Pain occurs in more than 80% of patients with chronic pancreatitis (CP) , which is most commonly caused by chronic alcohol use. Management of CP pain is challenging, and greater than 50% of patients with CP pain are placed on chronic opioids, contributing to the opioid epidemic. CP with ductal obstruction can be treated by endoscopic therapies. However, success rates vary, and despite technically successful procedures, some CP patients continue to experience pain. The lack of tools for predicting pain treatment response to endoscopic therapies presents challenges for clinical care, potentially delaying care, leading to more invasive therapies such as surgery, or unnecessarily exposing patients to opioids. In this proposal we aim to use machine learning to develop multimodal predictive biosignatures for pain response to endoscopic therapies utilizing electroencephalography (EEG), quantitative sensory testing (QST), and biopsychosocial variables. Our study is founded on the premise that while pancreatic pathology initially drives pain, over time alterations in central pain processing may become a dominant driver of pain in some patients with CP, making them resistant to therapies aimed at the periphery. Neuroimaging with EEG, sensory testing with QST, and psychosocial questionnaires assess central vs peripheral changes in pain processing. Combining these tools in a multimodal biosignature can improve sensitivity and specificity of prediction and advance s election of appropriate treatment of CP pain. In the UG3 phase of our proposal, we will measure EEG and QST and assess psychosocial factors in ~100 patients with alcohol-induced CP pain undergoing endoscopic therapy as part of standard clinical care. Using machine learning algorithms, we will extract features and develop candidate predictive biosignatures for pain treatment response to endoscopic therapy. In the UH3 phase we will validate and select the biosignature with the highest area under the curve met ric in a new cohort of patients. Our success will have direct clinical impact, improving care of this refractory chronic pain syndrome and enabling similar studies in other chronic abdominal pain syndromes.

PROJECT NARRATIVE The lack of predictive biomarkers for pain treatment response to chronic pancreatitis (CP) pain hinders clinical decision making, leading to treatment of more than 50% of patients with CP pain with opioids. Our proposal seeks to develop multimodal biosignatures, incorporating electroencephalography, quantitative sensory testing, and biopsychosocial factors, powered by machine learning, to predict pain response to endoscopic therapies for CP pain. Our proposal for development of a predictive biosignature can advance personalized pain medicine and treatment for CP pain, minimizing opioid use.