In this study, we develop an easily applied immunohistochemical surrogate for gene expression profile–defined luminal subtypes of breast cancer. We demonstrated that biological subtyping by use of this immunohistochemical surrogate panel of four biomarkers (ie, ER, PR, HER2, and Ki67) had statistically significant value that was independent of standard clinicopathological parameters (including age at diagnosis, tumor size, grade, lymphovascular invasion, and axillary lymph node status) in identifying high-risk women with hormone receptor–positive breast cancer in the settings of no adjuvant systemic therapy, adjuvant tamoxifen, and combined adjuvant tamoxifen and chemotherapy.
Gene expression profiling studies have consistently revealed biologically distinct breast cancer subtypes with different prognoses (39
). Luminal B breast cancers are a clinically important subgroup associated with poor outcome in both the presence and the absence of systemic adjuvant therapy. By use of two independent cohorts of invasive breast carcinomas, our study is, to our knowledge, the first to develop a four-marker surrogate immunohistochemistry panel, including ER, PR, HER2, and Ki67, to distinguish the luminal B subtype from the luminal A subtype. We developed the immunopanel against a gold standard definition for tumor subtype that used results from gene expression profiling and demonstrated clinically significant associations with breast cancer relapse and survival.
The luminal B subtype is characterized by having increased expression of HER2-associated genes (ie, ERBB2
) and a cell proliferation signature that includes the expression of MKI67
, and MYBL2
, which have been associated with tamoxifen resistance (16
). Efficient clinical identification of luminal B breast cancers would isolate a poor prognosis subgroup that could likely benefit from additional systemic therapy from among otherwise good prognosis, hormone receptor–positive tumors. As suggested from gene expression profiling, coexpression of HER2 and ER and/or PR can identify some luminal B tumors (ie, the luminal–HER2-positive group). However, only approximately 30% of luminal B tumors are HER2 positive, indicating that this clinical marker alone is not sensitive enough to identify most luminal B breast cancers. In this study, we categorized such tumors as luminal–HER2 positive because they require a distinct treatment approach involving HER2-targeted therapy (eg, trastuzumab). However, from a biological perspective (ie, in terms of their gene expression profile), these tumors belong to the luminal B subtype.
Ki67 is a well-established cell proliferation marker in cancer and an excellent candidate biomarker for luminal B tumors. Two recent meta-analyses have reported a statistically significant association between high Ki67 expression and increased risk of breast cancer relapse and death (12
). However, assessment of Ki67 has been a matter of controversy because some studies have used 10% (42
) or 20% (44
) cut points, whereas others dichotomized around the mean (46
) or median (30
) value. Our study is the first to apply quantitative Ki67 visual immunohistochemistry scores to breast cancer biological subtypes that were classified by gene expression profiling. An advantage of this approach is that the optimal threshold of Ki67 immunohistochemistry (in this case 14%) was determined against an important distinction in the underlying biology of breast cancer rather than against clinical outcome or the mean or median value of the Ki67 index in the study population. By this approach, the cut point will more likely be directly applicable in other cohorts of patients with different treatment regimens and risk distributions. Although gene expression profiling remains the most sensitive method, we have demonstrated that Ki67 can be added concurrently to the standard biomarker panel of ER, PR, and HER2 to identify additional luminal B tumors that would not be identified by these three markers. The addition of epidermal growth factor receptor and cytokeratin 5/6 to this panel allows identification of the basal-like subtype of breast cancer (20
We evaluated the prognostic value of our luminal B immunohistochemistry panel using an independent, regional population-based cohort of 4046 patients who were originally diagnosed with breast cancer between January 1, 1986, and September 30, 1992. These patients received adjuvant therapy according to guidelines developed and disseminated by the British Columbia Cancer Agency (24
). In general, adjuvant systemic treatment was less aggressive than in contemporary practice, with consequent higher event rates. We demonstrated the prognostic value of our luminal B definition within homogeneously treated patient subsets. Among patients with hormone receptor–positive tumors who received no adjuvant systemic treatment, luminal B and luminal–HER2-positive tumors were associated with increased risk of breast cancer relapse and death. In contemporary practice, almost all patients with hormone receptor–positive breast cancer are treated with hormonal therapy (tamoxifen or aromatase inhibitors), and in this study, luminal B and luminal–HER2-positive tumors were associated with increased risk of breast cancer relapse and death in the subgroup receiving adjuvant tamoxifen, in comparison with the more common luminal A subtype.
In multivariable analysis, the luminal B and luminal–HER2-positive subtypes provided statistically significant prognostic value beyond current standard clinicopathological parameters. The Cox regression models included tumor size, age at diagnosis, grade, lymph node involvement, and lymphovascular invasion, which include the compulsory variables for calculation of the Nottingham Prognostic Index (49
) and Adjuvant! Online, a computer software program that predicts breast cancer outcomes by use of SEER data and clinical trial meta-analyses to guide treatment decisions in clinical practice (50
). Indeed, almost half of our patient cohort was included in an earlier study confirming that in the British Columbia population, Adjuvant! predictions are associated with observed outcomes (51
), providing support that the conclusions in this study can be extended to other North American and UK populations. Luminal B status as defined by Ki67 labeling retained independent prognostic value in patients with lymph node–negative or lymph node–positive, hormone receptor–positive breast cancer who were treated with adjuvant tamoxifen. The recurrence score, a qRT-PCR–based measure of risk of breast cancer recurrence, is currently an available diagnostic test to predict distant recurrence for ER-positive breast cancers in patients with negative axillary lymph nodes who are treated with adjuvant tamoxifen (19
). This score uses the expression of 16 genes, weighted heavily on MKI67
and other proliferation-associated genes, to calculate a risk score. This recurrence score assay has not been applied to the BCCA series of tumors, which limits our capacity to do a head-to-head comparison between our immunopanel and this qRT-PCR assay (the cost of which is approximately 10-fold higher per tumor). However, Fan et al. (39
) have shown that breast cancer subtype as determined by gene expression profiling and the recurrence score have statistically significant agreement in outcome predictions. This result indicates that among patients with lymph node–negative, ER-positive disease who were treated with adjuvant tamoxifen, high recurrence scores appear to track largely with luminal B cancers.
The association between the luminal B subtype and response to adjuvant systemic chemotherapy has yet to be fully elucidated. A meta-analysis of 1521 patients with endocrine-responsive tumors enrolled in two randomized trials of adjuvant chemoendocrine therapy reported that Ki67 expression as a single marker was not associated with resistance or benefit from chemotherapy, beyond the benefit incurred with hormonal therapy alone (30
). The chemotherapy regimen used in these two trials was cyclophosphamide, methotrexate, and fluorouracil, and the median value—19%—was the cut point for the Ki67 index. In contrast, we assessed the Ki67 index only in the context of hormone receptor–positive, HER2-negative tumors and used a cut point of 14%. We found that the Ki67 index and HER2 expression could be used to stratify the risk for breast cancer relapse and death among patients with hormone receptor–positive breast cancer who were treated with both tamoxifen and chemotherapy as their adjuvant systemic therapy.
This study has several limitations. The main weaknesses of immunohistochemical approaches are limited technical reproducibility, subjective interpretation, and qualitative readouts (52
). To facilitate analysis of sufficiently large cohorts of samples, the immunostaining panel was trained and validated on tissue microarrays, whereas clinical implementation would likely occur on whole sections. It is possible that Ki67 index may demonstrate focally higher areas on whole sections that can be appreciated on tissue microarrays, although tissue microarray results have repeatedly been demonstrated to show excellent agreement with whole sections (54
), including in studies of breast cancer (55
) and of Ki67 (56
). The distinction between luminal A and luminal B tumors is somewhat difficult to achieve, and our panel has false-positive and false-negative rates of approximately 25% (). The need for long-term follow-up data necessitated the use of a historical cohort of old paraffin blocks, which could potentially differ from recently fixed and processed prospective specimens. Treatment recommendations at the time patients in this study were treated tended to be less aggressive than in contemporary practice, and treatment cohorts are not randomized in the population-based validation cohort.
Strengths of the study design include the use of large independent cohorts to separate development of the immunopanel (against a gold standard of gene expression profiling) from its application to a large patient series for which information on clinical outcomes was available. Our subtype definition was linked to breast cancer biology through a gene expression profiling approach in which tumor specimens were assigned to intrinsic breast cancer biological subtypes, which were previously shown to be reproducible across patient populations and gene expression platforms (25
), as opposed to being linked to an expression profile classifier optimized against patient outcome (which can be problematic to extrapolate to other cohorts of patients who in general will have differences in risk profile and specific treatment). The four-biomarker immunopanel is economical, antibodies against the biomarkers are readily available, and equivalent tissue microarray approaches can be used with existing legacy specimens from clinical trials to examine predictive values. Immunohistochemistry retains the advantage of assessing protein expression in the context of tumor morphology, can be applied to tiny core needle biopsy samples in clinical and research laboratories, and has a rapid turnaround time. Although we consider breast cancer molecular subtyping by gene expression profiling to be the gold standard, we nevertheless believe that there is an immediate need for a well-defined and validated immunopanels for worldwide clinical diagnostic use.