Decision-Analytic Models of Detection Strategies for Upper Gastrointestinal Cancers: A Methodological Systematic Review

Background: Early diagnosis of upper gastrointestinal (UGI) cancers is associated with better outcomes. However, decision-analytic models to evaluate UGI cancer detection, and diagnostic and treatment strategies vary widely in structure and underlying assumptions and have not been systematically appraised. Purpose: To develop a structured taxonomy of UGI early detection decision models and identify methodological gaps to guide future modelling. Data Sources: MEDLINE (Ovid), Embase, and the National Health Service Economic Evaluation Database (NHS EED) were searched to November 2024, supplemented by screening of relevant reviews and reference lists of included studies. Study Selection: Modelling studies evaluating UGI cancer early detection through screening, surveillance, or multi-cancer approaches were included. Data Extraction: A structured template captured study characteristics, model design, assumptions, and outcomes. Data Analysis: A four-dimension taxonomy was developed describing simulation level, structural form, time progression, and system interactions. Data Synthesis: Sixty-five studies were included, most focusing on gastric or oesophageal cancers (86%) and conducted in the United States and China (70%). Across studies, early detection strategies were most commonly modelled via cancer stage shift (60%), influencing outcomes through stage at diagnosis and downstream treatment effects. From these studies, 65 model structures were identified; 82% employed a state-transition structure, and none incorporated system-level dynamics or explicit resource constraints. Limitations: This review has several limitations, including restriction to English-language publications and limited applicability of findings to healthcare systems with more specialised or structurally distinct diagnostic pathways. Conclusion Research on decision modelling of early detection strategies for UGI cancers has been conducted predominantly in gastric and oesophageal cancers. Greater attention is needed for pancreatic, biliary tract, and gallbladder cancers, alongside more individual characteristics-based simulation approaches and considerations of resource constraints.