The Food and Drug Administration convenes advisory committees to take public votes on whether a drug’s benefits outweigh its risks, whether a label should change, and whether additional studies should be required. The committees are not the final decision-makers. Agency officials retain the authority to approve, reject, or delay. The record still shows that the votes are treated as a public test of the science and of the agency’s appetite for dissent.

Members of those committees are often academic clinicians, statisticians, and patient representatives. A subset also maintains consulting relationships with manufacturers, contract research organizations, or investment firms that cover the same therapeutic areas. When a vote concerns a product or a class in which a member has a financial interest, FDA conflict rules require disclosure and, in many cases, recusal. Recusal means the member may not vote. In some instances the member may not even sit in the room for the discussion.

The agency publishes, for many meetings, a public roster and a brief statement of financial interests and waivers. Those notices are the principal public record of who was asked to step aside. They do not always name the sponsor, the dollar amount, or the precise consulting contract. They do name the fact of recusal and, where a waiver was granted, the agency’s judgment that the need for the member’s expertise outweighed the appearance of a conflict.

Inspector-general and Government Accountability Office reviews over more than a decade have described the same pattern: incomplete or late disclosures, waivers granted for members with substantial industry income, and recusals that remove a voter but leave the rest of the panel intact. The reviews have not alleged a single conspiracy. They have alleged gaps in screening, uneven enforcement across centers, and a reliance on self-reporting.

According to public meeting materials, recusals cluster around high-stakes votes: oncology products with accelerated pathways, opioids and other controlled substances, rare-disease therapies with small trial populations, and follow-on products that would compete with a consultant’s existing clients. When several members recuse from the same session, the remaining quorum can be small. A close vote then rests on fewer independent voices.

FDA officials have said, in congressional testimony and in Federal Register notices, that the agency needs specialists who have seen the relevant data in practice, and that many such specialists have some industry contact. The public position is that recusal plus waiver is the lawful compromise. Critics in patient groups and in some academic societies have said the compromise still leaves the vote downstream of consulting markets.

The consulting itself is not hidden in every case. Open Payments data from the Centers for Medicare and Medicaid Services, company 8-K filings, and university conflict-of-interest websites often list the same names that appear on advisory rosters. The overlap is not proof that a recused member influenced a vote from outside the room. It is evidence that the labor market for drug expertise is shared between the regulator’s advisors and the firms whose products are under review.

Dockets for individual meetings show the mechanics. A member discloses a paid advisory board for a competitor. The agency recuses the member from a vote on a new molecular entity in the same class. Another member discloses equity in a small biotech. Recusal follows. A third member discloses speaking fees below a threshold; a waiver is granted and the member votes. The public sees the outcome of the vote, not the internal memo that weighed the waiver.

Procurement of advisory labor is not a contract in the ordinary sense. Members receive a modest honorarium and travel. The more valuable compensation, according to the same public payment databases, often arrives through separate consulting retainers, expert-witness work, and stock in companies that will rise or fall with a class of drugs. Recusal severs the vote. It does not sever the career.

Congressional committees have periodically asked the agency for counts: how many recusals per year, how many waivers, how often a recused member’s institution still received research funding from the sponsor. The answers, when provided, have been aggregate. They have not included a named mapping of every consultant to every vote. The dockets remain the more granular source, meeting by meeting.

The Centers for Drug Evaluation and Research and for Biologics Evaluation and Research run most of the product-specific votes. Device panels under the Center for Devices and Radiological Health follow a similar recusal template. Across centers, the public language is cautious: “potential conflict,” “appearance of a conflict,” “waived.” The verbs match the legal standard, not a finding of corruption.

When a recusal is announced at the opening of a session, the remaining members proceed. Transcripts show chairs reminding the room that recused members will not participate. They do not show whether the recused expert later briefed a sponsor, a hedge fund, or a trade association. Those later contacts, if they occur, fall outside the meeting docket.

The policy tension is structural. The United States concentrates drug-trial expertise in a relatively small set of academic medical centers and contract research networks. Manufacturers hire from the same set. The FDA, seeking votes that look scientifically current, draws from the same set. Recusal is the official filter. Public budgets for ethics staff and for conflict screening have been described in appropriations justifications as modest relative to the volume of meetings.

Nothing in the public record requires a reader to conclude that recusal is a sham. The record does show that recusal is frequent, that waivers remain available, and that the vote that follows is still cited in approval letters, in label negotiations, and in later litigation over whether the agency ignored a signal. The consultants who recuse are, by design, not on the tally. The consultants who remain, or who return at the next meeting on a different product, continue to occupy the chairs.

Subsequent meetings recycle many of the same names. A recusal on one sponsor’s application does not bar service on another. The revolving pattern is visible in successive Federal Register notices. It is not, by itself, a violation. It is the labor market the recusal rules were written to manage.

For reporters and for the public, the usable documents are the ones already posted: conflict waivers, meeting minutes, Open Payments, and IG summaries. They do not establish a secret vote. They establish a public procedure in which pharmaceutical consulting is treated as a recusable interest rather than a disqualifying one, and in which the remaining advisory votes still help determine what reaches the market.