A Prognostic Model Based on PAM50 and Clinical Variables (PAM50MET) for Metastatic Hormone Receptor–positive HER2-negative Breast Cancer

Abstract

Purpose: Predicting prognosis in HR+/HER2− metastatic breast cancer (MBC) might be clinically useful; however, no validated prognostic biomarkers exist in this setting to date. Patients and Methods: In phase III, EGF30008 trial, 484 patients with HER2− MBC who received letrozole and placebo or lapatinib were selected. PAM50 data, ECOG performance status, visceral disease, number of metastasis, biopsy type, and age were evaluated. A progression-free survival (PFS) Cox model was evaluated. The final model (PAM50MET) with a prespecified cutoff was validated in patients (n = 261) with HR+/HER2− advanced breast cancer (aBC) from BOLERO-2 (phase III trial that evaluated exemestane and placebo or everolimus). Results: In EGF30008, prognostic models with PAM50 plus clinical variables yielded higher C-index values versus models with only PAM50 or clinical variables. The PAM50MET model combined 21 variables: 2 PAM50 subtypes, basal signature, 14 genes, and 4 clinical variables. In EGF30008, the optimized cutoff was associated with PFS [HR = 0.37; 95% confidence interval (CI), 0.29–0.47; P < 0.0001] and overall survival (OS; HR = 0.37; 95% CI, 0.27–0.51; P < 0.0001). The median (months; 95% CI) PFS and OS were 22.24 (19.0–24.9) and not reached in PAM50MET-low versus 9.13 (8.15–11.0) and 33.0 (28.0–40.0) in PAM50MET-high groups, respectively. In BOLERO-2, the PAM50MET-low was associated with better PFS (HR = 0.72; 95% CI, 0.53–0.96; P = 0.028) and OS (HR = 0.51; 95% CI, 0.35–0.69; P < 0.0001). The median (months) (95% CI) PFS and OS were 6.93 (5.57–11.0) and 36.9 (33.4–NA) in PAM50MET-low versus 5.23 (4.2–6.8) and 23.5 (20.2–28.3) in PAM50MET-high groups, respectively. Conclusions: PAM50MET is prognostic in HR+/HER2− MBC, and further evaluation might help identify candidates for endocrine therapy only or novel therapies.

Gene expression profiles of breast cancer metastasis according to organ site

“…We then explored the association of 771 individual genes and 9 signatures with OS. We identified 1 signature score (i.e., luminal A signature score) and 51 genes whose high expression was significantly associated with better OS, and 2 signature scores (i.e., basal‐like signature score and PAM50MET signature score [ 27 ]) and 25 genes whose high expression was significantly associated with worse OS (Table S9 ). When adjusting for PAM50 subtype, 45 of the 771 genes (5.8%) were significantly associated with OS (Fig.…”

“…All tumors were assigned to an intrinsic molecular subtype of breast cancer (luminal A, luminal B, HER2‐enriched, basal‐like, and normal‐like) using the previously reported PAM50 subtype predictor [ 25 ]. For each sample, we calculated scores for 9 signatures including the 5 PAM50 signatures (luminal A, luminal B, HER2‐enriched, basal‐like, and normal‐like) [ 11 ], the proliferation signature [ 26 ], two risk of recurrence (ROR) signatures at 10 years: ROR score based on subtype (ROR‐S) and based on subtype and proliferation (ROR‐P) as described previously [ 26 ], and the previously reported PAM50MET signature, which is based on 17 variables [ 27 ]. Gene expression data will be deposited in the Gene Expression Omnibus under the accession number GSE175692.…”

Gene expression in metastatic breast cancer—patterns in primary tumors and metastatic tissue with prognostic potential

“…The relationship was more pronounced during the first 2 years of follow-up for both PFS (logrank test for trend p = 0.0008) and OS (logrank test for trend p = 0.0001). We also compared PAM50MET as quartile four compared to quartiles one to three, as reported in the original article ( Figures 9C,D ) ( Prat et al, 2020 ). The prognostic value during the full follow-up period was limited, but during the first 2 years of follow-up, quartile four associated to an inferior prognosis both in terms of PFS (HR 2 yrs FU = 1.70, 95% CI = 1.09–2.66, p = 0.019) and OS (HR 2 yrs FU = 2.28, 95% CI = 1.30–4.00, p = 0.0040).…”

“…The PAM50MET score was calculated as described in the original article ( Prat et al, 2020 ). The coefficients used in the calculations are found enclosed in Supplementary Table S4 .…”

“…Primarily, the PAM50MET score was calculated based on DM GEX. For patients where DM GEX data was not available, GEX data from PT was used ( n = 61), as described by the authors ( Prat et al, 2020 ). Complete data for calculation based on DM GEX was available for almost all DM samples ( n = 72).…”