Rationale and Design for the Diabetes RElated to Acute Pancreatitis and Its Mechanisms Study: A Prospective Cohort Study From the Type 1 Diabetes in Acute Pancreatitis Consortium.

The DREAM study includes the 10 clinical centers (with at least 3 additional satellite sites) of the Type 1 Diabetes in Acute Pancreatitis Consortium (T1DAPC) from throughout the United States. Investigators have formed a combination of working groups and committees that worked collaboratively in an iterative manner to develop the DREAM study protocol and execution plans, as previously introduced. 12 In addition to these teams, the Data Coordinating Center provides central administrative, regulatory, and statistical support for T1DAPC, including DREAM. 20 In accordance with NIH policies, the study is approved through a single institutional review board, and reviewed on a biannual basis by an NIH-appointed Data and Safety Monitoring Board (DSMB).

The primary endpoint of the DREAM study is the incidence of DM following the qualifying episode of AP defined as the combination of both concurrent newly diagnosed DM (ie, initially diagnosed at the time of the qualifying AP episode) and new-onset DM (ie, diagnosed during longitudinal follow-up). 19 The chronology of DM in relation to the qualifying episode of AP and definition for both DM and pre-diabetes for the DREAM study are described in Table 4 . DM status will be serially assessed using the combination of fasting glucose and HbA1c, according to the American Diabetes Association diagnostic criteria. 19 Furthermore, to identify individuals with lack of concordance of tests for diagnosing DM, we will also measure a 2-hour postprandial glucose level (during OGTT) as a sensitive method of diagnosis for participants in the serial OGTT arm. When participants partially satisfy diagnostic criteria for DM, they will be asked to repeat a fasting plasma glucose and HbA1c during the same time window. We will also assess the development of either diabetes or prediabetes as a secondary endpoint. Lastly, the frequency, distribution, and temporal emergence of islet autoantibodies will be evaluated in relation to DM, as described elsewhere 14 . While the primary objective is to evaluate the incidence and mechanisms of DM, the study design affords the opportunity to investigate other aspects of the natural history of AP. Specifically, we will evaluate the readmission rate, incidence of exocrine pancreatic dysfunction (serially measured with a combination of symptoms and fecal elastase-1 levels), changes in quality of life, healthcare utilization, and the recurrence of AP and/or progression to chronic pancreatitis. Additional secondary endpoints related to metabolic and immunological mechanisms of diabetes as well as imaging findings are discussed elsewhere. 13 – 15

Acute pancreatitis (AP) is an inflammatory disease of the pancreas that accounts for more than 300,000 hospitalizations and health care costs exceeding $2 billion annually in the United States. 1 , 2 The incidence of AP is increasing in both North America and Europe. 3 Complications can develop during hospitalization or subsequently during follow-up, and include diabetes mellitus (DM), exocrine pancreatic insufficiency, complications from walled-off pancreatic necrosis, and recurrent episodes of AP. 4 – 6 Hyperglycemia was historically considered a transient phenomenon of AP during hospitalization; however, meta-analyses have reported that approximately one-quarter of AP patients will develop DM within three years of discharge, suggesting it is a relatively common long term complication. 7 , 8 The suboptimal rigor of previous research has made it challenging to precisely define the incidence rate, accurately identify risk factors, and understand the underlying mechanisms of this problem. Defining an accurate incidence rate requires prospective ascertainment of DM status over a sustained period of time, preferably with testing methodology that is both highly sensitive and specific. Passive or retrospective determination of the development of DM is likely to underestimate the true incidence, and extended follow-up periods are necessary to understand the cumulative risk. Understanding the risk, both in the short and long term, is necessary to guide patient education and surveillance recommendations and to identify those at highest risk who may benefit from preventative strategies. Prospective data collection helps to minimize various forms of bias, including the ascertainment of DM status. One recent prospective study of 152 AP subjects without pre-existing DM reported a cumulative incidence of DM of 7% at 1 year and 11% at 2 years, as well as the composite endpoint of DM or pre-DM in 43% at 2 years. 9 This likely represents the most precise risk estimate to date. However, this study included few patients with moderate or severe disease. Various risk factors have been identified in prior studies, including the development of severe AP (particularly with pancreatic necrosis), exocrine pancreatic insufficiency, and a non-biliary etiology, but simultaneous assessment of comprehensive risk factors, including canonical risk factors for type 2 DM (T2D) has not been previously performed. 8 , 10 Considering these limitations, screening for DM is not currently recommended in routine clinical practice following AP discharge. In addition to estimating the risk of DM following AP, there are a number of additional knowledge gaps. Importantly, the pathophysiology is incompletely understood. The development of DM in patients with mild, interstitial AP suggests the role of factors other than mere reduction of islet cell mass from physical destruction of the pancreatic parenchyma due to necrosis. A number of potential mechanisms of disease have been proposed, including localized inflammation leading to insulin resistance and beta-cell autoimmunity; however, they are primarily based on inferential associations from clinical observations in AP and/or other forms of pancreatogenic DM. 10 , 11 In light of these gaps, it is not currently possible to provide tailored recommendations for prevention or treatment of DM following AP. The National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) supported Type 1 Diabetes in Acute Pancreatitis Consortium (T1DAPC) was formed in 2020, and has developed and launched the D iabetes RE lated to A cute pancreatitis and its M echanisms (DREAM) study to address these knowledge gaps and related scientific objectives. 12 The current article describes the design of the DREAM study and, along with other articles in this issue, outlines a series of primary and ancillary studies to permit a spectrum of metabolic, immunologic, imaging, and other analyses to better understand DM following AP.

There are multiple knowledge gaps in the understanding of AP and its complications. In the DREAM study, we will directly address important clinical questions related to the development of DM following AP and its underlying mechanisms, including beta-cell autoimmunity. In addition to the lack of robust data on the incidence of diabetes following AP, there is limiting understanding of the phenotype and pathophysiology of the disease, which we aim to investigate. Important questions include the factors associated with and the frequency at which islet cell autoimmunity occurs following exocrine pancreas injury and whether the kinetics of immune response and resolution relate to subsequent clinical phenotype. Further, the natural history of beta cell function in those who do and do not develop diabetes is unknown. Detailed metabolic testing will provide refined estimates of the incidence of DM and characterize the type and trajectory of metabolic and endocrine alterations, including beta cell function and alterations in pancreatic and incretin hormone responses. Serial assessments of clinical, immunological, cross-sectional imaging, and other data elements will be combined with paired biological specimens to create a comprehensive dataset that can be used to assess additional questions related to the natural history of AP and other disease-related complications. One of the guiding principles in the design of the DREAM study has been to create a rigorous protocol that addresses limitations of previous research. Importantly, prospective assessment of DM status is critical to minimize biased estimates of case ascertainment. In addition to fasting glucose and HbA1c values, we are implementing the use of serial OGTT to provide a sensitive estimate of DM. By performing systematic screening, we will increase the accuracy of identifying incident DM and its timing in relation to the qualifying AP episode. Similarly, the prospective and systematic collection of comprehensive data elements will allow investigation of the potential influence of covariates, including, but not limited to: etiology of AP, clinical severity of AP, imaging severity of AP, beta-cell autoantibodies, exocrine pancreatic dysfunction, changes in pancreas volume, and alterations in the intestinal microbiome. The assessment of patient-related outcomes (eg, quality of life) has not been attempted in a study of this scale in the United States, and will help identify patient-centric outcomes warranting additional investigation. Lastly, the large sample size will allow us to conduct robust subgroup and sensitivity analyses to understand the various interactions between patient and disease-related factors. Proactive measures have been undertaken to protect the scientific rigor and enhance the feasibility of the DREAM study, utilizing the interdisciplinary team of investigators from the disciplines of gastroenterology, endocrinology, immunology, and biostatistics. First, considering the critical role of biospecimen analyses for the planned translational studies, we convened a Biospecimen Committee to develop study-specific biospecimen standard operating procedures. 18 This guidance is intended to minimize heterogeneity across the clinical centers due to confounding related to variations in biosample collection, processing, storage, and/or shipping. Next, the Protocol Committee developed a series of study-specific Manual of Procedures to provide explicit guidance for completion of study assessments, including case report forms, for the participating sites. This manual will be updated on an as needed basis to address common questions or recurring situations where clarifying information is needed. We sought guidance from T1DAPC investigators and the NIH-representatives participating in recent and ongoing studies in pancreatitis to refine, and when possible harmonize, study definitions, data elements, case report forms, and research methodology. 21 – 25 Lastly, we have convened a Recruitment and Retention Committee concurrently with study planning to provide guidance and feedback regarding implications of decisions on participant burden. 26 There is not a precedent in the United States for a prospective cohort study in AP of this scale, so we will prospectively monitor recruitment and retention rates, completion of study assessments, and collaboratively discuss any needs to alter the sample size, number, and/or frequency of study assessments to maintain feasibility. Final decisions regarding modifications to the study protocol will be made by the T1DAPC Steering Committee and approved by the NIDDK, DSMB, and single institutional review board before implementation. In summary, the DREAM study is a prospective cohort study of patients with one or more episode of AP at risk for developing DM. The primary goal of the DREAM study is to investigate the incidence and mechanisms of DM following AP. Study results will generate an evidence base that will provide clinicians with further guidance regarding the surveillance of DM following AP, and generate the data needed to further investigate tailored preventative and treatment strategies to improve the lives of our patients.

We will apply a discrete time-to-event regression analysis with interval censoring because we anticipate the occurrence of DM for most participants will be detected at the scheduled visits (eg, 3, 6, 12, 18, 24, and 36 months). We will enroll up to 1000 participants (including up to 800 without pre-existing DM who will undergo longitudinal follow-up and up to 200 with pre-existing DM who will only undergo baseline assessment) into the DREAM study. We are estimating cumulative dropout rates of 5%, 10%, 15%, 20%, 25%, and 30% at 6, 12, 18, 24, 30, and 36 months of follow-up, respectively. Assuming cumulative DM incidence rates of 4.5%, 9%, 12%, 15%, 18%, and 21% at 6, 12, 18, 24, 30, and 36 months of follow-up, respectively, the 95% confidence interval for a sample size of 800 participants without pre-existing DM will have limits of ± 4% at 24 months based on a simulation study with 1000 Monte Carlo samples; confidence intervals for different enrollment sizes are depicted in Supplemental Figure 1 . Analyses for all primary and secondary outcomes will be repeated using a broader definition of new-onset dysglycemia, which includes the primary statistical endpoint of DM as well as pre-DM. We will also investigate relevant clinical factors (eg, AP severity, AP etiology, baseline body mass index, etc.) as to whether they significantly affect the discrete-time hazard function by including them in the regression model. We will perform secondary analyses using: 1) the discrete time-to-events hazards regression model, and 2) nested case-control designs with a 1:4 allocation ratio of cases to controls. Such nested case-control study designs with 800 participants (allowing for 20% loss to follow-up at 24 months) will yield 80% statistical power with a two-sided 0.05 significance level test to detect an odds ratio of 1.80 for a binary risk factor that has 50% prevalence in the controls; alternate power calculations for different enrollment sizes are shown in Supplemental Figure 2 .

Supplemental Figure 1 . The 95% confidence intervals for estimating the incidence rate of DM in the Diabetes Related to Acute Pancreatitis and its Mechanisms (DREAM) Study across different sample size estimates according to a discrete time-to-event analysis. Supplemental Figure 2 . Sample sizes of enrollment in the Diabetes Related to Acute Pancreatitis and its Mechanisms (DREAM) Study needed to identify odds ratios for incident diabetes with 80% statistical power using a nested case-control study design with 1:4 allocation ratio.