Background: Prognostically distinct subtypes of BC defined by gene expression include luminal A, luminal B, HER2-enriched, basal-like and normal breast-like tumors. Luminal subtypes express estrogen receptor (ER)-related genes; patients with luminal B tumors have a poorer prognosis and higher proliferation compared to those with luminal A tumors. For HR+ disease, accessible and affordable surrogates for identification of luminal subtypes may aid in selecting appropriate systemic therapy and assessing eligibility to novel agent clinical trials targeted toward specific biology. Recent studies support use of Ki67 positivity by immunohistochemistry (IHC) with a cut point of 15%, along with standard receptors, as an appropriate surrogate for luminal B. 

Methods: 116 I-SPY 1 TRIAL patients with intrinsic subtype assignments had Ki-67 IHC-stained pre-treatment whole tissue sections available for digital image quantification. The percentage of Ki67-positive nuclei was quantified using the Aperio Nuclear V9 (cell quantification) algorithm. Intrinsic subtype was previously determined by standard methods applied to Agilent 44K gene expression data. We selected the 49 patients with HR+/HER2- tumors for this analysis. Ki67 positivity was categorized into two groups: low (<15%) and high (≥15%). Association between intrinsic subtype and Ki67 was determined using Fisher's exact method. 

Results: The 49 pts with HR+/HER2- disease were classified by gene expression into intrinsic subtypes as follows: 26 (53%) luminal A, 15 (31%) luminal B, 2 (4%) HER2-enriched, 6 (12%) basal, and 0 normal. The fraction of Ki67 positive cells ranged from 0.28% to 86.8%, with a mean of 20.5%. The mean Ki67 positive fraction was 13% and 24% in LumA and LumB subgroups, respectively. Using the 15% cutoff, Ki67 was low in 25 (51%) and high in 24 (49%) cases. The majority of luminal A tumors have low Ki67 (17/26, 65%) and luminal A subtype was associated with Ki67 (p = 0.047). However, nearly a third of tumors with low Ki67 are not luminal A (8/25, 32%). High Ki67 tumors were distributed among luminal A (18%), luminal B (18%), basal (10.2%), and HER2-enriched (2%) subtypes; luminal B subtype was not associated with high Ki67 (p = 0.36) in this cohort. 

Conclusion: In this high-risk population, 16% of HR+/HER2- tumors were either basal or HER2-enriched by intrinsic subtyping. Low Ki67 (<15%) was associated with luminal A; this did not exclude other subtypes. In this well characterized group, high Ki67 (≥15%) in combination with ER status did not serve as a surrogate for luminal B subtype. These results suggest that intrinsic subtype classifications reflect information beyond that captured by hormone receptor status combined with Ki67 positivity. This research was partially supported by NIH CA31946 and CA33601.