PREDICT, an online prognostic tool developed from population registries, estimates the potential benefit of different treatments (i.e., surgery alone, chemotherapy, trastuzumab, endocrine therapy, and/or bisphosphonates) for early-stage BC patients (pts). We aim to investigate its prognostic performance among pts enrolled in randomized trials (RCTs) using individual patient data from the GIM and MIG adjuvant studies. We evaluated the performance of PREDICT on a population enrolled in five RCTs (i.e., MIG1, MIG5, GIM2, GIM3, GIM6) between 1992 and 2012 with complete baseline information. We used Uno's concordance index (C-index) to assess the model's discriminative power over all the time range, and Poisson regressions to assess the 5- and 10-year predictions calibration. A total of 6205 pts were included. The median age at BC diagnosis was 57 (IQR 48-65), with 5364 (86.5%) pts having estrogen receptor-positive, 2226 (35.9%) pts having node negative and 2088 (33.7%) pts having grade 3 tumors. Median tumor size was 18 mm (interquartile range (IQR) 13-25 mm). After a median follow-up of 9.3 years (IQR, 5.9–14.8), 1083 deaths were observed. PREDICT assigned higher risk to subjects who had earlier events 62% of the times (C-index 0.62 95%CI 58–65). Overall, the 5- and 10-year OS was underestimated by 5.1% (95%CI 3.9–6.4; observed 5-year OS was 92.2% vs. predicted 87.1%) and by 2.1% (95%CI 0.3–4.0; observed 10-year OS was 74.4% vs. predicted 72.3%), respectively; with 60% and 93% of the expected events respectively observed (calibration intercepts -0.50 and -0.08). The OS underestimation was consistent across subgroups, with worse predictive performance for pts with estrogen receptor-negative, large tumor size, or HER2-positive disease. Our analysis showed that PREDICT underestimates 5- and 10-year OS. Oncologists need to be aware of this limitation when they use this tool to provide prognostic information to pts with early-stage BC treated with modern chemotherapy and endocrine regimens.
No takes yet. Share an insight, caveat, or question.
Borea et al. (2024) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: