Correlative Analysis of Immunohistochemical and OncotypeDX® Derived PCR Data in Breast Cancer
Zulfia McCroskey, Girish Venkataraman, Patricia Robinson, Shelly Lo, Anna Szpaderska, Xiuzhen Duan, Cagatay Ersahin. Loyola University Medical Center, Maywood, IL
Background: OncotypeDX® assay analyzes the expression of several metastasis associated genes and estimates a distant disease Recurrence Score (RS). Testing is typically performed in estrogen receptor positive (ER+) breast cancer (BCA). We sought to correlate the concordance of immunohistochemical (IHC) stains with the OncotypeDX® PCR derived data on these markers.
Design: ER+ BCA cases (n=171) with available OncotypeDX® PCR data on ER, PR, and HER2 expression with concurrent available ER, PR Allred scores, HER2 and Ki67 IHC results were included. Ki67 and HER2 were modeled as ordinal covariates (0-10%, score 0; 10-20% score 1; and >20% score 2 for Ki67). Initial univariate correlations between RS, IHC data (ER, PR and Ki67) and PCR data were assessed. Univariable (u-) and multivariable (mv-) logistic regression (LR) with nested models were constructed and Receiver Operating Characteristic curves (ROC) were examined. Model comparison was performed using likelihood ratio test.
Results: ER, PR and HER2 IHC were strongly correlated with their PCR expression status (Pearson r=0,53, 0.72, 0.42 resply p<0.05) as expected. Concordance for ER and PR were 99% and 89% respectively between immunohistochemistry and qRT-PCR. Most discordant cases had low Allred scores. There was 100% concordance for HER2; however, all the cases were negative. In u-LR analyses, PR, HER2 and Ki67 explained 5.5, 1, 20.7% of variance in RS, while via PCR PR and HER2 explained 19.8 and 5% of the variance in RS. Interestingly, by IHC Ki67 explained most of the variance, whereas PR did not add any further information. Notably, PR explained most of the variance with small contributions by HER2 RT-PCR. We constructed a 2-parameter model using Ki67 and PR-PCR which explained nearly 34% of the variance in the RS (p<0.001)
Conclusions: There is some complementary information provided by combination of IHC of which Ki67 explains significant proportion of variance in RS. Nearly 65% of variance remains unexplained and it is most likely contributed by the other genes tested as part of the OncotypeDX® panel. However, discordant ER and PR results between validated immunohistochemistry results and qRT-PCR raise the doubt whether qRT-PCR correctly quantitates the gene expression levels in some cases.
Monday, March 4, 2013 9:30 AM
Poster Session I Stowell-Orbison/Surgical Pathology/Autopsy Awards Poster Session # 31, Monday Morning