Clinicomics for Predicting HER2 Expression in Metastatic Colorectal Cancer: A Multicenter Machine Learning Analysis on Real-World Data
Oncology · 2025
DOI: 10.1159/000549661 ↗Kurz gefasst
Multizentrische Studie in der Zeitschrift Oncology, 2025. Verfasst von Vincenzo Formica und 17 weiteren Autor:innen. Beteiligte Einrichtung: Medical Oncology Unit, Department of Systems Medicine, 'Tor Vergata' University, Rome, Italy, vincenzo.formica@uniroma2.it. Hauptthemen laut MEDLINE: Colorectal Neoplasms, Erb-b2 Receptor Tyrosine Kinases, Machine Learning. Die Arbeit nennt 45 Literaturangaben.
Abstract (Originalsprache)
Introduction
HER2 is expressed in a minority of patients with metastatic colorectal cancer (mCRC), yet it has proven to be a valuable therapeutic target for novel agents such as trastuzumab deruxtecan and tucatinib. Currently, testing is not mandatory in mCRC at baseline and a simple clinical tool to identify patients with a higher likelihood of being HER2 positive would be extremely helpful in guiding test requests and personalized medicine.
Methods
Two machine learning (ML) algorithms were applied to analyze 30 variables available in routine clinical practice (clinicomics) for the prediction of both overall HER2 expression (immunohistochemistry [IHC] score 1–3) and HER2 positivity (IHC score 3 or score 2 with ERBB2 gene amplification). Variables identified as relevant in a training cohort were selected to build an easy-to-use predictive model, whose utility was validated in a separate validation cohort.
Results
ML algorithms consistently showed that hemoglobin (Hb) 100 ng/mL, height >160 cm, and the presence of lymph node metastases were significantly associated with HER2 expression in the training cohort (n = 293). A model using these four variables had an area under the curve (AUC) of 67% (p = 0.0002). Patients with the presence of all predictive factors had a HER2 expression prevalence of 55%, compared to 15% in patients with none of the predictive factors (p < 0.0001), while HER2 positivity prevalence was 36% vs. 0%, respectively (p < 0.0001). The results were confirmed in the validation cohort (n = 96): AUC 68% (p = 0.004); difference in HER2 expression and positivity 58% vs. 12% (p = 0.005) and 47% vs. 0% (p = 0.0002), respectively.
Conclusion
Hb 100 ng/mL, height >160 cm, and the presence of lymph node metastases were associated with HER2 expression and positivity. HER2 testing should be considered mandatory when all these factors are present. The mechanisms linking these four factors to HER2 expression require further investigation.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Medical Oncology Unit, Department of Systems Medicine, 'Tor Vergata' University, Rome, Italy, vincenzo.formica@uniroma2.it.
- Medical Oncology Unit, Department of Systems Medicine, 'Tor Vergata' University, Rome, Italy.
- NEXT: Neurophysiology and Neuroengineering of Human-Technology Interaction Research Unit, Università Campus Bio-Medico di Roma, Rome, Italy.
- Medical Oncology, "Vito Fazzi" Hospital, Lecce, Italy.
- Ospedale Santa Maria Goretti, Latina, Italy.
- IRCCS Istituto Nazionale Tumori Regina Elena, Rome, Italy.
Zitieren
APA
Formica, V., Morelli, C., Rofei, M., Sansone, M., Vitale, J., Zurlo, I. V., Zoratto, F., Dell Apos Aquila, E., Lucchetti, J., Arrivi, G., Torsello, A., Picone, V., Emiliani, A., Cereda, V., Schirripa, M., Missori, F., Calegari, M. A., & Salvatore, L. (2025). Clinicomics for Predicting HER2 Expression in Metastatic Colorectal Cancer: A Multicenter Machine Learning Analysis on Real-World Data. Oncology. https://doi.org/10.1159/000549661
Vancouver
Formica V, Morelli C, Rofei M, Sansone M, Vitale J, Zurlo IV, et al. Clinicomics for Predicting HER2 Expression in Metastatic Colorectal Cancer: A Multicenter Machine Learning Analysis on Real-World Data. Oncology. 2025. doi:10.1159/000549661
BibTeX
@article{Formica2025,
author = {Formica, Vincenzo and Morelli, Cristina and Rofei, Michela and Sansone, Marta and Vitale, Jacopo and Zurlo, Ina Valeria and Zoratto, Federica and Dell Apos Aquila, Emanuela and Lucchetti, Jessica and Arrivi, Giulia and Torsello, Angela and Picone, Vincenzo and Emiliani, Alessandra and Cereda, Vittore and Schirripa, Marta and Missori, Francesca and Calegari, Maria Alessandra and Salvatore, Lisa},
title = {Clinicomics for Predicting HER2 Expression in Metastatic Colorectal Cancer: A Multicenter Machine Learning Analysis on Real-World Data},
journal = {Oncology},
year = {2025},
doi = {10.1159/000549661}
}
EUVIMED