Predicting Chronic Pain Following Breast Surgery

Overview

Breast surgery, which includes mastectomy, breast reconstructive surgery, or lumpectomies with sentinel node biopsies, may lead to the development of chronic pain and long-term opioid use. In the era of an opioid crisis, it is important to risk-stratify this surgical population for risk of these outcomes in an effort to personalize pain management. The opioid epidemic in the United States resulted in more than 40,000 deaths in 2016, 40% of which involved prescription opioids. Furthermore, it is estimated that 2 million patients become opioid-dependent after elective, outpatient surgery each year. After major breast surgery, chronic pain has been reported to develop anywhere between 35% - 62% of patients, while about 10% use long-term opioids. Precision medicine is a concept at which medical management is tailored to an individual patient based on a specific patient's characteristics, including social, demographic, medical, genetic, and molecular/cellular data. With a plethora of data specific to millions of patients, the use of artificial intelligence (AI) modalities to analyze big data in order to implement precision medicine is crucial. We propose to prospectively collect rich data from patients undergoing various breast surgeries in order to develop predictive models using AI modalities to predict patients at-risk for chronic pain and opioid use.

The primary objective of this is to develop predictive models using artificial intelligence algorithms to predict acute and chronic pain and opioid use in patients undergoing breast surgery. Development of these models will involve prospectively collecting data from this surgical population, including: 1) survey results from the Brief Pain Inventory, Fibromyalgia Survey Criteria, and PROMIS scales (depression scale, anxiety scale, physical function scale, fatigue scale, sleep disturbance scale); 2) pharmacogenomics (single nucleotide peptides from COMT, BDNF, SCN11a, OPRM1, ABCB1, CYPD26, and CYP34A, to name a few); 3) preoperative comorbidities (including but not limited to diabetes mellitus, chronic pain, psychiatric disorders, substance abuse history, obstructive sleep apnea, etc); 4) preoperative labs (i.e. hemoglobin); 5) demographic data (i.e. socioeconomic status, religion, ethnicity; primary language spoken, age, body mass index, sex, etc); 6) preoperative medication use; 7) primary surgical diagnosis; 8) surgery; and 9) social support system. Intraoperative data will include: 1) primary anesthetic type; 2) case duration; 3) total opioid use; 4) non-opioid analgesic use; 5) heart rate hemodynamics; and 6) blood pressure hemodynamics. Postoperative data will include: 1) total opioid use; 2) discharge medications; 3) hospital length of stay; 4) pain scores; 5) postoperative vital signs (blood pressure, heart rate); and 6) participation with physical therapy. The primary outcome measures will be opioid use in the acute period and chronic postoperative stage (30 and 90 days and 6 months) and development of chronic pain (up to 6 months after surgery). The model with the best performance will be used to develop a predictive analytic system aimed to identify high risk opioid patients in order to allocate expert pain management resources to those patients. We hypothesize that we can develop an accurate model for identifying high risk opioid users and patients at-risk for chronic pain in these surgical populations and subsequently implement a predictive analytic system that can detect these patients early-on.