The temporal dynamics of emotion dysregulation in prescription opioid misuse

Abstract

Background:

Opioid misuse is theorized to compromise the capacity to regulate positive and negative emotions. Yet, the temporal dynamics of emotion dysregulation in opioid misuse remain unclear.

Methods:

Participants with chronic pain on long-term opioid therapy (N=71) participated in an experiment in which they completed an event-related emotion regulation task while heart rate (HR) and galvanic skin responses (GSR) were recorded over a 5 second emotional picture viewing period. Participants were asked to passively view the images or to regulate their emotional responses via reappraisal (i.e., negative emotion regulation) and savoring (i.e., positive emotion regulation) strategies. Using a validated cutpoint on the Current Opioid Misuse Measure, participants were classified as medication-adherent or opioid misusers.

Results:

Medication-adherent patients were able to significantly decrease GSR and HR during negative emotion regulation, whereas opioid misusers exhibited contradictory increases in these autonomic parameters during negative emotion regulation. Furthermore, GSR during positive emotion regulation increased for non-misusers, whereas GSR during positive emotion regulation did not increase for misusers. These autonomic differences, which remained significant even after controlling for a range of covariates, were evident within 1 second of emotional stimulus presentation but reached their maxima 3–4 seconds later.

Conclusions:

Opioid misuse among people with chronic pain is associated with emotion dysregulation that occurs within the first few seconds of an emotional provocation. Treatments for opioid misuse should aim to remediate these deficits in emotion regulation.

INTRODUCTION

Patients with chronic pain are often treated with long-term opioid therapy (i.e., near daily use of opioids for more than 90 days (Korff et al., 2008). The initiation of opioid therapy following the onset of acute (e.g. surgery or injury) or chronic pain can develop into long-term use and escalate to opioid misuse and opioid use disorder (OUD); as such, long-term opioid therapy is a key vector of the current opioid epidemic (Compton & Jones, 2019). Approximately 25% of individuals suffering from chronic pain misuse their prescribed opioids (Vowles et al., 2015). Elucidating risk mechanisms underlying opioid misuse can improve assessment and intervention efforts, and emotion dysregulation represents one key mechanistic target presumably involved in the development and maintenance of opioid misuse.

Individuals suffering from an inability to effectively regulate positive and negative emotions display increased vulnerability to addiction (Garland, Bell, et al., 2018). Addictive drugs are often used as means of self-medicating negative affective states and compensating for a dearth of positive affect (Khantzian, 1997). In that regard, early onset of affective disorders such as anxiety and depression predicts the development of substance use disorders later in life (Bolton et al., 2009; Robinson et al., 2011). With respect to opioids, specifically, longitudinal research demonstrates that baseline mood and anxiety disorders predict opioid misuse and dependence years later (Martins et al., 2012). While emotional distress is often antecedent to opioid misuse, it might also be a consequence of opioid misuse. Indeed, prolonged, high-dose opioid use in the context of chronic pain is theorized to result in allostatic changes in brain stress and reward systems that amplify negative emotional reactions to stressors and blunt positive emotional responses to naturally rewarding objects and events in the social environment (Koob & Le Moal, 2001; Shurman et al., 2010). Moreover, by virtue of its putative effects on corticolimbic brain networks (Elman & Borsook, 2016), opioid misuse is thought to impair top-down cognitive regulation of autonomic responses during emotional experience (Garland et al., 2013). The neurovisceral integration theory (Thayer & Lane, 2000) posits that during emotion regulation, prefrontal cortical structures modulate limbic activation to fine-tune autonomic efference, manifested in visceral and peripheral physiological markers such as heart rate (HR) and galvanic skin response (GSR).

Insofar as the autonomic nervous system mobilizes energy resources in service of motivated goal attainment (Cannon, 1929; Lang et al., 1998; Lang & Bradley, 2011), HR and GSR fluctuate during emotional experience as the organism approaches appetitive objects and avoids aversive ones. Thus, HR and GSR have been shown in more than a century of psychophysiological research to reliably index emotional reactivity (Bradley et al., 2001; Cannon, 1914; Kreibig, 2010; Lacey, 1967; Lang et al., 1993). As such, conscious, top-down regulation of responses to emotional stimuli should produce changes in HR and GSR relative to conditions in which subjects simply attend and passively react to emotional stimuli. Indeed, GSR and HR have been used to assess emotion regulation for decades (Gross, 1998; Gross & Levenson, 1993, 1997; Jackson et al., 2000; Wegner & Zanakos, 1994) through experimental paradigms in which participants are instructed to consciously decrease their negative emotional response through cognitive strategies like reappraisal (i.e., reframing one’s interpretation of an emotional stimulus to decrease its negative emotional impact; see Gross, 2002) or increase their positive emotional response through strategies like savoring (i.e., focusing on the pleasant features of an emotional stimulus to increase its positive emotional impact; see Bryant & Veroff, 2007). To our knowledge, only one study has used such a task to examine emotion dysregulation in opioid misuse (Garland et al., 2017). In this study of patients on long-term opioid therapy, an emotion regulation task assessed parasympathetic responses (as indexed by high-frequency heart rate variability) over 195-second long stimulus blocks. Opioid misusing chronic pain patients (N=40) exhibited blunted autonomic regulation of positive and negative emotions relative to medication-adherent patients with chronic pain. However, the blocked task design precluded investigation of the temporal dynamics of regulating rapid sympathetic nervous responses during emotional experience.

In contrast, event-related designs can parse changes in autonomic indicators like HR and GSR on the order of seconds (Cacioppo et al., 2007). For example, among healthy individuals completing an event-related affective picture viewing task, differences between emotional and neutral stimuli can be observed in the temporal trajectory of HR and GSR within the first few seconds following stimulus onset (Bradley et al., 2005; Bradley et al., 2001). Among the few studies that have used an event-related design to investigate emotion regulation, Dan-Glauser and Gross (Dan-Glauser & Gross, 2011, 2015) found HR decreases during emotion regulation occurring as early as 1 second after stimulus onset. Conceptually, event-related designs are particularly significant - insofar as temporally dynamic changes in autonomic responses can reveal at what stage of emotion processing a given regulation strategy makes its impact (Thiruchselvam et al., 2011). For instance, emotion regulation might impact early engagement of attention to emotional information occurring within the first 1000 millseconds of stimulus exposure. Alternatively, emotion regulation might modulate later elaborative processing of the emotional stimulus occurring for several seconds or longer. However, to date, no studies have used an event-related emotion regulation task to examine the temporal dynamics of emotion regulation among chronic pain patients receiving LTOT.

Here we employed an event-related design to assess autonomic markers of emotion regulation (i.e., HR and GSR) in a sample of subjects with chronic pain undergoing LTOT to assess the temporal dynamics of emotion dysregulation associated with opioid misuse. We hypothesized that opioid misusers would have difficulties in regulating both positive and negative emotional states compared to medication adherent patients, as evidenced by aberrant psychophysiological responses (indexed by HR and GSR) during the emotion regulation task.

METHOD

Participants and Procedures –

Participants (N=71; see Table I for demographic/clinical characteristics) were recruited from a Veterans Health Care Administration Medical Center in the Intermountain West and met inclusion criteria if they reported having a chronic pain condition and had taken opioid analgesics daily or nearly every day for at least the past 90 days (Chou et al., 2009). Participants were instructed to take their prescribed opioid medication as usual on the day of the laboratory session to prevent cognitive, affective, or autonomic abnormalities due to opioid withdrawal. Following informed consent, participants completed self-report measures of opioid dose, pain severity and location, opioid misuse behaviors, and opioid craving. Then ECG electrodes were attached to the left and right pectoral muscles as well as under the lower left rib and GSR transducers were attached to the pointer and middle fingers of the non-dominant hand, and participants sat quietly for a 5-minute resting baseline period. After this resting baseline, participants then completed an event-related emotion regulation task while ECG and GSR were recorded. The study protocol was approved by the University of Utah IRB and the VA Salt Lake City Healthcare System Research and Development Committee. All procedures complied with standards propounded in the Helsinki Declaration of 1975.

Table I.

Demographic and Clinical Characteristics (N = 71)

Measure	Misusers	Non-Misusers	Test Statistic	p-value
N (%)	42 (59%)	29 (41%)
Female, N (%)	4 (10%)	16 (53%)	χ2 = .40	.82
Age, M ± SD	59.2 ± 11.0	63.2 ± 9.8	t = −1.6	.12
Race, N (%)			χ2 = 3.6	.47
African American	2 (4%)	2 (7%)
Hispanic/Latino	2 (4%)	-
White	34 (83%)	26(90%)
Native American/American Indian	3 (7%)	1(3%)
Asian/Pacific Islander	-	-
Other	1 (2%)	-
Years Education, N (%)	6.1 ± 2.0	5.7 ± 1.7	t = .92	.36
Primary Pain Location, N (%)			χ2 = 1.4	.85
Low back	17 (40%)	10 (34%)
Back	6 (14%)	4 (14%)
Neck	2 (5%)	3 (10%)
Joint	8 (6%)	5 (17%)
Other	7 (20%)	7 (25%)
Missing	2 (5%)	-
Pain Duration (in Years)	19.1 ± 13.1	20.7 ± 13.5	t = .98	.64
Primary Opioid Type			χ2 = .61	.92
Oxycodone	8 (19%)	6 (21%)
Hydrocodone	10 (24%)	6 (21%)
Morphine	4 (9%)	2 (7%)
Tramadol	13 (31%)	10 (34%)
Other	5 (12%)	5 (17%)
Missing	2(5%)	-
Morphine Equivalent Daily Dose, M ± SD	160.1 ± 379.7	92.7 ± 165.9	t = 1.0	.32
Pain Severity (BPI), M ± SD	5.1 ± 1.7	5.7 ± 1.3	t = −1.7	.09
Emotional Distress (DASS) M ± SD	46.2 ± 25.2	25.4 ± 17.3	t = −3.8	< .001
Opioid Misuse (COMM), M ± SD	20.3 ± 6.5	7.6 ± 2.6	t = −9.88	< .001

COMM = Current Opioid Misuse Measure; BPI = Brief Pain Inventory; DASS = Depression Anxiety Stress Scale

Prescription opioid misuse status was determined by scores on the Current Opioid Misuse Measure (COMM; Butler et al., 2007). A COMM score of 13 is suggestive of prescription opioid misuse among chronic pain patients in primary care (Meltzer et al., 2011). Using this validated COMM cut-point, participants were divided into two groups: chronic pain patients who engaged in opioid misuse behaviors (misusers, n=42), and chronic pain patients on long-term opioid therapy at low risk for opioid misuse (non-misusers, n=29).

Emotion Regulation Task –

This study utilized an emotion regulation paradigm (Jackson et al. 2000; Ochsner et al. 2002) that was comprised of 80 trials, each presented for 3 seconds, separated by a fixation cross with an intertrial interval of 500, 750, or 1000 ms. Four experimental task conditions were presented using a randomized, event-related design: “view negative,” “reappraise negative,” “view positive,” and “savor positive.” The “reappraise negative” condition instructed participants to reinterpret the image content to decrease emotional reactions to the image. The “view” conditions instructed participants to passively view the image without trying to change their emotional experience. The “savor positive” condition, consistent with standard “increase positive” instructions on emotion regulation tasks (e.g., Kim & Hamann, 2007), instructed participants to imagine experiencing the positive event occurring in the image, and to focus on the enjoyable aspects of the image and their own positive emotional response to the image. After each trial, participants used a five-button response box to rate the arousal of their current emotional state. Stimuli included positive images (e.g., social affiliation, athletic victories) and negative images (e.g., violence, injury) selected from the International Affective Picture System (Lang et al., 1997). Positive images had significantly higher valence than negative images (t(62) = 33.0, p < .0001, Mean±SD positive = 7.17±0.54; Mean±SD negative = 2.54±0.50), but did not differ in arousal ratings (t(62)=−.55, p = .59. Images of equivalent valence and arousal ratings were presented for the passive viewing and emotion regulation conditions.

Opioid use and misuse—

Morphine equivalent daily opioid dose (MEDD) was obtained by self-report and corroborated by medical chart review. The COMM (α=.81) was used to assess prescription opioid misuse. Participants responded to 17 items rated on a Likert scale (0 = never, 4 = very often) regarding how often in the past 30 days they had engaged in risk behaviors associated with opioid misuse or took opioid medication other than how it was prescribed.

Pain severity—

Pain severity was measured with the four-item pain severity subscale from the Brief Pain Inventory (BPI; α=.87), a well-validated measure widely used to assess acute and chronic pain (Cleeland, 1994) . Response options ranged from 0 (no pain) to 10 (pain as bad as I can imagine).

Emotional distress—

Distress was measured with the 21-item Depression Anxiety Stress Scale (DASS, Lovibond & Lovibond, 1995), which provides a composite score of negative emotional symptoms (α=.94).

Physiological Measures –

The impact of emotion regulation on autonomic responses including GSR and HR can be detected within a few seconds of exposure to an emotional stimulus (Dan-Glauser & Gross, 2011, 2015). In the present study, GSR and HR were assessed with a BIOPAC MP160 system using AcqKnowledge software (BIOPAC, Goleta, CA, USA) and a sampling rate of 2000 Hz. GSR was recorded with Ag/Ag-Cl electrodes attached to the non-dominant middle and pointer fingers, using a constant voltage of 0.5 V and a gain of 2 microsiemens/V. Electrocardiogram (ECG) data was recorded using Ag/Ag-Cl electrodes attached to the participants’ left and right pectoral muscles and the center of the bottom of the lowest left rib.

GSR and HR can be influenced by a host of factors. In that regard, images of equivalence valence and arousal ratings were employed for the passive viewing and emotion regulation conditions so that any observed differences in GSR and HR responses between experimental conditions could be attributed to regulation and not merely result from variability in the valence and arousing qualities of emotional stimuli. Moreover, the experimental conditions were presented during the same laboratory session so that the effects of potential unrelated confounders on GSR and HR (e.g., medication, diet, hormone fluctuations) would be equally distributed across the passive viewing and emotion regulation trials.

Data Reduction –

In order to extract heart rate (in BPM), we first processed the ECG channel using the fmrib_qrsdetect algorithm programmed for EEGLAB (Delorme & Makeig, 2004). Briefly, this algorithm uses combined adaptive thresholding to determine QRS peaks and reject artifacts. Using the artifact corrected RR-intervals, we calculated HR. Raw GSR data were analyzed. For both GSR and HR data, all 5 seconds of each trial were baseline-corrected for activity occurring within the 500 ms pre-stimulus baseline and binned into 500ms epochs. Although grand average waveforms were computed for display purposes, we analyzed epoched data from each individual trial using a multilevel modeling approach.

Statistical Analysis.

Multilevel models were employed for hypothesis testing. Multilevel modeling is pertinent to the aims of the study because it: (a) addresses data dependencies created by the nesting of epochs within trials and trials within participants, and (b) models both trials and participants as random factors (Raudenbush & Bryk, 2002). Baseline adjusted GSR and HR were examined separately under the positive and negative emotion regulation conditions. Each model was inspected for violations of multilevel modeling assumptions related to distributional form, normality of residuals and random intercepts, and influential cases; no violations were identified related to models of GSR and HR. The four multilevel models were of the following form (see Supplementary Materials for explanation of equations):

$$\begin{gathered} L1:Y_{ijk}={\pi }_{0jk}+{\epsilon }_{ijk} \\ L2:{\pi }_{0jk}={\beta }_{00k}+r_{0jk} \\ L3:{\beta }_{00k}={\gamma }_{000}+{\gamma }_{001}MisuseStatus+{\gamma }_{002}Condition+{\gamma }_{003}User*Condition+{\mu }_{00k} \end{gathered}$$

Emotion regulation was assessed by the condition factor in the linear mixed model that represented the passive viewing versus regulation conditions. A significant fixed effect of condition represents a difference in the autonomic dependent variable between the passive viewing and regulation (i.e., reappraisal or savoring) conditions. We also conducted sensitivity analyses to determine whether effects of misuse were robust in the presence of potential confounders. The variables MEDD, gender, distress, and pain severity were entered into each of the four models as grand mean centered covariates and were included to control for the effects of potential confounders on indices of emotional regulation (Garland et al., 2016). MEDD was standardized to make the scaling of this variable comparable to others in the model.

In contrast, we employed an “analysis of covariance” approach (within a linear mixed modeling framework) to handle valence ratings: valence ratings in the savor and reappraise conditions were regressed on misuser status after controlling for valence ratings during each of their respective passive viewing conditions. The analytic approach for these models differed from those for GSR and HR in which condition (view vs. regulation) was represented as a factor in the model because there were no baseline valence values measured in the absence of any stimuli. For that reason, and because misusers and non-misusers differed from one another in the view condition, the values in the view condition were used as baseline scores and were included in analytic models as a mean centered covariate. This provides a test of whether, controlling for differences in the view condition, misusers and non-misusers differ in their valence ratings when asked to reappraise or savor images (Aiken et al., 2002). Valence ratings were available at the level of trial rather than epoch; accordingly, models of valence ratings are two-level models rather than three-level models as for GSR and HR. Examination of the assumption of normality for the valence models indicated that valence ratings in the positive regulation condition were positively skewed. The outcome was, therefore, modeled using gamma regression with a log link. These models were of the basic form provided below. Though not shown in the equation for parsimony, the log link was included for the gamma regression model; the other model parameters maintain interpretations as presented above for the three-level models.

$$\begin{gathered} L1:Y_{ij}={\beta }_{0j}+{\epsilon }_{ij} \\ L2:{\beta }_{0j}={\gamma }_{00}+{\gamma }_{01}MisuseStatus+{\gamma }_{02}ViewValence+{\mu }_{0j} \end{gathered}$$

Multilevel model analyses were conducted using R 3.6.1 (R Core Team, 2019). Graphics were created with the ‘ggplot2’ package (Wickham, 2016). Multilevel models were analyzed with the ‘lme4’ package (Bates et al., 2015) The ‘lmerTest’ package (Kuznetsova et al., 2017) was used to obtain Satterthwaite degrees of freedom and p-values. Analyses of simple main effects were conducted to decompose interactions between condition and user status. Fixed and random effects from each model are reported in Tables 2 (GSR) and 3 (HR), with regulate – view difference waves plotted in Figure 1.

Table II.

Parameter Estimates for Multilevel Models of Baseline Adjusted GSR in the Positive and Negative Regulation Conditions

Fixed Effects	GSR Positive Regulation			GSR Negative Regulation
	Estimate (SE)	CI95a	pb	Estimate (SE)	CI95	p
Intercept	1.00 (0.72)	−0.41 – 2.41	.167	0.42 (0.36)	−0.28 – 1.12	0.238
Condition (Look)	0.98 (0.24)	0.51 – 1.45	<.001	−0.54 (0.21)	−0.96 – −0.12	0.012
User (Non-misuser)	−0.43 (0.93)	−2.25 – 1.41	.652	−0.28 (0.46)	−1.19 – 0.63	0.543
Condition by User	−1.25 (0.31)	−1.86 – −0.64	<.001	1.03 (0.28)	0.48 – 1.58	<0.001

Notes: N_Trial = 16, N_Subject = 71; reference group for fixed effects is provided in parentheses; raw GSR values were multiplied by 100

^aAll CIs are 95% and are bootstrapped values.

^bThe ‘lmertest’ package does not provide p-values for random effects, but lack of overlap with zero in the CIs indicates significance.

^cThe ‘lmertest’ package does not provide standard errors for random effects.

Table III.

Parameter Estimates for Multilevel Models of Baseline Adjusted HR in the Positive and Negative Regulation Conditions

Fixed Effects	HR Positive Regulation			HR Negative Regulation
	Estimate (SE)	CI95a	pb	Estimate (SE)	CI95	p
Intercept	6.03 (1.74)	2.61 – 9.45	.001	5.10 (1.54)	2.09 – 8.11	0.001
Condition (Look)	0.16 (0.68)	−1.18 – 1.50	.813	−3.02 (0.69)	−4.38 – −1.66	<0.001
User (Non-misuser)	2.47 (2.27)	−1.97 – 6.91	.280	−0.48 (2.00)	−4.39 – 3.44	0.812
Condition by User	−1.23 (0.89)	−2.97 – 0.51	.164	4.22 (0.90)	2.45 – 5.99	<0.001

Note: N_Trial = 16, N_Subject = 71; reference group for fixed effects is provided in parentheses; raw HR values were multiplied by 10

^aAll CIs are 95% and are bootstrapped values.

^bThe ‘lmertest’ package does not provide p-values for random effects, but lack of overlap with zero in the CIs indicates their significance.

^cThe ‘lmertest’ package does not provide standard errors for random effects.

Figure 1.

GSR and HR curves by opioid misuse status (Misusers, Non-misusers) are presented here for both Negative (left graphs) and Positive (right graphs) Emotion Regulation. Data (mean ± 1 standard error) represent difference curves between each passive viewing condition (Positive or Negative) and its respective regulation condition (Savor or Reappraise). Note: Neg.= Negative. Pos. = Positive.

RESULTS

Autonomic Indices of Positive Emotion Regulation

For GSR, we observed a significant misuse status by task condition interaction (b = −1.25, s.e. = 0.3, p < .001). Between-group differences in emotion regulation emerged as early as 1 second after stimulus presentation, reached their maxima around 2 second post-stimulus, after which they were maintained for the duration of picture viewing (see Figure 1). Among non-misusers, GSR was significantly higher during the savor positive condition (M= 1.98) than the view positive condition (M = 1.00), b = 0.98, s.e. = 0.24, p < .0001. Among misusers, GSR during the savor positive condition (M = 0.31) did not significantly differ from the view positive condition (M = 0.58), b = −0.27, s.e. = 0.20, p = .179.

For HR, there was no significant misuse status by task condition interaction, nor was there a main effect of condition.

Autonomic Indices of Negative Emotion Regulation

For GSR, we observed a significant misuse status by task condition interaction (b = 1.03, s.e. = 0.28, p < .001). Between-group differences in emotion regulation emerged as early as 1 second after stimulus presentation, but reached their maxima 4–5 seconds post-stimulus (see Figure 1). Among non-misusers, GSR was significantly lower during the reappraise negative condition (M = 0.12) than the view negative condition (M = 0.42), b = −0.54, s.e. = 0.21, p = .012. Among misusers, GSR was significantly higher during the reappraise condition (M = 0.63) than during the view negative condition (M = 0.14), b = 0.49, s.e. = 0.18, p = .006.

For HR, we observed a significant misuse status by task condition interaction (b = 4.22, s.e. = 0.90, p < .001). Between-group differences in emotion regulation emerged as early as 1 second after stimulus presentation, but reached their maxima 3–4 seconds post-stimulus (see Figure 1). Among non-misusers, HR was significantly lower during the reappraise negative condition (M = 2.08) than during the view negative condition (M = 5.10), b = −3.02, s.e. = 0.69, p < .001. Among misusers, HR was significantly higher during the reappraise negative condition (M = 5.82) than during the view negative condition (M = 4.62), b = 1.20, s.e. = 0.58, p = .037.

Sensitivity Analyses for HR and GSR

To determine whether the effects from each of the four models presented above were robust when potential confounders were included in the models, each model was reanalyzed with additional grand mean centered covariates: MEDD, gender, emotional distress, and pain severity. Sensitivity analyses did not change the significance or directionality of any of the significant interactions observed in original models.

Valence Ratings

Valence ratings did not differ by opioid misuse status for either positive (b = 0.02, s.e. = 0.03, p = .468) or negative emotion regulation (b = 0.21, s.e. = 0.22, p = .339).

DISCUSSION

The current study examined the psychophysiology of emotion dysregulation among opioid misusing chronic pain patients relative to medication-adherent patients with chronic pain. We observed significant differences in autonomic responses between the non-misuser (i.e., medication-adherent) and misuser groups during negative and positive emotion regulation. Specifically, medication-adherent patients were able to significantly decrease GSR and HR during negative emotion regulation, whereas opioid misusers exhibited contradictory increases in these autonomic parameters during the attempt to downregulate negative emotions. Furthermore, GSR during positive emotion regulation increased for non-misusers, but GSR during the attempt to upregulate positive emotions did not increase for misusers. Taken together, these data demonstrate that opioid misuse is marked by an inability to regulate both positive and negative emotional responses that is manifest in visceral and peripheral physiology.

Study results are in line with those from a previous investigation in our lab using a blocked version of the emotion regulation task, where opioid misusers demonstrated blunted high frequency heart rate variability (HRV) during negative and positive emotion regulation relative to medication-adherent patients (Garland et al., 2017). The present study replicates and extends these findings by revealing the temporal dynamics of emotion regulation deficits via GSR in addiction to cardiac-autonomic responses. Whereas HR is a summation of both sympathetic and parasympathetic influences, GSR is considered a pure measure of sympathetic nervous activity (Berntson et al., 2007; Dawson et al., 2007) and GSR magnitude is positively associated with emotional arousal (Lang et al., 1993). In the present study, GSR data indicate that medication-adherent patients were able to decrease sympathetic arousal to negative emotional stimuli via reappraisal. In contrast, opioid misusers evidenced increased sympathetic arousal during reappraisal, suggesting negative emotion regulation is effortful and may increase rather than reduce arousal. Elsewhere, we reported deficits in self-reported reappraisal among opioid misusing chronic pain patients that were associated with opioid craving (Garland, Hanley, et al., 2018), suggesting that opioid misusers may crave opioids as a means of self-medication when they are unable to successfully regulate negative emotions.

When medication-adherent patients engaged in positive emotion regulation, they successfully increased GSR. This heightened sympathetic arousal is indicative of increased reward responsiveness. In contrast, opioid misusers were unable to increase GSR during positive emotion regulation, indicating the presence of blunted reward responsiveness. These findings converge with our earlier psychophysiological work (Garland et al., 2017) and data from a recent large scale survey demonstrating heightened subjective anhedonia among opioid misusing pain patients (Garland et al., 2019). Taken together, these studies provide support for allostatic models for prescription opioid misuse (Elman & Borsook, 2016; Garland et al., 2013; Shurman et al., 2010), which assert that prolonged, high dose opioid use shifts set points in brain reward circuitry (e.g., orbitofrontal cortex and ventral stratum), resulting in an insensitivity to natural reward that drives opioid dose escalation as a means of obtaining hedonic equilibrium.

To our knowledge, this is the first study to examine psychophysiological responses during an event-related emotion regulation task in a population of chronic pain patients on long-term opioid therapy. The use of an event-related design allowed us to elucidate the temporal dynamics of emotion regulation in medication-adherent and opioid misusing patients. In medication-adherent patients, we observed autonomic activation decreases during reappraisal and increases while savoring. These effects were evident within the first second of emotional stimulus presentation, but reached their maxima several seconds later. With respect to downregulation of negative emotions, the observed pattern was consistent with that exhibited by healthy controls participating in a prior study (Dan-Glauser & Gross, 2011). However, among misusers, we observed an increase in sympathetic activation during reappraisal, suggesting that the attempt to downregulate negative emotions may have “backfired” by increasing physiological arousal. These results converge with findings from a study by Di Simplicio et al. (2012), where highly neurotic subjects exhibited reduced high-frequency HRV, an indicator of reduced parasympathetic regulation of sympathetic responses, during reappraisal relative to subjects with low levels of neuroticism. Moreover, the observed differences in autonomic responding remained significant even after controlling for opioid dosing, suggesting that opioid misuse, and not long-term opioid therapy by itself, may occasion deficits in emotion regulation.

We did not observe a convergence of psychophysiological responses and self-reported emotion regulation (i.e. valence ratings at end of trials). This discordance could be due to denial or a lack of self-awareness of problematic misuse, attributed to dysfunction in the anterior cingulate cortex (ACC) and prefrontal cortex (PFC), that is seen in over 80% of drug misusers (Goldstein et al., 2009). Indeed, denial and lack of self-awareness are associated with impaired executive function and mental speed in alcoholic patients, perhaps influencing the time-limited self-report used in our trials (Rinn et al., 2002). Furthermore, the ACC and PFC are integral to top-down regulation of negative emotional reactions (Ochsner & Gross, 2005), meaning the ‘backfire’ of psychophysiological response to negative emotion regulation we observed in misusers could be attributed to deficits in top-down regulation supported by these regions, the effects of which might also impair self-awareness or denial.

Due to the cross-sectional nature of this study, we do not know whether emotion dysregulation is a cause, correlate, or consequence of opioid misuse. In that regard, individuals suffering from preexisting emotion dysregulation distress may turn to opioids as a means of self-medicating a surfeit of negative affect or deficit of positive affect (Khantzian, 1997). In that regard, epidemiological studies of opioid misuse and dependence show that mood and anxiety disorders often predate opioid misuse (Green et al., 2011; Martins et al., 2012). Trait-like differences in the propensity to experience emotional distress (as a result of chronic pain and/or comorbid psychopathology) might explain emotion dysregulation among opioid misusers. Therefore, in the current study, we controlled for emotional distress in sensitivity analyses, and yet opioid misusers continued to exhibit autonomic indices of emotion dysregulation. While these analyses cannot provide causal inferences, they do argue against the notion that the observed emotion dysregulation is a mere epiphenomenon of the mood disturbances that often predate opioid misuse. Similarly, results from sensitivity analyses suggest that the emotion dysregulation we observed among opioid misusers is not simply the pharmacologic effect of opioid exposure nor is it a function of chronic pain, but rather represents a specific deficit in the capacity to exert top-down regulation of emotional responding. It is quite likely that trait deficits in emotion regulation constitute a diathesis that increases susceptibility to opioid misuse, which in turn exacerbates emotion dysregulation and propels a downward spiral leading to compulsive opioid use and OUD (Garland et al., 2013) – a hypothesis to be tested by future longitudinal studies. Moreover, the study was limited by the fact that we characterized patients as opioid misusers based on their score on a self-report instrument (the COMM). Because some patients may be reluctant to admit to opioid misuse on a self-report measure, reporting bias is certainly possible. For this reason, we may not have been able to correctly classify all the misusers in the sample. Nonetheless, participants were assured of confidentiality of the data collected in the study, and it appeared that the majority endorsed opioid misuse on the COMM. Future studies should triangulate self-reports of opioid misuse with clinical interviews and review of data from legally-mandated prescription monitoring programs. Finally, though the sample size was moderate for a psychophysiological study of this nature, larger samples would have greater statistical power to detect subtle differences in autonomic responding during emotion regulation.

In conclusion, the findings from the current study add to the body of literature underscoring the role of emotion dysregulation in opioid misuse both in and outside of the context of chronic pain. Furthermore, it extends the literature for the first time to document the temporal dynamic course of autonomic nervous responses during emotional regulation. Given the observed association between opioid misuse and emotion dysregulation, interventions that aim to strengthen emotion regulation might be especially efficacious treatments for opioid misuse. In that regard, Mindfulness-Oriented Recovery Enhancement (MORE; Garland, 2013), a behavioral intervention rooted in affective neuroscience that uses mindfulness training to potentiate reappraisal and savoring, has been shown to significantly decrease opioid misuse among chronic pain patients in two randomized controlled trials (Garland et al., 2019; Garland et al., 2014) by strengthening emotion regulation – and in particular, by enhancing positive emotion regulation (Garland et al., 2017; Garland et al., 2019). Future psychophysiological research should examine the extent to which MORE and other interventions for opioid misuse produce their therapeutic effects by enhancing emotion regulatory capacity in misusers.

Highlights.

• Opioid misuse is theorized to involve emotion dysregulation.

• Misusers demonstrated autonomic dysfunction during emotion regulation.

• Non-misusers demonstrated effective emotion regulation.

• Emotion dysregulation in opioid misusers is evident within one second of stimulus presentation.