In the present study we investigated the clinical outcome of patients with operable breast cancer classified to five well-established immunophenotypical breast cancer subtypes. These patients were included in two consecutively conducted adjuvant phase III trials with dose-dense sequential chemotherapy 
. The five subtypes were defined with the use of a four-marker panel (ER, PgR, HER2 and Ki67), as also reported by others 
. Notably, with the use of two additional biological markers, CK5 and EGFR we further separated triple negative tumors into two distinct subgroups, BCP and non-BCP.
The largest group of patients in our study was of the luminal B subtype (38.1%) followed by luminal A (24.8%). In contrast to the corresponding rate in large registry data sets, hospital-based or population-based case series 
, in which luminal B tumors represented only a small minority, about half of the tumors in our study were classified as luminal B/luminal-HER2. This is in keeping with the existing literature, for an “intermediate or high-risk” population entering a randomized trial with adjuvant chemotherapy in operable breast cancer 
. Conceivably, patients with more favorable prognosis, like those with luminal A tumors, are usually treated with less aggressive chemotherapeutic regimens followed by hormonal therapy or even with hormonal therapy alone.
Our study has several strengths, as well as limitations. It has been postulated that breast cancer subtypes are predictive and behave differently to specific treatments 
. If so, response evaluation of these subtypes to modern chemotherapeutic regimens, in the context of randomized trials, is of great importance. In our study, tissue blocks were obtained from patients participating in two such trials, investigating the role of epirubicin, paclitaxel and CMF-containing dose-dense sequential chemotherapy, which shared similar eligibility criteria. Furthermore, all IHC and in situ hybridization testing was carried out in a central “high-volume” laboratory, negating inter-laboratory assay result variability and, in parallel, scoring was done by experienced pathologists. Therefore, the possibility for misclassification of tumors was significantly reduced.
On the other hand, tumor tissue blocks were retrospectively collected in the HE10/97 trial and prospectively in the HE10/00 trial. Therefore, in the present analysis, patients with available blocks may not be representative of the entire patient population enrolled in the two trials, since there were differences in tumor size, number of metastatic lymph nodes and tumor grade, which may lead to biases. Nevertheless, these variables were taken into account in the multivariate analyses. Furthermore, the size of both trials is considered to be intermediate according to current standards, which may preclude the identification of subtle but still biologically meaningful differences in prognosis, especially in subtypes with relatively small number of patients, i.e. BCP or non-BCP TNBC subtypes, if examined separately.
Since a considerable number of patients in the present analysis were not treated with paclitaxel, while all received epirubicin and CMF, we sought for potentially differential responses of the different subtypes to this agent. Eventually, we were able to demonstrate a particular benefit from the use of paclitaxel only in patients with HER2-enriched tumors. In a previously published study on FFPE tissue samples from 394 patients randomized in the HE10/97 trial, an association of HER2 status and outcome was not detected, maybe due to the smaller sample size and HER2 status classification based mainly on IHC 
. However, in a large retrospective analysis of 1,322 FFPE tissue blocks from 3,121 women participating in the Cancer and Leukemia Group B 93441/INT0148 trial (four cycles of doxorubicin and cyclophosphamide followed by four cycles of paclitaxel or observation), Hayes et al 
reported that a survival benefit from the addition of paclitaxel was evident in patients with HER2-positive tumors regardless of ER status.
Regarding the behavior of TNBC patients in response to treatment with taxanes, results in the literature are equivocal at best. An analysis of the Groupo Espanol de Investigacion en Cancer de Mama (GEICAM) 9906 trial has shown a superiority of the fluorouracil, epirubicin, cyclophosphamide (FEC) combination followed by weekly paclitaxel over FEC alone, which was more prominent in patients with TNBC 
. In line with these data, Iwata et al 
reported that patients with TNBC were more likely to achieve pathological complete response (pCR) to neoadjuvant chemotherapy with four cycles of docetaxel followed by four cycles of FEC compared to those with other subtypes. Conversely, in the Breast Cancer International Research Group (BCIRG)-001 trial the combination of docetaxel (Taxotere®), doxorubicin (Adriamycin®) and cyclophosphamide (TAC) offered a significant improvement in 3-year DFS in patients with luminal B (defined as ER-positive, PgR-positive and either Ki67high
or HER-positive), but only a marginal trend in patients with TNBC or HER2-enriched tumors over treatment with fluorouracil, doxorubicin, cyclophosphamide (FAC). Intriguingly, among patients with node-positive operable breast cancer randomized in the PACS01 trial comparing six cycles of adjuvant FEC with three cycles of FEC followed by three cycles of docetaxel 
, those with luminal A did not benefit from docetaxel. Conversely, patients with luminal B, HER2-overexpressing and triple negative tumors had a significant reduction in their risk for relapse. Ki67 positivity was the most powerful predictor of benefit from docetaxel 
. We were not able to show a paclitaxel benefit in patients with TNBC. Whether this is due to differences in sample size, regimens used, patient populations or other unidentified factors cannot be revealed by the present analysis.
It has been shown that the pattern of first relapse varies among patients with different breast cancer subtypes 
. That was also the case for the randomized patients analyzed in our study, in which locoregional relapses as well distant relapses were more frequently seen in patients with TNBC and HER2-enriched tumors, a finding that has been also observed by others in hospital-based case series 
. The 6% rate of brain metastases, recorded in our patients with TNBC at the 5-year time point, was almost identical to that reported in other large studies 
. It is worth mentioning that HER2 is considered to be a robust risk factor for the development of CNS disease 
. A retrospective analysis of 9,524 patients enrolled in 10 adjuvant trials, led by the International Breast Cancer Study Group in the pre-trastuzumab era, reported that the 10-year cumulative incidence of brain as primary site of metastasis was 2.7% in patients with HER2-positive tumors and 1.0% in patients with HER2-negative tumors (p<0.01) 
. The respective rate in our patients with HER2-positive disease was slightly above 1%. Finally, the present analysis confirmed the significant liver predisposition of metastases associated with HER2-enriched tumors reported by others 
Notable differences in the behavior of breast cancer subtypes are not seen only in the pattern of metastases, but also in the time to first relapse. Data from 498 patients, enrolled in a randomized trial exploring the beneficial role of local cavity boost of RT to breast conserving therapy, indicated that median time to first event was significantly shorter for TNBC and HER2-enriched subtypes. Crude recurrence rates of luminal A tumors were less than one third of those seen in BCP tumors, regarding loco-regional relapses, and less than half regarding the development of distant metastases at 5 years 
. In the study by Blows 
, which pooled individual data from more than 10,000 patients with invasive breast cancer from 12 studies with known status of ER, HER2 and at least one basal marker (CK5/6 or EGFR), it was clearly shown that even though for HER2-positive and non-luminal subtypes mortality rates tended to peak within 5 years from diagnosis, with longer follow-up the prognosis became poorer in the luminal subtypes. Unfortunately, in that study, information on Ki67 status was not available and this might have led to misclassification of patients with luminal B as luminal A. These data fit perfectly with the cumulative hazard plots extracted from the present analysis, with patients with TNBC and HER2-positive tumors having an increased risk for relapse and death in the first 5 to 6 years from initial diagnosis. On the contrary, patients with luminal tumors had a constant risk for relapse and death across time. Further, at 5 years, the DFS and OS rates for patients with luminal A tumors were 83.5% and 92.1%, while for those with TNBC were 60.6% and 70.5%, respectively.
In summary, the present pooled analysis of individual data of more than 1,000 patients with “intermediate or high-risk” operable breast cancer, according to the immunophenotypical subtypes of their tumor, indicated that in the presence of dose-dense chemotherapy, patients with TNBC have the worse prognosis and luminal A the best. For patients with HER2-positive disease, a similar pattern emerges up to recurrence, in the adjuvant pre-trastuzumab era, while prolongation of time from recurrence to survival is noted, probably because of the use of trastuzumab. A predictive role of the subtypes for benefit from paclitaxel was not identified, except for patients with HER2-enriched tumors. Clearly, more data, similar to those reported here, from other randomized trials and eventually a meta-analysis, would be needed to delineate the predictive value of each subtype and especially of TNBC, for which a complementary targeted treatment to dose-dense sequential chemotherapy does not presently exist.