrepancies regarding eligibility. We recorded studies deemed ineligible in the 'Characteristics of excluded studies' table. Articles in languages other than English were translated where required. Only one study required translation, from Hungarian into English (Szanto 2001).

Data extraction and management

In the original review and 2013 review update, two review authors (original review: DG, ED; 2013 review update: MC, MW) independently extracted the data and resolved queries through discussion with a third review author (NW), National Health and Medical Research Council Clinical Trial Centre statisticians, and the CBCG's Statistical Editor. We extracted data on study accrual, randomisation methods, participants' baseline characteristics (that is age, first‐line/second‐line, prior anthracyclines/anthracycline naïve), chemotherapy regimens (number of cycles and duration), outcome definitions, follow‐up, and analyses conducted. We collected multiple publications on the same study and assigned the most complete report (that is the one with the outcomes most relevant to the review or the most recent outcomes) as the primary reference.

Assessment of risk of bias in included studies

We used The Cochrane Collaboration's 'Risk of bias' assessment tool to assess potential sources of bias in the included studies (Higgins 2011). In the 2013 review update, two review authors (MC, MW) independently assessed the potential risk of bias for each study; any differences in judgement were resolved through discussion. The domains assessed were random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. We assigned ratings of 'high', 'low', or 'unclear' risk of bias to each domain for each included study following the criteria outlined in theCochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). For phase III oncology studies, open‐label studies are common owing to difficulty in concealing different chemotherapy schedules, toxicities, etc. The blinding of outcome assessment domain was therefore lumped into those outcome measures most unlikely or most likely to be influenced by a lack of blinding. The outcomes were segregated into (a) overall survival (b) progression‐free survival, time to treatment failure, response rates, and toxicity and (c) quality of life.

Measures of treatment effect

We analysed overall survival, progression‐free survival, and time to treatment failure as time‐to‐event outcomes for which the hazard ratio (HR) is the most appropriate statistic. When possible, the HR and associated variances were extracted directly from the trial publication(s). If it was not reported, we obtained it indirectly employing the methods described by Tierney et al using either other available summary statistics (Tierney 2007), or from data extracted from published Kaplan‐Meier curves (Parmar 1998; Tierney 2007). In studies that did not report the relevant effect estimates and required curve extraction, we adjusted the numbers at risk based on estimated minimum and maximum follow‐up times. If these were not reported in any of the available reports, we estimated minimum follow‐up using the estimated time taken to complete treatment, and estimated maximum follow‐up using the last event reported in the relevant time‐to‐event curve (as per methods in Tierney 2007). We have recorded these follow‐up estimates in the 'Characteristics of included studies' table under 'Notes'. A HR less than 1.0 favoured regimens containing taxanes.

We analysed response rates as dichotomous variables (complete or partial versus stable disease or no response) and derived a pooled risk ratio (RR) with 95% confidence interval (CI). As trialists usually report RRs for both randomised and assessable participants, the same was done in the original and updated review. A RR larger than 1.0 favoured regimens containing taxanes.

We analysed toxicity data as dichotomous outcomes and added up the total number of grade 3 and 4 events and number at risk across trials. In the original review, a single odds ratio (with 95% CIs) was calculated in assessable and randomised participants. In the 2013 review update, we calculated the total number of grade 3 and 4 events and the number of assessable participants in each treatment arm and derived a pooled RR with 95% CI. The denominator was the number of assessable (not randomised) participants to ensure that toxicity outcomes were only measured in those participants who actually received the treatment. In those studies where the number assessable was not provided, we used the number of participants randomised to each treatment group. We have outlined any deviations from assessable number of participants as the denominator in the 'Characteristics of included studies' 'Notes' section. We extracted the total number of toxic events for treatment‐related death, leukopaenia, nausea or vomiting, neurotoxicity, and alopecia. If grade 3/4 nausea and vomiting were reported separately, we used data for vomiting.

We collected quality of life data using a variety of instruments across trials. These data were not statistically synthesised but were summarised and evaluated qualitatively.

Unit of analysis issues

Three studies were three‐arm trials (ECOG E1193: split into ECOG E1193 (A) and ECOG E1193 (B); JCOG9802; Rugo). The three treatment regimens in the ECOG E1193 study were eligible for both subquestions A (ECOG E1193 (A)) and C (ECOG E1193 (B)). This was taken into account when the overall effect of taxanes was calculated by halving the control group each time the trial was used (which was twice). In JCOG9802, data from two arms were used. We excluded the alternating treatment regimen of doxorubicin plus cyclophosphamide versus alternating doxorubicin plus cyclophosphamide and docetaxel. We included the treatment comparison doxorubicin plus cyclophosphamide versus docetaxel. In Rugo, there was one experimental arm (that is taxane‐containing regimen) and two control arms (ixabepilone plus bevacizumab, with different schedules for both drugs). In this case, the treatment arm where participants were randomised to receive bevacizumab 10 mg/kg every 14 days was the most appropriate control comparator arm for the taxane‐containing arm. We did not include the third control arm in this analysis.

Dealing with missing data

In the original review, no attempt had been made to contact most trial investigators for additional information. Many trials were in active follow‐up and others were still recruiting participants. The UK Medical Research Council had been contacted in relation to the UKCCCR AB01 trial, but additional information on this trial was not yet available at that time. Aventis was also contacted regarding the Nabholtz trial, but further details were not yet available.

In the 2013 review update, we contacted a number of trialists to obtain time‐to‐event data and clarification on whether or not analyses had been adjusted in the trial publication. The following trial reports provided additional information, or, in some cases, unpublished manuscripts: Blohmer, EU‐93011, JCOG9802, Lyman, TOG, and Yardley. We have provided further details of the data obtained in the 'Characteristics of included studies' table 'Notes' section.

Assessment of heterogeneity

We used the Chi² test and the I² statistic to test for heterogeneity over all trials, as well as visual inspection of forest plots (Higgins 2011). For the Chi² test, a P value of 0.10 indicated evidence of heterogeneity. We used the I² statistic as a rough guide to assess heterogeneity: 0% to 40% might not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; and 75% to 100% considerable heterogeneity. We evaluated the value of the I² statistic alongside the magnitude and direction of effects, and the P value for the Chi² test (Higgins 2011).

When appropriate, we used a fixed‐effect model for the primary analysis. We considered and discussed heterogeneity between results in the Discussion section of the review and tested where appropriate (see Subgroup analysis and investigation of heterogeneity and Sensitivity analysis sections).

Assessment of reporting biases

We assessed reporting bias using Cochrane's 'Risk of bias' tool (Higgins 2011). We used trial registers (WHO ICTRP and ClinicalTrials.gov) and published protocols (where available) to cross‐check the reporting of outcomes in the trial publications.

Data synthesis

For time‐to‐event outcome data, we obtained a pooled HR from the derived observed (O) ‐ expected (E) number of events and the variance for each trial using the fixed‐effect model (Yusuf 1985). The pooled HR represents the overall risk of an event on taxane‐containing chemotherapy versus non‐taxane‐containing chemotherapy.

For objective tumour response rates and treatment‐related death, we obtained a pooled RR using the fixed‐effect model (Mantel‐Haenszel analysis).

For leukopaenia, nausea/vomiting, neurotoxicity, and alopecia, we obtained a pooled RR using the random‐effects model (Mantel‐Haenszel analysis).

We have narratively described and presented quality of life results in Table 1.

1. Quality of life (QoL).

Trial ID	Instruments used	Summary of findings
303 Study Group	Participants completed EORTC QLQ‐C30 and physicians completed KPS	80% of assessable participants completed QoL assessments in both groups for the first 4 cycles, but was higher in the docetaxel group from cycle 6. There was no statistically significant difference between the 2 groups in mean decreases in global health and physical functioning scores from baseline
304 Study Group	Participants completed EORTC QLQ‐C30	72% of questionnaires returned for docetaxel and 68% for MV for baseline and cycle 2, but deteriorated to 59% for docetaxel and 61% for MV by cycle 8. Attrition higher in MV compared to docetaxel, and did not occur at random. Significantly higher proportion of participants in MV discontinued treatment due to deterioration in condition; trial authors concluded that participants in the poorest health did not complete QoL questionnaires, hence QoL may be underestimated in both groups. Groups similar at baseline for global health, physical functioning, and symptoms except for role functioning and diarrhoea (imbalance in favour of docetaxel). Results: No significant difference in global health status. Significant difference in favour of docetaxel for nausea/vomiting and loss of appetite, and in favour of MV for role and social functioning
306 Study Group	Participants completed EORTC QLQ‐C30 and QLQ‐BR23 (Breast cancer module) 3 days before first infusion then before every alternate cycle and at each visit during follow‐up until progression; and physicians completed KPS	Overall compliance was high through to cycle 6 (> 70%), then decreased during follow‐up (< 30%), although rates in both groups comparable. At cycle 8, more data were missing in AC group than in AT group. Baseline scores were comparable and remained constant during the study. There was no significant difference between groups in global health status/QoL score
ANZ TITG	Participants completed linear analog scales, and physicians completed Spitzer Quality of Life Index	Most QoL measures (physical well‐being, mood, nausea and vomiting, appetite, overall quality of life, and physician‐rated quality of life) were slightly better in the taxane arm. The exception is pain which was slightly better in the non‐taxane arm. Differences were not statistically significant
ECOG E1193: ECOG E1193 (A) and ECOG E1193 (B)	Participants completed FACT‐B	93% (687/738) of randomised participants and 94% (640/683) of eligible participants completed the baseline survey. 70% (451/683) of eligible participants completed the survey at week 16. There was no statistically significant difference in overall QoL score or in any of the subscales between any of the treatment groups
EORTC 10923	Participants completed EORTC QLQ‐C30 and Rotterdam Symptom Checklist	64% of randomised participants completed baseline EORTC QLQ‐C30 and 61% completed baseline Rotterdam Symptom Checklist. QoL comparisons were only performed for the first 3 cycles. There was no difference in global health status/QoL between the 2 groups. Doxorubicin was associated with significantly more nausea/vomiting, loss of appetite, and burden of disease and treatment, but less bone pain and rash than paclitaxel
EORTC 10961	Participants completed EORTC QLQ‐C30 and QLQ‐BR23 (Breast cancer module)	79% of participants completed the baseline questionnaire. Overall compliance (over 4 assessments) was 66%. There was no significant difference in health‐related QoL between the 2 treatment groups
Jassem	Participants completed EORTC QLQ‐C30 and QLQ‐BR23 (Breast cancer module)	81% of questionnaires returned for AT patients and 77% for FAC patients (throughout study and follow‐up), although compliance deteriorated over time. Information on non‐compliers not reported. No statistically significant differences in changes from baseline in functional scales for role, emotional, cognitive, social, global health status, body image, sexual enjoyment, or future perspective. Significant difference in favour of FAC for physical and sexual functioning scales, pain, fatigue, insomnia, and diarrhoea. Significant difference in favour of AT for nausea and vomiting. There was no significant difference in other symptoms.
JCOG9802	Participants completed FACT‐B	99% of the first 150 participants (i.e. 148/150) returned completed questionnaires at baseline, 89% at 6 weeks, and 87% at 18 weeks. There was no statistically significant difference between the 2 treatment arms of interest (for this review) at baseline, 6 weeks, or 18 weeks
Meier	Participants completed EORTC QLQ‐C30	102/120 participants completed QoL questionnaires. There was no significant difference between treatment groups at baseline. Compliance declined (54% at cycle 4), thereby making QoL comparisons difficult. A non‐significant trend for better scores in participants continuing on original docetaxel treatment was noted
Sjostrom	Participants completed EORTC QLQ‐C30	82% of questionnaires were returned over the entire study (overall compliance). Physical deterioration was greater in the methotrexate + fluorouracil group, hence possible bias in its favour. No statistically significant difference at baseline or by cycle 4 in any functional or symptom scale. No significant difference in median values of mean changes in QoL scores from baseline to cycle 6.
UKCCCR AB01	Participants completed FACT‐B	Abstract available in 2001 reported that QoL was "similar for both arms during treatment". No other results were available in the 2001 abstract or full trial publication in 2005

Question	Subgroups
Subgroups within Question B:	substitution of cyclophosphamide with taxane (306 Study Group, AGO, Blohmer, EORTC 10961, Lyman, UKCCCR AB01) substitution of fluorouracil + cyclophosphamide with taxane (Bonneterre, Bontenbal, Jassem) substitution of fluorouracil with taxane (Nabholtz) substitution of anthracycline with taxane (no studies identified)
Subgroups within Question C:	single‐agent taxane vs single‐agent anthracycline (303 Study Group, ECOG E1193 (A), EORTC 10923) single‐agent taxane vs non‐anthracycline single agent (Dieras, Talbot) single‐agent taxane vs anthracycline‐containing combination (no studies identified) single‐agent taxane vs non‐anthracycline‐containing combination (304 Study Group; ANZ TITG; Dieras; Nabholtz; Sjostrom; TOG; TXT)

Study	Variation in definitions and reporting of 'time to progression' used in 2013 Cochrane review update
303 Study Group	TTP: date randomised to date of progression or death
304 Study Group	TTP: date randomised to date of progression or death
306 Study Group	TTP: date randomised to date of first progression
AGO	PFS: no definition provided in the abstract.  Data for this outcome were not included in the review due to an inadequate amount of information presented in the abstract
ANZ TITG	PFS: date randomised to date of progression or death without progression
Blohmer	TTP: time from registration until disease progression
Bonneterre	TTP: not defined in the trial publication
Bontenbal	TTP: date of random assignment to the date of progression, death, or withdrawal
CECOG BM1	TTP: dates of randomisation until disease progression or death, whichever occurred first
Dieras	TTP: not defined in trial publication
ECOG E1193: ECOG E1193 (A) and ECOG E1193 (B)	TTF: date randomised to date of progression, toxic death, death attributed to breast cancer within 6 weeks of date last known alive with stable disease
EORTC 10923	PFS: date randomised to date of progression or death if it occurred before documentation of progressive disease
EORTC 10961	PFS: randomisation to date of progression or death or whichever occurred first
EU‐93011	TTP: i.e. progression‐free survival, the duration from randomisation until progressive disease, death, or last contact. Data presented for this outcome was incomplete (i.e. the number of events was not provided in the manuscript)
HERNATA	TTP: date of randomisation to date of documented progression with censoring for participants alive at last visit/date of death
Jassem	TTP: not defined in trial publication
JCOG9802	PFS: date of randomisation to the date of the first documentation of disease progression or death from any cause
Meier	TTP: not defined in trial publication. Inadequate information presented in the trial publication to allow accurate data extraction. Trial authors were contacted for additional information
Nabholtz	TTP: no definition provided in the abstract.  Data for this outcome were not included in the review due to an inadequate amount of information presented in the abstract
Rugo	PFS: time from randomisation to disease progression or death
Sjostrom	TTP: date randomised to date of progression or death or last follow‐up visit
Talbot	TTP: interval between first day of treatment and first recording of disease progression or death. Data for this outcome were not included in the review as limited information in the trial publication owing to premature discontinuation of the trial
TOG	TTP: duration between the first day of study treatment and date of progression
TRAVIOTA	TTP: defined according to the RECIST criteria. Data for this outcome were not included in the review as we were unable to accurately estimate the length of follow‐up
TXT	TTP: time of first treatment infusion to first objective evidence of tumour progression
UKCCCR AB01	PFS: time from random assignment to first appearance of progressive disease or death from any cause
Yardley	PFS: interval from first study treatment until the date that the first progression of breast cancer was documented

Trial ID	Outcome	Reason not included
303 Study Group	Toxicity: alopecia	Data reported but not as grade 3 or 4 toxicity, therefore it was not possible to calculate
304 Study Group	Toxicity: alopecia	Reported but no numerical data provided
AGO	Time to progression	Inadequate amount of information presented in abstract form; we contacted trialists but received no reply
Bonneterre	Time to progression (sensitivity analysis undertaken)	The number of events in each group for time to progression were not provided in the trial publication; individual participant data from the Piccart‐Gebhart 2008 systematic review were used instead
CECOG BM1	Overall survival	The trial publication stated that the data for this outcome are not yet mature
Nabholtz	Overall survival, time to progression	Inadequate amount of information presented in abstract form; we contacted trialists but received no reply
Rugo	Overall survival	Data for arms A and C (comparable control arm) were immature at the time of analysis
Talbot	Overall survival, time to progression	Data for outcome were not provided in the trial publication owing to premature discontinuation of the trial
TRAVIOTA	Time to progression, time to treatment failure	Duration of follow‐up not provided in the trial publication, therefore it was not possible to estimate the number of events in the taxane‐containing or non‐taxane‐containing arms, or hazard ratio
EU‐93011	Time to progression, toxicity: alopecia	Time to progression: unpublished manuscript did not provide the number of events in each treatment arm. Alopecia: reported but no numerical data provided

Date	Event	Description
14 February 2013	New citation required but conclusions have not changed	Ten new studies were included, adding 3228 participants. A further two studies 'awaiting classification' and 11 'ongoing studies' have been identified
14 February 2013	New search has been performed	Performed search for new studies on 14 February 2013
13 August 2008	Amended	Converted to new review format.
3 February 2005	New citation required and conclusions have changed	Substantive amendment

1	randomized controlled trial.pt.
2	controlled clinical trial.pt.
3	randomized.ab.
4	randomised.ab.
5	placebo.ab.
6	randomly.ab.
7	trial.ab.
8	groups.ab.
9	1 or 2 or 3 or 4 or 5 or 6 or 7 or 8
10	advanced breast cancer$.tw,sh.
11	advanced breast carcinoma$.tw,sh.
12	advanced breast tumour$.tw,sh.
13	advanced breast tumor$.tw,sh.
14	advanced breast neoplasm$.tw,sh.
15	metastatic breast cancer$.tw,sh.
16	metastatic breast carcinoma$.tw,sh.
17	metastatic breast tumour$.tw,sh.
18	metastatic breast tumor$.tw,sh.
19	metastatic breast neoplasm$.tw,sh.
20	10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19
21	taxane containing regimen$.tw,sh.
22	taxane containing chemotherapy regimen$.tw,sh.
23	exp Taxoids/
24	exp Paclitaxel/
25	docetaxel.tw,sh.
26	taxane$.tw,sh.
27	taxol.tw,sh.
28	taxotere.tw,sh.
29	paxene.tw,sh.
30	nsc‐125973.tw,sh.
31	anzatax.tw,sh.
32	4alpha.tw,sh.
33	7‐epi‐taxol.tw,sh.
34	21 or 22 or 23 or 24 or 25 or 26 or 27 or 28 or 29 or 30 or 31 or 32 or 33
35	9 and 20 and 34
36	limit 35 to (humans and yr="2010 ‐Current")

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Overall effect: Taxane‐containing regimens vs. not	23	6008	Hazard Ratio (95% CI)	0.93 [0.88, 0.99]
1.1 Regimen A plus taxane v Regimen A	2	518	Hazard Ratio (95% CI)	0.98 [0.81, 1.20]
1.2 Regimen A plus taxane v Regimen B	9	2645	Hazard Ratio (95% CI)	0.92 [0.84, 1.00]
1.3 Single agent taxane v Regimen C	12	2845	Hazard Ratio (95% CI)	0.94 [0.87, 1.03]
2 First‐line trials only: overall	16	4439	Hazard Ratio (95% CI)	0.93 [0.87, 0.99]
2.1 Regimen A plus taxane v Regimen A	2	518	Hazard Ratio (95% CI)	0.98 [0.81, 1.20]
2.2 Regimen A plus taxane v Regimen B	9	2645	Hazard Ratio (95% CI)	0.92 [0.84, 1.00]
2.3 Single agent taxane v Regimen C	5	1276	Hazard Ratio (95% CI)	0.93 [0.83, 1.05]
3 Subquestions A, B & C	23		Hazard Ratio (95% CI)	Subtotals only
3.1 Regimen A plus taxane v Regimen A	2	630	Hazard Ratio (95% CI)	1.00 [0.84, 1.18]
3.2 Regimen A plus taxane v Regimen B	9	2645	Hazard Ratio (95% CI)	0.92 [0.84, 1.00]
3.3 Single agent taxane v Regimen C	12	2957	Hazard Ratio (95% CI)	0.95 [0.87, 1.03]
4 Chemotherapy regimens	12		Peto Odds Ratio (95% CI)	Subtotals only
4.1 Single agent taxane vs single agent anthracycline	4	1212	Peto Odds Ratio (95% CI)	1.02 [0.90, 1.16]
4.2 Single agent taxane vs non‐anthracycline combination	8	1736	Peto Odds Ratio (95% CI)	0.94 [0.84, 1.06]
5 Type of taxane	23	6008	Hazard Ratio (95% CI)	0.93 [0.88, 0.99]
5.1 Paclitaxel containing	10	2834	Hazard Ratio (95% CI)	1.01 [0.93, 1.10]
5.2 Docetaxel containing	13	3174	Hazard Ratio (95% CI)	0.87 [0.80, 0.94]
6 Prior anthracyclines	23		Hazard Ratio (95% CI)	Subtotals only
6.1 Prior anthracyclines	6	1243	Hazard Ratio (95% CI)	0.97 [0.85, 1.11]
6.2 Anthracyclines naive	17	4765	Hazard Ratio (95% CI)	0.93 [0.87, 0.99]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Overall effect: Taxane‐containing regimens vs not	22	5960	Hazard Ratio (95% CI)	0.92 [0.87, 0.97]
1.1 Regimen A plus taxane v Regimen A	1	342	Hazard Ratio (95% CI)	0.97 [0.76, 1.25]
1.2 Regimen A plus taxane v Regimen B	10	2891	Hazard Ratio (95% CI)	0.90 [0.83, 0.98]
1.3 Single agent taxane v Regimen C	11	2727	Hazard Ratio (95% CI)	0.92 [0.85, 1.00]
2 First‐line trials only: overall	16	4509	Hazard Ratio (95% CI)	0.96 [0.90, 1.02]
2.1 Regimen A plus taxane v Regimen A	1	342	Hazard Ratio (95% CI)	0.97 [0.76, 1.25]
2.2 Regimen A plus taxane v Regimen B	10	2891	Hazard Ratio (95% CI)	0.90 [0.83, 0.98]
2.3 Single agent taxane v Regimen C	5	1276	Hazard Ratio (95% CI)	1.08 [0.97, 1.21]
3 Subquestions A, B & C	22		Hazard Ratio (95% CI)	Subtotals only
3.1 Regimen A plus taxane v Regimen A	1	454	Hazard Ratio (95% CI)	0.99 [0.81, 1.21]
3.2 Regimen A plus taxane v Regimen B	10	2891	Hazard Ratio (95% CI)	0.90 [0.83, 0.98]
3.3 Single agent taxane v Regimen C	11	2839	Hazard Ratio (95% CI)	0.93 [0.86, 1.00]
4 Subquestions A, B & C: first‐line only	16		Hazard Ratio (95% CI)	Subtotals only
4.1 Regimen A plus taxane v Regimen A	1	454	Hazard Ratio (95% CI)	0.99 [0.81, 1.21]
4.2 Regimen A plus taxane v Regimen B	10	2891	Hazard Ratio (95% CI)	0.90 [0.83, 0.98]
4.3 Single agent taxane v Regimen C	5	1388	Hazard Ratio (95% CI)	1.03 [0.93, 1.14]
5 Chemotherapy Regimens	11		Peto Odds Ratio (95% CI)	Subtotals only
5.1 Single agent taxane vs single agent anthracycline	4	1212	Peto Odds Ratio (95% CI)	1.08 [0.96, 1.22]
5.2 Single agent taxane vs non‐anthracycline combination	7	1618	Peto Odds Ratio (95% CI)	0.85 [0.76, 0.94]
6 Type of taxane	22		Hazard Ratio (95% CI)	Subtotals only
6.1 Paclitaxel containing	11	3080	Hazard Ratio (95% CI)	1.04 [0.96, 1.12]
6.2 Docetaxel containing	11	2880	Hazard Ratio (95% CI)	0.80 [0.74, 0.86]
7 Prior anthracyclines	22		Hazard Ratio (95% CI)	Subtotals only
7.1 Prior anthracyclines	5	1125	Hazard Ratio (95% CI)	0.76 [0.67, 0.86]
7.2 Anthracycline naive	17	4835	Hazard Ratio (95% CI)	0.96 [0.90, 1.02]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Subquestions A, B & C	5	1724	Hazard Ratio (95% CI)	0.90 [0.82, 0.98]
1.1 Regimen A plus taxane v Regimen B	2	713	Hazard Ratio (95% CI)	1.07 [0.93, 1.23]
1.2 Single agent taxane v Regimen C	3	1011	Hazard Ratio (95% CI)	0.78 [0.69, 0.88]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Overall effect: assessable patients	29	6999	Risk Ratio (M‐H, Fixed, 95% CI)	1.20 [1.14, 1.27]
1.1 Regimen A plus taxane v Regimen A	2	515	Risk Ratio (M‐H, Fixed, 95% CI)	1.52 [1.20, 1.92]
1.2 Regimen A plus taxane v Regimen B	14	3740	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [1.12, 1.26]
1.3 Single agent taxane v Regimen C	13	2744	Risk Ratio (M‐H, Fixed, 95% CI)	1.17 [1.05, 1.31]
2 Overall effect: randomised patients	29	7416	Risk Ratio (M‐H, Fixed, 95% CI)	1.20 [1.14, 1.27]
2.1 Regimen A plus taxane v Regimen A	2	543	Risk Ratio (M‐H, Fixed, 95% CI)	1.55 [1.22, 1.97]
2.2 Regimen A plus taxane v Regimen B	14	3953	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [1.12, 1.26]
2.3 Single agent taxane v Regimen C	13	2920	Risk Ratio (M‐H, Fixed, 95% CI)	1.17 [1.05, 1.31]
3 First‐line trials only: assessable patients	21	5512	Risk Ratio (M‐H, Fixed, 95% CI)	1.16 [1.10, 1.23]
3.1 Regimen A plus taxane v Regimen A	2	515	Risk Ratio (M‐H, Fixed, 95% CI)	1.52 [1.20, 1.92]
3.2 Regimen A plus taxane v Regimen B	14	3740	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [1.12, 1.26]
3.3 Single agent taxane v Regimen C	5	1257	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.75, 1.05]
4 Overall effect: randomised patients ‐ firstline only	21	5796	Risk Ratio (M‐H, Fixed, 95% CI)	1.16 [1.10, 1.23]
4.1 Regimen A plus taxane v Regimen A	2	543	Risk Ratio (M‐H, Fixed, 95% CI)	1.55 [1.22, 1.97]
4.2 Regimen A plus taxane v Regimen B	14	3953	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [1.12, 1.26]
4.3 Single agent taxane v Regimen C	5	1300	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.76, 1.06]
5 Subquestions A, B & C: assessable patients	29		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
5.1 Regimen A plus taxane v Regimen A	2	627	Risk Ratio (M‐H, Fixed, 95% CI)	1.47 [1.21, 1.79]
5.2 Regimen A plus taxane v Regimen B	14	3740	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [1.12, 1.26]
5.3 Single agent taxane v Regimen C	13	2856	Risk Ratio (M‐H, Fixed, 95% CI)	1.14 [1.03, 1.27]
6 Subquestions A, B & C: randomised patients	29		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
6.1 Regimen A plus taxane v Regimen A	2	665	Risk Ratio (M‐H, Fixed, 95% CI)	1.50 [1.23, 1.84]
6.2 Regimen A plus taxane v Regimen B	14	3953	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [1.12, 1.26]
6.3 Single agent taxane v Regimen C	13	3042	Risk Ratio (M‐H, Fixed, 95% CI)	1.16 [1.04, 1.29]
7 Subquestions A, B & C: assessable patients ‐ first‐line only	21		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
7.1 Regimen A plus taxane v Regimen A	2	627	Risk Ratio (M‐H, Fixed, 95% CI)	1.47 [1.21, 1.79]
7.2 Regimen A plus taxane v Regimen B	14	3740	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [1.12, 1.26]
7.3 Single agent taxane v Regimen C	5	1369	Risk Ratio (M‐H, Fixed, 95% CI)	0.90 [0.77, 1.05]
8 Subquestions A, B & C: randomised patients ‐ firstline only	21		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
8.1 Regimen A plus taxane v Regimen A	2	665	Risk Ratio (M‐H, Fixed, 95% CI)	1.50 [1.23, 1.84]
8.2 Regimen A plus taxane v Regimen B	14	3953	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [1.12, 1.26]
8.3 Single agent taxane v Regimen C	5	1422	Risk Ratio (M‐H, Fixed, 95% CI)	0.90 [0.77, 1.06]
9 Type of taxane: assessable patients	28	6932	Risk Ratio (M‐H, Fixed, 95% CI)	1.21 [1.15, 1.27]
9.1 Paclitaxel	14	3499	Risk Ratio (M‐H, Fixed, 95% CI)	1.06 [0.99, 1.14]
9.2 Docetaxel	14	3433	Risk Ratio (M‐H, Fixed, 95% CI)	1.40 [1.29, 1.51]
10 Prior anthracyclines: assessable patients	29	6999	Risk Ratio (M‐H, Fixed, 95% CI)	1.20 [1.14, 1.26]
10.1 Prior anthracyclines	7	1192	Risk Ratio (M‐H, Fixed, 95% CI)	1.43 [1.20, 1.72]
10.2 Anthracyclines naive	22	5807	Risk Ratio (M‐H, Fixed, 95% CI)	1.17 [1.11, 1.24]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Treatment‐related death: overall effect	22	5517	Risk Ratio (M‐H, Fixed, 95% CI)	1.00 [0.63, 1.57]
1.1 Regimen A plus taxane v Regimen A	0	0	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
1.2 Regimen A plus taxane v Regimen B	11	3195	Risk Ratio (M‐H, Fixed, 95% CI)	1.19 [0.61, 2.33]
1.3 Single agent taxane v Regimen C	11	2322	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.46, 1.59]
2 Leukopaenia: overall effect	28	6564	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.97, 1.17]
2.1 Regimen A plus taxane v Regimen A	2	512	Risk Ratio (M‐H, Random, 95% CI)	1.26 [0.96, 1.66]
2.2 Regimen A plus taxane v Regimen B	13	3209	Risk Ratio (M‐H, Random, 95% CI)	1.11 [1.02, 1.20]
2.3 Single agent taxane v Regimen C	13	2843	Risk Ratio (M‐H, Random, 95% CI)	0.99 [0.82, 1.21]
3 Leukopaenia: subquestions A, B & C	28		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
3.1 Regimen A plus taxane v Regimen A	2	624	Risk Ratio (M‐H, Random, 95% CI)	1.76 [1.11, 2.80]
3.2 Regimen A plus taxane v Regimen B	13	3209	Risk Ratio (M‐H, Random, 95% CI)	1.11 [1.02, 1.20]
3.3 Single agent taxane v Regimen C	13	2955	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.86, 1.34]
4 Nausea or vomiting: overall effect	26	6245	Risk Ratio (M‐H, Random, 95% CI)	0.62 [0.46, 0.83]
4.1 Regimen A plus taxane v Regimen A	2	512	Risk Ratio (M‐H, Random, 95% CI)	0.73 [0.35, 1.50]
4.2 Regimen A plus taxane v Regimen B	12	2990	Risk Ratio (M‐H, Random, 95% CI)	0.79 [0.57, 1.11]
4.3 Single agent taxane v Regimen C	12	2743	Risk Ratio (M‐H, Random, 95% CI)	0.43 [0.26, 0.71]
5 Nausea or vomiting: subquestions A, B & C	26		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
5.1 Regimen A plus taxane v Regimen A	2	624	Risk Ratio (M‐H, Random, 95% CI)	1.13 [0.55, 2.34]
5.2 Regimen A plus taxane v Regimen B	12	2990	Risk Ratio (M‐H, Random, 95% CI)	0.79 [0.57, 1.11]
5.3 Single agent taxane v Regimen C	12	2855	Risk Ratio (M‐H, Random, 95% CI)	0.46 [0.27, 0.78]
6 Neurotoxicity: overall effect	24	5783	Risk Ratio (M‐H, Random, 95% CI)	4.84 [3.18, 7.35]
6.1 Regimen A plus taxane v Regimen A	1	342	Risk Ratio (M‐H, Random, 95% CI)	6.09 [1.47, 25.24]
6.2 Regimen A plus taxane v Regimen B	12	2991	Risk Ratio (M‐H, Random, 95% CI)	4.89 [2.55, 9.38]
6.3 Single agent taxane v Regimen C	11	2450	Risk Ratio (M‐H, Random, 95% CI)	5.14 [2.50, 10.58]
7 Neurotoxicity: subquestions A, B & C	24		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
7.1 Regimen A plus taxane v Regimen A	1	454	Risk Ratio (M‐H, Random, 95% CI)	12.17 [2.92, 50.79]
7.2 Regimen A plus taxane v Regimen B	12	2991	Risk Ratio (M‐H, Random, 95% CI)	4.89 [2.55, 9.38]
7.3 Single agent taxane v Regimen C	11	2562	Risk Ratio (M‐H, Random, 95% CI)	5.99 [2.91, 12.31]
8 Alopecia: overall effect	11	2437	Risk Ratio (M‐H, Random, 95% CI)	2.37 [1.45, 3.87]
8.1 Regimen A plus taxane v Regimen A	0	0	Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
8.2 Regimen A plus taxane v Regimen B	6	1634	Risk Ratio (M‐H, Random, 95% CI)	1.17 [1.02, 1.34]
8.3 Single agent taxane v Regimen C	5	803	Risk Ratio (M‐H, Random, 95% CI)	4.12 [2.94, 5.77]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Overall survival	23	6008	Hazard Ratio (95% CI)	0.93 [0.88, 0.99]
1.1 Low risk of bias	18	4850	Hazard Ratio (95% CI)	0.91 [0.85, 0.97]
1.2 Unclear or high risk of bias	5	1158	Hazard Ratio (95% CI)	1.05 [0.92, 1.20]
2 Time to progression	22	5960	Hazard Ratio (95% CI)	0.92 [0.87, 0.97]
2.1 Low risk of bias	17	4815	Hazard Ratio (95% CI)	0.95 [0.89, 1.00]
2.2 Unclear or high risk of bias	5	1145	Hazard Ratio (95% CI)	0.80 [0.70, 0.90]

Methods	Accrual: 04/1994 to 01/1997  Multicentre, international  Centralised randomisation, method not specified  Slight imbalance in some baseline characteristics (see 'Risk of bias' table)  Median follow‐up: 23 months
Participants	Female  Age range 25 to 74 years, median age 52.0 years in both arms  100% metastatic breast cancer  100% > first‐line  All participants anthracycline naïve
Interventions	Arm 1: docetaxel 100 mg/m² Arm 2: doxorubicin 75 mg/m² Both arms q21 days for maximum 7 cycles
Outcomes	Primary: Time to progression, defined as from date randomised to date of progression or death Secondary: Overall survival, defined as date of randomisation until the date of death for any reason Time to treatment failure, defined as date of randomisation to the date of progression, death for any reason, withdrawal due to an adverse event, participant refusal, or further anticancer therapy before documentation of progression Response Toxicity Quality of life Toxic deaths
Notes	Follow‐up details not reported estimated minimum 5.5 months estimated maximum 34 months (OS), 20 months (PFS), 19.5 months (TTF) All randomised participants included in time‐to‐event analyses
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Patients were assigned randomly" and they used stratified randomisation
Allocation concealment (selection bias)	Low risk	Central allocation. Quote: "The randomization was centralized and stratified for treatment arm by institution"
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Non‐blinded study
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	A non‐blinded study. Unlikely that assessment of overall survival would be influenced by lack of blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Low risk	Imaging and tumour evaluation at the end of cycles 2, 4, and 7 or at least every 3 months. CR had to be confirmed by a second evaluation more than 28 days later. Tumour assessments were reviewed by an independent panel of 2 radiologists and an oncologist. Blood tests and scans (MUGA) or echocardiography conducted. Data "analysed directly from reported laboratory parameters"
Blinding of outcome assessment (detection bias)   QoL	High risk	EORTC QLQ‐C30 questionnaire completed by participants and Karnofsky Performance Status completed by physicians
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Study reported 159/161 participants in the docetaxel group and 163/165 participants in the doxorubicin group received treatment. All participants were included in the efficacy (survival) and safety analyses
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods are reported in the results section of the trial publication
Other bias	Low risk	Similar baseline characteristics in groups except for a slight imbalance in participants with bone metastases (docetaxel 55%, doxorubicin 63%; P = 0.012)

Methods	Accrual: 07/1994 to 02/1997  Multicentre, international  Randomisation "centralised ... with a block design by institution"  Baseline comparability: no significant imbalance apparent or reported except for number of organs involved  Median follow‐up: 19 months
Participants	Female  Age range 30 to 78 years, median age 51.0 (docetaxel) and 52.0 (mitomycin)  100% metastatic breast cancer  19% first‐line, 81% > first‐line  All women had failed previous anthracycline‐containing regimens
Interventions	Arm 1: docetaxel 100 mg/m² q21 days  Arm 2: mitomycin 12 mg/m² q6 weeks + vinblastine 6 mg/m² q21 days  Both arms for a maximum of 10 3‐week cycles
Outcomes	Primary: Time to progression, defined as from date randomised to date of progression or death Secondary: Overall survival, defined as date of randomisation until the date of death for any reason Time to treatment failure, defined as date of randomisation to the date of progression, death for any reason, withdrawal due to an adverse event, participant refusal, or further anticancer therapy before documentation of progression Response Toxicity Quality of life Toxic deaths
Notes	Follow‐up details not reported: estimated minimum 4.5 months estimated maximum 33 months (OS), 19 months (PFS) All randomised participants included in time‐to‐event analyses  Estimate for time‐to‐event outcomes obtained using exact P value and total events
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Patients were randomly assigned" and block randomisation by institution was used
Allocation concealment (selection bias)	Low risk	Central allocation. Quote: "The randomization was centralized at Rhone‐Pulene Rorer, Antony, France"
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Non‐blinded study
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	A non‐blinded study. Unlikely that assessment of overall survival would be influenced by lack of blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Low risk	Imaging and tumour evaluation at the end of cycles 3, 6, 8, and 10 or at discontinuation or at least every 3 months. CR had to be confirmed by a second evaluation more than 28 days later. Tumour assessments were reviewed by an independent panel of 2 radiologists and an oncologist in 10% of participants Blood tests and scans. "Drug safety was analysed directly from reported laboratory parameters"
Blinding of outcome assessment (detection bias)   QoL	High risk	EORTC QLQ‐C30 questionnaire completed by participants and Karnofsky Performance Status used to assess participant's condition from physician's perspective
Incomplete outcome data (attrition bias)   All outcomes	Low risk	Paper reported that 200/203 participants in docetaxel group and 187/189 participants in mitomycin group received treatment, and efficacy analyses used ITT principle. Safety analyses were conducted on all treated participants
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Low risk	Baseline characteristics of both groups were comparable except for the number of organs involved (i.e. ≥ 3 organs affected), however it was not mentioned if the difference was significant

Methods	Accrual: 06/1996 to 03/1998  Multicentre, international  Randomisation centralised (block design)  Baseline comparability: well balanced  Median follow‐up: 49 months
Participants	Median age 52.5 in doxorubicin + docetaxel group and 54 years in doxorubicin + cyclophosphamide group  100% metastatic breast cancer  100% first‐line  Anthracycline naïve
Interventions	Arm 1: doxorubicin + docetaxel (50/75 mg/m²)  Arm 2: doxorubicin + cyclophosphamide (60/600 mg/m²)  Both arms q21 days, maximum 8 cycles
Outcomes	Primary: Time to progression (defined as from date randomised to date of first progression) Secondary: Overall survival Time to treatment failure Response Toxicity Toxic death
Notes	Follow‐up details not reported  Report numbers at risk on survival curve  All randomised participants included in time‐to‐event analyses  Estimate for time to progression obtained from reported hazard ratio and 95% confidence intervals. Estimate for overall survival obtained from P value and total events. Estimate for time to treatment failure obtained from time‐to‐event curve
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote from abstract: "Patients were randomly assigned to receive doxorubicin". Randomisation was centralised with block design
Allocation concealment (selection bias)	Low risk	Quote: "The randomization was centralized..."
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Non‐blinded study
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	A non‐blinded study. Unlikely that assessment of overall survival would be influenced by lack of blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Low risk	Imaging and tumour evaluation after cycles 3, 6, 8 or study treatment discontinuation and then every 2 months until disease progression or death. "All tumour assessments from patients with radiologically assessable disease were reviewed by an independent expert panel" (3 radiologists and 1 medical oncologist) Weekly blood counts performed. Measurement of LVEF performed after cycles 3, 6, 8 and as clinically indicated
Blinding of outcome assessment (detection bias)   QoL	High risk	EORTC C30 and QLQ‐BR23 completed by participants and Karnofsky Performance Status completed by physicians
Incomplete outcome data (attrition bias)   All outcomes	Low risk	213/214 participants on AT and 210/215 participants on AC received treatment. Efficacy analyses performed on assessable and ITT populations. Safety analyses on all treated participants
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods are reported in the results section of the trial publication
Other bias	Low risk	Quote: "Baseline characteristics were well balanced and major negative prognostic factors were similar in both groups"

Methods	Accrual: 10/1996 to 05/1999  Multicentre, national (Germany)  Randomisation method not specified  Baseline comparability: no significant imbalance reported  Median follow‐up: 36 weeks
Participants	Median age 56 years in both arms  26% had "locally metastatic disease"  In the remainder the main site of metastases were liver or lung (although proportions do not add up)  100% first‐line  Anthracycline naïve (except for 3% of participants in epirubicin + paclitaxel group)
Interventions	Arm 1: epirubicin 60 mg/m² + paclitaxel 175 mg/m²)  Arm 2: epirubicin 60 mg/m² + cyclophosphamide 600 mg/m²) Both arms q21 days for at least 6 cycles
Outcomes	Outcomes were not separated into primary and secondary: Response Toxicity Progression‐free survival (no definition provided in the abstract)
Notes	Abstract only available. Full article has not yet been published
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "Patients were randomly assigned to receive either ...". No additional details were provided in the abstract
Allocation concealment (selection bias)	Unclear risk	Method of concealment was not described in the abstract
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information provided in the abstracts available
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Assessment of overall survival is unlikely to be influenced by no or incomplete blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	No information provided in the abstracts available. NCI‐CTC used. Comment: blood tests were assumed to be conducted
Incomplete outcome data (attrition bias)   All outcomes	Unclear risk	No reporting of attrition or exclusions in the available abstract
Selective reporting (reporting bias)	Unclear risk	Unable to assess from the abstract
Other bias	Low risk	Quote: "Prognostic factors were balanced between treatment arms"

Methods	Accrual: 30 September 1993 to 30 September 1995  Multi‐centre, international  Centralised, computer‐generated randomisation  Baseline comparability: no significant imbalance reported  Median follow‐up: 26 months
Participants	Age range 32 to 80 years, median age 54 years in both arms  Approximately 90% metastatic breast cancer  100% first‐line  Did not exclude prior anthracycline use, however only 1% of participants on CMFP and 0 participants on paclitaxel had received anthracyclines
Interventions	Arm 1: paclitaxel 200 mg/m²   Arm 2: cyclophosphamide 100 mg/m²+ methotrexate 40 mg/m²+ fluorouracil 600 mg/m²+ prednisone 40 mg/m² (CMFP)  Both arms q21 days for 8 courses
Outcomes	Outcomes were not separated into primary or secondary: Overall survival Progression‐free survival, defined as from date randomised to date of progression or death without progression Response Toxicity Quality of life
Notes	Follow‐up details reported minimum 17 months maximum 40 months All randomised participants included in time‐to‐event analyses  Estimate for time‐to‐event outcomes obtained using exact P value and total events
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Patients were stratified by institution and randomized to receive either...". Statistical methods section states "randomization was based on an adaptive biased coin procedure with a bias of 3n at each allocation in favor of the arm with n fewer patients"
Allocation concealment (selection bias)	Low risk	Quote: "Computer‐generated randomization charts were prepared for each center and held at the Statistical Centre at Peter MacCallum Cancer Institute, Melbourne, Australia"
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in trial publication
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Assessment of overall survival is unlikely to be influenced by no or incomplete blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	Weekly blood tests. Scans to evaluate disease repeated after 12 weeks and 24 weeks on therapy. Tests also repeated at time of suspected relapse or progression and at intervals no less than 4 weeks apart from PR or CR
Blinding of outcome assessment (detection bias)   QoL	High risk	QoL linear analog scales completed by participants and Spitzer QoL index completed by physicians (p. 2356)
Incomplete outcome data (attrition bias)   All outcomes	Low risk	105/107 participants received paclitaxel and 99/102 participants received CMFP. "All major endpoints were compared using ITT analysis that included all randomised patients"
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods are reported in the results section of the trial publication
Other bias	Unclear risk	Baseline characteristics (e.g. ECOG performance status) provided by treatment arm but no indication whether there were any significant differences between groups. 3 of 7 potential prognostic factors in this study (includes ECOG performance status) "were shown to have a significant influence on survival"

Methods	Accrual: 02/2000 to 11/2003  Multicentre, open‐label, randomised phase III trial at 49 centres in Germany  Randomisation “centralised with a block design by study centre”  Imbalance in the number of participants randomised to each arm (that is epirubicin + docetaxel group, n = 125; epirubicin + cyclophosphamide group, n = 111)  Median follow‐up: 24 months
Participants	Age range 31 to 73 years, median age 57 years (epirubicin + docetaxel group) and 56 years (epirubicin + cyclophosphamide group)  100% metastatic breast cancer  First‐line  No previous chemotherapy/anthracyclines allowed
Interventions	Arm 1: epirubicin 75 mg/m² (IV bolus or infusion for 10 min) + docetaxel 75 mg/m² (IV infusion over 1 hour) Arm 2: epirubicin 90 mg/m² (IV) + cyclophosphamide 600 mg/m² (IV over 30 min)  Treatment (both arms) q21 days, 6 cycles and in some cases 8 cycles if “maximum benefit had not been reached”
Outcomes	Primary: Time to progression, defined as the time from registration until disease progression Secondary: Overall survival, defined as date of registration to the date of death for any reason Objective response rates Adverse events and toxic effects
Notes	Trial prematurely stopped due to inadequate accrual and unlikelihood of reaching primary end point. We contacted the trialists (Peter Schmid) who provided the number of events in each treatment arm for the outcomes OS and PFS.  Method 4 (Tierney 2007) was then used to estimate O‐E, and V. For toxicity, the number of randomised women was used as the denominator.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: “patients were randomly assigned to receive either...”. Randomisation was centralised with a block design
Allocation concealment (selection bias)	Low risk	Quote: “The randomisation was centralized with a block design by study centre”
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Open‐label study
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Open‐label study. Unlikely that assessment of overall survival would be influenced by lack of blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	Tumour lesions were assessed at end of treatment cycles 2, 4, 6, 8 or at study discontinuation and then every 2 months. No description of how evaluations were done Adverse events and toxicity were assessed weekly and recorded for every cycle. Scans were used to assess cardiac function at baseline and after cycles 3 and 6, and at the end of the study
Incomplete outcome data (attrition bias)   All outcomes	Low risk	240 participants enrolled; 4 participants (1.7%) did not receive study medication and were excluded from ITT and safety analyses
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Unclear risk	Trial prematurely terminated due to inadequate accrual and results from the interim analysis

Methods	Accrual: 09/1998 to 11/2000  Multicentre, national (France)  Randomisation centralised and stratified according to centre  Baseline comparability: no significant imbalance reported, although ET arm had a higher proportion of participants with an original diagnosis of stage IV disease and more advanced disease than FEC arm  Median follow‐up: 23.8 months
Participants	Female  Age range 23 to 73 years, median age 54 years for both arms  100% metastatic breast cancer  100% first‐line  Slightly higher number of participants had ER or PR positive tumours in FEC group
Interventions	Arm1:  ET: epirubicin 75 mg/m² over 10 min + docetaxel 75 mg/m² Arm 2:  FEC: fluorouracil 500 mg/m² over 1 hour + epirubicin 75 mg/m² over 10 min + cyclophosphamide 500 mg/m²   All agents given once q21 days for up to 8 cycles
Outcomes	Primary: Objective response rate, taken as the best response obtained from the start of treatment until disease progression Secondary: Overall survival, defined as the time from beginning treatment to the time of death from any cause or the date of first contact if death was not recorded before the cutoff date Time to progression, not defined Duration of response, defined as the time from complete or partial response to the time that recurrent or progressive disease or death was first noted Toxicity, graded using the NCI‐CTC criteria
Notes	Abstract was only available in the previous version of this Cochrane review. The full article was published in 2004. For overall survival, hazard ratios were estimated using methods outlined by Parmar 1998; for TTP, individualised participant data from Piccart‐Gebhart 2008 was used and sensitivity analysis conducted
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: “centralised predefined schedule with randomisation stratified according to centre..”.
Allocation concealment (selection bias)	Low risk	Quote: “...centralised predefined schedule with randomisation stratified according to centre to one of two treatment arms”
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Open‐label study
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Unlikely that overall survival assessment was influenced by unblinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Low risk	Physical examinations and blood tests run at the beginning of each treatment cycle. Target lesions assessed every 2 cycles. Responses reviewed by an independent committee of radiologists Scans were performed every 4 cycles and then every 2 cycles once maximum cumulative dose reached (anthracyclines)
Incomplete outcome data (attrition bias)   All outcomes	Low risk	65/70 participants received ET and 67/72 participants FEC. ITT analysis undertaken on all (142) participants. For response rates, analyses used ITT population and assessable population
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Unclear risk	Baseline characteristics generally comparable; however ET group had greater metastatic involvement of lung, liver, skin, bone or > 3 organs; FEC group had a higher proportion of participants with oestrogen or progesterone receptor positive tumours

Methods	Accrual: 03/1997 to 04/2002  Multicentre, national (Netherlands)  Randomisation (by central telephone number) stratified for centre, previous chemotherapy, WHO performance status, and presence of bone metastases  Baseline comparability: no significant imbalance reported  Median follow‐up: 8.7 months
Participants	Age range 30 to 70 years, median age 53 years in AT group and 54 years in FAC group  100% metastatic breast cancer  100% first‐line
Interventions	Arm 1:  AT: doxorubicin + docetaxel 50/75 mg/m²   Arm 2:  FAC: fluorouracil, doxorubicin + cyclophosphamide 500/50/500 mg/m² Both arms q21 days for a maximum of 6 cycles
Outcomes	Primary: Objective response rate Secondary: Overall survival Time to progression, defined as the date of random assignment to the date of progression, death, or withdrawal Duration of response Toxicity, assessed before each new cycle and graded according to the NCI‐CTC
Notes	Phase III trial prematurely closed due to an unplanned interim analysis that indicated that the chance of finding a statistical difference in TTP between treatment arms had become too low. For TTP, method 3 was used to estimate O‐E and V (Tierney 2007). For toxicity, the number of randomised women was used as the denominator.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote from abstract: “patients were randomly assigned to either...” and involved stratification randomisation based on centre, prior chemotherapy, WHO performance status, and presence of bone metastases
Allocation concealment (selection bias)	Low risk	Quote: “randomisation performed by calling a central telephone number who stratified for center.”
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Non‐blinded study
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Unlikely that overall survival assessment was influenced by unblinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Low risk	X‐ray, ultrasonography, and MRI, etc. used to assess tumour status. Up to a max of 8 representable lesions were evaluated after cycles 2, 4, 6 then every 2 months for the first year, etc. Physical examination and biochemical tests and scans were performed weekly before each cycle
Incomplete outcome data (attrition bias)   All outcomes	Low risk	108/109 received AT treatment and 107/107 received FAC treatment. ITT analysis, and a separate per‐protocol analysis
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Unclear risk	Early study closure date: “An independent Data Monitoring Committee determined that the chance of finding a statistical difference in TTP between the treatment arms had become so low that they recommended study closure”. No differences in baseline characteristics

Methods	Accrual: 10/1999 to 11/2002  Multicentre phase III study, 29 centres in 12 countries  Centralised randomisation based on a minimising algorithm and stratified by prior adjuvant chemotherapy and centre  Baseline characteristics balanced between treatment arms except for menopausal status; FEC arm had a higher percentage of pre‐menopausal participants compared to GET arm (P = 0.024)  Median follow‐up: 24 months
Participants	Age range 29 to 74 years, median 53 years (GET) and 54 years (FEC)  100% metastatic breast cancer  100% first‐line All participants anthracycline naïve
Interventions	Arm 1:   GET: gemcitabine (1000 mg/m² day 1 and 4), epirubicin (90 mg/m² day 1), and paclitaxel (175 mg/m² day 1) Arm 2:  FEC: fluorouracil (500 mg/m² day 1), epirubicin (90 mg/m² day 1), and cyclophosphamide (500 mg/m² day 1) Both arms: q21 days for a maximum of 8 cycles No other anticancer drugs were allowed during the study including hormonal agents or immunotherapy or both
Outcomes	Outcomes were not separated into primary or secondary: Time to progressive disease, defined as from dates of randomisation until disease progression or death, whichever occurred first Time to response, defined as interval between the dates of randomisation and first documented complete response or partial response Overall survival, defined as dates of randomisation until death from any cause Toxicity Objective response rate
Notes	Data were not mature for overall survival. However, curve extraction using method 10 was undertaken to estimate TTP (Tierney 2007)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: “patients were randomly assigned to GET or FEC...based on a minimizing algorithm"
Allocation concealment (selection bias)	Low risk	Quote: “...via a centralised randomisation system based on a minimising algorithm”; stratified by centre and prior adjuvant chemotherapy
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in trial publication
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Unlikely that overall survival assessment was influenced by unblinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	Tumour imaging and response assessment carried out every cycle Blood tests weekly, so too ECG, echocardiogram (before cycles 5 & 7) and toxicity (after each cycle)
Incomplete outcome data (attrition bias)   All outcomes	Low risk	259 randomised participants; 124/124 on GET and 132/135 on FEC received study treatment. ITT was not used, instead assessable participants, for response and toxicity analyses. Survival data not yet mature
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Unclear risk	Baseline characteristics mainly balanced between groups except for menopausal status

PERMALINK

Taxane‐containing regimens for metastatic breast cancer

Davina Ghersi

Melina L Willson

Matthew Ming Ki Chan

John Simes

Emma Donoghue

Nicholas Wilcken

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Secondary outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

1. Quality of life (QoL).

Subgroup analysis and investigation of heterogeneity

2. Possible subgroups.

Sensitivity analysis

Results

Description of studies

Results of the search

1.

Included studies

Question A: regimen A plus taxane versus regimen A

Question B: regimen A plus taxane versus regimen B

Question C: single‐agent taxane versus regimen C

3. Definitions of time to progression.

4. Included RCTs, withdrawn by outcome, with reasons.

Excluded studies

Risk of bias in included studies

2.

Allocation

Blinding

Incomplete outcome data

Selective reporting

Other potential sources of bias

Effects of interventions

3.

Overall survival

Overall effect

1.1. Analysis.

4.

First‐line trials only (overall)

1.2. Analysis.

Subquestions: types of regimens

Question A: regimen A plus taxane versus regimen A

1.3. Analysis.

Question B: regimen A plus taxane versus regimen B

Question C: single‐agent taxane versus regimen C

1.4. Analysis.

Methods	Accrual dates cannot be determined  Multicentre, national, phase II  Randomisation centralised "performed by computerized log, without direct access of the investigator"  Baseline comparability: no significant imbalance reported  Median follow‐up: not reported
Participants	Female  Pre‐, peri‐, and post‐menopausal  Age range 29 to 69 years, median 52 years (arm 1) and 52.5 years (arm 2)  100% metastatic breast cancer  100% > first‐line  98% had prior anthracycline treatment
Interventions	Arm 1: paclitaxel 175 mg/m² by 3‐hour infusion q21 days  Arm 2: mitomycin 12 mg/m² in slow bolus q6 weeks  Both arms given for a minimum of 2 cycles (total cumulative dose of mitomycin limited to 60 mg/m²)
Outcomes	Outcomes were not separated into primary or secondary in trial report: Overall survival Time to progression (not defined) Response Toxicity Quality of life
Notes	Cross‐over to alternate regimen on progression. Many more cycles of paclitaxel were received than mitomycin. Follow‐up details not reported estimated minimum 1.5 months estimated maximum 23 months (OS), 12 months (PFS) Efficacy data available for 72/81 randomised participants. (4 participants did not receive allocated treatment and 5 received hormonal treatment while on study)  Estimate for time‐to‐event outcomes obtained from time‐to‐event curves
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: “Central randomization was performed by computerized log, without direct access of the investigator". Comment: random assignments were probably generated by computer
Allocation concealment (selection bias)	Low risk	Quote: “centralized randomization was performed by computerised log, without direct access of the investigator”
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in trial publication
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Unlikely that overall survival assessment was influenced by unblinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	No information provided in the trial publication about tumour evaluations Blood tests conducted prior to new course
Incomplete outcome data (attrition bias)   All outcomes	Low risk	No missing outcome data; all randomised participants included in analyses of baseline characteristics and efficacy, objective response rates also were analysed for evaluable participants
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Low risk	No baseline differences noted between treatment arms

Methods	Accrual: 2/1993 to 9/1995  Multicentre  Cross‐over (on progression)  Randomisation method not described  Baseline comparability: no significant imbalance reported  Median follow‐up: not reported
Participants	Female  Age range 25 to 79 years, median age 58 years (arm 1), 56 years (arm 2), and 56 years (arm 3)  Progressing regional disease (13% to 19% of participants) or metastatic disease  100% first‐line  All participants anthracycline naïve
Interventions	Arm 1: doxorubicin 60 mg/m²   Arm 2: paclitaxel 175 mg/m²/24 hour  Arm 3: doxorubicin 50 mg/m² + paclitaxel 150 mg/m²/24 hour + granulocyte‐colony stimulating factor q21 days. Doxorubicin for a maximum 8 cycles; paclitaxel until disease progression
Outcomes	Outcomes were not separated into primary and secondary: Overall response rate Overall survival Time to treatment failure, defined as from date randomised to date of progression, toxic death, or death attributed to breast cancer within 6 weeks of date participant last known alive with stable disease Toxicity Quality of life
Notes	Single agents cross‐over to alternate single agent on progression. Did not report as ITT 739 randomised (8 cancelled). Of 731 remaining, 33 excluded as ineligible (reasons explained). Data on 683 participants included in time‐to‐event analyses (reasons for exclusion of additional 15 participants not explained). Follow‐up details not reported, therefore estimated minimum months estimated maximum months (OS), months (PFS) The definition of TTF used in this trial was date of study entry to date progressive disease, toxic death, or death due to breast cancer within 6 weeks of date participant last known alive with stable disease. This meets the definition of PFS as used in this review.  The number of participants who received treatment cannot be determined  Estimate for time‐to‐event outcomes obtained from time‐to‐event curves ECOG E1193(A) compared arms 1 & 3
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: “patients were randomized to receive either...”. No additional details were provided on how random assignment was achieved
Allocation concealment (selection bias)	Unclear risk	Method of concealment was not described in the trial publication
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in trial publication
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Unlikely that overall survival assessment was influenced by unblinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	No information provided in the trial publication about tumour evaluations.  NCI toxicity criteria used. Comment: standard blood tests probably done
Blinding of outcome assessment (detection bias)   QoL	High risk	FACT‐B completed by participants at baseline and at week 16
Incomplete outcome data (attrition bias)   All outcomes	Low risk	33/731 randomised participants were excluded from analysis with ineligibility reasons provided. 70% of eligible participants completed the follow‐up assessment for QoL at week 16. Method of analyses not provided
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Low risk	Baseline characteristics well‐matched across the 3 arms

Methods	See E1193 (A)
Participants	See E1193 (A)
Interventions	See E1193 (A)
Outcomes	See E1193 (A)
Notes	See E1193 (A)  ECOG E1193(B) compared arms 1 & 2
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: “patients were randomized to receive either...”. No additional details were provided on how random assignment was achieved
Allocation concealment (selection bias)	Unclear risk	Method of concealment was not described in the trial publication
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in trial publication
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Unlikely that overall survival assessment was influenced by unblinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	No information provided in the trial publication about tumour evaluations. NCI toxicity criteria used. Comment: standard blood tests probably done
Blinding of outcome assessment (detection bias)   QoL	High risk	FACT‐B completed by participants at baseline and at week 16
Incomplete outcome data (attrition bias)   All outcomes	Low risk	33/731 randomised participants were excluded from analysis with ineligibility reasons provided. 70% of eligible participants completed the follow‐up assessment for QoL at week 16. Method of analyses not provided
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Low risk	Baseline characteristics well‐matched across the 3 arms

Methods	Accrual: 8/1993 to 5/1996  Multicentre, international  Cross‐over  Randomisation "performed centrally...by telephone, fax or computer"  Baseline comparability: no significant imbalance apparent or reported  Median follow‐up: not reported
Participants	Female  Age range 26 to 75 years, median age 54 years (paclitaxel group) and 55 years (doxorubicin group)  100% metastatic breast cancer in overt progression (73% had 2 or more metastatic sites)  100% first‐line  All participants anthracycline naïve
Interventions	Arm 1: paclitaxel 200 mg/m²   Arm 2: doxorubicin 75 mg/m²   Both arms q21 days for 7 cycles
Outcomes	Primary: Progression‐free survival, defined as from date randomised to date of progression or death if it occurred before documentation of progressive disease Response rate Secondary: Overall survival Toxicity Quality of life Toxic death
Notes	Cross‐over to alternate regimen on progression Follow‐up details not reported estimated minimum 5 months estimated maximum 46 months (OS), 46 months (PFS) All randomised participants included in time‐to‐event analyses  Estimate for time‐to‐event outcomes obtained using exact P value and total events
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: “patients were randomly assigned to receive...”. No details on the random component were described
Allocation concealment (selection bias)	Low risk	Central allocation. Quote: “randomization was performed centrally at the EORTC Data Center located in Brussels (Belgium) by telephone, fax or computer”
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in trial publication
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Assessment of overall survival is unlikely to be influenced by no or incomplete blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Low risk	Tumour response assessed per Union for International Cancer Control criteria. Serial evidence was documented by radiology or photography and assessed by external review. "All case report forms were reviewed by local investigators and to two independent radiologists who were blinded to treatment arm" Weekly blood tests and toxicity assessed using NCI‐CTC. MUGA or echocardiography for evaluating LVEF conducted at study entry and at completion of 5th and 7th course
Blinding of outcome assessment (detection bias)   QoL	High risk	EORTC QLQ‐C30 and Rotterdam Symptom Checklist completed by participants at baseline and at completion of cycles 3, 5, 7 and during follow‐up (every 2 months) until disease progression
Incomplete outcome data (attrition bias)   All outcomes	Low risk	164/166 randomised participants received paclitaxel; 163/165 randomised participants received doxorubicin. All randomly assigned participants were evaluated according to ITT
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Low risk	Baseline characteristics similar between the two groups; “no significant imbalance in classical prognostic factors”

Methods	Accrual: 11/1996 to 02/1999  Multicentre, international  Centrally randomised at EORTC data centre  Baseline comparability: no significant imbalance reported  Median follow‐up: 29.2 months
Participants	Pre‐ and post‐menopausal women aged 18 to 70 years  100% metastatic breast cancer  100% first‐line  All participants anthracycline and taxane naive
Interventions	Arm 1: doxorubicin 60 mg/m² + paclitaxel 175 mg/m²   Arm 2: doxorubicin 60 mg/m² + cyclophosphamide 600 mg/m²   Both arms q21 days to a maximum of 6 cycles
Outcomes	Primary: Progression‐free survival, defined as from randomisation to date of progression or death or whichever occurred first Secondary: Overall survival Response rate Toxicity Quality of life (EORTC QLQ‐C30 and QLQ‐BR23 (Breast cancer module))
Notes	Follow‐up details not reported estimated minimum 5.5 months estimated maximum 36 months (OS), 24 months (PFS) All randomised participants included in time‐to‐event analyses  Estimate for time‐to‐event outcomes obtained from reported hazard ratio and 95% confidence intervals
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: “...random assignment was undertaken using a minimization technique..."
Allocation concealment (selection bias)	Low risk	Quote: “patients were centrally randomized at the EORTC Data Center in Brussels”
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in trial publication
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Assessment of overall survival is unlikely to be influenced by no or incomplete blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	Tumour measurement performed every 6 weeks until tumour progression and assessed using WHO criteria. Comment: scans would have been done to use WHO criteria Physical exam and blood tests repeated before each cycle, hematological monitoring weekly, and MUGA scan or echocardiography at study entry, before cycles 3, 5, 6 and 3 months after last chemotherapy. NCI‐CTC used
Blinding of outcome assessment (detection bias)   QoL	High risk	EORTC QLQ‐C30 & BR23 completed by participants at baseline and before cycles 2, 4, and 6 and at first follow‐up
Incomplete outcome data (attrition bias)   All outcomes	Low risk	217/273 randomised participants received study treatment. Quote: “all randomized patients were evaluated for Response Rate, PFS and OS according to intention‐to‐treat principle”. Sensitivity analyses were performed to investigate missing data for QoL.
Selective reporting (reporting bias)	Low risk	Expected outcomes were reported in 2 separate publications
Other bias	Low risk	Participant characteristics between the 2 groups at baseline were balanced

Methods	Accrual: 06/1997 to 12/2001. Observed until the end of 2003  Multicentre study (Germany)  Randomisation carried out centrally using block sizes of various lengths  Participants stratified by age, treatment centre, disease‐free interval (> or < 18 months), hormone receptor status, prior adjuvant therapy with anthracyclines, presence of liver metastases and lung metastases  No baseline characteristics reported  Median follow‐up: 43.6 months
Participants	Participants < 80 years, no further details provided  100% metastatic breast cancer  First‐line  Anthracycline naïve
Interventions	Arm 1: mitoxantrone (12 mg/m² q21 days). Administered until complete response (plus 2 cycles) or cumulative dose 160 mg/m²   Arm 2: mitoxantrone (12 mg/m² q21 days) for up to 6 cycles plus docetaxel (80 mg/m² q21 days) until progressive disease for up to 6 cycles
Outcomes	Primary: Overall survival Time to progression (i.e. progression‐free survival, defined as the duration from randomisation until progressive disease, death, or last contact) Secondary: Best overall response Gain from treatment as measured by Modified Brunner's Score (MBS) Toxicity Quality of life
Notes	We contacted the trialists, and Martin Schumacher forwarded data from an unpublished manuscript for inclusion into this review
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "patients were randomly assigned using blocks of various lengths"
Allocation concealment (selection bias)	Low risk	Quote: "randomization was carried out centrally by fax or telephone in blocks of variable length"
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in unpublished manuscript
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Unlikely that overall survival assessment was influenced by unblinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	Assessed MBS that was composed of time to progression, PS/WHO (WHO performance status during chemotherapy as compared to that before the start of treatment), SUJ (Patient's rating of the treatment benefit) and TOX (toxicity component based on alopecia and nausea/vomiting during therapy). Comment: scans to assess tumour response probably done  Toxicity graded per WHO criteria. Comment: standard blood tests probably done
Incomplete outcome data (attrition bias)   All outcomes	Low risk	176/179 randomised participants and included in ITT analysis for primary outcomes. 3 participants excluded (1.7%) due to cerebral metastases and insufficient data (reasons provided in PRISMA flowchart)
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the unpublished manuscript
Other bias	Unclear risk	Recruitment stopped early because of poor accrual rate

Methods	Accrual: 05/2004 to 08/2008 Multicentre, phase III trial, conducted in Denmark, Sweden, and Norway Randomisation occurred centrally  Baseline characteristics were balanced  Median follow‐up: 34 months
Participants	Age range 29 to 72 years, median age 56 years (docetaxel + trastuzumab arm) and 57 years (vinorelbine + trastuzumab) 96.1% metastatic breast cancer; 3.9% locally advanced breast cancer  HER2 status: IHC 3+ = > 81%, FISH + = 35.2%, unknown 1.4% First‐line for metastatic breast cancer or locally advanced disease
Interventions	Arm 1: docetaxel 100 mg/m² IV over 60 minutes q21 days Arm 2: vinorelbine 30 mg/m² or 35 mg/m² IV as bolus injections days 1 and 8 q21 days Both arms q21 days for a median of 8 cycles (0 to 26 for docetaxel + trastuzumab) and 10.5 cycles (2 to 42 for vinorelbine + trastuzumab) Trastuzumab given before chemotherapy as IV infusion over 90 minutes with 8 mg/kg in the first cycle and subsequent cycles over 30 minutes with 6 mg/kg
Outcomes	Primary: Time to progression, defined as from date of randomisation to date of documented progression with censoring for participants alive at last visit/date of death Secondary: Overall survival (date of randomisation to date of death with censoring for participants still alive at last visit date) Time to treatment failure (time from randomisation to date of the last study chemotherapy administration, with censoring for participants still on treatment) Rate of response, assessed by investigators according the RECIST version 3.0 Toxicity (using NCI‐CTC version 3.0)
Notes	All randomised participants included in ITT analysis For OS and TTP, method 3 used to estimate O‐E and V (Tierney 2007)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "patients were randomly assigned". No further details provided in the trial report or retrieved case report form
Allocation concealment (selection bias)	Low risk	Central allocation. Quote: "...patients were, unstratified, randomly assigned centrally by the Danish Breast Cancer Cooperative Group Secretariat"
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in trial publication
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Assessment of overall survival is unlikely to be influenced by no or incomplete blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	Imaging and tumour evaluation every 3 months Lab tests (blood counts) repeated at each cycle
Incomplete outcome data (attrition bias)   All outcomes	Low risk	At time of analysis, 128/139 in the docetaxel group and 120/138 in the vinorelbine group discontinued therapy, including 11 and 15 participants, respectively, due to "other reasons" including lost to follow‐up. ITT analysis
Selective reporting (reporting bias)	Low risk	Pre‐specified outcomes in ClinicalTrials.gov record (http://clinicaltrials.gov/ct2/show/NCT00430001?term=HERNATA&rank=1) and the methods section of the trial publication are the same. All outcomes were reported in the results section
Other bias	Low risk	Quote: "baseline demographics and other variables were well balanced between the treatment groups"

Methods	Accrual: 11/1996 to 4/1998  Multicentre, international  "central randomisation"  Baseline comparability: no significant imbalance reported  Median follow‐up: 69 months
Participants	Age range 24 to 72 years, median age 50 years (in both groups)  100% metastatic breast cancer  100% first‐line  All participants anthracycline naïve
Interventions	Arm 1: doxorubicin 50 mg/m² followed 24 hour later by paclitaxel 220 mg/m² Arm 2: fluorouracil 500 mg/m² + doxorubicin 50 mg/m² IV + cyclophosphamide 500 mg/m²   Both arms q21 days for up to 8 cycles
Outcomes	Primary: Time to progression (not defined) Secondary: Overall survival Response rate Toxicity Quality of life
Notes	For OS and TTP, method 3 was used to estimate O‐E and V (Tierney 2007)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: “...patients were randomized in a 1:1 ratio to receive therapy with either...”. Involved stratified randomisation
Allocation concealment (selection bias)	Low risk	Quote: “Before central randomization, patients were stratified according to...”
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Open‐label study
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Unlikely that overall survival assessment was influenced by unblinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Low risk	All efficacy data was subjected to a blinded clinical and radiological independent review. Physical examination every cycle and imaging studies every other cycle Blood tests repeated before each cycle; LVEF assessed before cycles 5 and 7 and at end of study; toxicity assessed using WHO criteria
Blinding of outcome assessment (detection bias)   QoL	High risk	EORTC QLQ‐C30 and EORTC QLQ‐BR23 completed by participants
Incomplete outcome data (attrition bias)   All outcomes	Low risk	267 participants were enrolled and randomised, with 264 receiving treatment and 259 assessable for response. TTP and OS analyses based on ITT population. 264/267 (98.9%) of participants available for toxicity analysis
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Low risk	Quote: “patient characteristics were well‐balanced between two treatment arms”

Methods	Accrual: 01/1999 to 05/2003 Randomised, multicentre, non‐blinded phase III study that took place at 29 institutions (Japan) Participants were randomly assigned to 1 of 3 treatment groups by the minimisation method, balancing the arms according to disease status (stage IV versus recurrent disease), prior anthracyclines, liver metastasis, and institution All prognostic factors well balanced between the 3 treatment groups  Median follow‐up: not reported
Participants	Participant age range 26 to 75 years, median age 54 years (AC), 54 years (D), and 56 years (AC+D) 100% metastatic breast cancer No previous anthracyclines or taxanes were allowed
Interventions	Arm 1: AC: doxorubicin 40 mg/m² + cyclophosphamide 500 mg/m² q21 days for 6 cycles Arm 2: D: docetaxel 60 mg/m² Arm 3 (not used in review): AC+D: 3 cycles of AC and 3 cycles of D Treatment was administered q21 days
Outcomes	Primary: Time to treatment failure, from date of randomisation to the date of first documentation of discontinuation of first‐line chemotherapy, disease progression, or death from any cause Secondary: Overall survival, defined as the date of randomisation to the date of death from any cause Progression‐free survival, defined as the date of randomisation to the date of the first documentation of disease progression or death from any cause Response rate (based on the number of assessable participants) Adverse events (based on the number of assessable participants) Pilot study: quality of life
Notes	Only 2 arms of this 3‐arm trial were used. The sequential treatment comparison (i.e. AC vs AC+D) was excluded from analysis as it was not the focus of this review. This review compared AC vs D only (arm 1 vs arm 2) We contacted the trialists, and Noriyuki Katsumata provided the number of events in each group, the hazard ratio, confidence intervals, and P values for OS, TTP, and TTF. Method 3 was used to estimate O‐E and V (Tierney 2007)
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "...patients were then randomly assigned to one of the three treatment groups by the minimization method..."
Allocation concealment (selection bias)	Low risk	Quote: “The randomisation of treatment assignments was centralized. After confirming that candidate subjects met all the inclusion and exclusion criteria, the Japan Clinical Oncology Group (JCOG) Data Center was informed by telephone or fax”
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Non‐blinded study
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Unlikely that overall survival assessment was influenced by unblinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Low risk	Objective responses were assessed by central review at regular meetings. Response was classified using criteria similar to WHO criteria. Comment: scans would have been done to use WHO‐like criteria Blood tests conducted. Toxicity assessed using criteria similar to NCI‐CTC
Blinding of outcome assessment (detection bias)   QoL	High risk	FACT‐B questionnaire completed by first 50 participants in each treatment group
Incomplete outcome data (attrition bias)   All outcomes	Low risk	TTF, PFS, and OS were analysed using ITT population; response and safety used assessable participants. 68% of participants in the AC arm, 76% in the D arm, and 77% in alternating AC+D arm completed 6 cycles mainly due to disease progression
Selective reporting (reporting bias)	Low risk	Pre‐specified outcomes in ClinicalTrials.gov record (https://clinicaltrials.gov/show/NCT00193037) and the methods section of the trial publication are similar; the trial publication adds QoL. All outcomes were reported in the results section
Other bias	Low risk	All prognostic factors well balanced between the 3 treatment groups

Methods	Accrual: 02/1996 to 01/1997  Multi‐institutional cooperative group trial  Randomised phase II study  Baseline characteristics were balanced between treatment arms except that performance status was “somewhat better on the doxorubicin/paclitaxel arm” (p.146)  Median follow‐up: not reported
Participants	Age range 30 to 79 years, median age 53.7 years (doxorubicin and paclitaxel) and 55.9 years (doxorubicin and cyclophosphamide) Metastatic breast cancer or locally advanced breast cancer (number of participants with locally advanced breast cancer not provided) No prior therapy with anthracyclines, anthracenediones, paclitaxel, or docetaxel permitted
Interventions	Arm 1: AT: doxorubicin 60 mg/m² IV over 30 min followed by paclitaxel 200 mg/m² IV over 3 hour q21 days. Combined treatment was stopped after 6 cycles Arm 2: AC: doxorubicin 60 mg/m² IV over 30 min followed by cyclophosphamide 600 mg/m² IV. After 6 cycles, participants could be treated with any appropriate regimen
Outcomes	No distinction between primary and secondary outcomes provided: Complete response, defined as complete disappearance of all measurable and evaluable disease and no evidence of non‐evaluable disease Overall survival, not defined Toxicity
Notes	We contacted the trialists, and William Barlow provided the unadjusted hazard ratio, confidence interval, and P value for OS. Method 3 used to estimate O‐E and V
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: “Patients randomized between the two treatment arms in this study”. No additional details were provided on how random assignment was achieved
Allocation concealment (selection bias)	Unclear risk	Method of concealment was not described in the trial publication
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	No information in trial publication
Blinding of outcome assessment (detection bias)   Overall survival	Low risk	Assessment of overall survival is unlikely to be influenced by no or incomplete blinding
Blinding of outcome assessment (detection bias)   TTP, TTF, Response rate & Toxicity	Unclear risk	Responses assessed according to SWOG criteria, using markers and other lab values LVEF assessed on study and after 6 cycles. Comment: scans/tests probably done to grade toxicity
Incomplete outcome data (attrition bias)   All outcomes	Low risk	All participants accounted for in overall survival and objective complete response
Selective reporting (reporting bias)	Low risk	All pre‐specified outcomes in the methods were reported in the results section of the trial publication
Other bias	Unclear risk	Baseline characteristics for performance status differed between treatment groups with doxorubicin/paclitaxel arm somewhat better, but no statistics provided