| Summary: | Background: Fatty pancreas (FP) has become an increasingly encountered entity in recent years. Several studies have shown an association with several disease states. Aims: we aimed to generate a simple non-invasive scoring model to predict the presence of FP. Method: We performed a retrospective cross-sectional analysis at Galilee Medical Center. Inclusion criteria included patients who underwent endoscopic ultrasound (EUS) for hepatobiliary indications and who had either hyperechogenic pancreas consistent with FP or no sonographic evidence of fatty pancreas. Results: We included 569 patients. Among them, 78 patients had FP by EUS and 491 patients did not have FP. On univariate analysis, obesity (odds ratio (OR) 5.11, <i>p</i> < 0.0001), hyperlipidemia (OR 2.86, <i>p</i> = 0.0005), smoking (OR 2.02, <i>p</i> = 0.04), hypertension (OR 2.58, <i>p</i> = 0.0001) and fatty liver (OR 5.94, <i>p</i> < 0.0001) were predictive of FP. On multivariate analysis, obesity (OR 4.02, <i>p</i> < 0.0001), hyperlipidemia (OR 2.22, <i>p</i> = 0.01) and fatty liver (OR 4.80, <i>p</i> < 0.0001) remained significantly associated with FP. We developed a diagnostic score which included three parameters that were significant on multivariate regression analysis, with assignment of weights for each variable according to the OR estimate. A low cut-off score of ≤1 was associated with a negative predictive value (NPV) of 98.1% for FP, whereas a high cut-off score of ≥2 was associated with a positive predictive value (PPV) of 35–56%. Conclusion: We recommend incorporating this simple score as an aid to identify individuals with FP.
|
|---|