Aim
The primary aim of our study was to quantify and compare the vascular anomalies of pancreas-bound arteries of patients with type 2 diabetes and of control subjects. The secondary objective was to compare the volume and density of the pancreas between the two groups.
Design and setting
We performed a retrospective monocentric study in a tertiary referral center for type 2 diabetes (Cochin Hospital, Paris), comparing patients with type 2 diabetes and matched control subjects. This was a retrospective study, using existing clinical data only.
Patients were included if they were older than 18 years old, had type 2 diabetes, and had had an abdominal CT before and after iodinated contrast media intravenous injection between January and December 2017 in our institution. Using the data available from a study performed on postmortem pancreatic angiography [15], we estimated that we needed at least 39 subjects in each group in order to detect a difference between the groups (OR of decreased vascularity in patients with type 2 diabetes versus in control subjects: 5.4, expected proportion of exposed controls: 20%, alpha risk: 0.05, study power: 90%).
The list of patients with type 2 diabetes was extracted from our electronic medical records system by selecting patients who had an International Classification of Disease code for type 2 diabetes (ICD code: e11xx) and a billing code of abdominal CT scan. Non-inclusion criteria were as follows: CT scans performed without contrast media injection, or CT scans acquired only after injection of contrast media (hence lacking images with or without contrast, respectively). 98 patients out of 223 were pre-selected at random in order to retain enough patients after exclusion.
Exclusion criteria were as follows: type 1 diabetes or other causes of diabetes mellitus (MODY, cystic fibrosis…); known pancreatic disease (acute pancreatitis in the 6 months prior to the CT scan, chronic pancreatitis, pancreatic adenocarcinoma, history of pancreatic surgery); liver or biliary disease (hepatocellular carcinoma, advanced cirrhosis, cholangiocarcinoma, biliary prosthesis); cardiorespiratory arrest or multi-visceral failure in the month prior or in the week following the CT scan; lack of recent medical reports (maximum 2 years). After exclusion, 48 patients with type 2 diabetes were finally included in the study.
Control subjects were selected from the list of patients for whom an abdominal CT before and after iodinated contrast media intravenous injection had been performed from January 1st 2016 to December 31st 2017 in our hospital. The total number of potentially eligible control patients was 3267. For each included patient with type 2 diabetes, a control subject was selected. The control subjects were matched with the patients for gender and for age ± 1 year.
For each patient with T2D, a first potentially eligible control subject was picked at random. The medical records of the eligible control subjects were then assessed in order to check for the presence or absence of exclusion criteria. If the first eligible control subject was excluded, the records of another potentially eligible control were assessed, etc. 177 potentially eligible control subjects were excluded (Additional file 1: Table S1).
The exclusion criteria were the same for the patients and for the control subjects. A supplementary exclusion criterion for control subjects was diabetes of any kind. The medical files of all control subjects were checked to ensure that they did not have diabetes, i.e. that they did not have a history of fasting hyperglycemia (> 7 mmol/L) or of hyperglycemia at any time (> 11 mmol/L) and that they did not take any antidiabetic medication.
Processes, interventions and comparisons
Clinical and biological characteristics
All clinical and biological data were extracted from the subjects’ electronic medical records.
The presence or absence of hypertension, dyslipidemia, and coronary disease was recorded for all the subjects. Hypertension was deemed present if a history of hypertension was noted in the patient’s medical record AND/OR if the patient was taking any antihypertensive medication.
Dyslipidemia was deemed present if a history of dyslipidemia, hypercholesterolemia or hypertriglyceridemia was noted in the patient’s medical record AND/OR if the patient was taking statin or fibrate medication AND/OR if the patient had a known history of LDL cholesterol concentration > 4.95 mmol/L or a known history of serum triglyceride concentration > 1.71 mmol/L.
Coronary disease was deemed present if a history of coronary disease, myocardial infarction, coronary angioplasty or coronary bypass surgery was noted in the patient’s medical record.
Serum creatinine concentration was recorded. Estimated glomerular filtration rate (eGFR) was calculated using MDRD formula [19]. Renal failure was defined as an eGFR < 60/mL/min/1.73 m2. Severe renal failure was defined as an eGFR < 30 mL/min/1.73 m2.
For the patients with type 2 diabetes, the presence or absence of peripheral neuropathy, nephropathy and retinopathy was recorded. Peripheral neuropathy was deemed present if a history of peripheral neuropathy was noted as such in the patient’s medical record, AND/OR if lower limbs paresthesia, lower limb neuropathic pain, or abnormalities at the monofilament testing were noted in the patient’s medical record.
Nephropathy was deemed present if a history of diabetic nephropathy was noted as such in the patient’s medical record AND/OR if there was a history of microalbuminuria > 3 mg/mmol of creatininuria or a history of proteinuria, or if the MDRD glomerular filtration rate was < 60 mL/min without another known cause for renal dysfunction.
Retinopathy was deemed present if a history of diabetic retinopathy was noted as such in the patient’s medical record AND/OR if there was a history of macular edema or of retinal photocoagulation.
Computed tomography
All the patients had abdominopelvic computed tomography (CT) using a Somatom Sensation 64® (Siemens Healthcare, Forchheim Germany). The following scanner parameters were used: 279–450 mm field of view, 38.4 mm beam collimation (64 × 0.6 mm collimator setting), 120 peak kVp tube potential, 0.5–0.8 s gantry revolution time and 46 mm per gantry rotation table speed resulting in a beam pitch of 1.2 and 310–500 mm field of view, and 55 mm per gantry rotation table speed resulting in a beam pitch of 1.38. Online, real time, anatomy-adapted, attenuation-based tube current modulation techniques (Care Dose 4D®, Siemens Medical Solution) were used with a tube current set to 120–170 effective mAs. At the start of the procedure, 120 mL of non-ionic iodinated contrast material (Iomeprol, Iomeron®, Bracco Imaging SpA, Milan, Italy or iopamidol, Iopamiron®, Guerbet, Roissy-Charles de Gaulle, France) were injected intravenously through a 20-Gauge catheter into an antecubital vein, at the rate of 3 mL/s by using an automated power injector. Two pass imaging sets were obtained 25 s and 70 s after the start of the contrast material administration. All CT examinations were performed from the hepatic dome to the lower margin of the symphysis pubis, using a cephalocaudal direction after breath hold instruction was given.
After acquisition, CT data were reconstructed at 1 mm thickness at 0.5 mm intervals for transverse and multiplanar reconstructions and 3D imaging. All data were stored on internal picture archiving and communication system (PACS, Directview, V12.1.5.1156, Carestream Health Inc., Rochester, NY, USA).
Pancreas volume
For all patients, transverse CT images were analyzed along with multiplanar and 3D images using the PACS workstation. Pancreas volume calculation (Fig. 1) was performed on venous phase CT images using a combination of contour drawing, thresholding and region growing. Interpolation between the marked slices was performed semi-automatically using a linear algorithm [20, 21]. The splenic, mesenteric and hepatic arteries, celiac trunk, splenic vein, and the superior mesenteric vein were excluded from pancreatic segmentation.
Vessels assessment
The vessel assessment (Fig. 2) was performed on arterial phase CT images. The diameter of the splenic, hepatic and gastroduodenal arteries was measured one centimeter after their origin. The presence and the total number of visible pancreas-bound branches emerging from the splenic artery or sometimes from the celiac trunk or from the superior mesenteric artery in case of anatomical variations (i.e. the greater pancreatic artery, the dorsal pancreatic artery and the inferior pancreatic artery) were noted (from 0 to 3 visible branches). Their diameters were measured one centimeter after their origin. Simultaneously, the identification of intrapancreatic first- and second-order vessel division was performed.
Calcifications
Calcifications in the splenic artery and abdominal aorta walls were evaluated on the images without contrast and classified according to the following score:
0: no calcification; 1: scarce interspersed calcifications; 2: linear calcifications with intervals of normal artery; 3: circumferential calcifications for the abdominal aorta, and continuous linear calcifications for the splenic artery (Fig. 3).
Pancreas and liver density
Pancreas and liver density were quantified on the images without contrast. The density of the pancreas, in Hounsfield Units (HU), was defined as the mean density of 3 Regions Of Interest (ROI) delineated in the head, body and tail of the pancreas, respectively. Each ROI was carefully delineated in order to exclude vessels and calcifications.
Interobserver reliability
All CT images analyses were performed by a unique investigator blinded to the diabetes or control status of the subjects. To validate our image analysis methods, two different sets of randomly selected patients were quantified by a second, independent investigator, who was also blinded to the diabetes or control status of the subjects.
The characteristics of the pancreas (pancreas density and volume) were validated in a subgroup of 21 random subjects (11 patients with type 2 diabetes and 10 control subjects).
The characteristics of the vessels (abdominal aorta and splenic artery calcium scores, splenic artery diameter, number of pancreas-bound arteries and of intrapancreatic arterial subdivisions) were validated in another subgroup of 22 random subjects (11 patients with type 2 diabetes and 11 control subjects).
Duration of diabetes
In order to know if the characteristics of the pancreas observed in patients with type 2 diabetes correlated with the duration of diabetes, we divided the patients in whom the duration of diabetes was known into two subgroups: the patients of the first group had a duration of diabetes of 0 to 10 years, and the patients of the second group had a duration of diabetes of more than 10 years. Age at first diagnosis of diabetes, from which diabetes duration was determined, was established according to patients and/or physician reports, noted in the medical records of the patients. We then compared the characteristics of the pancreas and its vessels between the two subgroups.
Statistical analysis
Data are presented as mean with standard deviation (SD) for the normally distributed variables, and as median with interquartile range (IQR) for the variables with non-normal distribution. For continuous variables, t-tests (paired or unpaired) were used when 2 groups of normally distributed variables were compared, Mann-Whitney test was used when 2 groups of unpaired variables with non-normal distribution were compared, Wilcoxon matched-pair signed rank test was used when 2 groups of paired variables with non-normal distribution were compared. ANOVA was used when more than 2 groups of normally distributed variables were compared, Kruskal–Wallis comparison was used when more than 2 groups of variables with non-normal distribution were compared.
For categorical variables, McNemar test was used when 2 paired groups were compared, Fisher’s exact test was used when 2 unpaired groups were compared, and Chi Square was used when more than 2 groups were compared.
After performing a log likelihood ratio comparison of different models, we tested if there was an independent association between the presence or absence of type 2 diabetes and pancreas and liver characteristics using multivariate logistic regression with adjustment for age, gender, BMI and the presence or absence of hypertension. We did the same to test the association between a duration of diabetes of less or more than 10 years and pancreas and liver characteristics.
Interobserver reliability was assessed using intraclass correlation coefficient for continuous variables, and Cohen’s weighted kappa for categorical variables.
Analyses were performed using Graphpad Prism 5 (GraphPad Software, Inc., San Diego, CA) and the Real Statistics Resource Pack software (Release 6.8), copyright (2013–2020) (www.real-statistics.com).