Clinical and Imaging Predictors for the Development of Diabetes Mellitus Following a Single Episode of Acute Pancreatitis in Youth

Abstract

Background:

Acute pancreatitis (AP) increases the risk of diabetes mellitus (DM). Our aim was to identify clinical, laboratory and imaging predictors of preDM/DM in youth_post index AP.

Methods:

This was a prospective cohort study of patients ≤21 years-old with an index admission for AP and follow up at 3 and/or 12 months. Clinical laboratory values, imaging findings, admission course, and plasma chemokine and cytokine measures collected at index admission were tested for association with preDM/DM development. A multivariable regression model was used to predict preDM/DM.

Results:

Among 187 enrolled participants, 137 (73%) and 144 (77%) underwent DM screening at 3 and 12 months respectively, and 137 (73%) had imaging available. PreDM/DM occurred in 22/137 (16%; preDM n=21, DM n=1) at 3 months and 23/144 (16%; preDM n=18, DM n=5) participants at 12 months.

Univariate associations with preDM/DM at 12 months included: severe AP (SAP) (52% preDM/DM vs. 17% no DM; p=0.0008), median [IQR] IL-6 (910 pg/ml [618–3438] vs. 196 pg/ml [71–480], p<0.05) and CRP (4.16 mg/L [1.67–10.7] vs. 1.55 mg/L [0.4–3.68], p=0.1) at time of AP attack.

The optimal multivariable model to predict preDM/DM included with clinical variables was severe acute pancreatitis (SAP), c reactive protein (CRP), interleukin-6 (IL-6), and age [AUC=0.80; (0.70, 0.88)]. Including imaging markers, the ideal model included SAP, CRP, IL-6, subcutaneous fat area, age and presence of autoimmune disease with an AUC [0.82 (0.71, 0.90)].

Conclusions:

Development of preDM/DM following an index AP episode can be predicted by baseline AP severity, baseline CRP, IL-6 levels, and subcutaneous fat area.

Introduction

Acute pancreatitis (AP) is an increasingly recognized in the pediatric patients, with an estimated incidence of approximately 1:10,000 [1]. While the initial phase of AP in youth is typically mild, long-term sequelae remain unknown, including diabetes mellitus (DM)) [1, 2]. DM is defined by the American Diabetes Associated (ADA), as having one of the following: (1) glycosylated hemoglobin (HgA1c) ≥6.5% (≥48 mmol/mol), (2) fasting plasma glucose ≥ 126 mg/dL (≥7.0 mmol/L), (3) 2-hour plasma glucose ≥ 200 mg/dL (≥11.1 mmol/L) during an oral glucose tolerance test (OGTT) (4) random plasma glucose of ≥ 200 mg/dl (≥11.1 mmol/L) in conjunction with hyperglycemic symptoms [3]. Pre-DM is diagnosed with one of the following: (1) HgA1c 5.7–6.4% (2) fasting plasma glucose 100 mg/dl (5.6 mmol/L) – 125 mg/dL (6.9 mmol/L) (3) 2-hour plasma glucose during 75 g OGTT (4) 140 mg/dl (7.8 mmol/L) to 199 mg/dL (11.0 mmol/L) [3] DM in its early stages is clinically silent disease despite having a high morbidity and mortality. Thus, early identification is critical to allows for appropriate resource allocation and clinical management.

Acute recurrent pancreatitis (ARP) and chronic pancreatitis (CP), as defined by the International Study Group of Pediatric Pancreatitis: In Search for a Cure (INSPPIRE), are known risk factors for preDM/DM (also termed pancreatic endocrine insufficiency) in youth, [5–7]. In adult studies, the risk of preDM/DM following a first time (index) AP episode is approximately 20%; up to 60% will have an abnormal oral glucose tolerance testing (OGTT) [8–10]. Youth have significantly different etiologies for AP, including drugs, metabolic, viral/systemic illness, traumatic, genetic, anatomic, and biliary, compared to biliary and alcohol in adults [11–13]. Thus, the risk factors for preDM/DM development following AP cannot be extrapolated from adult literature. For AP in youth, one prior study reported that preDM/DM occurred in 15% after a single AP episode, 46% after a second AP episode, and 38% after onset of CP, demonstrating some increased risk in youth following even a single episode of AP, with severe AP (SAP) and obesity as identified risk factors [6]. Clinical and role of inflammatory signaling factors that put the patients at risk for preDM/DM after AP are poorly understood.

Adult studies show a bidirectional relation between obesity and severe AP [14]. Further, obesity can lead to relative skeletal muscle loss (sarcopenia) due to the lipotoxic effect on myocytes. [15–17]. Body composition parameters, including subcutaneous fat mass and sarcopenia can be derived from cross-sectional images [e.g., abdominal computed tomography (CT) or magnetic resonance imaging (MRI)], with several studies showing prognostic value of multiple parameters for SAP [18, 19]. With pediatric obesity rates now on the rise in the United States, there is an increasing need to understand the impact of obesity, muscle composition and sarcopenia on AP in youth [20, 21]. To our knowledge, there are no prior studies exploring the link between AP severity and preDM/DM following AP with image derived body composition parameters. Thus, risk models for DM development using imaging parameters constitute a huge knowledge gap.

In this single center prospective cohort study, we followed youth following an index episode of AP at 3 months and 12 months for development of preDM/DM and hypothesized that there is an increased risk of preDM/DM in patients with certain clinical, inflammatory markers and imaging characteristics. Our aim was to evaluate clinical, chemokines/cytokine signals and imaging data involved in the index AP attack and to develop risk models for predicting preDM/DM development.

Methods

Patient Enrollment

Patients 21 years of age and younger with first time (index) AP episode presenting to Cincinnati Children’s Hospital Medical Centre (CCHMC, Cincinnati, Ohio USA) were enrolled into an institutional review board (ethics) approved prospective registry AP database [CCHMC IRB 2012–4050]. Informed consent was obtained from all patients and/or parents/guardians. AP diagnosis was made based on previous published criteria from The INSPPIRE consortium: 2 of the following 3 criteria: abdominal pain, serum amylase and/or lipase ≥3 times upper limit of normal, and imaging findings consistent with AP [7]. SAP in our cohort included the moderately severe and severe AP cases as per the NASPGHAN classification [22]. PreDM/DM diagnosis was made in any one criteria set forth by the American Diabetes Association was met [3]. The AP database included clinical, laboratory imaging data starting from the index AP case and demographics and glucose measures from 3- and 12-month follow ups, recorded via REDCap (Research Electronic Data Capture, Nashville, Tennessee, USA). Our patient cohort was prospectively enrolled in the AP registry between March 2012 to February 2023. Data from the AP database, including the patients in this study, have been previously reported, however those studies had different goals, hypotheses, and objectives [23, 24].

Laboratory methods:

Initial blood samples were obtained within 48 hours of admission for the index AP attack, with the majority (80%) obtained within 24-hours of presentation. Plasma was extracted, aliquoted and stored at −80° C.

Chemokines/Cytokines Assays

This study contained two data sets, one from a prior publication [23], with chemokines assay obtained via Luminex assays. Luminex assays were performed by the Hummune Immune Monitoring Centre (HIMC) at Stanford University (http://iti.stanford.edu.stanford.idm.oclc.org/himc/protocols.html). Human 62-plex Procarta kits were purchased from eBioscienses/Affymetrix/Thermo Fisher (Santa Clara, California, USA) and used per manufacturer’s recommendations with the following modifications. Beads were added to a 96 well plate and washed in Biotek ELx405 washer. Sample containing mixed antibody-linked beads were added to the plate and incubated at room temperature (RT) for 1 hour. They then underwent overnight incubation with shaking (orbital shaker 500–600 rpm) at 4°C. Plates were then washed in Biotek ELx405 washer. Biotinylated detection antibody was added at RT with shaking for 75 minutes. The plates were then washed with Biotek ELx405 washer and streptavidin-PE. They were then incubated for 30 minutes at RT, followed by washing (as mentioned above) and a reading buffer was added to the well. Each sample was measured twice. Plates were read utilizing a Luminex 200 or FM3D FlexMap instrument with a lower bound of 50 beads per sample per cytokine. Custom Assay Chex control beads were purchased from Radix Biosolutions (Georgetown, Texas, USA) and added to all wells.

The second data set was performed via enzyme-linked immunosorbent assay (ELISA), which was used to detect Interlukin-6 (IL-6) and Monocyte Chemotactic Protein-1 (MCP-1), in subsequently enrolled patients, as they were the most significantly elevated chemokines in severe courses of the index AP. Using the R&D Systems D6050 kit for IL-6 and then R&D System DCP00 Kit for MCP-1, the plasma samples collected during the index admission underwent IL-6 and MCP-1 ELISA analysis. Briefly, all the working standards and reagents were prepared as directed by the kit. The samples were thawed at RT.

For IL-6, 100 μl of assay diluent was added to each well, followed by 100 μl of standard, control, or sample. Samples were then incubated for 2 hours at RT. Each well was then aspirated and washed for a total of four washes, ensuring complete removal of liquid. After the final wash, 200 μl of human IL-6 conjugate was added to each well, and then covered with an adhesive strip and incubated for 2 hours at RT. The wash steps were repeated, and 200 μl of substrate solution was added to each well. Samples were then incubated in the dark for 30 minutes at RT. Subsequently 50 μl of stop solution was added to each well as the color changed from blue to yellow. Determination of the density of each well occurred within 30 minutes, using a microplate reader set to 450 nm.

For MCP-1 50 μl of assay diluent was added to each well, followed by 200 μl of standard, control, or sample to their respective wells. Samples were incubated for 2 hours at RT. Each well was aspirated and washed with 400 μl of wash buffer, for a total of four washes, ensuring complete removal of liquid. After the final wash, 200 μl of human MCP-1 conjugate was added to each well, and then covered with an adhesive strip and incubated for 2 hours at RT. The wash steps were repeated, and 200 μl of substrate solution was added to each well. Samples were then incubated in the dark for 30 minutes at RT. Subsequently, 50 μl of stop solution was added to each well as the color changed from blue to yellow.

C Reactive Protein (CRP) Assay

CRP assay was performed with the Simens Atellica Analyzer. This utilized a CH C Reaction Proteion_2 (CRP_2) latex reagent which was composed of a suspension of uniform polystyrene latex particles with anti-CRP antibody. If a sample contained CRP is agglutinated with the reagent, resulting in an increase in turbidity which was measured at 571 nm. The final CRP concentration was determined from a generated calibration curve.

Imaging data

The closest CT or MRI examination within 90 days of the index AP admission was selected for review. A single pediatric radiologist (N.A.A., 2 years of experience) reviewed the examinations to identify the L3/L4 disk level. Using the identified L3/L4 disk level a research fellow (P.D.) manually segmented psoas muscle cross sectional area and subcutaneous fat area. Specifically, a geometric region of interest measurement tool (Intellispace; Philips Healthcare) was used to trace the peripheral margins of the left and right psoas muscles and the inner and outer boundaries of the subcutaneous fat defined as the outer surface of the abdominal wall muscles and the skin surface, respectively. A board-certified pediatric radiologist (A.T.T. 11 years of experience) reviewed all segmentations prior to data entry. Figure 1 illustrates the measurements made on an MRI of a patient.

Figure 1:

Axial computed tomography (CT) image from a 17-year-old male with severe acute pancreatitis. Yellow highlight indicates subcutaneous fat area and red highlight indicates psoas muscle area. This patient has a relatively high subcutaneous fat standardized to height (52.1 cm²/m²) but low psoas muscle z score (−2.68). The patient went on to develop preDM at follow-up.

Measured total psoas muscle area (PSMA) was entered into an online tool (Pediatric tPMA z-score calculator, The Hospital for Sick Children, Toronto, Ontario) to generate a define a PSMA z-score for each patient, which takes into account patient’s sex and age [25]. The subcutaneous fat area was standardized to the square of patient heights (cm²/m²).

Biostatistics

The primary analysis was to identify important risk factors for preDM/DM in 12 months follow-up. We tabulated and summarized patient’s characteristics, baseline lab measures, and imaging data. Continuous variables were reported in median with interquartile range (IQR) and tested by Wilcoxon rank sum tests. Categorical variables were tested by Fisher’s exact tests or Chi-square test (where relevant). Pearson correlations were calculated between BMI percentiles and segmentation measurements from imaging along with 95% confidence intervals (CI) and p values. Strength of correlations was classified as follows: 0–0.19, very weak; 0.2–0.39, weak; 0.40–0.59, moderate; 0.60–0.79, strong; and 0.80–1.0, very strong [26]. Patients who had missing primary outcomes were excluded from analysis. We examined covariates missing pattern and validated the missing at random assumption by comparing all available baseline risk factors between patients who had missing observations or not. For each biomarker or imaging feature with missing values, we created an indicator which takes the value 0 if the biomarker/imaging feature is missing and value 1 otherwise. We included indicator and interaction terms of indicator with original biomarker/imaging features in model to account for the potential impact of missing values on the relationship between outcomes and risk factors. Data were randomly split into training and test cohort in an 8:2 fashion. We developed a statistical learning process and built multivariable logistic regression models on the training data. We initiation considered variables with a p<0.2 in combination with variables from prior pediatric studies that have significantly associated with preDM/DM development (i.e., older age, sex, and BMI) [6]. Notable, while subcutaneous fat area was not significantly different in patients with preDM/DM it was moderately correlated with BMI percentile (r=0.62), which is known to have a significant role in preDM/DM development, and thus was included in the analysis. We selected the important risk factors and generated optimal based on area under the receiver operating characteristic (AUROC) curve. Test cohort was used to validate the model performance and stability. All analyses were performed using SAS Version 9.4.

Results

Clinical Factors and Demographics

A total of 187 subjects completed follow-up visits at 3 months and/or 12 months. Imaging results within 90 days of index AP were available for 73% (137/187). At 3 months, 73% (137/187) patients had follow-up with DM screening; 15% (21/137) had preDM, and 1% (1/187) had DM (Table 1). At 12 months, 77% (144/187) had follow up with DM screening, of whom 13% (18/144) had preDM, and 3% (5/144) had DM. Of the 18 patients with preDM at 12 months, 11 had normal glucose metabolism values, 3 had preDM, and 4 had no available follow-up data at 3 months. Of the 5 patients with DM at 12 months, 1 had normoglycemia, 1 had preDM, 1 had DM, and 2 had no data at 3 months. PreDM/DM at any time point occurred in 21% (40/187) of subjects (preDM n=35, DM n=5).

Table 1.

Comparison of demographics and clinical features of study participants who did or did not develop pre-diabetes mellitus or diabetes mellitus (preDM/DM) during the follow up period of 12 months.

Patient characteristic	Developed preDM/DM in 12 months % (count)		P-value a
	No (n = 121)	Yes (n = 23)
PreDM/DM in 3 Months			0.1323
Normal	53.7 % (n = 65)	52.2 % (n = 12)
PreDM	9.9 % (n = 12)	17.4 % (n = 4)
DM	0.0 % (n = 0)	43 % (n = 1)
Unknown	36.4 % (n = 44)	26.1 %(n = 6)
Severe AP			0.0008
No	82.6 % (n = 100)	47.8 %(n = 11)
Yes	17.4 % (n = 21)	52.2 % (n = 12)
Age, in year b	12.6 [8.4, 15.6] (n = 121)	13.9 [9.4, 15.9](n = 23)	0.4937
Sex			0.3632
Male	47.9 % (n = 58)	60.9 % (n = 14)
Female	52.1 % (n = 63)	39.1 %(n = 9)
BMI percentile b	75.0 [31.1. 94.1]	69.9 [38.8. 96.6]	0.6906
	(n= 121)	(n = 23)
Weight Status			0.7016
<85 %	58.7 %(n = 71)	56.5 % (n = 13)
85 % − <95 % (Overweight)	18.2 %(n = 22)	13.0 %(n = 3)
≥95 % (Obesity)	23.1 % (n = 28)	30.4 % (n = 7)
Family history of AP			0.5165
Yes	10.9 % (n = 16)	10.0 % (n = 4)
No	72.1 % (n = 106)	65.0 % (n = 26)
Unknown	17.0 % (n = 25)	25 % (n = 10)
Etiology of AP
Idiopathic	32.7 % (n = 48)	25.0 % (n = 10)	0.3956
Toxic/Metabolic/Autoimmune/ Drugs	31.3 % (n = 46)	42.5 % (n = 17)	0.1836
Obstructive	21.8 % (n = 32)	17.5 % (n = 7)	0.5557
Traumatic/post-ERCP	6.8 % (n = 10)	2.5 % (n = 1)	0.3052
Insulin used during attack			0.5055
No	97.5 % (n = 118)	95.7 % (n = 22)
Yes	2.5 % (n = 3)	43 % (n = 1)
ARP			0.8012
No	72.7 % (n = 88)	69.6 % (n = 16)
Yes	27.3 % (n = 33)	30.4 % (n = 7)

DM = Diabetes Mellitus; AP = Acute Pancreatitis: Body mass index = BMI: ARP = Acute Recurrent Pancreatitis.

^aexact test or Chi-square tests were used for categorical variables. Wilcoxon rank sum tests were used for continuous variables. Fisher’s.

^bContinuous variables presented as medians and interquartile ranges.

Patients were divided into two groups: no preDM/DM at 12 months (normoglycemia) and preDM/DM at 3 and 12 months (Table 1). At 3 months only severity of the index AP was found to be significant with moderate and severe AP in 17% vs. 40% (p=0.0052) in no DM vs. preDM/DM respectively. The frequency of SAP was significantly hher (52% vs. 17%) in patients with preDM/DM vs. no DM (p=0.0008). There were no significant differences in age, sex, BMI percentile, or ARP frequency between patient groups. Additionally, a linear mixed model showed statistical significance in BMI change from baseline through 3 months (p=0.85) and 12 months (p=0.9) follow ups. Insulin use during the index AP, indicating significant hyperglycemia, was not significantly different, with 2.3% and 4.2% in no DM vs preDM/DM groups respectively. At initial AP presentation, significantly higher levels of IL-6 (pg/ml) (preDM/DM: 910 [IQR: 618–3438] vs. no DM: 196 [71–480], p=0.004) were seen in patients with preDM/DM compared to patients without DM. CRP was also higher in the preDM/DM group but not statistically significant (p=0.11). The frequency of laboratory values in each group is shown in Table 2.

Table 2.

Chemokines, Cytokines and Imaging Biomarkers of study participants who did or did not develop pre-diabetes mellitus or diabetes mellitus (preDM/DM) during the follow up period of 12 months.

Chemokines/imaging markers	Developed preDM/DM in 12 months median [IQR]		P-value a
	No (n = 121)	Yes(n = 23)
IL-6 (pg/ml)	196 [71. 480] (n = 24)	910 [618, 3438] (n = 6)	0.0040
CRP (mg/dL)	1.55 [0.40, 3.68] (n = 62)	4.16 [1.67, 10.70] (n = 9)	0.1115
BUN mg/dL	11.0 [8.0, 15.0] (n = 111)	10.5 [7.5, 18.0] (n = 20)	0.9387
MCP-1 (pg/mL)	1471 [450, 2766] (n = 25)	3438 [840, 4509] (n = 6)	0.3295
MMP-9 (ng/mL)	66.002 [25,505. 165,640] (n = 22)	149,463 [37,082, 418,425] (n = 5)	0.3031
PSMA Z-score	−1.14 [−2.11, −0.20] (n = 60)	−1.60 [−2.68, −0.74] (n = 17)	0.1748
Subcutaneous fat area standardized to height (cm2/m2)	44.6 [26.3, 78.3] (n = 55)	52.1 [25.9, 96.7] (n = 13)	0.8515

IQR = Interquartile range; 1L-6 = lnterleukin-6; CRP = C Reactive Protein; BUN = Blood Urea Nitrogen; MCP-1 = Monocyte Chemoattractant Protein-1; PSMA = Psoas Muscle Area; MMP-9 = Matrix Metalloproteinase-9.

^aWilcoxon rank sum tests were used for continuous variables. Chi-square tests were used for categorical variables.

Imaging

BMI percentile was weakly correlated with PSMA z-score (r=0.27; 95% CI: 0.09 to 0.54; p=0.0039) and moderately correlated with subcutaneous fat area standardized to height (r=0.62, 95% CI: 0.49 to 0.72; p<0.0001). There was no statistically significant correlation between PSMA z-score and standardized subcutaneous fat (r=0.13, 95% CI: −0.20 to 0.44; p=0.441). Supplemental Figure 1 illustrates scatter diagrams showing the correlation between the variables. There were no significant differences observed in the PSMA Z-score (p=0.1748) and subcutaneous fat area standardized to height (p=0.8515) between patients with and without preDM/DM (Table 2).

Univariate Analysis

At 3-month follow up, only AP severity was found to be significantly associated with preDM/DM development (p=0.0052). At 12 months both AP severity (p=0.0008) and baseline IL-6 (p=0.0040) were significantly associated with preDM/DM (Table 2). While not significant, baseline CRP and PSMA z-score both had a p<0.2, suggesting some predictive performance. While age (p=0.268), and BMI percentile (p=0.8061) were not significant, they were considered as potential predictors in the multivariable regression due to prior studies showing significant relationship with post AP preDM/DM development.

Multivariable Regression

A statistical learning process was developed using variables with a p<0.2 in combination with variables from prior pediatric studies that have significantly associated with preDM/DM development (i.e, older age, sex, and BMI) [6]. Notably, while subcutaneous fat area was not significantly different in patients with preDM/DM it was moderately correlated with BMI percentile, which is known to have a significant role in preDM/DM development, and thus was included in the analysis., First a clinical, cytokine and lab model was developed, which showed optimal performance for predicting preDM/DM at 12 months with AP severity, CRP, IL-6, and age included, with a combined training and test dataset AUC of 0.80 (Figure 2). When imaging variables were included, the optimal model included AP severity, CRP, IL-6, and standard subcutaneous fat, with combined AUC of 0.80 (Figure 2). All other variables were not included as they did not improve the model’s performance or AUC.

Figure 2:

Area under the receive operater curves (AUROC) of multivariable prediction model for preDM/DM with curves of the univariate variables: Panel A: Clincial multivariable model including: severe acute pancreatitis (SAP), C Reactive Protein (CRP), Interleukin-6 (IL-6) and age. AUROC of predictive model is 0.8. Panel B: Multivariable model including imaging: SAP, CRP, IL-6 and most proximal subcutaneous fat area standardized to height (Sub-Q fat) with final model AUROC 0.8. Both AUROC of these predictive models are better than the AUROC of each individual parameter alone.

Discussion

This study reports novel data on predictors for the development of preDM/DM following an index AP attack in a prospective cohort of youth. We found that AP severity, baseline CRP, baseline IL-6, age, and subcutaneous fat are predictive of risk for development of preDM/DM over 12 months. Specifically, we generated two predictive models, one that could be applied when imaging data is available, and one that could be applied when only clinical data is available. Our predictive model may have a role during the index AP attack, which is when the initial insult that leads to the development of preDM/DM is thought to occur). Application of our model may help with planning follow up and future testing of specific patients who are at higher risk of developing preDM/DM.

Pancreatogenic DM (i.e., Type 3C DM) is hypothesized to occur due to pancreatic exocrine tissue diseases (e.g., AP) [5]. Specifically, due to the anatomical proximity of exocrine (acinar and ductal cells) and endocrine cells (islets of Langerhans), exocrine cell inflammation and atrophy leads to scarring and irreversible, islet cell loss and beta cell dysfunction, and consequently dysglycemia [3, 27]. There is a well-established relationship with pancreatogenic DM and ARP and CP, with a prevalence of 4–9% of pancreatogenic DM compared to 0.25% of any form of DM in youth [28]. One study evaluating tissue samples resected from adult patients with CP showed that patients with pancreatogenic DM have significantly increased pancreatic cytokines, making the hypothesis that inflammation may be the inciting event of B-cell dysfunction prior to frank loss of islet tissue [29].

Pancreatic polypeptide (PP) is produced and secreted from islet cells post-prandially, and functions to inhibit gastric emptying and exocrine pancreatic secretions. Multiple cross-sectional studies have shown that PP is deficient in patients with pancreatogenic DM, and positively correlated with hepatic insulin sensitivity [5, 30]. While pancreatogenic DM can occur after any episode of AP, SAP is correlated to pancreatic cell damage, and consequently islet cell damage with resultant decreased PP production, and increased risk of pancreatogenic DM. Other risk factors include exocrine insufficiency, pancreatic atrophy, and obesity [5]. While we did not measure PP in our patient cohort, our models reinforce this relationship of pancreatogenic exocrine cell damage correlated with hepatic insulin sensitivity.

Our results are similar to prior work in youth with AP, ARP and CP that showed increased prevalence of preDM/DM, and identified obesity, male sex and SAP to be associated risk factors for abnormal glucose testing [6]. We were able to demonstrate that laboratory markers of inflammation (i.e., CRP, IL-6) have a significant predictive value, and that dysglycemia is not exclusively a transient event, and can be sustained for at least 12 months.

Prior adult literature has shown imaging derived body composition measures of muscle and fat mass had increased predictive value for mortality and severity in AP when compared to weight or BMI [18, 31–34]. Specifically, sarcopenia and sarcopenic obesity were found to be independent predictors for development of SAP [17]. Despite patients with preDM/DM having a higher median of PSMA z-score in our study, there was not a statistically significant difference (p=0.1748) compared to normoglycemic patients. This may be secondary to the fact that sarcopenic obesity (i.e., low muscle mass and high subcutaneous fat) is only seen in a small portion of our cohort (shown in the bottom right quarter of the scatter quadrant in Supplemental Figure 1C). Although PSMA z-score was not included in our final multivariable model, we do show standardized subcutaneous fat to predict preDM/DM in youth with AP.

Our study has some limitations. All prospective data was obtained from a single pediatric tertiary care center, meaning a validation cohort, perhaps from a multi-center source, will be needed in the future. Secondly, while this is the largest known prospective cohort of youth with AP, the small sample size may have limited the understanding of contribution of some variables, specifically PSMA z-score and standard subcutaneous fat area.

In conclusion, we report a predictive model for development of preDM/DM following an index AP episode using AP severity, CRP, IL-6, age, and standard subcutaneous fat as determinants. Further studies are needed to validate and optimize our model for preDM/DM, as well as investigate the application of this model in both larger pediatric and separate adult AP cohorts.

What is Known:

• Youth with acute recurrent and chronic pancreatitis are at increased risk of Diabetes mellitus (DM).

• After the initial AP episode, risk of DM increases in adults.

What is New:

• This is a prospective study identifying predictors of preDM/DM in youth with index AP.

• Clinical predictors include AP severity, C Reactive Protein, and Interleukin-6 levels at initial AP presentation.

• The subcutaneous fat area in imaging has also been shown to predict pre-DM/DM.

Abbreviations:

AP

acute pancreatitis

ARP

acute recurrent pancreatitis

CP

chronic pancreatitis

DM

diabetes mellitus

PreDM

pre-Diabetes Mellitus