Heavy marijuana use but not deprivation is associated with increased stressor reactivity

Abstract

Although stressors appear to motivate marijuana use and marijuana use, in turn, is believed to induce stress system neuroadaptations, relatively little empirical work has explicitly tested for stress neuroadaptations associated with heavy marijuana use. We examined stressor reactivity to threat of unpredictable electric shock via startle potentiation among heavy marijuana users and a control group that reported minimal history of marijuana use. Heavy marijuana users were randomly assigned to 3 days of marijuana deprivation or no deprivation.This design allowed us to test contrasts for heavy (vs. minimal) use and deprivation (vs. no deprivation) on stressor reactivity. Heavy marijuana users (both deprived and non-deprived) displayed increased startle potentiation during threat of unpredictable electric shock relative to minimal use controls. In contrast, marijuana deprivation had no effect on startle potentiation. Startle potentiation was also increased among users who reported greater stress coping motives for their marijuana use and users with cannabis use disorder diagnoses. To our knowledge, this is the first study to demonstrate increased reactivity to unpredictable stressors among heavy marijuana users. However, comparable increased unpredictable stressor reactivity among patients with alcohol and other substance use disorders has been previously documented. This relationship to heavy marijuana use is consistent with predictions from rodent addiction models regarding stress neuroadaptations following heavy, chronic drug use, but could also represent an etiologically relevant pre-morbid risk characteristic. Finally, the clinical import of unpredictable stressor reactivity is reinforced by its relationships with stress coping motives and cannabis use disorder status.

Marijuana, the most commonly used illicit drug, has been perceived as less harmful than other substances (Nutt, King, Saulsbury, & Blakemore, 2007). However, approximately half of those who use marijuana daily become addicted (Anthony, 2006), calling for increased efforts to elucidate causes and consequences of heavy, frequent use. In this study, we examine the relationship between heavy marijuana use and response to stressors, given the key role stressors play in addiction etiology. Although stressors appear to motivate marijuana use (Hyman & Sinha, 2009), and marijuana use, in turn, is believed to induce stress system neuroadaptations (Fox, Tuit, & Sinha, 2013; Koob & Le Moal, 2008), relatively little empirical work has explicitly tested for stress neuroadaptations among heavy marijuana users.

Stressors figure prominently in addiction etiology and relapse across drug types (Kaye, Bradford, Magruder, & Curtin, 2017); specifically, heavy and/or chronic drug use contributes to heightened behavioral and affective responses to stressors via central nervous system (CNS) mechanisms including corticotrophin-releasing factor (CRF) and norepinephrine (NE) in the central extended amygdala (Koob & Le Moal, 2008). These stress neuroadaptations in turn result in dysregulated emotional responses to stressors upon cessation of drug use, which provides potent motivation to resume use via craving, increasing relapse risk (Baker, Piper, McCarthy, Majeskie, & Fiore, 2004) across major classes of addictive drugs.

Of particular relevance for marijuana, the endocannabinoid system also regulates CNS response to stressors via modulation of CRF, NE, serotonin, and other stress hormones (Volkow, Hampson, & Baler, 2016). Additionally, cannabinoid receptor agonists and Δ⁹-tetrahydrocannabinol (THC) activate peripheral stress mechanisms via the hypothalamic-pituitary-adrenal (HPA) axis in humans (D’Souza et al., 2004) and animals (Martín-Calderón et al., 1998), during both acute administration and withdrawal (Rodríguez de Fonseca, Carrera, Navarro, Koob, & Weiss, 1997). Repeated activation of the endocannabinoid system is thought pivotal to both CNS and peripheral stress system adaptations occurring in heavy marijuana users (Fox et al., 2013; Volkow et al., 2016), but empirical evidence of stress neuroadaptations in human marijuana users is lacking. Thus, the first aim of the present study was to test for this predicted increase in stressor reactivity among heavy marijuana users relative to minimal use controls, consistent with this Stress Neuroadaptation Model.

Tonic alterations in mood and/or affective response have been clearly documented during withdrawal from marijuana. The robust marijuana withdrawal syndrome as observed in humans (Budney & Hughes, 2006) and animals (Lichtman & Martin, 2002) involves increased negative affective symptoms (irritability, anger or aggression, nervousness or anxiety) (American Psychiatric Association, & DSM-5 Task Force, 2013; Budney & Hughes, 2006; Budney, Hughes, Moore, & Vandrey, 2004; Crippa et al., 2009). These negative affective symptoms during withdrawal are reliably reduced or eliminated by the administration of cannabinoid receptor agonist, the main psychoactive component in marijuana. Furthermore, individuals with a marijuana use disorder report this anxiety and negative affect contributes to difficulty abstaining from use, and cite relief from withdrawal as a motivation for use (Bonn-Miller & Moos, 2009; Budney et al., 2004; Budney & Hughes, 2006; Coffey et al., 2002; Copersino et al., 2006). However, it remains unclear if these affective symptoms result from stress neuroadaptation mechanisms that are enhanced during withdrawal. Therefore, a second aim was to test if marijuana deprivation increases stressor reactivity among heavy users, consistent with a withdrawal-related impact on stress neuroadaptation mechanisms.

If heavy marijuana use produces stress neuroadaptations that contribute to the etiology and maintenance of cannabis use disorder, individuals who use marijuana to cope with stressors or negative affect may be at heighted risk. Mounting evidence suggests that stress-coping motives are common among marijuana users (Fox et al., 2013; Green, Kavanagh, & Young, 2003; Hathaway, 2003; Hyman & Sinha, 2009), and are associated with DSM-5 Cannabis Use Disorder, depression and perceived stress (Maitra, Christopher, Anderson, & Stein, 2015). More than half of experienced users report blowing off steam, feeling less anxious, and forgetting worries as important reasons for using marijuana (Hathaway, 2003), while nearly three-quarters of daily users report using marijuana to relax or relieve tension (Johnston & O’Malley, 1986). Moreover, heavy marijuana users endorse coping-related motives for their use to a greater extent than do infrequent users, which may itself result from stress neuroadaptations (Hyman & Sinha, 2009). Therefore, a third aim of this study was to conduct exploratory analyses to determine if marijuana use motives and cannabis use disorder diagnostic status were related to stressor reactivity.

To accomplish these three goals, we quantified stressor reactivity via startle potentiation to unpredictable shock threat, a non-invasive, psychophysiological index of stressor reactivity that is sensitive to putative stress neuroadaptations associated with drug use (Davis, Walker, Miles, & Grillon, 2010; Hogle, Kaye, & Curtin, 2010; Moberg, Bradford, Kaye, & Curtin, 2017). We chose to use unpredictable shock threat because of translational evidence that implicates CRF and NE mechanisms in the central extended amygdala specifically in response to unpredictable (vs. predictable) threats (Davis et al., 2010; Kaye et al., 2017). We assessed startle potentiation to these unpredictable threats among deprived (three days of abstinence from marijuana) and non-deprived heavy marijuana users and a control group on individuals with minimal history of marijuana use. This allowed us to evaluate differences in mean stressor reactivity across Heavy Marijuana User (deprived and non-deprived heavy users vs. minimal use controls) and Deprivation (deprived vs. non-deprived heavy users) contrasts. Furthermore, we assessed individual differences in characteristics regarding marijuana use to evaluate relationships between these individual differences and stressor reactivity.

Method

Study procedures were reviewed and approved by the University of Wisconsin-Madison’s Institutional Review Board (protocol: SE-2008-0164). We also support emerging open science guidelines (Nosek et al., 2015) providing the data, analysis scripts, questionnaires, and other study materials associated with this report publicly via Open Science Framework at this URL: https://osf.io/sgx5c.

Participants

We used flyers and online advertisements to recruit 104 heavy marijuana users (52 female) who reported marijuana use ≥5 days per week, ≥twice per day on days when used for ≥6 months (Arnone et al., 2008; Pope & Yurgelun-Todd, 1996) and 52 minimal use controls (26 female) who reported a history of lifetime marijuana use >1 but <50 times and no use in the past month. Participants who reported any recreational drug use (other than alcohol, marijuana, and nicotine) resulting in intoxication once or more per week or with current diagnoses of alcohol or other drug dependence (other than marijuana or nicotine) were excluded from all groups. Paricipants reporting use of psychotropic medication(s), engagement in psychological treatment (within the last 6 months) or current or past diagnosis of psychotic disorders were also excluded from all groups.

A priori power analyses for sample size planning demonstrated that N=156 provided approximately 86% power to detect a moderate sized between subjects effect (partial eta² =.06) with a two-tailed alpha of 0.05 within the general linear model.

Procedure

Preliminary screening was accomplished when prospective participants called the laboratory to indicate interest in the study. Prospective participants were informed on the phone about the protections of the NIH Certificate of Confidentiality associated with this study, and provided verbal consent to assess their medical and drug use history. Those meeting preliminary eligibility criteria during the phone contact were scheduled for a subsequent formal screening session in our laboratory. During this screening session, all participants provided informed consent, and a study clinician verified inclusion/exclusion criteria by administering a standardized structured interview to assess their medical history, past and current drug use history (using the alcohol and drug use disorders components of the Mini-International Neuropsychiatric Interview (MINI) adapted to assess past use; Sheehan et al., 1998), and marijuana use patterns. All participants were asked to provide a urine sample to verify drug use. Minimal use controls were required to test negative on an immediate-result qualitative test of marijuana use using a cut-off score of 50 ng/ml (Key Operated Drug Test Cup; Drug Test Systems, Dover, NH). Urine samples from heavy users were collected for quantitative comparison analysis.

Heavy users were randomly assigned to one of two Deprivation Groups (deprived or non-deprived) in this screening session. Participants assigned to the deprived group were instructed to abstain from any marijuana use (e.g., smoked, ingested) for three days prior to their experimental session (described below). Those assigned to the non-deprived group were instructed to maintain their typical frequency and quantity of marijuana use. However, they were instructed to refrain from using marijuana for at least one hour before the experimental session to avoid acute intoxication effects in that session. All participants were asked to avoid using alcohol for 24 hours and other recreational drugs (excluding marijuana) for one week prior to their experimental session. Those reporting use of nicotine were asked to continue their usual quantity and frequency of use. Minimal use controls were also asked to continue to avoid marijuana use. Participants scheduled their experimental session to occur within 4–10 days of the screening session.

On arrival for the experimental session, participants reported the date/time of their last marijuana use and any recent (i.e., past 24 hour) alcohol or other (one week) drug use. Breath alcohol concentration (BAC) was also assessed. Deprived participants who were non-compliant with abstinence instructions based on urine drug test or their self-report were dismissed and given the opportunity to reschedule once. Any participant indicating non-compliance with the alcohol and/or other drug use requirements was dismissed¹.

We measured participant’s subjective shock tolerance to a series of 200 ms electric shocks of increasing intensity (7 mA maximum) using standard procedures (Curtin, Patrick, Lang, Cacioppo, & Birbaumer, 2001; Hefner & Curtin, 2012; Hefner, Moberg, Hachiya, & Curtin, 2013). We administered electric shocks using a custom shock stimulator (Bradford, Magruder, Korhumel, & Curtin, 2014) via stainless steel electrodes across the distal phalanges of the index and ring fingers of the left hand. The procedure was discontinued once participants reached the maximum level of shock they could tolerate. We set shock intensity during the main task to each participant’s subjective maximum tolerance threshold to minimize individual differences in shock tolerance.

Shock Threat Manipulation

Participants were administered unpredictable shocks in the context of a reward decision-making task² in which they pressed buttons to indicate preferred rewards. This task was designed to address other questions that are not a focus of the present investigation (Hefner, Starr, & Curtin, 2016). Participants completed the task in alternating blocks of unpredictable ‘Unpredictable Shock Threat’ or ‘No-Shock.’ In Unpredictable Shock Threat blocks, participants were instructed that electric shocks were unpredictable and could occur at any point during the block, while in No-Shock blocks, shocks would never be administered. To increase manipulation salience, the experimenter verbally identified block type and reiterated shock contingencies at the beginning of each block. Finally, a block label (e.g., “NO SHOCKS”) remained on the screen in the top left corner throughout the block. Participants completed 4 total blocks of trials (2 each) in one of four between-subjects counterbalanced orders. A total of 5 shocks were administered throughout the main session; 3 in the first unpredictable shock threat block, and 2 in the second. Similar procedures have been used successfully and reliably produce robust startle potentiation during Unpredictable Shock Threat blocks relative to No-Shock blocks (Hefner & Curtin, 2012; Hefner et al., 2013; Hogle et al., 2010; McCarthy, Gloria, & Curtin, 2009).

Measures

Stable Individual Difference Measures

The MINI was used to diagnose marijuana or other drug abuse and dependence, and was adapted to assess past abuse and dependence in all participants. Participants also self-reported any history of 1) anxiety, 2) depression or 3) other psychiatric diagnoses or treatment. All participants completed the Depression Anxiety and Stress Scales-21 (DASS-21; Lovibond & Lovibond, 1995), adapted to assess symptoms over the past 30 days rather than past week. Cronbach’s alphas were 0.65, 0.82, and 0.80 for Anxiety, Depression, and Stress subscales, respectively, in this sample. All participants completed the Marijuana Motives Measure (MMM; Simons, Correia, Carey, & Borsari, 1998) to assess motivations for using marijuana. Cronbach’s alphas were 0.80, 0.92, 0.86, 0.93, 0.83 for Conformity, Stress Coping, Enhancement, Expansion, and Social subscales, respectively, in this sample. Heavy marijuana users reported their approximate quantity of weekly use.

Deprivation-related Manipulation Checks

All participants completed the Marijuana Withdrawal Checklist (MWC; Budney, Novy, & Hughes, 1999; Vandrey, Budney, Moore, & Hughes, 2005) and Marijuana Craving Questionnaire (MCQ; Heishman, Singleton, & Liguori, 2001) at both screening and experimental sessions in order to assess changes in withdrawal symptoms and craving. Cronbach’s alphas were 0.52 and 0.86 for the MWC and 0.88 and 0.89 for the MCQ at screening and experimental sessions, respectively. Heavy marijuana users also reported the time of their last marijuana use.

Startle Response

Electromyographic activity in the orbicularis oculi muscle was sampled at 2000Hz with a bandpass filter (0.05–500 Hz) from electrodes placed under the right eye according to published guidelines (Blumenthal et al., 2005; van Boxtel, 1998). Eyeblink startle response was measured in response to startle-eliciting noise probes (50ms of 102dB white noise with near instantaneous rise time). Twenty-four noise probes were presented during a subset of trials during unpredictable and no-shock blocks). Data reduction and processing followed published guidelines (Blumenthal et al., 2005). We performed offline processing in Matlab using EEGLab (Delorme & Makeig, 2004) and PhysBox plugins (Curtin, 2011). This processing included high pass filtering (28Hz, 4th order Butterworth, zero phase shift), rectification, smoothing (30Hz, 2nd order Butterworth low pass filter, zero phase shift), epoching (−50 to 250 milliseconds surrounding probe), and baseline correction. Startle magnitude was scored as the peak response between 20–100 milliseconds post-probe onset. We rejected trials containing artifact consistent with standard practices from our laboratory (Kaye, Bradford, & Curtin, 2016). This included trials with deflections greater than ±20μV in the 50ms pre-probe baseline (i.e., unstable baseline) and trials with mean activity≤ −10μV between 150–250 milliseconds post probe onset (i.e., baseline over-correction due to pre-epoch artifact). Startle potentiation was calculated as the increase in startle magnitude during unpredictable shock relative to no-shock blocks. We also calculated General Startle Reactivity as mean startle magnitude to six noise probes presented prior to the start of the task. Following previous recommendations (Bradford, Kaye, & Curtin, 2014), we use General Startle Reactivity as a covariate in all analyses of startle potentiation to increase power to detect focal effects.

Urinalysis of Marijuana Use

Urine samples from heavy users were obtained at both screening and experimental sessions for quantitative analysis to verify compliance with the abstinence instructions for the deprived users. A specimen ratio of creatinine-normalized samples at two different time points (Creatinine-Normalized Specimen 2 / Creatinine-Normalized Specimen 1) was used to detect recent use among marijuana users assigned to abstain from marijuana (Huestis & Cone, 1998). Following previously established procedures and cut scores to detect recent drug use among abstaining individuals, deprived users with specimen ratios exceeding 1.5 were considered non-compliant with the abstinence requirement. (Huestis & Cone, 1998; Manno, Ferslew, & Manno, 1984). This specimen ratio has not been used previously to confirm continued use among regular (non-abstinent) marijuana users and no information about its sensitivity or specificity for that purpose exists to our knowledge. Therefore, we limited its use to confirm no recent use among deprived heavy users; however, we collected and calculated specimen ratios in all heavy users to allow quantitative comparisons across groups as a manipulation check.

Results

Data analysis and figure preparation were accomplished with R (R Core Team, 2015) within R-Studio using the lmSupport (Curtin, 2015) package. We use general linear models (GLMs) for all analyses unless otherwise noted. Marijuana Group effects were analyzed with two planned orthogonal contrasts: the Heavy Marijuana User contrast (deprived and non-deprived heavy users vs. minimal use controls) and the Deprivation contrast (deprived vs. non-deprived heavy marijuana users). We report 95% confidence intervals for raw parameter estimates (b) and also partial eta² ( ${\eta }_p^2$) from the GLMs to document effect sizes.

Group Comparisons on Affect and Marijuana-Relevant Individual Differences

Means, standard deviations, and p-values for the two Marijuana Group contrasts for the affect and marijuana-relevant individual difference measures are reported in Table 1. None of the Deprivation contrasts were significant for any of the individual difference measures, as expected given that heavy users were randomly assigned to the deprived and non-deprived groups. Significant Heavy Marijuana User contrasts were observed for all individual difference measures. On the DASS-21, heavy marijuana users reported higher Depression [ ${\eta }_p^2=0.04$, b=2.6, 95% CI=[0.6,4.5], t(153)=2.56, p=0.011], Anxiety [ ${\eta }_p^2=0.09$, b=3.3, 95% CI=[1.6,5.1], t(153)=3.84, p<.001], and Stress [ ${\eta }_p^2=0.05$, b=3.2, 95% CI=[0.9,5.4], t(153)=2.77, p=0.006] than controls in the past month. On the MMM, heavy marijuana users endorsed greater Stress coping [ ${\eta }_p^2=0.36$, b=4.7, 95% CI=[3.7,5.7], t(153)=9.37, p<.001], Enhancement [ ${\eta }_p^2=0.40$, b=5.5, 95% CI=[4.4,6.5], t(153)=10.02, p<.001], Expansion [ ${\eta }_p^2=0.31$, b=5.9, 95% CI=[4.5,7.4], t(153)=8.28, p<.001], and Social Motives [ ${\eta }_p^2=0.33$, b=4.7, 95% CI=[3.6,5.8], t(153)=8.62, p<.001], and less Conformity [ ${\eta }_p^2=0.03$, b=−0.7, 95% CI=[−1.4,−0.1], t(153)=2.16, p=0.032], than controls. Finally, heavy marijuana users were positive for marijuana abuse and dependence diagnoses (assessed via the MINI) at higher rates than controls [χ ²(4)=77.9, p<.001]

Table 1.

Participant Characteristics by Marijuana Group

	Deprived	Non-Deprived	Control	Deprivation contrast	Heavy User contrast
Age	22.3(3.6)	22.1(3.3)	21(2.9)	.743	.030*
Sex				1.0	1.0
Female	50%	50%	50%
Male	50%	50%	50%
Race				.524	.082
Asian	0%	1.9%	5.8%
Black	19.2%	23.1%	11.5%
White	80.8%	75%	82.7%
Hispanic				.674	1.0
No	92.3%	96.2%	94.2%
Yes	7.7%	3.8%	5.8%
Depression Anxiety Stress Scales
Anxiety	6.9(5.2)	6.8(5.8)	3.5(4.3)	.939	<.001***
Depression	6.2(5.8)	6.7(7.1)	3.8(4.5)	.665	.011*
Stress	10.4(7.4)	9.7(6.9)	6.9(5.7)	.599	.006**
Self-report of previous anxiety diagnosis				1.0	.536
No	96.2%	98.1%	100%
Yes	3.8%	1.9%	0%
Self-report of previous depression diagnosis				.674	.580
No	96.2%	92.3%	90.4%
Yes	3.8%	7.7%	9.6%
Self-report of any previous psychiatric diagnosis or treatment				.625	.611
No	82.7%	76.9%	84.6%
Yes	17.3%	23.1%	15.4%
Young Adult Alcohol Problems Scale	4.3(3.0)	5.3(3.9)	4.2(2.7)	.138	.230
Alcoholic drinks per week	10.1(10.7)	9.8(12.6)	9.0(9.4)	.883	.631
Cigarettes per week	22. 9(40.5)	19.4(32.1)	0.0(0.0)	.558	<.001***
Smoker status				.290	<.001***
Non-smoker	32.7%	44.2%	88.5%
Former smoker	9.6%	1.9%	1.9%
<daily smoker	23.1%	19.2%	9.6%
Daily smoker	34.6%	34.6%	0.0%
Marijuana Use Motives Measure
Conformity	0.7(1.4)	0.8 (1.7)	1.5 (2.5)	.762	.032*
Enhancement	11.1(2.6)	11.9 (2.3)	6.0 (4.3)	.190	<.001***
Expansion	7.6 (4.7)	8.2 (4.8)	1.9 (3.0)	.487	<.001***
Social	7.1 (3.5)	7.9 (3.6)	2.8 (2.5)	.193	<.001***
Stress coping	5.4 (3.6)	5.8 (3.1)	0.9 (2.0)	.490	<.001***
Cannabis Use Disorder Diagnosis(current)				.204	<.001***
No diagnosis	25.0%	11.5%	90.4%
Marijuana abuse	25.0%	30.8%	9.6%
Marijuana dependence	50.0%	57.7%	0.0%

NOTES:

N=156; Table entries are mean (standard deviation) unless otherwise indicated.

p-values for the Deprivation contrast and Heavy Marijuana User contrasts are reported in the two columns to the right.

p-values are from GLM for quantitative variables; chi-squared tests for categorical variables.

^*p<.05;

^**p<.01;

^***p<.001

Marijuana Deprivation Manipulation Checks

Means, standard deviations, and p-values for the Deprivation contrast on self-report and biological confirmation of marijuana use, marijuana withdrawal, and marijuana craving are reported in Table 2. Deprived and non-deprived heavy marijuana users reported comparable marijuana use at the screening session, as expected by random assignment, ${\eta }_p^2<0.01$, b=0.2, 95% CI=[−2.6,3.1], t(102)=0.17, p=0.866. Deprived marijuana users reported greater time since last marijuana use (M=3.7 days, SD=0.8, Range=3–8) relative to non-deprived users (M=0.5, SD=0.4, Range=0–2) at the experimental session [ ${\eta }_p^2=0.86$, b=3.2, 95% CI=[3.0,3.5], t(102)=24.97, p<.001], with descriptive statistics confirming that both groups adhered to instructions. Biological verification of deprivation status via urinalysis confirmed that deprived marijuana users displayed lower creatinine-normalized THC at the experimental session (controlling for their baseline values at screening) [ ${\eta }_p^2=0.26$, b=−0.9, 95% CI=[−1.2,−0.6], t(102)=6.02, p<.001], and a lower specimen ratio (experimental / screening session ratio) [ ${\eta }_p^2=0.20$, b=−179.6, 95% CI=[−251.5,−107.7], t(101)=4.95, p<.001] relative to non-deprived users. Furthermore, no deprived user’s specimen ratio exceeded the 1.5 cut-point advocated by Huestis & Cone (1998) for detecting recent use. Deprived users reported higher scores than non-deprived users on the total score [ ${\eta }_p^2=0.15$, b=4.2, 95% CI=[2.2,6.3], t(101)=4.19, p<.001] and negative affect subscale [ ${\eta }_p^2=0.09$, b=0.8, 95% CI=[0.3,1.4], t(101)=3.09, p=0.003] of the MWC at the experimental session, controlling for their respective baseline scores at screening. Deprived users also reported higher marijuana craving than non-deprived users at the experimental session, controlling for baseline craving at screening [ ${\eta }_p^2=0.05$, b=4.9, 95% CI=[0.7,9.2], t(101)=2.29, p=0.024].

Table 2.

Marijuana Deprivation Manipulation Checks

	Deprived	Non-Deprived	Deprivation contrast
Average marijuana use(grams/week from screening session)	7.2(7.8)	7.0(6.8)	.866
Days since last marijuana use(experimental session)	3.67(0.82)	0.46(0.43)	<.001***
Urinalysis
Creatinine-Normalized THC(experimental session)a	123(158)	303(202)	<.001***
Specimen ratio (experimental/screening session)	0.4 (0.2)	1.3 (1.1)	<.001***
Marijuana Withdrawal Checklist(experimental session)
Total scorea	7.4(6.5)	3.1(3.3)	<.001***
Negative affect subscalea	1.3(1.8)	0.5(0.6)	.003**
Marijuana Craving Questionnaire(experimental session)a	4.5(0.7)	4.2(0.6)	.024*

NOTES:

N=104; Table entries are mean (standard deviation).

p-values for the Deprivation contrast are reported in the right column.

Analyses and descriptive statistics for creatinine-normalized THC and the scales from marijuana withdrawal and craving measures control for baseline scores on these measures from the screening session.

^*p<.05;

^**p<.01;

^***p<.001

Marijuana Group Effects on Startle Potentiation

Startle Potentiation³ was analyzed in a GLM with between-subjects regressors for the Heavy Marijuana User and Deprivation planned orthogonal contrasts. General Startle Reactivity (mean-centered) and Order (order 1 vs. order 2) were included as interactive covariates. As expected, significant non-zero startle potentiation was observed, ${\eta }_p^2=0.49$, b=37.8, 95% CI=[31.4,44.2], t(141)=11.69, p<.001 such that startle response was significantly potentiated during unpredictable shock threat as compared to no-shock blocks. The Heavy Marijuana User contrast was significant, ${\eta }_p^2=0.03$, b=15.0, 95% CI=[1.5,28.5], t(141)=2.20, p=0.029, indicating that heavy marijuana users displayed approximately 15μV greater startle potentiation to threat of unpredictable shock as compared to controls (Figure 1). The Deprivation contrast was not significant, ${\eta }_p^2<0.01$, b=−1.4, 95% CI=[−17.2,14.4], t(141)=0.18, p=0.861, indicating comparable startle potentiation across deprived and non-deprived heavy users.

Figure 1. Startle Potentiation by Marijuana Group.

This figure displays the predicted values (i.e., means) for startle potentiation to unpredictable shock threat by Marijuana Group from the general linear model (GLM). Error bars respresent ±1 standard error for these predicted values. This GLM adjusted for all covariates.

Figure © 2017 Kathryn Hefner, Mark Starr, and John Curtin, under Creative Commons Attribution 4.0 International Public License CC-By

Supplemental robustness analyses evaluated the above reported significant Heavy Marijuana User contrast while controlling for all non-marijuana use relevant individual differences listed in Table 1 (i.e., age, sex, race, Hispanic ethnicity, DASS scales, self-report of anxiety, depression, or any psychiatric diagnosis or treatment, drinks/week, cigarettes/week, smoker status) by including each of these individual differences as a covariate in separate analyses. Cannabis use disorder diagnoses and Marijuana Use Motives were not considered as covariates in these analyses because variance on these measures is inherent to group membership in the heavy user groups (Miller & Chapman, 2001). The Heavy Marijuana User contrast remained significant across all models controlling for these covariates (individual p-values for this contrast ranged from .022–.047).

Marijuana-Relevant Individual Differences and Startle Potentiation

To clarify the observed significant Heavy Marijuana User contrast, we conducted follow-up analyses of marijuana-relevant individual differences measures (i.e., marijuana use motives, diagnostic status). Each of these individual difference variables was examined in separate GLMs that included General Startle Reactivity and Order as interactive covariates per the primary analysis of Marijuana Group reported above. Given the exploratory nature of these follow-up analyses, we report both raw and false discovery rate (FDR) corrected p-values (Benjamini & Hochberg, 1995) for these statistical tests.

Participants who reported greater Stress Coping motives for their marijuana use displayed significantly increased startle potentiation, ${\eta }_p^2=0.05$, b=2.4, 95% CI=[0.7,4.1], t(145)=2.81, p=0.006, FDR-corrected p=0.031 (Figure 2A)⁴. No significant effects were observed for Conformity [ ${\eta }_p^2=0.02$, b=−3.5, 95% CI=[−8.1,1.1], t(145)=1.50, p=0.137, FDR-corrected p=0.274], Enhancement [ ${\eta }_p^2=0.01$, b=0.8, 95% CI=[−0.7,2.3], t(145)=1.07, p=0.288, FDR-corrected p=0.432], Expansion [ ${\eta }_p^2<0.01$, b=0.2, 95% CI=[−1.1,1.5], t(145)=0.33, p=0.744, FDR-corrected p=0.744], or Social motives for marijuana use [ ${\eta }_p^2=0.01$, b=0.8, 95% CI=[−0.9,2.4], t(145)=0.90, p=0.367, FDR-corrected p=0.441]. A significant linear effect for Marijuana Use Disorder diagnosis was observed, ${\eta }_p^2=0.05$, b=18.8, 95% CI=[4.5,33.1], t(141)=2.60, p=0.010, FDR-corrected p=0.031, with increasing startle potentiation observed across participants with no diagnosis vs. marijuana abuse vs. marijuana dependence (see Figure 2B).

Figure 2. Startle Potentiation by Marijuana-Relevant Individual Differences.

Panel A: This panel displays predicted values for startle potentiation to unpredictable shock threat by Stress Coping Motives from the general linear model (GLM). The shaded region respresent ±1 standard error for these predicted values. This GLM adjusted for all covariates.

Panel B: This panel displays predicted values for startle potentiation to unpredictable shock threat by Marijuana Use Disorder diagnoses from the general linear model (GLM). Error bars represent ±1 standard error for these predicted values. This GLM adjusted for all covariates.

Figure © 2017 Kathryn Hefner, Mark Starr, and John Curtin, under Creative Commons Attribution 4.0 International Public License CC-By

Discussion

Stress Neuroadaptation in Heavy Marijuana Use

We investigated stressor reactivity among deprived and non-deprived heavy marijuana users and minimal use controls to test for predicted stress neuroadaptations that may result from heavy marijuana use. We observed that heavy marijuana users (both deprived and non-deprived) displayed increased startle potentiation during threat of unpredictable electric shock relative to minimal use controls. To our knowledge, this is the first study to demonstrate increased reactivity to unpredictable stressors among heavy marijuana users in a manner comparable to exaggerated stressor reactivity among users of other drugs. Specifically, Moberg and colleagues (2017) reported that alcohol dependent participants in early abstinence (<8 weeks) displayed increased startle potentiation to unpredictable (vs. predictable) shock threat relative to healthy controls. Similarly, Gorka and colleagues (2013) demonstrated increased startle potentiation during unpredictable shock threat in participants with co-morbid alcohol dependence and panic disorder relative to both participants with only panic disorder and healthy controls. Furthermore, Grillon and colleagues (2007) confirmed increased startle potentiation during blocks of unpredictable but not predictable air blast cues in overnight deprived smokers relative to non-smokers.

Synthesis of these studies provides emerging support that is consistent with the Stress Neuroadaptation Model in the etiology of alcohol and other drug (AOD) addiction (Kaye et al., 2017). Rodent addiction models posit that repeated homeostatic adjustments in brain stress systems to acute periods of AOD intoxication eventually lead to long-lasting, compensatory allostatic changes in CNS structures and circuits involved in behavioral and emotional response to stressors (i.e., stress neuroadaptations; Koob & Le Moal, 2008). Substantial evidence has suggested that specific adaptations in CRF and NE sensitive pathways in the central extended amygdala contribute importantly to increased stressor reactivity in rodents, including stress-induced reinstatement of drug use (for reviews see Mantsch, Baker, Funk, Lê, & Shaham, 2016; Silberman & Winder, 2013; Smith & Aston-Jones, 2008). Importantly, basic affective science research with both humans and rodents have validated startle potentiation selectively to unpredictable stressors as a sensitive index of these stress mechanisms. Thus, increased startle potentiation to unpredictable threat among heavy users of marijuana, alcohol and/or tobacco in humans is consistent both with increased stressor reactivity and the neural mechanisms proposed by the rodent Stress Neuroadaptation Model. Future research in humans can strengthen explore neural mechanisms by direct pharmacologic manipulation of these neurotransmitter systems while measuring unpredictable stressor startle potentiation in AOD dependent users (e.g., NIAAA R01-AA024388). Moreover, future research should explore other marijuana use characteristics (e.g., chronicity, patterns of episodic use such as binge/withdrawal) that may contribute to neuroadaptations in stressor reactivity.

Although results from recent studies are consistent with the Stress Neuroadaptation Model, the exclusive use of cross-sectional designs with pre-existing groups across these studies does not provide definitive evidence of drug-related changes, and allows for other plausible interpretations. In particular, increased startle potentiation to unpredictable stressors may represent a pre-morbid risk factor for AOD use disorders rather than merely a consequence of chronic AOD use (Gorka, Lieberman, Phan, & Shankman, 2016; Rasmussen & Kincaid, 2015). Moreover, these factors may work in tandem, such that pre-morbid risk involving heightened startle potentiation contributes to AOD use, which subsequently exacerbates stressor reactivity over time. Additional research including longitudinal designs to document within-person change in stressor reactivity following heavy and/or chronic use is necessary to adjudicate between alternative but mutually tenable interpretations. Such research would also address ambiguity regarding other possible confounding variables that may differ between the heavy and minimal use groups that might explain their differences in stressor reactivity. Our use of covarites to control for some key individual differences reduces but cannot eliminate concern about such confounds.

Stressor Reactivity during Marijuana Deprivation

With respect to our second aim, we did not observe any impact of marijuana deprivation on startle potentiation to unpredictable shock threat among heavy users. This null result was observed despite clear evidence of a successful deprivation manipulation, wherein compliance was verified by both self-report and urinalysis. Furthermore, deprived heavy users reported significantly more intense withdrawal symptoms and negative affect symptoms, as well as greater craving, than did non-deprived heavy users. These results support that our deprived heavy users experienced a robust withdrawal syndrome including expected increases in negative affect. Nonetheless, marijuana withdrawal did not appear to influence their stressor reactivity.

The null effect of marijuana deprivation on unpredictable startle potentiation suggests potentially distinct mechanisms may be responsible for a) the increased stressor reactivity observed among all heavy marijuana users, and b) the increased subjective negative affect displayed by deprived heavy users only, respectively. This distinction has support from rodent models; specifically, NE and CRF mechanisms proposed to experience neuroadaptations following chronic drug use support “dynamic, active response to an acute stressor” rather than tonic, persistent negative mood states (Koob & Zorrilla, 2012, p. 309; also see Heilig, Goldman, Berrettini, & O’Brien, 2011).

Thus, it appears possible that heavy marijuana users may experience two forms of affective disruption that could motivate further use or prompt relapse after cessation of use. Unpredictable stressors may elicit punctate exaggerated reactivity that increases the reward value of further marijuana use during stressful periods. These phasic increases in drug use motivation may further be layered on top of tonically increased negative affect resulting from deprivation and also provide strong motivation further use and/or relapse (Baker et al., 2004; Curtin, Mccarthy, Piper, & Baker, 2006).

Of course, increased confidence in this dual mechanism thesis for marijuana and other drugs requires additional investigation. For example, it is possible that we did not observe a deprivation effect because reactivity to unpredictable threat may represent a stable risk marker that does not vary across periods of marijuana use and abstinence. Alternatively, this null effect for changes in stressor reactivity during marijuana deprivation could be a type 2 error, though it should be acknowledged that we had adequate power to detect such an effect (86% power to detect a moderate sized between-subjects effect). Nonetheless, confidence in this null effect would be increased with replication that used potentially more powerful within subject manipulations of deprivation. This effect must also be scrutinized for other commonly used illicit drugs. For example, to our knowledge, short-term drug deprivation effects on unpredictable stressor reactivity have not been examined for alcohol or opiates, two drugs that produce potent withdrawal syndromes. Two studies have examined the effects of tobacco deprivation on unpredictable stressor reactivity using startle potentiation but the results were inconsistent. Hogle (2010) and colleagues observed increased startle response to unpredictable shock threat among deprived relative to non-deprived smokers, but they did not include non-smokers. In contrast, Grillon and colleagues (2007) reported increased startle potentiation to unpredictable air blasts among deprived smokers vs. non-smokers but no differences between deprived and non-deprived smokers. Given the paucity of evidence, we consider it premature to draw definitive conclusions about the impact of drug deprivation across drug classes on stressor reactivity (see Kaye et al., 2017 for further discussion of possible deprivation/withdrawal mechanisms).

Individual Differences in Stressor Reactivity

Our third aim was to examine relationships between individual differences in stressor reactivity and measures of marijuana use motives and marijuana use diagnostic status. As expected, heavy marijuana users reported stronger motives for marijuana use than minimal use controls across all motives (i.e., enhancement, expansion, social, and stress coping) other than conformity. However, stress coping motives was unique in its relationship with startle potentiation to unpredictable shock threat. Participants reporting greater stress coping motives for marijuana use displayed increased startle potentiation during unpredictable stressor exposure. This relationship could emerge through varied etiologic pathways. If heavy marijuana use had increased stressor reactivity per the Stress Neuroadaptation Model, users who may have experienced this neuroadaptation would be expected to use marijuana to dampen or otherwise cope with increased reactivity as it emerged over time. Support for this hypothesis would be strengthened by evidence that acute marijuana use dampened unpredictable threat startle potentiation, as has been demonstrated for other anxiolytic drugs (Baas et al., 2002; Bradford, Motschman, Starr, & Curtin, 2017; Bradford, Shapiro, & Curtin, 2013; Grillon et al., 2006; Hefner et al., 2013). Alternatively, if increased stressor reactivity was a pre-morbid risk factor for heavy marijuana use, use of marijuana to cope may represent an initial motive that itself leads to heavy use. Notwithstanding, the likely clinical importance of unpredictable threat startle potentiation is made salient by its unique relationship with stress coping motives. tartle potentiation to unpredictable threat also varied systematically by diagnostic status, with higher means observed for participants with marijuana dependences vs. marijuana abuse vs. no diagnosis, further underscoring its clinical import.

Limitations and Future Directions

Synthesis of our results indicates that increased startle potentiation a) was observed among heavy users vs. minimal use controls, b) was associated with increasing stress coping motives, and c) was associated with increased diagnostic severity for marijuana use disorder. Similar increased startle potentiation to unpredictable threat has been confirmed among individuals with AOD use disorders. This highlights heightened phasic stressor reactivity as a potential etiologically relevant mechanism across addictive drugs as proposed by the Stress Neuroadaptation Model. Furthermore, it makes salient a potentially important negative consequence that may result from heavy use of any addictive drug, including marijuana. However, cross-sectional designs like that of the present study are not sufficient to establish stress neuroadaptation, and do not exclude the possibility that the observed effects result from pre-morbid risk-related individual differences. As noted earlier, increased confidence in these claims will require longitudinal designs to clearly document within person change following heavy, chronic drug use. Nonetheless, we believe that confirmation of these group differences in stressor reactivity as predicted by the Stress Neuroadaption model in a cross-sectional design was a necessary pre-requisite to justify subsequent more costly longitudinal designs to definiteively test for within person change.

Research with other drugs has also specifically implicated reactivity to unpredictable (vs. predictable) stressors. Unfortunately, the current study did not include a predictable stressor condition. Future research should also use methods (e.g., pharmacological challenge with neurotransmitter agonists and antagonists) and measures (e.g., fMRI, PET) that enable more precise specification of the neural mechanisms accounting for increased stressor reactivity. For now, inference about neural mechanism rests primarily on the tight translation of both measure (startle potentiation) and methods (unpredictable shock threat) from preclinical research with rodents (Davis et al., 2010). Furthermore, it should be acknowledged that stressor reactivity was assessed in this study in the context of a reward decision-making task. Reward related processes should not have confounded heavy user or deprivation contrasts because all groups completed this decision making task. However, this leaves open questions about whether these stressor reactivity effects generalize to situations without concurrent focus on rewards.

More broadly, NIMH has begun to advocate for an “experimental therapeutics” approach in clinical research that simultaneously evaluates intervention efficacy and mechanism in clinical trials research (Insel, 2015; Insel & Gogtay, 2014). This approach requires that clinical trials explicitly measure pertinent mechanism(s) of either the disease process or the action of the intervention. In particular, the use of “surrogate endpoints” – early markers of disease process mechanisms with high predictive validity for later clinical outcomes – offer high promise to increase the pace of clinical trials research to develop and refine medications and other interventions for psychopathology broadly, including substance use disorders (Insel, 2012; Lerman et al., 2007; Litten, Falk, Ryan, & Fertig, 2016; McKee, 2009). We believe that data from this and other recent studies support the use of startle potentiation to unpredictable threat as a surrogate endpoint to study stress mechanisms in addiction using an experimental therapeutics approach (for review, see Kaye et al., 2017).

Finally, the focus on an etiological pathway involving increased stressor reactivity, either pre-morbidly or as a consequence of heavy use, highlights an important target for treatment. Clinical trials are underway to examine the use of NE alpha1 antagonists medications (e.g., prazosin, doxazosin; NIAAA R01-AA024388) to target stress induced relapse in substance use disorders, consistent with mechanisms we propose here (for review, see Mantsch et al., 2016). More precise targeting of sources and coping strategies for unpredictable stressors may increase the efficacy of psychological interventions for addiction. Relapse prevention programs can help patients to better identify risks by explicit, personalized assessment of stressors in their lives that are both potent and characterized by a high degree of uncertainty. These programs can also help patients develop tools to reshape their environments and social interactions to reduce unpredictability (e.g., problem-solving to address uncertain financial, housing, or interpersonal stresses; practicing direct communication with others to clarify stressful but ambiguous interpersonal exchanges). This focus on reactivity to unpredictable stressors also reinforces the potential benefits offered by existing but often unavailable or underfunded harm reduction approaches that robustly reduce uncertainty for drug dependent users regarding some of the most potent and unpredictable stressors involving housing, health, and other basic needs (Newman & Goldman, 2008). We believe that such multi-faceted approaches involving medications, psychosocial interventions, and public health initiatives offer the most promise to address the high costs of AOD use disorders generally, including any increase in rates of marijuana use disorder that may result from changing perceptions and regulations of this drug.

General Scientific Summary.

Animal models of addiction etiology propose that heavy, chronic drug use causes changes in the brain that result in heightened stressor reactivity. Consistent with this thesis, heavy marijuana users displayed exaggerated response to the threat of unpredictable electric shock compared to individuals with minimal history of marijuana use. Furthermore, stressor reactivity was increased among users who reported greater stress coping motives for their marijuana use and users with cannabis use disorder diagnoses.