Prognostication and treatment predictions for estrogen receptor positive early-stage breast cancer: incorporating the 70-gene signature into the PREDICT prognostication model
Ellen G. Engelhardt,
Mary Ann E. Binuya,
Paul D.P. Pharoah,
Coralie Poncet,
Emiel J.T. Rutgers,
Martine Piccart,
Fatima Cardoso,
Laura J. van ‘t Veer,
Ewout W. Steyerberg,
Sabine C. Linn,
Marjanka K. Schmidt
Affiliations
Ellen G. Engelhardt
Division of Molecular Pathology, Netherlands Cancer Institute, Amsterdam, the Netherlands; Division of Psychosocial Research and Epidemiology, Netherlands Cancer Institute, Amsterdam, the Netherlands
Mary Ann E. Binuya
Division of Molecular Pathology, Netherlands Cancer Institute, Amsterdam, the Netherlands; Department of Biomedical Data Sciences, Leiden University Medical Center, Leiden, the Netherlands; Department of Clinical Genetics, Leiden University Medical Center, Leiden, the Netherlands
Paul D.P. Pharoah
Department of Computational Biomedicine, Cedars-Sinai Medical Center, Los Angeles, CA, USA
Coralie Poncet
European Organisation for Research and Treatment of Cancer Headquarters, Brussels, Belgium
Emiel J.T. Rutgers
Department of Surgical Oncology, Netherlands Cancer Institute, Amsterdam, the Netherlands
Martine Piccart
Oncology Department, Institut Jules Bordet and l'Université Libre de Bruxelles (U.L.B.), Hôpital Universitaire de Bruxelles (HUB), Brussels, Belgium
Fatima Cardoso
Breast Unit, Champalimaud Clinical Center, Champalimaud Foundation, Lisbon, Portugal
Laura J. van ‘t Veer
UCSF Helen Diller Family Comprehensive Cancer Center, San Francisco, CA, USA
Ewout W. Steyerberg
Department of Biomedical Data Sciences, Leiden University Medical Center, Leiden, the Netherlands
Sabine C. Linn
Division of Molecular Pathology, Netherlands Cancer Institute, Amsterdam, the Netherlands; Division of Medical Oncology, Netherlands Cancer Institute, Amsterdam, the Netherlands; Department of Pathology, University Medical Center Utrecht, Utrecht, the Netherlands
Marjanka K. Schmidt
Division of Molecular Pathology, Netherlands Cancer Institute, Amsterdam, the Netherlands; Department of Clinical Genetics, Leiden University Medical Center, Leiden, the Netherlands; Corresponding author. Division of Molecular Pathology Netherlands Cancer Institute, Plesmanlaan 161, 1066CX, Amsterdam, the Netherlands.
Background: The 70-gene signature (70-GS) has been shown to identify women at low-risk of distant recurrence who can safely forgo adjuvant chemotherapy. Incorporating this GS into the well-validated and widely used PREDICT breast cancer model could improve the model's ability to estimate breast cancer prognosis, and thereby further reduce overtreatment and its long-term impact on patients' quality of life. We incorporated the 70-GS into PREDICT-v2.3 and assessed the new PREDICT-GS model's ability to predict 5-year risk of breast cancer death. Methods: Data from the MINDACT trial (N = 5920) was used to estimate the 70-GS's prognostic effect (coefficient = 0.70), which was then incorporated into PREDICT-v2.3. Netherlands Cancer Registry (NCR) data (N = 3323) was used to assess PREDICT-GS's discrimination (area under curve (AUC)), calibration and clinical utility. Results: Compared to PREDICT-v2.3 (AUC: 0.71 (95 % CI: 0.63–0.79)), PREDICT-GS (AUC: 0.76 (95 % CI: 0.69–0.83)) had better discrimination. Both models tended to overestimate the 5-year risk of breast cancer death in the NCR cohort, but the absolute overestimation was smaller for PREDICT-GS. Regarding clinical utility, only at the 10 % decision threshold did we find modest improvement: four extra patients per 1000 tests were correctly classified as not needing chemotherapy by PREDICT-GS compared to PREDICT-v2.3. Conclusion: Extending PREDICT-v2.3 with 70-GS led to modest improvement in its ability to predict 5-year risk of breast cancer death. Future research should focus on assessing the added value of the 70-GS for longer-term prediction of recurrence and death with the incorporation of quality of life in risk prediction tools.
