**4. Discussion**

While pain after injury is an expected outcome, continued, unresolved pain is a troubling consequence of traumatic injury. This study explored relationships among biomarkers of stress and pain severity measured at two-time points (hospitalization and five to ten months post-injury) in a sample of 147 traumatically injured subjects to explore biological relationships underlying the transition from acute to chronic pain. Biomarkers, clinical measures, and pain scores were determined within days of the traumatic injury and again five to ten months later-a time period when physical injuries are healed. Moderate to severe chronic pain was present in 34% of subjects 5–10 months after injury, which is a lower percentage than found in a previous study of an injured patient population in a community-based sample [3]. In that study, which was carried out four months after the injury, 43% of patients had moderate to severe pain, 50% had moderate to severe life interference associated with the development of chronic pain, and 50% continued to use opioids to treat their chronic pain four months after a traumatic injury [3].

In the current study, those without a committed relationship were significantly more likely to have chronic pain. This is aligned with the results of a recently published large study (>900 participants) which found that those with limited social support were significantly more likely to exhibit chronic pain, functional limitations, and poor mental health outcomes following moderate-severe traumatic injury [40].

In our study, those injured due to a gunshot wound were more likely to develop chronic pain. Given that civilian gunshot injuries are associated with interpersonal violence, these data support the role of distress as a risk factor for developing chronic pain [41].

As has been reported previously [42], the path analysis demonstrated a significant, positive effect of the female sex on both acute and chronic pain. Likely because our sample was only about one-third women, this difference did not reach significance in univariate analyses, but it did go in the same direction as the path analysis. This difference was most likely because the path analysis had more power with pain scores as a continuous variable and the path model explained some of the variability in the pain scores reducing the unexplained variability and making it easier to detect other effects.

Numerical pain scores at the time of injury were positively associated with the severity of chronic pain. This is in accord with the well-accepted notion that acute pain causes

changes in the sensory pathway and pain-related brain circuits resulting in sensitization and chronic pain [43]. Both univariate and path analyses also identified a significant positive correlation between the severity of the physical injury (measured using the ISS) and acute pain at the time of injury and path analysis revealed a moderate but significant positive association between ISS and chronic pain. This result differed from findings in a previous study by our group that did not find ISS to correlate with the development of chronic pain [44], which suggests that ISS has a weak influence on chronic pain. Interestingly, the path model did not identify a significant association between ISS and acute pain, suggesting that ISS contributes to the risk for chronic pain beyond solely an enhancement of acute pain.

Cortisol concentrations were significantly negatively correlated with contemporaneous pain measures both in the univariate analyses and in the path analysis. For the path analysis, we made the assumption that the relationship between pain and cortisol would be independent of sampling time, reasoning that the biological relationship between them would not be altered by time since injury. A rerun of the model without this assumption did not appreciably change the relationships between cortisol and pain at either time, suggesting that this is a valid assumption.

While acute pain tends to increase concentrations of cortisol in healthy individuals, we found a significant, negative relationship between contemporaneous measures of circulating cortisol and pain both in the days after the traumatic injury and at 5–10 months follow-up. The expectation of high cortisol during acute pain was not seen in this sample. There are several possible explanations; the first is the timing of the cortisol measurements, which were several days on average after the injury. It is possible that cortisol was depleted at this stage due to the significant stress of the injury itself. Interestingly, previous studies have found that Blacks and Whites differ in cortisol responses to pain and stress, with Blacks exhibiting a lower response than Whites [45]. Given that our study population was enriched in individuals from racial and ethnic minorities, it is possible that the demographics of our sample contributed to this relationship. On the other hand, hypocortisolism has been linked to chronic pain disorders, including fibromyalgia, chronic fatigue syndrome, chronic pelvic pain, and temporomandibular disorder [46,47]. As suggested by Hannibal and Bishop, hypocortisolemia can potentiate and prolong chronic pain due to increased inflammation, which can increase pain and increase the risk for depressed mood, an additional risk factor for chronic pain [48].

The lack of association between levels of eCBs at hospitalization and acute pain refutes our hypothesis that pain measures correlate with eCBs at baseline. However, the positive correlation between cortisol and 2-AG at the time of hospitalization supports our hypothesis and is in accord with preclinical data demonstrating that glucocorticoid receptor (GR) activation increases 2-AG synthesis in the brain [49] and periphery [50], although the possibility that cortisol and 2-AG are elevated independently cannot be ruled out. In this regard, ISS and 2-AG were positively associated, suggesting that the severity of the traumatic injury may contribute to 2-AG concentrations independently from cortisol.

Our primary goal in this study was to test the predictive value of the biomarkers measured at the time of injury for the development of chronic pain. Bivariate analyses indicated that 2-AG concentrations at the time of hospitalization were positively associated with the degree of pain interference with activities of daily living measured using the Brief Pain Inventory at follow-up. The path analysis also identified a significant and positive association between circulating 2-AG concentrations at the time of injury and pain severity at the follow-up visit and a nearly one-to-one correspondence between pain severity and pain interference. These data support our hypothesis that circulating concentrations of 2-AG have value as a biomarker for the risk of developing chronic pain.

On the other hand, neither analysis demonstrated significant associations between measures of pain severity or interference at follow-up and circulating concentrations of 2-AG measured at the same time. This finding contrasts with studies in which circulating 2- AG was found to be higher compared to pain-free control groups in individuals with several types of chronic pain, including fibromyalgia, irritable bowel syndrome, and neuropathic

pain [17,18,51]. Differences in the duration and type of chronic pain could underlie the difference in findings.

It is ye<sup>t</sup> unknown if elevated 2-AG concentrations during the time of injury *per se* are mechanistically involved in the severity of pain months later. Given the preclinical data that CB1R activation is associated with reduced pain in many models and that highly elevated 2-AG concentrations result in reduced CB1R density [52], it is possible that CB1R signaling is down-regulated by the high 2-AG concentrations that occur following injury. This could result in increased pain perception at the time of injury, a known risk factor for the development of persistent pain [53]. Our finding of a significant relationship between pain severity at hospitalization and follow-up supports this notion, as do the extremely high concentrations of 2-AG during the peritraumatic period. However, pain severity was not related to circulating 2-AG concentrations at hospitalization, which would be expected if this were the mechanism. An alternative hypothesis, based upon the findings that chronic pain is accompanied by widespread changes in brain circuits [43] and 2-AG/CB1 signaling affects synaptic activity throughout the brain [54], is that excessive 2-AG-mediated signaling at the time of injury contributes to lasting changes in circuits that subserve chronic pain. Further studies are needed to explore these and other possible mechanisms.

There were no interactions between circulating concentrations of the second endocannabinoid, AEA, and the other measures in this study. While 2-AG and AEA are both endogenous ligands of the cannabinoid receptors, there are differences in the triggers for their mobilization and their reported associations with psychological indicators in humans [55]. Previous studies in humans have found that AEA concentrations are more likely to be associated with anxiety while 2-AG is more likely to be associated with depression [56,57]. Interestingly, 2-AG concentrations at hospitalization were also associated with increased risk for depression at follow-up [32], while AEA concentrations at hospitalization were associated with risk for the development of chronic PTSD [33] in the participants of this study.

This study was not without limitations. First, we measured the endocannabinoids at only two time points (hospitalization and at least 5 months later), which provided partial longitudinal data over a period of time when the participants were recovering from their injury and were undergoing important changes in symptoms. However, we do not know the trajectory, and importantly, we do not know when the elevated 2-AG concentrations returned to normal values. Our hypothesis that excessive elevation of eCB signaling suggests that individuals with a prolonged increase in 2-AG would be at greater risk for chronic pain. Measuring the eCBs over multiple time points would allow us to test this hypothesis. In addition, the follow-up visits occurred over a broad range of times (5.2 months to nearly 10 months), which could add significant variability to the results. The blood collection times were dictated by the situation of the participant, which resulted in a fairly wide range of elapsed time since injury and did not differ between the NCP and CP groups. Similarly, the time of day for the blood collections was not controlled for, which is a limitation because of the circulating concentrations of both cortisol and 2-AG exhibit strong circadian rhythms [58]. Moreover, the literature has noted that the microbiota can affect the ECSS [59–61]. However, this was not evaluated within our study. Finally, we did not screen participants for pre-existing chronic pain, which could be a possible confound for assessing chronic pain associated with the current traumatic injury. However, the participants were asked at the follow-up to evaluate the severity of their pain related to the injury a few months earlier.
