Decision impact and feasibility of different ASCO-recommended biomarkers in early breast cancer: Prospective comparison of molecular marker EndoPredict and protein marker uPA/PAI-1

Abstract

Background: Adjuvant therapy decisions in early breast cancer are based on accurate risk assessment. Urokinase plasminogen activator (uPA) and plaminogen activator inhibitor-1 (PAI-1) have been the first biomarkers in hormone receptor (HR) positive breast cancer to reach highest level of evidence. The EndoPredict test (EPclin) combines gene expression information with nodal status and tumor size. The aim of this prospective study was to compare uPA/PAI-1 and EPclin as prognostic biomarkers with regard to feasibility, risk stratification and impact on adjuvant therapy recommendation.

Materials and method: 395 patients with HR positive, HER2 negative, intermediate risk breast cancer were enrolled. Relations and concordance of histologic grading as well as EPclin and uPA/PAI-1 values were assessed by Spearman's rank correlation coefficient and Cohen's Kappa. To compare decision impact of EPclin and uPA/PAI-1 three independent case discussions were held: One with known uPA/PAI-1 and EPclin results, one blinded to EPclin alone and another one blinded to both EPclin and uPA/PAI-1.

Results: EPclin could be determined in all 395 (100%), uPA/PAI-1 in 190 (48%) of the tumor samples. EPclin allocated 250 patients (63%) to the low-risk group and 145 patients (37%) to the high-risk group, whereas uPA/PAI-1 allocated 88 patients (46%) to the low-risk group and 102 patients (54%) to the high-risk group. In 59% of cases, both tests showed concordant results. EPclin resulted more frequently in a change of therapy recommendation than the uPA/PAI-1 test (46% vs 24%). Recommendation of adjuvant chemotherapy (CTX) was abandoned twice as often by EPclin (45%) compared to uPA/PAI-1 (22%).

Conclusion: In this first prospective comparison of EPclin and uPA/PAI-1 we found, that EPclin is superior to uPA/PAI-1 with respect to feasibility and decision impact. This leads to substantial avoidance of adjuvant CTX in endocrine-sensitive, HER2-negative breast cancer. Data collection for patients´ clinical outcome is ongoing.

Conflict of interest statement

Competing Interests: JE, EK, AH, KG, SP, LW, KS, HH declared no competing interests. CP is a former employee of Sividon Diagnostics, the manufacturer of the EndoPredict test, and received a salary as well as a bonus from the sales of the EndoPredict test. CP is an inventor on patents relating to EndoPredict but has no financial interest of such patents. MK received honorarium for lectures about gene expression tests from Sividon Diagnostics. This does not alter our adherence to PLOS ONE policies on sharing data and materials.

Figures

Fig 1. Distribution of risk classes based on EPclin and uPA/PAI-1 test results.

Fig 2. EPclin shows a stronger correlation with grading than uPA/PAI-1.

(A) Distribution of the EPclin class as a function of histopathological parameter of grading. Spearman's correlation rho = 0.32; p

Fig 3. Moderate correlation between EPclin and uPA.

Relations are quantified by Spearman’s rank correlation coefficient (r). Allocation to risk classes is indicated by dashed lines. Corresponding concordance is measured by Cohen's kappa (κ).

Fig 4. Very weak correlation between EPclin and PAI-1.

Relations are quantified by Spearman’s rank correlation coefficient (r). Allocation to risk classes is indicated by dashed lines. Corresponding concordance is measured by Cohen's kappa (κ).

Fig 5. Decision impact by EPclin in the overall study population.

Interdisciplinary tumor conference was aware of both EPclin and uPA/PAI-1 results.

Fig 6. Decision impact by EPclin is stronger compared to decision impact by uPA/PAI-1.

(A) Decision impact by uPA/PAI-1. (B) Decision impact by EPclin.