*3.1. Demographic Results*

Demographic information, pain data, and biospecimens were collected from hospitalized participants an average of 2.5 days following injury (range 1–10 days; hospitalization time point). Pain data and blood samples were also obtained 5–10 months post-injury (average 192 days; range 156–286; follow-up time point). The demographic and clinical information related to the injury for the total population of participants are shown in the second column of Table 2. The sample was predominately male (69.4%) and spanned the entire adult age range. Forty-five percent of the sample self-identified as Black or African American and 7.5% as Hispanic or Latino.

**Table 2.** Demographic and clinical data of the study participants.


NCP: no chronic pain and CP: chronic pain subgroups determined at follow-up. Age, time between injury and follow-up, severity scores (ISS), and acute pain scores were compared between the NCP and CP groups using Mann–Whitney non-parametric *t*-tests; other comparisons were made using the Chi-squared test.

The presence of pain was assessed at the follow-up visit and chronic pain (CP) was defined as an NPS of 4 or greater and was endorsed by 50 individuals (34%). The demographic and clinical characteristics of the no chronic pain (NCP; NPS < 4) and CP subgroups of the sample were determined and compared (Table 2). There were no significant differences between the NCP and CP subgroups with regard to age or sex. There was no difference in the time that elapsed between the injury and follow-up visit between the NCP and CP subgroups. The ISS and NPS scores at the time of hospitalization were both significantly higher in the CP group than in the NCP group. Those not in a committed relationship were more likely to have chronic pain. One hundred forty-five of the 147 subjects in the study were treated with opiate analgesics at the time of injury, so the impact of opiates on chronic pain could not be studied in this cohort. Individuals were asked about cannabis use at both the time of hospitalization and at the follow-up assessment; there was no difference in reported use between the NCP and CP subgroups (data not shown).

#### *3.2. Mechanisms of Injury*

The mechanisms of traumatic injury were examined in the entire sample and compared between the NCP and CP subgroups. The three most prevalent mechanisms of injury were motor vehicle crashes (32%), falls (17%), and gunshot wounds (16%). A significantly greater proportion of those in the CP than the NCP subgroup was injured by gunshot wounds (28%, *p* = 0.023).

#### *3.3. Correlational Analyses of Biomarkers and Pain*

Correlational analyses were used to test our initial hypotheses that circulating concentrations of the eCBs and cortisol are correlated with acute and chronic pain (Table 2). For these exploratory analyses, False Discovery Rate (FDR) <0.1 was considered a significant correlation. At the time of hospitalization, cortisol concentrations were significantly, negatively correlated with acute pain; neither 2-AG nor AEA concentrations were correlated with acute pain. There was a modest, positive correlation between cortisol and 2-AG at hospitalization.

At follow-up, none of the biomarkers were correlated with pain measures. However, a significant, positive relationship between 2-AG and cortisol occurred while the concentrations of AEA and cortisol were negatively correlated.

To test the predictive value of the biomarkers, correlational analyses were carried out between biomarker concentrations at the time of hospitalization and indices of chronic pain 5–10 months after injury. 2-AG concentrations at the time of hospitalization were positively correlated with the degree of pain interference with activities of daily living 5–10 months after the injury.

#### *3.4. Model Incorporating Biomarkers, Demographics, and Pain Measures*

A path model was developed to analyze the relationships among the biomarkers, pain measures, and covariates (Figure 1). Table 1 shows the goodness of fit indices of the initial and reduced models, indicating an excellent fit for both.

The strongest relationships identified using this model were: (1) a positive relationship between serum concentrations of 2-AG at hospitalization and pain severity at follow-up; and (2) a negative relationship between cortisol and pain severity at both time points. Both 2-AG and cortisol concentrations at hospitalization were positively correlated with 2-AG concentrations at follow-up, and cortisol concentrations at hospitalization and followup were positively correlated with each other. ISS influenced pain severity at follow-up directly and indirectly through a relationship with 2-AG concentrations at hospitalization. The female sex exerted a significant, positive effect on pain severity at both time points, independent of other factors in the model. Cortisol at hospitalization affected cortisol at follow-up, and pain severity affected interference with activities of daily living at follow-up.

**Figure 1.** Model of the Relationship of Traumatic Pain, Cortisol, and Endocannabinoid Levels. The numbers on the arrows are regression coefficients, representing the effect of a 1-unit increase in the predictor on the outcome. For the log-transformed biomarkers, a 1 unit increase on the log scale corresponds to a 10-fold increase in the original value. All coefficients shown are significant with *p* < 0.05.
