Editor’s Note: This piece originally appeared on the Mad in America Substack.
Two positive studies out of 13 conducted. Rejected by the FDA four times. An FDA advisory committee vote against approval. Yet in 2023, gepirone (Exxua) was approved for treating depression. As of December 2025, it’s now commercially available.
How does this keep happening? How can the FDA leaders keep approving drugs against the advice of its scientific advisory boards (see the aducanumab controversy)? How can they keep approving drugs that fail to show efficacy in the majority of their trials, with the positive results likely due to chance or methodological bias (see the esketamine controversy)?
And shouldn’t the product label be required to contain all the evidence on the drug’s supposed efficacy? The data from the 11 failed studies? The data showing the drug performed worse than existing antidepressants? The data showing lack of a clinically significant effect? Isn’t that information vital for doctors to be able to make informed prescribing decisions?
In a new article in top-tier psychiatry journal JAMA Psychiatry, Erick H. Turner and his co-authors John H. Powers III, Rosa Y. Ahn-Horst, and Aaron S. Kesselheim bring the whole process to light. Turner is a former FDA reviewer, now professor emeritus of psychiatry at Oregon Health and Science University. He sat on the advisory committee for esketamine, and has since become an outspoken advocate for FDA reform.
Doctors and the public perceive FDA approval to mean that a drug has been proven safe and effective. As Turner writes, “Among surveyed US adults, almost half mistakenly believe that the FDA approves only extremely effective drugs.” And this is not just true of laypeople: “Among surveyed physicians, 70% mistakenly believed that, for the FDA to approve a drug, clinical trial results must show both statistical and clinical significance, and 73% believed that new drugs must be as effective as drugs already approved for the same condition. Moreover, when physicians were asked whether new drugs must be more effective than already-approved drugs, nearly 40% responded affirmatively.”
“This lack of insight into the data underlying FDA-approved drugs may help explain why clinicians tend to overestimate interventions’ benefits and underestimate their harms,” Turner adds.
In truth, the actual FDA approval process does not ensure clinically significant effects. It doesn’t ensure drugs that equal or beat existing drugs. In many cases, it doesn’t even ensure overall efficacy.
And the FDA’s approval of gepirone provides a perfect case study of regulatory corruption and industry manipulation in service of selling ineffective drugs.
Gepirone’s story begins in 1986, when it was first synthesized by Bristol-Myers Squibb. By 1993, they had stopped working on the drug and sold the rights to Fabre-Kramer. In 1998, Fabre-Kramer made an agreement with Organon to work on the drug together, and Organon soon submitted a New Drug Application (NDA) to the FDA. The FDA rejected it with a Refuse to File letter (see pages 62-64), taking issue with the methodology of the studies, the fact that several were negative, and more. Organon resubmitted in 2001, and the FDA rejected this as well, citing lack of evidence for efficacy in a Not Approvable letter. In 2003, Organon added further data from the maintenance trial that they claimed supported the drug, but the FDA found this unconvincing and noted that the company appeared to have manipulated the data by reclassifying relapsed patients after unblinding, among other issues (see page 21). The FDA considered this a negative trial, and issued another Not Approvable letter.
In 2005, Fabre-Kramer took back all the rights to the drug from Organon and soon submitted another NDA to the FDA. This NDA included 12 short-term trials and one maintenance trial. The FDA issued another Not Approvable letter in 2007 and, in a meeting with Fabre-Kramer, again told them that their data had not shown sufficient evidence of gepirone’s efficacy.
One might assume that the story should end there. The fourth NDA had included all 13 trials that were conducted and now the FDA had rejected the drug four times (with three Not Approvable letters), stating that these 13 trials did not provide sufficient evidence of efficacy. What more could Fabre-Kramer possibly do to change the FDA’s decision? Indeed, they were quiet for several years.
Then, in 2011, Fabre-Kramer formally asked the FDA to reconsider, and in 2012 they submitted an NDA amendment suggesting that the FDA should ignore the negative trials and focus on only the two positive trials (and that the maintenance trial should be considered positive, despite the FDA’s judgment that it was negative). The FDA remained unconvinced. In 2014, the FDA issued a General Advice Letter restating that the studies had not shown sufficient evidence of gepirone’s efficacy.
Fabre-Kramer then submitted a Formal Dispute Resolution Request and met with the FDA in 2015. An advisory committee of scientific experts was formed to analyze the evidence. The advisors voted 9 to 4 against approval, saying that Fabre-Kramer had not shown sufficient evidence of efficacy. The advisors were not as concerned about safety (they voted 11 to 2 in favor of the motion that Fabre-Kramer had shown the drug was safe).
Again, one might assume the story should end there. Four submissions, four rejections, and now an advisory committee vote against approval, with experts saying Fabre-Kramer hadn’t proved that gepirone worked; eleven failed studies with only two showing marginal efficacy against placebo.
But the story continued. In 2016, the FDA granted Fabre-Kramer’s appeal, saying that they would consider approving the drug but only if the drugmaker showed more evidence of the drug’s safety. This was odd because the advisory committee had already said there was enough evidence of safety—it was the lack of efficacy that concerned them.
The letter granting Fabre-Kramer’s appeal came from the Director of the FDA Office of New Drugs, John Jenkins (see pages 40-46). Jenkins argues in favor of the drug, stating that some of the failed trials “numerically favored gepirone over placebo on the primary endpoint and in the case of Study 008 there was a trend (p=0.2).” (This p value is not even close to statistically significant. Requiring statistically significant results is a basic tenet of evidence-based medicine. Some argue that we should go farther, requiring larger effect sizes or other measures, but statistical significance is the first bar that should be cleared.)
Jenkins also discounts four failed trials that were stopped early after gepirone appeared worse than placebo (and active comparators) by blaming “business reasons” for their failure. He argues that the FDA should ignore the negative maintenance trial too, since “a negative maintenance trial does not prove that gepirone does not work for this indication.”
Jenkins acknowledges that FDA experts “raised concerns that the two positive trials could have occurred by chance (i.e., may be false positives),” and notes that they “concluded based on the counting analysis that the probability that the two positive trials were false positives was unacceptable and that the meta-analysis of the non-positive trials provided no evidence to support the efficacy of gepirone.”
But Jenkins then rejects the analysis of the FDA’s experts. He dismisses 5 of the 11 negative trials as simply having “failed” and argues that they should not be counted in the analysis. He then reanalyzes the probability of false positives and claims that (contrary to the expert opinion) that probability is low enough to consider the two positive trials to be “true positive” results.
Jenkins further argues that existing antidepressants often don’t beat placebo, too, so it actually doesn’t matter if gepirone beats placebo or not. “It is not uncommon for effective anti-depressants to fail to beat placebo in adequate and well-controlled trials,” he writes. “An analysis presented by Dr. Mitchell Mathis, Director of DPP, at the advisory committee meeting showed that this happens on average 50% of the time for the approved anti-depressants.”
To paraphrase: according to Jenkins, we know that antidepressants are “effective” despite the fact that they “fail to beat placebo,” so a new drug failing to beat placebo is not a problem.
Jenkins then goes further, dismissing several studies that showed that gepirone actually performed worse than existing antidepressants. “Even if we were to conclude that gepirone is inferior to one or more active control,” Jenkins writes, “that would not necessarily preclude approval with appropriate labeling.”
Thus, according to Jenkins, gepirone should be approved despite not beating placebo, and despite not even showing equal effectiveness to existing antidepressants that also fail to beat placebo.
Yet in the 2014 General Advice Letter (see pages 55-60), Deputy Director Robert Temple made it clear that it was a serious problem that gepirone kept showing up as significantly worse than existing antidepressants. Four studies, he writes, “included an active-control arm (ORG134004, ORG134006, CN134017 and CN105053), in which the active control performed statistically significantly better than gepirone ER or placebo on the HAMD-17 scale.” In these studies, gepirone performed significantly worse than the “active control”—existing antidepressants fluoxetine (Prozac), paroxetine (Paxil), and imipramine.
Temple is also quite dismissive of Fabre-Kramer’s attempts to call two further studies “positive” despite the fact that gepirone failed to beat placebo. “Per your own analysis,” he writes, “gepirone ER did not reach statistical significance over placebo either on the primary endpoint or on almost every secondary variable. We continue to interpret these studies as negative gepirone ER trials.”
Ultimately, Temple writes, “Although two short-term trials favor gepirone ER for the treatment of MDD, the seven negative short-term studies and one negative maintenance trial with gepirone ER raise considerable doubts about the effectiveness of gepirone in the acute or sustained treatment of depression. The 2 positive studies could represent chance findings, given the absent, negative, or minimal findings in 8 other studies.”
So how did Jenkins come to such a different conclusion? The breakpoint for Jenkins, it seems, came from a meta-analysis conducted by Fabre-Kramer. (A meta-analysis is used to combine and analyze all the data from various trials. By pooling the smaller trials into a larger sample, this type of study can be used to detect a smaller effect that might still be meaningful.)
Yet with 11 failed trials and only two positive studies, it would still be unlikely that a meta-analysis would show the drug to be effective. So how did Fabre-Kramer come up with a positive conclusion? Simple: they removed more than half of the failed studies from the analysis—”finding, based on 5 trials, that there was an even slightly greater separation from placebo,” Turner writes.
Even this heavily manipulated analysis found that the drug barely reduced depression scores, beating placebo by only 1.32 points. (Thus, they did not meet even their own definition of clinical significance, a 2-point difference.)
It appears to be this meta-analysis that swayed Jenkins toward approval. After explaining the meta-analysis, Jenkins writes, “After considering all the data and analyses I conclude that Fabre-Kramer has provided data to support a finding of substantial evidence of effectiveness for gepirone in the short-term treatment of MDD.”
In 2016, the Director of Center for Drug Evaluation and Research, Janet Woodcock, appeared to agree with Jenkins and supported the ultimate approval of gepirone, despite Temple’s concerns (see pages 2-17). With the support of Woodcock, gepirone was approved on September 22, 2023.
By 2025, the drug—branded Exxua—became commercially available, at a price point of about $1,788 per month. It’s projected to make Fabre-Kramer $46 million in 2026, increasing to $165 million a year by 2029.
Now that Exxua is being sold to the public, does the product label at least clarify the data used by the FDA to approve the drug? Surely, for doctors to be able to prescribe this drug they must be informed about its evidence base.
Well, not so. In fact, the label implies that only two studies were ever conducted, stating that the drug “was evaluated in two eight-week randomized, double-blind, placebo-controlled, flexible-dose studies in adults.” These two studies, of course, are the positive ones. The 11 failed trials are not mentioned.
Worse, the label only includes safety data from the two positive trials.
Exxua is being sold as less likely to induce weight gain and sexual side effects than the SSRI antidepressants, but this is based on very limited data from these short-term studies and so it is unclear how much to trust this conclusion. The FDA’s main safety concern was that Exxua may cause heart arrhythmias. (Exxua is considered a selective serotonin 5HT1a receptor agonist, so it still works on the serotonin system, but through a somewhat different pathway than regular SSRIs.)
Turner writes, “Provided with this information, clinicians may understandably assume that the drug worked (and was safe) in all trials conducted.”
The problem is even larger than that, though, because the labeling information is what gets disseminated across other media: it determines how the industry can market the drug, and it gets used in drug information tools. As Turner writes, “Incomplete trial data in the labeling are important even for clinicians who do not read the labeling because they are incorporated into other widely used point-of-care tools, such as UpToDate and Micromedex. Further, the label forms the framework for allowable sponsor advertising.”
Turner writes that at the very least, the FDA must start requiring the product label to include all the data from all the efficacy trials.
“Using such information, clinicians could make better-informed prescribing decisions and convey more realistic expectations of therapeutic benefit to patients, including comparisons to other drugs. More complete and transparent labeling would also mitigate against the selective reporting of positive trials in the published literature, which would benefit systematic reviewers, clinicians, payors, and patients.”
Another solution, not mentioned by Turner, is that the FDA could be mandated to abide by the decision of its own advisory committees. When the experts vote against approval, that could be the end of it.
–
This post was originally published on Mad In America.