Apparent Effects of Opioid Use on Neural Responses to Reward in Chronic Pain

Neural responses to incentives are altered in chronic pain and by opioid use. To understand how opioid use modulates the neural response to reward/value in chronic pain, we compared brain functional magnetic resonance imaging (fMRI) responses to a monetary incentive delay (MID) task in patients with fibromyalgia taking opioids (N = 17), patients with fibromyalgia not taking opioids (N = 17), and healthy controls (N = 15). Both groups of patients with fibromyalgia taking and not taking opioids had similar levels of pain, psychological measures, and clinical symptoms. Neural responses in the nucleus accumbens to anticipated reward and non-loss outcomes did not differ from healthy controls in either fibromyalgia group. However, neural responses in the medial prefrontal cortex differed, such that patients with fibromyalgia not taking opioids demonstrated significantly altered responses to anticipated rewards and non-loss outcomes compared to healthy controls, but patients with fibromyalgia taking opioids did not. Despite limitations including the use of additional non-opioid medications by fibromyalgia patients taking opioids, these preliminary findings suggest relatively “normalized” neural responses to monetary incentives in chronic pain patients who take opioids versus those who do not.


S2
and we used the computerized adaptive test (CAT) version 1 7

. The Fibromyalgia Assessment
Form includes 19 body locations that can be denoted as painful and these are derived from the American College of Rheumatology revised 2011 criteria for fibromyalgia 8 .
Note: The original BIS/BAS questionnaire by Carver and White uses a 4-point response scale (1 = strong agreement, 4 = strong disagreement) with no option for a neutral response 1 . Our BIS/BAS questionnaire unintentionally used a 5-point response scale which included a neutral response option. Thus, our BIS/BAS measurements may have been influenced by central tendency effects (i.e., preferential selection of the neutral response option).

Regarding Power for MID Task fMRI Analysis
Peak MID task fMRI NAcc reward anticipatory activation (+$5 versus +$0) results in the large reported effect size of f 2 = 3.07, and at least 6 participants are required to detect group effects with power equal to .80 (p < 0.05) 9,10 . Previous investigations support between-group differences for NAcc MID task fMRI data with sample sizes of 12 to 19 per group 11,12 .
Therefore, we estimate that our group sample sizes should be large enough to detect group differences in fMRI activation in response to the MID task design. While our sample sizes are modest, we believe that they are sufficient to provide meaningful results for the generation of data to support future hypotheses and subsequent/replication investigations. We also note that reasonably powered comparisons have been made in schizophrenia using the MID task (group sizes of: 22 patients with schizophrenia, 24 with patients with depression, and 21 controls), so we believe it is reasonable to generate similar hypotheses with our somewhat smaller sample 13,14 .

Region of Interest Additional Details
The NAcc and MPFC regions of interest (ROI) used in the present analysis were identical to those used in our previous publication 15 . Specifically, for the MPFC ROI selection, we created a S3 MPFC ROI as two conjoined spheres (4mm radius) at +/-4, 50, -3 (Brodmann Area 10 / frontal pole). We chose these coordinates primarily from a previous study using the MID task and analyzing a MPFC ROI in healthy controls and major depressive disorder patients (MDD) which used the coordinates: 4, 50, -4 11 . We used slightly inferior coordinates for our ROI with the goal of overlapping with a MPFC region shown to have altered functional connectivity to the NAcc in patients with chronic pain which used coordinates: +/-2, 52, -2 16 . Additionally, we chose more lateralized (x = +/-4) and more posterior (y = 50) coordinates from the previous (MDD patients) study in order to extend lateral coverage of the area, reduce the amount of overlap of the spheres, and to avoid potential prefrontal signal dropout (which is typically found more anteriorly). Our selected ROI was superior and anterior to the MPFC results identified in a previous meta-analysis 17 , however, we believe that our selected ROI region was less susceptible to any potential fMRI artifacts within the orbitofrontal cortex (which is a region that often suffers from fMRI data signal drop out due to proximity to the orbital and sinus cavities) 18 .
To create our ROI, we first created a MPFC mask image by drawing spheres (AFNI) over a TT_N27+tlrc structural image. Then, we resampled the mask image to match the our fMRI data image dimensions (using AFNI's 3dfractionize, with clip 0.1) which resulted in a mask containing 42 voxels.

FMRI Beta and Questionnaire Variable Correlation Analysis
Arousal (ratings to +$5 cues) was included as a behavioral measure in the correlational analyses because of the group differences identified for arousal (see Fig. 1, whereas valence was not included in the correlational analyses because no group differences were observed).
Correlations between ROI fMRI beta values and questionnaire variables were corrected for multiple comparisons based on the inclusion of 7 independent (not correlated) measures.
Several of the questionnaire variables were correlated with each other (using data from the fibromyalgia groups, N=34) and represented 4 independent measures: (1) arousal (not correlated with other measures); (2) behavioral drive and behavioral fun seeking (all p < 0.001); (3) behavioral inhibition (not correlated with other measures), (4) behavioral reward responsiveness, positive affect, negative affect, trait anxiety, state anxiety, total mood disturbance, depression, pain severity, pain interference, fatigue, and number of painful body regions (all p < 0.05). Only the NVLout ROI fMRI betas were correlated with each other (using data from the fibromyalgia groups, N=34, r = 0.531, p = 0.001). Therefore, the four ROI fMRI betas represented 3 independent measures: 1) NAcc GVNant, 2) MPFC GVNant, and 3) combined independent measure for NAcc NVLout and MPFC NVLout. Thus, all correlations between ROI fMRI betas and questionnaire variables were Bonferroni corrected for a total of 7 multiple comparisons and considered significant at the level of p < 0.007 (corrected threshold).

Analysis for Motion Effects
Motion can be a confounder in any fMRI experiment, and this is especially important to consider when comparing patients to healthy controls because individuals experiencing pain (e.g., chronic pain patients) during scanning may have difficulty remaining still during MRI scans.
Because of this we conducted additional post-hoc analyses to confirm that group differences and individual differences in motion during scan acquisition were not confounders of or S5 contributors to our results. To determine average motion during the MID task scans for each participant, Euclidean norm values (e-norm, calculated square root of the sum of squares across 6 rigid-body motion parameters) were calculated for each fMRI volume. These motion estimates were included as a covariate of no interest in post-hoc analyses (ANCOVA, SPSS) and were found to not change the results. The extracted beta values were statistically analyzed across all three groups using two-tailed bivariate Pearson correlation of ROI fMRI beta values and motion (SPSS). Additionally, all volumes with signal exceeding > 4 standard deviations from the mean activity were removed from raw preprocessed time course data prior to extraction of beta values and statistical analysis of fMRI data 19 .

Post-hoc Whole Brain Confirmatory and Supplementary Analysis
A post-hoc whole brain analysis across the 3 groups was conducted to 1) confirm the ROI findings, and 2) potentially inform future research. A whole brain mask excluded all voxels outside of the brain from statistical analysis of the images performed in AFNI (3dANOVA).

Confirmatory Non-significant Relationships between Beta Values and Motion
Post-hoc correlation analyses between motion and ROI fMRI responses revealed no significant

Confirmatory Post-hoc Whole Brain Analyses
Whole brain fMRI data were compared across the 3 groups for both reward anticipatory (GVNant contrast) and non-loss outcome response (NVLout contrast) and generally confirmed the observed group differences from the ROI analyses (extracted beta values). The overall group differences for the whole brain analyses were consistent with the ROI analyses, but no clusters survived correction for multiple comparisons (results reported here are uncorrected only). Uncorrected F-stat images showed widespread group differences in the MPFC for both GVNant and NVLout contrasts (uncorrected p < 0.05). Post-hoc between group whole brain analyses (T-stat images, uncorrected p < 0.05) confirmed these overall findings and the ROI results; these images showed regional (e.g., MPFC) differences during gain anticipation and noloss outcome between non-opioid fibromyalgia and control groups as well as between opioid and non-opioid fibromyalgia groups, but minimal differences between opioid fibromyalgia and control groups ( Supplementary Fig. S1 and S2). Additional post-hoc whole brain analyses for the LVNant (loss versus non-loss anticipation) and GVNout (gain versus no gain outcome) contrasts were also conducted to inform future investigations ( Supplementary Fig. S3 and S4).