The Covington examples here are rightly called out as weak i.e. p-values and propensity score concerns are textbook statistics, not trade secrets. But the "roadmap" argument deserves more scrutiny in edge cases: imagine a first-in-class therapy targeting a novel mechanism where the FDA's feedback on a specific biomarker selection reflects genuine scientific uncertainty that hasn't yet entered the public literature. In that scenario, a competitor could plausibly free-ride on the pioneering company's costly trial-and-error without having to do the underlying work. The fix probably isn't confidentiality by default, but a more granular redaction standard that distinguishes routine statistical critique from feedback that reveals genuinely novel scientific judgment calls.
This is the right hill to die on. I think the harshness is earned (though I don't know if it will be effective). What strikes me about Covington's "roadmap to competitors" argument is that it quietly reclassifies a public good as a private asset. The FDA's standard for what counts as a valid endpoint, or whether a propensity-score model survives contact with selection bias, isn't proprietary science; it's infrastructure. A standard everyone can see is a road; a standard only the incumbent can see is a toll booth. The petition is really a request to keep the toll booth.
There's a lot of irony in the fact that the confidentiality claim is only plausible because the FDA mostly governs through case-by-case decisions rather than published rules. The opacity is what manufactures the proprietary feeling. Release enough CRLs and the "secret" evaporates—not because anyone leaked anything, but because the de facto rules finally become legible. That's not a loss to the applicant; it's the difference between an agency that recites its obscure standards and one that publishes them. As you say, even Covington's own client comes out ahead in a world where it can read everyone else's letters.
The roadmap objection quietly concedes the whole point: if firms can't currently see that the FDA cares about selection bias or a p-value, that opacity is the inefficiency, not a trade secret. The deeper win is for the evidence base itself. Right now the public record skews toward approvals, and the reasons drugs fail stay in a drawer, which is the same publication-bias problem Erick Turner documented in antidepressants. Releasing CRLs turns those failures into shared knowledge.
The issues that the law firm objects to are the same kinds of issues one can easily find in any medical or statistical review within an FDA drug approval package. Perhaps the release of drug approval packages isn't objected to because their release has been routine for so long (eFOIA of 1996). If statistical concerns don't constitute commercial confidential information/trade secrets in FDA reviews, there is no obvious reason why they should be confidential when they appear in CRLs.
The Covington examples here are rightly called out as weak i.e. p-values and propensity score concerns are textbook statistics, not trade secrets. But the "roadmap" argument deserves more scrutiny in edge cases: imagine a first-in-class therapy targeting a novel mechanism where the FDA's feedback on a specific biomarker selection reflects genuine scientific uncertainty that hasn't yet entered the public literature. In that scenario, a competitor could plausibly free-ride on the pioneering company's costly trial-and-error without having to do the underlying work. The fix probably isn't confidentiality by default, but a more granular redaction standard that distinguishes routine statistical critique from feedback that reveals genuinely novel scientific judgment calls.
This is the right hill to die on. I think the harshness is earned (though I don't know if it will be effective). What strikes me about Covington's "roadmap to competitors" argument is that it quietly reclassifies a public good as a private asset. The FDA's standard for what counts as a valid endpoint, or whether a propensity-score model survives contact with selection bias, isn't proprietary science; it's infrastructure. A standard everyone can see is a road; a standard only the incumbent can see is a toll booth. The petition is really a request to keep the toll booth.
There's a lot of irony in the fact that the confidentiality claim is only plausible because the FDA mostly governs through case-by-case decisions rather than published rules. The opacity is what manufactures the proprietary feeling. Release enough CRLs and the "secret" evaporates—not because anyone leaked anything, but because the de facto rules finally become legible. That's not a loss to the applicant; it's the difference between an agency that recites its obscure standards and one that publishes them. As you say, even Covington's own client comes out ahead in a world where it can read everyone else's letters.
The roadmap objection quietly concedes the whole point: if firms can't currently see that the FDA cares about selection bias or a p-value, that opacity is the inefficiency, not a trade secret. The deeper win is for the evidence base itself. Right now the public record skews toward approvals, and the reasons drugs fail stay in a drawer, which is the same publication-bias problem Erick Turner documented in antidepressants. Releasing CRLs turns those failures into shared knowledge.
The issues that the law firm objects to are the same kinds of issues one can easily find in any medical or statistical review within an FDA drug approval package. Perhaps the release of drug approval packages isn't objected to because their release has been routine for so long (eFOIA of 1996). If statistical concerns don't constitute commercial confidential information/trade secrets in FDA reviews, there is no obvious reason why they should be confidential when they appear in CRLs.