The Myth of the Miracle Cure: Is Ketamine an Efficacious Antidepressant?

This blog post is the result of a collaboration between Dr Lucy Robinson (Twitter, email) & me.

Life with depression can be miserable, painful and sad. The suffering it causes is indubitable and it is imperative we do a better job at helping people feel better, with effective, practical, and long-lasting treatments.

The promise of resolution of symptoms within hours that has recently been offered by trials of ketamine and its derivatives¹ has raised hope that swift effective treatment could soon be a reality. But anything more than a cursory glance at the literature tells a different story. The effects observed have not been shown to last. Single-group, non-randomised, uncontrolled studies in very small samples dominate the picture. And we must not forget that we are talking about ketamine—a known drug of abuse.

The New Evidence

This blog looks more closely at a recent study by Canuso et al. 2018², published in the American Journal of Psychiatry (full text). The paper raised the standard of the existing evidence by publishing a double-blind placebo controlled trial of esketamine (a type of ketamine) in patients needing hospital admission for suicidality.

Sixty-eight patients assessed at the point of presentation to hospital—all judged to require hospital admission due to high risk of suicide—were randomised to a nasal spray of esketamine twice a week for 25 days or a suitably embittered placebo. Thirty-five patients received esketamine and 31 placebo. The randomisation was well-conducted, good attempts were made to blind participants and researchers to treatment allocation, and the sample size was one of the largest available in a controlled ketamine study. This is no easy population to target, and one desperately in need of new treatment options. The team’s efforts are truly commendable, and we are well aware that such a study with a sample of about 70 participants enrolled under such difficult circumstances equals years of intense effort. Here is the abstract:

At the same time, the gravity of the topic also means we should do our best to present the data as faithfully as possible to the public. So what do Canuso et al. find? They report three main outcomes (depression severity measured on a rating scale; a single suicide item of that scale; and clinician’s judgement of suicide risk) at three timepoints (4 hours, 24 hours, and 25 days after ketamine/placebo administration). On a very positive note, there was a rapid, large drop in depressive symptoms in both groups, suggesting that the care they were getting was of great benefit. What were the group-specific effects?

In the first 4 to 24 hours of receiving the nasal spray, the ketamine group showed a steeper drop in scores on the depression scale than the placebo group. But at day 25, there was no difference in depressive symptoms between the groups. The one suicide item of the depression scale showed differences between groups 4 hours in, but not 24 hours later or at day 25. Finally, clinicians did not observe any difference between the two groups in terms of suicidal ideation at any time point. Side-effects more common in the ketamine group were nausea, dizziness, dissociation, unpleasant taste, and headache.

A case had been building for a rapid reduction in suicidality following ketamine administration³, and the FDA have issued a Breakthrough Therapy Designation for ketamine for the treatment of suicidality. But this study bucks that trend. In a prior blog, we have already heralded caution on the over-interpretation of small studies with methodological weaknesses, and others have raised this point with specific regard to the ketamine literature⁴. We know that small underpowered studies are at risk of false positive results that fail to replicate, especially in the face of publication bias, file drawer effects, and bias introduced by industry sponsorship and conflicts of interest, all of which results in non-significant trials being published less often.

In case of the present study, Canuso et al. 2018 did a very thorough job measuring suicidality with multiple different measures from the perspective of both the patient and the clinician; they used 3 scales—the Beck Scale for Suicide Ideation, the Beck Hopelessness Scale, and the Suicide Ideation and Behavior Assessment Tool, a new scale designed to be a more sensitive measure of changes in suicidal thoughts and behaviour. In principle, this is not a bad idea, because it increases the robustness of a scientific result if it can be replicated within the same study across multiple measures, and can additionally be helpful to cover the whole breadth of a construct. In the study, these 3 measures did not show any differences between groups, but the score on the suicide item of the depression questionnaire dropped more rapidly in the 4 hours after the first dose for the participants who received ketamine. As far as we can tell, this particular measure was not identified in advance in the trial protocol⁵ as a primary or secondary outcome, and neither was a 24-hour timepoint.

This calls for considerable caution in interpreting the finding, as the analysis was not registered in advance, and is inconsistent with other assessments of suicidal ideation—looking at the protocol in more detail, all of the specific changes in suicidality that the authors registered before the study were negative. Interestingly, it appears that some alternative outcomes were reported in the study that were, as far as we can tell, not declared in advance, such as 2-day follow up, and % of patients scoring 0 or 1 on the clinician global judgement of suicidality⁶; and some preregistered analyses remained unreported, such as change in Clinician-SIBAT after 4 hours, 25 days, and 81 days, and change in patient-SIBAT after 25 days.

Additional Challenges

We see some additional challenges for research in this area. First, measuring the effects of ketamine taxes our current measurement methods. Depressive symptoms are typically measured over time periods of days to weeks. We do not yet know enough about the hour-by-hour dynamics of depressive symptoms to understand the implications of shortening the timescale of these measures to just a few hours⁷. Researchers, including Canuso et al. 2018⁸, have had to be creative in solving this problem and some have opted to carry over scores from symptoms that cannot change much over a short timescale (e.g. sleep, weight, appetite, libido). In doing so, we are no longer measuring the whole of the depressive syndrome as defined in current diagnostic nomenclature, which might lead to problems in terms of content validity⁹. It will be important for researchers to be fully transparent about specifically which items are being carried over and for how long so this can be factored into understanding their results. Complementary research into idiographic symptom dynamics and measurement methods will also help here.

Another difficult problem to solve in ketamine research is the powerful short-term dissociative effects it induces, which might explain the potential effects on depression symptoms a few hours after administration; note that a 4-hour dissociative drug effect is not the same thing as therapy or treatment. From a trial methodology point of view, it is also challenging to overcome the potential unblinding effects this could have. Canuso et al. give this point good airtime in their discussion and very helpfully provide clinician-rated dissociation scores 40-minutes after administration—the differences between groups are marked. Some of the key outcomes are other clinician-rated scales though, and functional unblinding is a definite concern, which could mean reported effects are overestimates. Even psychoactive placebos are not a complete solution, as few other drugs have such a potent dissociative effect. But we need to keep in mind that successful blinding is necessary by definition to successfully carry out a randomized control trial. This is by no means criticism of Canuso et al. who took this problem very seriously, and provide a much more careful discussion about the topic than most prior work. But remains a general challenge that is worth thinking about in future work.

Then there is the question of whether the observed effects on mood are specific to ketamine–boiling it down to the bare bones, studies to date have shown that an acute dose of a potent psychoactive drug has a temporary effect on mood. This is well known, of course, and there are licit and illicit substance industries based entirely on this principle. Although administration of esketamine continued throughout the 25-day period, the effects had worn off by the end on a steady dosing regimen. Do people need more? Would the results be the same if other psychoactive substances were administered? As Joanna Moncrieff points out in her 2015 World Psychiatry letter¹⁰:

“Most drugs with psychoactive effects—including many antipsychotics, benzodiazepines, stimulants, buspirone and opiates—produce the same changes as so-called antidepressants in randomized trials in people diagnosed with depression. Moreover, antidepressants themselves come from a wide array of chemical classes, and produce diverse pharmacological effects. Unsurprisingly, it seems that the experience of taking some sort of mind-altering substance produces a slightly different result from taking an inert placebo, when you attempt to measure people’s thoughts and feelings.”

Finally, the study by Canuso et al. was funded by Janssen Research and Development, which belongs to Johnson & Johnson. Several authors, including the first author, are either employees at Janssen Research and Development, hold stock or stock options in Johnson & Johnson, or are inventors of patents for technology used in the study. All of this is properly declared in the study, and there is no principled concern of conflicts of interests specifically for this study. But in general, this raises the question what impact such funding has on the conduct of a study and the reporting of the results. This is most important when researchers have considerable degrees of freedom in their measurements or analyses that can potentially be exploited, as demonstrated clearly by Steegen et al. in 2017¹¹, and happens regularly (industry-sponsored studies are more likely to report favourable results for the drug manufacturer than independently-financed trials¹²).

Where does this leave us?

Understandably desperate though we are for fast, effective treatments, the standard of evidence needs to be very high—and the risk/benefit profile absolutely clear—before potent psychoactive drugs become a realistic clinical option. Whilst the quality of the evidence is inching upwards, there remains a distance to go. At the moment, the conclusion one should draw from the ketamine literature is largely unaltered: The drug provides temporary relief from some depressive symptoms, perhaps in similar ways other illicit substances could. But it represents something very far short of an actual treatment, i.e. a lasting resolution of symptoms, or even—until the risk profile is clearer—a safe treatment in a crisis situation.

The iatrogenic harm suffered by those who became addicted to benzodiazepines is not yet out of living memory and stands steadfastly as a cautionary tale—in the face of profound distress, clinicians have great power, especially when offering ‘cures’ with dramatic effects and apparently limited consequences. We do those in distress no justice by blinding ourselves to methodological weaknesses that permit a rose-tinted view of a landscape that might hold something far less rosy in store when the real world leaves us nothing to hide behind. Ketamine treatment is worth exploring, like many other treatments. It’s a difficult research topic to tackle, and Canuso et al. have done very well in navigating some of the big challenges. But we owe it to patients, our funders, and the public, to limit research degrees of freedom as much as possible (e.g. by registering all outcomes and timepoints before data data are in, and by faithfully reporting all outcomes and timepoints that were registered), and to communicate results as clearly and faithfully as possible to facilitate optimal clinical decisions.

1. https://www.ncbi.nlm.nih.gov/pubmed/24688759
2. https://doi.org/10.1176/appi.ajp.2018.17060720
3. https://www.ncbi.nlm.nih.gov/pubmed/27232360; https://www.ncbi.nlm.nih.gov/pubmed/25773961
4. e.g. https://www.ncbi.nlm.nih.gov/pubmed/24688759
5. https://clinicaltrials.gov/ct2/show/NCT02133001
6. Nothing wrong with doing exploratory analyses, obviously, as long as they are declared as such
7. Although there is a lot of new research on the way regarding the dynamics of emotion and symptom time-series; see the work by Laura Bringmann, Marieke Wichers, Aaron Fisher, and others
8. https://www.ncbi.nlm.nih.gov/pubmed/16894061; https://www.ncbi.nlm.nih.gov/pubmed/20679587; https://www.ncbi.nlm.nih.gov/pubmed/22297150; https://doi.org/10.1176/appi.ajp.2018.17060720
9. And depression scales already vary a lot in terms of symptom contentin the first place: https://www.ncbi.nlm.nih.gov/pubmed/27792962
10. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4592647/
11. https://www.ncbi.nlm.nih.gov/pubmed/27694465
12. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC156458/; https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2872821/