Optimizing Nutritional Care with Machine Learning: Identifying Sarcopenia Risk Through Body Composition Parameters in...

Area of Science:

• Oncology
• Nutritional Science
• Biostatistics

Background:

• Cancer and treatments negatively impact body composition, increasing malnutrition and sarcopenia risks.
• These conditions are linked to poor prognosis and reduced Health-Related Quality of Life (HRQoL).
• Accurate interpretation of body composition (BC) is crucial for managing cancer patients.

Purpose of the Study:

• To develop a predictive model using unsupervised approaches (PCA and k-means clustering) for sarcopenia risk assessment in cancer patients.
• To identify distinct body composition profiles.
• To enhance the interpretation of Bioelectrical Impedance Analysis (BIA) data.

Main Methods:

• Utilized Principal Component Analysis (PCA) and k-means clustering on Bioelectrical Impedance Analysis (BIA) data.
• Assessed malnutrition and sarcopenia risk using NRS-2002 and SARC-F questionnaires.
• Collected data on anthropometrics, HRQoL (EORTC QLQ-C30), and patient demographics.

Main Results:

• Identified three body composition clusters: High Muscle Profile (HMP), Moderate Muscle Profile (MMP), and Low Muscle Profile (LMP).
• Patients in the LMP cluster were older, had more comorbidities, and a higher risk of malnutrition and sarcopenia.
• Multivariable analysis showed age, lung cancer, diabetes, and malnutrition risk were associated with sarcopenia; LMP patients had a 62% increased sarcopenia risk.

Conclusions:

• The NUTRISCREEN project offers a personalized nutritional pathway for early malnutrition and sarcopenia screening in cancer patients.
• Unsupervised analysis provides distinct BC profiles and identifies key factors associated with sarcopenia risk.
• This data-driven approach can improve clinical practice by defining risk categories and optimizing nutritional strategies for better patient outcomes.