A Bayesian hierarchical assessment of gastric emptying with the linear, power exponential and modified power exponential models

Background Many studies assessed gastric retention over time utilizing different models, mostly with scintigraphic measures at varied endpoints from limited number of normal volunteers. With a standardized 4‐h gastric emptying (GE) protocol, we compared model fit by the linear, power exponential (PE), and modified power exponential (MPE) models to contrast differences in GE among different groups based on clinical diagnosis and gender. Methods We retrospectively collected 320 patient records with four consecutive hourly scintigraphic measures of percent intragastric residual at the Kansas University Medical Center. We obtained parameter estimates with the Bayesian hierarchical models using informative priors from previous research. Key Results The PE or MPE model captured the time dependent GE rate better than the linear model. The estimated GE rates more than doubled for those without gastroparesis compared to patients diagnosed with gastroparesis. Males tended to empty gastric content faster but were not significantly different from females at the 5% level. Conclusions & Inferences The point estimates and 95% credible interval for GE rates obtained with the PE and MPE models may provide an alternative diagnostic tool for clinicians since it utilizes gastric emptying scintigraphy measures at multiple endpoints which may be sensitive to different aspects of the disease. No agreement in lag phases was obtained by the three models based on respective definitions from previous researches, but similar results would be obtained with the PE and MPE models if both defined lag phase by back projecting the regression lines to the same gastric retention level.