A Randomized Controlled Trial of Repeated Ketamine Administration for Chronic Posttraumatic Stress Disorder

This was a randomized double-blind parallel-arm controlled study conducted between June 2015 and January 2020 at the ISMMS Depression and Anxiety Center for Discovery and Treatment. The study involved three phases: 1) screening, during which participants were assessed for eligibility and any required medication tapering took place; 2) a 2-week treatment phase, during which participants received six infusions of study drug approximately three times per week and were assessed on each infusion day, as well as 24 hours and 2 weeks after the first infusion; and 3) a follow-up phase for participants who showed treatment response, defined as ≥30% symptom improvement 2 weeks after the first infusion (19–21), assessed with the Clinician-Administered PTSD Scale for DSM-5 (CAPS-5) (22), using a past-week recall period (see also Figure S1 in the online supplement). Treatment responders were followed naturalistically, weekly for up to 4 weeks and then monthly until loss of responder status or up to 6 months if there was no loss of response. To assess for eligibility, a trained study rater administered the Structured Clinical Interview for DSM-5 (SCID-5) (23) and the CAPS-5. Additionally, the principal investigator, a psychiatrist, reviewed each patient’s ratings, medical history, and concomitant medications, conducted a clinical evaluation with each patient, and administered the Columbia-Suicide Severity Rating Scale (C-SSRS) (24) and the Clinical Global Impressions severity scale (CGI-S) (25). All patients underwent a physical examination, basic laboratory screening, including routine hematologic, biochemical, and urine toxicology testing, and an ECG to rule out unstable medical conditions and current substance use other than marijuana.

Participants were eligible if they were between ages 18 and 70, had a current primary diagnosis of chronic PTSD as defined by DSM-5 criteria (of at least 3 months’ duration) and determined by the SCID-5, and had a score ≥30 on the CAPS-5, which is several points into the moderate severity range. Exclusion criteria were any serious unstable medical condition, active suicidal or homicidal ideation (having a plan or inability to contract a no-harm agreement), lifetime history of psychotic or bipolar disorder, current anorexia or bulimia nervosa, alcohol or substance use disorder within 3 months of screening, history of recreational ketamine or phencyclidine use on more than one occasion or any use in the past 2 years, and current treatment with a long-acting benzodiazepine or opioid medication. Stable dosages of other psychotropic medications (e.g., antidepressants, antipsychotics, or mood stabilizers) for at least 3 months before randomization were permitted. For patients taking a short- acting benzodiazepine, morning doses were held on infusion days. Although individuals with marijuana use disorder were excluded, current use of marijuana or cannabis derivatives was permitted (Table 1), given its prevalence among patients with PTSD (for further details on inclusion and exclusion criteria, see ClinicalTrials.gov NCT02397889). Participants who were taking a contraindicated psychotropic medication that was not effective or medically necessary were given the opportunity to taper off the medication before randomization, under the supervision of their treating psychiatrist.

A randomization schedule was generated by the ISMMS research pharmacy, employing randomly permuted blocks, to assign participants to receive six infusions of either ketamine hydrochloride (0.5 mg/kg) or midazolam (0.045 mg/kg). Eligible participants were randomly assigned on the morning of their first infusion. All participants and study personnel were blinded to treatment assignment, which was known only to the research pharmacy. Trained raters administered safety ratings during infusions and 40 and 120 minutes after infusions. A different set of trained raters, blinded to the ratings conducted during and after infusions on infusion days, administered the CAPS-5 and the Montgomery-Åsberg Depression Rating Scale (MADRS) at baseline, at the 1-week and 2-week assessments, and during the follow-up assessments. These same raters also administered the MADRS 24 hours after the first infusion.

Both study drugs (ketamine and midazolam) were administered intravenously using the same procedure at each infusion day. Participants were scheduled to receive a total of six infusions approximately three times per week over 2 consecutive weeks. On the morning of each infusion, participants were admitted to an outpatient research unit. After confirming that participants were nil per os, tested negative on point-of-care urine pregnancy (in females of child-bearing potential) and on toxicology tests with the exception of marijuana, and met the cutoff for baseline blood pressure (no more than two readings above 140/90 mmHg or one above 160/90 mmHg), an indwelling catheter was placed in the antecubital vein of the nondominant arm. The study drug was administered over 40 minutes via an infusion pump. Continuous ECG monitoring was conducted from infusion start to end. Pulse, blood pressure, and oxygen saturation were monitored every 5 minutes during infusion and every 15 minutes postinfusion as appropriate (for further details, see reference 14). A study physician (anesthesiologist or psychiatrist trained in sedation procedures) oversaw all infusions. A nurse monitored the patient during each infusion and up to 120 minutes after the start of the infusion. A study physician assessed all participants before their discharge from the research unit.

The primary outcome was change in PTSD symptom severity, compared between the two groups, as measured with the CAPS-5, administered at baseline just before the first infusion, at 1 week (just before the fourth infusion), and at 2 weeks (after completion of all six infusions, usually on the Monday after the second week of infusions), with a past-week recall period. The CAPS-5 with past-week recall was additionally administered at follow-up visits. Secondary outcome measures were the CAPS-5’s four PTSD symptom clusters and the MADRS with past-week recall (26), administered at the same time points; the Clinical Global Impressions severity (CGI-S) and improvement (CGI-I) scales, administered by a study clinician at each time point to evaluate overall clinical status at that moment; and the Impact of Event Scale–Revised (IES-R, self-report) (27) and the MADRS, both modified for past 24-hour recall, administered at baseline and 24 hours after the first infusion.

On infusion days, potential psychotomimetic, dissociative, and manic symptoms were assessed before infusion and at 40 and 120 minutes after start of infusion with the Clinician‐Administered Dissociative States Scale (28), the 4-item positive symptom subscale of the Brief Psychiatric Rating Scale (29), and the first item (elevated mood) of the Young Mania Rating Scale (30). The Patient-Rated Inventory of Side Effects (PRISE) was used to assess side effects in eight organ systems and general side effects. Acute infusion-related side effects were assessed with the PRISE 120 minutes after start of infusion, covering the period since start of infusion. Side effects were also assessed with the PRISE before each infusion, covering the period since the previous assessment. Additionally, the C-SSRS was administered, with past-week recall.

A priori power calculations were conducted on the basis of results from our previously published single-dose ketamine infusion randomized controlled trial (18). In that study, which employed a crossover design, a mean decrease of 27 points (SD=20) on the CAPS (DSM-IV) total score was observed in the ketamine arm at 7 days, compared with a mean decrease of 13 points (SD=18) in the midazolam arm (d=0.74). Thus, assuming a paired two-sample t test, we calculated that a sample size of 40 would provide >80% power to detect a between-group difference of this magnitude, assuming a type I error rate of 0.025 (using the Bonferroni adjustment to hold the family-wise error rate at 0.05).

Although we had originally planned for a total sample size of 40 participants, partially unblinded interim analyses (drug A, drug B), also planned, were conducted by an independent statistician with 30 participants. For this interim analysis, an intent-to-treat mixed-effects model approach was used to compare the primary outcome measure (CAPS-5) between the two treatment groups at baseline and at 1 week and 2 weeks. From this model, adjusted means were calculated at the three time points, and two-sample t tests were used to compare the two treatment groups at each of these time points. The robust, large-magnitude, and clinically significant difference (d=1.13; 66.7% compared with 20.0% treatment responders) between the two groups led to our decision to stop the trial with 30 participants, in consultation with the study data safety monitoring board.

For all outcome measures, we utilized intent-to-treat mixed-effects model analyses with time, treatment, and their interaction, assuming compound symmetry for the correlation matrix across time points. For the CAPS-5 total, CAPS-5 subscale, and MADRS total scores with past-week recall and for the CGI-S score, the time points were baseline, 1 week, and 2 weeks. Time points for the CGI-I score were 1 and 2 weeks, and for the IES-R and MADRS with past-24-hour recall, they were baseline and 24 hours after the first infusion. A chi-square test was conducted to compare percentages of responders in the two treatment groups. We produced Kaplan-Meier plots for probability of loss of response by treatment, with right-censoring for those who did not lose response. All analyses were conducted at an alpha of 0.0304 per the Pocock alpha spending function and in SAS, version 9.4, using linear mixed models (PROC MIXED) and cumulative logit mixed models (PROC GLIMMIX) for ana- lyses of CGI-S and CGI-I scores.