CanDetect Workshop: Progress Toward Earlier Detection
The recent CanDetect workshop in late October brought together researchers, clinicians, and patient and public involvement (PPI) members to review progress across the programme’s workstreams, all focused on improving earlier detection of upper gastrointestinal (GI) cancers.
Workstream 1: Symptoms, Risk Patterns, and AI Models
Workstream 1 (WS1) reported strong progress analysing coded primary care data to better understand symptoms, medications, and other health factors linked to upper GI cancers. The team is initially including a wide range of symptoms, with modelling used later to identify those most important for early detection. A key theme was the value of looking at patterns over time, not just individual “red flag” symptoms. WS1 also presented new work using AI models to analyse patient histories and estimate cancer risk at different future time points.
Workstream 3.3: Health Economics and Diagnostic Pathways
Workstream 3.3 (WS3.3) shared updates on modelling how new early-detection tools could affect NHS resources. Because most people investigated will not have cancer, the team is carefully assessing the impact of false positives on costs and patient wellbeing. WS3.3 also discussed capturing the complexity of real-world referral pathways, which do not always follow a simple sequence.
Workstreams 3.1 & 3.2: AI in GP Consultations
Workstreams 3.1/3.2 presented timely findings on AI “scribes” used in GP consultations. PPI members raised important questions about consent and communication, which will shape future research. The group also discussed how best to communicate risk to patients identified as high risk.
Workstream 2.2: Experiences of Patients Without Cancer
Workstream 2.2 outlined plans to study the experiences of patients urgently referred for suspected cancer who ultimately do not have it, including their diagnoses and feelings about the process.
The workshop highlighted meaningful progress across all workstreams and the vital role of PPI in guiding this research.