Targeting white matter neuroprotection as a relapse prevention strategy for treatment of cocaine use disorder: Design of a mechanism-focused randomized clinical trial

Abstract

Cocaine use continues to be a significant public health problem with limited treatment options and no approved pharmacotherapies. Cognitive-behavioral therapy (CBT) remains the mainstay treatment for preventing relapse, however, people with chronic cocaine use display cognitive impairments that are associated with poor response to CBT. Emerging evidence in animal and human studies suggests that the peroxisome proliferator-activated receptor-gamma (PPAR- γ) agonist, pioglitazone, improves white matter integrity that is essential for cognitive function. This project will determine whether adjunctive use of pioglitazone enhances the effect of CBT in preventing relapse during the early phase of recovery from cocaine use disorder. This paper describes the design of a mechanism-focused phase 2 randomized clinical trial that aims first to evaluate the effects of pioglitazone on targeted mechanisms related to white matter integrity, cognitive function, and cocaine craving; and second, to evaluate the extent to which improvements on target mechanisms predict CBT response. Positive results will support pioglitazone as a potential cognitive enhancing agent to advance to later stage medication development research.

1. Introduction

The U.S. is facing a fourth wave in the overdose crisis, marked by a surging rise in the availability and use of psychostimulants, including cocaine, following a previous period of decline (1, 2). Although strides have been made in medication development for the treatment of cocaine use disorder (CUD), no FDA-approved pharmacotherapies are currently available. Effective behavioral treatments exist; still, the majority of patients relapse within one year.

Cognitive-behavioral therapy (CBT) remains the mainstay of treatment for preventing relapse (3). CBT targets behavior change through learning theory principles. Implicit to the effectiveness of learning theory applications are intact cognitive functions that mediate behavior (e.g. executive control, memory, attention) such that limitations in these constituent abilities may diminish robustness of CBT treatment response. These cognitive functions have their basis in the neurobiological integrity of the brain. It follows that CUD patients with greater cognitive impairments are less likely to benefit from CBT. This is particularly challenging, given that chronic cocaine use is closely associated with deficits across a wide range of cognitive domains. As such, pharmacological interventions with cognitive-enhancing effects have been recommended to improve response to CBT (4). Medications that act on catecholamine neurotransmitters have been considered, but remain in the nascent stage of development. Medication approaches that focus on alternative pathways for treating cognitive deficits in CUD are needed.

A growing empirical literature supports the rationale for treatments targeting white matter (WM) structure in the brain to improve or help preserve cognitive function in substance use disorders. Chronic cocaine exposure alters WM structural integrity as measured by diffusion tensor imaging (DTI) and these alterations compromise cognitive function in many neurological and psychiatric disorders, including CUD (5, 6). Moreover, better WM integrity predicts better CUD treatment outcome (7). Peroxisome proliferator-activated receptors (PPARs) are ligand-activated transcription factors involved in the regulation of energy homeostasis and inflammation. Therapeutically, PPAR-γ agonists exert protective effects in preserving or restoring WM integrity in diseases of the CNS, such as stroke and Parkinson’s disease (8–10). The FDA-approved PPAR-γ agonist pioglitazone has shown potentially beneficial effects in preclinical models of addiction and cognition (8, 11), leading to recent pilot testing of pioglitazone in the treatment of CUD (12). While the exact neurobiological mechanisms are not yet clear, it is plausible to consider the neuroprotective role of PPAR-γ agonism as a cognitive-enhancing strategy for improving the effectiveness of CBT in preventing relapse. Here we describe the design of a phase 2 mechanism-focused randomized clinical trial to evaluate pioglitazone as an adjunct to CBT for the treatment of CUD.

2. Study rationale and methods

Behavioral therapy has been the mainstay of treatment for relapse prevention in CUD. The CBT model of relapse, introduced by Marlatt and Gordon in 1985, is arguably the most empirically supported and widely used approach for reducing the probability of relapse (13). CBT for relapse prevention involves teaching patients how to manage or cope with high-risk situations that trigger drug use. In learning these skills, the individual gains awareness of the causal relationship between antecedents and consequences of their use (i.e., functional analysis), while gaining cognitive control over problematic thoughts and impulsive responses. Findings from systematic reviews and large-scale treatment outcome studies have generally supported the clinical effectiveness of CBT (14–16), although effect sizes are modest, bolstering the call for research aimed at improving effects with integrative treatments. Recent examples support the paradigm of medication-facilitated treatment to boost CBT response (for review, see 4).

Deficits in performance across a range of executive function tasks have been repeatedly demonstrated in CUD and linked to dysfunction in prefrontal cortex regions (17–20). We have conducted a series of studies showing significant differences in brain structure and function in patients with CUD compared to non-drug-using controls while performing tasks of working memory (21), decision-making (22), and inhibitory behavioral control (23). Importantly, the literature supports an association between cognitive deficits at baseline and subsequent treatment outcome (retention and abstinence) in the context of clinical trials using CBT as the platform (21, 24–29). Longitudinal evidence suggests that cognitive deficits covary with changing cocaine use. In a study by Vonmoos and colleagues (30), increased cocaine use predicted additional cognitive decline within 1 year, whereas decreased cocaine use was associated with cognitive improvements within 1 year, particularly in attention and memory domains. Importantly, in patients who ceased using cocaine within 1 year there was improvement in cognitive function (comparable to the control group), suggesting that cocaine-induced deleterious neuroadaptations may be in part modifiable though psychotherapy and/or pharmacotherapy interventions.

Thus, cognitive enhancement as a pharmacotherapy target for the treatment of CUD is an emerging and novel area of addiction research. The present project takes an experimental therapeutics approach by first establishing medication effects on cognitive function as the hypothesized target mechanism (31). Pilot results from our proof-of-concept trial in a sample of treatment-seeking patients with CUD demonstrated evidence of target engagement by showing that pioglitazone produced measurable improvement (pre-to-post treatment changes) on diffusion tensor imaging (DTI) of WM integrity as measured by fractional anisotropy (FA) and radial diffusivity (RD) values in four target regions-of-interest (12). By modulating processes underlying WM integrity, pioglitazone might enhance cognitive processes required for effective CBT, as WM integrity is strongly associated with a broad array of cognitive factors and neurological disease processes (32).

Unlike traditional medications that target classic neurotransmitter systems, pioglitazone’s activation of the PPAR pathway has broad spectrum anti-inflammatory and anti-oxidative effects (9). These effects have made pioglitazone an attractive treatment option for CNS injuries and other “white matter” diseases, e.g., multiple sclerosis (MS), involving mechanisms of demyelination where targeting PPAR promotes neuroprotection and myelin formation (33–36), consistent with the present proposed therapeutic goal in treatment of CUD relapse. Importantly, changes in WM in neurological disease correspond to changes in symptom presentation, including cognitive function (37–39).

The proposed study will be the first of its kind to exploit WM integrity as a pharmacologic target for improving CBT outcomes in the context of relapse prevention. For the sizeable majority of patients with CUD who present with significant cognitive impairments, identifying a novel pharmacologic agent that confers clinical benefit via improving WM integrity would have high impact on the field of medication development.

2.1. Study aims and hypothesis

We propose an early-stage randomized clinical trial that is consistent with an experimental therapeutics approach. The overarching goal of this trial is to demonstrate that pioglitazone treatment directed at changing targeted mechanisms will be associated with clinical efficacy.

Specific study aims and corresponding hypotheses are twofold. Aim 1 will examine the effects of pioglitazone versus placebo on targeted mechanisms of change in recently abstinent patients with CUD. We hypothesize that pioglitazone will: (1) increase WM integrity indexed by DTI metrics; (2) improve performance on measures of cognitive function; (3) reduce self-reported craving for cocaine. Aim 2 will demonstrate evidence linking clinical efficacy with mechanism engagement. We hypothesize that: (1) pioglitazone will be associated with continuous abstinence (reduced relapse) and improved functional health outcomes at the end of treatment and follow-up; and (2) improvement on targeted mechanisms will be associated with abstinence and functional health at the end of treatment and follow-up.

2.2. Trial design overview

The trial will employ a standard relapse prevention paradigm such that participants achieve initial abstinence from cocaine while residing at a facility for brief (5-day) detoxification (detox) program during study medication induction, with relapse to cocaine use assessed during subsequent outpatient treatment. As shown below in Figure 1, prior to undergoing inpatient detox, eligible participants who complete screening and baseline assessment will be randomly assigned to one of the two treatment conditions, CBT + pioglitazone or CBT + placebo, using urn randomization to ensure balance between groups on cocaine severity, sex, age, and other (non-cocaine) substance use status. Upon discharge from detox, participants will continue to receive 12 weeks of their assigned treatment. All participants will attend 3 study visits per week to receive study medication and individual CBT. Measurement of cocaine use by self-report and urine drug screen (UDS) results will occur at each visit and at the one-month follow-up visit.

Figure 1.

Overview of study design. Note. CBT = Cognitive Behavioral Therapy.

2.3. Recruitment and eligibility

The study will enroll treatment-seeking individuals, 18 to 60 years old, who meet current DSM-5 criteria for CUD of at least moderate severity (≥4 symptoms). Recruitment strategies will include print and radio advertisements in the local media, flyers posted in public settings, and through clinical referrals. Eligible subjects must submit at least one positive UDS for the cocaine metabolite, benzoylecgonine (BE ≥ 150 ng/ml) during intake to ensure enrollment of individuals actively using cocaine. To be included, individuals need to be willing to be admitted to a 5-day inpatient detox program. Those meeting moderate or severe diagnostic criteria for substance use disorders other than cocaine, marijuana, or nicotine will be excluded. Other exclusion criteria will include having a significant and unstable medical/psychiatric disorder or taking medications (e.g., CYP2C8 inhibitors or inducers, antihyperglycemic medications) which are contraindicated for pioglitazone pharmacotherapy. The following medical conditions will be exclusionary: medication- or insulin-dependent diabetes, congestive heart failure, edema, clinically significant liver disease, hypoglycemia, or history of bladder cancer. Having medical contraindications to MRI scans, e.g., history of pacemaker or metal implants, will be exclusionary. No pregnant women will be permitted in the study. Females of childbearing potential must agree to use an acceptable method of birth control during study participation and for one month after discontinuation of the study medication.

The UTHealth Committee for the Protection of Human Subjects (local IRB) approved informed consent will be obtained from all participants. This NIH-funded trial is protected by a Certificate of Confidentiality and registered at ClinicalTrials.gov [NCT04843046].

2.4. Treatments

2.4.1. Cognitive Behavioral Therapy (CBT)

Provision of relapse-prevention focused CBT (13) during the 12-week outpatient treatment phase will assist patients in ongoing recovery to maintain abstinence. Session 1 will take place on the day of discharge from the inpatient detox facility. Scheduled 1-hour individual therapy sessions will occur twice weekly during weeks 1–4, followed by once per week sessions through weeks 5–12. This front-loading of therapy intensifies treatment during the initial month post-discharge; a high-risk period for relapse (40, 41). The CBT manual will parallel existing protocols that have used this behavioral therapy platform in pharmacotherapy trials (42–46). CBT focuses on identifying situations that precipitate relapse and teaching cognitive and behavioral skills to reduce risk. CBT-trained masters-level therapists will deliver this intervention. Salient guidelines for ensuring treatment fidelity include audiotaping and review of all sessions, assessing adherence and competency by supervisor and independent raters, and providing ongoing training to prevent deviation or drift from the therapy manual (47).

2.4.2. Pharmacotherapy

Pioglitazone is an FDA-approved treatment for diabetes mellitus type 2. In addition to its anti-diabetic properties, pioglitazone has anti-inflammatory, neuroprotective, antioxidative, and anti-excitotoxic properties (48, 49). Given the wide variety of actions, pioglitazone has been studied as a treatment agent in patients with neuropsychiatric diseases including Alzheimer’s disease and multiple sclerosis (e.g., 50, 51).

The medication schedule is based on results from our pilot trial showing that pioglitazone 45 mg was associated with acceptable levels of tolerability, safety, and compliance (12). Thus, we will use the highest acceptable dose according to existing guidelines, with the primary goal of replicating target engagement. While it is possible that the optimal dosing parameters are higher or lower than 45 mg, testing a broader dose range would be considerably more expensive and require a much larger study.

We will follow recommended adult initial dosing at 30 mg (days 3–5 of detox) to reach maintenance dose of 45 mg by discharge (start of outpatient week 1). This more rapid dose titration is expected to provide a faster onset of therapeutic benefit to coincide with transitioning to outpatient CBT. This titration schedule will allow achievement of steady-state based on pioglitazone’s half-life (24 hrs.), and is within standard titration parameters as per the Package Insert. Inactive placebo will be administered in identical capsules on the same schedule. All investigators and staff, except the pharmacist, will be blind to medication assignment. At the end of the treatment we will assess the integrity of the study blind by having participants and the study physician judge to which medication group the participant had been assigned (52).

2.5. Measures

2.5.1. Screening and eligibility

Consenting subjects will receive a comprehensive medical and psychiatric evaluation including laboratory chemistries (blood chemistry screen, complete blood count, urinalysis and serum pregnancy test) and electrocardiogram (ECG). Masters-level clinicians will conduct the Structured Clinical Interview for DSM-5 (SCID: 53), the Addiction Severity Index (ASI: 54) the Kreek-McHugh-Schluger-Kellogg (KMSK: 55) assessment of lifetime substance use interview, and the Timeline Followback (TLFB: 56) interview.

2.5.2. Target mechanisms

Change in WM integrity, measured at baseline and week 12, will serve as a primary target mechanism. DTI scans will be acquired on a Philips Ingenia 3T magnet with a 32 channel. DTI analyses, including segmentation and region of interest (ROI) determination, will follow atlas-based methods and quality assurance protocols for serial stability (57–59). ROI analyses of FA will focus on commissural fibers (genu and splenium of corpus callosum), projection fibers (anterior and posterior thalamic radiations), and association fibers (cingulum and the external capsule). Additionally, we will conduct a whole brain DTI analysis using tract-based spatial statistics methodology to explore associations between whole brain WM integrity, treatment condition, duration of abstinence, and cognitive function. In addition to FA we will analyze all primary DTI metrics, including axial, radial, and mean diffusivities in order to examine microstructural specificity associated with pioglitazone treatment, as per our pilot study (12).

Cognitive function will be measured at baseline, week 4 and week 12, using the NIH Cognition Toolbox test battery (60). The Toolbox includes Picture Sequence (Memory), Flanker (inhibitory control and attention), List Sorting (working memory), Picture Vocabulary, Reading Recognition (reading decoding skills), Dimensional Change Card Sort (cognitive flexibility/executive function), and Pattern Comparison (Processing Speed). The Toolbox Cognition Battery yields three summary scores: Crystallized Cognition Composite, Fluid Cognition Composite, and Overall Cognition Composite (61). Consistent with recommendations by Kwako (62, 63), our analyses will specifically examine the List Sorting, Flanker Task, and the Dimension Change Card sort tests from the NIH Toolbox to parse out these critical addiction-related cognitive domains.

Craving for cocaine will be measured once weekly at outpatient study visits. The Brief Substance Craving Scale (BSCS: 64) is a 16-item, self-report instrument assessing craving for cocaine and other substances over a 24 hour period. Intensity, duration, and frequency of craving are rated on a 4-point Likert scale. The sum of the 3 scores yields a craving composite measure. A Visual Analogue Scale (VAS) will be used to measure cocaine craving intensity experienced during the previous week. Three rating scales (100 mm line, anchored by 0 “not at all” and 100 “extremely”) will ask questions related to craving right now, craving on average in the past week, and the worst craving in the past week.

2.5.3. Clinical outcomes

Cocaine use will be assessed at each clinic visit using objective (UDS) and self-report (TLFB) measures. The primary outcome of abstinence (non-relapse) will be defined as continuous abstinence in the final three weeks of treatment via self-reported non-use days confirmed by negative UDS results. Additional outcome measures that have been recommended as indicators of treatment success (65–68) will be assessed, including (1) total percentage of days abstinent during 12-week treatment; (2) total percentage of cocaine-negative UDS during 12-week treatment; (3) time to relapse. For this third outcome variable, “relapse” will be defined as 3 or more consecutive cocaine-positive UDS during treatment. The occurrence of this event will be coded at the time of the fourth positive UDS. Thus, in addition to sustained abstinence (primary outcome), we will obtain a more nuanced interpretation of the ability of pioglitazone to delay relapse, consistent with other cocaine trials in recently abstinent CUD patients (e.g., 69).

Functional health status will be assessed using the Patient-Reported Outcomes Measurement Information System (PROMIS). The PROMIS is supported by the NIH roadmap effort to standardize and promote a common measurement system for clinical research (70). Specifically, we will use the PROMIS-29 to assess seven domains: physical function, anxiety, depression, fatigue, sleep disturbance, satisfaction with social role, and pain interference/intensity, with four questions per domain and an additional pain intensity 0–10 numeric rating scale. Total raw scores representing the sum of the values of the response to each question within each domain are converted into standardized T-scores. The PROMIS provides global health summary scores representing physical and mental health (71).

2.5.4. Medication monitoring

Multiple methods of monitoring medication compliance will be used. Riboflavin will be added to study medication capsules as an oral medication ingestion marker (72). Nurse-observed dosing will occur at thrice-weekly outpatient clinic visits. Pill counts of returned blister packs will be conducted. A novel cell-phone assisted remote observation of medication adherence (CAROMA: 73) approach will be used in which participants will be video-called by research staff who will observe consumption of study medication on non-clinic visit days.

2.6. Follow-up

All participants, including those who do not complete the entire treatment regimen, will be invited to attend a follow-up assessment at four weeks post treatment. Primary clinical outcome data, described above (Sect. 2.5.3) will be collected at this visit. Recommended retention strategies will be used to achieve a ≥80% follow-up rate (74). The research team will stay in regular contact with participants throughout the study and at the point of completion. Seven days prior to the follow-up appointment, participants will receive text reminders of their scheduled visit. As-needed employment of additional outreach procedures may include contacting locators, sending letters, emails, and searching social media networks.

3. Statistical analysis plan and power

Given recommendations that all clinical trials should be analyzed in Bayesian and conventional Frequentist fashion (75–77), we will implement parallel analyses for evaluation of Aims 1 and 2, with Bayesian approach as the primary outcome analysis. The two approaches will provide complementary information in this early phase II clinical trial where our primary goals are to increment the precision of effect-size estimates and evaluate the probability that the alternative hypothesis is true. Whereas frequentist results yield the probability of the observed data, or data more extreme, given that the null hypothesis (H₀) holds, Bayesian results yield a straightforward and direct estimate of the probability of the alternative hypothesis (H₁) existing, given the data, thus permitting use of posterior distributions to quantify the evidence that treatment confers benefit of some magnitude (78). Bayesian approaches have been recommended by the FDA as a means to improve methodological efficiency and inform decision making, even in the context of relatively small sample sizes (79, 80). In the current trial, Bayesian models will be evaluated via posterior probability (PP) threshold guidelines in the literature (81, 82), suggesting that PP = 75% to 90% indicates moderate evidence, PP = 91% to 96% indicates strong evidence, and PP ≥ 97% indicates very strong to extreme evidence. Consistent with probability thresholds stipulated for go/no-go decisions in other medication trials (83–87), a posterior probability PP ≥ 75% that an effect exists (i.e. PP(θ> 0) ≥ 0.75) will be considered a minimum threshold of support for the effect of pioglitazone.

Prior to hypothesis testing, generalized linear modeling (GLM) will evaluate relationships between predictors, outcomes, and baseline sample characteristics. Any baseline characteristic that demonstrates a relationship with both the predictor and the outcome in a given model will meet criteria for being a potential confounder (88, 89). Subsequent models will be tested with and without adjustment for any identified potential confounding variables; if inferences are influenced by the inclusion of these variables, both models will be reported; otherwise, the model without adjustment will be reported. Other covariates may be considered intractable from the relationships under investigation, including cocaine severity, sex, age, and other substance use. These variables will be included by default in each model as essential covariates.

Issues of multiplicity differ for Bayesian as opposed to Frequentist methods because the former observes the “Likelihood Principle”(90, 91). Fundamentally, we want estimates from the current study to replicate in future studies. The key lies in evaluating the data in the context of a family of priors (92). More specifically, priors are essentially assumptions and the robustness of the results to these assumptions may be evaluated by specifying optimistic, neutral, or pessimistic levels of and informativeness for the prior distribution (91). While the proposed priors are neutral and extremely vague (i.e. coefficients ~Normal(μ = 0, σ = 1000), evaluating the data in the context of more skeptical priors, as well as priors informed by our previous pilot study, permits evaluation of the robustness of the findings (12, 93). More skeptical priors tend to regularize the resulting parameter estimates to more conservative values and ostensibly increase the chance of replicating with the new but comparable sample of participants (94). We will report the results of such a sensitivity analysis if they vary from the analysis with the extremely vague priors. Please see Supplementary Materials for Frequentist estimates of power using an FDR correction for the evaluation of DTI measures, and maintaining α = 0.05 (type I error) for each of the cocaine craving measures.

Hypothesis testing for the effect of pioglitazone on targeted mechanisms (Aim 1) will use generalized linear mixed modeling (GLMM). The primary endpoint will measure change in WM integrity indexed by DTI metrics. Secondary endpoints will measure change in cognitive function (NIH Toolbox) and self-reported craving for cocaine (BSCS, VAS). Multilevel models will evaluate each mechanism (WM integrity, cognitive function, craving) as a function of time, treatment, and the interaction of time and treatment, with level 2 terms (i.e., random effects) to account for correlated data. If the interaction is supported (PP ≥ 75%) by the data, follow-up models will evaluate change over time within each treatment condition. Otherwise, models will drop the interaction term and evaluate the unique fixed effects of time and treatment condition.

Hypotheses for Aim 2 involve analysis of the association between improvement on targeted mechanisms and clinical outcomes. The primary abstinence endpoint will be continuous cocaine-negative UDS in the final three weeks of treatment. Secondary endpoints will include total percentage of days abstinent, total percentage of cocaine-negative UDS, and relapse, defined as > 3 consecutive cocaine-positive UDS during treatment. The endpoint for change in functional health status will be measured by the PROMIS. Multilevel analyses for Aim 1 hypothesis testing will permit the output of random effects for each participant in the analysis. These random effects (e.g., intercepts and slopes) will then serve as predictors in a Cox proportional hazards model. Initially, to improve interpretability, each participant’s slope will be coded as indicating improvement (slope < 0) or non-improvement (slope ≥ 0) on the targeted mechanism. Having adjusted for each participant’s intercept (baseline), Cox proportional hazards regression will evaluate each outcome measure (e.g., continuous abstinence, PROMIS) as a function of (non)/improvement. Subsequent analyses will evaluate slopes as continuous predictors to identify any potential non-linearities in the relation between improvement and outcome.

Power/precision estimates are based on effect sizes and intra-class correlations from our pilot trial of pioglitazone on WM integrity, our primary outcome (12). Estimates are conservatively based on N=48 “completers”, i.e., the minimum expected sample size. Assuming N = 48 completers and a posterior probability threshold of 0.75 for the existence of an effect (PP (θ > 0) ≥ 0.75), K = 10,000 Monte Carlo simulations indicate that the current design will detect improvements in WM integrity for the anterior and posterior thalami as well as genu and splenium 95% - >99% of the time. Supplementary information provides power/precision estimates for all WM areas measured, which are consistent with our pilot study (12). While these power estimates appear to be quite high, we should note that power calculations generally assume that hypothesized effect sizes are estimated without error which likely results in overly optimistic power estimates (95, 96). [The SAS simulation source code for the trial design is provided in Supplementary Materials].

4. Conclusion

Over one million American adults suffer from CUD with recent trends showing an increase in cocaine-related deaths since 2010. For people with chronic cocaine use, significant changes in brain function and structure set the stage for relapse that, unfortunately, continues to be the most common outcome following treatment. CUD currently lacks efficacious medication treatment options. Behavioral therapies, such as CBT, are a mainstay with relatively modest treatment effects for preventing relapse. A recommended way forward is through the identification of medications that facilitate or optimize response to behavioral interventions (4). We base the current trial on converging lines of evidence suggesting that WM integrity supports neural and cognitive functions that influence a patient’s ability to respond well to CBT. Using an inpatient-to-outpatient RCT design, results are expected to confirm that pioglitazone is an effective cognitive enhancing treatment to use in combination with CBT for preventing relapse in patients with CUD.