The OncoMasTR Test Predicts Distant Recurrence in Estrogen Receptor–Positive, HER2-Negative Early-Stage Breast Cancer: A Validation Study in ABCSG Trial 8

Several prognostic gene signatures have been developed to estimate the risk of breast cancer recurrence after surgery and to guide adjuvant treatment decisions. The MammaPrint Score, the Oncotype DX Recurrence Score, the EndoPredict Score, the Prosigna PAM50 Risk of Recurrence Score, and the Breast Cancer Index have previously been validated as prognostic tests (1–7).

OncoMasTR is a novel prognostic test with the aim of accurately estimating the risk of recurrence for patients with estrogen receptor (ER)–positive, human epidermal growth factor receptor 2 (HER2)–negative, early-stage breast cancer. The signature was developed using a novel bioinformatic approach that identified a shared network of master transcriptional regulators (MTR) from which the combination of only 3 genes, namely Forkhead Box M1 (FOXM1), Pituitary Tumor Transforming Gene 1 (PTTG1), and Zinc Finger Protein 367 (ZNF367), formed the OncoMasTR Molecular Score (8, 9). The final OncoMasTR Risk Score combines the OncoMasTR Molecular Score with nodal status and tumor size. The OncoMasTR test has been analytically validated in accordance with industry standard guidelines from the Clinical Laboratory Standards Institute (CLSI; ref. 10).

The clinical performance of the OncoMasTR test was initially validated in 646 patients with ER-positive, HER2-negative, early-stage breast cancer with 0 to 3 involved lymph nodes enrolled into the ATAC trial (11). These results suggested that the OncoMasTR test was highly prognostic for the overall population and for node-negative disease as well as for both early (0–5 years) and late (5–10 years) distant recurrence (DR). In addition, the OncoMasTR Risk Score added significant prognostic value to Ki67, the Nottingham Prognostic Index (NPI), and the Clinical Treatment Score (CTS; ref. 11).

In the present study, we independently validated the clinical performance of the OncoMasTR Risk Score in a large cohort of patients with early-stage breast cancer who were enrolled into the Austrian Breast & Colorectal Cancer Study Group (ABCSG) Trial 8, a large, prospective, randomized clinical trial with long-term follow-up comparing the efficacy of two adjuvant endocrine treatment regimens (12, 13).

The objective of ABCSG Trial 8 was to compare the efficacy of the sequence of adjuvant tamoxifen (TAM) for 2 years followed by anastrozole (ANA) for 3 years with tamoxifen alone for 5 years. From January 1996 to June 2004, a total of 3,901 postmenopausal women with hormone receptor–positive, early-stage, grade 1 (G1) or grade 2 (G2) breast cancer who had undergone primary surgery with or without radiotherapy were randomized (12, 13). Patients in the trial did not receive adjuvant chemotherapy and/or anti-HER2 treatment (patients with high-grade tumors were excluded). The results from this study indicate that postmenopausal women already receiving adjuvant tamoxifen for the treatment of hormone receptor–positive early breast cancer benefit from switching to adjuvant anastrozole, rather than continuing with tamoxifen (12, 13). The study was conducted in accordance with the Declaration of Helsinki, approved by the responsible ethics committees, and all patients gave written informed consent.

All patients (n = 3,901) included in ABCSG Trial 8 were eligible for the translational study and participating centers were requested to provide a tumor block of their patients. Tumor specimens were obtained at the time of surgery prior to the adjuvant therapy. Details with regard to the collection of the samples and the preparation of sections were previously published as part of translational studies in this population (5, 7, 14–18). Formalin-fixed, paraffin-embedded (FFPE) surgical breast cancer specimens from 1,200 patients were retrospectively collected and used for the OncoMasTR test. A flowchart illustrating patient and sample selection is shown in Fig. 1. The collection of test data was performed in the ABCSG central research lab at the Medical University of Vienna, blinded to patient clinical and outcome parameters.

One hematoxylin and eosin (H&E)–stained slide was prepared from each paraffin block and reviewed by an experienced breast pathologist (M. Rudas) to confirm the presence of invasive breast carcinoma. If more than 30% invasive tumor was present, macro-dissection was performed and tumor areas were scraped off the slide. In addition, in cases with extensive intraductal component, ductal carcinoma in situ (DCIS) was removed before performing the assay. Five adjacent sections of 5 μmol/L thickness from the FFPE block, or necessary tissue quantity (if the tumor-cell content was less than the prespecified 30%), was extracted using the Qiagen RNeasy FFPE Kit (Qiagen), followed by an optimized DNase digestion step to remove the genomic DNA background. The extracted RNA was read on NanoDrop for its concentration and purity.

The extracted and quality-controlled RNA was mixed with 1-Step TaqPath Master Mix (OncoMark) and RT-PCR grade water and applied to the OncoMasTR test plate, which is a 96-well plate prespotted with dried oligonucleotides for analyzing the expression levels of individual MTRs and reference genes. The RNA was analyzed in triplicates per gene target. There was an in-line positive control and negative control for each individual run. The OncoMasTR Test was performed on an Applied Biosystems 7500 Fast Dx Real-Time PCR instrument (Applied Biosystems). The OncoMasTR test reagents (including OncoMasTR plate, Reagent Kit, Instructions for Use), Qiagen RNeasy FFPE Kit, and the Applied Biosystems 7500 Fast DX Real-Time PCR instrument were provided by OncoMark for the duration of the analyses. The relative normalized expressions of the prognostic genes were translated into the OncoMasTR Risk Score using a clinically validated algorithm incorporating tumor size and lymph node status (8, 11). The OncoMasTR Molecular Score ranges from 0 to 100 and the OncoMasTR Risk Score ranges from 0 to 10, where a score of <5 corresponds to low risk and a score of ≥5 corresponds to high risk of DR within 10 years of initial diagnosis.

All data analyses were carried out according to a predefined statistical analysis plan. The prospectively defined primary endpoint of the study was time to DR defined as time from randomization until the occurrence of distant metastasis (ignoring loco-regional recurrences). All secondary carcinoma including contralateral breast cancer, and all deaths were treated as censoring events. In case of no event, the outcome was censored at the date of last contact. Follow-up was censored at 10 years.

Time to DR was analyzed by univariable and multivariable Cox proportional hazards regression models and described by HRs with 95% confidence intervals (CI), likelihood ratio tests, and by Kaplan–Meier survival curves.

Our primary objectives were to evaluate whether the OncoMasTR Molecular Score as a continuous variable, the OncoMasTR Risk Score as a continuous variable, and the OncoMasTR Risk Category as a categorical variable are prognostic for time to DR in all patients (0–3 involved nodes). The prognostic value of the OncoMasTR Scores (molecular score, risk score, and risk category) was also assessed in patients who were node negative and patients with 1 to 3 involved nodes. Likelihood ratio (LR) tests from uni- and multivariable Cox models were used to evaluate whether additional prognostic information was provided when the OncoMasTR Scores was added to age (as continuous variable), tumor size (as continuous variable), nodal status (pN1 vs. pN0), grading (G2 vs. G1), ER expression (as continuous variable), progesterone receptor (PR) expression (as continuous variable), Ki67 expression (as continuous variable), treatment arm (TAM/ANA vs. TAM), the NPI, or the CTS (19–22). The proportional hazards assumption was checked with weighted Schoenfeld residuals. The continuous relationship between DR risk and numerical risk scores was estimated using Royston/Parmar models (23).

Categorical baseline data according to OncoMasTR Risk Category were compared in univariable analysis using the χ² or Fisher exact test depending on the expected cell frequencies. Continuous baseline data were compared using Wilcoxon test. All reported P values are from two-sided tests. All results with P values <0.05 were considered statistically significant. Statistical analyses were performed by members of the biostatistics group at ABCSG using statistical analysis system (SAS) software (SAS version 9.4). This study meets the Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK; ref. 24).

Patient and sample selection is shown in a consort diagram (Fig. 1). Main clinical and laboratory parameters of all patients included in the present study (also compared with trial patients without tumor blocks) are summarized in Supplementary Table S1.

Mean patient age was 63.3 years. All patients had ER-positive and HER2-negative tumors, 69.5% had tumors smaller than 2 cm, 73.2% had node-negative disease, and 77.6% had G2 grading. Breast conservation was achieved in 82.1%. Data from 1,087 of the 1,200 samples (90.6%) met the prespecified OncoMasTR data quality criteria and were included in the analysis. Within the 1,087 patients, 850 (78.2%) were categorized as low risk and 237 (21.8%) as high risk (Table 1). High-risk patients were significantly older, had larger tumors, more frequently node-positive disease, more tumors with G2 grading (as compared with low grade), more progesterone receptor–negative tumors, and a higher mastectomy rate (Table 1).

Within the 10-year follow-up period, 86 of 1,087 (7.9%) patients of the study population had developed distant metastases, 48 of 796 (6.0%) node-negative patients, and 38 of 291 (13.1%) patients with 1 to 3 positive nodes.

The estimated probability of 10-year DR as continuous functions of the OncoMasTR Molecular Score (A) and the OncoMasTR Risk Score (B) are shown in Fig. 2.

At 10 years, the DR rate for all patients in the low-risk group was 5.8% versus 21.1% for patients in the high-risk group (HR 4.33, 95% CI, 2.83–6.61, P < 0.0001; absolute risk reduction 15.3%; Table 2; Fig. 3A). The absolute risk reduction at 10 years was 12.9% in patients that were lymph node negative (HR 4.19, 95% CI, 2.35–7.48, P < 0.0001; Table 2; Fig. 3B) and 14.7% in patients with 1 to 3 positive lymph nodes, respectively (HR 3.15, 95% CI, 1.61–6.15, P = 0.0005; Table 2; Fig. 3C).

In univariable analyses, the OncoMasTR Risk Score was significantly associated with time to DR in all patients (HR 1.91, 95% CI, 1.62–2.26, P < 0.0001; Table 2), in lymph node–negative patients (HR 1.79, 95% CI, 1.43–2.24, P < 0.0001; Table 2), in patients with 1 to 3 positive nodes (HR 1.93, 95% CI, 1.44–2.58, P < 0.0001; Table 2), in years 0 to 5 (HR 1.98, 95% CI, 1.60–2.45, P < 0.0001; Table 2), and in years 5 to 10 (HR 1.81, 95% CI, 1.39–2.38, P < 0.0001; Table 2). Similar significantly prognostic results were obtained for the OncoMasTR Molecular Score and the OncoMasTR Risk Category (Table 2). In addition, we conducted univariate analyses of the three scores using competing risk models and obtained similar results. We also tested the proportional hazards assumption and none of the assumptions for the OncoMasTR variables were violated.

Next in a series of bivariable models, we evaluated whether the OncoMasTR Scores provided additional prognostic value when added to existing prognostic parameters. In years 0 to 10, OncoMasTR Molecular Score, OncoMasTR Risk Score, and OncoMasTR Risk Category added significant prognostic information to age, tumor size, stage, grading, Ki67, NPI, and CTS in the overall population (Supplementary Fig. S1). Again, we observed similar significantly prognostic results in node-negative and in node-positive patients.

We performed three independent multivariable Cox models combining age, tumor size, nodal status, tumor grade, ER expression, PR expression, Ki67 expression, and treatment arm with either the OncoMasTR Molecular Score, the OncoMasTR Risk Score, or the OncoMasTR Risk Category, respectively (Table 3). Each of the three OncoMasTR scores were significantly prognostic in these three multivariable models. In the first multivariable model, tumor size, nodal status, and OncoMasTR Molecular Score (HR 1.49, 95% CI, 1.20–1.86, P = 0.0004) were independent prognostic factors for time to DR (Table 3). In the second multivariable model, only tumor size and OncoMasTR Risk Score (HR 1.58, 95% CI, 1.26–1.97, P < 0.0001) remained independently associated with time to DR (Table 3). In the third multivariable model, tumor size, nodal status, Ki67 expression, and OncoMasTR Risk Category (HR 2.37, 95% CI, 1.41–3.98, P = 0.001) were highly prognostic for prediction of DR (Table 3).

The OncoMasTR gene signature was developed with the aim of accurately estimating the risk of DR in the target population: patients with ER-positive, HER2-negative, early-stage breast cancer with 0 to 3 positive lymph nodes. In contrast to currently available gene signatures in this setting, which were discovered using gene-expression profiling, OncoMasTR was trained using a novel bioinformatic approach that identified a shared network of MTRs from two already validated prognostic gene signatures (8, 9). In this approach, FOXM1, PTTG1, and ZNF367 have been identified to play a key role in breast cancer biology and progression and formed the OncoMasTR Molecular Score (8, 11). The OncoMasTR Risk Score combines the molecular score with tumor size and lymph node status and the OncoMasTR Risk Category classifies patients into the two clinically relevant low- or high-risk groups (11).

The first independent validation of the OncoMasTR Risk Score was carried out in 646 patients from the TransATAC cohort (11). That dataset included postmenopausal women with ER-positive, HER2-negative, breast cancer who received 5 years of endocrine therapy. In the TransATAC study, the OncoMasTR Risk Score was highly prognostic for prediction of DR and provided additional prognostic information beyond NPI and CTS in the overall cohort as well as in node-negative patients. In patients with 1 to 3 positive lymph nodes in that cohort, the OncoMasTR Risk Score was significantly prognostic on its own but did not provide significant additional prognostic value beyond NPI and CTS, possibly due to the relatively small number of patients in this group (11).

The performance of the OncoMasTR test has also been evaluated in a subset of Irish patients enrolled in the TAILORx study (n = 404), showing that OncoMasTR was highly prognostic for DR in this cohort (25). There was only a moderate correlation between OncoMasTR and Oncotype DX: OncoMasTR reclassified a majority of Oncotype DX “intermediate risk” patients to “low risk” and was particularly accurate in reclassifying patients with distant recurrences to “high risk” (25).

In the present study, we report the second independent validation of the clinical performance of the OncoMasTR Risk Score in a large cohort of patients who received endocrine therapy only and who were enrolled into ABCSG Trial 8, a prospective, randomized clinical trial with long-term follow-up. Similar to the TransATAC results, all three OncoMasTR scores (molecular score, risk score, and risk category) significantly predicted DR in year 0 to 10 in all patients, and in patients who were node negative. When compared with the TransATAC results, the OncoMasTR scores show improved performance for the prediction of DR in patients with 1 to 3 affected lymph nodes. Moreover, all three OncoMasTR scores provided significant additional prognostic information when added to standard prognostic variables, NPI, and CTS. As a result, together with the data from TransATAC, there is now level 1B evidence for the OncoMasTR Risk Score in predicting DR.

The strengths of our study are a large, well-documented patient cohort representing the precise clinical target population of the test, the use of analytically validated assays, prospectively defined scores and thresholds, and the blindness of laboratory staff to clinical data. Also, all statistical analyses were carried out according to a predefined statistical analysis plan.

Limitations of this work are the retrospective character and the fact that patients based on the ABCSG Trial 8 inclusion criteria did not receive adjuvant chemotherapy. The latter makes it even theoretically impossible to potentially predict chemotherapy benefit in this cohort. However, if a molecular risk prediction tool accurately defines low risk of distant relapse (as it is the case with the current results: less than 5% DR at 10 years for patients who were node negative, and less than 10% for 1–3 positive nodes), it is highly unlikely that any additional adjuvant therapy (including cytotoxic chemotherapy) could improve those outcomes with reasonable risk/benefit ratio.

The prognostic and/or predictive value of the OncoMasTR Risk Score in other breast cancer subsets such as premenopausal patients with breast cancer will have to be determined in future studies.

Several other gene signatures such as EndoPredict, Prosigna, and MammaPrint were also validated in ABCSG Trial 8 (5, 7, 15, 16, 26, 27). Therefore, we can numerically compare the performance of the OncoMasTR test with other tests. The OncoMasTR test and MammaPrint classified 78% and 71% of ABCSG Trial 8 patients with 0 to 3 involved lymph nodes as “low risk”, respectively (26). In the combined analysis of ABCSG Trial 6 and 8 and in the combined analysis of patients from ABCSG Trial 8 and ATAC, EndoPredict and Prosigna categorized 72% and 58% of the patients with 0 to 3 involved lymph nodes in the low-risk group, respectively (16, 27).

In addition, we performed a cross-study analysis that allows for general comparisons between Prosigna and OncoMasTR in terms of performance. Supplementary Table S2 shows results for patients with ER-positive, HER2-negative breast cancer from ABCSG Trial 8. The patient populations with Prosigna and OncoMasTR test data are not completely identical but, for 1,030 of 1,397 (74%) patients, both Prosigna and OncoMasTR results are available. In contrast to the OncoMasTR subset, which only includes patients with 0 to 3 nodes, the Prosigna subset includes patients with 0 to 9 nodes (7). Since the study endpoints differ in both studies [DRFS (distant recurrences and deaths due to breast cancer) for Prosigna, time to DR for OncoMasTR], we also analyzed DFRS for OncoMasTR. C-Indices are for the continuous scores and are almost identical for both tests. The comparison shown in Supplementary Table S2 demonstrates that both Prosigna and OncoMasTR performed similarly well in ABSCG Trial 8 with regard to distinguishing patients at “high risk” from patients at “low risk”.

However, the present results suggest that the OncoMasTR test classifies a large proportion of patients as “low risk”; the ultimate goal of obtaining additional molecular prognostic information is to identify as many patients with DR risk levels low enough to argue to spare them adjuvant chemotherapy. A formal comparison of the OncoMasTR test with other molecular tests is planned by ABCSG and the next logical step in the development of this signature.

In summary, the OncoMasTR Risk Score is highly prognostic for DR in postmenopausal women with ER-positive, HER2-negative primary breast cancer with 0 to 3 involved lymph nodes. This independent validation in ABCSG Trial 8 provides level 1B evidence for the prognostic role of the OncoMasTR Risk Score.

M. Filipits reports grants from OncoMark Ltd. and European Union during the conduct of the study, as well as personal fees from AstraZeneca, Bayer, Biomedica, Boehringer Ingelheim, Eli Lilly, Merck Sharp & Dohme, Myriad Genetics Inc., Pfizer, and Roche outside the submitted work. C.F. Singer reports grants, non-financial support, and other support from Novartis; grants, personal fees, and non-financial support from Amgen; grants and personal fees from AstraZeneca; personal fees and non-financial support from Roche; and grants from Myriad outside the submitted work. F. Fitzal reports personal fees from Novartis, Pfizer, Roche, and AstraZeneca outside the submitted work. Z. Bago-Horvath reports personal fees from Roche, as well as other support from MSD outside the submitted work. R. Greil reports personal fees and other support from AstraZeneca during the conduct of the study, as well as personal fees and other support from Roche, Amgen, Janssen, AstraZeneca, Novartis, MSD, Celgene, Gilead, BMS, AbbVie, Daiichi Sankyo, and Sanofi outside the submitted work. M. Balic reports grants and other support from Medical University of Graz during the conduct of the study. P. Regitnig reports personal fees and non-financial support from Roche and Novartis, as well as personal fees from Diaceutics outside the submitted work. W. Hulla reports personal fees from Biomedica outside the submitted work. D. Egle reports personal fees from AstraZeneca, Daichii Sankyo, MSD, Novartis, and Seagen, as well as personal fees and non-financial support from Pfizer, Roche, and Lilly outside the submitted work. S. Barron was an employee of OncoMark Limited. T. Loughman reports personal fees from OncoMark Limited during the conduct of the study. D. O'Leary reports personal fees from OncoMark Limited outside the submitted work. W.M. Gallagher reports other support from OncoMark Limited during the conduct of the study, as well as other support from Carrick Therapeutics outside the submitted work; in addition, W.M. Gallagher has a patent for EP3194621B1 issued and licensed to OncoMark Limited. D. Hlauschek reports grants from Oncomark Limited during the conduct of the study. M. Gnant reports personal fees from Amgen, Daiichi Sankyo, AstraZeneca, Eli Lilly, LifeBrain, Nanostring, Novartis, and TLC Biopharmaceuticals outside the submitted work; in addition, an immediate family member of M. Gnant is employed by Sandoz. P. Dubsky reports grants from Oncomark Limited during the conduct of the study; P. Dubsky also reports grants from Cepheid/Danaher, Agendia, and Myriad, as well as other support from Myriad, Merck, AstraZeneca, and Roche outside the submitted work. No disclosures were reported by the other authors.

M. Filipits: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. M. Rudas: Investigation, methodology, writing–review and editing. V. Kainz: Investigation, methodology, writing–review and editing. C.F. Singer: Resources, investigation, writing–review and editing. F. Fitzal: Resources, investigation, writing–review and editing. Z. Bago-Horvath: Resources, investigation, writing–review and editing. R. Greil: Resources, investigation, writing–review and editing. M. Balic: Resources, investigation, writing–review and editing. P. Regitnig: Resources, investigation, writing–review and editing. S. Halper: Resources, investigation, writing–review and editing. W. Hulla: Resources, investigation, writing–review and editing. D. Egle: Resources, investigation, writing–review and editing. S. Barron: Conceptualization, formal analysis, writing–review and editing. T. Loughman: Conceptualization, resources, investigation, methodology, writing–review and editing. D. O'Leary: Conceptualization, resources, methodology, writing–review and editing. W.M. Gallagher: Conceptualization, investigation, methodology, writing–review and editing. D. Hlauschek: Conceptualization, formal analysis, writing–review and editing. M. Gnant: Resources, investigation, writing–original draft, writing–review and editing. P. Dubsky: Conceptualization, data curation, formal analysis, investigation, methodology, project administration, writing–review and editing.

This work was supported by OncoMark Ltd. The work was supported by European Union's Horizon 2020 research and innovation program awarded to OncoMark under grant agreement number 698630.

We thank our patients who contributed to ABCSG Trial 8, the ABCSG investigators, ABCSG centers, and the ABCSG central office for ongoing support. We thank Jana Link and Magdalena Schwarz for assistance in project coordination, Victoria Kügler and Katharina Hofbauer for technical assistance, and Chan-Ju Angel Wang for administrative assistance.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.