“Two esketamine PhIII trials were completed three months ago, and nothing has been disclosed from their results. It seems odd that a company like JNJ/Janssen, which has pioneered prosocial alliance-building via JLabs and their Innovation Centers, has now gone against common practice and regressed to a stance of corporate omerta, not even releasing top-line results. Trying to read between these lines is a fool’s errand. We do not know if JNJ is indulging itself in the unhurried luxury of positive results, or shoveling dirt from the corpses of clinical failures to see if data-mining can produce some kind of resurrection.” —NeuroPerspective, March 2018 Depression review
Or there is a mishmash of clinical data that, while only minimally disclosed thus far, looks to be less definitive than was expected. What was also unexpected is that JNJ/Janssen, having had months to massage the data and its disclosure, was haphazard and careless with a standard press release.
To recap briefly, two trials were done using intranasal esketamine, one in adults with treatment resistant depression (TRD=at least two failed trials with oral antidepressants), one in elderly patients with TRD. Dosing was twice weekly for four weeks, a flexible dosing paradigm was used. Both trials used as newly initiated co-therapy one of four oral antidepressants, chosen from two SSRIs, two SNRIs. The criteria/proportions of selection were not disclosed, which begs the question of whether the specific drug, or mechanism, might have accounted for any ‘noise’.
One-tailed p values were reported. This would allow reporting smaller p values, but in a drug testing scenario, two-tail analysis is generally done. The criterion for ‘success’ was thus p=.025. It is puzzling why they did it this way.
At 28 days, the esketamine adjunct group did separate from the placebo adjunct group, the one-tail significance was p=.01. This indicates a durable effect of esketamine over those four weeks, its clinical magnitude is unclear. The percentage of 28 day responders in the esketamine group was 69.3% (50% or better improvement on MADRS: 52% for placebo), remission rates also showed numerical superiority, 52.5% versus 31%. Statistical significance was not
reported for those comparisons.
However, with esketamine billed as a rapidly-acting antidepressant, the other critical criterion besides durability is speed. Here, the esketamine group had numerical superiority on a secondary endpoint whose criterion was improvement at 24 hours, carried out for 28 days, but the one-tailed p value was p=.161, far from establishing statistical significance.
At 28 days, the esketamine group was numerically superior on the primary, but just missed significance (p=.029). A median change of 3.6 points was reported, which JNJ then compared to historical data regarding changes in MADRS scores, in order to imply superiority. Nothing was said about speed or responder/remission rates.
The adverse event reporting for both trials simply listed the numerous dissociative and more side effects one would expect from ketamine or esketamine that achieved greater than a 10% incidence rate. More color on the actual frequencies of side effects would have been much more useful.
The most egregious errors were in the comments by Husseini Manji, JNJ’s Global Head for Neuroscience. He was quoted in the press release as saying that:
“they (the results) mark the first time an antidepressant has achieved superiority versus an active comparator in any clinical trial for major depressive disorder.”
Uh…no. These trials were both adjunctive, both groups had an oral drug on board: The comparator was placebo, not an active drug.
2) “the response was rapid”: With a p value of .161, it really pushes the envelope to claim that such an effect was demonstrated. There is an 84% chance that the difference at 24 hours in the adult trial was ‘real.’ Nothing regarding 24 hr endpoints was reported for the elderly trial, which does not augur good news.
We have two words for JNJ: Data tables. They would have been helpful. We know, JNJ wanted to save something juicy for APA. Perhaps data will be presented this week that will provide additional validation for the speed claim, but the inconsistency of what was reported for the two trials was sloppy, there is no justification for the ‘active comparator’ statement, and this reduces our temptation to give JNJ the ‘benefit of the doubt’ on these other points.
The bottom line is that esketamine did, to some degree, make its case on the 28 day data, but the magnitude of benefit in terms of speed is more ambiguous than we would have expected, and the effects on treatment response and remission were less dramatic than might have been hoped. The high profile of these trials may have set the stage for a high placebo effect, reducing the signal strength; depression trials are indeed hard to execute. But press releases should not be.