Use of a sequential multiple assignment randomized trial to test contingency management and an integrated behavioral economic and mindfulness intervention for buprenorphine-naloxone medication adherence for opioid use disorder

Based on operant conditioning, CM is a behavioral treatment where prizes or vouchers are provided to patients based upon proof of a desired behavior such as abstinence or adherence to a medication. Importantly, CM may be ideal for individuals with low motivation to change or limited environmental support for abstinence or treatment engagement given that it provides an immediate tangible reward such as money or a prize that is contingent upon achieving the target outcome (e.g., chemically verified abstinence or medication compliance) [14, 15]. A meta-analysis of 30 studies that evaluated the use of CM to promote opioid abstinence in patients receiving methadone showed that CM resulted in a greater number of drug-negative urinalysis screens (effect size r= 0.25) compared to a control group and that this effect was stronger for immediate rewards [14]. Other meta-analyses have shown that CM is not as effective at improving clinic attendance (r = 0.15) [16] but has greater efficacy for opioid-specific abstinence (d = 0.65) than general substance use abstinence (d= 0.42) [17]. Interestingly, larger effect sizes appear to be associated with shortertreatment duration, suggesting that the effects of CM may diminish over time [17].Although no studies have explored the use of CM for buprenorphine-naloxone adherence, across other forms of MOUD (e.g., methadone), the effect size of CM on adherence was found to be medium to large (d= 0.75) and it is thus a promising approach [18].

Although both CM and CBT methods have the potential to improve medication adherence, CBT-based strategies have greater potential for dissemination than CM. Although CM is widely accepted as an efficacious treatment for substance use disorder and remains an accepted program for substance use cessation in the Veteran’s Administration system [35], this strategy has not often gained a foothold in community settings [36].

In addition, it is possible that these complimentary strategies are differentially effective at different treatment stages. Evidence suggests that CM reward effects diminish, whereas CBT strategies gain momentum over time [37, 38]. One study of individuals with cocaine dependence (N= 120) directly examined the long-term abstinence outcomes of CM vs. CBT over a period of 12 months [37]. Cocaine urinalysis results indicated that for those in the CM group, abstinence declined substantially over time (63%, 40%, and 44%, across 17, 26, and 52 weeks, respectively). Conversely, the CBT group showed significant increasesin abstinence over the follow-up time periods (33%, 64%, and 81%, respectively) [37]. Another study comparing these same treatment conditions was conducted with individuals with concomitant cocaine use who were receiving methadone maintenance [39]. Although CM was found to be the superior treatment at 16 weeks, the percentage of CBT group participants with cocaine-free urine samples (53%) significantly exceeded percentages in CM (47%) at 26 weeks [39].

While these long-term abstinence results are promising, the traditional CBT for substance use tested in these studies did not explicitly address mindfulness, delay discounting, or engagement with substance-free alternative activities and did not address medication adherence. Based on behavioral economic theory, brief motivational interviewing plus substance-free activities session plus mindfulness (BSM) could alter time orientation to value the future more and changing behavior to pursue drug-free rewards and manage cravings that are prevalent during treatment initiation. BSM is likely to be a critical mechanism to support medication adherence and promote recovery and can be delivered efficiently in a brief format.

At this point, there is considerable evidence of the effectiveness of CM and the CBT strategies in BSM for substance use abstinence, but less compelling or complete evidence of how these methods address adherence to buprenorphine-naloxone, the current state of the science treatment for OUD [1, 2]. Given that adherence to buprenorphine-naloxone is closely related to long-term treatment success [40], the specific and objective evaluation of buprenorphine-naloxone adherence is paramount to evaluating intervention utility.

Our research team seeks to compare the effectiveness of these two different interventions for MOUD adherence by conducting a sequential multiple assignment randomized trial (SMART) to assess sequential, individual, and combined effects of these two forms of treatment across MOUD initiation and maintenance. We will randomize 280 buprenorphine-naloxone initiating patients to receive either CM or BSM for their first 4 physician visits for MOUD. Adherence will be established via attendance and toxicology screen. Participants who are adherent will enter maintenance intervention (remaining in their assigned intervention condition for up to 6 months). Those who are not adherent for the initial intervention period will be re-randomized to receive either the other intervention or a combined intervention and will then continue to receive this intervention for an additional 6 months.

This is randomized superiority clinical trial design with individual parallel block-randomization (1:1) to 2 active intervention groups at the first stage (Stage 1), CM or BSM, and subsequent (Stage 2) randomization for initially non-adherent participants (known as SMART design). Figure 1 presents the participant flowchart, including information on the timing of re-randomization, length of intervention, and follow-up periods.

Due to the high likelihood of relapse during MOUD treatment initiation, we are recruiting patients at their first visit to the clinic and providing adjunctive intervention (CM or BSM) only for the first 4 visits for MOUD with the addiction medicine physician. However, because there is risk of illicit opioid use for individuals who switch MOUD providers, we consider all new patients to the clinic eligible, regardless of prior experience with MOUD. Those who are non-adherent will be re-randomized (described below).

We will engage with referred MOUD patients at the time of their initial visit to the university-based addictions clinic. Interested patients will be informed about the study in a private room and, if eligible, will be consented and randomized to condition.

Randomization for Stage 1 and Stage 2 is a 2-arm, parallel, random assignment with 1:1 allocation ratio using block-randomization (blocks of 4) developed by the study statistician with separate blocks for males and females to avoid coincidental assignment imbalance over time and to ensure that males and females are comparatively represented in both of the respective randomization points. Randomization sequence will be stored in a database where the sequence is concealed until participants are enrolled and interventions are assigned by the study coordinators. Buprenorphine providers and the outcomes assessor will be blinded to condition.

Following randomization, participants will start their assigned intervention at their physician visits for MOUD. Notably, patients’ MOUD visit schedules vary during treatment initiation and, depending upon circumstances, can occur weekly, bi-weekly, or monthly.

At each of these 4 MOUD physician visits (known as treatment initiation regardless of time period between visits), participants will complete assessments, then engage in intervention activities. Those who are adherent to buprenorphine-naloxone (i.e., attend physician visits and demonstrate presence of buprenorphine in urine screen) during this 4-visit period will remain in their intervention condition for up to 6 months (maintenance intervention). A final assessment will be conducted at 8 months from Stage 1 randomization.

Those who are not adherent to buprenorphine-naloxone, defined as missing their physician visit or absence of buprenorphine in urine screen, will be re-randomized to either switch to the other intervention condition or add the other intervention (CM + BSM combined). We expect that about 50% of the participants will be initially non-adherent and will qualify to be re-randomized. For re-randomized participants, intervention will start from the beginning.

To encourage participation in the follow-up assessments, all participants will receive $50 gift certificates to Amazon for the completion of each of the assessment contacts. All participants will receive appointment reminders via their preferred modality (text message, phone call, or email) the day before their appointments. For participants who miss an appointment or are lost to follow-up, we will attempt to contact them at least three times a week when they are in an open visit window (i.e., due for a visit). We will use multiple different modalities including phone call, text messages, email, and contacting the participant’s alternate contact person. We will also attempt to contact the participant at different times of the day.

Those randomized to the BSM will engage in a weekly discussion centered around increasing engagement in substance-free activities that are enjoyable and/or consistent with the individual’s goals or values. A related focus is enhancing the salience of future goals and of the association between daily patterns of behavior and achieving these outcomes, as well as the potential impact of relapse on those goals. This includes an episodic future thinking exercise where the individual vividly imagines and describes a positive future event associated with treatment adherence and achieving sobriety, as well as ongoing planning to engage in specific substance-free activities, and the SOBER breathing space exercise [34, 42, 43]. Each session of this 4-session intervention will take 45 min. Further descriptions of these intervention components can be found in Table 1.

Demographics will include age, contact information, gender, race, ethnicity, dependency status, marital status, education, and household income. These characteristics will only be assessed at baseline.

Contact information will include name, address, email address, and emergency contact persons and contacts for retention purposes. Participants will provide updated contact information at each study visit to aid in retention.

Opioid use disorder will be assessed at baseline using symptom criteria from the Diagnostic and Statistical Manual of Mental Disorders-5^thedition [44].

Delay discounting will be measured using the Money Choice Task [45], a 27-item questionnaire which asks the participant to decide between immediate, small rewards, and delayed, but larger rewards. The individual’s “discount rate” determines the steepness of the reduction in present value of delayed rewards. As the discount rate increases, the duration of the window of vulnerability increases, as does the strength of preference for the impulsive choice within the window. A participant’s discounting curve may be calculated according to the following function: V = A/(1 + kD) where V is the present value of the delayed reward A at delay D, and k is the rate of discounting. k typically falls between 0.0 and 0.5, with smaller values indicating a lack of discounting and preference for delayed rewards and higher values indicating strong discounting and a preference for immediate rewards. Thus, higher values of kare indicative of high levels of impulsivity. The Money Choice Questionnaire has demonstrated good test–retest reliability over 3 years and external validity as evidenced by relations to impulsive behaviors, such as crime and substance use [45, 46].

The Behavioral Economic framework suggests that addiction is based on a recurrent and persistent pattern of elevated motivation for drugs [24]. Reinforcing efficacy of the substance is often quantified using hypothetical purchase tasks in which individuals provide values for the drug of choice across a range of escalating prices (i.e., $0, $0.25, $0.50, $1, $1.50, $2, $2.50, $0.3, $0.3.50, $4, $4.50, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $20, $25, and $30). From these tasks, demand and expenditure curves can be plotted, which yield several indices of relative value including intensity of demand (consumption at lowest price, usually zero); breakpoint (price at which consumption first reaches zero); O_max (maximum expenditure); P_max(price corresponding to maximum expenditure); and elasticity (i.e., relative change in consumption in response to change in commodity price) [48, 49]. The measurement of reinforcing efficacy using hypothetical purchase tasks among young adults has been validated using the Alcohol Purchase Task (APT) [50] to assess alcohol demand. Indices derived in this hypothetical task demonstrate good test–retest reliability [51] and are associated with demand indices derived from actual consumption choices in a laboratory setting [52]. They also evidence associations with alcohol problems beyond consumption [51] and predict treatment outcomes following brief drinking interventions [53, 54]. Modified hypothetical purchase tasks have been used to assess young adult marijuana demand [55], cigarette demand [56], and non-prescribed opioid demand [47]. The Opioid Purchase Task will be assessed at each visit.

This is a 47-item measure which asks participants to rate the extent to which the following factors have been helpful in their recovery from opioids either currently or during periods in the past when they stopped using. Items are scored on a 3-point Likert scale ranging from 1 = Not at all helpful to 3 = Very helpful.

The BPI Short Form [58] assesses pain severity and impact on daily life function. The short form is a 9-item measure that assessed multiple areas in a Likert scale, open-ended, and visual analog scale type questions. Research with chronic pain patients indicates that the BPI Short Form is internally reliable, test–retest reliable and is valid in terms of construct, convergent, and predictive validity [58].

The 5-level EQ-5D version (EQ-5D-5L) [59] assesses quality of life across 5 dimensions: mobility, usual activities, self-care, pain/discomfort, and anxiety/depression. Each dimension is scored on a 5-point scale: 1 = no problems, 2 = slight problems, 3 = moderate problems, 4 = severe problems, and 5 = extreme problems. The patient is asked to indicate his/her health state by ticking the box next to the most appropriate statement in each of the five dimensions. This decision results in a 1-digit number that expresses the level selected for that dimension.

The Generalized Anxiety Disorder 2-item (GAD-2) is a brief screening tool for generalized anxiety disorder [60]. The two items are scored on a 4-point scale ranging from 0 (Not at all) to 3 (nearly every day). The total score is obtained by adding the scores from the two items. Research suggests that the GAD-2 has a sensitivity of 86% and specificity of 83% for predicting generalized anxiety disorder.

The Pain, Enjoyment, General Activity scale (PEG for Chronic Pain) [61] is a 3-item measure that assesses pain intensity and functional interference. It has demonstrated good reliability and validity in large samples of chronic pain patients in primary care, VA hospitals, and pain management centers [61]. The PEG is scored by averaging the items; scores range from 0 to 10. Its excellent psychometric properties and responsiveness to treatment compare favorably to other established measures such as the PROMIS and SF-36 pain scales [62].

The Patient Health Questionnaire-2 (PHQ-2) is a 2-item questionnaire that measures [63] the frequency of depressed mood and anhedonia over the past 2 weeks. The two items are scored on a 4-point scale ranging from 0 (Not at all) to 3 (nearly every day). The total score is obtained by adding the scores from the two items. Research suggests that the PHQ-2 has a sensitivity of 97.6% and specificity of 59.2% for predicting major depressive disorder.

The Tobacco, Alcohol, Prescription medication, and other Substance use (TAPS) Tool [64] for substance use screening is a 4-item screen for tobacco, alcohol, illicit drugs, and non-medical use of prescription drugs, followed by a substance-specific assessment of risk level for individuals who screen positive. Items are scored on a 6-point Likert scale, ranging from 0 = never to 5 = daily. In a sample of 2000 medical patients, the TAPS Tool had a sensitivity of 0.93 (95% CI 0.90–0.95) and specificity of 0.87 (95% CI 0.85–0.89) for tobacco, and a sensitivity of 0.74 (95% CI 0.70–0.78) and specificity of 0.79 (95% CI 0.76–0.81) for alcohol [64].

The Timeline Followback (TLFB) [65] is a method that can be used as a clinical and research tool to obtain a variety of quantitative estimates of marijuana, cigarette, and other drug use. The TLFB can be administered by an interviewer, self-administered, or administered by computer. It involves asking clients to retrospectively estimate their drug, marijuana, or cigarette use 7 days to 2 years prior to the interview date [66].

Craving is a cardinal feature of substance use disorder and has been shown to predict future drug use. Because of this, craving is often assessed in treatment settings as a marker of risk for subsequent drug use. The Opioid Craving Scale, a modification of the Cocaine Craving Scale [68] consists of three items rated on a visual analog scale from 0 to 10: (1) How much do you currently crave opiates? (rated from not at all to extremely), (2) In the past week, please rate how strong your desire to use opiates has been when something in the environment has reminded you of opiates (rated from no desire to extremely strong), and (3) Please imagine yourself in the environment in which you previously used opiates. If you were in this environment today and if it were the time of day that you typically used opiates, what is the likelihood that you would use opiates today? (rated from not at all to I’m sure I would use opiates). This scale has demonstrated strong internal reliability, and greater craving was found to be associated with higher odds of prescription opioid use in the following week. For each one-unit increase on this scale, the odds of using opioids in the subsequent week was 17% higher [67].

Perceived social support [69] is assessed via a 40-item scale which asks about support from friends and family. Each item is scored as Yes/No/I Don’t Know. Three studies affirmed the internal reliability of the PSS and confirmed the validity of the measure via relations with measures of supportive relationships, positive life events, anxiety, and depression.

The participant’s toxicology screen will be abstracted from the medical record.

Attendance at clinic visit is reported as yes or no for each visit window.

Dosage of buprenorphine-naloxone prescribed will be obtained from medical record at each visit.

Self-reported adherence to buprenorphine will assessed by the participant reporting how many days they skipped their medication since their last buprenorphine care provider visit in conjunction with a calculation of the total number of days since their last buprenorphine care provider visit.

The full schedule of assessments and intervention delivery can be found in Table 2.

This manuscript uses SPIRIT reporting guidelines [70]. A SPIRIT checklist is included as an additional file.

Moderators that will be examined fall into three categories. Behavioral economics-related variables will be assessed at each study visit using self-report questionnaires. These include measures of delay discounting [45], opioid demand [47], engagement in substance-free activities [57], adherence self-efficacy [71], and locus of control [72].

Additionally, measures of mental health diagnoses will be examined via self-report measures, (Patient Health Questionnaire-2 [63] and Generalized Anxiety Disorder 2-item [60]) and transcription from the medical file. Further, we will measure opioid and other substance use using the Timeline Followback [73], measures of opioid cravings [67], and opioid use disorder criteria [44]. We will also assess adverse events (e.g., overdose) and pain [61].

All data entry, processing, and management will occur via a relational database. All of the informatics structures within this system will operate in a client/server network environment and provide methods to secure data [33]. Only study personnel with appropriate access privileges will have access to the informatics system.

Data will be stored on a HIPAA-compliant cloud software. These datasets will be stored with read-only privileges to prevent any changes to the file. The statistician will be able to download the data as needed. Data will be coded and a codebook developed.

Interventionists will be master’s level counselors trained in the execution of the study interventions. Following didactic training on session content, interventionists will be required to “test out” of the intervention sessions by enacting them with the clinical supervisor of the study, who will serve as the practice patient.

All treatment sessions will be audio recorded in order to conduct treatment fidelity on 20% of all sessions (224 of a possible total 1120 intervention sessions recorded) by the clinical supervisor of the study.

Participants will be enrolled at the time of their first visit to the buprenorphine treatment provider. Participants will be seen up to four times in Stage 1 randomization at their next four provider visits. Assessments and intervention will occur at each of these visits. If a participant is re-randomized into Stage 2, they will be seen at the next four visits. Assessments and intervention will also occur at each of these visits. There will be a final follow-up assessment 8 months after enrollment.

Investigative Team is comprised of the Principal Investigator and six Co-Investigators with specific expertise required for the execution of this trial. Two investigators provide medical oversight and safety review, three provide theoretical and cognitive behavioral oversight, and two provide data management and data acquisition oversight.

A data safety monitoring board (DSMB) unaffiliated with the study team and funding agency will evaluate procedures for participant safety and study futility. This board will report directly to the funding agency every 6 months throughout the funding period. The DSMB has no competing interests.

The primary data analysis will follow an intent-to-treat analysis [79]. The principal approach with respect to missing data is foremost to make every effort to keep missing data to a minimum [80]. If missingness is unavoidable, contemporary multiple imputation [81, 82] will be used in the main analysis. We will initially summarize and compare participants’ anthropometric, demographic, and other baseline characteristics and will test for systematic differences in these for participants who do meet or do not meet adherence after Stage 1 and, again, Stage 2 using t-test/Wilcoxon-test, chi-square/Fisher’s exact test, and weighted approaches when indicated by the SMART design [77].

We will also conduct a descriptive heterogeneity of treatment effect analysis (HTE) [83] with respect to sex, gender, race, socioeconomic status, and age. We will test for HTE by modeling the interaction between treatment arm and covariate and declare HTE present if the interaction term is statistically significant at the 5% level based on a likelihood ratio test. We will conduct and report the HTE-analyses (including 95% confidence intervals) irrespective of the determination of a statistically significant overall effect to facilitate the planning of future clinical trials and subsequent meta-analytical approaches investigating possible sex-, gender-, or race-specific effects.

Hypothesis 1: We predict that a higher percentage of those in the CM condition will be adherent after 4 weeks of intervention than those in BSM: Comparison of proportions and logistic regression with adherent yes/no as dependent variable includes baseline demographics, etc., as explanatory variables. We are adequately powered to test this hypothesis as well as the question whether there are differences in adherence after the 6-month maintenance period depending on initial treatment assignment.

Hypothesis 2: Based upon evidence of diminished effects of CM over time, we predict that those in CM (vs. BSM) will show decreases in adherence during the 8-month follow-up: Approach is as for Hypothesis 1 above but based on all entering the maintenance phase after being adherent for 4 weeks. We have adequate power only if the relative difference between these two groups is large (80% power to detect, e.g., 10% vs. 32% subsequent non-adherence, assuming stage-1 adherence proportions of γ₁ = γ₂ = 0.5 and accounting for 20% attrition; smaller differences can be detected if initial adherence proportions are higher or attrition is lower). Even here, we will report our findings including 95% confidence intervals to facilitate the planning of future clinical trials and subsequent meta-analytical approaches.

Hypothesis 3: We predict that the combined CM + BSM will have the highest treatment effect among those who are initially non-adherent, given the greater intensity of intervention. Comparison of the relevant final adherence proportions in the participants re-randomized, using Tukey’s Honest Significant Difference (HSD) procedure to adjust for the multiplicity in the pairwise comparisons. We expect to have a minimum of 23 individuals in the smallest of these re-randomized groups, but do not expect to be powered to authoritatively conclude about this hypothesis. We will report our findings including 95% confidence intervals to facilitate the planning of future clinical trials and subsequent meta-analytical approaches.

The Principal Investigator (PI) will evaluate security of data and safety of participants monthly. Reports will be sent to the PI and statistician each month by the Study Coordinator regarding recruitment and retention.

A Data Safety Monitoring Board (DSMB) has been established for this study. While the interventions used in this pilot are believed to pose minimal risk, the high-risk nature of the study population (Opioid Use Disorder patients) makes additional protections a necessity. This committee comprised of four qualified persons independent of competing interests who will meet a minimum of once every 6 months. Types of credentials our DSMB members possess are as follows: (1) experience conducting behavioral intervention studies; (2) experience with treating opioid use disorder; and (3) biostatistical expertise in intervention studies. As DSMB members, the DSMB will review the reports sent by the study statistician to determine whether there is any corrective action, trigger of an ad hoc review, or stopping rule violation that should be communicated to the study investigator, the IRB, or the funder. The DSMB will be headed by a Chair member who will send the DSMB’s recommendation to approve, halt, or discontinue the study to the funder.

Interim analyses will be conducted solely for the DSMB reports and annual reports to the funder. The analyses include subject accrual, adverse event (AE), and serious adverse event (SAE) rates by arm, overall retention and retention by arm, adherence by study arm, and stopping rules regarding statistical power implications of drop-outs and missing data. All of these analyses with the exception of adherence by study arm will be available to all investigators and DSMB members. The investigative team, except for the statistician compiling the report, will not have access to adherence by arm since it is a primary outcome.

A physician Co-Investigator will be responsible for overseeing AE and SAE reporting. Research coordinators will take initial reports from participants and the physician will review for safety and accuracy and recommend follow-up as needed. All AEs and SAEs will be reviewed within five business days. Any SAEs possibly related to study intervention will be reported to the IRB, DSMB, and funding agency within five business days.

While the funder does not plan regular audits, the study team is prepared for any possible audits from the funder or the approving IRB.

Protocol amendments of any nature will be submitted to the IRB for approval. Major protocol changes will also be approved by the funder before implementation.

Informed consent will be obtained from all participants before any research procedures take place. The consent form is approved by the reviewing IRB and contains all basic and additional elements of informed consent as regulated by 45CFR46.

Trained study staff will obtain informed consent primarily in-person at the participants’ initial buprenorphine provider visit. However, the IRB approved an alteration of consent for electronic consent when needed. The informed consent discussion takes place via telephone while the participants views and signs the consent electronically.

All personally identifying information (PII) and personal health information (PHI) needed for recruitment, study involvement, and tracking will be obtained and maintained by the project personnel in a secured server with limited access. All computer files and systems are password protected and accessible by authorized personnel only. Data entry and transfer will be performed by the study staff shared only with those persons authorized to have access.

All paper records will be locked in a secure file cabinet behind at least two locked doors. Only necessary staff will have access.

The datasets from the study will not contain any PHI. Participants will have an assigned study ID number and their data will be associated with this number only.

The final trial dataset will be available initially to the study’s Investigators. After the main outcomes are published, public access will be provided.

This project was funded through the NIH HEAL (Helping End Addiction Long-term) Initiative, which requires award recipients to share data via the HEAL Initiative Data Ecosystem in compliance with the HEAL Initiative Public Access and Data Sharing Policy, which is in line with the new NIH policy for Data Management and Sharing. Data shared here will be de-identified.

Other researchers who are interested in using this data for publication will be able to submit a request to the investigative team.

The protocol for this study will be published on ClinicalTrials.gov at the completion of the study. The dataset and statistical code will be available by request to the investigative team.

Recruitment for this trial began on February 1, 2022. We anticipate enrollment to be completed by the end of 2023. At the time this manuscript is being written, protocol version 2.0 from 11/15/2022 is in use.