Moore, Dan H.

10586 Background: Teschendorff et al. (Breast Cancer Res 2008) reported a 7-gene, immune response-enriched gene expression classifier using microarrays that was prognostic in women with ER-negative breast cancer. To facilitate further evaluation of the prognostic value of a classifier in archival fixed tumor tissue, we developed two multiplex TaqMan assays to quantify the expression levels of these genes, then trained and validated a 7-gene immune response score (IRS). Methods: RNA was extracted from FFPE sections of 185 ER-negative breast cancer patient samples from Mayo Clinic (n=81), Guy’s Hospital (n=50) and California Pacific Medical Center (CPMC) (n=27). Patients (N+ and N-) did not receive adjuvant therapy and were followed for a minimum of 5 years. Amplification of i) LY9, TNFRSF17, HLA-F, and IGLC2 and ii) XCL2, SPP1, C1QA and two reference genes NUP214 and PPIG were carried out in two multiplex RT-PCR TaqMan reactions and relative expression levels were determined using the DDCT method. The training set, comprised of the Mayo Clinic samples, was used to develop an IRS by using a summation of expression levels of the 7 genes. Association of the IRS with distant metastasis was assessed using a Cox proportional hazards model. The score was subsequently tested in a validation set (Guy’s + CPMC). Results: The amplification efficiency of each gene within the multiplex reactions was comparable to singleplex reactions. Duplicate amplifications showed excellent assay reproducibility and sensitivity (R2=0.98). The hazard ratio for a single standard deviation decrease in the IRS was 1.45 (95% CI 0.99-2.12, p=0.056) in the training set and 2.18 (1.32-3.61, p=0.002) in the validation set. ROC curves of continuous IRS to predict distant metastasis within 5 years had AUC of 0.74, 95% CI = 0.60-0.88. Conclusions: Robust multiplex TaqMan assays for FFPE tissue extracted RNA were developed for a 7-gene IRS. Increases in risk are associated with decreases in the IRS. Immune related gene expression clearly plays an important role in the spectrum of risk of ER-negative tumors. Additional signatures are in development that further support this conclusion.

5534 Purpose: A number of patients with advanced/recurrent cervical cancer do not respond to cisplatin-based chemotherapy. A pool analysis of three published phase III GOG studies was undertaken to identify the predictive factors and develop a model predictive of (non-) response to chemotherapy. Methods: The study population consisted of patients who received single-agent cisplatin or a cisplatin-containing combination in GOG protocols 110, 169 and 179. Prognostic variables (age, race, performance status, stage, histology, grade, disease site, prior chemotherapy—with primary radiation, time to recurrence, single-agent versus combination) were analyzed and multivariate analysis was conducted to identify factors independently predictive of response and survival. These analyses were used to establish a predictive model. Results: 816 patients were evaluable for response. In addition to single-agent treatment, multivariate analysis identified six factors (age, African-American, PS > 0, pelvic disease, prior radiosensitizer, recurrence ≤ one year) independently predictive of poor response. Those factors, but not age and African-American, were also independently associated with increased risk of death. 428 patients treated with a cisplatin- containing combination were classified into three risk groups based on the total number of risk factors (low risk: 0–1 factor; mid risk: 2–3 factors; high risk: 4–5 factors). Patients with 4–5 of risk factors were predicted to have a treatment response of only 13% (observed 10%), and median progression-free and overall survival of PFS of 2.8 months and 5.5 months, respectively. This subgroup of patients consist ∼14% of the target population in clinical practice. The predictive model was externally validated using GOG protocol 149 data that were not used for model development and further supported the predictive accuracy. Conclusions: High risk patients should be directed to non-cisplatin chemotherapy or investigational trials. No significant financial relationships to disclose.