Intragastric Meal Distribution During Gastric Emptying Scintigraphy for Assessment of Fundic Accommodation: Correlation with Symptoms of Gastroparesis

Abstract

Impaired fundic accommodation (FA) limits fundic relaxation and the ability to act as a reservoir for food. Assessing intragastric meal distribution (IMD) during gastric emptying scintigraphy (GES) allows for a simple measure of FA. The 3 goals of this study were to evaluate trained readers' (nuclear medicine and radiology physicians) visual assessments of FA from solid-meal GES; develop software to quantify GES IMD; and correlate symptoms of gastroparesis with IMD and gastric emptying. Methods: After training to achieve a consensus interpretation of GES FA, 4 readers interpreted FA in 148 GES studies from normal volunteers and patients. Mixture distribution and κ-agreement analyses were used to assess reader consistency and agreement of scoring of FA. Semiautomated software was used to quantify IMD (ratio of gastric counts in the proximal stomach to those in the total stomach) at 0, 1, 2, 3, and 4 h after ingestion of a meal. Receiver-operating-characteristic analysis was performed to optimize the diagnosis of abnormal IMD at 0 min (IMD0) with impaired FA. IMD0, GES, water load testing, and symptoms were then compared in 177 patients with symptoms of gastroparesis. Results: Reader pairwise weighted κ-values for the visual assessment of FA averaged 0.43 (moderate agreement) for normal FA versus impaired FA. Readers achieved 84.0% consensus and 85.8% reproducibility in assessing impaired FA. IMD0 based on the division of the stomach into proximal and distal halves averaged 0.809 (SD, 0.083) for normal FA and 0.447 (SD, 0.132) (P < 0.01) for impaired FA. On the basis of receiver-operating-characteristic analysis, the optimal cutoff for IMD0 discrimination of normal FA from impaired FA was 0.568 (sensitivity, 86.7%; specificity, 91.7%). Of 177 patients with symptoms of gastroparesis, 129 (72.9%) had delayed gastric emptying; 25 (14.1%) had abnormal IMD0 Low IMD0 (impaired FA) was associated with increased early satiety (P = 0.02). Conclusion: FA can be assessed visually during routine GES with moderate agreement and high reader consistency. Visual and quantitative assessments of FA during GES can yield additional information on gastric motility to help explain patients' symptoms.

Trial registration: ClinicalTrials.gov NCT01696747.

Keywords: Gastroparesis Cardinal Symptom Index; Patient Assessment of Upper Gastrointestinal Symptoms questionnaire; fundic accommodation; gastric emptying; gastroparesis.

© 2018 by the Society of Nuclear Medicine and Molecular Imaging.

Figures

FIGURE 1.

Examples of normal FA (A) and abnormal FA (B) assessed by GES. In normal FA, most radiolabeled solids appeared in proximal stomach immediately after meal ingestion (time, 0 min). Over time, solids progressed into distal stomach. In abnormal FA, most radiolabeled solids appeared in distal stomach at 0 min.

FIGURE 2.

Three methods for dividing stomach into proximal and distal portions. (A) Illustration of how computer-generated regions of interest (ROIs) for proximal and distal stomach (solid line) were defined by dividing stomach at one-half the distance along long axis of stomach (dotted line). (B) Illustration of how computer-generated ROIs were defined by selecting equal one-third divisions along long axis of stomach. (C) Stomach incisura angularis is site of formation of acute angle on lesser curvature (arrow) to form localized “notch.” Location of incisura varies depending on degree of gastric distension; therefore, consistent localization is difficult.

FIGURE 3.

IMD over time after meal ingestion for normal volunteers (A) and for patients with abnormal FA (B), as indicated by readers’ assessment of FA. Values are means ± 1 SD at each recorded time.

FIGURE 4.

Logistic regression and receiver-operating-characteristic (ROC) curve for 99 test subjects, with IMD0 or percentage of proximal gastric retention at baseline being used as predictor of abnormal/impaired FA. Area under curve (concordance statistic) for this ROC curve was 0.934; this value implied that corresponding logistic regression model offered excellent fit to data (25).