Toward a Patient-Specific Digital Twin for Gastroesophageal Reflux Disease: A Longitudinal Framework for Symptoms, Reflux Physiology, and Daily Context

Gastroesophageal reflux disease (GERD) affects a large population, yet the path from recurring symptoms to a confident explanation can extend across repeated treatment trials, endoscopy, ambulatory reflux monitoring, and changes in daily behavior. These observations are collected at different times and measure different aspects of the disease, which makes a single static label an incomplete representation of the patient. This paper asks whether a patient-specific digital twin could organize these observations as an evolving state rather than as disconnected clinical records. We propose a longitudinal framework that keeps symptoms, measured reflux physiology, treatment and exposure, daily context, and uncertainty as related but distinct components. The framework specifies how intermittent clinical tests and higher-frequency observations can update the patient state, how candidate scenarios such as meal timing, sleep position, and medication use can be represented without being treated as established treatment effects, and when the model should abstain because evidence is insufficient. Published evidence is used to motivate the design and evaluation targets rather than to claim performance from an untested model. We then define an evaluation plan covering temporal forecasting, personalization, scenario comparison, calibration, missing-data behavior, abstention, and external generalization. The resulting framework treats a GERD digital twin as an updateable and testable patient model whose predictions remain tied to the observations and uncertainty that produced them. It is intended as a foundation for future validation rather than as a clinical dosing or treatment recommendation system.