Single, Repeated, and Maintenance Ketamine Infusions for Treatment-Resistant Depression: A Randomized Controlled Trial

Male and female outpatients with treatment-resistant depression (age range, 18–65) completed a single-center randomized controlled trial of ketamine at the Royal’s Institute of Mental Health Research in Ottawa between January 2013 and December 2017. Participants were recruited from physician referrals and advertisements. Participants met DSM-IV-TR criteria for major depressive disorder (18), single or recurrent episode without psychotic features, confirmed with the Mini-International Neuropsychiatric Interview (19). Treatment resistance was defined as failure to respond to at least two antidepressant medications of different pharmacological classes and two augmentation strategies at adequate dosages for at least 6 weeks each during the present episode, as recorded in the Antidepressant Treatment History Form (20). Inclusion criteria were a score ≥25 on the Montgomery-Åsberg Depression Rating Scale (MADRS) (21) at screening and at randomization, with no more than 20% improvement between these visits. Participants remained on stable dosages of concomitant psychotropic medication with treatment durations of at least 6 weeks with no changes to their treatment regimens during the trial. Exclusion criteria included a history of drug abuse or dependence as defined by DSM-IV-TR criteria (18) or confirmed by positive urine toxicology screen, a body mass index ≥35, a history of mania or hypomania, and unstable medical conditions identified by physical examination and measurement of vital signs, ECG, standard blood tests, and urinalysis. Female participants of childbearing potential required a negative pregnancy test at enrollment and use of adequate birth control throughout the study. The study protocol was approved by the Royal Ottawa Mental Health Centre Research Ethics Board. All participants provided written informed consent.

The data derive from a three-phase clinical trial (Figure 1). The purpose of phase 1 was to test the efficacy of ketamine compared with an active control. In phase 1, participants received a single ketamine infusion during a randomized double-blind crossover with midazolam (a short-acting benzodiazepine that serves as an active placebo control for ketamine) (5). Phase 1 infusions occurred at least 7 days apart, and participants were required to have a return of 80% of their baseline MADRS score to proceed to the second phase 1 infusion and to begin phase 2. The goal of phase 2 was to test reinstatement of antidepressant response after relapse and to evaluate the efficacy of repeated infusions. In phase 2, participants received a course of six open-label ketamine infusions, administered thrice weekly over a 2-week period. The aim of phase 3 was to test maintenance of antidepressant effects when the frequency of infusions was reduced. In phase 3, participants who obtained an antidepressant response after repeated administration received once-weekly infusions for an additional 4 weeks. Antidepressant response was defined as a ≥50% decrease in MADRS total score from baseline (before the first infusion in phase 1) to the end of phase 2. Nonresponders exited the study after phase 2.

In phase 1, participants were randomly assigned, in a 1:1 ratio, to receive midazolam or ketamine first. Ketamine and midazolam doses were prepared by a nurse in correctly labeled separate bags. To maintain the study blind, medication bags were then relabeled as drug A and drug B by an independent randomizer in accordance with a randomization log. Blinded study staff were subsequently provided with the correctly coded randomly assigned bag for infusion, and the unused bag (labeled A or B) was sent to the pharmacy to be destroyed. All study personnel and participants remained blind to phase 1 data until study completion.

Ketamine hydrochloride was administered at a dose of 0.5 mg/kg, diluted in 0.9% saline, over a 40-minute period by intravenous pump. The dose of 0.5 mg/kg was selected because this dose has been used in almost all previous clinical trials of ketamine for depression. Midazolam was administered at a dose of 30 µg/kg, to obtain an approximate dose of 2 mg, diluted in saline. The midazolam dose was selected on the advice of a consultant anesthesiologist. Infusions were administered in an outpatient clinic by a study physician and a research nurse under cardiorespiratory monitoring. Blood pressure and pulse were monitored at 5-minute intervals throughout each infusion and afterward until return of preinfusion levels. Participants were discharged approximately 2 hours postinfusion. On infusion days, participants were required to abstain from consuming grapefruit juice (a potent 3A4 cytochrome inhibitor that may slow the elimination of midazolam and possibly ketamine) (22) and to avoid taking any benzodiazepines the preceding day, as they attenuate ketamine response (23).

Measures included the 10-item clinician-administered MADRS and the 16-item patient-administered Quick Inventory of Depressive Symptomatology–Self Report (QIDS-SR) (24). For phase 1, the primary outcome measure was change in MADRS total score from preinfusion to 24 hours postinfusion. Secondary outcome measures included MADRS scores at additional postinfusion time points (2 hours and 7 days) and QIDS-SR scores (assessed preinfusion and 24 hours, 4 days, and 7 days postinfusion). For phase 2, the primary outcome measure was change in MADRS score throughout the course of repeated infusions. Additional study measures included the proportion of individuals who met antidepressant response and remission criteria (MADRS total score ≤10) after single and repeated infusions. Change in MADRS score over phase 3 was also examined. Follow-up measures for phases 2 and 3 were obtained 3 days after the final infusion in each phase. Ratings were conducted by study physicians who trained together to establish interrater reliability.

The safety and tolerability of ketamine were assessed at regular intervals. General adverse events were recorded with the Systematic Assessment for Treatment Emergent Events (25). Dissociative effects were assessed in a subgroup of participants with the Clinician-Administered Dissociative States Scale (CADSS) (26). The CADSS was administered during both phase 1 infusions and during the final ketamine infusion in phase 2 at three time points: preinfusion, immediate postinfusion, and 2 hours postinfusion.

The baseline characteristics of responders and nonresponders were compared with independent-samples t tests for continuous variables and chi-square tests for categorical variables. For phase 1, a random-effects linear mixed model was used to evaluate changes in MADRS score between ketamine and midazolam infusions across four time points from preinfusion to 7 days postinfusion. The model included participants as a random effect, drug and time as fixed within-subject factors, interaction between drug and time, and a fixed intercept. Baseline MADRS score and order of drug administration were included as covariates with fixed within-subject and between-subject effects, respectively. An unstructured variance-covariance matrix best fit the data using Akaike’s information criterion. Restricted maximum likelihood estimation was used. Significant effects were examined with simple-effects tests. For phases 2 and 3, random-effects models were used to quantify changes in MADRS score over time, adjusting for phase-specific baseline MADRS score. Analyses of QIDS-SR data were performed as described above for MADRS data. CADSS scores were analyzed with paired t tests and Pearson’s correlation coefficient. Statistical analyses were performed with IBM SPSS Statistics, version 24.0 (IBM, Armonk, N.Y.). Results were considered significant at a two-tailed p value of 0.05.

Sixty-three individuals were prescreened for eligibility through consultation with a study physician. Of these, 46 signed consent forms and underwent formal screening (see the CONSORT diagram in Figure SF1 in the online supplement). There were three screen failures. Forty-three participants met criteria for the intent-to-treat sample and received at least one infusion. Four participants withdrew during the course of the study. Participants’ baseline characteristics are summarized in Table 1.

Forty-one participants completed the randomized double-blind crossover comparison of single infusions of ketamine and midazolam. Participants received the two infusions an average of 10 days apart (SD=6, range=7–36 days). There was no difference in the elapsed time between phase 1 infusions for participants who received ketamine first compared with those who received midazolam first (t=0.25, df=39, p=0.80). Using random-effects modeling, after adjustment for baseline MADRS score and order of drug administration, there were significant main effects for drug (F=8.84, df=1, 40, p<0.001) and time (F=30.77, df=3, 40, p<0.001) and a significant drug-by-time interaction (F=13.15, df=3, 40, p<0.001). Simple-effects analyses revealed that participants had significantly lower MADRS total scores at each postinfusion time point after ketamine infusion compared with midazolam infusion (Figure 2A). At the a priori primary efficacy endpoint 24 hours after kettamine infusion, participants had a mean decrease of 10.9 points (SD=8.9) in MADRS total score relative to preinfusion scores compared with a mean decrease of 2.8 points (SD=3.6) with midazolam.

Examination of the model covariates revealed significant main effects for baseline MADRS total score (F=359.95, df=1, 37, p<0.001) and order of drug administration (F=4.24, df=1, 37 p=0.047), indicating carryover effects between phase 1 infusions. Although participants who received midazolam first had similar preinfusion MADRS total scores for both phase 1 infusions (mean, 35.4 [SD=5.6] compared with 34.9 [SD=4.5]; t=0.36, df=20, p=0.72), those who received ketamine first had slightly lower preinfusion scores at their second phase 1 infusion (mean, 34.6 [SD=4.1] compared with 32.6 [SD=5.6]; t=3.44, df=20, p=0.006).

Twenty-four hours after the single ketamine infusion, 11 participants (27%) met antidepressant response criteria, and two participants (5%) achieved remission. Single-infusion responders had a mean decrease of 22.3 points (SD=5.3) in MADRS total score (Figure 2B) 24 hours postinfusion, and nonresponders had a mean decrease of 6.7 points (SD=5.6) (Figure 2C). No participants met antidepressant response criteria with midazolam at any postinfusion time point.

The primary outcome for phase 2 was change in MADRS scores over the course of six repeated infusions. A random-effects model that was adjusted for participant and depression severity at the start of phase 2 revealed a significant main effect for time (F=11.16, df=6, 39, p<0.001) (Figure 3A). On average, participants’ MADRS total score decreased by 2 points with each infusion.

Of the 41 participants treated in phase 2, 39 completed the full course of infusions. At the post–phase 2 follow-up visit, 23 participants (59%) met antidepressant response criteria, and nine (23%) achieved remission. Responders had a mean decrease of 21.6 points (SD=5.8) in MADRS total score overall, and nonresponders had a mean decrease of 3.1 points (SD=5.7) (Figure 3B). Responders included nine (82%) of the 11 participants who met response criteria after the single phase 1 ketamine infusion, and 14 additional phase 1 nonresponders. The median number of infusions participants needed to first meet response criteria was three. Seventy-seven percent of phase 2 responders received three or more infusions before meeting response criteria (Figure 3C).

Participants who had at least a 50% improvement in MADRS scores after repeated infusions (N=23) continued to phase 3, the maintenance phase. A linear mixed model that was adjusted for the random effect of participant and phase-specific baseline depression severity revealed no main effect of time on change in MADRS total score during maintenance infusions (F=0.88, df=4, 22, p=0.49). This indicates no further change in MADRS score once ketamine infusions were reduced to once-weekly administration (Figure 4). Examination of individual-level data revealed that 21 responders (91%) met antidepressant response criteria throughout maintenance infusions.

Linear mixed models were used to evaluate participants’ self-reported change in depression severity by using QIDS-SR total scores. For phase 1, there were significant main effects for drug (F=6.60, df=1, 40, p=0.01) and time (F=6.76, df=3, 36, p=0.001) and a significant drug-by-time interaction (F=3.46, df=3, 37, p=0.03). Simple-effects analyses revealed that participants had significantly lower QIDS-SR total scores with ketamine compared with midazolam at 24 hours and 4 days postinfusion (see Figure SF2 in the online supplement). Consistent with the analyses of clinician-administered MADRS scores, there was a significant fixed effect of time on QIDS-SR total score in phase 2 (F=7.67, df=6, 39, p<0.001) (see Figure SF3 in the online supplement) and no effect for time in phase 3 (F=1.29, df=4, 22, p=0.30) (see Figure SF4 in the online supplement).

No serious adverse events were reported during the trial. The most common side effects associated with ketamine were cardiorespiratory effects, numbness or tingling, dissociation, dizziness, and visual disturbances. Vital signs were continuously evaluated at each infusion throughout the study. Mean ketamine doses and overall changes to cardiorespiratory values were calculated for participants’ first ketamine infusion. The mean absolute dose of ketamine administered was 40 mg (range, 26–60 mg). During ketamine infusions, participants experienced transient elevation of blood pressure (maximum mean change, systolic, 25.3 mmHg; diastolic, 15.7 mmHg) and heart rate (maximum mean change, 10.2 bpm). On average, values returned to preinfusion levels at 24 minutes postinfusion (range, 5–40 minutes). No infusions were discontinued as a result of cardiorespiratory effects, and no rescue medications were administered during the study.

The dissociative effects of ketamine were formally assessed in a subgroup of study participants (N=22). A paired t test comparing participants’ preinfusion change with their immediate postinfusion change in CADSS total score during their phase 1 infusions revealed significantly higher CADSS scores with ketamine compared with midazolam (t=5.70, df=21, p<0.001). The dissociative effects associated with ketamine returned to baseline levels by 2 hours postinfusion (t=1.70, df=21, p=0.10). In phase 1, participants’ dissociative side effects (change in CADSS score) during the ketamine infusion were significantly correlated with antidepressant response at 24 hours postinfusion (change in MADRS total score, Pearson’s r=–0.46, p=0.03) (see Figure SF5 in the online supplement). Comparison of change in CADSS scores before and after participants’ first ketamine infusion (mean=18.0, SD=14.3) and their final ketamine infusion in phase 2 (7th infusion overall; mean=4.6, SD=7.3) revealed a decrease in dissociative side effects with repeated infusions (t=4.06, df=19, p=0.001).

Although drug craving was not formally assessed during the study, there were no spontaneous reports of craving or drug-seeking behavior among the participants during the trial nor in any participants seen since in follow-up.