Interactions Between Endocannabinoid and Endogenous Opioid Systems Prospectively Influence Postoperative Opioid Use in Pregnant Patients Undergoing Cesarean Delivery.

This study used a prospective cohort design and was conducted between June 2021 and April 2022 at a single academic center. Eligible participants were pregnant patients aged 18 and older undergoing scheduled cesarean delivery under regional anesthesia. Participants receiving general anesthesia, presenting in labor or with unplanned cesarean delivery, having a prior diagnosis of chronic pain or opioid use disorder, or currently using methadone or buprenorphine were excluded. Vanderbilt University Medical Center serves a diverse community, and individuals of all races and ethnicities were eligible to participate. All study procedures were approved by the Vanderbilt University Medical Center Institutional Review Board. A sample size of at least n = 113 participants with complete data was targeted to provide a statistical power of ≥.90 using a 2-sided P < .05 as the criterion for significance to be able to detect a moderate effect size (r = .30). Participants were recruited in the outpatient clinic setting or upon inpatient admission for their scheduled cesarean delivery during preoperative assessment. Eligibility was confirmed by electronic REDCap, or in-person survey conducted by the study team. All participants provided informed consent prior to enrollment. Two 4 mL samples of blood were collected preoperatively in ethylenediaminetetraacetic acid tubes for plasma EC assays at the time of routine i.v. catheter insertion. Upon collection, blood samples were immediately processed by centrifuge for 10 minutes at 3,500 rpm under 4 °C within 20 minutes of sample collection, and the supernatant plasma was extracted and stored at −80 °C. At the time of regional spinal or combined spinal-epidural placement, a sample of participant CSF (~3.5–4.0 mL) was collected by the anesthesiologist for BE and EC assays prior to administration of any analgesic medication. CSF samples were collected and checked for blood contamination, then centrifuged to discard cellular debris. All samples were processed for 12 minutes at 3,000 rpm under 4 °C within 30 to 60 minutes of sample collection. The supernatant was extracted and stored at −80 °C. At discharge, participants were provided with 30 tablets of ibuprofen 600 mg and 30 tablets of hydrocodone 5 mg-acetaminophen 325 mg per study protocol, unless a different analgesic was preferred by their delivering provider. Relevant surgical and clinical information was extracted from the electronic health record. All participants were compensated for their time ($60 for completion of the study). At a 2-week postoperative follow-up, participants were asked via electronic REDCap survey to count and report the number of unused opioid tablets remaining and the number of days they had used opioids since discharge. The number of opioid tablets used during the 2-week follow-up period (primary opioid outcome) was calculated from the known initial prescription quantity and the number of unused opioid tablets. Our prior work found a high correspondence (intraclass correlation = .85) between opioid use based on self-reports and electronic pill cap-derived opioid use in the 2 weeks following cesarean delivery. 12 Secondary outcomes included number of days of opioid use and, given some variability in actual opioid analgesic prescribed (oxycodone vs hydrocodone as detailed in Sample Characteristics , Table 1 ), total amount of opioids used expressed in milligram morphine equivalents (MMEs) during the 2-week follow-up period. Pain intensity was assessed via REDCap both preoperatively (to characterize the sample) and at 2-week postoperative follow-up (as a secondary outcome) using a 0 to 10 numeric rating scale (NRS) measure of average pain intensity in the last 7 days taken from the Patient-Reported Outcomes Measurement Information System (PROMIS) Profile 29 Version 2.0. 13 , 14 BE concentrations in CSF samples were quantified via enzyme-linked immunosorbent assay using a competitive enzyme inhibition immunoassay technique ( antibodies-online.com , No. ABIN6955552). Duplicate assays were carried out according to manufacturer instructions, with a reported assay sensitivity of 5.11 pg/mL, detection range of 12.35 pg/mL to 1,000 pg/mL, and inter- and intra-assay coefficients of variation (CV%) of < 30%. The standard was further diluted (.69–1,000 pg/mL) given that the majority of samples were in the lower range of detection. After determining that the standard curve was linear at lower concentrations, the specimens were measured undiluted using the adjusted standard curve. Samples with CV% > 30% were reprocessed and reanalyzed to reduce errors related to sample preparation or assay issues. Final data analysis used the mean of the duplicate samples and excluded participants whose CSF assay results had CV% > 30%. Participants with CSF BE concentrations below the lower limit of detection were included in the final analysis and were assigned a value of .69 pg/mL. In the current study, the ECs of interest included 2-arachidonoylglycerol (2-AG) and anandamide (AEA). These were assayed using liquid chromatography coupled with tandem mass spectrometry (LC-MS) with methods similar to those described by Bruehl et al. 8 HPLC-grade methanol, acetonitrile, and water for LC-MS analysis were purchased from ThermoFisher Scientific (Waltham, MA). Formic acid, butylated hydroxytoluene (BHT), toluene, cyclohexane, ethyl acetate, trifluoracetic acid, phenylmethanesulfonyl fluoride (PMSF), sodium flouride (NaF), and sodium chloride (NaCl) were purchased from Sigma-Aldrich (St. Louis, MO). EC authentic standards and deuterated internal standards (AEA-d4 and 2-AG-d5) were purchased from Cayman Chemicals (Ann Arbor, MI). All LC-MS analyses were performed on a Shimadzu Nexera system in-line with a SCIEX 6500 QTrap (Framingham, MA). The QTrap was equipped with a TurboV Ionspray source and was operated in the positive-ion mode and analyte detection was accomplished via multiple reaction monitoring. SCIEX Analyst software (ver 1.6.2, Framingham, MA) was used to control the instruments and then acquire and process the data. Prior to LC-MS analysis, plasma samples were thawed, vortexed, and 1 mL transferred to new glass vials. Deuterated internal standards (AEA-d4 and 2-AG-d5) were added to each sample, as well as a 10 μL aliquot of methanolic BHT (conc = .2 M), PMSF (conc = .12 M), and a 10 μL aliquot of aqueous NaF (1.0 M). A 4 mL solution of 1.0% trifluoracetic acid (aqueous) and .4 mL of saturated NaCl was also added and the samples were vortexed. The analytes were extracted by the addition of 5.0 mL of 20:1 (v:v) toluene:cyclohexane, after which the mixture was mixed well, centrifuged to promote phase separation, and the upper layer was removed to a clean vessel and dried under N 2 . Samples were stored at −20 °C if they were not going to be analyzed immediately. CSF samples were processed by slightly differing methodology. Samples were removed from −80 °C storage and thawed on ice. Samples were mixed, and .7 mL were transferred to 12 × 75 mm test tubes. Samples were then spiked with deuterated internal standards (AEA-d4 and 2-AG-d5), as well as a 10 μL aliquot of methanolic BHT (conc = .2 M) and PMSF (conc = .12 M) and a 10 μL aliquot of aqueous NaF (1.0 M). Next, the samples were diluted with an equal volume of water (containing 1% acetic acid), mixed well, and extracted with a 3× volume of ethyl acetate:cyclohexane (9:1) (v:v). The samples were capped, mixed, and centrifuged at 3,000 rpm for 15 minutes at 4 °C. The upper organic layer of each sample was transferred into a new vial and dried under N 2 for storage at −20 °C until LC-MS analysis. Dry samples were prepared by reconstitution in 70 μL MeOH and 40 μL H 2 O. Details of the analysis of 2-AG, AEA, and their internal standards have been described in a prior study. 8 All analytes were quantitated against deuterated internal standards, and all concentrations were given in pmols/mL. Preliminary data revealed that concentrations of the EC AEA in CSF were undetectable in most samples. Therefore, the final analysis was limited to CSF 2-AG alone. All analyses were conducted using SPSS for Windows version 29 (IBM Inc, Armonk, NY). Descriptive analyses used means (SD) and percentages for continuous and categorical data respectively. Although CSF 2-AG was normally distributed, distributions for CSF BE and plasma 2-AG and AEA were significantly non-normal ( P ’s < .001) based on Kolmogorov-Smirnov tests specifying a normal distribution. To reduce the influence of extreme values, which might tend to overestimate parameter estimates, we conservatively applied log 10 transformations to CSF BE and plasma 2-AG and AEA values prior to analysis. Spearman nonparametric correlations were used to examine zero-order correlations between CSF BE, EC measures, and the pain and opioid use outcomes at follow-up in order to display the pattern of effects when not adjusted for potential confounders. Although participant ethnicity was not associated with BE or EC levels, preliminary analyses indicated that these biomarkers did vary as a function of maternal age and body mass index (BMI) at delivery, and/or self-reported race (dummy coded as: 1 = Black, 2 = Asian, 3 = Multiple Races, 4 = White [referent group]). These variables were therefore included as control variables in all analyses to permit examination of BE and EC effects on outcomes, independent of these potential confounds (ie, type III sums of squares). To optimize testing of interactions in primary analyses and ensure accurate graphic display of the source of interactions, the BE and EC measures, as well as continuous control variables, were mean-centered prior to analyses (the mean of each variable was subtracted from the actual value 15 ). Primary analyses of the opioid outcomes (number of tablets used, number of days used, MMEs of opioids used) were generalized linear model (GLM) analyses employing a Poisson distribution due to the highly skewed and dispersed nature of opioid outcomes. Primary analyses for the postoperative pain outcome were a series of hierarchical linear regressions with the control variables above entered in step 1, the main effects of CSF BE and the targeted EC measure entered in step 2, and the multiplicative interaction of these biomarkers entered in step 3. All analyses used a criterion of P < .05 (2-sided) for significance. Effect sizes for all significant effects are provided as Cohen’s w (from Chi-square values) or Cohen’s d (from t values). Solely for the purpose of characterizing the source of significant interactions identified, simple effects analyses were conducted to examine the effect of the EC measure on the targeted outcome within low versus high CSF BE sub-samples (based on median splits 15 ). To portray the source of interactions for primary opioid use and secondary pain outcomes graphically, we employed the approach recommended by Aiken and West 15 and used publicly available Excel (Microsoft Corporation, Redmond, WA) macros optimized for Poisson regression and linear regression, respectively. 16 These figures display predicted outcome values as a function of hypothetical biomarker values one standard deviation below and above the sample mean for both BE and the EC measures. This simple effects analytic approach and method of graphical presentation for interactions is based on best practice recommendations. 15 , 17

A STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) flow chart is provided in Fig 1 . Complete study data were available for a total of 137 participants. Of these, 112 (81.8%) were included in the final analysis, which consisted of all enrolled participants with valid CSF BE and EC assay data necessary to test hypothesized interactions. Of note, we were unable to identify detectable concentrations of the EC AEA in CSF for most participants; therefore, analysis of the impact of EC concentrations in the central nervous system on pain and opioid use outcomes was limited to CSF 2-AG. Sample characteristics are summarized in Table 1 . Participants, on average, were relatively young, White, and of non-Hispanic ethnicity. Nonetheless, the sample displayed diversity reflecting the local community, with > 20% of participants identifying as non-White. The majority of participants had undergone cesarean delivery previously. Current use of psychiatric medication was uncommon. The standard opioid analgesic prescribed at discharge was hydrocodone/acetaminophen 5/325 mg, with a small proportion (14%) receiving oxycodone (per preference of their delivering provider). The number of opioid tablets used over the 2-week follow-up period showed substantial variability and ranged from no opioid use to using the full prescribed amount (30 tablets). The number of days opioids were used also displayed substantial variability. Eleven patients (9.8%) used all prescribed opioids. Table 2 displays zero-order Spearman nonparametric correlations among CSF BE, CSF and plasma ECs, and measures of opioid use and pain at 2-week postoperative follow-up. These zero-order correlations display the associations between preoperative biomarkers and opioid use and pain outcomes when not adjusted for identified confounds, all of which were nonsignificant ( P ’s > .10). Table 2 also displays the associations among preoperative BE concentrations and preoperative EC measures. Higher CSF BE was associated with significantly higher CSF 2-AG, but also with significantly lower plasma AEA levels. No associations were noted between CSF BE and plasma 2-AG. Higher plasma 2-AG was associated with significantly higher plasma AEA as well. Finally, we examined associations between 2-AG from central and peripheral sources (the only EC for which both CSF and plasma values were available). The correlation between CSF and plasma measures of 2-AG indicated a significant positive association that was small in magnitude. GLMs (summarized in Table 3 ) were used to assess the main and interactive effects of CSF BE and EC measures on outpatient opioid use primary outcomes over the 2-week follow-up period (based on pill counts). All analyses controlled for the potential confounding effects of maternal age, BMI, and race. Model fit for all analyses was adequate (Pearson Chi-Square value/df > 3.23 for all analyses). In GLM analyses targeting the CSF 2-AG measure of ECs, there was a significant CSF BE × CSF 2-AG interaction for the number of opioid tablets used during the follow-up period (Cohen’s w = .42). Simple effects analyses were conducted in the high versus low BE subgroups (based on median split) to further characterize the interaction. In the low BE group, there was a significant inverse association between CSF 2-AG and number of opioid tablets used, that is, those with higher CSF 2-AG levels used fewer opioid tablets at 2 weeks postoperatively compared to those with lower CSF 2-AG (beta = −.243, standard error (SE) = .04, P < .001). In contrast, there was a nonsignificant association noted in the high BE group (beta = −.077, SE = .044, P = .079) that included zero in the 95% confidence interval (−.163 to .009). Fig 2 portrays the source of this interaction by plotting the predicted number of opioid tablets used for hypothetical CSF BE and CSF 2-AG values one standard deviation above and below the sample mean. Inspection of this figure indicates that elevated concentrations of 2-AG in CSF were associated with reduced opioid use, but only in people simultaneously displaying low concentrations of CSF BE. There was a similar significant CSF BE × CSF 2-AG interaction noted for the number of days opioids were used over the follow-up period (Cohen’s w = .23). In the low BE group, there was a significant inverse association between CSF 2-AG and number of days opioids were used (beta = −.186, SE = .065, P = .004) with no association in the high BE group (beta = .014, SE = .064, P = .820). Finally, the CSF BE × CSF 2-AG interaction was significant for the total amount of opioids used in MMEs over the follow-up period (Cohen’s w = .52). In a pattern similar to the findings above, there was a significant inverse association between CSF 2-AG and opioid MMEs used in the low BE group (beta = −.250, SE = .021, P < .001) with a much smaller inverse association in the high BE group (beta = −.123, SE = .019, P < .001). Fig 3 indicates that elevated CSF 2-AG was associated with the lower total amount of opioids used primarily in individuals also having low BE levels in CSF. GLM analyses of the plasma 2-AG measure also revealed a significant interaction with CSF BE on the number of outpatient opioid tablets used (Cohen’s w = .23). As for CSF 2-AG, simple effects analyses indicated that elevated plasma 2-AG was associated with fewer opioid tablets used in the low CSF BE group (beta = −1.068, SE = .299, P < .001). There was no significant association observed in the high BE group (beta = −.064, SE = .186, P = .730), with a 95% confidence interval for the beta weight including zero (−.430 to .301). As was noted for 2-AG concentrations in CSF, inspection of Fig 4 indicates that higher plasma 2-AG was associated with reduced opioid use most strongly among participants also displaying low CSF BE. A significant interaction between CSF BE and plasma 2-AG on outpatient postoperative MMEs was also noted (Cohen’s w = .53). This interaction was due to a significant inverse association between plasma 2-AG and MMEs in the low CSF BE group (beta = −1.267, SE = .129, P < .001) but a nonsignificant association observed in the high BE group (beta = −.102, SE = .082, P = .210). The 95% confidence interval for the latter beta weight included zero (−.262 to .058). GLM results for the days of opioid use outcome only approached significance for the BE × Plasma 2-AG interaction ( P = .054), so this interaction was not further interpreted. There were no significant main effects in the latter model ( P ’s > .10). GLM analyses of the plasma AEA measure also revealed a significant interaction with CSF BE on the number of outpatient opioid tablets used (Cohen’s w = .34). Simple effects analyses revealed a larger positive association between plasma AEA and the number of opioid tablets used in the high CSF BE group (beta = .760, SE = .195, P < .001) than in the low BE group (beta = .600, SE = .199, P = .003). These effects are summarized in Fig 5 . In contrast to results for both CSF and plasma 2-AG, the lowest opioid use was observed among participants with low plasma AEA concentrations irrespective of BE concentrations. The interaction between CSF BE and plasma AEA on total MMEs of opioid use over the 2-week follow-up period was also significant (Cohen’s w = .69). In a pattern similar to the opioid tablets outcome, simple effects analyses revealed a larger positive association between plasma AEA and outpatient MMEs used in the high CSF BE group (beta = .774, SE = .086, P < .001) than in the low BE group (beta = .558, SE = .085, P .10). For all significant race effects noted in the models above, participants who identified as White reported lower opioid use than did participants in other racial groups (Black, Asian, Multiple Races). Given the number of analyses reported for related opioid outcomes, we also examined the results when adjusting for false discovery rate using the Benjamini-Hochberg procedure. All significant BE × EC interactions described above, except for the BE × plasma 2-AG interaction, on the number of opioid tablets used remained significant even after false discovery rate correction. In secondary analyses, hierarchical linear regression models (see Table 4 ) were used to assess the main and interactive effects of preoperative CSF BE and EC measures on the past week’s average NRS pain at the 2-week follow-up. All analyses controlled for the potential confounding effects of maternal age, BMI, and race. Results for CSF 2-AG revealed only a nonsignificant trend for interaction between CSF BE and CSF 2-AG ( P = .091), so this interaction was not pursued further. There was also a nonsignificant trend ( P = .072) suggesting a main effect association between elevated preoperative CSF 2-AG and decreased past week average NRS pain intensity at 2-week follow-up. All main and interaction (with CSF BE) effects for plasma 2-AG were nonsignificant ( P ’s > .10). However, analyses revealed a significant CSF BE × Plasma AEA interaction on average pain at follow-up (Cohen’s d = .39). Simple effects analyses indicated that there was a significant inverse association between plasma AEA and average pain intensity (ie, higher preoperative AEA was associated with lower postoperative pain) only in the low CSF BE group (beta = −3.009, SE = 1.298, P = .025) whereas there was a positive association approaching significance noted in the high BE group (beta = 3.003, SE = 1.557, P = .060). Fig 6 indicates that participants displaying both high CSF BE and high plasma AEA at preoperative baseline reported the highest pain at a 2-week postoperative follow-up.

This study evaluated the impact of interactions between CSF BE and EC concentrations (CSF and plasma) on outpatient opioid use in pregnant patients following cesarean delivery. Our hypotheses were guided by prior placebo-controlled laboratory work in chronic back pain patients, which suggested interactions between central endogenous opioid function and plasma ECs on opioid analgesic responsiveness. 8 Specifically, we prospectively tested the hypothesis that pregnant patients with both lower preoperative CSF BE and EC concentrations would display greater opioid use following cesarean delivery. Our findings did not support this hypothesis. In the context of acute pain following cesarean section surgery, higher rather than lower 2-AG was linked to reduced outpatient opioid analgesic use in individuals with low CSF BE levels. In contrast, higher plasma AEA was linked to greater outpatient opioid use in individuals with high CSF BE concentrations. Differing patterns of findings between 2-AG and plasma AEA could potentially reflect their differing affinity for CB-1 and CB-2 receptors (full agonist vs partial agonist, respectively), 18 , 19 with CB-2 being the primary peripheral EC receptor targeted by plasma ECs. 20 We note that effect sizes for observed endogenous opioid X EC interactions were generally medium to large in magnitude based on Cohen’s standard interpretation guidelines. 21 Our prior findings 8 suggest a reinforcement model might explain the current pattern of results. In that prior work, individuals with low plasma EC concentrations and low endogenous opioid function exhibited the greatest opioid analgesic efficacy under placebo-controlled laboratory conditions. Assuming the converse was also true, we might, by extension, hypothesize that individuals with both high EC and BE concentrations would use fewer opioids postoperatively if they experienced lower analgesic efficacy and, therefore, lower reinforcement of their opioid use. While this is consistent with findings for both CSF and plasma 2-AG in the current work, the opposite was observed for plasma AEA. Because it is difficult to extend past laboratory findings regarding placebo-controlled opioid efficacy to these clinical data regarding ad libitum opioid use, this hypothesis remains speculative. Several methodological differences could also have contributed to differing findings between the present postoperative pain study and past laboratory work in chronic pain patients. The current participants were females free of chronic pain in the final stages of pregnancy, which has a specific hormonal milieu. Preclinical studies suggest direct interactions between sex hormones and both the EC 22 and endogenous opioid systems. 23 Thus, it is possible that endogenous opioid and EC systems interact differently in pregnant females than in mixed-sex chronic pain or other postoperative populations due to fluctuations in pregnancy-related sex hormones. Furthermore, it is important to consider that placebo-controlled opioid antagonist laboratory methodology was used to assess endogenous opioid function in prior human studies 8 , 24 whereas CSF BE concentrations were used to index endogenous opioids in the present work. CSF BE concentrations may not directly reflect endogenous opioid inhibitory function due to un-assessed influences of receptor availability, 24 , 25 but nonetheless provide clues regarding central endogenous opioid function. We also hypothesized that individuals low in preoperative BE and ECs would exhibit higher pain intensity at follow-up, given previously reported pain inhibitory effects of both BE and ECs. 10 , 11 However, analyses of pain intensity at 2-week follow-up indicated a significant interaction in which higher BE concentrations were associated with lower subsequent pain only for individuals with low plasma AEA. Surprisingly, individuals with high concentrations of both BE and AEA reported the highest postoperative pain levels at 2 weeks postoperative follow-up. Although contrary to our original hypothesis, this is consistent with one of the only prior studies examining CSF ECs in the postoperative pain context. 26 In osteoarthritis patients undergoing total knee arthroplasty, higher preoperative CSF 2-AG predicted more intense postoperative pain. 24 Although speculative, it is plausible that elevated EC concentrations in our participants prior to cesarean delivery triggered receptor down-regulation, resulting in reduced functional effects of EC agonists at follow-up. This would be consistent with work suggesting that EC-related receptor downregulation affects endometriosis-associated pain. 27 Although questions remain regarding underlying mechanisms, this study has potential research and clinical implications. First, we present a unique female pain model for jointly evaluating EC and endogenous opioid system influences on opioid analgesic responses and pain by using pregnant patients undergoing a surgical procedure common both in the United States and globally. 28 In a purely research setting, CSF samples are infrequently obtained due to rare but serious potential complications (pain, dural puncture headache, spinal infection, or bleeding). However, CSF is routinely obtained and discarded during clinical regional anesthesia procedures. Pregnant patients undergoing cesarean delivery who have already agreed to the risks of neuraxial anesthesia provide an optimal opportunity to obtain CSF samples for research studies without additional risk. Second, plasma and CSF concentrations of 2-AG in the current study both similarly predicted opioid use outcomes, suggesting the possibility that blood sampling could provide a pragmatic means of assessing biologically relevant EC concentrations for research studies. However, given the small magnitude of the significant association between CSF and plasma 2-AG in the current study, it is possible that these represent separate central and peripheral EC mechanisms that influence opioid use outcomes via different pathways. Because extracting and analyzing ECs is complex, requiring mass spectrometry, future work identifying phenotypic markers reliably associated with ECs (and endogenous opioid activity) could help with the translation of this work to the clinical environment. Finally, the apparently different nature of interactions between endogenous opioids and ECs on opioid analgesic outcomes depending on the type of pain (acute vs chronic) and methodology used to assess endogenous opioids (opioid blockade responses vs. agonist concentrations) and opioid analgesic outcomes (placebo-controlled efficacy vs clinical use) highlights the extent to which we are still discovering ways in which these 2 systems operate in clinical human pain conditions. As pharmacotherapy trials seek to evaluate the synergistic effects of exogenous cannabinoids and opioid analgesics, it will be important to consider potential differences in acute versus chronic pain states, sex differences, and individual differences in receptor availability. Several limitations warrant consideration. First, we were unable to identify detectable concentrations of the EC AEA in CSF to examine the impact of AEA in the central nervous system on pain and opioid use outcomes. Recent work suggests that AEA concentration are generally low in CSF and further refinement of our methodology may be necessary to obtain detectable concentration of AEA in future studies. 29 , 30 A second potential limitation is that the sample in this study was relatively young and only included female patients who were pregnant and undergoing a single major abdominal surgery at term. Studies on female reproductive processes have demonstrated a close link between the EC system and the hypothalamic-pituitary-ovarian axis, with receptors identified in the hypothalamus, anterior pituitary, and ovary. 31 While dynamic changes in endogenous opioid and EC plasma concentration are known to occur around the time of parturition, changes in the CSF are not as well studied and likely do not vary. 32 While standardization in this study of the surgical procedure and indication enhanced homogeneity and may have enhanced the ability to detect endogenous opioid and EC effects, the unique nature of the current sample does not permit generalization to males, nonpregnant females, older individuals, or other surgical procedures. Generalizability remains to be evaluated in subsequent work. Third, race and ethnicity were assessed only by single-item self-report measures, and examining the influence of these on outcomes as broader social constructs was not possible with the data available. The finding that race was a significant covariate and associated with opioid use outcomes independent of endogenous opioid and EC influences is important to consider in future work. Finally, whether associations between ECs and BE and pain intensity outcomes would have differed (ie, inverse associations instead of positive) if they had been assessed concurrently at 2-week follow-up cannot be addressed—biomarker assays were only available preoperatively due to the reliance on clinical sampling procedures. Moreover, pain intensity measures reflected only the last week of the 2-week follow-up period, whereas opioid use was assessed over the entire follow-up period. This difference may have affected the pattern of findings observed across the opioid use and pain intensity outcomes.

Our study presents unique evidence regarding the interactive effects between the EC and endogenous opioid systems on analgesic responses and pain outcomes by directly assessing CSF agonist levels in the postoperative setting of patients undergoing cesarean delivery. The consistency of the predictive findings for 2-AG in both plasma and CSF suggests a reliable association between higher ECs and reduced postoperative opioid use for individuals with low endogenous opioid inhibition. Further exploration of interactions between these 2 inhibitory systems as they influence responses to opioid analgesics in other clinical pain populations may help guide the development of precision pain management approaches. Given the substantial variability in study design and findings regarding the analgesic efficacy of exogenous cannabinoids in clinical pain studies, 5 , 33 continuing to explore how endogenous opioids and other inhibitory systems may alter cannabinoid analgesic effects appears warranted.