Background:We recently reported that the GHI Recurrence score (GHI-RS) provided additional information regarding distant recurrence beyond that obtained from classical clinical factors (age, nodal status, tumour size, grade, randomised treatment) in 1231 patients with ER+ primary breast cancer in the ATAC trial who did not receive chemotherapy and were randomized between anastrozole vs. tamoxifen (Dowsett et al, SABCS 2008). Here we develop a score based on 4 standard laboratory assays (IHC4 score) and compare its prognostic value to that of the GHI-RS.Methods:Quantitative IHC scores were obtained for ER, PgR, and Ki67, and HER2 was scored as positive/negative in 1125 of those women from TransATAC for whom the GHI-RS and sections were available; 793 of those women were node negative. An IHC4 score was obtained by a proportional hazards regression using classical variables and the 4 IHC values. This was done separately for all patients and node negative patients for recurrence (TTR) and distant recurrence (TTDR), but the scores were very similar (pairwise correlation ρ > 0.93). The data set was then split at random into two equal parts. Models were created on each half with the classical and IHC4 variables to create an IHC4 score, which was then applied to the other half (validation) and compared with the GHI-RS. Results were summarized by likelihood ratio (LR) χ2 test which combined results from the two halves. This was repeated for 100 random splits of the data.Results: Each of the 4 IHC variables added significantly to a model containing the classical variables and the overall model was highly significant (χ2(4df)=34.9, P < 0.0001, for TTDR on all pts). Of the 4 variables PgR was the most predictive for all patients. When restricted to node negative patients, ER and HER2 were jointly the strongest predictors.The IHC4 score was modestly correlated with GHI-RS (ρ=0.70), and in sample splitting evaluations provided a similar amount of information as the GHI-RS score (GHI-RS slightly more for TTR and IHC4 slightly more for TTDR, see Table).

Mean change in LR χ2 for addition of IHC4 or GHI-RS or both to classic model (C) in the validation halves of 100 random splits of the data. Higher values indicate more information.

  TTR TTDR 
Model All Patients Node Neg. All Patients Node Neg. 
C+IHC4 vs C (1df) 19.7 21.2 26.9 26.6 
C+GHI-RS vs C (1df) 26.7 25.6 25.6 19.4 
C+IHC4+GHI-RS vs C (2df) 30.0 29.9 32.5 29.7 
  TTR TTDR 
Model All Patients Node Neg. All Patients Node Neg. 
C+IHC4 vs C (1df) 19.7 21.2 26.9 26.6 
C+GHI-RS vs C (1df) 26.7 25.6 25.6 19.4 
C+IHC4+GHI-RS vs C (2df) 30.0 29.9 32.5 29.7 

The predicted 9y distant recurrence proportion for an No, T<2cm, poorly differentiated tumour receiving anastrozole at either the 25th or 75th percentile of each score was 6.7% vs 12.3% for IHC4 and 7.8% vs 11.2% for GHI-RS.

View Large
None of these variables provided predictive information regarding the choice between tamoxifen or anastrozole.Conclusion:These data suggest that 4 standard IHC assays performed in a high quality laboratory can provide similar amounts of prognostic information for endocrine treated ER+ breast cancer patients as the GHI-RS. External validation in another data set is needed to confirm these results.

Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 74.

2009