Artificial intelligence-based muscle analysis risk assessment of treatment-related toxicity in metastatic colorectal cancer

Abstract

Introduction

Up to 60% of adults with metastatic colorectal cancer (mCRC) receiving combination cytotoxic chemotherapy may experience loss of skeletal muscle mass and function. This study explores associations of artificial intelligence (AI)-based skeletal muscle assessment with hematologic toxicity and chemotherapy relative dose intensity (RDI) in adults with mCRC receiving standard combination chemotherapy.

Methods

We conducted a retrospective analysis of adults with mCRC receiving first-line FOLFOX or FOLFIRI over 6 months (≤12 cycles) from 1/2011 to 11/2018. We used a validated AI-based skeletal muscle assessment on baseline (prior to starting chemotherapy) computed tomography scans to determine skeletal muscle index (SMI, cm²/m²), categorizing low SMI using independent sex-specific cut-off values. We sought to evaluate the association between low SMI and the incidence of grade ≥3 (G ≥ 3) cytopenia over 6 months. Secondary endpoints included time to G ≥ 3 cytopenia and RDI.

Results

Overall, 126 adults met inclusion (median age = 61 years [range, 29–85]; 56 [44%] female) with a median BMI of 26.6 kg/m² (IQR, 24.1–30.5 kg/m²), including 59 (47%) with a low SMI. G ≥ 3 neutropenia incidence was higher in adults with low SMI (31% vs. 15%, p = 0.036). There was no difference for other G ≥ 3 cytopenias (39% vs 24%, p = 0.067) or median time to G ≥ 3 neutropenia (p = 0.053). Patients with a low SMI had a lower 5FU-bolus RDI (p = 0.045).

Conclusion

Adults with mCRC receiving first-line chemotherapy with low SMI experienced more G ≥ 3 neutropenia and decreased 5-FU bolus RDI.