Besides apprising us of the height at which FDA currently sets the bar for approval of new drugs in psychiatry, the recent approval of the s-enantiomer of ketamine for treatment-resistant depression provides grist for conjecture about what may be an increasing resemblance between criteria for approval of prescription drugs, which traditionally have required compelling evidence (before approval) of both safety and effectiveness for the indication applied for, and non-prescription drugs, whose approval has depended principally upon safety and permitted “truth in advertising” and post-approval sales to identify products that live up to their efficacy claims.
Emerging to compensate for apparently diminished influence of evidence-based effectiveness with respect to approval of prescription drugs have been a “precedent effect,” such as that of s-ketamine, claimed to represent the first candidate of a novel class of antidepressant drugs submitted for approval for an indication (treatment-resistant depression) characterized by a dearth of treatment options. .
In two of its three preapproval efficacy trials, ketamine proved no more effective than placebo with respect to a priori primary outcome criteria—a finding of evidence dismissed on the basis of a rationale derived from findings of non-comparable provenance, excerpted from a review of studies wherein up to fifty percent of previously approved and “established” antidepressants had been found no more effective than placebo in comparisons from larger studies designed to investigate hypotheses other than comparative efficacy of placebo and the established antidepressant.
Diminishing emphasis on critical scrutiny of preapproval trials of ketamine that were interpreted to substantiate efficacy despite failure to separate from placebo on primary outcome criteria in two of three studies, was insult added to the accompanying injury of a dismissive investigation of what had been interpreted to be depressive relapse in the ketamine discontinuation study, wherein an unusual concentration of early relapses consistent with a discontinuation or withdrawal syndrome of ketamine (documented to act on mu opioid receptors, to have opioid-like analgesic effects, to reduce opioid dose requirements postsurgically, and to be antagonized clinically by the opioid antagonist naltrexone) prompted a desultory follow-up investigation, despite three suicides in the discontinuation group (abruptly withdrawn from ketamine after 16 weeks). A fair question broached by the momentum to approve “a new class of antidepressants” that subsequently failed to demonstrate efficacy for depression as a syndrome (and instead detected a transient state change 24 hours after infusion, when assessment of changes in sleep, energy, appetite, concentration, and stable mood change cannot be assessed, and elsewhere reported acute, transient remission of suicidal ideation, distinguishable from and reported to occur independently of syndromic improvement) is whether diminishing emphasis on effectiveness as a criterion for approval may conduce to disinclination to detect and analyze critically significant irregularities.
Like ketamine a candidate for adjunctive use in treatment-resistant depression is a combination of buprenorphine (a partial agonist of both the alpha and kappa opioid receptor subtypes) and samidorphan (a mu opioid receptor inhibitor).
Distinguished, as is ketamine, by equivocal efficacy, the combined formulation of buprenorphine and samidorphan too has proved no more effective than placebo, and has so in three clinical trials (including that published most recently (M. Fava and colleagues, American Journal of Psychiatry, published on line , ahead of print, in May, 2016)). We abstract it herein, however, by virtue of how candidly and unselfconsciously its authors reanalyzed and reinterpreted initial evidence of inefficacy by de-randomizing patient assignments to treatment. As such, their study accents the importance of randomization in contolled trials, and additionally reminds us that what evidence shows and what conflicted interest may compel its authors to assert may differ (with the cause of the difference hidden in plain sight).
The authors’ findings derive from a subgroup of subjects in a 31-site study tasked to demonstrate the effectiveness for treatment-resistant depression of an adjunctively prescribed formulation of two drugs with opioid activity: buprenorphine (a partial agonist at both sigma and kappa opioid receptors) and samidorphan (an inhibitor of the mu opioid receptor). The rationale for combining them is not disclosed (on any scale), but the purpose is (to convey augmented antidepressant effectiveness that does not cause opioid dependence). Why they believe that a combination of buprenorphine and samidorphan will do so, is not elaborated.
The authors divide their study into two stages, which are in actuality a first, randomized, double-blind comparison in patients with purportedly treatment-resistant major depressive disorder. To one group of 95 patients, they administer placebo, and to two others (of twenty patients each) the combination formulation of buprenorphine and samidorphan (2 mg of each in one group and 8 mg of each in the other, both sublingually and daily), for four weeks, assessing response using the Hamilton Depression Rating Scale as the primary outcome criterion. The study ends after an additional one week during which the combined opioid formulation is tapered and stopped.
Each group shows change in the direction of improvement, but no statistically significant differences are observed among them. (The abstract of the study gives the impression that this first, fair study, designated “Stage 1,” showed statistically significant superiority of buprenorphine and samidorphan (2 mg each), relative to placebo.) Neither dosage of the combination was associated with greater reductions of depression ratings than was the placebo group.
The authors then performed a second study (designated, “Stage 2,” wherein they randomly assigned only patients who did not respond to placebo in the Stage 1 study to placebo and each dosage of buprenorphine-samidorphan, adjunctively. After four weeks of treatment, the placebo group showed significantly less improvement than did the group receiving 2 mg of burprenorphine with 2 mg of samidorphan.
No statistically significant difference was observed between placebo-treated patients and patients treated with 8 mg (each) of buprenorphine and samidorphan.
Adverse effects occurred in 85 percent, the most common of which were sedation and somnolence. No patient suffered serious adverse effects or laboratory abnormalities, nor were any patients reported to manifest withdrawal symptoms after four weeks of daily administration.
(Obiter: The authors’ findings—of no difference during their first stage following randomization, followed by statistically significant superiority of their lower dosage formulation—was predictable after enrichment of their control group for lower placebo response (15 to 20 percent after selection for no response during the first study.) They did not study the comparability of their drug-treated groups during the second, non-randomized experiment, and could find no prior literature in which placebo-nonresponders during a first, “screening” experiment were exposed to active drug during a second trial with a naturalistic drug-treated group with normal potential for response to inactive compounds.
The artifact introduced by derandomization exemplifies the importance of random assignment to treatment in drug trials. Consider a manufacturer of Drug A who observes no statistically significant difference between its drug and competing manufacturer B’s Drug B. Wishing to introduce an artifact that would make its Drug A appear superior to the competition, what could be more transparently prejudicial than to select for patients who do not respond to Drug B and assign them to treatment with Drug B during a second trial comparing Drugs A and B? The authors’ introduced the same prejudicial condition when they refined their Stage 2 placebo group with placebo-unresponsive subjects.)