A Single Ketamine Infusion Combined With Motivational Enhancement Therapy for Alcohol Use Disorder: A Randomized Midazolam-Controlled Pilot Trial

During screening, participants were assessed with the Mini-International Neuropsychiatric Interview for DSM-IV and underwent medical and psychiatric evaluation, including serum collection (electrolytes, CBC, and liver function tests) and other diagnostic tests, such as ECG and vital sign assessment. Study applicants were considered eligible if they were <70 years old, had no medical illness or psychiatric comorbidity, and met DSM-IV criteria for alcohol dependence and minimum daily (at least four heavy drinking days over the past 7 days) or weekly (35 drinks per week for men and 28 drinks per week for women) use while not using other substances. Alcohol and other drug use were determined by self-report and urine toxicology. Individuals with a history of severe withdrawal symptoms (e.g., seizures, cardiac instability, and delirium) were excluded, as were those with a history of psychotic or dissociative symptoms and with current depressive symptoms indicative of a DSM-IV disorder. The study was approved by the institutional review board at New York State Psychiatric Institute, and written informed consent was obtained from all participants.

All procedures and outpatient visits took place at New York State Psychiatric Institute on the Columbia University Medical Center campus. All staff involved in this trial were blinded to treatment condition.

Participants were engaged in motivational enhancement therapy delivered by staff trained in the manual used in Project MATCH. Motivational enhancement therapy is a behavioral treatment for various substance use disorders (11) that involves a variety of strategies to promote motivation and self-directed change (12). In previous trials (13, 14), this therapy demonstrated only modest efficacy; the authors therefore considered it unlikely to obscure medication efficacy.

Six motivational enhancement therapy sessions were provided over 5 weeks. In week 1, participants engaged in an initial session, during which goals were explored and motivational statements elicited. Weekly subsequent sessions during weeks 2–5 were provided to achieve these goals. An additional session was provided during week 2, 24 hours after infusion, in order to capitalize on the hypothesized motivation-enhancing effects of ketamine, which may be most pronounced within the 24- to 48-hour postinfusion period. Sessions were audiotaped and supervised by a psychiatrist (E.D.) for fidelity to the manual.

Leading up to the infusion, participants were counseled to reduce the number of drinks per day in preparation for abstinence initiation and infusion administration during week 2. Participants came to the clinic twice weekly for motivational enhancement therapy and to meet with a psychiatrist, with visits spaced by 3–4 days, except during week 2, which involved 3 consecutive days. The timeline followback assessment was administered at each visit to quantify the number of drinks for each calendar day since the last visit. A positive urine ethyl glucuronide test among individuals who reported abstinence for this period led to the days being marked as drinking days. Measures related to alcohol-related vulnerabilities, such as craving and arousal (assessed with the visual analogue scale), withdrawal (assessed with the Clinical Institute Withdrawal Assessment), self-efficacy (assessed with the Alcohol Abstinence Self-Efficacy Scale and the Drug-Taking Confidence Questionnaire) (15, 16), perceived stress (assessed with the modified Perceived Stress Scale) (17), mindfulness (assessed with the Five Facet Mindfulness Questionnaire) (18), and impulsivity (assessed with the Barrett Impulsiveness Scale) (19), were collected at each visit, along with urine samples for toxicology (six-panel dipstick, ethyl glucuronide). Participants were provided with referrals at the end of the trial. Telephone follow-up was conducted 6 months after the trial.

Participants were compensated with $10 on screening and appointment days to defray the costs of travel, as well as $25 for the screening itself.

Participants were asked to abstain from alcohol for ≥24 hours before the infusion, as well as to fast after midnight. They were informed that they might receive any of several medications, in addition to ketamine or midazolam. This blinding procedure was intended to disguise what drug was specifically given so as to minimize expectancy effects.

Participants were randomly assigned (1:1) by a statistician using randomly sized blocks to receive an intravenous infusion of ketamine or midazolam. Medication assignment and preparation were performed at the New York State Psychiatric Institute pharmacy.

Active control (a 2-minute saline bolus followed by a 50-minute slow-drip intravenous infusion of midazolam, 0.025 mg/kg) or ketamine hydrochloride (a 2-minute 0.11-mg/kg bolus in saline followed by a 50-minute slow-drip intravenous infusion of 0.6 mg/kg) was administered between 10:00 a.m. and 12:00 p.m. on the second appointment day of week 2. This dose of ketamine was selected because it was the highest dose tolerated by participants in preliminary studies (8, 9) and may be more efficacious than lower doses for alcohol dependence given evidence that chronic alcohol use blunts glutamate receptor sensitivity (6). A bolus was provided before the infusion in order to obtain a potent subanesthetic serum level of ketamine at the start of medication administration. Midazolam was chosen as the active control because it alters consciousness without any known persistent (>8 hours) effect on alcohol dependence. Blood pressure, heart rate, and blood oxygen saturation were continuously monitored. Medical coverage was provided for up to 3 hours postinfusion, and a brief psychiatric evaluation was conducted before discharge.

As in our previous studies (8–10), we provided relaxation and mindfulness-based exercises before and during infusions. After the infusion, participants completed a subjective-effects assessment battery.

The study participants’ baseline demographic characteristics are summarized in Table 1.

The primary outcome of self-reported drinking days by timeline followback after the infusion was dichotomized so that each day was defined as abstinent (0 drinks) or nonabstinent (at least one drink). Days with missing data were treated as nonabstinent days. A longitudinal logistic mixed-effects model with a logit link and a random intercept was used to account for between-subject variances. The fixed effect of time (days postinfusion), treatment, and time-by-treatment interaction, adjusted by baseline total drinks, was used to analyze the longitudinal primary outcome: abstinent days during the 21 days postinfusion. Additionally, because the observed data do not follow a linear trend, time was also tested as a quadratic effect by testing whether the effect of time squared was significant. The Akaike information criterion and Bayesian information criterion summary fit statistics were used to identify which model best fit the observed data.

Reduction in heavy drinking days and other secondary measures were analyzed with longitudinal mixed-effects models, with the effect of study week, treatment, and study week-by-treatment two-way interaction adjusted by the corresponding outcome measures at baseline. A random intercept was used to account for between-subject variances, and a generalized estimating equation structure was included to account for within-subject correlations over time as an autoregressive (AR[1]) process.

We used the logit link to model binary outcomes (i.e., heavy drinking day) and the log link function to model outcomes (i.e., craving, arousal, and withdrawal) with right-skewed distributions. All other secondary outcomes were approximately normally distributed and modeled with identity-link functions.

The secondary outcome, time to relapse, was defined as time to the first heavy drinking day or time to dropout, whichever came first, and was analyzed by using Kaplan-Meier survival curves and the log-rank test. Time to first alcohol use and time to first heavy drinking day were similarly analyzed.

All analyses were performed in SAS, version 9.4, with all tests two-sided at a 5% level of significance.

Ninety-five individuals were screened for participation in the study, 50 were enrolled, and 40 were randomly assigned in an intent-to-treat analysis (Figure 1). Participants were heavy drinkers, with minimal psychiatric comorbidity (Table 1). Infusions were well tolerated, with the most common adverse effects being sedation (midazolam, N=12; ketamine, N=8) and headache (midazolam, N=4; ketamine, N=6), persisting about 12 hours postinfusion. Scores on the Clinician-Administered Dissociative States Scale were significantly higher in the ketamine group immediately following infusions (median=19, interquartile range, 9–30.75) compared with the midazolam group (median=2, interquartile range, 0.25–9.25; χ²=7.87, p=0.005). Two participants in the ketamine group experienced mild agitation lasting 1 hour postinfusion. There were no incidents of persistent psychoactive effects or initiation of drug misuse (benzodiazepines, opioids, or ketamine) in either study group.

Across the 21 days after infusion, 47.1% (N=8/17) of participants in the ketamine group used alcohol, and 17.6% (N=3/17) had a heavy drinking day. Among participants in the midazolam group, 59.1% (N=13/22) used alcohol, 40.9% (N=9/22) had a heavy drinking day, and 52.2% (N=12/23) relapsed. Throughout the entire study period, six participants dropped out of the midazolam group. Of these, one participant did not return after the infusion, and five dropped out after the first week. Four participants had heavy drinking days before dropping out, and one dropped out while still abstinent (as assessed during the participant’s final visit). No participants dropped out of the ketamine group.

The longitudinal logistic mixed-effects model for alcohol-abstinent days with time treated as a quadratic term yielded the best-fit model (Akaike information criterion=569.04, Bayesian information criterion=580.87) compared with the initial model with time treated as a linear term (Akaike information criterion=576.18, Bayesian information criterion=586.31). The quadratic effect of time was significant (F=8.21, df=1, 797, p=0.004), as well as the time-by-treatment interaction (F=25.1, df=1, 797, p<0.001), adjusted for total baseline drinks (F=1.09, df=1, 797, p=0.297). The observed and model-estimated proportions of alcohol abstinence across the 21 days postinfusion are shown in Figure 2. The modeled proportion of abstinence in the ketamine group remained stable while decreasing significantly in the control (midazolam) group. The observed number needed to treat was 6, and the modeled number needed to treat was 4. The score on the Clinician-Administered Dissociative States Scale was not significantly associated with end-of-study abstinence (odds ratio=1.11, p=0.068).

A sensitivity analysis without the assumption that missing days were nonabstinent days was performed for the primary outcome (abstinent days during the 21 days postinfusion). When treating missing days as missing, we found results similar to those reported above, suggesting that the assumption about missing days did not affect the obtained results. An analysis without adjustment for baseline total drinks also led to similar results.

Telephone interviews were conducted 6 months after the trial. Data were collected for 19 of the 40 participants (47.5%); 21 participants (52.5%) could not be reached by telephone. Of those participants who could be reached, eight were in the ketamine group and 11 in the midazolam group. Seventy-five percent of participants (N=6) in the ketamine group reported abstinence, compared with 27% (N=3) in the midazolam group.

When modeling heavy drinking days, the interaction between treatment group and time was significant (F=12.34, df=1, 798, p<0.001) (Figure 3). In the midazolam group, the probability of a heavy drinking day increased with each postinfusion day (odds ratio=1.19, 95% CI=1.14–1.25, p<0.001), while the odds of a heavy drinking day did not change significantly for the ketamine group (odds ratio=0.98, 95% CI=0.89–1.08, p=0.74).

When modeling the secondary outcomes (craving, withdrawal, mindfulness, impulsivity, stress sensitivity, and self-efficacy), the two-way interaction of study week-by-treatment group was not significant in any of the models.

On the basis of the log-rank test, participants in the ketamine group had significantly longer time to relapse (defined as the first heavy drinking day or dropout) (χ²=4.2, p=0.04), compared with participants in the midazolam group (Figure 4). However, there were no significant differences between groups in time to first use or to first heavy drinking day.

These results are preliminary, and little is known about the utility and safety of ketamine infusions in the treatment of substance use disorders outside of specialized research settings. Caution and restraint in clinical settings are warranted. Several other limitations underscore the preliminary nature of these findings. First, our sample was both small and homogeneous, with inadequate power for evaluating secondary outcomes and with minor discrepancies between treatment arms. The hazards of inferring treatment effects from small samples are well known (34). Second, the duration of treatment was short (5 weeks), with only 21 days of alcohol use monitoring by using the timeline followback method after infusions. Study follow-up occurred several months after the final study visit, and some participants could not be reached, resulting in a responder rate lower than that of other clinical trials involving more closely spaced follow-up visits. A longer follow-up with more frequent assessments would have allowed for better assessment of the duration of ketamine’s efficacy. Third, more than a quarter of participants in the control group dropped out. This is consistent with our hypothesis that ketamine promotes engagement with behavioral modification. Although the approach we used for missing observations is standard in clinical trials for substance use disorders, namely, to treat dropout as relapse, this assumption may not be entirely accurate, and with a small sample size, the findings could be sensitive to just a few unobserved violations of this assumption. Hence, missing data for the primary outcome were analyzed as both “using” and “missing,” with no difference in results. Fourth, fidelity to the motivational enhancement therapy manual was monitored through supervision and review of audiotaped sessions only. Lastly, the sample was selected for minimal psychiatric comorbidity, which limits generalizability.

The minor deception that we employed to further protect the blind and to minimize expectancy and placebo effects constitutes another limitation; informing individuals that they might receive any of a variety of medications would not pertain to clinical settings, and it is not clear whether there would have been additional risks stemming from participants knowing that they received ketamine. As in other clinical procedures involving substance abuse liability (e.g., provision of postoperative opioids), various safeguards were employed that work to minimize the risk of illicit use, such as careful patient selection, preparation, monitoring, and postprocedure support. With these safeguards in place, there has been no incidence of ketamine misuse, to our knowledge, in any of the research to date (although follow-up was short in these studies), suggesting that its propensity to be approached problematically can be effectively managed even when administered to substance users (8–10, 35).