This meta-analysis combines the results from fourteen methodologically satisfactory trials that prospectively enrolled and randomly assigned 5747 patients with high-risk primary breast cancer to therapy with either HDCT or conventional chemotherapy. Our pooled results suggests that HDCT is associated with a modest advantage in DFS ((HR
0.89; 95% CI
0.79–0.99) but not in OS (HR
0.91; 95% CI
0.062). However, it was found that age and hormone receptor status had a significant interaction with prolonged OS.
Retrospective research had suggested that HDCT in high-risk primary breast cancer demonstrated significant favorable outcomes compared with historical data, but a prospective randomized clinical trial is the accepted standard for comparing different treatments such as HDCT for high-risk primary breast cancer. In fact, the results of prospectively randomized clinical trials were conflicting. Finally, other trials have shown nonsignificant trends in favor of HDCT. Only in the WSG trial, which employed tandem HDCT, did the EFS advantage translate into an OS benefit 
. An important factor that might account for the superiority of HDCT in the WSG trial is the double autologous stem-cell transplantation. High-dose therapy with tandem autologous stem-cell rescue is effective for treating high-risk neuroblastoma 
and multiple myeloma 
with encouraging long-term survival. If we excluded the WSG trial 
from our meta-analysis, the final conclusion that age and hormone receptor status had a significant interaction with prolonged OS would not have changed (data not shown).
In an effort to shed some light on the impact of HDCT as a first-line treatment for high-risk primary breast cancer, the data were pooled from available published trials for meta-analysis. However, two previous meta-analyses 
provided different evidence of the impact of HDCT on high-risk primary breast cancer outcome. The combined HRs, which are the preferred summary statistics for reporting time-to-event data 
, were not used in the previously published meta-analyses 
. The most widely recommended approach for summarizing the effect of treatment from time-to-event data in clinical trials is to use a hazard ratio. The best statistic to use is the hazard ratio (HR) in Meta-analyses of published time-to-event outcomes. HRs given in trial reports can be used directly, or if sufficient summary statistical information or Kaplan-Meier curves are presented, then HRs can be estimated indirectly 
Another meta-analysis was based on individual patient data (IPD) for the OS. For time-to-event outcomes, the gold standard approach is to obtain IPD from each included study. IPD should overcome problems of within-study selective reporting 
and should allow a more complete analysis including the potential to investigate treatment–covariate interactions 
. IPD meta-analyses are difficult to perform because of challenges in collecting the patient-level data. This analysis was displayed as an abstract, and no details were shown. We could get the data of median age, PR status and menopausal (MP) status of the whole population 
, so we used a median age of 47, a PR positive rate of 50% and a premenopausal rate of 70% as cutoff values to complete the subgroup analyses. Therapeutic strategies are generally based on the endocrine responsiveness and the estimated risk of relapse defined by tumor size, axillary lymph node involvement, histologic and nuclear grade, lymphatic and/or vascular invasion, HER2/neu-overexpression and age 
. Our pooled results suggest that HDCT is associated with a modest advantage in DFS; however, the EFS advantage did not translate into an OS benefit. When we subgrouped the studies according to age and hormone receptor status, we found a prolonged OS while performing HDCT in high-risk primary breast cancer. The analyses of the Dutch 
and WSG 
trials now suggest a predictive value for HDC benefit for HER2-negative and triple-negative (ER/PR/HER2-negative) status 
. The hypothesis-generating observations suggest that this breast cancer category presents increased sensitivity to dose intensification of alkylating agents and should remain the subject of clinical HDCT studies 
Treatment-related-mortality (TRM) may account for our finding that improved DFS did not translate into improved OS. Patients who received HDCT had a greater risk of dying during remission than patients who received non-myeloablative chemotherapy, primarily because of the toxicity associated with the regimen resulting in patients' protracted pancytopenia, which results in a prolonged risk of infection or bleeding. Although the DFS may be prolonged by HDCT, the benefit was offset in part by treatment-related deaths. Two studies reported highly significant TRM during HDCT 
, which was higher than HDCT in other studies. A considerable reduction in TRM would be needed to demonstrate a survival benefit from HDCT. Women 50 years and older appeared to have a higher risk of TRM than younger women if randomly assigned to HDCT 
. Young age will result in a reduced TRM.
Heterogeneity is a potential problem when interpreting the results of a meta-analysis. However, no evidence for statistically significant heterogeneity was found in any of the models used. This result indicates that using an overall estimation of the value of HDCT may be appropriate. To eliminate heterogeneity, we divided the 14 studies into subgroups as far as possible; subsequently, heterogeneity decreased for subgroups of age and hormone receptor status, which revealed that most of the studies could not be grouped helpfully according to age and hormone receptor status.
Quality assessment was based on the reporting of the study methods and results, namely: randomisation, allocation concealment, intention to treat, defined inclusion and exclusion criteria, extent of follow-up described clearly, balanced prognosis. Ad hoc scores may lack demonstrated validity, and the results may not be associated with quality 
. Overall, these studies included in the analysis were considered of good quality, typically prospective multicenter trials that reported outcomes analyzed as ITT; were performed at the national level; and were published in peer-reviewed journals.
From our analysis, age and hormone receptor status seem to have a significant interaction with prolonged OS. However, we must explicitly state that caution is highly advisable when interpreting subgroup analyses. These cannot be used for recommendations on treatment selection for individual patients. Nevertheless, with appropriate care, they can be used in the development of new, empirically based research.
Several limitations should be considered when interpreting the results of our analysis. First, our results were based on unadjusted estimates, while a more precise analysis could be conducted if individual data were available, which would allow for adjustment by other co-variates. Second, this analysis does not use primary patient data, but rather relies on information available in prior publications. Because of the lack of original data, one RCT was excluded because no information on HR was available 
. Third, only published studies were included in this meta-analysis. Nonsignificant or negative findings may be unpublished. Farquhar et al.
reported 6 ongoing RCT of HDCT in primary breast cancer; however, we only found two such articles. In addition, our analyses did not clarify whether double unit grafts would influence the outcome of HDCT in primary breast cancer.
In conclusion, HDCT has a benefit on DFS and OS compared to SDC in some special patients with high-risk primary breast cancer. Because of the limitations that we mentioned above, our results may not be used as a guideline for primary breast cancer treatment. Further study is needed to determine whether specific subgroups of patients, such as those who are HER2-negative or triple-negative, also benefit from HDCT. Alternatively, these questions could be addressed by combining individual patient data from the completed trials, but such an endeavor would require a large investment of resources as well as multinational cooperation.