Introduction
The pancreas remains one of the least studied organs in type 2 diabetes despite its central role in determining both onset and progression. This is largely a consequence of the difficulty of studying this organ in vivo due to its anatomical position surrounded by visceral fat deep in the abdominal cavity [
1]. Only a small number of studies have employed MRI to quantify pancreas volume [
2‐
5], including one study providing a validation of MRI quantification against direct water displacement measurement in mini-pigs [
4]. We developed a robust MRI-based method for precise quantification of pancreas volume and observed a 33% decrease in type 2 diabetes compared with matched volunteers with normal glucose tolerance [
6]. We also observed that the borders of the pancreas were markedly irregular compared with healthy individuals. The question must now be answered of whether these abnormalities are secondary to a decline in the trophic effects of mealtime spikes of intrapancreatic insulin concentration, or whether they may be primary. Given that insulin secretory function can now be returned to normal in type 2 diabetes with re-establishment of normal glucose control [
7‐
9], it is now possible to address this question.
The original observation of irregular pancreas borders was made by simple visual inspection of images [
6]. Although the etiolated appearance with serrated borders was striking, it was necessary to develop a quantitative method to allow precise description. Fractal dimension (FD) analysis has been used in biomedical imaging studies to address questions of biological relevance based on irregularity of borders with varied image complexity [
10‐
12]. It has been successfully applied to histological study of the liver [
13‐
16], pancreas [
17] and spleen [
18], and to CT- and MRI-generated images [
11,
19‐
22]. The nature of the pancreas border introduces challenges to the application of FD analysis, and these have been overcome in this study.
The Counterbalance (Counteracting BetA cell failure by Long term Action to Normalize Calorie intakE) study observed the effect of a very low energy diet (very low calorie diet [VLCD]) on reversal of type 2 diabetes with follow-up over 6 months [
9]. It demonstrated that those individuals achieving fasting plasma glucose <7 mmol/l after acute weight loss regained normal first-phase insulin secretion. MRI studies of the pancreas were carried out, allowing us to report for the first time on the effect on pancreas morphology of the duration of type 2 diabetes and of sustained restoration of insulin secretory function.
Discussion
This study has established that pancreas volume is a marker of reversibility of type 2 diabetes, being higher in responders. However, in those individuals who regain normal insulin secretion after weight loss, there is no detectable increase in pancreas volume over 6 months. The irregularity of the pancreas border has been quantified and shown to change over 6 months in individuals who show a reversal of their diabetes, but to remain unchanged despite a similar weight loss in those who remain diabetic after their weight loss.
In type 2 diabetes, studies using ultrasound, CT and MRI have suggested a 7–22% decrease in pancreas volume [
5,
27‐
30]. Previously, we observed a 33% decrease in pancreas volume in individuals who had type 2 diabetes compared with matched controls with normal glucose tolerance [
6]. In the present study, which included patients with a longer duration of type 2 diabetes, we observed that the pancreas volume decreased by almost 50% relative to the previously reported non-diabetic control group (44.4 ± 2.8 vs 82.6 ± 4.9 cm
3;
p < 0.001). Reflecting this, a significant inverse correlation between pancreas volume and duration of diabetes was defined within the present study population. Although a large study of individuals up to the age of 100 years showed an age-related decline in pancreas volume [
28], within the 35–70 year age range of the present study this did not explain the duration-related decrease. Despite the restoration of acute insulin responsiveness to plasma glucose concentration, pancreas volume based on manual delineation of the pancreas borders did not change.
One aim of the present study was to determine whether the decrease in volume of the pancreas was a consequence of the ongoing pathological processes or whether individuals with a small pancreas might be predisposed to developing type 2 diabetes. Pancreas volume could secondarily be affected by a loss of the normal postprandial rise in insulin, acting by a paracrine effect. Insulin is a potent stimulator of growth at concentrations around ten-fold greater than those required for metabolic effects [
31]. Pancreas tissue is likely to be exposed to very high concentrations of insulin after meals, when the local rise in concentration must exceed the 10–15-fold rise achieved in plasma concentration by at least two orders of magnitude [
32]. When there is no local insulin production in type 1 diabetes, pancreas volume is known to be decreased by one third [
33] and this is evident from just after diagnosis [
3]. In type 2 diabetes, basal insulin levels are raised, but there is an absence of the immediate and major insulin release after meals that could confer local growth-promoting effects of the hormone. There are few animal studies of pancreas volume, although a study in non-human primates observed a modest decrease in pancreas volume 2–6 months after inhibition of insulin secretion by low-dose streptozotocin [
34]. The hypothesis that the restoration of large meal-related increases in intra-pancreatic insulin concentration might restore the trophic effects of insulin and hence normal pancreas volume has been disproved by the present study, at least over a 6 month period. As pancreas volume decreased with increasing duration of diabetes, it appears likely that loss of volume is indeed secondary to the disease rather than being a factor increasing susceptibility to developing the condition. However, future studies must define pancreas volume in populations at risk of developing type 2 diabetes.
There is a marked variation of pancreatic morphology in the general population, with a more serrated boundary of the pancreas generally ascribed to ageing [
35]. Anecdotally, radiologists recognise that the serrated appearance is also associated with diabetes. We previously reported a significantly greater irregularity of the pancreas border in type 2 diabetes using a semi-quantitative visual inspection method [
6]. To quantify the irregularity in an objective manner, we employed FD analysis based on a standard box counting method [
19‐
22]. By a combination of three-dimensional volume segmentation of the pancreas and FD analysis, we developed a new platform that can mathematically define the morphology of the pancreas borders. Unexpectedly, FD analysis showed a significant increase in pancreas complexity after acute weight loss only in the responder group. This was followed by a significant decrease, such that by the end of the study the pancreas border had become smoother. No change in the pancreas border was seen in the non-responder group. Further work is required to determine whether this pattern of change relates to an acute loss of intrapancreatic fat with underlying slow hypertrophy of pancreas tissue continuing over 6 months.
There was a negative correlation between FD value and pancreas volume in the total study population at baseline, while FD increased in proportion to duration of diabetes. The latter can explain the significant difference in FD value between the groups with short-term and long-term diabetes. Regeneration of pancreatic tissues has been reported to affect both the endocrine and exocrine elements [
36]. The present observation that FD change occurred only in the responder group excludes the possibility that change in FD is solely the result of a loss of visceral fat. Longer term follow-up after reversal of type 2 diabetes will be required to determine the extent and prognostic significance of a decrease in irregularity of the pancreas border.
The limitations of this study need to be highlighted. First, although we have previously demonstrated good precision of the method in quantifying pancreas volume [
6], accuracy is more difficult to establish in humans. Volumetry by MRI has been shown to correspond almost exactly to that by water displacement for the pancreas of mini-pigs [
4]. Second, it is likely that the volume of the pancreas had declined very slowly over a decade or more, and follow-up observation for longer than 6 months is now required. Third, it is conceivable that a true change in pancreas volume after reversal of type 2 diabetes may be below the precision power of this quantification method. Work is underway to further optimise image acquisition for the three-dimensional volume-rendering method. The latter technique requires less human input and could potentially generate more accurate volumetric data than the contour-based ImageJ method. Fourth, FD analysis has not previously been applied to study of the pancreas. In the current study, we measured FD on only part of the pancreas surface exposed as the edge of the pancreas on each projection. Further work is required to optimise the method by applying fractal analysis to the whole three-dimensional rendered pancreas to measure the global FD change.
In conclusion, this study quantifies the extent of a decrease in pancreas volume and irregularity of the pancreas borders in people with type 2 diabetes defined in terms of the capacity to regain beta cell function following weight loss. Increasing duration of diabetes was associated with both lower pancreas volume and greater irregularity of the pancreas border. Reversal of type 2 diabetes over a 6 month period brought about smoothing of the pancreas border only in individuals who achieved a return to normal metabolic regulation. No detectable increase in pancreas volume occurred after weight loss irrespective of the metabolic improvement.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.