Subjects
Enrollees were recruited by advertisements and clinical referrals from 25 sites in six countries (Bulgaria, Croatia, Finland, Latvia, Poland, and Sweden). Enrollees were diagnosed by trained clinicians as having AUD according to the Diagnostic and Statistical Manual of Mental Disorders, Revised Fifth edition71. .
Inclusion criteria were as follows: aged ≥ 18 years; reported current heavy alcohol consumption (≥ 60 g alcohol/day and ≥ 40 g alcohol/day for men and women, respectively) on six or more days in the four-week period preceding the Screening visit, and were currently drinking above the WHO International guide for monitoring alcohol consumption and related harm (Johnson BA, Addolorato G, Lesch O, Liu L, Rodd ZA. A critical scientific evaluation of a purportedly negative data report – response to Seneviratne et al. 2022 [manuscript submitted for publication]) - defined as drinking an average of > 40 g of ethanol/day for males and > 20 g of ethanol/day for females for the 4 weeks prior to the Screening visit; willing to provide a blood sample for DNA analysis, and had selected genotypes:
-
LL genotype of the insertion-deletion polymorphism in the 5’regulatory region plus rs1042173-TT SNP in the 3′untranslated region of the SLC6A4 gene that encodes the serotonin transporter;
-
and/or one, two, or three of the SNPs on the genes for the 5-HT3 receptor subunits: rs1150226-AG or rs1176713-GG in the gene that encodes the 5-HT3A receptor subunit, and rs17614942-AC in the gene that encodes the 5-HT3B receptor subunit.
Finally, subjects had to have a negative screen for narcotics, amphetamines, and sedative hypnotics at enrollment, and had signed a written informed consent form to participate. Abstinence was not a study outcome criterion; however, all subjects reported a desire to stop or reduce their alcohol consumption, and to participate in the BBCET (Copyright PEPCO LLC, 2009) to enhance compliance and pill-taking66.
Exclusion criteria were as follows: a current substance use disorder diagnosis other than alcohol or nicotine use disorder; alcohol withdrawal symptoms necessitating inpatient detoxification; clinically significant somatic abnormalities (i.e., on physical examination), electrocardiography including a QTcF interval > 450 ms, malignant or hematologic disease, liver enzymes ≥ 3 times the upper limit of normal (except for gamma glutamyl transferase), elevated bilirubin, pregnancy, lactation, taking medications with a potential effect on alcohol consumption, and the receipt of treatment for their AUD within 28 days prior to enrollment.
Study personnel were trained to administer the rating scales TLFB and BBCET.
Ethics approval was provided by the committee for the protection of human subjects at each participating site. The trial was conducted in compliance with the protocol, the International Conference of Harmonization Good Clinical Practice, the applicable regulatory requirement(s) and the Declaration of Helsinki.
General Procedures
Trial enrollment occurred between February 2020 and August 2021. The clinical sites were predominantly affiliated with main hospital units, universities, or academic centers. The clinical sites were monitored and audited by Optimapharm d.o.o (https://optimapharm.eu/).
At Screening (Visit 1), after providing written informed consent, subjects were assessed on: (1) physical health status, including medical history and vital signs, 12-lead ECG, and a comprehensive set of hematologic and biochemical studies including a urine pregnancy test; (2) breath alcohol concentration (BAC); (3) urine drug and biochemical screens; (4) DSM-5R71 diagnosed AUD status (mild, moderate, or severe) and the number of AUD criteria as well as psychiatric diagnosis;(5) self-reported drinking over the past 28 days based on the TLFB72;(6) suicidality based on the C-SSRS73; (7) AE profile; (8) medication history; and (9) DNA analysis to identify the target genotypes i.e., rs4795541 (InDel) encoding the LL/rs1042173 and TT genotypes and/or one, two, or three of the SNPs on the genes for the 5-HT3 receptor subunits: rs1150226-AG or rs1176713-GG in the gene that encodes the 5-HT3A receptor subunit, and rs17614942-AC in the gene that encodes the 5-HT3B receptor subunit.
Eligible subjects were invited back for Visit 2 (baseline) approximately 28 days after Visit 1. Participants were re-assessed against the inclusion and exclusion criteria, and those evaluations also included physical status through the conduct of a physical examination and vital signs, BAC < 0.02%, medication history since Visit 1, AEs, urine pregnancy test, drug screen, TLFB, quality of life using the WHO Quality of Life BREF63, general well-being using the 9-item Patient Health Questionnaire (PHQ-964), and non-pathological 12-lead ECG without a prolonged QTc (< 450 ms), interval, which had been previously associated with the effects of intravenous but not oral ondansetron46,47.
Participants were then randomly assigned to receive double-blind treatment (i.e., ondansetron 0.33 mg (AD04) or placebo, morning and evening), and attended their first session of a structured brief psychosocial intervention to enhance compliance with the study and pill-taking (BBCET, see details below).
After Visit 2, subjects received 24 weeks of double-blind treatment (Weeks 1 to 24). At bi-weekly intervals, subjects were administered BBCET and the clinical team collected the TLFB, AEs, pill-taking frequency, concomitant medications, and AE profile. At monthly intervals between Weeks 1 and 24, subjects were assessed on vital signs, BAC, urine pregnancy test for women, PHQ-9, CSSRS, and pill-taking frequency. At eight-week intervals during the same treatment period, subjects were assessed for safety via ECG recordings. At Weeks 12 and 24, subjects were assessed for safety by physical examination, examination of hematologic and biochemical tests, urine drug screen; and the severity of AUD illness via the DSM-VR criteria; and quality of life via the WHO Quality of Life BREF. Finally, at Week 24, to establish safety at discharge, a physical examination was done. A 4-week safety follow-up period occurred after discontinuation of treatment at Week 24, and the study concluded at Week 28.
Notably, this study was done during the global COVID-19 pandemic during which all European countries placed stringent restrictions on the movement of people. To facilitate the successful completion of the study, a provision was made such that visits, excluding those conducted for safety, could be performed remotely.
This study was registered in www.clinicaltrials.gov (NCT04101227).
Supply, Dosing, Blinding, Compliance for Medication, and Randomization
AD04 and matching placebo were manufactured for Adial Pharmaceuticals by Tedor Pharma, Inc. AD04 dosing was based on the findings from the previous Phase 2 study, whereby 4 mcg/kg was approximated for an average body weight of about 80 kg to render a dose of about 0.33 mg. Both active medication and placebo were identical in size, weight, shape, and color.
The pill-taking frequency was calculated by subtracting the amount of medication returned from the amount dispensed. Medication was supplied by the contract research organization (CRO) responsible for packaging and distribution (Catalent) to the Investigative Sites. Catalent received the demographic and alcohol consumption data for the purposes of pre-stratified randomization. Participants were pre-stratified and randomized through an IBM clinical database managed by Optimapharm based on the demographic and alcohol consumption data. Participants were pre-stratified by two characteristics: gender and drinking level dichotomized into heavy drinkers (< 10 DDD) and very heavy drinkers (≥ 10 DDD) prior to initiating the Screening visit. Because a standard drink was defined as 10 g of ethanol, the distinction between heavy and very heavy drinkers equates to < 100 g or ≥ 100 g of ethanol per drinking day. The clinical relevance of this distinction is underscored in a defining manuscript where binge drinkers were shown to drink an average of eight standard US drinks (13 g/drink) per occasion55 (i.e., drinking day), which is equivalent to 10 standard European drinks. A comparative risk assessment analysis indicated that there are distinct differences between heavy drinkers and very heavy drinkers for alcohol-related injuries and mortality and for the overall disease burden of AUD56.
Combining the stratification of subjects by level of alcohol consumption and genotype results in specific essential factors. We have learned that studying for genetic markers within subgroups requires a study of ample sample size, a criterion if not fulfilled can lead to equivocal or possibly false-negative results51. We feel that this report is a verification of the need to have proper statistical power (sample size) to conduct accurate precision medicine clinical research.
As described earlier, the pre-stratification by alcohol consumption level was based on a previous study, whereby ondansetron’s effects were only seen in heavy drinkers24,38. The randomization strategy also was done to assure similar numbers of heavy drinkers and very heavy drinkers in each of the two medication groups; however, the assignment was not forced artificially to have equal numbers of either category in each group. Pursuant to pre-stratification, subjects were assigned to receive either AD04 or placebo using a permuted block design. Medication boxes containing the blinded assignment for enrolled subjects were delivered by the CRO to the pharmacy at each clinical site.
Brief Behavioral Compliance Enhancement Treatment
BBCET, a brief psychosocial intervention, was based on the principles of brief counseling66 used in generic practices in Europe but focused on using behavioral and motivational techniques to enhance medication taking and attendance at the participating clinical site. The BBCET manual for this study was translated into five non-English European languages (i.e., Bulgarian, Finnish, German, Polish, and Swedish), and all those who administered BBCET were trained by the neuropsychobiological group (i.e., Psychological Education Publishing Company, PEPCO) that was led by the primary author. BBCET interviews were done in the native language of the subject and BBCET administrator. Training included didactics, video interviews, and an electronic examination with a minimum pass rate of 80%. Trained administrators of BBCET were provided with certificates. The fidelity and consistency of BBCET administration was ensured by every interview between the BBCET administrator being taped. Approximately 15% of these taped interviews were reviewed by BBCET Supervisors at PEPCO, who provided performance feedback to the clinicians. Delivering BBCET in a structured format provided a stable and consistent platform against which to measure the effects of the pharmacological intervention. Previously, structured BBCET had been used in several clinical trials67,68 in Europe and the USA, with currently over 700 trained practitioners (PEPCO, LLC).
Data Monitoring Safety Committee
Study safety oversight was done by an independent Data Monitoring Committee (DMC) who reviewed unblinded data (approximately 3-monthly) to ensure adequate safety of study subjects. The DMC was tasked with ensuring that procedures for discontinuation from the study, unblinding during an emergency, the use of prohibited medication during the trial, and how to handle the use of other concomitant medications or serious relapse or alcohol withdrawal during the study that were defined in the clinical protocol were adhered to, and that any protocol deviations were reported and evaluated. The operational details were covered by the DMC Charter. Following each data review, the DMC recommended the study continue without modification.
STATISTICAL METHODS
General Methods
Data quality was supervised by Optimapharm. Optimapharm was also responsible for data cleaning and delivered the full and locked dataset to the statistical analysis group at PharPoint Research Inc. for safety and efficacy determination. Additional data and statistical verification were done by Professor Lei Liu and Dr. Joe Hirman (Pacific Northwest Statistical Consulting, Inc.).
Efficacy data were analyzed consistent with the intent-to-treat method, i.e., according to the treatment to which subjects were randomized.
Analysis of Safety and Tolerability Data
Demographic and baseline characteristics at the Screening visit for all subjects included in the full analysis set were described using summary statistics of means and standard deviations and presented by the baseline stratification factor into heavy drinkers and very heavy drinkers. Statistical analyses were done to confirm that the groups were balanced.
Serious adverse events, adverse events, attendance, study compliance, pill-taking frequency, and suicidality were collated by treatment assignment and the alcohol consumption stratification factor. Data were summarized as means and standard deviations and compared statistically to examine for group differences. Study completion occurred when only safety data and no drinking data were collected at week 28. All safety data collected have been included in the analysis.
Analysis of Efficacy Data
Efficacy was measured as the PHDD/Study Month. A heavy drinking day was defined as > 40 g and > 60 g of ethanol for women and men, respectively.
The PHDD was calculated per study month starting at the date of Day 1 through the last date of study treatment, resulting in a maximum of six measures per subject, from Study Month 1 to Study Month 6. The endpoints were calculated using the TLFB as captured in the records and converted to standard drinks. If the same date was recorded multiple times in the TLFB (e.g., due to overlap between visits), the first record was used in the calculation, as the first record would be subject to less recall bias. TLFB data prior to the last date of study treatment were used in the modeling. TLFB data captured after the subject terminated treatment but prior to discontinuing from the study were not used in the primary analysis.
The formula for the calculation of PHDD is PHDD = (total number of HDD in period of interest / total number of days in period of interest) * 100.
The primary efficacy variable was the mean change in PHDD from baseline to study Months 5 and 6 combined, using a Mixed Model for Repeated Measures (MMRM) with an unstructured covariance structure for repeated measures. Restricted maximum likelihood estimation and the Kenward-Roger approximation for degrees of freedom were employed. Covariates in the model included the stratification factors of gender and baseline DDD category (i.e., heavy drinkers vs. very heavy drinkers), together with baseline PHDD, study month (as a categorical variable), and a study month by treatment interaction.
The secondary efficacy variable, the mean change from baseline in PHDD to study Month 6 (i.e., study end) was also calculated. This was important because it has been commonly used to validate efficacy for regulatory agencies such as the European Medicines Agency.
Additional secondary efficacy variables included the change in mean AUD symptoms and clinical status using the DSM-5 and the categorical shift level in the WHO drinking risk levels of alcohol consumption. AUD symptoms and clinical status according to DSM-5 were measured at three time points: baseline, Week 12, and Week 24. At each time point, the severity of AUD was classified into four continuous levels: 0: No diagnosis (< 2 symptoms); 1: MILD: Presence of 2 or 3 symptoms; 2: MODERATE: Presence of 4 or 5 symptoms; 3: SEVERE: Presence of six or more symptoms. Next, the change in the DSM-5 scale at Week 12 (and Week 24) from baseline was considered as the outcome of interest in an MMRM with similar specification to PHDD. The baseline period was defined as the time encompassed by the TLFB taken at the Screening visit. Logistic models were used to describe the proportion of subjects with a significant categorical shift from baseline in the modified WHO Quality BREF (DRL) at study Month 6. There were six possible levels: very high risk, high risk, medium risk, low risk, very low risk, and abstinence. The downward shift was described by three outcome variables: at least one level (e.g., from “very high risk” at baseline to “high risk” at month 6); at least two levels (e.g., from “very high risk” at baseline to “medium risk” at month 6); and at least three levels (e.g., from “very high risk” at baseline to “low risk” at month 6), respectively.
Thus, our efficacy variables included a direct measure of self-reported drinking (i.e., mean change in PHDD) along with measures to assess the presence of an AUD diagnosis and clinical symptom severity (i.e., change in the number of AUD criteria) and psychosocial improvement (i.e., change in WHO Quality of Life BREF).
Missing data were accounted for in the following ways: for the primary endpoint of PHDD, study months with at least 7 days of TLFB data had PHDD calculated based upon the available data. If a study month was found to have less than 7 days of available TLFB data, the study month was considered missing. MMRM statistical analysis was used to analyze the available study month data, with the assumption that missing data were missing at random. For missing data related to the alcohol use symptoms and clinical status according to DSM-5 and for the DRL shift, MMRM also was used to account for missing data.
Incomplete dates were imputed. An incomplete date was any date for which either the day, month or year was unknown, but all three fields were not unknown. An incomplete date happened when the exact date of an event occurring or ending could not be obtained from a subject. For many of the analyses, a complete date was necessary to determine if the event should have been included in the analysis (i.e., if the event was treatment emergent) or to establish the duration of an event. In such cases, incomplete dates were imputed. To minimize bias, the project statistician imputed dates in a systematic and reasonable manner. If the month/year of the incomplete date was the same as the Day 1 month/year, then the date was set to the date of Day 1. In other cases, missing days were imputed as the day component of Day 1; missing months/years were imputed as the month/year of Day 1. For non-existent dates that occurred at the end of a month created by this imputation method, the first date of the next month was used (e.g., Day 1 = 31JAN2021, incomplete date = XXFEB2021, imputed date = 01MAR2021). If, however, another date precluded the possibility (e.g., an end date for an adverse event could not be prior to the start date, or medication date(s) linked to an AE could only occur after the start of an adverse event), the closest date to Day 1 that was possible for the incomplete date was used. If a date had an unknown day, month, and year, the event was assumed to have started/ended on the date of Day 1, unless another date precluded the possibility, in which case the closest date to Day 1 was to be used.
Additional planned analyses using other models to examine treatment effects of AD04 vs. placebo for the heavy drinking subgroup were also done and are presented as supplementary information.
Finally, subgroup analysis with pre-specified baseline DDD as the stratum to dichotomize heavy drinkers vs. very heavy drinkers was done for all the above models.