Opioid use disorder (OUD) is a major problem in America, currently reaching epidemic levels. Unfortunately, OUD is especially prevalent among Veterans, as it is common that Veterans need pain treatment and the liberal use of opioids in medicine is one of the major reasons why the OUD problem keeps growing. There are good treatment options for OUD: Both buprenorphine and methadone can be used in maintenance therapies in which, as long as the patient stays in treatment, they will not likely truly abuse opioids. This is extremely important as one major reason for death in OUD is death by fentanyl overdose, and a patient in maintenance therapies will likely avoid that fate. However, it is very common that patients discontinue treatment. An important gap in knowledge arises from the fact that we have no means to predict which patients are more likely to drop from treatment. Such prediction would be of great interest as limited resources could be optimally allocated. In addition, an understanding of the brain circuitry behind both OUD and OUD treatment outcomes is necessary for the rational design of the next wave of therapeutic approaches. Big data approaches to scientific questions are increasingly common, however in psychiatry advances are (as usual, psychiatry likely being the most complex field in medicine) lagging. We have shown that using a machine learning approach to human brain imaging analysis, we can classify psychiatric patients according to past suicide attempt and high suicide ideation. We propose to use a similar (albeit improved) approach to the prediction of buprenorphine treatment in Veterans with OUD. We propose to use different MRI modalities (structure, white matter, resting state functional connectivity) and limited genotyping (two single nucleotide polymorphisms in the opioid receptor and the ? 5 nicotinic acetylcholine receptor subunit known to be associated with OUD risk) in machine learning algorithms to predict OUD treatment outcomes. MRI and genetics will be collected before treatment and MRI again within 10 days of treatment initiation (with a smaller group imaged at 6 months also), and Veterans will be followed up to study outcomes. If successful, this proposal would provide both mechanistic data (brain circuitry and function, including a genetic component) about OUD and OUD treatment outcome, and an unbiased approach to OUD treatment prediction.
Opioid use disorder (OUD) is a major problem, including high death risk, in returning Veterans. Although therapies are available (methadone and buprenorphine), often these don?t work especially because patients discontinue treatment. Our understanding of the brain mechanisms of OUD and OUD treatment success is lacking, which precludes the development of improved therapies. We propose to use brain imaging and limited genetics to understand the brain circuitry behind OUD and treatment outcomes, and to create a machine learning algorithm to predict which patients entering the clinic are most and less likely to succeed. This knowledge would allow for improved personalized therapy and for optimal resource allocation.
