Prediction of Co-Morbid Chronic Pain and Posttraumatic Stress: Results of a Pilot Analysis of Clinical and MicroRNA Data From a Longitudinal Cohort of African American Trauma Survivors
Recommended Citation
Kim R, Pan Y, Kurz M, Hendry P, Pearson C, O'Neil B, Lewandowski C, Datner E, Liu Y, McLean SA, and Linnstaedt S. Prediction of Co-Morbid Chronic Pain and Posttraumatic Stress: Results of a Pilot Analysis of Clinical and MicroRNA Data From a Longitudinal Cohort of African American Trauma Survivors. Biological Psychiatry 2020; 87(9):S212.
Document Type
Conference Proceeding
Publication Date
5-2020
Publication Title
Biological Psychiatry
Abstract
Background: Co-morbid chronic musculoskeletal pain and posttraumatic stress (CMSP/PTS) is a common outcome of trauma exposure and is associated with greater disability than either outcome alone. Identification of CMSP/PTS vulnerable individuals would aid in preventative treatment decisions. In the current study, we performed analyses to identify significant predictors and build a prediction tool for CMSP/PTS based on clinical and biological data. Methods: African American men/women presenting to the emergency department (ED) within 24 hours of motor vehicle collision were enrolled. Sociodemographic and psychological/cognitive characteristics, and blood (PAXgeneRNA) for microRNA-seq were collected in the ED. Six-month surveys identified individuals with CMSP (≥4, 0-10 Numeric Rating Scale)/PTS (≥33, Impact of Events Scale-Revised). The prediction tool was built using regularized logistic regression with feature selection, where significant predictors were identified via 1,000x repetitions of Monte Carlo cross-validation. Results: 30% (n=222/741) of the full cohort reported CMSP/PTS and 27% (n=198/741) reported neither outcome. Clinical and demographic variables were identified using a subset of individuals without miRNA data (n=332); selected variables showed good reliability in predicting CMSP/PTS (AUC=0.76+/-0.008). miRNA data alone (n=88) yielded weak reliability (AUC=0.64+/-0.009). Combining clinical, demographic, and miRNA variables (n=88) improved prediction versus either subset alone (AUC=0.79+/-0.008). Top predictors included initial pain severity, fear of pain getting worse, feeling frustrated or angry, socioeconomic status and microRNAs miR-199a, miR-339, let-7d, miR-192, and miR-29. Conclusions: These analyses suggest that supplementing clinical prediction with microRNA moderately improves accuracy of identifying vulnerable individuals. Future studies should aim to replicate these findings in additional trauma cohorts.
Volume
87
Issue
9
First Page
S212
Comments
Supported By: K01AR071504, R01AR060852, The Mayday Fund Keywords: PTSD, Chronic Pain, Machine Learning, Cross-Validation, microRNA