Fatty Pancreas

Abstract

The metabolic consequences of visceral fat deposition are well known, and the presence of intrapancreatic fat (IPF) has been recognized for decades. However, our knowledge about the distribution of fat in the pancreas and its clinical implications is in a nascent stage. Various terms have been proposed to describe IPF; for the purpose of this narrative review, we chose the general term fatty pancreas. Herein, we describe the radiologic, endoscopic, and histopathologic aspects of diagnosing fatty pancreas and provide an overview of the diseases associated with this condition. Our purpose is to highlight diagnostic challenges and identify specific clinical questions that would benefit from further study. As evident in this review, IPF is associated with various metabolic diseases, pancreatitis, pancreatic cancer, and precancer—yet establishing causality needs careful, further study.

Obesity results in visceral fat deposition that in some cases has profound metabolic consequences, directly contributing to the risk of cardiovascular disease and insulin resistance and to increased risk of malignancy. Nearly 70% of the US adult population is either obese or overweight.¹ Obese persons are at increased risk of fatty liver disease, which can lead to steatohepatitis and, subsequently, advanced liver fibrosis. Similarly, an increased likelihood of fat deposition in the pancreas, also called fatty pancreas (FP), has been described in obese persons, but our knowledge about the health consequences of pancreatic fat deposition is at a rudimentary stage.

Debate continues on whether FP is clinically relevant or innocuous. Increased pancreatic fat is not infrequently reported on abdominal imaging studies. However, no consensus currently exists on terminology, prevalence, normal thresholds, and diagnostic criteria of this condition. This lack of consensus has led to speculation on whether FP represents normal peripancreatic fat interdigitating with the pancreatic lobules or true fat deposition in adipocytes within the pancreas, or a combination of both. In addition, various terms have been used to describe intrapancreatic fat (IPF), including pancreatic steatosis, nonalcoholic fatty pancreas disease, lipomatous atrophy of the pancreas, pancreatic lipomatosis, pancreatic adiposity, fatty replacement of pancreas, and fatty infiltration of pancreas.^2–6 In this review, we use the general term fatty pancreas to describe accumulation of IPF and summarize key studies that have described clinical associations of this condition with the aim to highlight areas where the need for further study is critical.

MATERIALS AND METHODS

We searched the MEDLINE database through the Ovid interface using the aforementioned terms. We limited the search to English-language articles indexed from database inception to December 2015. Our literature search retrieved a total of 87 abstracts that were carefully reviewed by 2 investigators (N.A.P. and S.M.) for clinical relevance. Case reports and narrative reviews were excluded. For inclusion in this review, 61 full-text articles were selected, and the bibliographies of these articles were manually searched to identify additional studies of relevance to the topic.

RESULTS

Diagnosis of FP

Before the availability of medical imaging, FP was studied at necropsy or at the time of pancreatic surgery.^7–9 The clinical approach to diagnosing FP was modeled on approaches for quantifying hepatic fat. The retroperitoneal location of the pancreas makes it difficult to visualize on transabdominal ultrasonography (TUS), and obtaining pancreatic tissue for histologic assessment can be technically challenging. Only recently have studies started correlating imaging and histologic findings of IPF.¹⁰ Pancreatic resection specimens are mostly procured in the context of a pathologic state and do not allow for accurate assessment of the “normal” amount of IPF. Not surprising, studies correlating imaging and histology of FP are scarce, and no criterion standard is available for FP diagnosis.

Imaging

A wide range of techniques have been described for imaging IPF, including magnetic resonance imaging (MRI), computed tomography (CT), TUS, and endoscopic ultrasonography (EUS).

Transabdominal Ultrasonography

Traditionally, pancreatic echogenicity has been compared with the liver, kidney, and spleen,^11–14 and a comparatively hyperechoic pancreas has been labeled as FP. Transabdominal ultrasonography is an inaccurate tool for the assessment of IPF. It is subjective; the highly operator-dependent results are influenced by imaging presets (gain and frequency), as well as the patient’s body habitus. Moreover, another inherent limitation to this approach is that hepatic echogenicity is altered by steatosis. Because TUS is widely available, some of the largest studies of FP have used data from TUS evaluations. A large Chinese study reported the presence of FP in 1297 persons (16%) of 8097 healthy volunteers.¹⁵ These results have yet to be validated by subsequent studies.

Endoscopic Ultrasonography

Endoscopic ultrasonography has been used to diagnose and grade FP.^13,16,17 A hyperechoic pancreas is considered synonymous with FP (Fig. 1). Sepe et al¹⁶ used clarity of pancreatic parenchyma and duct margins, in addition to comparative echogenicity, and proposed an EUS grading system (Grades I–IV). However, this observational study did not incorporate tissue sampling or pancreatic fat estimation with CT or MRI. Interobserver agreement is poor for all EUS features of chronic pancreatitis, and the same is likely true for FP.¹⁸ Clearly defined and well-validated EUS criteria with histologic correlation may allow for a more definitive diagnosis of FP.

FIGURE 1.

Computed tomography and MRI of normal and FP. A, Normal attenuation of pancreas (arrow) on non–contrast-medium CT. B and C, No signal change (arrows) between opposed-phase (B) and in-phase (C) images in the pancreas. Of note, the signal of liver drops (asterisks) in opposed-phase image (B) because of diffuse hepatic steatosis. D, Low CT attenuation in the pancreatic body with marbling pattern (arrow) suggestive of FP. E and F, Signal of pancreas drops (arrows) in the opposed-phase (E) compared with in-phase (F) images correlate with CT finding of pancreatic parenchymal fat deposition.

Computed Tomography

Adipose tissue shows negative (−150 to −30 HU) attenuation on non–contrast-medium CT¹⁹ (Fig. 2). Attenuation of the pancreas is expected to decrease with fatty infiltration. For glands that appear diffusely hypoattenuating, mean attenuation from regions of interest across different areas of the gland is used. A reasonable correlation seems to exist between CT and histologic quantification of IPF (r = 0.67).¹⁰ The difference between pancreatic and splenic attenuation on CT has also been used to quantify IPF. In a cohort of 62 patients, Kim et al²⁰ demonstrated reasonable correlation between histologic characteristics and CT attenuation indices. A population-based study of 1721 healthy volunteers that used volumetric histography showed a mean (standard deviation [SD]) pancreatic fat volume of 27.9 (15.7) cm³ and a mean (SD) fat to parenchyma ratio of 0.69 (0.44).²¹ This approach is probably the most accurate CT method for quantifying IPF because it evaluates individual pixel values. However, the fat to parenchyma ratio in that study seems to be abnormally high, most likely due to exclusion of pancreatic tissue with intermediate attenuation.

FIGURE 2.

Focal pancreatic fat mimicking mass lesion. A, Contrast medium–enhanced CT showing hypoenhancing area in the pancreatic head (arrow). B, MRI scan (opposed phase) showing the suspected mass lesion in the pancreatic head (arrow). C, MRI (in phase) scan of the same region showing signal drop (arrow) suggestive of focal fat deposit.

Although non–contrast-medium CT is widely available and relatively inexpensive, it has limitations when used to assess FP. A hypoattenuating mass or cyst can occasionally mimic focal fatty change (Fig. 3). In addition, the attenuation range for fat is based on adipose tissue measurement, which is exclusively adipocytes, and the same threshold is unlikely to be accurate for measuring IPF where adipocytes and pancreatic tissue are interspersed.^7,8,22

FIGURE 3.

Endoscopic ultrasonography image of the pancreas showing a hyperechoic gland, suggestive of FP.

Magnetic Resonance Imaging

Magnetic resonance imaging has superior soft tissue resolution compared with CT and identifies IPF without relying solely on attenuation. Use of the difference in resonance frequency to discriminate between fat and water protons was first described by Dixon²³ in 1984. Although this technique has been subsequently modified, it continues to be the basic principle for the most widely used method for MRI quantification of pancreatic fat (Fig. 2). Wong et al²⁴ studied 202 healthy volunteers and found 90% of this population to have a pancreatic fat content ranging from 1.8% to 10.4%. In another MRI study of 1241 volunteers, mean IPF was 4.5%.²⁵ In a smaller study of 36 healthy volunteers, median total pancreatic fat was 2.7% (interquartile range, 1.0%–6.5%).²⁶ Magnetic resonance imaging–measured IPF needs to be studied and validated in large healthy cohorts to define an accurate diagnostic cutoff value for the reference IPF level.

Magnetic resonance imaging–estimated proton density fat fraction (MR-PDFF) is a novel technique that has been used for quantification of hepatic fat. Results from initial studies suggest good correlation with liver histologic characteristics.^27–29 A recent study demonstrated the feasibility of using MR-PDFF to estimate IPF.³⁰ Magnetic resonance spectroscopy (MRS) is another technique that can accurately quantify triglyceride content of the liver, as well as skeletal and cardiac muscle,^31–33 with high sensitivity.^26,33 In a study of Zucker diabetic fatty rats, MRS-estimated pancreatic triglyceride content had good correlation with biochemically determined tissue triglyceride content, and the investigators demonstrated technical reproducibility with human volunteers.³⁴ Interest is considerable in a 3-dimensional reconstruction algorithm termed iterative decomposition with echo asymmetry and least squares estimation (IDEAL).^35,36 Studies comparing IDEAL with MRS have shown variable correlation between the 2 techniques.^37,38 However, IDEAL has higher spatial resolution, requires a shorter image acquisition time, and overall seems to be better suited for clinical use.

Histology

Currently, no validated objective scoring system is available for histologic assessment of FP. Early studies used a subjective grading score from 1 to 4, where 1 represented few scattered adipocytes in the exocrine pancreas and 4 represented partial or complete replacement by fat.³ More recently, studies have used morphometric analysis and software-aided calculation of percentage pixel area of adipocytes to gain a more objective assessment.^10,21

In contrast to the liver, where intrahepatocyte fat accumulation has been described, IPF is located within adipocytes. Human autopsy studies have not shown intra-acinar staining for the adipocyte marker perilipin-1 (Fig. 4).¹⁰ Interestingly, cytoplasmic vacuoles in the exocrine pancreas of mice fed a high-fat diet have been noted to stain positively for adipose differentiation–related protein but were negative for perilipin.³⁹ The mechanistic pathways involved in possible infiltration of adipocytes or transformation of human exocrine pancreas are being investigated.

FIGURE 4.

Immunohistochemistry of human pancreatic section showing the adipocyte staining with perilipin-1. The perilipin-1 is stained brown. Inset, the vacuoles in adjacent pancreatic acinar cells (black arrows) and islets (red dash–outlined structure) do not show staining. Scale bar represents 50 μm.

Associations and Clinical Consequences of FP

Various conditions have been associated with FP (Table 1). Fatty pancreas can be an end result of a disease process affecting the pancreas (eg, cystic fibrosis [CF]), parallel fat infiltration elsewhere (eg, obesity), or a determinant of clinical outcome (eg, acute pancreatitis). In this section, we summarize the available evidence on the conditions associated with FP.

TABLE 1.

Conditions Associated With FP

Physiologic

Advanced age

Genetic syndromes

CF

Shwachman-Diamond syndrome

Johanson-Blizzard syndrome

CEL gene mutation

Metabolic

Obesity

Diabetes mellitus

NASH

Metabolic syndrome

Inflammatory

Acute pancreatitis

Chronic pancreatitis

Malignant

Pancreatic intraepithelial neoplasia

PDAC

Medication use

Gemcitabine

Rosiglitazone

Miscellaneous

Postoperative pancreatic fistula

HIV infection

Cushing syndrome

HIV indicates human immunodeficiency virus.

FP and Age

Pancreatic fat content increases with age and seems to be highest in the third and fourth decades.²¹ Pancreatic parenchymal volume declines beyond the sixth decade and, as a result, the fat to parenchyma ratio increases in that age group, although total pancreatic fat volume remains stable.²¹ In a recent study of 126 healthy male volunteers, the MRI-estimated IPF was noted to be greater in men aged 50 to 70 years than in younger men (mean [SD], 6.32% [1.18%] vs 2.8% [0.66%], P < 0.01).⁴⁰

FP and Genetic Syndromes

Complete fatty replacement is the most common pancreatic imaging finding in adult patients with CF.⁴¹ The degree of fatty change differs with allelic variation and disease duration.⁴² Fat replacement of the pancreatic parenchyma in CF is a response to acinar injury resulting from ductal obstruction by inspissated secretions. It is unclear whether fat deposition in CF pancreas precedes the development of exocrine and endocrine insufficiency.

Other rare genetic syndromes associated with FP include Shwachman-Diamond syndrome, Johanson-Blizzard syndrome, and heterozygous carboxyl ester lipase (CEL) mutations.^43–46 In a small case-control study, pancreatic fat deposition was found to precede the onset of pancreatic endocrine insufficiency in carriers of heterozygous CEL mutations.⁴⁶

FP and Obesity

Compared with lean persons, obese persons have more IPF.²¹ This association is independent of age and glycemic status. In a cohort of age- and sex-matched healthy volunteers (lean, n = 460; overweight, n = 460; and obese, n = 230), Saisho et al²¹ demonstrated a linear incremental relationship between body mass index and IPF in both diabetic and nondiabetic participants (no diabetes, r = 0.4, P < 0.0001; type 2 diabetes mellitus [DM], r = 0.5, P < 0.0001). Visceral adipose tissue area has also been noted to be an independent predictor of FP.¹⁷ Interestingly, a metabolomic analysis comparing the fat composition of pancreata in obese and lean mice showed a greater content of triglycerides and free fatty acids and less content of phospholipid and cholesterol in the obese animals.⁴⁷ This finding is consistent with the lipid content of adipocytes being predominantly triglyceride. In addition, evidence suggests that the release of nonesterified fatty acids from pancreatic adipocytes may potentiate local pancreatic injury during acute pancreatitis.¹⁰

The effect of weight loss on FP and its downstream metabolic consequences has been an area of interest in recent years. In a study of 23 patients who underwent bariatric surgery (Roux-en-Y gastric bypass and sleeve gastrectomy), IPF decreased significantly after surgery (14% vs 20%, P < 0.01). Of note, changes were unrelated to change in body weight and intra-abdominal fat. Patients with DM who achieved normoglycemia postoperatively had a significantly greater decrease in PF percentage.⁴⁸ Another recent bariatric surgery study found similar improvement in glycemic outcomes associated with a decrease in IPF independent of postoperative change in body mass index.⁴⁹ Decreased insulin resistance alone or in combination with a change in IPF quantity and composition may be contributory. The association between bariatric procedures, quantity and composition of IPF, and beta cell function needs further study.

FP and Nonalcoholic Fatty Liver Disease

In a large cross-sectional study involving 8097 patients, the prevalence of nonalcoholic fatty liver disease (NAFLD) was higher in patients with FP than those without it (67.2% vs 35.1%).¹⁵ Study of a cohort of 60 patients reported a similar high prevalence (57%) of NAFLD.¹³ Increased prevalence of FP in patients with NAFLD has also been observed (Table 2).^50,51 However, a recent study of 41 patients with histologically proven NAFLD did not show a significant correlation between MRI pancreatic fat fraction and the histologic grade of hepatic steatosis (r_s = 0.212).³⁰ Although studies indicate a trend toward FP predicting a higher risk of NAFLD and vice versa, reaching a firm conclusion based on currently available evidence is difficult.

TABLE 2.

Reported Prevalence of FP in Patients With NAFLD

First Author	Year of Publication	Country of Origin	Patients With NAFLD, No.		Patients With FP, No.		Modality for Diagnosis of NAFLD	Modality for Diagnosis of FP	Reported Prevalence, %
			Male	Female	Male	Female
Uygun et al50	2015	Turkey	64	20	31	12	Histology	USG	51.2
Della Corte et al51*	2015	Italy	73	48	33	25	USG, histology	USG	47.9

^*Contained children and adolescents.

USG indicates ultrasonography.

FP and DM

The amount of IPF is greater in persons with DM than those with prediabetes and increases with the duration of DM^52–56 (Table 3). A recent study found less IPF in new-onset DM than in long-standing DM (mean % [SD], 4.3 [4.0] vs 9.3 [7.7]).⁵⁴ It is unclear whether this finding represents a cumulative increase in IPF or fat replacement of atrophic pancreatic parenchyma secondary to diabetic exocrine pancreatopathy.⁵⁷

TABLE 3.

Pancreatic Fat Percentage in Patients With DM and Prediabetes Compared With Normoglycemic Controls

First Author	Publication Year	Country of Origin	Cases (DM/Prediabetes*), No.	Controls, No.	Matching Criteria	Modality for Diagnosing FP	Pancreatic Fat Cases vs Controls
Tushuizen et al55	2007	Netherlands	12/0	24	Age, BMI	MRS	20.4% vs 9.7%† (P = 0.03)
Saisho et al21‡	2007	United States	165/0	660	Age, BMI	CT, histology	30.3 (15.2) cm3 vs 31.6 (16.6) cm3§
Heni et al56	2010	Germany	0/23	28	Age, sex	MRI	8.3% vs 7.4%§ (P = 0.40)
Lim et al54‖	2014	Korea	52/0‖	50	Age, BMI	CT	9.3% vs 2.9%§ (P < 0.01)
Begovatz et al52	2015	Germany	14/14	28	NA	MRI, MRS	DM, 8.35% vs 1.95%† (P < 0.05); prediabetes, 4.79% vs 1.95%† (NS)
Idilman et al30	2015	Turkey	5/0	41	NA	MRI	12.2% vs 4.8%§ (P = 0.03)

^*Prediabetes includes impaired fasting glucose and impaired glucose tolerance.

^†Median.

^‡Included adults and children.

^§Mean (SD).

^‖Patients with long-standing DM (>5 years).

BMI indicates body mass index; NA, not applicable; NS, not significant.

In another large case-control study, patients with FP had a higher prevalence of DM than the non-FP controls (12.6% vs 5.2%), and DM was independently associated with FP in an adjusted logistic regression analysis.¹⁵ In addition, patients with nonalcoholic steatohepatitis (NASH) and FP have been shown to have a higher prevalence of DM and prediabetes than NASH patients without FP (74.4% vs 41.4%, P = 0.004).⁵⁰ In the cohort of 41 patients with biopsy-proven NASH, MR-PDFF patients with DM had more IPF than patients without diabetes (12.2% vs 4.8%, P = 0.028).³⁰

To date, the relationship between IPF and beta cell dysfunction is unclear. Magnetic resonance spectroscopy–estimated median pancreatic triglyceride content has been shown to be greater in patients with impaired fasting glucose, impaired glucose tolerance, or type 2 DM compared with healthy volunteers.³⁴ However, a more recent study found no significant difference in pancreatic triglyceride content among normoglycemic patients versus diabetic patients.⁵⁸ Although animal studies have demonstrated the direct toxic effect of triglycerides on beta cell function and some human studies have reported a negative association between FP and beta cell function, the evidence is not convincing that IPF contributes to the pathogenesis of DM.^56,59,60 Glycemic and cardiovascular risks associated with the combination of DM and FP are poorly understood. It is unclear whether FP is a harbinger of DM in normoglycemic persons.

FP and Acute Pancreatitis

In patients in whom acute pancreatitis develops, obesity is an independent risk factor for its severity.⁶¹ Severe acute pancreatitis (SAP) is associated with a considerable mortality risk,⁶² and this risk seems to be potentiated by obesity irrespective of the cause of the pancreatitis. A human autopsy study of persons who died of SAP suggested an incremental association between IPF and more extensive pancreatic necrosis.¹⁰ In vitro studies have shown that unsaturated fatty acids generated from IPF are proinflammatory and cause pancreatic acinar cell necrosis through a mechanism that involves release of intracellular calcium and inhibition of mitochondrial complexes I and V.¹⁰ Animal studies have indicated possible presence of intra-acinar fat in obese mice.^63,64 The presence and composition of fat in human pancreatic acinar cells need to be studied in greater detail to understand the role of IPF in this pathogenetic pathway. Therapy targeted at lowering the circulating levels of unsaturated fatty acid may have a future role in improving outcomes of SAP.

FP and Pancreatic Fistula

Pancreatic fistula is a frequent complication after pancreatoduodenectomy and has been reported in up to 25% of all patients.^65,66 Several studies have found FP to be an independent predictor of postoperative pancreatic fistula.^67–70 These studies used variable criteria for histologic assessments of IPF in the resected specimen. Lee et al⁷⁰ combined histologic assessment with preoperative MRI and demonstrated reasonable correlation between imaging and histologic characteristics for IPF assessment (r² = 0.560, P = 0.013). In their study, patients in the pancreatic fistula group (n = 20) had a greater amount of IPF than the healthy controls (mean [SD], 35.5 [14.7] vs 23.3 [10.6], P = 0.006). A more recent study using pancreatic attenuation on preoperative CT to assess FP did not show a risk association between FP and eventual development of pancreatic fistula. That study, however, did not incorporate histologic assessment of pancreatic fat.⁷¹

The role of FP as an independent risk factor for pancreatic fistula continues to be controversial. A study aimed at identifying the relative risk of pancreatic fistula in patients with both radiologic and histologic evidence of FP would be useful.

FP and Pancreatic Ductal Adenocarcinoma

Strong evidence from case-control and cohort studies suggests an association between obesity and the increased risk of pancreatic ductal adenocarcinoma (PDAC) and cancer death.^72–74 The association of FP and PDAC has been assessed in only small case-control studies.^75–78 Hori et al⁷⁵ quantified FP using image analysis software in carefully selected tumor-free areas of resected pancreas and reported a significantly greater median percentage of IPF in cases than in controls (26% vs 15%, P < 0.001). In another study that defined FP as higher than 5% fatty change in a resected section without malignant infiltration, PDAC cases had a greater proportion of IPF than controls without cancer (72% vs 44%).⁷⁶ Studies also have linked IPF to a higher risk of lymph node metastasis in patients with PDAC.⁷⁸ Evidence is not clear whether the increased IPF of patients with PDAC is a primary process or is secondary to acinar atrophy as a result of either underlying chronic pancreatitis or tumor-associated ductal obstruction. Rebours et al⁷⁹ recently demonstrated an association between IPF and pancreatic intraepithelial neoplasia lesions. Interestingly, the same study did not note incremental correlation between body mass index and the number of pancreatic intraepithelial neoplasia lesions, which is suggestive that IPF has a possible independent role in pancreatic tumorigenesis.

Some evidence suggests that the proinflammatory milieu induced by obesity contributes to pancreatic oncogenesis through the activation of K-ras signaling pathways.⁸⁰ The underlying mechanisms of adiposity-driven pancreatic tumorigenesis need to be carefully examined.

CONCLUSIONS

The existence of fat in the pancreas has been known for decades. The pathophysiologic mechanisms that lead to IPF deposition and the clinical and metabolic implications are beginning to be understood. The effect of excess IPF on pancreatic exocrine or endocrine function and fibrosis is unclear. Potential benefits and methods to reduce pancreatic fat content need further exploration. It is safe to conclude on the basis of available evidence that IPF is not innocuous and may have a role in causing or aggravating metabolic, inflammatory, and neoplastic disease processes.

Abbreviations:

CEL

carboxyl ester lipase

CF

cystic fibrosis

CT

computed tomography

DM

diabetes mellitus

EUS

endoscopic ultrasonography

FP

fatty pancreas

IDEAL

iterative decomposition with echo asymmetry and least squares estimation

IPF

intrapancreatic fat

MRI

magnetic resonance imaging

MR-PDFF

magnetic resonance–estimated proton density fat fraction

MRS

magnetic resonance spectroscopy

NAFPD

nonalcoholic fatty pancreas disease

NASH

nonalcoholic steatohepatitis

PDAC

pancreatic ductal adenocarcinoma

SAP

severe acute pancreatitis

TUS

transabdominal ultrasonography