Clinicomics for Predicting HER2 Expression in Metastatic Colorectal Cancer: A Multicenter Machine Learning Analysis...

Area of Science:

• Oncology
• Translational Research
• Biomarker Discovery

Background:

• HER2 expression is a target for novel therapies in a subset of metastatic colorectal cancer (mCRC) patients.
• Current HER2 testing is not mandatory at baseline for mCRC.
• A predictive tool for HER2 positivity could enhance personalized medicine strategies.

Purpose of the Study:

• To develop and validate a simple clinical tool to predict HER2 expression and positivity in mCRC.
• To identify routinely available clinical variables associated with HER2 status.

Main Methods:

• Machine learning algorithms analyzed 30 clinicomic variables to predict HER2 expression (IHC scores 1-3) and positivity (IHC score 3 or IHC score 2 with ERBB2 amplification).
• A predictive model was built using significant variables from a training cohort and validated in a separate cohort.

Main Results:

• Hemoglobin <12 g/dL, CEA >100 ng/mL, height >160 cm, and lymph node metastases were significantly associated with HER2 expression.
• The predictive model demonstrated good performance (AUC 67% in training, 68% in validation).
• Patients with all predictive factors showed a substantially higher prevalence of HER2 expression (55%) and positivity (36%) compared to those with none.

Conclusions:

• Four clinical variables (Hb, CEA, height, lymph node status) are associated with HER2 expression and positivity in mCRC.
• Consideration for mandatory HER2 testing is suggested when all four predictive factors are present.
• Further research is needed to elucidate the biological mechanisms linking these clinical factors to HER2 expression.