The relationship between pancreas steatosis and the risk of metabolic syndrome and insulin resistance in Chinese adolescents with concurrent obesity and non‐alcoholic fatty liver disease

Summary Background The incidence of childhood obesity and associated comorbidities are on an increasing trend worldwide. More than 340 million children and adolescents aged between 5 and 19 years old were overweight or had obesity in 2016, from which over 124 million children and adolescents (6% of girls and 8% of boys) had obesity. Objective To describe the relationship between pancreas steatosis, body fat and the risk of metabolic syndrome, insulin resistance in Hong Kong Chinese adolescents with both obesity and non‐alcoholic fatty liver disease (NAFLD). Methods Fifty two adolescents with obesity and NAFLD were analysed (14‐18 years), stratified into fatty and non‐fatty pancreas groups using chemical shift encoded MRI‐pancreas proton density fat fraction ≥5%. Pancreatic, abdominal subcutaneous adipose tissue (SAT)/visceral adipose tissue (VAT) volumes, biochemical and anthropometric parameters were measured. Mann‐Whitney U test, multiple linear/binary logistic regression analyses and odds ratios were used. Results Fifty percent had fatty pancreas, 38% had metabolic syndrome and 81% had insulin resistance. Liver proton density fat fraction (PDFF) and VAT were independent predictors of insulin resistance (P = .006, .016). Pancreas and liver PDFF were both independent predictors of beta cells dysfunction (P = .015, .050) and metabolic syndrome (P = .021, .041). Presence of fatty pancreas in obesity was associated with insulin resistance (OR = 1.58, 95% CI = 0.39‐6.4) and metabolic syndrome (OR = 1.70, 95% CI = 0.53‐5.5). Conclusion A significant causal relationship exists between fatty pancreas, fatty liver, body fat and the risk of developing metabolic syndrome and insulin resistance. Key Points Fatty pancreas is a common finding in adolescents with obesity, with a prevalence rate of 50% in this study cohort. Liver PDFF and VAT are independent predictors of insulin resistance while pancreas PDFF and liver PDFF are independent predictors of both beta cells dysfunction and metabolic syndrome. Presence of fatty pancreas at imaging should not be considered as a benign finding but rather as an imaging biomarker of emerging pancreatic metabolic and endocrine dysfunction.

Conclusion: A significant causal relationship exists between fatty pancreas, fatty liver, body fat and the risk of developing metabolic syndrome and insulin resistance.

Key Points
• Fatty pancreas is a common finding in adolescents with obesity, with a prevalence rate of 50% in this study cohort.
• Liver PDFF and VAT are independent predictors of insulin resistance while pancreas PDFF and liver PDFF are independent predictors of both beta cells dysfunction and metabolic syndrome.
• Presence of fatty pancreas at imaging should not be considered as a benign finding but rather as an imaging biomarker of emerging pancreatic metabolic and endocrine dysfunction.

K E Y W O R D S
fatty liver, insulin resistance, magnetic resonance imaging, metabolic syndrome, pancreas

| INTRODUCTION
Childhood obesity and its associated comorbidities are increasing. 1 The excess fat tends to accumulate in undesired areas such as the liver, pancreas, heart, skeletal muscle and visceral adipose tissue. 2 It is this distribution of fat that plays a critical role in the development of complications 3 and is understood to pose a risk for insulin resistance. 4 Lee et al 5 showed that a third of children and adolescents with obesity have glucose intolerance and relative β-cell failure. Fatty pancreas for instance has been shown to be a significant risk factor for insulin resistance/diabetes in children and adults. 6 In adults, fatty pancreas is found to be significantly correlated with β-cell dysfunction and decreased insulin secretion. 7 Fatty pancreas has also been associated with metabolic syndrome, which is characterized by central obesity, hypertension, impaired glucose tolerance and dyslipidemia. 8  creatic fat was present in adolescents with obesity who also had metabolic syndrome. Likewise, fatty pancreas is further associated with central obesity, which is linked to both insulin resistance and metabolic syndrome. 8 Interestingly, studies have shown that in people suffering from impaired glucose metabolism there was decreased pancreatic volume and increased pancreatic fat. 10,11 Other studies also showed that individuals with type 2 diabetes mellitus (T2DM) had a smaller pancreatic volume and higher pancreatic fat when compared to people without T2DM. 12 Suggesting that a large pancreatic volume may indicate a larger reservoir of beta cells and greater capacity to withstand the various factors that contribute to the development of diabetes. 13 Chemical shift encoded MRI (CSE-MRI) is an excellent quantitative method to calculate fat in the body. It is robust, accurate, reproducible, vendor and operator independent method that is able to quantify body, pancreatic and hepatic fat content. 14 Very limited studies have examined the relationship among fatty pancreas, other ectopic fat deposition areas in the abdomen and the risk of developing metabolic syndrome and insulin resistance in adolescents using magnetic resonance imaging. Most of the available studies used Ultrasound. To the best of our knowledge, to date only four studies involving predominantly European Caucasian, 1,15,16 African American/Latino 17 children and adolescents with obesity, with and/or without NAFLD used MRI to evaluate the afore-mentioned relationship. Based on these findings, the purpose of our study was to utilize CSE-MRI (mDixon method) to evaluate the relationship of fatty pancreas, whole abdominal subcutaneous/visceral adipose tissues and the risk of developing metabolic syndrome and insulin resistance in Chinese adolescents with both obesity and non-alcoholic fatty liver disease (NAFLD).

| Study population
This study was a substudy of Chan et al 18 reported previously, which evaluated the efficacy of dietitian-led lifestyle modification programme to reduce non-alcoholic fatty liver disease (NAFLD) in adolescents with obesity. The study was approved by our institutional review board and written informed consent was obtained from the parents or guardians and all participants assent to participate in the study. Seventy-nine (79) children with obesity were screened between February 2014 and March 2014 for the presence of liver fat content level of ≥5% by proton magnetic resonance spectroscopy to determine fatty liver. 19 Fifty two participants were finally enrolled in the study ( Figure 1). In order to be more accurate with liver fat measurements, we re-evaluated the liver fat content in all the 52 participants using chemical shift encoded MRI method.
NAFLD was defined as liver proton density fat fraction (PDFF) of ≥3.5% in children and adolescents. 20

| Physical examinations and anthropometrics
Physical examination and anthropometric measurements were carefully taken at least within 24 hours of performing abdominal MRI scan. These included body weight in kilograms (kg), while the child was in light clothes and bare feet. The height in metres was measured using a flexible non-stretchable measuring tape with the subject standing upright, extended knees, hips, waist and neck. BMI was calculated as weight (kg) divided by height in metre squared and z-scores were derived using World Health Organization references. 22 Obesity was diagnosed if the BMI was ≥95th percentile for age and sex. Waist circumference (WC) was measured at the mid-point between the lower costal margin and the iliac crest over the unclothed abdomen in the standing position, bare feet, at the end of normal expiration with the measuring tape stretched all around the body in the horizontal position.

| Blood pressure (BP)
Participants were allowed to be seated in a quiet room for 3-5 minutes before measurement to reduce anxiety, with the back supported and feet uncrossed on the floor. 23 Talking was not allowed during BP measurements. Systolic and diastolic blood pressure was measured from the right arm in all studied children three times, at 2 minutes' intervals using a standard commercially available automated blood pressure machine. The results were recorded as necessary using the lowest reading of the three obtained readings. Hypertension was defined as blood pressure in subjects above 10 years old with systolic and diastolic blood pressure ≥130/85 mm Hg. 24   The mean signal intensities from the three ROIs were averaged to get the mean pancreatic fat fraction as the final result. The interclass correlation coefficient was calculated to assess the reliability of the measurements from the two readers. In cases of discrepancies, a consensus was reached by repeating the measurements and then an average was obtained for final analysis. Contrasting to NAFLD, there is no well-recognized threshold to determine the upper bound of pancreatic fat for healthy individuals or adolescents. However, a study by Maggio et al 1 recommended that the upper bound normal pancreatic fat fraction in adolescents is 5%, therefore, this cut off was adopted in our study.

| Liver proton density fat fraction (PDFF)
Reader 1 determined liver PDFF for each participant using the same RadiaAnt viewer. Nine elliptical regions of interest (ROIs) set to 4 cm 2 were placed into all nine Couinaud liver segments localized on PDFF maps (obtained in at least two slices) avoiding the hepatic blood vessels, bile ducts and motion artefacts. 29 The average liver PDFF from all the nine segments was used as the final measurement. All measurements were repeated thrice to define the intraclass correlation coefficient. and VAT volumes were extracted from CSE-MRI proton density fat fraction images of the whole abdomen, that is, region from the dome of the diaphragm to the symphysis pubis as shown in Figure 4. Briefly, this algorithm utilized Bresenham's Line method and Midpoint Circle method to construct a spoke-like template, and this template was applied to the scan over the adipose tissue to separate SAT and VAT.

| Statistical analysis
Normally distributed data was expressed as means ± SD, unless stated otherwise. Differences between two groups were analysed using Mann-Whitney U test. Kruskal-Wallis test was used to compare the fat distribution in the three regions of the pancreas. Interclass correlation coefficient was used to evaluate the inter and intra reader agreement of ROIs with 1 to 2 weeks' interval between measurements.
Pearson's correlation coefficients were used to assess linear relationships between variables. To evaluate the causation relationships between variables, multiple linear and binary Logistic regression analyses with correction for multiple comparisons were used. Relative risk was determined by odds ratios. All tests were two-sided and P-values <.05 were considered statistically significant. Statistical analyses were performed by using the SPSS statistical package software (version 25.0; SPSS, Chicago, IL).

| Blood biochemistry analysis
No statistically significant differences were observed between groups in all blood biochemical markers, lipid profiles, HOMA-IR and HOMA-B. However, there was a trend of higher HOMA-IR, serum insulin and lower HOMA-B (higher beta cells dysfunction) in the fatty pancreas group. 79% of all the participants were found to have insulin resistance, that is, HOMA-IR ≥2.6 32 but all had normal QUICKI ≥0.36. 33 The proportion of participants with insulin resistance was not different between the two groups, that is, 77% vs 81% P = .390, in the non-fatty pancreas group vs fatty pancreas group respectively.
Metabolic syndrome was diagnosed in 27% vs 39% of the participants in the non-fatty pancreas group vs fatty pancreas group respectively, P = .38. Hypertension was diagnosed in 35% vs 58% of the participants in the non-fatty pancreas group vs fatty pancreas group, P = .098.

| Radiological analysis
The mean liver PDFF between groups was not statistically different Comparison of anthropometric and patient characteristics in participants with and without fatty pancreas   Table 3.

| DISCUSSION
Excess body fat tends to accumulate in ectopic areas, and is associated with metabolic diseases. 34 Unlike the liver that has been widely studied in relation to obesity related comorbidities, limited studies utilizing MRI are available that demonstrated the relationship among fatty pancreas, body fat and the risk of metabolic syndrome and insulin resistance in adolescents with both obesity and non-alcoholic fatty liver disease. In this study it has been demonstrated that fatty pancreas is a common finding (50%) among Chinese adolescents with concurrent obesity and NAFLD. Fatty pancreas, fatty liver and visceral adipose tissue (VAT) were shown to be interrelated, mediated by general and central obesity and were significant risk factors in the development of insulin resistance and metabolic syndrome.
The prevalence of fatty pancreas in this present study is in agreement with the literatures range between 44% and 58% 6,33,35  This study showed that the independent predictors of fatty pancreas are BMI, fasting plasma glucose and total cholesterol, in agreement with a previous study. 37 Of note, BMI as the highly significant independent factor (over glucose and total cholesterol) to the development of fatty pancreas could be another probable explanation why the prevalence rate of fatty pancreas was 50% in our study, especially that the definition of obesity in the Chinese population uses lower BMI (Kg/m 2 ) cut offs. 38,39 It was also shown that both pancreas and liver PDFF were independent predictors of beta cells dysfunction.
Additionally, liver PDFF showed a significant linear association with HOMA-B in agreement with a previous study. 40 Tushuizen  This study showed that the odds ratio of developing insulin resistance in adolescents with both obesity and fatty pancreas was nearly 2 folds than in those without fatty pancreas, similar to the findings of and ultimately leading to insulin resistance. 49 Accordingly, the mechanism underlying ectopic fat distribution in the liver and pancreas with resultant insulin resistance may be different. Therefore, these results suggest that increased liver PDFF and VAT play a primary and critical role in the development of insulin resistance vis-à-vis type 2 diabetes mellitus (T2DM) while increased pancreas PDFF plays an additional role. However, despite the finding that the role of fatty pancreas in the development of insulin resistance is an adjunct one, its presence appears to indicate a "worsening metabolic condition" in an individual.
As opposed to our findings, other studies have shown that increased pancreatic fat plays a primary role in the development of insulin resistance vis-à-vis T2DM. 6,7,15,[50][51][52] Binary logistic regression showed that both pancreas and liver PDFF were independent predictors of metabolic syndrome similar to other studies. 6,53 We further showed that the odds ratio of metabolic syndrome in adolescents with obesity and fatty pancreas was 2 folds than in those without fatty pancreas, similar to findings of Singh et al. 9 These findings support the hypothesis that fatty pancreas and fatty liver are a part of the metabolic syndrome. They further reiterate the essential role of pancreas and liver in glucose and energy homeostasis/lipid metabolism. Any disruption of the anatomical integrity such as ectopic fat infiltration has the potential to distort organ function resulting in metabolic disorders and associated complications such as T2DM and cardiovascular diseases. Therefore, in our opinion, the presence of fatty pancreas at imaging should not be considered as a benign finding but rather as an imaging biomarker of pancreatic metabolic and endocrine dysfunction. This could serve as a wake-up call to the clinicians to prioritize these patients under early interventional programme with best possible treatment option to reverse the vulnerable metabolic situation.
Strangely, no direct correlations among pancreas PDFF, liver PDFF and VAT were noted in our study similar to the findings of van der Zijl et al. 42 However, BMI and central obesity correlated significantly with all these parameters including SAT in agreement with a previous study. 51 These findings imply that the relationship that exists among these parameters is mediated by both general and central obesity. In view of the above inter-related associations and correlations, we postulate that any early intervention aimed at reducing excess body fat would have a ripple effect in reducing ectopic fat within the liver, pancreas and other visceral organs like kidneys, resulting in reduced risks of metabolic syndrome, insulin resistance and possibly minimize long term complications such as T2DM.
Thus, we would recommend "screening" in this population group once the waist circumference and BMI/BMI z-scores are outside the normal range. Furthermore, as total cholesterol and plasma fasting glucose were found to be independent predictors of fatty pancreas, we would recommend initiating "screening" if these parameters are elevated. Moreover, as the presence of fatty pancreas was as high as 50% in this Hong Kong Chinese cohort with fatty liver, the presence of fatty liver could be an indicator of the presence of fatty pancreas.
Since fatty liver and fatty pancreas can be assessed during the same scanning session using chemical shift encoded MRI (CSE-MRI) PDFF, the assessment of pancreatic fat can be performed in addition to assessment of hepatic fat.
For limitation, this study only involved a cohort of adolescents with both obesity and NAFLD without healthy controls. Insulin resistance and beta cells function were not directly measured. Instead we used HOMA-IR and HOMA-B, acceptable surrogates to measure insulin resistance and beta cells function respectively. Besides, a relatively small sample size may also reduce the statistical power of our measurements. Finally, our study cohort was predominantly Chinese, therefore caution should be taken in the generalization of the results.
In the future, it will be valuable to provide longitudinal follow up for those participants with high risk of developing insulin resistance and metabolic syndrome into their adulthood, to monitor their bio- lation study in Hong Kong. As the procedure is simple and straightforward, such a study can be extended to study groups with individuals who do not have obesity and/or without NAFLD.
In conclusion, fatty pancreas is a common finding in Chinese adolescents with both obesity and non-alcoholic fatty liver disease. A significant causal relationship exists between fatty pancreas, fatty liver, body fat and the risk of developing metabolic syndrome and insulin resistance.