External validation and clinical utility assessment of PREDICT v2.2 prognostic model in young, node-negative, systemic treatment-naïve breast cancer patients

Y Wang, GMHE Dackus, A. Broeks, D Giardiello, M Hauptmann, K Jozwiak, E.A. Koop, M Opdam, S Siesling, G.S. Sonke, Nikolas Stathonikos, N.D. ter Hoeve, E van der Wall, C.H. van Deurzen, Paul van Diest, A.C. Voogd, W. Vreuls, Sabine Linn, M.K. Schmidt

Research output: Contribution to conferencePosterAcademic

Abstract

Background: The PREDICT breast prognostic model is widely used by oncologists for decision-making about systemic treatment for breast cancer patients. However, whether PREDICT could provide accurate predictions before systemic treatment in young patients remains unclear. This study assessed the validity and clinical utility of the latest version of PREDICT(v2.2) in young, node-negative, breast cancer patients who did not receive systemic treatment.
Methods: We selected all women from the Netherlands Cancer Registry, diagnosed with node-negative breast cancer under 40 years of age between 1989 and 2000; a period when national guidelines did not recommend the use of systemic treatment for node-negative patients. The validity of PREDICT to predict all-cause mortality was assessed through calibration and discrimination, calculated as the ratio of observed and expected all-cause mortality (O/E ratio), and the area under the receiver-operating-characteristic-curve (AUC) at 10 years, respectively. Clinical utility of PREDICT was evaluated using decision curve analysis and compared to the clinical utility of the European Society of Medical Oncology (ESMO) guideline. Predefined thresholds for estrogen receptor (ER)-positive and ER-negative patients were based on the MINDACT trial, where adjuvant chemotherapy was recommended to patients with a predicted 10-year all-cause mortality ≥8% (ER-negative) or ≥12% (ERpositive). Clinical utility was represented by net benefit, calculated as the rate of correctly predicted high-risk patients who should receive chemotherapy minus the weighted (odds of the threshold) rate of falsely predicted high-risk patients who should not receive chemotherapy.
Results: We included 2,264 patients with a median age at diagnosis of 36 years. Most patients had ER-positive (70.9%), and grade 1–2 tumors (56.2%); the median tumor size was 16 mm. Observed 10-year all-cause mortality for all patients was 32% higher than the predicted value (table), which was likely due to earlier years of diagnosis of the study population compared to the PREDICT derivation cohort. PREDICT had a 65% chance (AUC) to correctly separate patients who would and would not die within 10 years. Compared to the ESMO guideline, PREDICT only showed slightly higher net benefit in ER-positive patients.
Conclusions: PREDICT on average underestimated all-cause mortality in young, node-negative breast cancer patients. However, the model showed higher clinical utility for ER-positive patients than the ESMO guideline, although the difference was small.
Original languageEnglish
PagesS7
Number of pages2
Publication statusPublished - 2022

Fingerprint

Dive into the research topics of 'External validation and clinical utility assessment of PREDICT v2.2 prognostic model in young, node-negative, systemic treatment-naïve breast cancer patients'. Together they form a unique fingerprint.

Cite this