Study design

We performed a retrospective analysis in localized breast cancer female patients addressed for curative primary surgical treatment at the Institute of Oncology, Bucharest from January 2005 to December 2012. Only data from patients with invasive histology, for whom a surgical specimen was evaluated by IHC and who were followed up for at least 36 months, were retained for the study. All patients provided written consent for invasive procedures as well as for the prospective collection of their data. The study was conducted in accordance with the Declaration of Helsinki and approved by the review board of Institute of Oncology “Prof Dr Alexandru Trestioreanu,” Bucharest, Romania. Patients with a history of metachronous breast cancer, at a metastatic stage of the disease, with an in situ-only histology, or for whom surgery was of a palliative intent or sequential after neoadjuvant treatment were excluded from our analysis.


The general population was classified into five intrinsic subgroups on the basis of pathological parameters and according to the 2015 St Gallen Consensus Conference, as follows: luminal A, ER and PR positive, HER2 negative, tumor grade 1–2, Ki67 ≤20%; luminal B, ER and/or PR positive, HER2 negative, tumor grade 3, Ki67 >20%; luminal B HER2 overexpressing, ER, PR, and HER2 positive; HER2 overexpressing, ER and PR negative, HER2 positive; and triple-negative, ER, PR, and HER2 negative.7

The threshold for the definition of TNBC was <1% immunopositivity of either ER or PR and an immunoscore of 0 or 1+ for HER2 expression or 2+ in the absence of amplification by fluorescent ISH. A comparative subgroup analysis was conducted between the TNBC population and all other patients in the overall population displaying hormonal and/or HER2 overexpressing features, conventionally named non-TNBC.

We considered 2005 to be the starting year of our analysis, because it was then that the TNBC phenotype was recognized as a distinct breast cancer subgroup. Regarding therapeutic considerations, systemic adjuvant treatment was proposed for all patients according to European guidelines available at the time, and the clinical stage was updated according to the American Joint Committee on Cancer, seventh edition.28 External radiotherapy was provided according to institutional guidelines. Anti-HER2 monoclonal antibody therapy with trastuzumab only became available for systemic therapy in 2007, when it was approved for general use in Romania.

We analyzed several independent variables such as demographics (age, menopausal status, family history of cancer, Charlson comorbidity index, and body mass index [BMI]), clinical tumor characteristics (laterality, involved quadrant, tumor size, clinical stage, and type of surgery), histopathological features (histopathological type, pathologic stage of regional lymphnode (pTN) stage, size of invasive component in millimeters, multifocality/multicentricity status, lymph node status, lymphovascular invasion, and in situ component), IHC characteristics (expression of ER and PR, HER2, and Ki67), and treatment-related features (chemotherapy regimen and radiotherapy provided).

Histopathological and IHC analysis and scoring

Histopathological and IHC tumor specimen assessments were carried out on formalin-fixed paraffin-embedded tissue sections selected to include representative sections of carcinomas and adjacent normal breast tissue. Tumor grading was performed, as described by Elston and Ellis, in relation to tubule formation, nuclear grade, mitotic status, and peritumoural vascular invasion according to the Rosen score.29,30 Standard IHC analysis was performed using standard procedures on 4-µm sections of paraffin-embedded tissue specimens stained with the corresponding antibodies. The tumor cell staining was compared with that of the surrounding normal breast epithelium, used as the negative control. The slides were scored according to the percentage of positive cells vs total cell number, regardless of staining intensity for non-standardized biomarkers. The immunostaining scores for ER, PR, and Ki67 and the algorithm for HER2 scoring were determined according to ASCO and CAP guidelines.10,11 The antibodies used for ER, PR, and HER2 staining, as well as those used for the other biomarkers assessed, are listed in Supplementary materials.

Cell proliferation (Ki67) was assessed by nuclear staining in at least 500 tumor cells using a mouse monoclonal antibody, clone MIB1 (Dako Denmark A/S, Glostrup, Denmark) at a 1/100 dilution. By convention, we considered the expression level of Ki67 to be low if the percentage of nuclear staining was <20%, intermediate if between 21% and 60%, and high if ≥60%.

For each of the biomarkers (cytokeratin 5/6 [CK5/6], EGFR, epithelial cadherin [E-cad], tumor protein p53, B-cell lymphoma 2 [Bcl2], topoisomerase II alpha [TOP2A], and cyclooxygenase-2 [Cox-2]), the assessment was considered to be “positive” if at least 10% of the tumor cell specimens were stained according to the internal criteria of the laboratory. The number of invasive tumor cells showing definitive immunoreactivity for the corresponding antigen was >2,000 neoplastic cells at least. The studied biomarkers were assessed regarding correlations with other clinical, pathological, and treatment-related variables as well as with a relapse/no relapse status. These correlations were further assessed to be used as a prognostic score in the specific population.

Study outcome

The primary outcome of interest was the association of biomarker expression with the estimated event-free survival (EFS) in the specific population, where EFS is the time elapsed between the date of surgery and the date of first recurrence, diagnosis of a second malignancy, or death resulting from any cause. Further, biomarkers that might predict relapse and EFS in the subpopulations were stratified according to biomolecular clustering and differences in expression according to the relapsed or non-relapsed status evaluated at the study’s end, in both the general and the specific populations.

Statistical analyses

Statistical analyses were conducted using SPSS version 22 (IBM Corporation, Armonk, NY, USA). The chi-squared test determined the correlation between biomarkers and other clinical and histological variables, and logistic regression predicted relapse based on these correlations. Fisher’s exact test was employed when there were fewer than five cases in any cell of the 2×2 table. The association between variables and survival was analyzed using Cox proportional hazard regression analysis. Survival curves were generated using the Kaplan–Meier method and compared using the log-rank test. HRs and 95% CIs for the variables were calculated using the Cox proportional hazard model.31

All statistical tests were two-sided, and a P-value <0.05 was considered to be significant. All analyses were based on the observed data with the assumption that missing data were randomly distributed.