Accepted Manuscript

Title: SOCIAL COGNITIVE PREDICTORS OF

TREATMENT OUTCOME IN CANNABIS DEPENDENCE

Author: Matthew J. Gullo Marya Matveeva Gerald F.X.

Feeney Ross McD. Young Jason P. Connor

PII: S0376-8716(16)30981-4
DOI: http://dx.doi.org/doi:10.1016/j.drugalcdep.2016.10.030
Reference: DAD 6239

To appear in: Drug and Alcohol Dependence

Received date: 18-7-2016

Revised date: 18-10-2016
Accepted date: 24-10-2016

Please cite this article as: Gullo, Matthew J., Matveeva, Marya, Feeney, Gerald
F.X., Young, Ross McD., Connor, Jason P., SOCIAL COGNITIVE PREDICTORS
OF TREATMENT OUTCOME IN CANNABIS DEPENDENCE.Drug and Alcohol
Dependence http://dx.doi.org/10.1016/j.drugalcdep.2016.10.030

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 1

SOCIAL COGNITIVE PREDICTORS OF TREATMENT OUTCOME IN CANNABIS

DEPENDENCE

Matthew J. Gullo1,2,3, Marya Matveeva1,3, Gerald F.X. Feeney1,2, Ross McD. Young2,4,5,

Jason P. Connor1,2,6

1
Centre for Youth Substance Abuse Research, The University of Queensland, Brisbane, QLD
4029, Australia
2
Alcohol and Drug Assessment Unit, Division of Medicine, Princess Alexandra Hospital,
Brisbane, QLD 4102, Australia
3
School of Psychology, The University of Queensland, Brisbane, QLD 4072, Australia
4
Faculty of Health, Queensland University of Technology, Brisbane QLD 4059, Australia
5
Institute of Health and Biomedical Innovation, Queensland University of Technology,
Brisbane QLD 4059, Australia
6
School of Medicine, The University of Queensland, Brisbane, QLD 4029, Australia

*Correspondence to:

Matthew J. Gullo, Centre for Youth Substance Abuse Research, The University of

Queensland, Brisbane, QLD 4029, Australia

E-mail: m.gullo@uq.edu.au
 2

Highlights

 Expectancies and self-efficacy are key cognitive constructs in cannabis dependence

 First study to measure both constructs as treatment outcome predictors

 Emotional relief refusal self-efficacy was robust predictor of treatment outcome

 Negative cannabis expectancy predicted lower likelihood of lapse

 Self-efficacy may mediate risk conveyed by positive cannabis expectancy

ABSTRACT

Background: Drug-related outcomes expectancies and refusal self-efficacy are core

components of Social Cognitive Theory. Both predict treatment outcome in alcohol use

disorders. Few studies have reported expectancies and refusal self-efficacy in cannabis

dependence. None have examined both, although both constructs are key targets in

Cognitive-Behavioural Therapy (CBT). This study tests the predictive role of expectancies

and refusal self-efficacy in treatment outcome for cannabis dependence. Design: Outpatients

completed a comprehensive assessment when commencing cannabis treatment and predictors

of treatment outcome were tested. Setting: A university hospital alcohol and drug outpatient

clinic. Participants: 221 cannabis-dependent patients participated in a 6-week CBT program

where the goal was abstinence. Measurements: Cannabis Expectancy Questionnaire and

Cannabis Refusal Self-Efficacy Questionnaire, cannabis dependence severity [Severity of

Dependence Scale], psychological distress [General Health Questionnaire] at baseline; the

timeline follow-back procedure at baseline and each session. Findings: Patients reporting

lower confidence in their ability to resist cannabis during high negative affect (emotional

relief refusal self-efficacy) had a lower likelihood of abstinence (p = .004), more days of use

(p < .001), and larger amount used (p < .001). Negative cannabis expectancies predicted
 3

greater likelihood of abstinence (p = .024). Higher positive expectancies were associated with

lower emotional relief self-efficacy, mediating its association with outcome (p < .001).

Conclusions: Emotional relief refusal self-efficacy and negative expectancies are predictive

of better treatment outcomes for cannabis dependence. Positive expectancies may indirectly

predict poorer outcome because of a negative association with self-efficacy, but this

conclusion remains tentative as directionality could not be established.

Keywords: cannabis dependence, expectancies, self-efficacy, treatment, CBT

1. INTRODUCTION

Cannabis is the most widely used illicit drug with 2.8 to 4.5% of the adult global

population estimated as cannabis users (Degenhardt and Hall, 2012). In the most recent

Australian survey, lifetime prevalence of cannabis dependence is approximately 3% when

using criteria from the fourth edition of the Diagnostic and Statistical Manual of Mental

Disorders (DSM-IV; Mewton et al., , 2013). Compared to the fifth edition, DSM-IV cannabis

dependence is equivalent to moderate-to-severe cannabis use disorder (4+ criteria met;

American Psychiatric Association, 2013). Cannabis dependence is associated with a wide

variety of adverse physical and mental health consequences (Hall and Degenhardt, 2009; Hall

and Pacula, 2003). No medications have been approved for the treatment of cannabis

dependence (Justinova et al., 2013). Psychological therapies are the first line of treatment

(Davis et al., 2015; Litt et al., 2008). Cognitive-behavioural therapy (CBT) is among the most

effective (Babor and The Marijuana Treatment Project Research Group, 2004; Budney et al.,

2006; Carroll et al., 2006; Hoch et al., 2014). Relapse rates remain high. Determining

predictors of outcome is required to inform more effective treatment (McRae et al., 2003;

Moore and Budney, 2003).

 4

Social Cognitive Theory (SCT) emphasizes the importance of two domains of

cognition in bringing about behaviour change, both of which can be affected by vicarious

experience or modeling (Bandura, 1986). Outcome expectancies describe an individual’s

belief that a particular behaviour will produce certain outcomes (Bandura, 1977, 2001; Jones

et al., 2001). According to SCT, expectancies about the effects of substance use play an

important role in consumption, dependence, and treatment. Evidence links expectancies to all

of these outcomes (Boden et al., 2013; Connor et al., 2007; Jones et al., 2001; Young et al.,

2011). Expectancies have been shown to predict cannabis use and dependence (Connor et al.,

2014). Positive cannabis expectancies are predicted to increase motivation to consume

cannabis, and negative expectancies to decrease it (Connor et al., 2011).

Self-efficacy forms the second key component of SCT. Refusal (or abstinence) self-

efficacy has received most research attention. This is belief in the ability to refuse an abused

substance. Low refusal self-efficacy is hypothesised to undermine motivation for abstinence

and predict poorer outcomes (Bandura, 1999; Oei and Baldwin, 1994). In their review,

Kadden and Litt (2011) reported low refusal self-efficacy consistently predicted poorer post-

treatment outcomes across substances. Self-efficacy for avoiding cannabis after completing

treatment predicted cannabis use at 12 months over-and-above past consumption, peer use,

temptation, and stress (Stephens et al., 1995). In a comparison of different psychological

treatments for cannabis dependence, Litt and colleagues (2008) reported increases in refusal

self-efficacy to be a primary common mechanism through which 12-month outcomes were

achieved.

Most previous studies have employed ad hoc global measures of refusal or abstinence

self-efficacy. Young et al. (2012) developed the Cannabis Refusal Self-Efficacy

Questionnaire (CRSEQ), a theoretically-driven and psychometrically robust assessment of

 5

self-efficacy. The measure assesses situational confidence to refuse cannabis consumption for

emotional relief, social facilitation, and opportunistically. These subtypes of refusal self-

efficacy may be differentially related to clinical outcomes. In a large sample of court-referred

cannabis users, Young et al. reported all three subtypes were associated with lower

consumption, but only high emotional relief self-efficacy was related to lower severity of

cannabis dependence.

Despite evidence that both cognitive domains individually predict substance use

outcomes (Boden et al., 2013; Connor et al., 2007; Jones et al., 2001; Kadden and Litt, 2011;

Young et al., 2011), refusal self-efficacy and outcome expectancies have rarely been

examined together, particularly in clinical populations. According to SCT, refusal self-

efficacy should mediate the relationship between expectancies and treatment outcomes

(Bandura, 1999; Oei and Baldwin, 1994). Those holding more positive (or less negative)

beliefs about using cannabis should find it more difficult to refuse it in cued situations. Low

refusal self-efficacy has been found to mediate the association between positive alcohol

expectancies and problematic drinking in patients undergoing treatment, college students, and

adolescents (Connor et al., 2011; Gullo et al., 2010). Given the significant correlation

between the two, not including both expectancies and self-efficacy in predictive models could

have obscured previous findings and the identification of priority targets for CBT (Connor et

al., 2014).

Connor and colleagues (2014) investigated the combined role of expectancies and

refusal self-efficacy in 1,115 cannabis users referred for assessment by the courts as an

alternative to prosecution. In this cross-sectional study, refusal self-efficacy fully mediated

the relationship between negative cannabis expectancies and weekly consumption. It partially

mediated the effects of positive expectancies on weekly consumption. Expectancies and

 6

refusal self-efficacy are also likely to be associated with treatment outcome. To date, no study

has investigated the influence of both expectancies and refusal self-efficacy as predictors of

treatment outcome in cannabis dependence.

The current study investigated the relationship between outcome expectancies, refusal

self-efficacy and treatment outcome among cannabis-dependent outpatients. Outcomes of

interest were cannabis abstinence, number of days of use, and amount used. The secondary

aim of the study was to test the hypothesised mediational relationship between these

constructs as outlined in SCT. According to SCT, refusal self-efficacy should mediate the

relationship between expectancies and treatment outcomes. It was predicted that greater

positive cannabis expectancies would be associated with lower refusal self-efficacy and, in

turn, predict poorer treatment outcomes; fully mediating the relationship between

expectancies and outcomes. A similar relationship was hypothesised for negative

expectancies, refusal self-efficacy and treatment outcomes, but with lower negative

expectancies being associated with lower self-efficacy and greater use.

2. METHOD

2.1. Participants and procedures

Data were obtained from 221 treatment-seeking cannabis users who presented to an

outpatient alcohol and drug clinic at an Australian metropolitan public hospital. All patients

attended treatment voluntarily. An initial intake assessment was conducted by a clinical nurse

or social worker prior to referral to a cognitive-behavioural cannabis treatment program

where the goal was abstinence. The program comprised five 1-hr sessions delivered over six

weeks, with the final session taking place one fortnight after session four. The program was

delivered one-on-one by Masters- or Doctoral-qualified clinical psychologists. It included

CBT and elements of motivational enhancement; specifically, craving management, cognitive

 7

restructuring, relapse prevention, and motivational interviewing. Patients were not excluded

from program if they lapsed, so long as they maintained abstinence as their goal.

Questionnaires assessing refusal self-efficacy, cannabis expectancies, severity of dependence

and psychosocial functioning were completed at the first treatment session. Dependence

severity and psychosocial functioning could impact treatment response and were included as

potential covariates (Stephens et al., 1993; White et al., 2004). Abstinence, numbers of days

used, and amount of cannabis used were recorded at each session. Therapists were not aware

of study aims. Hospital and university human research ethics approval was obtained.

2.2. Measures

2.2.1 Cannabis Expectancy Questionnaire (CEQ; Connor et al., 2011). The 45-item CEQ

assessed positive (18 items, e.g., ‘Smoking cannabis makes me feel outgoing and friendly’)

and negative outcome expectancies (27 items, e.g., ‘Smoking cannabis makes me confused’).

Responses were rated on a 5-point Likert scale (1 = Strongly disagree to 5 = Strongly agree).

Both subscales have excellent internal reliability (α = 0.89 and 0.93 for negative and positive

expectancies, respectively (Connor, Gullo, et al., 2011). The factor structure and criterion

validity of the CEQ has been confirmed across two samples (Connor et al., 2011).

2.2.2. Cannabis Refusal Self-Efficacy Questionnaire (CRSEQ; Young et al., 2012). The 14-

item CRSEQ measured levels of cannabis refusal self-efficacy. Responses were rated on a 6-

point Likert scale (1 = I am very sure I could NOT resist smoking to 6 = I am very sure I

could resist smoking. It comprises three subscales: Emotional relief (six items, e.g., ‘When I

feel sad’), Opportunistic (five items, e.g., ‘When I am at a party’), and Social facilitation

(three items, e.g., ‘When I want to feel more accepted by friends’). The measure has good-to-

excellent internal consistency (α = 0.97, 0.91 and 0.84 for each subscale, respectively) and its

factor structure and criterion validity has been previously established (Young et al., 2012).
 8

2.2.3. Severity of Dependence Scale–Cannabis (SDS-C; Swift et al.,, 1998). This five-item

scale assesses degree of cannabis dependence experienced by users (e.g., Have you ever

thought your cannabis use is out of control?). Responses are rated on a 4-point Likert scale (0

= Never to 3 = Always). The SDS-C has good test-retest reliability and is sensitive to severity

of cannabis dependence (Swift et al., 2000). Using Australian normative data, a score of ≥3 is

indicative of DSM-IV cannabis dependence (Swift et al., 1998).

2.2.4. Cannabis consumption. The Timeline Followback (Sobell and Sobell, 1992) was used

to measure session-by-session cannabis consumption and abstinence status. This

retrospective, calendar-based assessment is one of the most psychometrically robust measures

of daily substance use (Robinson et al., 2014). Outcomes of interest were abstinence status (0

= not abstinent, 1 = abstinent), number of days used, and total amount used (in grams).

2.2.5. Psychological distress. The Anxiety, Depression, and Somatic Symptoms scales of the

General Health Questionnaire-28 (GHQ-28) were used to assess psychological distress

(Goldberg and Williams, 1988). Items assess recent changes in perceived health and

wellbeing (e.g., Felt that life is not worth living) and rated on a 4-point Likert-type scale (0 =

Not at all to 3 = Much more than usual). It has strong psychometric properties (Goldberg et

al., 1997; Goldberg and Williams, 1988; Werneke et al., 2000). Higher scores reflect poorer

functioning.

2.3. Data analysis

Multi-level modelling (MLM) in MLwiN (version 2.30) was used to analyse the

influence of expectancies and self-efficacy on treatment outcomes. MLM is particularly

suited to analysing longitudinal clustered data; in this instance, sessions nested within

patients (Hox, 2002). This is superior to analysing endpoint summary statistics, such as

percentage days abstinent, as it allows for modelling of individual trajectories of change over
 9

time. MLM is also suited to naturalistic settings where the number and frequency of sessions

may vary across individuals. MLM utilises full information maximum likelihood (FIML)

estimation, which are optimal for handling missing data (Graham, 2009), which is substantial

in treatment studies. FIML produces less biased estimates than other missing data

approaches, such as assuming relapse or carrying forward the last observation (Hallgren and

Witkiewitz, 2013). Full iterative generalized least squares (IGLS) estimation, a type of FIML,

was employed for days used and amount used models. Abstinence was analysed with

generalized linear models utilising a logit link function and Taylor series expansion. In these

models, coefficients represent logit transformed probabilities. To calculate effect size, they

were converted to odds by finding the natural antilog (ex). Assumptions of linearity and

normality were assessed by examination of residuals. Baseline models included a random

intercept (constant, β0j), session number, and controlled for time (days) between sessions.

Potential covariates were included (β3 - β10), but only retained if statistically significant:

Abstinence/Days Used/Amount Usedij = β0j + β1 Time (days) between sessionsij + β2

Session number j + β3 High school completion:Yesj + β4 Severity of cannabis

dependencej + β5Gender:Malej + β6 Agej + β7 Employment:Yesj + β8 Somatic

Symptomsj + β9 Anxietyj + β10 Depressionj + eij.

SCT predictors were added to the model on Step 2 (grand mean-centered), followed by SCT

x session number interactions to investigate time-dependent effects. Predictors were tested for

significance using the Wald test. Mediation was tested using the joint significance procedure,

which is less prone to Type II error (MacKinnon et al., 2002) than the commonly-used

‘causal steps’ procedure (Baron and Kenny, 1986). There is support for mediation where

there is a significant association between IV and mediator (path a) and a significant

association between mediator and DV (path b). The primary MLM analyses test path b of
 10

mediation. Because IV and mediator were measured once, standard multiple regression was

implemented to test path a (expectancies→self-efficacy). Mediation effects were estimated

with the product-of-coefficients method using the PRODCLIN software to calculate 95%

confidence intervals (MacKinnon et al., 2007). When estimating mediation effects for the

abstinence outcome variable (dichotomous), a and b path coefficients were first standardized

using procedures outlined in MacKinnon and Dwyer (1993) to correct for differences in

mediator/outcome distributions and variance of the residual.

3. RESULTS

3.1. Descriptive statistics and missing data

Nearly the entire sample was cannabis dependent (98.5%), with only three

participants scoring below the ≥3 cut-off on SDS-C (Swift et al., 1998). The average number

of sessions attended was three out of the total five, with 99 (44.8%) patients completing the

treatment program. Of those, 66 (66.7%) had been abstinent for at least the past two weeks.

Multi-level models are well equipped to account for missing values under most missing data

conditions, especially for longitudinal designs (Graham, 2009; Tasca and Gallop, 2009), and

Little’s (1988) Missing Completely At Random (MCAR) test on baseline data was not

significant, χ2 (734) = 792.598, p = .066. Grand mean probability of abstinence was .68,

95%CIs [.608, .748]. Grand mean amount of cannabis consumed between sessions was 1.31

grams (SD = 4.09) and participants used the drug on 1.22 days between sessions (SD = 3.33).

Sample descriptive statistics are reported in Table 1.

3.2. Abstinence

In the baseline model, no covariate was statistically significant and were omitted.

Session number predicted abstinence, with probability of abstinence increasing over the

course of treatment. Patients were 1.20 times more likely to be abstinent with each session,
 11

95%CIs [1.003, 1.43]. When SCT variables were added to the model, higher negative

expectancies at baseline predicted greater abstinence. A +1 SD increase in negative

expectancies increased the odds of abstinence by 40.4%. Emotional relief refusal self-

efficacy was also significant, with a +1 SD increase in self-efficacy associated with an

increase in the odds of abstinence by 80%. SCT x session number interactions added on Step

3 were not significant (ps > .05) and not retained. Therefore, the slope of increase in

abstinence over the course of treatment was not moderated by baseline expectancies or self-

efficacy. The final model is summarised in Table 2.

3.3. Number of days cannabis used

In the baseline model, no covariate was statistically significant. Session number was

significant, with patients reporting fewer days of cannabis use as treatment progressed.

Significant effects of SCT variables were found. Emotional relief self-efficacy predicted

fewer days of cannabis use during treatment (Table 3). SCT x session number interactions

added on Step 3 were not significant and not retained (ps > .05). The final model is

summarised in Table 3.

3.4. Amount of cannabis consumed

In the baseline model, no covariate was statistically significant. Session number was

significant, with patients consuming, on average, 0.13 grams less cannabis per session.

Significant effects of SCT variables were found. Emotional relief self-efficacy predicted a

lower amount of cannabis use during treatment (Table 4). SCT x session number interactions

added on Step 3 were not significant and not retained (ps > .05). The final model is

summarised in Table 4.
 12

3.5. Mediating role of refusal self-efficacy

Emotional relief refusal self-efficacy was the only domain of self-efficacy predictive

of cannabis use during treatment. To investigate its role as a potential mediator of expectancy

effects, a standard multiple regression was conducted. Overall, 11% of the variance in

emotional relief refusal self-efficacy was accounted for by expectancies (path α), F (2,207) =

12.95, p < .001. Positive expectancies (β = -.33, p < .001) uniquely accounted for 10.5% of

variance, but negative expectancies (β = -.12, p = .072) did not reach significance. Because

positive expectancies were directly associated with emotional relief refusal self-efficacy (path

a) and refusal self-efficacy predicted treatment outcome (path b), there was evidence for

mediation according to the joint significance procedure. Furthermore, all mediation effects

involving positive expectancies (predictor) and emotional relief refusal self-efficacy

(mediator) estimated using the product-of-coefficients method were statistically significant

(see Figure 1).

4. DISCUSSION

This is the first study to test the unique role of cannabis outcome expectancies and

refusal self-efficacy in treatment outcome. Results show emotional relief refusal self-efficacy

was a consistent predictor of improved outcomes in a 'real world' outpatient setting. There

was also preliminary support for its role as a mediator of the effects of positive expectancies

on outcome, as predicted by Social Cognitive Theory (SCT; Bandura, 1986). High negative

expectancies had a direct, protective effect on probability of abstinence during treatment, but

not number of days used or amount used during a lapse. Patients who attended more sessions

showed greater improvements across all outcomes. Findings provide new insights for health

professionals administering cognitive-behavioral therapy (CBT).

 13

Findings support the critical role of self-efficacy in treatment, as predicted by SCT

(Bandura, 1986; Connor et al., 2014; Oei and Baldwin, 1994). They are also in line with

previous reports on the role of self-efficacy in substance use treatment outcome more broadly

(Adamson et al., 2009; Kadden and Litt, 2011) and specifically in cannabis treatment

outcome (Litt et al., 2008; Stephens et al., 1995). The current study sought to extend this

literature by testing the unique contribution of three subtypes of refusal self-efficacy and, of

those, only emotional relief refusal self-efficacy was related to treatment outcomes. Young et

al. (2012) previously reported that while all subtypes of refusal self-efficacy were cross-

sectionally related to weekly cannabis consumption, only emotional relief refusal self-

efficacy was associated with severity of dependence. Copeland et al. (2001) reported stress

relief was the most commonly cited reason for cannabis use in a sample of 229 treatment

seekers. The current prospective study builds on these findings to show that only emotional

relief refusal self-efficacy is predictive of cannabis use during CBT treatment where the goal

is abstinence. According to SCT, with increased specificity self-efficacy becomes more

predictive of subsequent behavior. Given this, targeting patients’ confidence to refuse

cannabis use to alleviate negative affect is likely to be more effective than improving self-

efficacy more broadly. Considering the rate of treatment dropout, this may be an important

early goal in treatment.

The regulation of negative affect is a primary motivator for breaching abstinence in

cannabis dependence treatment. Similar results have been reported in the alcohol treatment

literature (Cooney et al., 1997; Law et al., 2016; Miller et al., 1996). However, the early

treatment context is important to consider here. Many patients report disposing of available

cannabis and avoiding cannabis-using peers when commencing treatment, potentially

reducing the impact of social facilitation and opportunistic refusal self-efficacy beliefs on
 14

abstinence. It is possible that these other domains of self-efficacy play a greater role in

longer-term abstinence and this requires further investigation. Nevertheless, the value in

distinguishing between refusal self-efficacy subtypes is clear. These results suggest treatment

should focus on building emotional relief refusal self-efficacy to reduce the likelihood of

early lapse. This could be achieved through practicing adaptive coping strategies, such as

cognitive reappraisal or relaxation, as alternative means of regulating affect. Successful

implementation of these strategies would likely increase emotional relief self-efficacy, reduce

lapses, and serve to maintain motivation for change (Bandura, 1986).

High positive cannabis expectancies were associated with lower refusal self-efficacy,

consistent with the hypothesis that positive expectancies increase substance use by

undermining self-efficacy (Connor et al., 2014; Gullo et al., 2010). The more rewarding

cannabis use is believed to be, the more difficult a patient will find it to refuse in cued

situations. Importantly, expectations of negative reinforcement (tension reduction) are

included in the positive expectancies scale and are likely to be most relevant to emotional

relief self-efficacy. These findings support the proposal that self-efficacy acts as the final

pathway to human behaviour (Bandura, 1999), including cannabis use. Litt et al. (2008)

reported that increases in refusal self-efficacy was the primary common mechanism through

which different psychological treatments achieved abstinence at 12 months post-treatment.

Thus, there are several ways refusal self-efficacy could be strengthened. Our results suggest

that, for patients reporting strong positive expectancies, challenging these exaggerated beliefs

about the rewarding effects of cannabis may be an effective method of improving self-

efficacy (Gullo et al., 2010). However, caution is required when inferring directionality here,

given expectancies and self-efficacy were measured at the same time point.
 15

Negative expectancies predicted a lower likelihood of cannabis lapse, consistent with

Boden et al.’s (2013) analysis of self-initiated abstinence in military veterans. Boden et al.

also reported negative expectancies predicted average amount consumed, but their study did

not control for self-efficacy. The present study found no association and this is in line with

Connor et al.’s (2014) study of court-referred cannabis users not engaged in treatment, who

also controlled for self-efficacy. Connor et al. reported a positive association between

negative expectancies and severity of cannabis dependence, but the prognostic implications

of this were limited by their study’s cross-sectional design. The present results and those of

Boden et al. demonstrate clearly that negative expectancies predict lower likelihood of lapse.

The present study further demonstrates that this protective effect is independent of refusal

self-efficacy and, at least for patients undergoing CBT, is unrelated to the severity of a lapse

if it occurs. It is likely that pre-treatment negative expectancies are an important determinant

of motivation for abstinence. Strategies that highlight the negative impact of cannabis on

patient health will likely further reinforce this motivation (Copersino et al., 2006),

irrespective of a patient’s confidence in their ability to refuse cannabis. Strategies may

include cost-benefit analyses and evoking cognitive dissonance over how continued cannabis

use interferes with long-term goals (Beck et al., 1993; Miller and Rollnick, 2012). The large

sample and longitudinal design of the present study allows for stronger inferences as to the

direction of effects between cognition and cannabis use in clinical settings.

This study has limitations. Social cognitive variables were only measured together at

the start of treatment. This limits inferences about direction of effects between expectancies

and self-efficacy (including mediation), as well as how treatment-related changes in these

domains affect treatment outcome. Social Cognitive Theory predicts that, in the context of

substance use, outcome expectancies influence self-efficacy beliefs (Bandura, 2001; Oei and
 16

Baldwin, 1994). Tracking session-by-session changes in refusal self-efficacy and

expectancies could provide valuable insights into which components of CBT more effectively

impact upon these beliefs and, ultimately, lead to better outcomes (Gwaltney et al., 2005).

Although pre-treatment assessment is common in addiction research, cannabis withdrawal

may have impacted on cognition and influenced completion of instruments (Hall, 2015).

Future research employing independent measures of psychopathology will provide a better

indication of the role of comorbid mood and anxiety symptoms. This study also relied on

self-reported cannabis use and abstinence. Future studies could benefit from corroboration

from biological markers of cannabis metabolites. Lastly, treatment was delivered in a public

hospital outpatient clinic. While the abstinence program was manualized, fidelity checks

would have been desirable as treating psychologists were free to diverge from the program in

accordance with their clinical judgment. On the other hand, this freedom likely increases the

generalizability of findings to the ‘real world’.

In conclusion, this is the first study to investigate the unique contribution of outcome

expectancies and refusal self-efficacy to the prediction of treatment outcome in cannabis

dependence. Emotional relief self-efficacy was the most consistent predictor of outcome and

may mediate the effects of positive expectancies on cannabis use, making it an important

target for psychological treatment. Negative expectancies directly predicted greater likelihood

of abstinence. Overall, findings provide further support for the utility of Social Cognitive

Theory in the assessment and treatment of cannabis use disorder. Future studies need to

employ more frequent assessments of expectancies and self-efficacy to elucidate the dynamic

association between treatment-related changes in cognition and treatment outcome.

 17

Contributors: All authors were involved in designing the study. MM and MJG conducted

the statistical analyses. All authors contributed to drafting the manuscript and approved the

final submission.

AUTHOR DISCLOSURES

Role of Funding Source: MJG is supported by a National Health and Medical Research

Council (NHMRC) Early Career Fellowship (1036365). JPC is supported by a NHMRC

Career Development Fellowship (1031909). The Funder had no input into the design,

analysis or interpretation of results.

Conflicts of interest: none

Acknowledgments: We acknowledge the assistance of Annie McPherson, Jane Tucker and

Karen Dillman at the ADAU for their involvement in the QIDDI program.
 18

REFERENCES

Adamson, S. J., Sellman, J. D., Frampton, C. M. A., 2009. Patient predictors of alcohol

treatment outcome: a systematic review. J. Subst. Abuse Treat. 36, 75–86.

American Psychiatric Association, 2013. Diagnostic And Statistical Manual Of Mental

Disorders (5th ed.). APA, Washington, DC.

Babor, T. F., The Marijuana Treatment Project Research Group, 2004. brief treatments for

cannabis dependence: findings from a randomized multisite trial. J. Consult. Clin.

Psychol. 72, 455–466.

Bandura, A., 1977. Self-efficacy: toward a unifying theory of behavioral change. Psychol.

Rev. 84, 191–215.

Bandura, A., 1986. Social Foundations Of Thought And Action: A Social Cognitive Theory.

Prentice-Hall, Englewood Cliffs, NJ.

Bandura, A., 1999. A sociocognitive analysis of substance abuse: an agentic perspective.

Psychol. Sci. 10, 214–217.

Bandura, A., 2001. Social cognitive theory: an agentic perspective. Ann. Rev.Psychol. 52, 1–

26.

Baron, R. M., Kenny, D. A., 1986. The moderator–mediator variable distinction in social

psychological research: conceptual, strategic, and statistical considerations. J. Person.

Soc. Psychol. 51, 1173.

Beck, A. T., Wright, F. D., Newman, C. F., Liese, B. S., 1993. Cognitive Therapy Of

Substance Abuse. Guilford Press, New York.

Boden, M. T., McKay, J. R., Long, W. R., Bonn-Miller, M. O., 2013. The effects of cannabis

use expectancies on self-initiated cannabis cessation. Addiction 108, 1649–1657.

 19

Budney, A. J., Moore, B. A., Rocha, H. L., Higgins, S. T., 2006. Clinical trial of abstinence-

based vouchers and cognitive–behavioral therapy for cannabis dependence. J. Consult.

Clin.Psychol. 74, 307–316.

Carroll, K. M., Easton, C. J., Nich, C., Hunkele, K. a., Neavins, T. M., Sinha, R., Ford, H.L.,

Vitolo, S.A., Doebrick, C.A., Rounsaville, B. J., 2006. The use of contingency

management and motivational/skills-building therapy to treat young adults with

marijuana dependence. J. Consult. Clin. Psychol. 74, 955–966.

Connor, J. P., George, S. M., Gullo, M. J., Kelly, A. B., Young, R. M., 2011. A prospective

study of alcohol expectancies and self-efficacy as predictors of young adolescent alcohol

misuse. Alcohol Alcohol. 46, 161–169.

Connor, J. P., Gudgeon, E. T., Young, R. M., Saunders, J. B., 2007. The relationship between

alcohol expectancies and drinking restraint in treatment seeking alcohol dependent

patients. Addict. Behav. 32, 1461–1469.

Connor, J. P., Gullo, M. J., Feeney, G. F. X., Kavanagh, D. J., Young, R. M., 2014. The

relationship between cannabis outcome expectancies and cannabis refusal self-efficacy

in a treatment population. Addiction 109, 111–119.

Connor, J. P., Gullo, M. J., Feeney, G. F. X., Young, R. M., 2011. Validation of the Cannabis

Expectancy Questionnaire (CEQ) in adult cannabis users in treatment. Drug Alcohol

Depend. 115, 167–174.

Cooney, N. L., Litt, M. D., Morse, P. a., Bauer, L. O., Gaupp, L., 1997. Alcohol cue

reactivity, negative-mood reactivity, and relapse in treated alcoholic men. J. Abnorm.

Psychol. 106, 243–250.

Copeland, J., Swift, W., Rees, V., 2001. Clinical profile of participants in a brief intervention

program for cannabis use disorder. J. Subst. Abuse Treat. 20, 45–52.
 20

Copersino, M. L., Boyd, S. J., Tashkin, D. P., Huestis, M. A., Heishman, S. J., Dermand, J.

C., Simmons, M.S., Gorelick, D. A., 2006. Quitting among non-treatment-seeking

marijuana users: reasons and changes in other substance use. Am. J. Addict. 15, 297–

302.

Davis, M. L., Powers, M. B., Handelsman, P., Medina, J. L., Zvolensky, M., Smits, J. A. J.,

2015. Behavioral therapies for treatment-seeking cannabis users: a meta-analysis of

randomized controlled trials. Eval. Health Profess. 38, 94–114.

Degenhardt, L., Hall, W. D., 2012. Extent of illicit drug use and dependence, and their

contribution to the global burden of disease. Lancet 379, 55–70.

Goldberg, D., Gater, R., Sartorius, N., Ustun, T. B., Piccinelli, M., Gureje, O., Rutter, C.,

1997. The validity of two versions of the GHQ in the WHO study of mental illness in

general health care. Psychol. Med. 27, 191–197.

Goldberg, D., Williams, P., 1988. A user’s guide to the General Health questionnaire. NFER-

Nelson, Windsor, UK.

Graham, J. W., 2009. Missing data analysis: making it work in the real world. Annual

Review of Psychol. 60, 549–576.

Gullo, M. J., Dawe, S., Kambouropoulos, N., Staiger, P. K., Jackson, C. J., 2010. Alcohol

expectancies and drinking refusal self-efficacy mediate the association of impulsivity

with alcohol misuse. Alcohol. Clin. Exp. Res. 34, 1386–1399.

Gwaltney, C. J., Shiffman, S., Balabanis, M. H., Paty, J. A., 2005. Dynamic self-efficacy and

outcome expectancies: prediction of smoking lapse and relapse. J. Abnorm. Psychol.

114, 661–675.

Hallgren, K. A., Witkiewitz, K., 2013. Missing data in alcohol clinical trials: a comparison of

methods. Alcohol. Clin. Exp. Res. 37, 2152–2160.

 21

Hall, W. D., 2015. What has research over the past two decades revealed about the adverse

health effects of recreational cannabis use? Addiction 110, 19–35.

Hall, W. D., Degenhardt, L., 2009. Adverse health effects of non-medical cannabis use.

Lancet 374, 1383–1391.

Hall, W. D., Pacula, R. L., 2003. Cannabis Use And Dependence: Public Health And Public

Policy. Cambridge University Press, Cambridge, UK.

Hoch, E., Bühringer, G., Pixa, A., Dittmer, K., Henker, J., Seifert, A., Wittchen, H. U., 2014.

CANDIS treatment program for cannabis use disorders: findings from a randomized

multi-site translational trial. Drug Alcohol Depend. 134, 185–193.

Hox, J., 2002. Multilevel Analysis: Techniques and Applications. Lawrence Erlbaum

Associates, Mahwah, NJ.

Jones, B. T., Corbin, W., Fromme, K., 2001. A review of expectancy theory and alcohol

consumption. Addiction 96, 57–72.

Justinova, Z., Mascia, P., Wu, H.-Q., Secci, M. E., Redhi, G. H., Panlilio, L. V., Scherma,

M., Barnes, C., Parashos, A., Zara, T., Fratta, W., Solinas, M., Pistis, M., Bergman, J.,

Kangas, B.D., Ferre, S., Tanda, G., Schawarcz, R., Goldberg, S. R., 2013. Reducing

cannabinoid abuse and preventing relapse by enhancing endogenous brain levels of

kynurenic acid. Nat. Neurosci. 16, 1652–1661.

Kadden, R. M., Litt, M. D., 2011. The role of self-efficacy in the treatment of substance use

disorders. Addict. Behav. 36, 1120–1126.

Law, B., Gullo, M. J., Daglish, M., Kavanagh, D. J., Feeney, G. F. X., Young, R. M., Connor,

J. P., 2016. Craving mediates stress in predicting lapse during alcohol dependence

treatment. Alcohol. Clin. Exp. Res. 40, 1058–1064.

Little, R. J. A., 1988. A test of missing completely at random for multivariate data with
 22

missing values. J. Am. Stat. Assoc. 83, 1198–1202.

Litt, M. D., Kadden, R. M., Kabela-Cormier, E., Petry, N. M., 2008). Coping skills training

and contingency management treatments for marijuana dependence: exploring

mechanisms of behavior change. Addiction 103, 638–648.

Mackinnon, D. P., Dwyer, J. H., 1993. Estimating mediated effects in prevention studies.

Eval. Rev. 17, 144–158.

MacKinnon, D. P., Fritz, M. S., Williams, J., Lockwood, C. M., 2007. Distribution of the

product confidence limits for the indirect effect: program PRODCLIN. Behav. Res.

Methods 39, 384–389.

MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., Sheets, V., 2002. A

Comparison of methods to test mediation and other intervening variable effects.

Psychol. Methods 7, 83–104.

McRae, A. L., Budney, A. J., Brady, K. T., 2003. Treatment of marijuana dependence: a

review of the literature. J. Subst. Abuse Treat. 24, 369–376.

Mewton, L., Slade, T., Teesson, M., 2013. An evaluation of the proposed DSM-5 cannabis

use disorder criteria using Australian national survey data. J. Stud. Alcohol Drugs 74,

614–621.

Miller, W. R., Rollnick, S., 2012. Motivational Interviewing: Helping People Change.

Guilford Press, New York.

Miller, W. R., Westerberg, V. S., Harris, R. J., Tonigan, J. S., 1996. What predicts relapse?

Prospective testing of antecedent models. Addiction 91 Suppl. S155–72.

Moore, B. A., Budney, A. J., 2003. Relapse in outpatient treatment for marijuana

dependence. J. Subst. Abuse Treat. 25, 85–89.

Oei, T. P. S., Baldwin, A. R., 1994. Expectancy theory: a two-process model of alcohol use
 23

and abuse. J. Stud. Alcohol 55, 525–534.

Robinson, S. M., Sobell, L. C., Sobell, M. B., Leo, G. I., 2014. Reliability of the Timeline

Followback for cocaine, cannabis, and cigarette use. Psychol. Addict. Behav. 28, 154–

162.

Sobell, L. C., Sobell, M. B., 1992. Alcohol Timeline Followback Users’ Manual. Addiction

Research Foundation, Palo Alto.

Stephens, R. S., Wertz, J. S., Roffman, R. A., 1993. Predictors of marijuana treatment

outcomes: the role of self-efficacy. J. Subst. Abuse 5, 341–353.

Stephens, R. S., Wertz, J. S., Roffman, R. A., 1995. Self-efficacy and marijuana cessation: a

construct validity analysis. J. Consult. Clin. Psychol. 63, 1022–1031.

Swift, W., Copeland, J., Hall, W., 1998. Choosing a diagnostic cut-off for cannabis

dependence. Addiction 93, 1681–1692.

Swift, W., Hall, W., Copeland, J., 2000. One year follow-up of cannabis dependence among

long-term users in Sydney, Australia. Drug Alcohol Depend. 59, 309–318.

Tasca, G. A., Gallop, R., 2009. Multilevel modeling of longitudinal data for psychotherapy

researchers: I. the basics. Psychothera. Res. 19, 429–437.

Werneke, U., Goldberg, D. P., Yalcin, I., Ustün, B. T., 2000. The stability of the factor

structure of the General Health Questionnaire. Psychol. Med. 30, 823–829.

White, A. M., Jordan, J. D., Schroeder, K. M., Acheson, S. K., Georgi, B. D., Sauls, G.,

Ellington, R.R., Swartzwelder, H. S., 2004. predictors of relapse during treatment and

treatment completion among marijuana-dependent adolescents in an intensive outpatient

substance abuse program. Subst. Abuse 25, 53–59.

Young, R. M., Connor, J. P., Feeney, G. F. X., 2011. Alcohol expectancy changes over a 12-

week cognitive-behavioral therapy program are predictive of treatment success. J. Subst.

 24

Abuse Treat. 40, 18–25.

Young, R. M., Gullo, M. J., Feeney, G. F. X., Connor, J. P., 2012. Development and

validation of the Cannabis Refusal Self-Efficacy Questionnaire (CRSEQ) in adult

cannabis users in treatment. Drug Alcohol Depend. 125, 244–251.

 25

Figure Captions

Figure 1. Mediation of positive cannabis expectancies effect on treatment outcome.

**p < .01.

***p < .001.

Mediation effect on abstinence: CI 95% -0.550, -0.086;

Mediation effect on days used: CI 95% 0.011, 0.052;

Mediation effect on amount used: CI 95% 0.012, 0.043.

 26

Table 1. Descriptive statistics of the sample (N = 221).

Mean SD Range
Age (years) 30.10 8.72 18 - 59

Positive Expectanciesa 51.59 11.84 21 - 86

Negative Expectanciesa 80.44 19.97 24 - 133

Emotional Relief Self-Efficacyb 19.58 9.96 6 - 36

Opportunistic Self-Efficacyb 15.05 7.37 0 - 30

Social Facilitation Self-Efficacyb 12.94 4.51 0 - 18

Severity of Dependencec 9.54 2.85 0 - 15

Anxiety (GHQ-28)d 9.62 5.40 0 - 21

Depression (GHQ-28)d 6.64 5.93 0 - 21

Somatic Symptoms (GHQ-28)d 7.34 4.37 0 - 21

n %
Gender
Male 163 74
Female 58 26
Completed High School
107 48.4
Yes
No 114 51.6
Currently Employed
Yes 175 79.2
No 46 20.8
Note. GHQ-28 = General Health Questionnaire-28 (Goldberg & Williams, 1991).
a
Higher scores reflect higher expectancy.
b
Higher scores reflect greater refusal self-efficacy.
c
Higher scores reflect greater dependence severity.
d
Higher scores reflect poorer psychosocial functioning.
 27

Table 2

Social cognitive predictors of cannabis abstinence (N = 221).

Parameter Unstandardized SE z p
coefficient
Fixed effects

Constant, β0j 0.318 0.338

Step 1

Time between sessionsij -0.009 0.009 1.00 .32

Session numberij 0.181 0.091 1.99 .047

Step 2

Time between sessionsij -0.011 0.009 1.22 .22

Session numberij 0.177 0.097 1.82 .07

Positive expectanciesij 0.023 0.015 1.53 .13

Negative expectanciesij 0.018 0.008 2.25 .024

Emotional relief self-efficacyij 0.074 0.026 2.85 .004

Opportunistic self-efficacyij 0.028 0.032 0.88 .39

Social facilitation self- -0.050 0.060 0.83 .41

efficacyij

Random effects

Ωu 2.023 0.412
 28

Table 3

Social cognitive predictors of number of days cannabis used (N = 221).

Parameter Unstandardized SE z p
coefficient
Fixed effects

Intercept, β0j 1.699 0.375

Step 1

Time between sessionsij 0.014 0.009 1.56 .12

Session numberij -0.225 0.097 2.32 .02

Step 2

Time between sessionsij 0.015 0.010 1.50 .13

Session numberij -0.226 0.101 2.24 .03

Positive expectanciesij 0.007 0.020 0.35 .73

Negative expectanciesij -0.010 0.011 0.91 .36

Emotional relief self- -0.118 0.035 3.37 < .001

efficacyij

Opportunistic self-efficacyij 0.053 0.042 1.26 .21

Social facilitation self- 0.022 0.080 0.28 .78

efficacyij

Random effects

σ2 e 4.771 0.386

σ2u0 4.599 0.736

Deviance (-2*log likelihood) 2240.788

 29

Table 4

Social cognitive predictors of cannabis consumption (in grams, N = 221).

Parameter Unstandardized SE z p
coefficient
Fixed effects

Intercept, β0j 1.019 0.212

Step 1

Time between sessionsij 0.005 0.006 0.83 .41

Session numberij -0.133 0.051 2.61 .01

Step 2

Time between sessionsij 0.005 0.006 0.83 .41

Session numberij -0.132 0.052 2.54 .01

Positive expectanciesij -0.014 0.015 0.93 .35

Negative expectanciesij -0.002 0.007 0.29 .77

Emotional relief self- -0.105 0.024 < .001

efficacyij 4.38

Opportunistic self-efficacyij 0.039 0.030 1.30 .19

Social facilitation self- 0.075 0.056 .18

efficacyij 1.34

Random effects

σ2 e 0.807 0.080

σ2u0 1.946 0.297

Deviance (-2*log likelihood) 1090.330