Repurposing semaglutide as an adjunctive treatment for cocaine use disorder: protocol for a randomised controlled trial

Abstract

Introduction

Cocaine use disorder (CUD) is a significant public health concern in the USA, with considerable prevalence and mortality and no Food and Drug Administration (FDA)-approved pharmacotherapies. Recent advances in addiction science emphasise the need for novel, mechanism-based treatments. Glucagon-like peptide-1 receptor agonists, such as semaglutide, have shown promise in modulating reward-related behaviours and may offer therapeutic benefits for CUD. We present a study protocol evaluating semaglutide, as an adjunct to cognitive behavioural therapy (CBT), as a novel approach for treating CUD.

Methods and analysis

This is a randomised, double-blind, placebo-controlled trial enrolling 75 treatment-seeking adults with CUD. Participants will be randomised 1:1 to receive either once-weekly semaglutide (0.25–1.0 mg) or placebo injections over 14 weeks, alongside weekly individual CBT. Primary outcomes include changes in neurophysiological reactivity to drug-related and non-drug-related motivationally relevant cues (late positive potential), behavioural economics (cocaine demand), craving (Cocaine Craving Questionnaire) and cocaine use (self-report, urine drug screens). Exploratory aims assess associations between mechanistic changes and cocaine use, consumption of other substances (ie, tobacco, alcohol and cannabis) and dose–response relationships. Data will be analysed using Bayesian statistical methods using an intention-to-treat approach.

Ethics and dissemination

The study has been approved by the UTHealth Committee for the Protection of Human Subjects (HSC-MS-25-0412) and is registered on ClinicalTrials.gov. All participants will provide written informed consent. Findings will be disseminated through peer-reviewed publications and scientific conferences.

Trial registration number

NCT07227948.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This trial will examine the glucagon-like peptide 1 receptor as a novel treatment target for cocaine use disorder (CUD), a condition for which no Food and Drug Administration (FDA)-approved pharmacotherapies currently exist.

• The trial will utilise a randomised, double-blind, placebo-controlled design with prespecified outcome measures and intent-to-treat analysis.

• Elucidating potential mechanisms by which semaglutide may ‘work’ in the context of CUD will inform future translational research and accelerate the development of more effective, mechanism-based therapies for cocaine addiction.

• By repurposing an existing medication, this approach, if effective, offers a rapid and pragmatic path to clinical implementation, potentially shortening the timeline from discovery to impact in a population with urgent unmet needs.

• A potential limitation of the trial is its relatively short duration and exclusion criteria, which may limit the ability to detect the long-term effects of semaglutide and the generalisability of the findings to broader clinical populations.

Introduction

According to the 2023 National Survey on Drug Use and Health, approximately 1.3 million individuals in the USA met the diagnostic criteria for cocaine use disorder (CUD) within the past year.¹ This represents some decline from the 1.5 million cases reported in 2022.² Despite this recent decrease, long-term trends remain concerning. For instance, cocaine-related overdose deaths increased dramatically from 6784 in 2015 to 29 449 in 2023,³ marking a more than fourfold rise over an 8-year period. These patterns underscore the persistent and evolving public health challenges associated with cocaine use and the urgent need for effective pharmacological treatments for CUD. Although no medication has yet received Food and Drug Administration (FDA) approval for this indication, decades of research have yielded valuable insights and promising therapeutic candidates. Additionally, the integration of pharmacological treatments with behavioural interventions has shown promise in the treatment of CUD, with cognitive-behavioural therapy (CBT) being a leading psychotherapeutic approach for behavioural-pharmacological combined treatment for the disorder.⁴

Lessons learnt from past efforts suggest several important approaches for CUD medication development.⁵ First, there is a growing recognition of the need to explore novel targets beyond traditional monoaminergic pathways. Second, repurposing existing FDA-approved medications would offer an expedited route to CUD treatment development. Third, the utilisation of an experimental therapeutics framework linking target engagement with clinically meaningful outcomes would facilitate the development of more effective treatments. Fourth, recent FDA guidance acknowledges reductions in cocaine use frequency as a valid endpoint, shifting focus from total abstinence to more achievable treatment goals.⁶ Finally, the pursuit of standalone pharmacotherapies has evolved towards integrating adjunctive medications with evidence-based behavioural interventions.

Amid these developments, glucagon-like peptide-1 receptor agonists (GLP-1RAs) have emerged as a promising class of compounds for addiction treatment.⁷ GLP-1RAs mimic the action of GLP-1, an incretin hormone and satiation factor produced in the intestinal L-cells and central nervous system, including the nucleus tractus solitarius and olfactory bulb.⁸ Through GLP-1R signalling, these agents regulate glucose metabolism, delay gastric emptying and enhance satiety, making them effective treatments for type 2 diabetes (T2D) and obesity.⁹ Notably, GLP-1Rs are also expressed in brain regions implicated in reward processing and hedonic behaviour, suggesting a potential role of GLP-1R signalling in modulating addictive behaviours.¹⁰,¹² Consistent with this, in preclinical studies, GLP-1RAs attenuated cocaine self-administration,¹³ cocaine-induced locomotor stimulation,¹⁴ conditioned place preference¹⁵ and cocaine seeking behaviour¹⁶,¹⁸ and reduced cocaine-evoked dopamine release in the nucleus accumbens.^{13 14} An analysis of the US Department of Veterans Affairs databases showed that, compared with other medications for T2D, GLP-1RA use was associated with a lower risk of stimulant use disorders in patients with T2D.¹⁹ These preclinical and early epidemiological findings suggest that GLP-1RAs may offer therapeutic benefits for CUD, warranting further exploration of their efficacy and mechanisms of action.

This paper presents the design of a randomised clinical trial (RCT) evaluating the efficacy of once-weekly semaglutide, a GLP-1RA, as an adjunct to CBT for the treatment of CUD. This investigation aligns with current research priorities and contributes to the growing body of work aimed at identifying effective, mechanism-based treatments for stimulant use disorders.

Methods and analysis

Study design

This randomised, double-blind, two-arm, placebo-controlled trial will be conducted at the University of Texas Health Science Center at Houston (UTHealth), Center for Neurobehavioral Research on Addiction (CNRA). A total of 75 treatment-seeking individuals with CUD will be enrolled and randomised in a 1:1 ratio to receive a 14-week treatment of either once-weekly semaglutide (with escalating doses from 0.25 mg to 1.0 mg/week) or placebo (sterile saline, 0.9%) as an adjunct to individual CBT. Participants will attend study visits two times weekly to allow for urine drug testing, aligning with the typical detection window for cocaine use. Multimodal assessments, including neurophysiological, behavioural, self-report measures, of hypothesised target mechanisms will be conducted at baseline (pretreatment), during treatment and at the end of treatment (EOT, week 15). A follow-up visit for safety assessments is scheduled for week 23.

Study aims

Specific aims

Aim 1: to determine the effects of semaglutide on targeted mechanisms of change in the treatment of CUD.

Hypothesis 1a: compared with placebo, semaglutide will attenuate event-related potential (ERP) reactivity (late positive potential (LPP)) specific to cocaine-related stimuli.

Hypothesis 1b: semaglutide (vs placebo) will reduce motivation to purchase and consume cocaine as measured by behavioural economic indices (drug demand).

Hypothesis 1c: semaglutide (vs placebo) will reduce cocaine craving as measured psychometrically using the Cocaine Craving Questionnaire (CCQ).

Aim 2: to examine the effects of semaglutide on cocaine use.

Hypothesis 2: compared with placebo, semaglutide will reduce cocaine use, as measured by the percentage of cocaine-negative urine drug screens (UDS) during the final 2 weeks of treatment.

Exploratory aims

Aim 3: explore the association between changes in target mechanisms and cocaine use.

Aim 4: explore the effects of semaglutide on the use of other drugs, including tobacco smoking, alcohol intake and cannabis use.

Aim 5: characterise dose–response effects during semaglutide dose escalation.

Study participants

Eligibility

Participants will be adults 18 years and older meeting DSM-5 criteria for CUD and reporting recent cocaine use (verified by at least one positive UDS for the cocaine metabolite benzoylecgonine (BE), during intake). A full listing of inclusion and exclusion criteria is found in box 1.

Box 1. Eligibility criteria.

Inclusion criteria

1. Ability to provide informed consent before any study-related activity, willing to comply with all study procedures, and be available for the duration of the study.

2. Meet DSM 5 diagnostic criteria for CUD and report recent cocaine use (verified by at least one positive UDS during intake.

3. Body mass index of ≥25 kg/m^2.

4. Agree (if the participant is female and of child-bearing potential) to use effective contraceptive methods.

5. Have a medical and psychiatric history and a brief physical examination demonstrating no clinically significant contraindications for study participation.

6. Be able to provide the names of at least two persons who can consistently locate their whereabouts.

Exclusion criteria

Medical exclusions

1. Personal or first-degree relative(s) history of medullary thyroid carcinoma (MTC) or multiple endocrine neoplasia syndrome type 2.

2. History or presence of chronic pancreatitis or recent acute pancreatitis.

3. Type 1 or type 2 diabetes mellitus (previously diagnosed or indicated by glycosylated haemoglobin (HbA1C) ≥48 mmol/mol (6.5%) as measured at screening).

4. History of malignant neoplasms within the past 5 years prior to screening. Basal and squamous cell skin cancer and any carcinoma in situ are allowed.

5. Severe gastrointestinal disease (ie, severe gastroparesis), history of severe cardiovascular disease, end-stage renal disease.

6. Alanine aminotransferase or aspartate aminotransferase ≥3 times upper limit of normal (ULN).

7. Systolic blood pressure >180 mm Hg and/or diastolic blood pressure >105 mm Hg.

8. History of retinopathy.

9. Known or suspected hypersensitivity to semaglutide, excipients or related products.

10. History of seizure or elevated risk of seizure.

11. Women who are or plan to become pregnant, or lactating, or of childbearing potential and not using medically accepted forms of contraception.

12. Any medical illness or condition which would preclude safe and/or successful completion of the study.

13. Any foreseeable procedure requiring general anaesthesia or deep sedation.

Psychiatric/substance use exclusions

1. Current ≥moderate substance use disorder (SUD) aside from alcohol, nicotine or marijuana, or a SUD requiring medical detoxification (eg, alcohol, opioid, benzodiazepine).

2. Current or recent suicidal ideation.

3. Homicidal ideation that requires immediate attention.

4. Any psychiatric illness or condition, which would preclude safe and/or successful completion of the study.

Weight-related exclusions

1. Gained/lost ≥4.5 kg (10 lb) over the past 6 months (prior to screening).

2. Uncontrolled thyroid disease at screening.

Medication-related exclusions

1. Currently using sulfonylureas, insulin and insulin products or medication used for weight management.

2. Any otherwise not specified concomitant medication that could compromise participant safety or treatment.

General exclusions

1. Current, anticipated or pending enrolment in another addiction treatment programme and/or research study that could potentially affect participant safety and/or the study data/design.

2. Not planning to live in the area for the duration of the trial.

3. Surgery is scheduled for the duration of the trial, except for minor surgical procedures.

4. Unable to communicate (read, write and speak) fluently in English.

Discontinuation/withdrawal criteria

The participant may be discontinued at any time during the study at the discretion of the PI and/or the Study Physician for safety, behavioural, compliance or administrative reasons. Efforts must be made to have the participants who discontinue the study medication continue in the study. Only participants who withdraw consent will be considered withdrawn from the study.

Screening

Participants will be recruited using various strategies, including flyers, posters, newspapers and companies that provide recruitment services for clinical research using online platforms. Potential participants will complete a brief telephone prescreen to determine initial eligibility. The prescreen assessment includes questions assessing medical and psychiatric history, psychoactive substance use and cocaine use history. Based on the initial prescreen, individuals who appear eligible will be invited for a face-to-face screening visit.

The screening will begin with the completion of the informed consent, which will be obtained by the principal investigator (PI) or designee. Informed consent procedures will include details of the study, potential risks, study timeline and voluntariness of participation and dropping out. Consenting subjects will receive a comprehensive medical and psychiatric evaluation, a medical history questionnaire, physical examination, laboratory tests (complete blood count, complete metabolic panel, thyroid stimulating hormone, HbA1C, UDS and urine pregnancy test) and an ECG. Masters-level clinicians will conduct the Structured Clinical Interview for DSM-5,²⁰ the Addiction Severity Index²¹ and the Timeline Followback (TLFB)²² interview.

Study visits

Baseline assessment of target mechanisms, drug use and other measures will take place before treatment initiation. Study visits throughout the 14-week treatment period will occur two times weekly with at least a 2-day interval between visits. The study medication/placebo will be administered once weekly. All participants will be assessed at end-of-treatment (week 15) to collect outcome measures related to target mechanisms, drug use, treatment acceptability and other measures. Participants will return to the clinic 8 weeks after treatment discontinuation (week 23) for a safety check. Please refer to table 1 for the schedule of assessments performed at each study visit.

Table 1. Schedule of assessments.

Assessments and procedures	Screen	Baseline	Treatment (week)			EOT	Follow-up
	−5–0	0	1–4	5–8	9–14	15	23
Eligibility
Informed consent	X
Medical history/physical exam	X
Psychiatric Evaluation (SCID-5)	X
Addiction Severity Index (ASI)	X
Interventions
Medication
Semaglutide (mg/week)			0.25	0.5	1.0
Placebo			X	X	X
CBT			X	X	X
Target mechanisms
EEG (LPP)		X		W5	W9	X
Cocaine demand (CPT)		X		W5	W9	X
Cocaine Craving Questionnaire (CCQ)		X		W5	W9	X
Drug use
Timeline follow back (TLFB)	X	X	X	X	X	X	X
Urine drug screen (UDS)	X	X	XX	XX	XX	X	X
Safety
Lab (CBC, HbA1C, etc.), ECG	X
Vital signs, body weight	X	X	X	X	X	X	X
Fingerstick blood glucose			X	X	X
Body mass index (BMI)	X					X	X
Urine pregnancy test	X		X	X	X	X	X
Suicide severity (C-SSRS)	X		X	X	X	X	X
AE monitoring		X	X	X	X	X	X
PHQ-9	X		X	X	X	X	X
Treatment acceptability
Satisfaction (CSQ-9)						X
HRQoL		X				X	X
Other
Cigarette Evaluation Questionnaire (mCEQ)		X		W5	W9	X
Alcohol Craving Experience (ACE)		X		W5	W9	X
Cannabis Experiences Questionnaire (CEQ)		X		W5	W9	X
Food Craving Inventory (FCI)		X		W5	W9	X
Reward-based Eating Drive (RED)		X		W5	W9	X

CBC, complete blood count; CBT, cognitive behavioural therapy; CPT, cocaine purchase task; CSQ-9, Client Satisfaction Questionnaire; C-SSRS, Columbia Suicide Severity Rating Scale; EEG, electroencephalogram; EOT, end of treatment; HbA1C, glycosylated haemoglobin; HRQoL, Health-Related Quality of Life; LPP, late positive potential; PHQ-9, Patient Health Questionnaire.

Randomisation

Participants will be randomly assigned in a 1:1 ratio to receive either semaglutide or placebo. To ensure balance across treatment groups on key prognostic variables, an urn randomisation procedure will be employed. Stratification factors will include baseline cocaine use severity (≥ vs <15 days of use in the past 30 days), sex (male vs female) and concurrent use of other substances (tobacco, alcohol and cannabis). ‘Other substance use’ will be operationalised as a binary variable (yes/no) based on the use of all three substances and included as a stratification factor.

Blinding will be maintained throughout the study. Participants, investigators, outcome assessors and data analysts will remain blinded to treatment allocation. A designated unblinded study nurse will retain a hard copy of the randomisation codes in a secure, locked location. Additionally, a password-protected file containing the unblinding codes will be stored in REDCap to allow for emergency unblinding if necessary.

Intervention

Semaglutide/placebo

We will use FDA-approved semaglutide (Ozempic) available in single-use subcutaneous injection pens at doses 0.25/0.5 mg and 1.0 mg. Dosing will follow a standard titration schedule: 0.25 mg/week (weeks 1–4), 0.5 mg/week (weeks 5–8) and 1.0 mg/week thereafter (weeks 9–14). This schedule is designed to achieve steady-state plasma concentrations of semaglutide at the 1.0 mg dose during the final phase of treatment, allowing for evaluation of primary outcomes (eg, cocaine use during the last 2 weeks of treatment).

The reasons supporting our decision to investigate the effects of semaglutide at the target dose of 1.0 mg, which is lower than the 2.4 mg dose utilised for weight management, are: (1) this early phase project aims to detect a signal of target engagement and clinical effect while balancing time and feasibility. The higher, 2.4 mg, dose would require a 16-week dose escalation period, followed by at least 4 weeks needed to achieve steady state, posing logistical challenges for the study population both in terms of research and eventual clinical practice. (2) Higher doses of GLP-1RAs are associated with an increased occurrence of gastrointestinal (GI) side effects and with more profound weight loss, which could be problematic for this population. (3) Semaglutide at a dose of 1.0 mg has been shown to influence ingestive behaviours,²³ another reward-related behaviour. If this lower dose is adequate to modify reward-related mechanisms and cocaine use, it could have significant clinical implications.

All injections will be prepared and administered by an unblinded study nurse who will not be involved in behavioural counselling, data collection or outcome assessment. To maintain blinding, participants will be blindfolded during each injection. This administration protocol has been successfully implemented in our previous RCTs involving injectable GLP-1RAs.^{24 25}

Sterile saline (0.9%) will serve as the placebo for semaglutide. The placebo will be administered in the same blinded manner as semaglutide and will be volume-matched using insulin syringes.

Contraindications of semaglutide (exclusionary in this study) include personal or family history of MTC or in patients with multiple endocrine neoplasia syndrome 2 and serious hypersensitivity reaction to semaglutide or any of its excipients. Common side effects include gastrointestinal symptoms, mainly nausea, vomiting, diarrhoea, constipation and abdominal cramps. Semaglutide’s warnings and precautions listed in package insert include acute pancreatitis, diabetic retinopathy complications, increased risk of hypoglycaemia when used concomitantly with insulin secretagogue or insulin, acute kidney injury due to volume depletion, severe gastrointestinal adverse reactions, hypersensitivity reactions, acute gallbladder disease and pulmonary aspiration during general anaesthesia or deep sedation.

Study medication discontinuation criteria include suspicion of pancreatitis, excessive weight loss, pregnancy or intention of becoming pregnant, suicidal thoughts or behaviour, or any safety concern, as judged by the PI and/or Study Physician. Participants who discontinue the study medication will be encouraged to continue with the scheduled visits and assessments to ensure continued counselling and data collection.

Cognitive behavioural therapy

All participants will receive weekly 1 hour individual CBT, allowing us to evaluate the effects of semaglutide as an adjunct to CBT, the standard treatment for CUD.⁴ The CBT manual will be modelled after existing protocols used by our team²⁶,²⁸ and others.²⁹ CBT focuses on (1) identifying situations that precipitate drug use and (2) preventing relapse by teaching cognitive and behavioural skills to reduce risk. CBT-trained master-level therapists will be supervised by Dr Schmitz (PI), a licensed clinical psychologist. We will follow recommended guidelines for ensuring treatment fidelity, including use of written manuals, audiotaping and review of sessions, adherence and competency ratings by supervisor and independent raters, and ongoing training and review to prevent deviation or drift from the therapy manual.^{30 31}

Data collection and measures

Target mechanisms

Electroencephalogram (EEG) will be collected to measure LPP responses to a wide array of cocaine-related and non-drug-related motivationally relevant stimuli. The picture viewing protocol will be based on published standardised parameters^{32 33} and on our preliminary study of LPP responses in individuals with CUD.³⁴ EEG will be collected while participants passively view a slideshow of images depicting scenes with varying emotional content (pleasant, unpleasant, neutral and cocaine related). Images will be taken from the International Affective Pictures System³⁵ and from picture sets used in previous LPP cocaine studies.³⁴³⁶,⁴⁰

EEG acquisition and analysis

EEG data will be collected with a 64-channel actiCAP electrode cap, amplified with BrainAmp MR, and digitised using Brain Vision Recorder (Brain Products, Munich). The sampling rate will be 500 Hz and data will be filtered with 1 Hz high-pass and 100 Hz low-pass filters. Data reduction will be performed with Brain Vision Analyzer 2, MATLAB and BESA.⁴¹ We will define the LPP as the mean amplitude between 400 and 800 ms poststimulus over the central-parietal electrodes, and we will use individual LPP responses to estimate the motivational relevance of drug-related and non-drug-related emotional stimuli.⁴²,⁴⁴

Cocaine demand will be assessed using the hypothetical Cocaine Purchasing Task,^{45 46} which we have found to be significantly associated with clinically relevant measures of cocaine use and treatment outcome. To help ground responses, participants will first be asked what the highest number of cocaine rocks they have previously consumed in a 24-hour period and what the total cost was for that amount of cocaine.⁴⁷ Participants will be asked how many rocks of cocaine they would purchase at various prices (ie, free, 1, 2, 5, 10, 20, 50, 100, 200, 500, 1000 dollars). Participants will be asked to assume the following: that their income and savings are what they usually are; the quality of cocaine is the type they usually purchase; there are no other sources of cocaine; any cocaine purchased has to be used within that day; and their craving and desire for cocaine is how they currently feel. The instructions and range of prices for cocaine are similar to those previously reported in the literature and successfully utilised by our group.^{46 48}

Demand data collected from the CPT will first be assessed for general orderliness and systematicity using established criteria from the field.^{49 50} Data will subsequently be fitted to the exponentiated demand model, which is widely used in the field.⁵¹ The exponentiated demand model produces the following demand metrics: Q₀—intensity of demand—purchase and consumption of drug at or near price zero; α—a fitted parameter characterising the rate of change in the slope of the demand function; O_max—the maximum output or expenditure observed (units of drug X cost of drug at a given price point); P_max—price at which O_max is observed; and breakpoint (BP)—the lowest price at which drug is no longer purchased and consumed. For greater ease of interpretation and statistical analysis, α will be converted to essential value (EV) using an inverse transformation.⁵² Data from zero responders (individuals reporting zero cocaine consumption at all prices) will have all demand metrics (Q₀, EV, O_max, P_max, BP) defined as zero as previously established in the field^{53 54} and followed by our group.^{45 46}

Craving will be assessed using the CCQ⁵⁵ from the NIH PhenX Toolkit.⁵⁶ The CCQ includes 10 statements about the participant’s current thoughts and feeling about using cocaine with responses ranging from ‘strongly disagree’ to ‘strongly agree’. Total craving intensity is derived as the mean response to 10 statements with higher scores indicating greater craving.^{57 58}

Drug use

TLFB²² interviews will be conducted by trained research assistants (masked to treatment condition) at each clinic visit to evaluate the participant’s self-reported substance use since the previous visit. Participants will be asked about their consumption of cocaine, nicotine/tobacco, alcohol, cannabis and other psychoactive drugs for each day. Information regarding quantity and form of substance used will be collected, however, for the purpose of this study, the primary outcome variable of interest will be the frequency of use (days used).

UDS will be collected under observed conditions at each clinic visit. Cocaine use will be tested for the presence of BE with a 150 ng/mL cut-off considered to be positive.

Safety

Safety measures listed in table 1 will be consistently monitored throughout the treatment phase. In addition to medical assessments (vital signs, body weight, blood glucose, potential side effects of the study medication, etc), we will assess for reports of suicidality and changes in mood/behaviours using the Patient Health Questionnaire⁵⁹ and the Columbia Suicide Severity Rating Scale.⁶⁰

Treatment acceptability

Treatment acceptability data will be collected to inform future medication development efforts. The Client Satisfaction Questionnaire (CSQ-9)⁶¹ will assess how much the participant perceived treatment as helpful, meeting their needs and worth recommending or receiving again, if needed. The CSQ-9 will be adapted slightly to assess satisfaction with the once-weekly subcutaneous injection mode of treatment. Each item is scored from 1 (poor) to 4 (excellent) with higher scores, indicating higher treatment satisfaction.

Other measures

Health-Related Quality of Life (HRQoL) will be assessed using the Patient-Reported Outcomes Measurement Information System-29 (PROMIS⁶²), a 29-item multidimensional measure of health, consisting of seven domains: Physical Functioning, Anxiety, Depression, Fatigue, Sleep Disturbance, Satisfaction with Participation in Social Roles, and Pain Interference. Raw scores for each domain are calculated by summing the item scores and transforming to T scores (M=50; SD=10). Higher T scores imply a higher level of the domain being measured.⁶³

The use of other drugs, including tobacco smoking, alcohol and cannabis will be assessed via TLFB procedures. A set of validated measures assessing the subjective, reinforcing and aversive effects of smoking (Modified Cigarette Evaluation Questionnaire (MCEQ)⁶⁴), alcohol (Alcohol Craving Experience (ACE) Questionnaire)⁶⁵ and cannabis (Cannabis Experiences Questionnaire (CEQ))⁶⁶ will be also included. Additionally, two measures of food craving, the Food-Craving Inventory (FCI)⁶⁷ and the Reward-based Eating Drive (RED) scale⁶⁸ will be administered to assess the effects of semaglutide on food craving and consumption.

Outcome measures

Primary outcomes

• LPP (EOT vs baseline).

• Cocaine Purchasing Task demand indices (EOT vs baseline).

• CCQ score (EOT vs baseline).

• Proportion of cocaine-negative UDS during the final 2 weeks of treatment (weeks 13 and 14).

Secondary outcome

• Treatment response as a dichotomous outcome, defined as at least three cocaine-negative UDS during the final 2 weeks of treatment (weeks 13 and 14).

Safety and monitoring

Medical monitoring

Participants’ vital signs, body weight, overall well-being and potential side effects of semaglutide will be assessed at each visit, and finger stick blood glucose levels will be checked prior to each dose of semaglutide. Additionally, participants will be closely monitored for the emergence or worsening of depression symptoms, suicidal thoughts or behaviour and/or any unusual changes in mood or behaviour. Appropriate clinical follow-up will be initiated if acute pancreatitis is suspected, if the participant experiences suicidal thoughts or behaviours, or if there are other safety concerns as judged by the PI and study physician.

Adverse event assessment and management

All adverse events (AE) and serious AE (SAE) occurring during the study will be collected and documented. The PI, in consultation with the Study Physician, will review any AEs as soon as they are reported. Each week the PI and/or Study Physician will review the AE forms from the previous week for events that were reported as new or continuing. All AEs will be followed up to the point of satisfactory resolution. A study participant may have their medication discontinued or may be withdrawn from the study if the PI and/or Study Physician determines it is the best decision in order to protect the safety of the participant. AEs deemed to be SAE will be reported to the IRB, NIDA and the FDA.

Data collection, entry and storage

Checklists will guide study personnel in all procedures for data collection and data management. Whenever possible, questionnaire data are entered directly into the computer by the participant, reducing data entry errors. All neurobehavioural data are collected automatically by the computer using professional programmes designed for this purpose and do not allow missing or impossible values. All hand-entered data (miscellaneous study forms) are initially reviewed by the study coordinator for general accuracy and completeness. They are then double-entered in REDCap, with any inconsistencies examined and resolved. Data quality will be monitored continuously by the Quality Control Manager and any problems detected will be discussed with the PIs. If necessary, retraining of data collectors will be conducted. Only investigators will have access to the data.

Data security and plan for protecting confidentiality

All data collected on paper forms will be stored in locked cabinets, while electronic data will be stored in REDCap database on a secure password-protected server maintained by the UTHealth Medical School Information Technology Department. Individual participants and their research data will be identified by a unique study identification number (study ID). The study data entry and study management systems will be secured and password-protected. The participants’ contact information will be securely stored for internal use during the study. At the end of the study, all records will continue to be kept in a secure location for as long a period as dictated by IRB and Institutional regulations.

Data monitoring

The PI and research team will monitor data on an ongoing basis. In addition, a data safety and monitoring board (DSMB) will oversee the ongoing progress of the study. The board consists of two physicians, a psychologist, and a biostatistician who are not affiliated with the study or the sponsor. Each member will provide written documentation attesting to the absence of conflict of interest. The initial responsibility of the DSMB is to review and approve the initiation of the trial. Thereafter, DSMB evaluations will be conducted on an annual basis. On study completion, the DSMB will perform an end-of-study evaluation. An emergency meeting of the DSMB (open, closed or executive session) may be called at any time by the Chairperson should questions of patient safety arise.

Trial stopping rules

The trial may be suspended or prematurely terminated if there is sufficient reasonable cause. Circumstances that may warrant termination or suspension include, but are not limited to, determination of unexpected, significant or unacceptable risk to participants; insufficient compliance with protocol requirements; and/or data that are not sufficiently complete and/or evaluable. If the study is suspended, it may resume once concerns about safety, protocol compliance and data quality are addressed and satisfy the suspending party and the IRB.

Protocol modifications

The study PI will receive IRB approval before initiating any changes, including those required by the sponsor, which would affect study participants, such as changes in methods or procedures, the number or types of study participants or revisions to the informed consent document or procedures. All protocol revisions will be submitted to the primary sponsor of research.

Project timeline

The recruitment for this study is projected to begin in November 2025; and the estimated primary completion date (final data collection for primary outcome measures) is January 2029. It is anticipated that 2–3 participants will be randomised each month for a total of 75 participants during the 33-month enrolment period. Once randomised, participant duration in the study is 23 weeks.

Patient and public involvement

None.

Statistical analysis plan

Inferential paradigm

The goal of this initial proof-of-concept trial is to make robust and accurate decisions about whether (or not) to move forward with a promising new medication treatment for CUD. In this context, the Bayesian approach is inherently well-suited, with advantages over classic frequentist analysis that have been widely published.⁶⁹,⁷¹ Most notably, as opposed to the classical focus on p values to reject the null hypothesis, Bayesian statistics focuses on evidence supporting the alternative hypothesis, quantified by probability distributions.

Detailed descriptions of Bayesian statistical reasoning exist elsewhere.^{69 72} Succinctly, Bayesian probability estimates incorporate prior information about plausible parameter values (ie, the prior distribution) and the observed data (ie, the likelihood). Combining these two distribution forms, the posterior distribution, which permits evaluation of the probability that the true value of the parameter falls in some range. Bayesian methods have been incorporated in clinical trial designs^{70 73 74} and FDA guidelines.⁷⁵ Our group has published several studies using Bayesian methods to evaluate treatments for substance use disorders.⁷⁶,⁸³

General data analytic strategy

Preliminary data analyses will characterise the sample in terms of demographic and baseline variables. Evaluation of variables showing a correlation with treatment and outcomes will be conducted to identify potential confounds, which will then be used as covariates in the analysis. Broadly, the data analytic strategy will use generalised linear and multilevel models. Multilevel approaches will account for correlations due to repeated measures within participants via random effects (R V.4.2.1 and Stan,V.2.30; packages rstanarm and brms) for both continuous and discrete outcomes.²,⁴ All analyses will use intention-to-treat principles. Bayesian approaches will implement joint modelling of observed outcomes and the missing data, which is robust to ignorable missingness (ie, MCAR and MAR).^{84 85} Sensitivity analyses will evaluate the robustness of analytic conclusions to missing data. Non-ignorable missing data patterns will be addressed through pattern-mixture modelling methods.⁸⁶ Specification of weakly informative, neutral priors will result in conservative, regularised estimates of effect. For generalised linear and multilevel models, priors for regression coefficients will be specified as ~normal (µ=0, σ=1 x 10²) for continuous outcomes (with some adjustment for scaling), and ~normal (µ=0, σ=0.56) in the log-form for discrete outcomes (this latter prior centres risk-ratios on the null hypothesis while assuming a 95% chance that the risk ratio will be in the interval 0.33–3.00). Level 1 and 2 error/random effect variances will be specified as ~Half T (df=3, mean=0, SD=1×10²). Choice of prior distribution for level two variances will follow recommendations by Gelman.⁸⁷

Power analysis

Assumptions

Calculations focus on the effect of treatment (semaglutide vs placebo) on the primary outcomes in hypothesis 1 (target mechanisms) and hypothesis 2 (cocaine use). All calculations are based on generating k=1000 Monte Carlo simulations. Statistical power is then conceptualised as the percentage of these posterior distributions that meet or exceed a specific probability threshold of interest. Given the early stage of research on GLP-1RAs in treating CUD and that the danger at this juncture is making a ‘type II’ error, which might bring investigation to a premature halt, we stipulate that, in the absence of differential adverse events, a Bayesian posterior probability >0.70 that an effect exists, favouring semaglutide, will constitute sufficient evidence to warrant investigation in a larger, confirmatory trial, consistent with probability thresholds stipulated for go/no-go decisions in other medication trials.⁸⁸,⁹²

Power calculation and sample size

Due to the novelty of semaglutide (and GLP-1RAs) in the treatment of CUD, there are currently no published between-group effect sizes for the primary outcome measures of aim 1 and aim 2. Therefore, we obtained estimates comparable data. Regarding ERP reactivity and cocaine craving, the Parvaz study⁴⁰ showed LPP reversal from baseline (drug >pleasant cues) to follow-up (pleasant >drug) in 19 treatment-seeking individuals with CUD (vs 18 controls) (t₁₈=2.66, p=0.016), with the change in drug-pleasant cue LPPs associated with a reduction in cocaine craving (r_s=−0.477, p=0.045). For drug demand, we published the only study to date showing a reduction in cocaine demand over the course of CUD treatment with behavioural therapy among treatment responders, with effect sizes ranging from d’s of 0.24 to 0.47 for demand metrics.⁴⁶ For cocaine use outcomes (aim 2), the recent Amin-Esmaeili et al study⁹³ which analysed data from 13 RCTs of pharmacological interventions found that 20.9% of participants transitioned to reduced frequency of cocaine use, verified by cocaine-negative UDS at the end of trial. Conservatively, we propose that a minimum absolute difference of 10% in proportion of cocaine-negative UDS during the final 2 weeks of treatment, favouring semaglutide over placebo, would constitute a clinically meaningful effect.

Hypothesis testing

Primary aims 1 and 2

Aim 1: determine the effects of semaglutide (vs placebo) on targeted mechanisms of change in the treatment of CUD.

Hypothesis 1 a: semaglutide will attenuate ERP reactivity (LPP) specific to cocaine-related stimuli. Primary outcome: LPP relative amplitude differences between nondrug pleasant images and cocaine images at week 15 (EOT).

Hypothesis 1b: semaglutide will reduce motivation to purchase and consume cocaine as measured by behavioural economic indices (drug demand). Primary outcome: CPT demand indices (Q₀, EV, O_max, P_max, BP) at week 15 (EOT).

Hypothesis 1c: semaglutide will reduce cocaine craving as measured psychometrically using the CCQ. Primary outcome: CCQ score at week 15 (EOT).

Primary analysis: these target mechanism outcomes will be analysed using GLM with the week 15 endpoint value as the dependent variable, treatment assignment as the primary predictor, and baseline values of the respective outcome measures and stratification variables as covariates.

Aim 2: examine the effects of semaglutide (vs placebo) on cocaine use.

Hypothesis 2: semaglutide will reduce cocaine use, as measured by the percentage of cocaine-negative UDS. Primary outcome: proportion of cocaine-negative UDS during the final 2 weeks of treatment (weeks 13–14). Secondary outcome: response rate, defined as at least 3 cocaine-negative UDS out of the 4 samples obtained during the final 2 weeks of treatment (weeks 13 and 14).

Primary analysis: the proportion of cocaine-negative UDS during the final 2 weeks of treatment (weeks 13–14) will be analysed using GLM with proportion of cocaine-negative UDS as the dependent variable, treatment assignment as the primary predictor, and stratification variables as covariates.

Secondary analysis: GLMs will evaluate response (yes/no) as a function of treatment after adjusting for stratification variables.

Exploratory aims analyses

Exploratory aim 3: explore the association between changes in target mechanisms and cocaine use.

Primary analysis: although modelling change in one longitudinal process (ie, LPP, demand, craving) as a predictor of some parallel process (cocaine-negative UDS) might best be accomplished using structural equation modelling, the proposed sample size suggests a more modest approach. We will model treatment mechanisms longitudinally using GLMMs for repeated measures at weeks 5, 9 and 15. The resulting output will provide random effects for each participant in the analysis. Random effects (eg, intercepts and slopes) will then serve as predictors in a GLM that will evaluate cocaine use at weeks 13 and 14. Initially, to improve interpretability, each participant’s slope will be coded as indicating improvement (slope <0) or non-improvement (slope >0) for each target mechanism measure. Having adjusted for each participant’s intercept (baseline), GLMs will evaluate primary and secondary cocaine use outcomes (Hyp 2) as a function of (non)/improvement.

Exploratory aim 4: explore the effects of semaglutide on the use of other drugs, including tobacco smoking, alcohol intake and cannabis. Primary outcome: frequency of use defined using TLFB self-report percentages of days of drug use during treatment, examined separately for cigarette smoking, alcohol use and cannabis use.

Primary analysis: GLMs will estimate the proportion of days used for each drug separately as a function of time, treatment and the interaction of time and treatment, after adjusting for stratification variables.

Secondary analysis: GLMs will evaluate subjective effects of tobacco smoking (MCEQ), alcohol (ACE), cannabis (CEQ) and food (FCI, RED) as a function of time, treatment, and their interaction, after adjusting for stratification variables.

Exploratory aim 5: characterise dose–response effects during semaglutide titration.

Primary analysis: in addition to EOT effects (aims 1 and 2), longitudinal models will apply generalised multilevel approaches to evaluate each primary outcome (LPP, demand, craving, cocaine use) as a function of time, treatment (as a time-varying covariate) and the interaction of time and treatment. Balancing the magnitude of this difference against the report of any adverse events will permit consideration of dosing regimens in future clinical trials.

Interim analyses

Not applicable for this trial.

Ethics and dissemination

On ClinicalTrials.gov, the study can be identified by the ID NCT07227948. The study approval has been received from the UTHealth Committee for the Protection of Human Subjects (HSC-MS-25-0412). All participants will sign informed consent. The study results will be disseminated via peer-reviewed publications and conference presentations.

Discussion

This phase IIa RCT is designed to evaluate semaglutide, a GLP-1RA, as an adjunctive pharmacotherapy for CUD. Supported by a growing body of evidence suggesting that GLP-1R signalling may modulate reward-related behaviours, the study has several notable strengths. First, this trial addresses a critical gap in treatment options for CUD, a condition with no FDA-approved pharmacotherapies despite its significant public health burden. Furthermore, the repurposing of an existing medication represents a pragmatic and potentially expedited path to clinical implementation. Second, the study uses an experimental therapeutics framework, examining target engagement alongside clinical outcomes. This mechanistic insight is essential for refining treatment strategies and identifying biomarkers of response. Third, the inclusion of exploratory aims allows for a broader understanding of the impact of semaglutide, including its effects on the use of other substances and food-related behaviours. These data may inform future research on the transdiagnostic utility of GLP-1RAs for polysubstance use as well as cooccurring addictive behaviours and metabolic dysregulation. Fourth, evaluating responses across dose levels of semaglutide will provide initial data that would inform the design and optimisation of future clinical trials, including dosing strategies, efficacy assessments and safety monitoring.

An additional strength of the study is its rigorous methodological features, including stratified randomisation, blinded outcome assessment and a comprehensive statistical analysis plan using Bayesian inference. These elements enhance the internal validity of the trial and support robust interpretation of findings.

However, several limitations should be acknowledged. The relatively short treatment duration (14 weeks) may limit the ability to detect long-term effects or sustained abstinence. Additionally, the exclusion criteria, while necessary for safety, may reduce generalisability to broader clinical populations, particularly those with comorbid medical or psychiatric conditions. The reliance on self-reported substance use, although supplemented by UDSs, may also introduce reporting bias.

Despite these limitations, this trial represents a significant step forward in the development of pharmacotherapies for CUD. By combining a novel pharmacological agent with a well-established behavioural intervention and by rigorously evaluating both clinical and mechanistic outcomes, this study has the potential to yield actionable insights that can inform future research and clinical practice.