The single positive efficacy trial

The only positive study found a difference of 4 points on the Montgomery–Åsberg Depression Rating Scale (MADRS) favouring esketamine over placebo.^{Reference Popova, Daly, Trivedi, Cooper, Lane and Lim22} The MADRS scale goes from 0 to 60; the average score for participants at baseline was 37. The response to placebo treatment (a nasal spray with an embittering agent) was a 15.8-point reduction in the MADRS score. The response to esketamine was a reduction of 19.8 points.²¹ A reduction of 7–9 points on the MADRS has been found to correspond to a clinically noticeable (‘minimally improved’) change on the Clinical Global Impression scale (CGI);^{Reference Leucht, Fennema, Engel, Kaspers-Janssen, Lepping and Szegedi23} ‘much improved’ requires a reduction of 16–17 points. A 4-point difference therefore corresponds to less than ‘minimal’ change and is one-quarter the size of the placebo response, suggesting doubtful clinical relevance.^{Reference Turner18} Furthermore, participants would have been unmasked (‘unblinded’) by the noticeable psychoactive effects of esketamine (dissociation was reported by the majority of participants); expectation effects might therefore inflate the apparent difference between placebo and esketamine.

The discontinuation trial

As Janssen could not provide two positive efficacy trials, the FDA loosened its rules and allowed a discontinuation trial to provide evidence of efficacy.^{Reference Turner18} This trial randomly allocated patients who demonstrated ‘stable remission’ following a 16-week esketamine treatment (a highly selected group of participants) to either continue or stop esketamine, and measured subsequent relapse.^{Reference Daly, Trivedi, Janik, Li, Zhang and Li24}

This study design is problematic because withdrawal effects from the drug can be mistaken for relapse of depression. Ketamine is recognised to have withdrawal effects, including lowered mood (dysphoria), fatigue, poor appetite and anxiety.^{Reference Chen, Huang and Lin11} Although the study report suggests that there was no evidence of a withdrawal syndrome on the Physician Withdrawal Checklist, scores for the different groups are not reported and it is not clear how items in the checklist such as ‘insomnia’, ‘anxiety–nervousness’, ‘dysphoric mood–depression’, ‘difficulty concentrating, remembering’, ‘fatigue’ and ‘lack of appetite’ were distinguished from almost identical items in the MADRS (‘apparent sadness’, ‘reported sadness’, ‘inner tension’ ‘reduced sleep’, ‘reduced appetite’, ‘concentration difficulties’, ‘lassitude’).

As half (48.7%) of relapses occurred in the first 4 weeks following esketamine cessation, the time most likely for withdrawal effects to occur, and as the relapse rate in the placebo group became ‘closer to esketamine with each week’, as highlighted by the FDA, confounding of ‘relapse’ by withdrawal seems likely.²¹ Further evidence of a withdrawal effect is also suggested by the marked ‘relapse prevention’ effect of a drug with minimal antidepressant effects in the short term. This pattern is similar to what might be seen in a trial of a benzodiazepine for anxiety: modest effects in the short term, but marked ‘relapse prevention’ effects on discontinuation, if confounding by withdrawal effects are ignored.

The FDA also highlighted another problem with this study design: ‘unblinding’,²¹ as in the acute efficacy studies. The absence of esketamine's psychoactive effects would be noticed by participants randomised to placebo and consequent negative expectations would tend to increase their chance of relapse.²¹ Higher dissociation scores while on treatment were correlated with shorter time to relapse, consistent with this hypothesis.

Importantly, the FDA also raised the concern that the positive results of this study were driven by a single site in Poland. At this site there was a 100% relapse rate in the placebo arm (n = 16 participants), compared with a 33% relapse rate in this arm in all other sites (n = 81 participants).²¹ It has been demonstrated that if this outlier site is excluded there is no difference between esketamine and placebo (the P-value changes from 0.012 to 0.48),^{Reference Turner18} leading to the conclusion that the findings are ‘not robust’.

The safety of esketamine in trials submitted to the FDA

When licensing esketamine, the FDA recommended a risk evaluation and mitigation strategy (REMS) because of concerns about sedation, dissociation and diversion. This requires the drug to be administered in a healthcare setting, and for patients to be monitored for resolution of sedation, dissociation and changes in vital signs for at least 2 h. There were, however, some concerning safety signals that many commentators felt were not adequately addressed by these measures.

In the Janssen licensing trials there were six deaths,²¹ all occurring in the participants who were assigned to esketamine. Three were suicides (Table 1).

Table 1 Deaths reported in esketamine trials^a

a. These data are derived from the Food and Drug Administration review of data submitted by Janssen.^{Reference Kim and Chen30}

b. Total number of participants exposed in Stage 2 and Stage 3 trials up to 4 September 2018.^{Reference Kim and Chen30}

One death was a motorcycle accident 26 h after esketamine administration.²¹ One death was a myocardial infarction, 6 days after esketamine.²¹ The third death was due to acute heart and lung failure 5 days after esketamine, in an individual with risk factors. These causes of death have been observed previously in ketamine use: blood pressure spikes cause heart failure and myocardial infarction in patients at risk; impairment of hand–eye coordination and its dissociative effects increase the risk of motor vehicle accidents.^{Reference Morgan and Curran3,Reference Suleiman, Ik and Bo4,Reference Schifano, Corkery, Oyefeso, Tonia and Ghodse8,Reference Cheng, Chan and Mok9} However, the FDA argued that one participant's cardiovascular parameters had been normal before and after their last administration of esketamine, and in the other that esketamine produces only short-lasting blood pressure rises, concluding that ‘it is difficult to consider these deaths as drug-related’.²¹ In addition to the deaths, one participant on esketamine experienced a non-fatal cerebral haemorrhage,^{Reference Cristea and Naudet19} also recognised as a risk in ketamine use owing to its blood pressure spikes,^{Reference Morgan and Curran3} and five further participants in the esketamine arms had non-fatal motor vehicle accidents.²⁵

The three suicides occurred 4, 12 and 20 days after the last dose of esketamine.^{Reference Schatzberg26} The FDA attributed these deaths to ‘the severity of the patients’ underlying illness’.²¹ However, two of the participants who died by suicide showed no previous signs of suicidal ideas during the study, either at entry to the study or at the last visit (data were not available for the third participant).²¹ The FDA accepted Janssen's assessment that the suicides were not ‘drug-related’.²¹

However, others have argued that these cases might fit with a pattern of a severe withdrawal reaction, consistent with other reports of suicide associated with recreational ketamine,^{Reference Schifano, Corkery, Oyefeso, Tonia and Ghodse8,Reference Cheng, Chan and Mok9} and are significant enough in number to constitute a worrying signal.^{Reference Schatzberg26}

An increase in depression and suicidality was also observed during esketamine treatment, compared with placebo (Table 2). In the short-term trials, six participants (1.4% of the total) in the esketamine group became more depressed, compared with only one on placebo (0.2%); five participants expressed increased suicidal ideas in the esketamine group (0.9%), compared with two on placebo (0.5%).²¹ The drug will be marketed with a ‘black box’ warning including a risk of suicidal ideas and behaviour,^{Reference Cristea and Naudet19} but it is not clear whether this measure is stringent enough.

Table 2 Selected adverse events reported in esketamine trials^a

n.r., not reported.

a. These data are derived from the submission by Janssen²⁵ and the Food and Drug Administration (FDA) review of the data.^{Reference Kim and Chen30} However, the FDA^{Reference Kim and Chen30} noted that there was misclassification, underreporting and misrepresentation of the adverse events, so that the presented data likely represent an underestimation of the incidence of adverse events. Differing dosages of esketamine have been combined for simplicity.

b. Total number of participants exposed in trials 3001, 3002, 3005 (4-week efficacy trials) and trial 3003 (discontinuation trial).^{Reference Kim and Chen30}

c. Reported in Janssen submission to FDA without relevant sample size.²⁵

d. Total number of participants in completed Stage 3 studies.²⁵

Potentially paradoxically, Janssen is seeking an expansion of the indications for use of esketamine to include acutely suicidal patients (on the basis of two studies that showed a reduction of suicidality on a subscale of the MADRS at 24 h, but no effect on the pre-specified measure of suicidality, the CGI Severity of Suicidality (Revised)).^{Reference Jancin27}

A large range of other side-effects occurred: half of participants experienced dissociation and one-third dizziness; increased blood pressure, vertigo, hypoaesthesia, nausea and sedation were each present in between 10 and 30% of participants (Table 3).²¹

Table 3 Most common adverse events from esketamine compared with placebo^a

a. Data from the 4-week studies 3001 and 3002.^{Reference Kim and Chen30} Differing dosages of esketamine have been combined for simplicity.

A significant number of participants on esketamine developed signs of bladder irritation, reminiscent of ‘ketamine bladder’: urinary tract infections, pain, discomfort, cystitis and nocturia.²¹ In the 60-week study, with weekly or fortnightly esketamine administration (less frequent than the short-term trials), a fifth of participants reported bladder effects. Some cognitive impairment was detected in the long-term study among older participants. The FDA, however, argued that the lack of a comparison arm and individual variability made it difficult to conclude that these effects were attributable to esketamine, even though they are well-known consequences of ketamine use.²¹