A dataset linking 45,988 finished phase 2–3 drug trials on ClinicalTrials.gov (2010–2026) to subsequent FDA approval of the drugs tested
This paper presents a comprehensive dataset linking 45,988 completed phase 2–3 drug trials from ClinicalTrials.gov (2010–2026) to subsequent FDA approval outcomes, providing critical data for benchmarking trial design and analyzing drug attrition rates.
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
The journey of a new medicine is a long, winding road filled with checkpoints. Before a drug can reach a pharmacy shelf, it must pass through rigorous tests on human volunteers known as clinical trials. These studies are registered in a public database called ClinicalTrials.gov, where scientists and companies post details about what they are testing, who they are testing it on, and when they plan to finish. For years, this public record has told the story of the trial itself, but it has remained silent on the most important question of all: did the drug actually make it to the market? While the registry knows if a study finished or stopped early, it does not record whether the medicine being tested eventually received official permission from the US Food and Drug Administration to be sold. This missing link has made it difficult for researchers to understand the true success rates of drug development or to see which trial designs are most likely to lead to a finished product.
To bridge this gap, an independent researcher named Lars Kullman has created a massive new dataset that connects the dots between these public trial records and the official FDA approval lists. By matching the names of the drugs tested in nearly 46,000 finished trials against the government's public records of approved medicines, Kullman has built a clear picture of what happened to the drugs after the trials ended. The work covers every interventional drug trial in phases two and three that started in 2010 or later and had finished by October 2026. This effort transforms a scattered collection of trial reports into a single, searchable history, allowing anyone to see not just how a trial was run, but whether the drug inside it eventually became an approved treatment.
The resulting dataset is a comprehensive map of 45,988 trials, ranging from small studies to large, multi-country investigations. It includes details on who sponsored the research, whether the drug was a small molecule, an antibody, or a more complex biological product, and the specific conditions the drug was meant to treat. Kullman's work goes a step further by classifying the final fate of the drugs tested. He distinguishes between trials that tested brand-new drugs and those that tested old drugs for new uses. For the new drugs, the dataset tracks whether they eventually received FDA approval for any use, and specifically whether they were approved for the exact condition the trial was studying. This distinction is crucial, as a drug might succeed in getting to market for one disease but fail to get approval for the specific illness a particular trial was investigating.
The numbers reveal a landscape of both high ambition and significant attrition. Of the trials that tested a novel drug and had enough time to reach a conclusion, roughly one in five led to an approval for some use of that drug. However, when looking strictly at whether the drug was approved for the specific condition the trial was designed to test, the success rate drops to about one in nine. The data shows that the stage of the trial matters greatly; phase three trials, which are the final large-scale tests before a drug can be sold, had a much higher chance of leading to an approval for any use compared to earlier phase two studies. Industry-sponsored trials also showed higher success rates than those led by universities or government groups. Interestingly, even when a trial was stopped early or labeled as a failure, the drug tested in it still went on to receive approval in about one out of every six cases, often because it succeeded in a different trial or for a different medical condition.
The timeline of these successes offers another layer of insight. For the drugs that did get approved, the median time from the end of the trial to the final FDA approval was about ten months. This suggests that the regulatory process, while thorough, often moves relatively quickly once a trial is complete. However, the data also highlights a significant number of trials that finished too recently to have a known outcome yet, as the process of approval can sometimes take years. The researcher also noted that for many approved drugs, the first approval actually happened before the trial even finished, usually because the drug was already being tested or approved for a different disease. This reality complicates the simple idea that a trial's success is solely determined by its own results, as a drug's fate is often tied to its performance across the entire development program.
This work is not without its limitations, and the researcher is careful to note where the data might be incomplete. The matching process relies on drug names, which means that drugs known only by code names or those approved by regulators outside the United States are not captured in this specific dataset. Additionally, the data reflects the information as it was entered by the trial sponsors, which can sometimes contain errors or placeholders. Despite these constraints, the dataset provides a rare, open window into the full lifecycle of drug development. It allows scientists, policymakers, and the public to move beyond guessing about success rates and instead rely on a concrete, auditable record of what has actually happened. By linking the public record of the trial to the public record of the approval, this dataset turns a fragmented history into a clear narrative, showing exactly how many new medicines make the journey from the testing ground to the pharmacy.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.