In this pilot study, utilising a novel MRI-technique that allows the non-invasive quantification of pancreatic fat in a cohort of well-characterised patients with biopsy-proven NAFLD, we demonstrate that pancreatic steatosis is common in patients with NAFLD, and pancreatic fat content positively correlates with liver histology-determined steatosis grade. Furthermore, the MRI-determined pancreatic fat content is higher in patients who had increased NAFLD activity score (NAS ≥ 5) on liver histology. Finally, we also found that pancreatic fat content is lower in NAFLD patients who had advanced fibrosis. In summary, these findings suggest that steatosis accumulation in patients with NAFLD is probably also occurring in other organs including the pancreas as shown in this pilot study. It is also likely that pathophysiological changes that occur in the liver due to fat deposition also occur in other organs in which fat accumulation is occurring.
These finding suggest that perhaps a similar mechanism may be involved in steatosis of the liver and pancreas. Similar to risk factors for NAFLD, prior studies have also established that obesity, age and insulin resistance are associated with pancreatic steatosis.[18, 19, 21, 23, 24, 38] Pancreatic fat accumulation results in β-cell dysfunction, which may also contribute to hepatic steatosis. The correlation between pancreatic and hepatic steatosis was highlighted recently in 36 healthy participants using MRI. Sijens et al. found that unlike kidney fat content, MRI fat content of the liver and pancreas are coupled and correlate with BMI in healthy patients. In an autopsy study, Van Geenen et al. recently compared postmortem liver and pancreatic histology in 80 patients without known pancreatic or liver disease and noted that pancreatic fat correlated with histology-determined NAS, suggesting that pancreatic fat may play a role in the pathogenesis of NASH. It should be noted, however, that death may lead to inflammatory changes in the fat cells of the pancreas. Therefore, an autopsy study may not be a reliable indicator of in vivo changes in pancreatic and liver fat content in humans.
No correlation with BMI, age or diabetes was noted in our cohort of patients with NAFLD, which differs from findings in prior studies. This may be explained by the fact that prior studies focused on healthy patients without known liver disease. In contrast, all patients in this study had NAFLD, 93.0% were overweight (BMI >24.9 kg/m2) and 67.1% were obese (BMI >29.9 kg/m2). Although only 14 of 43 patients had diabetes, 64.3% of the remaining patients were pre-diabetic (A1c >5.7). In addition, the role of diabetes is unclear, as Saisho et al. reported no association between pancreatic fat and diabetes in a postmortem analysis of 1886 adults.
One of the concerns with measuring pancreatic fat is the need for non-invasive testing as a biopsy cannot be performed on living subjects to evaluate pancreatic fat, inflammation and fibrosis. Some prior studies have relied on postmortem histological analysis of the pancreas[18-20, 40, 41]; however, inflammatory changes with death can make this analysis unreliable as noted previously. Ultrasonography has also been used, but provides a relatively insensitive measure of fat content.[24, 42, 43] More recently, various MRI and MRS techniques have been used to measure pancreatic fat.[12, 13, 23, 38] We chose to use a novel chemical shift-based gradient-echo MRI technique to measure PDFF because of its improved accuracy over traditional techniques and because it has been validated in measuring fat content non-invasively in human tissue.[28, 34, 35, 44] A second concern is the uneven accumulation of fat in the pancreas, which differs from the relatively homogenous steatosis of the liver in NAFLD. Focal accumulation of fat in the pancreas, particularly in the tail and anterior aspect of the head, has previously been described using ultrasonography, computed tomography (CT) and MRI techniques.[42, 45-48] Li et al. used a similar MRI technique as was used in our study to measure fat content in the head, body and tail of the pancreas in healthy subjects and noted no significant different in fat content across regions. Our results are consistent with Li and colleagues and showed that there was no significant difference in fat content between the head, body and tail of the pancreas.
In our study, patients with histology-determined liver fibrosis had significantly less pancreatic fat than those without evidence of liver fibrosis. It is possible that pancreatic steatosis may have a similar mechanism of causing fibrosis in the pancreas as the development of liver fibrosis in patients with NASH. Therefore, the reduced degree of pancreatic steatosis in these patients may be related to increased pancreatic fibrosis. Although the concept of pancreatic fibrosis in non-alcoholics has not been studied extensively, Pitchumoni et al. noted that fibrosis was present in 29% of non-alcoholics in a postmortem analysis. Our study did not use histology or imaging techniques to evaluate fibrosis of the pancreas; however, it has been established previously that lower liver steatosis is associated with greater liver fibrosis in patients with NAFLD. In addition, obesity and pancreatic steatosis have been shown to result in increased cytokine production and fibrosis in the pancreas in studies in which mice were fed a high fat diet.[50, 51]
With the increasing prevalence of NAFLD worldwide, pancreatic steatosis will probably also become increasingly common. Pancreatic fat may induce local effects in the liver that affect the progression of NAFLD. Clinicians performing endoscopic ultrasounds have noted a significant prevalence of pancreatic steatosis and many of these patients may have undiagnosed NAFLD; however, there is little information to guide what clinical management, if any, is required in these patients. There are no data about pancreatic fat in patients with biopsy-proven NAFLD, and this study fills that gap. This study illustrates that there is a strong association between pancreatic fat and liver steatosis. In addition, it suggests that steatosis and lipotoxicity may lead to fibrosis of the pancreas as well as the liver.
Strengths and limitations
The major strengths of this study include the use of an MRI technique that has been well validated to measure fat content in the liver, histological assessment of the liver, a patient population exclusively comprised of subjects with biopsy-proven NAFLD and detailed biochemical and demographic data. As mentioned previously, no prior studies have reviewed pancreatic fat in patients with biopsy-proven NAFLD. In addition, measuring pancreatic fat in all anatomic areas of the pancreas allowed for detailed measurements and confirmation of the homogenous nature of fat distribution in the pancreas. Both the pathologist and radiologist were blinded to the clinical data, and radiology or pathology data respectively. However, we acknowledge following limitations of the study. We did not have a control and it is neither feasible nor ethical to obtain a biopsy of the pancreas to confirm pancreatic steatosis, and evaluate for changes in co-existent pancreatic fibrosis. In addition, we do not have longitudinal data to help clarify whether pancreatic steatosis affects the progression of NAFLD. We also acknowledge that an MRI slice thickness of 8 mm may not provide optimal spatial resolution for measurement of fat in the pancreas. We adopted the spectral model of fat derived from human liver in vivo by Hamilton et al. While the spectral model of fat in human pancreatic tissue is likely to be similar to that in liver tissue, this has not yet been experimentally verified. A refinement for future studies will be the integration, if possible, of a spectral model of fat derived from human pancreas in vivo. Finally, there was a small time interval on average of 42.9 days between liver biopsy and MRI assessment in this study, which could theoretically allow for a change in patient behaviour or management before both studies were completed.