2025/03/16 by Abidin, Nurdiana Zainol
#Artificial Intelligence Machine Learning Digital Biomarkers Systematic Review Meta-analysis Obesity Metabolic Factors Pancreatic Cancer Colorectal Cancer Breast Cancer Oncology Predictive Modeling #Medicine and Health Sciences
paper · doi:10.17605/osf.io/sc3tg
This project aims to systematically review and quantitatively synthesize the evidence on the use of artificial intelligence (AI) and machine learning (ML) models, potentially integrated with digital biomarkers, in predicting outcomes related to pancreatic, colorectal, and breast cancers. Specifically, the review will assess: Cancer incidence - The ability of AI/ML models to predict new cancer cases in at-risk populations. Disease course - How these models forecast disease progression, recurrence, survival, and treatment response in diagnosed patients. Clinically significant endpoints The review follows PRISMA 2020 guidelines and will include a comprehensive search of PubMed, Scopus, and Web of Science. Eligible studies include randomized controlled trials, non-randomized interventional studies, and observational studies that incorporate AI-based predictive modeling. Data extraction will cover study characteristics, patient demographics, details of the AI/ML models (including algorithms, feature selection, and performance metrics), the role of digital biomarkers, and obesity/metabolic variables. Risk of bias will be assessed using established tools such as Cochrane RoB 2.0, ROBINS-I, and PROBAST. Where appropriate, meta-analyses using a random-effects model will be conducted, and heterogeneity will be evaluated with I² and Cochran’s Q tests. This systematic review and meta-analysis will inform clinical decision-making and risk stratification in oncology by clarifying the utility of AI-driven models and the impact of metabolic factors on predictive accuracy.