This review concluded that there were many options for first and subsequent lines of therapy, including chemotherapy and targeted therapies, that have comparable efficacy in advanced breast cancer. Overall, the authors' conclusions appear valid, but the lack of any assessment of trial quality and the use of both direct and indirect evidence, suggests that the findings should be interpreted cautiously.

To assess different chemotherapy regimes and targeted therapies for the treatment of advanced breast cancer and determine their relative effects on survival.

PubMed and the Cochrane Central Registry of Controlled Trials were searched up to October 2007. Search terms were reported. Reference lists of primary studies and retrieved reviews were screened for further studies. Three major journals (2004-2007) in the topic area were handsearched, as were the conference abstracts and presentations of major meetings (2006-2007) of the American Society of Clinical Oncology, the European Society of Clinical Oncology and the European Cancer Conference. Searches were not restricted by the language or publication status.

Randomised controlled trials (RCTs) comparing the effects on survival of at least two different regimes of chemotherapy and/or targeted therapies for the treatment of advanced breast cancer, were eligible for inclusion in the review. Advanced breast cancer was defined as metastatic or recurrent disease, which was not amenable to surgery. Trials including any concomitant therapies such as surgery, radiotherapy, hormonal treatments or radioisotopic treatments had to apply the therapies similarly to all treatment arms.

Trials included in the review compared various drug regimens from the following groups of therapies: anthracyclines, anthracenediones, non-taxane novel chemotherapies (capecitabine, gemcitabine or vinorelbine), taxanes, thalidomide, marimastat, trastuzumab, lapatinib, bevacizumab and older miscellaneous treatments. Two thirds of the trials assessed therapies in first-line treatment and the remainder in second-line and subsequent lines of treatment. The majority of trials compared standard or low-dose anthracyclines with either combinations of older drug therapies or standard-dose anthracyclines. Trials were mainly performed in the USA or Europe and compared two different treatment arms. Further details of the included trials are available online. Detailed trial design information is available on request from the authors.

The authors did not state how papers were selected for review or how many reviewers performed the selection.

The authors did not state that they assessed validity.

Median overall survival was extracted for each trial arm and any significant differences recorded. Unadjusted (or adjusted multivariate) hazard ratios with 95% confidence intervals from Cox regressions were extracted (where available) and used to calculate the logarithm hazard ratio and its variance. Authors were contacted for missing data, or the data were estimated from the number of events and patients per treatment arm, with the log hazard ratio being estimated so that it had the p-value of the log rank test. Alternatively, where p-values were unavailable hazard ratios were estimated from the ratio of median survivals.

Two authors independently extracted data from the included studies and discrepancies were discussed with a third reviewer and cross-checked with the original articles.

Pooled hazard ratios with 95% confidence intervals were calculated for head-to-head comparisons of the different therapies using a random-effects model. Statistical heterogeneity was calculated using the I² statistic; values greater than 50% were considered to be high, 25-50% modest and below 25% low. The studies were also pooled using a multiple-treatment network meta-analysis, using both head-to-head and indirect treatment comparisons. The primary assumption of coherence required for the network meta-analysis was tested by assessing the level of incoherence between the network's indirect and direct data. Further analyses were conducted to examine the effect of single versus combination treatments, standard versus low-dose anthracycline treatments, year of publication and line of therapy.

Three hundred and seventy randomised controlled trials (RCTs) were included in the review (n=54,189 patients); 172 trials (n=31,552 patients) compared different treatments, of which 128 trials (n=26,031 patients;148 comparisons in total) reported survival data and were included in the analyses.

Direct comparisons of therapies: Statistically significant differences were reported for nine direct head-to-head comparisons of different therapy categories. Standard doses of combinations of anthracycline therapies were favoured over standard dose single anthracyclines and combination therapies with older agents. Combinations of older therapies were favoured over low dose single agent anthracycline therapy. Standard doses of single agent anthracyclines were favoured over older single agent therapies. Novel single agent non-taxane therapies were favoured over combination and single agent regimes of older therapies. Novel non-taxane therapies combined with taxane were favoured over taxanes alone. Combinations of older therapies were favoured over older single agent therapies. Taxanes combined with trastuzumab were favoured over single agent taxanes.

Survival: Evidence from the network meta-analysis suggested that the largest reduction in mortality was associated with taxanes in combination with novel agents trastuzumab or older agents. Taxanes in combination with lapatinib, anthracyclines or both anthracyclines and novel non-taxane therapies, were more effective at prolonging survival than single therapy with an older agent. Combination trastuzumab and standard dose anthracyclines offered similar benefits in survival. Single agent taxanes and standard dose anthracycline combinations reduced mortality risk by approximately one third. Older combination therapies without anthracyclines or taxanes appeared to produce a 25% reduction in mortality risk.

Further analyses of the effects of year of publication suggested that the introduction of newer therapies over the last 20 years have produced little benefit in terms of survival. The line of therapy also appeared to have little effect.

Details of the effect sizes are reported in the review.

The evidence suggested that there are many options for first and subsequent lines of therapy that have comparable efficacy for the treatment of advanced breast cancer, including anthracyclines, taxanes, novel non-taxane therapies and molecular targeted therapies.

This review answered a clear question using a broad range of eligible interventions. A number of databases and other sources were searched for trials and no restrictions were placed on language or publication status, suggesting that the risk of language and publication bias was low. The author did not state how trials were selected for inclusion in the review, but data were extracted and checked by at least two reviewers, suggesting that some attempts were made during the review process to reduce reviewer error and bias. It was difficult to assess the reliability of the trial data as the authors did not appear to have considered the quality of the included trials. Individual trial data were also lacking in the review, but this was understandable given the large number of trials included. The authors gave further trial details in online appendices. Given the lack of direct head-to-head comparisons between some of the interventions of interest, the authors also analysed the data using a network meta-analysis, which used both direct and indirect data. For this type of analysis to be valid, the study characteristics must be homogeneous between trials and there must be coherence between the direct and indirect data. Both of these factors were adequately addressed by the authors, who also advised a cautious interpretation of their findings from this analysis, given that indirect data are only suggestive of relative effectiveness and still liable to bias. The use of hazard ratios to summarise the outcome effect sizes was also appropriate, given that the primary outcome was survival which used censored time to event data. Overall, the authors' conclusions appear valid, but the lack of any assessment of trial quality, and the use of both direct and indirect evidence, suggests that the findings should be interpreted cautiously.

Practice: The authors stated that treatments should be tailored to each individual patient taking into account both the relative benefits and the potential adverse effects of treatments.

Research: The authors stated that further trials are required to investigate direct head-to-head comparisons of treatments for which there is currently little or no data, including newer treatments such as vinorelbine, gemcitabine, capecitabine, taxanes and molecular targeted therapies.

Not stated.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.