Effects of Low-Dose and Very Low-Dose Ketamine among Patients with Major Depression: a Systematic Review and Meta-Analysis

1Aside from electro-convulsive therapy (ECT), there are no widely used treatments that provide large or rapid benefits for patients suffering from severe depression. This presents a major clinical challenge, especially for patients who have not responded fully to existing therapy or if symptoms include suicidality. In 2000 a small crossover trial suggested that a single subanaesthetic dose of ketamine could provide a large antidepressant benefit, starting within a few hours and lasting for at least several days. Ketamine appeared to improve specific depressive symptoms such as sadness, suicidality, and helplessness, rather than induce a nonspecific mood-elevating effect. Since then, several more trials have been published, but existing reports and reviews do not easily allow a direct comparison of how treatment effects persist in the days following treatment, given differences in graphical and tabular reporting methods and outcome scales. Three very recently reported trials (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data) evaluated lower doses of ketamine than previously tested, providing an opportunity to compare treatment efficacy and tolerability of different doses. One other key issue is the extent of bias resulting from the use of a crossover design in all but one previous trial. Potential biases from crossover designs, such as carryover of treatment effect into the second treatment period, can be addressed by restricting analyses to the first treatment period but often requires unpublished data. We therefore conducted a systematic review of trials of ketamine in patients with treatment-resistant depression to evaluate antidepressant efficacy, suicidality, safety, and tolerability.

2We considered all relevant randomized trials (either crossover or parallel) in which ketamine was used specifically for the treatment of a Major Depressive Episode (DSM-IV diagnosis) and was compared with placebo, including active placebo. Eligible trials included participants with either unipolar or bipolar affective disorder. We excluded trials conducted in the context of ECT and surgery, but there were no restrictions on concomitant pharmacological or psychological treatments.

3We included any trial that attempted to evaluate a comparison between single administration of ketamine and placebo for the treatment of major depressive disorder. There was no restriction on ketamine regimen used (eg, dose or route).

4The primary endpoint was change in depression severity scores from baseline on depression scales such as the Hamilton Depression Rating Scale (HAM-D; ) and/or the Montgomery Åsberg Depression Rating Scale [MADRS]) at the following times: day 1 (24 hours after first dose), day 3, and day 7. Other outcomes were: clinical remission (defined as HAM-D <7 or MADRS <10); clinical response (defined as ≥50% reduction in depression severity score); suicidality measures, including the suicidality item of HAM-D or MADRS; safety and tolerability as assessed by reported adverse events and dropouts.

5Reports were restricted to English language publications. Publications for this review were identified by searching Medline, Embase, and PsycINFO databases before August 2014 using the following search terms as free text or subject headings as appropriate for each database: (“ketamine” OR “NMDA receptor antagonist”) AND (“depression” OR “major depressive disorder” OR “bipolar depression” OR “depressive disorder” OR “dysthymic disorder” OR “treatment resistant depression”) AND (“ randomized controlled trial” OR “controlled clinical trial” OR “randomized” OR “placebo”) (full details available in supplementary Table 1 in the online data supplement). If additional studies cited in these articles met with these criteria, then they were also included.

6Two authors (A.R. and Y.X.) reviewed the search strategy and selected studies for inclusion in the review. In case of disagreement, M.H. arbitrated.

7When relevant publications were selected, 2 authors (A.R. and Y.X.) independently extracted information on year of publication, geographic location, sample size, loss to follow-up, mean age, study design, major inclusion and exclusion criteria, concomitant treatments, and outcome measures as well as safety and tolerability.

8Potential sources of bias were identified for each trial using criteria recommended in the Cochrane Handbook. For random sequence generation, we categorized trials as “low risk” if random-number charts or coin tossing was used, as “unclear” if there were no details of sequence generation, and as “high risk” if sequence was generated by odd/even date of birth. For allocation concealment, we categorized trials as low risk if opaque or sealed envelopes were used, as unclear if there was no detailed information and as high risk if assignments could possibly be foreseen. For blinding, we categorized studies as low risk if blinding was ensured and unlikely to be broken, as unclear if an inactive placebo was used, and as high risk if outcome measurement was likely to be influenced by lack of blinding. For incomplete outcome data, we categorized trials as low risk if missing outcome data balanced in numbers and with similar reasons across intervention groups, as unclear if reports of drop- outs were insufficient, and as high risk if there were imbalances in numbers or reasons for missing data across intervention groups. For selective reporting, we categorized trials as low risk if there was a published protocol and the outcome measures listed in the protocol were reported, as unclear if protocol could not be found and as high risk if a described outcome measure in the protocol or methods was not reported in results. Assessment of period and carryover effects were listed under “other sources of bias.” We categorized crossover trials as high risk and parallel trials as low risk.

9Depression severity score (HAM-D and/or MADRS) at each follow-up time point was calculated for each trial, for ketamine or placebo groups, and a placebo-corrected value for each time point was calculated. If not available in tabular form, HAM-D and/or MADRS results at different time points were estimated independently from Figureures in the articles by 2 authors. Placebo-corrected HAM-D scores were plotted separately for each trial with that outcome measure, and placebo-corrected MADRS scores for all trials with that outcome measure. In addition, we plotted a percentage reduction in depression severity score for all trials based on MADRS score where available and HAM-D for other trials. We conducted a meta-analysis of mean MADRS score at day 1, 3, and 7, with day of administration defined as day 0. These analyses were conducted initially to include the data from the single parallel trial and the first period of all crossover trials. We also compared the relative risk (RR) during the first period against the second period in crossover trials in order to examine the direction and magnitude of carryover effects. Meta-analysis of continuous outcomes was conducted using Stata13 (www.stata.com), Windows 7. Meta-analyses of response rates and clinical remission at days 1, 3, and 7 were conducted using Review Manager Version 5.3. software to estimate RR and number needed to treat (NNT). All meta-analyses were done using a random effects model. The I2 statistic was used to assess heterogeneity of trial results, and subgroup heterogeneity was tested using a chi-squared statistic. Data were available only on day 4 for one trial, and this was included in the day 3 category. Data were sought from the crossover trials for each treatment period separately.

10The search results and study selection process are summarized in a PRISMA flowchart (Figure 1) showing the number of unique references identified by the search, the number of records excluded, and the number of full-text records retrieved. One placebo-controlled trial was not included because the sequence was not randomized and one randomized trial was not included because ketamine was compared with ECT. One unpublished trial of the effects of intra-operative ketamine on depression was identified, but data were not available (M. Bastos, personal communication).

11Nine randomized, placebo-controlled trials were identified, and a total of 201 patients (105 females and 96 males) were included (Table 1) (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data). Eight studies (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data) were crossover trials with a 1- to 2-week washout period between treatments, and one was a parallel group design with 1 week of follow-up. Two trials included only patients with bipolar disorder (BD), and 7 trials (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data) included patients with only unipolar depression (although one trial included one patient with BD). All patients were treatment resistant, variously defined (Table 1). Patients with history of psychosis or recent substance use disorders were generally excluded. Patients in the 2 BD trials were drug free except for lithium or valproate. In the other 7 trials, patients were drug free in 2, allowed to take only a nonbenzodiazepine hypnotic in one, and taking stable doses of other psychotropic medications in 4 trials (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data). Six trials assessed a single 40-minute infusion of i.v. ketamine at a subanaesthetic low dose of 0.5mg/kg (defined here as low dose). Three trials tested lower doses (defined here as very low-dose trials): 50mg intranasally (as 10mg every 5 minutes), estimated to achieve plasma concentrations equivalent to about 0.3mg/kg i.v.; 0.1 to 0.4mg/kg i.v., or 0.1 to 0.5mg/kg i.v., intramuscular, or s.c. in an ascending dose design (with placebo randomly inserted) (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data). Seven trials used 0.9% saline as placebo and 2 used 0.045mg/kg or 0.01mg/kg (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data) midazolam as an active placebo. Potential sources of bias are summarized in Figure 2.

12Effects on depression severity over time are presented in Figure 3, with data redrawn from original trials and placed on a uniform linear scale, and shown in Figure 4 with data from the single parallel trial and from the first period of crossover trials. A large reduction in depression severity was evident within 4 hours in all but one small trial of very low-dose ketamine, and treatment effects were largest at day 1. In the trials of very low-dose ketamine (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data), the reduction in overall severity appeared smaller and shorter lived. A test for heterogeneity indicated a significant difference in treatment effects at day 3 for low-dose compared with very low-dose trials (P=.02). In the 4 trials conducted with ketamine 0.5mg/kg i.v. for patients with unipolar depression, the large reduction in average mood score at 24 hours remained evident, though moderately attenuated, up to day 7. In the 2 trials among patients with BD, the treatment effect largely dissipated by days 4 to 7. A smaller treatment effect in BD patients compared with unipolar patients on the HAM-D was evident by 24 hours. There was no evidence of greater benefit in trials that aimed to attain high peak concentrations, either with an i.v. push during a few minutes or an initial loading dose of 0.27mg/kg during the first 10 minutes.

13Analyses of treatment effects on response and remission, including only the single parallel group trial and the first treatment period of crossover trials, are summarized in Figure 5. At day 1, there was a large treatment effect, with 50% of those in the ketamine group compared with 13% in the placebo group meeting criteria for response (RR 2.6, 95% CI 1.6 to 4.4, P=.0003; NNT 2.9, 1.9–7.1). There was a similarly large effect on remission of symptoms at day 1 (RR 5.2, 95% CI 2.1–12.9, P=.0003; NNT 4.5, 3.1–8.3) among those receiving ketamine. At day 7, response rates were substantially increased in the ketamine group and to a larger degree (RR 3.4, 95% CI 1.6 to 7.1, P=.001; NNT 6.3, 3.4–25). Remission rates were still substantially increased in the ketamine group, although to a lesser degree (RR 2.6, 95% CI 1.2–5.7, P=.02; NNT 9.1, 5–50). While the pooled results of 3 trials of very low-dose ketamine showed no significant effect on response or remission on days 1, 3, or 7 and the treatment effect appeared smaller compared to low-dose ketamine, formal subgroup heterogeneity tests were not significant (P=.09 for response and P=.12 for remission on day 3).

14One of the major potential biases in crossover trials is dropout after the first treatment, and this occurred in 12/46 patients who received ketamine initially compared with 5/55 who received placebo initially in 5 trials. Dropout was for improved mood for 4 patients in the ketamine group and 2 in the placebo group. In terms of treatment effects in the second period of the crossover trials, overall 0/34 patients responded at day 1 who received placebo as their second treatment compared with 22/50 patients responding at day 1 after receiving ketamine as their second treatment. Remission was achieved in 0/34 patients at day 1 who received placebo as their second treatment compared with 8/50 patients after receiving ketamine as their second treatment. Thus, it did appear that carryover effects were not marked, although this cannot be assessed reliably since fewer patients in the ketamine groups proceeded to the second period.

15Measures of suicidality were published in 4 trials and these are summarized in Table 2. Each trial reported a significant reduction in suicidality assessed using different measures. Seven trials provided unpublished data on the suicide item component of a depression scale, 1 for HAM-D, and 6 for MADRS (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data). Overall, reported suicidality scores were low at baseline, with an average of 1.6 on the MADRS suicidality item (which ranges from 0 to 6) for trials that reported this outcome, as shown in Table 3. Nonetheless, a significant reduction in suicidality severity score was observed for the ketamine group at days 1 and 3, as seen in Figure 6.

16Measures of tolerability and side effects were summarized in Table 4. Low-dose ketamine at 0.5mg/kg, infused over 40 minutes, was generally well tolerated, with transient, mild-to-moderate dissociative symptoms and blood pressure and/or heart rate increases in a minority of patients. For example, in the largest trial, 17% of patients had significant dissociative symptoms immediately after ketamine infusion, but these all resolved within 2 hours and no severe psychotic symptoms (paranoia, hallucinations, or delusions) occurred in either arm. The characteristic symptomatic side effects (eg, confusion, blurred vision) and the increases in heart rate and blood pressure resolved within 4 hours of administration in all trials. These side effects were less marked with very low-dose ketamine delivered intranasally for 20 minutes but were more marked with very low-dose ketamine given as an i.v. push for a few minutes. Among trials that reported a full listing of adverse events in both groups, there was no excess in the ketamine vs placebo after administration day (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data).

17Eleven events were reported as serious adverse events in the ketamine group: hypotension and bradycardia occurred in 1 person, which resolved in <1 minute and was considered to be due to a vaso-vagal episode; suicide attempt occurred in 1 patient while tapering off of psychotropic medication; tachycardia (>150 bpm) occurred in 2 patients; and mean arterial pressure elevations >20% from baseline in 5 patients (C. K. Loo, V. Galvez, E. O’Keefe, unpublished data). Also, while not reported as serious adverse events, one ketamine-treated patient developed manic symptoms that resolved within 80 minutes, and an affective switch occurred for one patient in the placebo group. All the hemodynamic side effects resolved within a few hours with no lasting effects. Thus, it could be summarized that 3 major psychiatric events (suicide attempt before treatment, transient manic symptoms, and affective switch) were reported in the studies, of which only one was potentially attributable to ketamine (transient manic symptoms). No major medical events occurred in the trials.

18Diverse measures were used to test possible psychotomimetic, dissociative, or mood elevation symptoms, including Visual Analogue Scale score for intoxication “high”, Brief Psychiatric Rating Scale, Young Mania Rating Scale, and Clinician Administered Dissociative States Scale. The results showed that Visual Analogue Scale score for intoxication “high” scores returned to baseline within 110 minutes of the infusion. In no case did euphoria, derealization, or depersonalization persist beyond 110 minutes.

19Continuation into the second phase of the crossover trial was also assessed as a proxy of tolerability, and most dropouts were reported as being due to changes in mood rather than adverse events. In 5 crossover trials with first-phase data available, 12 of 46 patients did not proceed to placebo treatment after receiving ketamine first. For 4 patients, this was due to improved mood, for 4 it was due to worsening mood, and other reasons were involved for an additional 4.

20This systematic review of trials of single-dose ketamine compared with placebo for treatment-resistant patients in a major depressive episode confirms a large reduction in depression severity and suicidality that is apparent within 4 hours after ketamine administration. The data suggest benefits are smaller and shorter lived with very low-dose ketamine. There is also a suggestion of short-lived effects in patients with BD. There is substantial heterogeneity in clinical response, with benefits lasting <1 week for most patients, but approximately one-fifth of patients remaining in remission at 1 week.

21This systematic review has a number of methodological limitations. Most important is the small sample size (average of only 23 patients per trial) and 201 patients in total. The use of a crossover design in all but one trial is a potential issue. While a crossover design improves study power to partly mitigate problems associated with small sample size, it can have limitations for mood disorder trials. Crossover designs are most reliable when assessing a fully reversible treatment, with outcomes that return to baseline levels after a suitable washout period (ie, short-lived interventions that do not affect the natural history of a stable illness). For treatments that do not have these criteria, one can observe period effects (ie, the treatment effect in the first period is systematically different from that in the second period) and/or carryover effects (ie, the treatment effect in the first period is still operating during the second period). Statistical tests to detect period and carryover effects are relatively insensitive, and analytic approaches to overcome these issues are generally unsatisfactory. However, these issues can be addressed with a separate analysis of first period data, which was possible in this meta-analysis. Finally, all trials involved one-off treatments and varying measures of side effects and had relatively short follow-up; hence, the safety and efficacy of long-term treatment remain uncertain.

22Strengths of this review include the reporting of additional unpublished data from numerous trials and analyses of only the first treatment period, increasing the reliability. It is also the first review to compare results of low-dose and very low-dose ketamine, suggesting larger benefits with the former. The consistency of the findings described in this review suggest a true treatment effect in improving depressive symptoms, albeit temporarily for most. Several additional strands of evidence indicate that the benefit observed in these trials is a result of depression treatment, rather than nonspecific mood elevation or a ‘high’ following ketamine administration: there is improvement in core depressive symptoms such as sadness, suicidality, and helplessness; the time course of benefits is in the following days and sometimes weeks, whereas mood elevation with drugs of abuse is generally limited to the time of intoxication; the response pathway can be interrupted in animal models of depression treatment; and postoperative mood improvement occurs among patients given ketamine intra-operatively, who would have been unaware of immediate subjective effects. Recent trials have also suggested possible benefits among patients with posttraumatic stress disorder and in obsessive compulsive disorder, and there is encouraging initial evidence comparing ketamine with ECT.

23Most clinicians, drug regulators, and health funders will require further research evidence before low-dose ketamine is adopted in clinical practice. The results of this review have several implications for future research.

24First, research is urgently required on efficacy and safety of longer treatment regimens. A key issue is whether the ketamine response can be maintained or even enhanced by repeated dosing, such as the 2 to 3 times weekly schedule that has been assessed in some case series with promising findings. Safety, tolerability, and abuse potential of repeated dosing is a critical issue for future trials. While some reassurance can be taken from the much more extensive use of i.v., oral, and sublingual ketamine in chronic pain control, future longer term trials must assess safety in this patient population. Trials should assess the likelihood of abuse potiential in subjects with major depression, while incorporating strategies to mitigate this.

25Second, trials should adopt parallel designs rather than crossover designs, since benefits remain for more than 1 week in a significant proportion of patients; hence, this situation does not meet the fundamental criterion for crossover trials to have a stable disease baseline and short-lived treatment effect in all patients. Third, trials should recruit many times more patients, so that the effects in different subgroups can be assessed reliably given the considerable heterogeneity in response. Fourth, trial designs should reflect the possibility of different effects among patients with bipolar and unipolar disorder. Fifth, there is no evidence concerning first-line treatment of major depressive disorder: ketamine may have a role to play in patients with severe presenting symptoms, including suicidality, and could conceivably cover the lag phase before SSRI efficacy onset (and reduction of suicidal ideation may be particularly important in young adults). Further evidence of reduction in suicidality was recently provided by an additional trial published after our search cut-off date, conducted among 27 patients with mood and anxiety spectrum disorders who presented with clinically significant suicidal ideation. Ketamine may also augment the response from standard treatments. For example, while ECT is indicated in severe depression with acute suicidality it may take a week or longer to alleviate symptoms and ketamine may play an important role by acting more quickly in this setting. Finally, further data on safety and efficacy of use in ‘real world’ clinical settings is clearly required, given that many of the trials were conducted in highly medically controlled environments. While i.v. infusions may be practical in service settings used for ECT, other formulations and delivery routes, such as intranasal spray, s.c., intramuscular, or oral, are likely to be more broadly applicable to clinical practice. Minimizing dissociative and hemodynamic side effects is important for widespread applicability, and this is likely to involve avoiding high peak plasma levels. However, very low doses are associated with lower efficacy, suggesting multiple low doses delivered gradually or sequentially may be optimal. Future trials could also assess ketamine enantiomers, metabolites, or other glutamatergic agents.

26This systematic review confirmed a large, rapid benefit in response and remission following a single dose of ketamine in patients with treatment resistant depression. There was also a reduction in suicidality. Transient psychotomimetic and heemodynamic effects occurred. There were no major medical events. Collectively, these data suggest ketamine has considerable promise for the acute treatment of major depression and provide the rationale for a large, parallel group, randomized, placebo-controlled trial to assess safety and efficacy of a longer course of ketamine in patients with severe mood disorders.

27Dr. Lapidus has received consulting fees from LCN consulting and serves on the advisory board for Halo Neuro Inc. In addition to performing ketamine research, he provides clinical treatments with ketamine. Until September 2014, Dr. Lapidus was at the Icahn School of Medicine at Mount Sinai, which has been named on a use patent on ketamine for the treatment of depression. The Icahn School of Medicine has entered into a licensing agreement for the use of ketamine as therapy for treatment-resistant depression and could potentially benefit if ketamine were to gain approval for the treatment of depression. Dr. Loo has received an honorarium from Lundbeck to speak at independent educational sessions organised by psychiatrists, for which Lundbeck provided funding. Dr. Glozier has received speaking honoraria from Lundbeck, Janssen, and Servier Laboratories, and advisory board fees from Lundbeck.