Imaging abnormalities of the pancreas in diabetes: Implications for diagnosis and treatment

Abstract

Purpose of review

Radiographic imaging of the pancreas has drawn recent interest as pancreas volume may serve as a biomarker in identifying the likelihood of diabetes development, subtyping diabetes, and identifying prognostic indicators of poor ultimate outcomes. In this review, the role of pancreas imaging is discussed in various forms of diabetes including type 1 diabetes (T1D), type 2 diabetes (T2D), and diabetes of the exocrine pancreas, particularly diabetes following acute or chronic pancreatitis.

Recent findings

Recent literature of quantitative pancreatic imaging correlating with various forms of diabetes was reviewed. Imaging-derived pancreas volumes are lower in individuals with diabetes, in particular those with T1D. Additionally, morphologic changes, enhancement characteristics, fat content, and MRI signal changes have been observed in different diabetes subtypes. These characteristics, as well as potential confounding variables, are reviewed. Additionally, future areas of research in MRI, CT radiomics, and pancreatitis-related imaging predictors of diabetes are discussed.

Summary

Increased understanding of pancreas imaging features which predict diabetes and gauge prognosis has the potential to identify at-risk individuals and will become increasingly important in diabetes care. This article reviews the current knowledge of common pancreas imaging features as well as future directions of ongoing research in diabetes imaging.

Introduction

Diabetes mellitus by definition results in abnormally high blood glucose (hyperglycemia) due to inadequate insulin production by pancreatic beta cells. The diagnosis and management straightforwardly rely on measuring hyperglycemia, either directly by fasting glucose levels/glucose tolerance testing, or indirectly by hemoglobin A1c (HbA1c), which correlates with blood glucose levels during the preceding 2–3 months. However, the emergence of hyperglycemia in an individual is a lagging indicator of beta cell dysfunction, because hyperglycemia is present only after a majority of beta cells have been lost, or beta cell mass becomes insufficient to meet metabolic demands¹. There is, therefore, significant interest in identifying biomarkers that identify both the likelihood of development of diabetes in those at high risk as well as prognostic indicators of poor ultimate outcomes. Conceivably, imaging could also aid in identifying subtypes of diabetes, something that is sometimes elusive based on biomarkers and clinical characteristics (for instance pancreatitis-related diabetes vs type 2 DM). Recent improvements in anatomic and functional cross-sectional imaging of the pancreas have increased enthusiasm for using diagnostic imaging to predict diabetes. In this review, we discuss current knowledge surrounding the latest pancreas imaging tools as they pertain to diabetes, and review potential uses in type 1 diabetes (T1D), type 2 diabetes (T2D), post-pancreatitis diabetes, and other rare etiologies of diabetes.

Imaging in T1D

T1D is caused by the autoimmune destruction of beta cells in the pancreas, resulting in a deficiency of insulin production. Currently, it is estimated that it accounts for approximately 5–10% of diabetes cases². Identifying patients at risk for T1D can be difficult, as approximately 90% of patients do not have a first-degree relative with diabetes³. However, it is believed that early detection could potentially improve outcomes as it may reduce the likelihood of the common presentation of life-threatening diabetic ketoacidosis. Additionally, immunomodulatory therapy with teplizumab can delay the onset of T1D onset in at-risk individuals, therefore justifying further research to optimize methods that can identify subjects in pre-clinical stages who would be the best candidates for a treatment that modulates the natural history of the disease⁴.

Recent evidence suggests diagnostic imaging of the pancreas has potential to act as a biomarker to identify at-risk patients and may serve other roles in diabetes management. Imaging-derived measurements are generally very reproducible, and results are similar with CT or MRI imaging^5–7. A metanalysis has shown that pancreas volume is significantly reduced in T1D identified on CT, MRI, or US⁸. The degree of exocrine volume loss exceeds the totality of the endocrine pancreas compartment, which accounts for only 1–3% of the pancreas volume. Surprisingly, imaging-derived patient volumes are reduced not only in those with recent-onset T1D, but also in their first degree relatives without diabetes⁹. The reason is unclear but suggests that a reduction in pancreatic volume is not completely explained by loss of insulin production and progression to diabetes. Pancreas volume declines minimally through the course of the disease, and is most pronounced in the early phases, suggesting a future potential role in disease prediction and possibly management ^9–12,. In one study, pancreas volume information helped predict the onset of the hyperglycemic phase of T1D, and its prognostic value was additive to that obtained by the oral glucose tolerance test¹². However, volume alone is a crude marker which is confounded by growth of the pancreas in youth and atrophy in later life¹³. Additionally, pancreas volumes are higher in those with higher BMI and higher visceral fat content, as well as taller individuals and those with wider vertebral bodies¹⁴. Therefore, any volume analysis must adjust for these factors appropriately.

There is a need for precise imaging biomarkers of disease activity in the exocrine and endocrine pancreas, and one potential target is noninvasive estimation of beta cell mass. Unfortunately, the small size of the islets, distribution throughout the gland and lack of differentiation from the exocrine pancreas have hindered cross-sectional islet cell imaging^15,16. Multiple imaging studies have evaluated the use of PET radiotracers to monitor beta cell mass which may help guide treatment, including FDG PET to monitor metabolism and the use of direct radiolabeled monoclonal antibodies. While some have shown early promise, these strategies remain limited by currently unfavorable biodistributions and background signal and are not yet effective enough to guide clinical practice^2,17. MRI has been another area of investigation, with functional parameters and imaging of newer agents. For instance, one agent is GdDOTA-diBPEN, which can be detected in tissue with high levels of Zn^{2+ 18,19}. It is known that Zn²⁺ is packaged in beta cell granules during insulin secretion, and it has been demonstrated in mice that these elevated levels can be detected with GdDOTA-diBPEN MRI, during insulin secretion but not in the fasting pancreas¹⁹. However, this has not yet been tested in humans. Further investigations into efficacy and cytotoxicity are needed prior to clinical application of any beta cell mass agents currently under investigation.

Imaging in T2D

T2D is by far the most common form of diabetes and is usually diagnosed later in life. In T2D the primary pathophysiologic derangement is insulin resistance, but hyperglycemia develops because pancreatic beta cells partially lose their capacity to secrete insulin and compensate for insulin resistance. There are known changes in the pancreas that can be identified when imaging T2DM patients. Meta-analysis has shown that pancreas volumes are reduced in T2D, though to a lesser extent than T1D⁸ (Figure 1). Additionally, pancreatic fat content can be fairly easily quantified on MRI and has been shown to be higher in T2D on a metanalysis⁸ (Figure 2). Pancreatic volume loss and increased fat content are seen in the early phases of disease and appears to worsen with longer disease duration²⁰. However, these findings on the fat fraction have not been universally repeated, with some studies failing to find differences^8,21,22. The fat fraction is highly correlated with age and BMI²². Therefore, like volume, any studies evaluating fat fraction must account for substantial confounders.

Figure 1: Pancreas changes on MRI with diabetes.

Figure A demonstrates normal pancreas (arrow) on sequence in a 21-year-old male on T1-weighted fat suppressed MRI. Volume is best measured using 3d volumetric analysis but can be estimated by single plane measurements. In this case, the pancreatic body thickness is 25 mm. Note that the T1 signal in normal pancreas is higher than reference organs of muscle (star), spleen (*), and liver(X). Figure B demonstrates a mildly reduced pancreas volume in a 44-year-old patient with T2D with 17 mm maximum thickness. Additionally, the T1 signal is less hyperintense than the normal pancreas (figure A). In this case it is near the liver signal. This mild reduction can be better characterized with T1-mapping (not performed here). Figure C demonstrates a markedly atrophic pancreas with T1D with a maximal body thickness of 6 mm. T1D is associated with more pancreas atrophy than T2D.

Figure 2: MRI imaging of pancreatic fat content.

In this 44-year-old T2D patient the pancreatic head and uncinate (arrows) demonstrates hyperintense signal on fat suppressed T1-weighted series (A) with interspersed areas of increased fat seen as dark regions. Fat is confirmed by the substantially decreased signal on opposed phase T1-weighted imaging (B) relative to in-phase images (C). Increased fat content has been observed in T2D patients, though interpretation is complicated by confounding factors of obesity and increased age which also increase pancreatic fat. Note this patient also demonstrates increased levels of visceral fat and hepatic steatosis (*).

There are several unique pancreas imaging features that can be detected on MRI that may further aid in identifying and monitoring T2D. The pancreas generally demonstrates high T1 signal on MRI in normal individuals, which is thought to be due to normal aqueous proteins present in functional pancreas tissue that can be lost in settings where the pancreas develops fibrosis or degradation of normal function. The T1 signal can be readily estimated on routine imaging or more accurately calculated by using dedicated sequences to generate T1 maps. It has been shown that T1 is longer in patients with prediabetes and significantly longer in individuals with diabetes relative to those without diabetesn²³. However, diagnostic accuracy using T1 signal alone is not high, as many factors can affect pancreatic T1 signal including chronic pancreatitis, advancing age, and fatty replacement itself. Other complementary methods may therefore help improve accuracy. One such method is the identification of the pancreatic extracellular volume (ECV) which can be calculated with T1 mapping before contrast and 5 minutes after intravenous gadolinium contrast administration. It has been shown that ECV measured in delayed imaging is increased in processes that produce fibrosis in the heart and in chronic pancreatitis^24,25. Preliminary studies have found that ECV may be more effective than T1 signal in differentiating patients with T2D relative to controls or prediabetes²¹. However, this was a small cohort and additional larger studies are needed to better understand the potential for ECV in monitoring or identifying T2D.

Additionally pancreatic morphology itself has been studied on both CT and MRI, with numerous studies reporting more pancreatic contour irregularity in patients with T2D versus euglycemic patients^26,27. It is not yet clear the extent to which the abnormal size and morphology of the pancreas relates to the disease process in T2D versus causal factors in pancreas pathophysiology itself. It has been shown that with weight loss induced by diet modification, the abnormal size increases and morphological irregularity seen in T2D resolves in those who no longer exhibit clinical manifestations of T2D relative to those with similar weight loss but continued T2D symptoms²⁶. It is hypothesized that this return to normal occurs as a result of more normal insulin secretion in post prandial peaks²⁶, though the exact etiology is unclear and it is also unclear why some patients fail to show resolution of T2D despite similar degree of weight loss.

Radiomics is an active area of imaging research in which numerous quantitative features embedded in digital imaging data in an area of interest are collected and correlations are used to predict outcomes or diagnoses. These features go beyond what can be captured by the human eye, and thousands of variables can be evaluated. There have been single-center studies evaluating the use of radiomics in CT to predict T2D risk, and have shown that models incorporating numerous radiomic features can be predictive of developing T2D^28–30. However, radiomics models can be scanner and institution dependent and multicenter investigations using a pre-defined model can be hard to reproduce and have not yet been attempted in diabetes imaging. Nevertheless, these radiomic studies suggest that morphologic characteristics of the pancreas may hold the key to recognizing individuals at risk for diabetes.

Diabetes of the exocrine pancreas

Diabetes of the exocrine pancreas is an increasingly recognized form of diabetes that results from prior pancreatic disease or resection. Etiologies that fall into this group include diabetes due to acute or chronic pancreatitis, pancreatic malignancy, pancreatic resection, hemochromatosis, and cystic fibrosis (CF). This is a frequently unrecognized and misdiagnosed entity, and by one estimate 87.8% of cases are incorrectly labeled as T2D³¹. However, this is clearly a distinct entity with studies showing patients typically manifest worse glycemic control, higher likelihood to require insulin, and higher rates of diabetes-related complications and all-cause mortality relative to T2D³². While these causal etiologies are often grouped together, each may have differing pathophysiologic processes and disease courses, and some have proposed a naming system focusing on the type of injury: pancreatic-cancer related diabetes, CF-related diabetes, and post pancreatitis diabetes mellitus (PPDM)³³. PPDM has been further divided into diabetes following acute or chronic pancreatitis (PPDM-A and PPDM-C respectively)³⁴. Imaging the pancreas has theoretical potential to improve diagnostic precision, prediction models, and explain the natural history of progression of some subtypes of this entity.

PPDM-C

Studies in China and Japan have demonstrated a significant increase in the incidence of PPDM-C depending on the duration of chronic pancreatitis, showing PPDM-C incidence of 4–10% at diagnosis of chronic pancreatitis but rising to 52–83% after 20–25 years following initial diagnosis^35,36. The cause of PPDM-C is likely multifactorial, but insulin insufficiency is present due to pancreatic damage. However, it has been suggested that insulin resistance may also be present, though the evidence is limited³⁷.

Many imaging features are known to occur in chronic pancreatitis and can serve as markers of disease severity. While this work is ongoing, it is believed that in early stages parenchymal findings can precede ductal changes and in later stages pancreatic calcifications and reduced secretin-induced exocrine secretions can be identified. The parenchymal findings are best detected by MRI, due to its unique excellent soft tissue contrast differentiation relative to CT^38,39. The best studied early and potentially quantifiable parenchymal findings mirror those discussed in T2D above, including T1 lengthening, decreased ECV, and increased fat fraction^40,41. Ductal changes of irregularity, dilatation of main duct and side branches, and response to secretin can be seen on MRCP, but may be absent or normal in early-stage disease. CT is less sensitive to the changes above but can identify pancreas atrophy and is superior to MRI in identification of late-stage parenchymal calcifications (Figure 3). While there has been recent consensus on how to best to identify and report key imaging findings in chronic pancreatitis, there is not yet good longitudinal data identifying key features which predict the occurrence or severity of PPDM-C⁴². This is an active area of ongoing research.

Figure 3: Characteristic imaging findings in a patient with chronic pancreatitis and PPDM-C.

CT images (A) demonstrate a pancreas with atrophy, ductal dilation, and parenchymal calcifications indicating advanced chronic pancreatitis. MRI images (B) additionally demonstrates markedly decreased T1-weighted signal.

PPDM – A

An increasingly recognized cause of diabetes is acute pancreatitis (PPDM-A). Several recent systematic reviews and meta-analyses have placed the incidence of new DM and DM requiring insulin therapy following pancreatitis at 23% and 15%, respectively^43,44, which appears to be higher than the background rate in the general population⁴⁵. Data indicates the incidence of PPDDM-A increases to 40% in with longer term follow-up (>60 months). On a meta-analysis, those with severe pancreatitis or necrotizing pancreatitis have a 37–39% risk of developing DM, compared with a 11–14% risk in mild pancreatitis⁴⁴.

At present, the underlying mechanisms for DM following AP are poorly understood, but increasing evidence suggests that its etiology may be heterogeneous. Recent data suggest that individuals with PPDM-A may share features of Type 2 DM⁴³. Understanding the underlying phenotypes and pathophysiology are key to guiding effective treatment and potential preventive strategies in this population. The Type 1 Diabetes in Acute Pancreatitis Consortium (T1DAPC) was recently formed to address these knowledge gaps⁴⁶. The primary longitudinal study developed by the T1DAPC is the Diabetes Related to Acute pancreatitis and its Mechanisms (DREAM) study⁴⁷, which is described elsewhere in this issue.

One major unmet clinical need in PPDM-A being investigated by this group and others is the ability to accurately risk stratify patients following an episode of AP for the subsequent development of DM. Cross-sectional abdominal imaging may play a pivotal role in this regard. Studies have demonstrated that pancreatic necrosis, recurrent acute pancreatitis, pancreatitis severity, male sex, C-reactive protein, obesity, pancreatic necrosis are associated with increased risk of PPDM-A^44,48–50. However, detailed evaluations of the imaging related risk factors are sparse. A single institution retrospective study from China demonstrated that CT radiomics may be predictive of PPDM-A beyond clinical features⁴⁸. However, radiomics data can be hard to reproduce, and prospective data identifying key identifiable factors on CT and MRI is currently nonexistent in the literature.

Other diabetes etiologies

In other causes of diabetes of the exocrine pancreas, the diagnosis is often straightforward by clinical history and identifiable on imaging, though there is a paucity of data on the role of imaging. Hemochromatosis involving the pancreas can be readily identified on in-phase and opposed-phase imaging which is standard in clinical abdominal MRI, and is quantifiable with iron-sensitive sequences⁵¹. Parenchymal atrophy, calcifications, and fatty replacement in cystic fibrosis (CF) are easily identified by CT. A retrospective study of 94 patients with CF in which 41 had CF -related diabetes clarifies the utility of CT in this setting. All 9 patients with pancreatic calcifications also had CF-related diabetes, and CF-related diabetes was more likely in those with complete fatty replacement of the parenchyma⁵². However, in hemochromatosis and CF the imaging findings have not to our knowledge been studied vis-à-vis diabetes-related incidence or outcomes.

Post-resection pancreatitis deserves special mention as an important iatrogenic cause of pancreatitis. Complete discussion of indications, techniques, and types of pancreatic surgery is beyond the scope of the article. However, there are some studies in post-surgical patients where imaging may shed light on the likelihood of diabetes. For instance, there is convincing evidence that distal pancreatectomy carries a substantially higher risk of developing diabetes (approximately 40%) than pancreaticoduodenectomy (approximately 20–30%) ^53–57. One potential reason for this is that cadaveric analysis has identified that islet cells are twice as prevalent in the pancreatic tail relative to the head and body.⁵⁸ Unsurprisingly, pancreatic resection volume is positively correlated with diabetes risk⁵⁵, and CT generated volumes can be used to quantify residual pancreas and assess diabetes risk⁵⁹. CT resection characterization can therefore potentially be clinically useful in the postoperative setting to gauge future risk of diabetes by identifying the volume and location of pancreatic tissue resected.

Rare hormonal causes of diabetes

Rarely, diabetes is caused by neuroendocrine tumors of the pancreas, which can easily be identified on diagnostic imaging⁶⁰. Glucagonomas and somatostatinomas deserve special mention as they are best identified on pancreas protocol CT, where they appear most commonly as hyper-enhancing masses on arterial phases and can be unrecognized causes of diabetes. Both tumors are extremely rare, though the often-reported prevalences of 1 in 20–40 million are almost certainly significant underestimates, as the incidence of neuroendocrine tumors has risen nearly 7 fold from 1973 to 2012 as high resolution CT has become more widely available⁶¹. Somatostatinomas manifest clinically with malabsorption, steatorrhea, gallstones, and diabetes. In cases leading to diabetes, they are found in the pancreas in 75% of the time, and less commonly in the duodenum or elsewhere in the GI tract⁶². Glucagonoma occurs almost exclusively in the pancreas, and typically presents as a large (>3 cm) mass with clinical potential clinical symptoms of glossitis, necrolytic migratory erythema, and mild diabetes or prediabetes⁶².

Conclusion

Imaging has demonstrated promising potential to influence the diagnosis and management of diabetes mellitus. In some cases, such as in etiologies of diabetes due to pancreas damage or hypersecreting tumors, the clinical role is clear and outlined above. There are key features that may be seen in other cases of diabetes, such as altered size, morphology, and fat content that may increase clinical suspicion of diabetes and should be widely recognized among all physicians treating diabetes or pancreatic diseases. This article outlines these clinically relevant factors as well as key emerging uses in imaging that are undergoing active research with potential to predict, diagnose, and monitor diabetes.

Key points:

• Imaging-derived pancreas volumes are significantly lower in T1D compared to normal individuals, and are also lower in T2D, though to a lesser degree.

• MRI studies have shown that increased fat content, decreased T1 signal, and increased delayed enhancement are observed more often in T2D than normal individuals.

• These imaging features and pancreas volume differences can be confounded based on patient age, sex, BMI, and other factors, and research is underway for more direct biomarkers

• The predictors of diabetes on imaging in pancreatitis is an area of active research with longitudinal studies currently underway.