The FDA has the antidepressant trial data. It won't publish it.

Two decades after the original Kirsch meta-analysis put a question mark next to the antidepressant literature, the cleanest answer to that question is sitting on a server in Silver Spring, Maryland, and the agency that owns it has decided the public does not need to see it.

The FDA holds the patient-level datasets behind every approved selective serotonin reuptake inhibitor on the American market. Those datasets are the raw material of efficacy: who improved, by how much, on what dose, against what comparator, for how long. They are also the only place where an independent reviewer can test whether the published summary of a trial matches what the trial actually found. By the agency's own rule book, that material is disclosable. In practice, it is the rare request that gets past the redactions.

What Kirsch actually said, and what he didn't

Irving Kirsch's 2008 re-analysis, published in the open-access journal PLOS Medicine, pulled together the data the FDA had made available at the time on behalf of regulators. The headline number was uncomfortable: mean effect size on the Hamilton Depression Rating Scale, in the trials the agency had reviewed for approval, was below the threshold most guidelines treat as clinically meaningful. The paper did not say SSRIs are sugar pills. It said the average benefit was small in the population studied, that the distribution of response was skewed, and that the trials themselves systematically excluded the very patients most likely to be prescribed the drug in routine practice.

What Kirsch could not do, then or now, is run the same arithmetic on the trials the FDA has approved since. The 2008 paper used the data the agency had released under its then-prevailing policy. Modern SSRIs are approved on the basis of trials for which the FDA holds primary datasets whose disclosure status is decided case by case, under a framework designed to protect sponsor interests rather than to facilitate replication.

The disclosure question, and why the regulator defaults to closed

The statutory answer to "can I see this" is: usually yes, with conditions. The FDA Amendments Act of 2007 expanded the kinds of data the agency can release and clarified that Clinical Study Reports, the structured documents trial sponsors submit to regulators, are releasable in response to a Freedom of Information Act request. The Center for Drug Evaluation and Research, the division inside the FDA that reviews psychiatric drugs, processes those requests, often heavily redacted, sometimes denied outright on commercial-confidentiality grounds.

Compare this with the European Medicines Agency, which in 2014 began publishing the full Clinical Study Reports for drugs it had approved, with only narrowly tailored redactions for genuinely personal data. The EMA's transparency policy was the result of a deliberate institutional choice, fought for by civil society groups and academic researchers and partially walked back under sponsor pressure in 2018. The FDA made the opposite choice, and made it earlier, and never reversed it.

What the wire is reporting, and what it isn't

Pharmaceutical coverage on the business pages treats antidepressant efficacy as a settled, occasionally contested, scientific question. The standard story: yes, they work; here are the effect sizes; here is the debate about magnitude. That story is true at the level of published summaries. It is incomplete at the level of data. The summaries come from the sponsors; the data come from the regulator; and the regulator's default is silence.

This is the gap the original draft distinguished itself by foregrounding. Kirsch's finding is not the story. The story is why his successor has not been able to reproduce the analysis on the next generation of trials.

The procedural mechanics of an FOIA request for a CSR

A researcher who wants the raw data files for an FDA-approved SSRI files a request under the Freedom of Information Act, naming the application number, sometimes the trial identifiers, and the documents sought. The request is routed to CDER. CDER's disclosure office reviews each line of the Clinical Study Report against two statutory exemptions: exemption 4, which protects trade secrets and confidential commercial information, and exemption 6, which protects personal privacy.

In practice, exemption 4 does most of the work. Sponsors argue, credibly in some respects, that integrated summaries and certain analytic datasets reveal commercial strategy. The agency's default position has been to accept that argument generously. The result, for antidepressants, is that the patient-level listings and full protocols that would let a meta-analyst reproduce the company's analysis are routinely redacted to the point of uselessness, while the regulator keeps the unredacted originals on a server accessible only to its own reviewers.

What a transparency default would actually change

The argument for disclosure is not that SSRIs are unsafe or that the published trials are fraudulent. The argument is methodological, and it has nothing to do with the marketing question. A regulator that releases the data it relied on invites independent replication. Independent replication, when it confirms a finding, closes the loop of public confidence. When it doesn't, the public learns early and adjusts. Either way, the regulator's authority is reinforced, not eroded, because the authority rests on the auditability of its decisions.

The Denmark population-level data referenced in recent weeks, comparing treated and untreated depression outcomes at national scale, is the kind of evidence base that can only complement a transparent regulator. It cannot substitute for one. Population registries answer questions about real-world use; the FDA's file answers questions about what the trials actually showed, as opposed to what their summaries said they showed. Both are needed. Only one is in hand.

What to watch

The next two years will bring at least three pressure points. First, the ongoing federal litigation over the FDA's interpretation of exemption 4, which has the potential to force broader release of CSRs across therapeutic areas if the plaintiffs prevail. Second, the agency's own reassessment of its 2014 release policy, frozen in place since 2018 and subject to renewed pressure from academic researchers through the National Academies' standing committee on emerging science. Third, the next generation of antidepressant approvals, esketamine's successors and the 5-HT2A-targeted compounds now in pivotal trials, will arrive on the market carrying the same disclosure posture as their predecessors unless the policy changes.

The agency does not need new statutory authority to release the data. It needs a policy choice, made by political leadership at HHS, that treats clinical trial transparency as the rule and commercial confidentiality as the narrow exception. That choice has been on the table, in various forms, since at least 2007. It has not been made.