Celiac Disease
Conditions
Brief summary
The goal of this observational study is to learn about an adult's chance of having celiac disease based on blood testing and symptoms. The main question it aims to answer is: Can a blood test and symptom information separate patients into 3 groups of low, intermediate, and high risk for celiac disease? Participants already being evaluated for celiac disease as part of regular medical care will answer online survey questions about symptoms and have laboratory data collected from charts. The investigators hypothesize that a clinical prediction model integrating clinical data with TTG-IgA antibody levels can accurately identify patients with celiac disease offering a personalized approach. The investigators anticipate this prediction model would classify patients into 3 risk groups for celiac disease: 1) Low likelihood (no further testing required), 2) Intermediate likelihood (biopsy required for confirmation), and 3) High likelihood (biopsy can be avoided based on the model's accuracy) thereby reserving endoscopy and biopsies for cases of intermediate probability to improve diagnosis, reduce invasive testing, increase patient focus, and decrease costs.
Detailed description
The specific aims of this study are to: 1) develop and validate a clinical prediction model for celiac disease probability (external validation will be performed by site and time), 2) evaluate the implementation potential of the model, and 3) pilot the model and determine its impact on patient experience.
Interventions
None listed
Sponsors
The Cleveland Clinic
National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK)
Study design
Observational model
COHORT
Time perspective
PROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
For Aim 1: Inclusion Criteria: * Patients ≥18 who underwent duodenal biopsy during upper endoscopy and had a TTG-IgA antibody test 3 months before or 1 month after biopsy
Exclusion criteria
* Patients with a prior diagnosis of celiac disease undergoing biopsy and TTG-IgA antibody testing for follow-up care * Children and vulnerable populations (e.g. pregnant women or prisoners) * Patients with IgA deficiency * Patients already following a gluten-free diet For Aim 2: Inclusion Criteria: * Physicians (primary care or subspeciality) who test or evaluate patients for celiac disease * Patients already diagnosed with celiac disease or undergoing evaluation for celiac disease
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Aim 1 Prediction Model | 1 year | The primary outcome being predicted is biopsy-confirmed celiac disease, defined as villous atrophy on duodenal histopathology. A prediction model will be built and after final model construction, the investigators will report its performance using five established measures: sensitivity, specificity, positive predictive value, negative predictive value, and F-measure. A calibration plot will be produced to illustrate if the model's predicted probabilities of an outcome reflect the true outcome probability. The investigators will use the SHapley Additive exPlanation (SHAP) method to provide a list of all model features ranked according to relative importance. |
| Aim 2 Interview Transcript Coding | Years 2-3 | Interview transcripts will be uploaded into NVivo software, a qualitative data analysis tool that facilitates coding of source data and identification of similarities in coded concepts indicative of themes. A research assistant and the PI will independently inductively code interviews in NVivo. Data-driven codes will be combined with a priori codes corresponding to the PRISM domains to develop the study codebook and summarize themes. We will map these codes to the PRISM framework to understand how the intervention, recipients, implementation structure, and external environment interact to support implementation of a prediction model for celiac disease diagnosis. |
| Aim 3 Model Performance and Patient Preferences | Years 4-5 | For aim 3, the primary outcome of interest will be model accuracy reported as AUC, AUPRC, sensitivity, and specificity. We will describe patient-reported preferences for communication and display of the prediction model, as well as ranking of decisional attributes (e.g. discomfort, certainty in results). Best practices for survey reporting will be used. |
Contacts
CONTACTClaire Jansson-Knodell, MD
Outcome results
None listed