Back

Physiology-guided quantitative symptom analysis for gastroduodenal disorders

Schamberg, G.; Varghese, C.; Uren, E.; Calder, S.; O'Grady, G.; Gharibans, A. A.; BSGM Consortium,

2023-06-12 gastroenterology
10.1101/2023.06.07.23291112 medRxiv
Show abstract

BackgroundCurrent approaches to symptom-based classifications in gastroduodenal disorders are binary and substantially overlapping. We aimed to develop a standardized and quantitative system for classifying patient-level symptom profiles guided on physiological principles. MethodsA large database (n = 787) of 4.5 h (30 min baseline; 4-h postprandial) Gastric Alimetry (Alimetry, NZ) recordings were used to identify, and quantify distinct symptom patterns based on established gastroduodenal physiology concepts. Tests comprised a standardized meal challenge and symptoms were simultaneously recorded at minimum 15 minute intervals using a 10-point likert scale with pictograms encoded in a validated digital App. Key ResultsSix symptom profiles were defined. The meal change metric was used to define meal-induced and meal-relieved symptom profiles, defined as an increase (+2) or decrease (-2) in the average symptom severity between the first post- and pre-prandial hours of recordings. The continuous profile was defined as a reduced range (<3; i.e., difference between the 95th and 5th percentile symptom severity), and thresholded to the 5th percentile of symptom severity being > 2. The symptom/amplitude correlation metric defined the sensorimotor profile, thresholded when the correlation was >0.5. The symptom/amplitude time lag metric was used to define activity-relieved and post-gastric symptom profiles, defined as negative (< -0.25) or positive (>0.25) average difference between the cumulative distribution functions of the symptom and amplitude curves. Conclusions & InferencesStandardized quantification of symptom profiles in relation to a meal-stimulus and gastric amplitude offer a novel classification scheme based on gastroduodenal physiology.

Matching journals

The top 1 journal accounts for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.