Comparable endpoint decisions in accelerated approval: a public docket and a duty to reconcile
This paper argues that the FDA's accelerated approval process lacks a structured mechanism for comparing and reconciling endpoint decisions across different applications, and proposes establishing a public docket of endpoint acceptances and refusals alongside an independent annual reconciliation review to address this architectural deficit without impeding treatment access.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
When a new medicine is tested, the gold standard for approval is usually a simple, hard fact: did the patients live longer, or did they feel better? For many serious illnesses, however, waiting for that final answer takes years, and patients may not have that long. To solve this, regulators have a special pathway called accelerated approval. This allows a drug to reach the market earlier if it changes a measurable sign in the body—like a blood test result or an image on a scan—that scientists believe is reasonably likely to predict that the patient will actually benefit. This sign is called a surrogate endpoint. The system relies on a crucial judgment: deciding that this specific marker, in this specific disease, is a trustworthy stand-in for the real outcome. Without this judgment, the shortcut cannot exist. But because every disease and every drug is different, making this judgment is a complex, case-by-case task that happens behind closed doors, leaving the public to wonder if the rules are applied fairly or consistently.
A researcher named Phani Kurada set out to see how this judgment is actually made and recorded. The study did not look at whether the drugs worked or failed, but rather at the paperwork itself. The author examined the detailed review documents for three recent drugs that received this accelerated approval. These cases were chosen to represent different situations: one involved a rare disease with no other treatments, another used a new type of marker for a fatal condition, and the third involved a marker that had been used successfully before. The goal was to see if the agency, the Food and Drug Administration, recorded its reasoning in a way that allowed the public to compare one decision with another.
The investigation found that while the agency does write down its reasoning, the records are not built for comparison. In each of the three cases, the reviewers explained why they accepted or rejected a specific marker. They discussed the biological logic, the strength of the evidence, and the level of uncertainty they were willing to accept. However, these explanations were buried in different parts of different documents, written in different styles, and organized in unique ways. Most importantly, none of the reviews explicitly compared its decision to a previous decision on a similar marker made by the agency in a different drug application. Even when a marker had been used before, the current review did not say, "We are using the same standard as we did last time," or "We are accepting this with more uncertainty because the situation is different." The reasoning was there, but it was isolated. It was like having three separate maps of different cities, each drawn by a different cartographer, with no legend to show how the scales matched up.
The paper argues that this is not a problem of hidden secrets or missing evidence, but a flaw in the system's design. The agency publishes its decisions, but it does not publish them in a format that lets anyone see if the standards are consistent. The author notes that the agency has a council created by Congress to ensure these decisions are consistent, but that council only reports on general topics like "recent approvals" or "ongoing studies." It does not publish the actual reasoning behind specific endpoint decisions, so the public cannot check if the same logic was applied to different drugs. The study also points out that the current system has a strange imbalance: if a drug fails later and needs to be removed from the market, there is a formal, public process with hearings and written explanations. But when a drug is first approved based on a surrogate marker, there is no such public record of the specific judgment that allowed it.
To fix this, the paper proposes a new, structured public record. This would be a simple, standardized list where the agency records every time it accepts or rejects a surrogate endpoint. For each entry, the agency would state the marker, the disease, the evidence used, and why it decided to accept or reject it. Crucially, this record would also require the agency to explain how this decision relates to past decisions. If the agency accepts a marker today that it rejected five years ago, or accepts it with less evidence than before, the record would force a public explanation of why the standard changed. This would not stop any drug from being approved or delay treatment for patients. Instead, it would simply require the agency to show its work, much like a judge writing an opinion that can be compared to previous rulings.
The author suggests that this transparency would not require a rigid set of rules that might block good drugs. Instead, it would create a system where the agency must answer for its choices. If the reasoning is sound, the public can see it. If the reasoning is inconsistent, the inconsistency becomes visible and must be explained. The study concludes that the missing piece in the current system is not the evidence itself, but the ability to compare decisions side by side. By creating a public docket where these judgments are recorded in a common language, the system could maintain its flexibility to help sick patients while ensuring that the trust placed in these shortcuts is earned through consistent, visible reasoning.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.