Abstract 215P
Background
Gene expression profiling has a significant impact on understanding the biology of breast cancer (BC). Over the past 15 years, four main intrinsic molecular subtypes of BC have been identified. At the 13th Saint Gallen International BC Consensus, a surrogate classification of BC molecular subtypes by immunohistochemistry (IHC) was established. The most controversial point was the difference between the luminal A (LA) and luminal B (LB) subtypes according to the Ki67 values. Commonly, 14% is the Ki67 cut-off that has been established for differentiating BC subtypes; however, in later studies, this value has been questioned and a cut-off of 20% has been proposed. This study aimed to analyse the correlation between the surrogate BC subtypes using IHC and PAM50 gene expression assay, considering Ki67 as an independent factor and identify the best Ki67 cut-off.
Methods
We included women diagnosed between 2015 and 2020 with early stage luminal BC whose samples underwent genomic testing using PAM50/Prosigna (NanoString Technologies, Seattle, WA, USA). A total of 143 samples were analysed at a single institution. The IHC subtypes were classified using two independent Ki67 cut-offs, 14% and 20%, and these were compared to the subtypes identified by PAM50.
Results
Using the Ki67 cut-off >14%, we observed a correlation of 70.6% with a sensitivity of 79.1% and a specificity of 55.8%, as well as a positive predictive value (PPV) of 75.8% and a negative predictive value (NPV) of 60.4%. By modifying the Ki67 cut-off to be >20%, the percentage of well-classified tumours as determined by IHC was 76.2%, improving the agreement by 6.2%. The sensitivity was 93.4%, but the specificity was 46.1%. The PPV was 75.2% and the NPV was 80%, which suggests that IHC has a high probability of diagnosing LA and LB.
Conclusions
Based on the study’s results, we demonstrated that modifying the Ki67 cut-off to >20% provides a better surrogate classification by IHC and a higher sensitivity for classifying the luminal subtypes than ≥ 14%. We propose that the Ki67 cut-off should be globally modified to >20% as an independent factor; however, due to the low specificity, other factors, such as progesterone receptor expression, should be considered.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
All authors have declared no conflicts of interest.