The FDA Hasn't Changed How It Runs Clinical Trials in 60 Years. That Ended Monday.

FDA Commissioner Marty Makary says 45% of drug development time is administrative dead time. J&J cut clinical trial report preparation from 700 hours to 15 minutes with AI. Now the FDA has launched a pilot with AstraZeneca and Amgen to stream aggregated trial signals to regulators in real time, a structural reform that could eliminate years of delay from the $2.6 billion, 10-to-15-year process of bringing a single drug to market.

Forty-five percent. That is the fraction of total drug development time that FDA Commissioner Marty Makary describes as "dead time," consumed by administrative tasks and paperwork rather than actual science. In a process that averages 10 to 15 years from discovery to approval and costs $2.6 billion per drug according to the Tufts Center for the Study of Drug Development, 45% translates to roughly 5.4 years of waiting, filing, compiling, and formatting per drug program. Not testing. Not treating patients. Filing.

On April 28, 2026, the FDA announced two proof-of-concept real-time clinical trials and an open Request for Information inviting pharmaceutical companies to join a broader pilot. It is the first structural reform to the clinical trial data pipeline since the current framework solidified in the 1960s under the Kefauver-Harris amendments, and if you want to understand why it matters, you need to see the math nobody is running on what "dead time" actually costs in human terms.

The Dead-Time Arithmetic

ClinicalTrials.gov lists approximately 48,000 active studies in the United States across all phases. Roughly 6,000 of those are interventional drug trials in active enrollment. If the 45% dead-time figure applies broadly and each trial program carries between 1 and 2 years of eliminable administrative delay between phase transitions, that represents 6,000 to 12,000 trial-years of pure bureaucratic latency distributed across the active American drug pipeline at any given moment.

For oncology, the numbers are worse, and the stakes are not abstract. Median time from Phase 1 initiation to FDA approval for cancer drugs runs approximately 7.3 years (Kola & Landis, Nature Reviews Drug Discovery). Apply Makary's 45% figure and you get 3.3 years of dead time per cancer drug program. Consider that a cancer drug treating 100,000 U.S. patients per year at a 30% to 50% response rate means that each year of administrative delay denies access to 30,000 to 50,000 patients who might have responded to that therapy. Not all of them would survive regardless. But the scale is staggering, and the cause of the delay is paperwork, not science, not safety review, not data collection. Paperwork.

What the FDA Actually Built

Two trials are now streaming aggregated safety and efficacy signals to the FDA through a technology intermediary called Paradigm Health. AstraZeneca's Phase 2 TRAVERSE trial for treatment-naïve mantle cell lymphoma, running at MD Anderson and UPenn, has already transmitted its first signals. Amgen's Phase 1b STREAM-SCLC trial for limited-stage small cell lung carcinoma is the second.

The privacy architecture matters because it determines whether this thing actually scales or collapses under its own regulatory weight. Individual patient records stay with the trial sponsor. Always. The FDA receives only aggregated signals: adverse event rates, tumor response percentages, enrollment velocity. FDA Chief AI Officer Jeremy Walsh confirmed the privacy-preserving design publicly, a decision that removes the single largest objection pharmaceutical companies have raised against real-time data sharing for the past decade.

The practical goal is to eliminate the "hiatus" between trial phases. Right now, a Phase 2 trial ends, the sponsor compiles data for months, submits an analysis, the FDA reviews the submission over additional months, and then Phase 3 design begins. Walsh says real-time monitoring could enable "continuous" trials where the regulator sees signals as they emerge and can engage in rolling dialogue with sponsors about phase transitions. Comments on the RFI close May 29, 2026, with pilot selections scheduled for August.

J&J's 99.96% Document Compression

One day before the FDA announcement, Johnson & Johnson's CIO Jim Swanson told Reuters that AI had cut clinical trial report preparation from 700 to 900 hours down to approximately 15 minutes. That is a 99.96% reduction in a process that constitutes the bulk of the document-preparation workload in clinical development.

Run the industry extrapolation, because the per-company numbers are revealing. A mid-size pharmaceutical company operates roughly 20 active trials per year, each generating approximately 10 major regulatory documents: clinical study reports, interim safety analyses, protocol amendments, and regulatory submissions. At 700 hours per document, that is 140,000 hours annually per company devoted to document preparation alone. At a loaded cost of $150 per hour for regulatory affairs professionals, each mid-size pharma company spends approximately $21 million per year writing documents that nobody reads for pleasure and that exist primarily to satisfy a bureaucratic cadence established before the internet existed. Across the top 20 pharmaceutical companies globally, aggregate document preparation costs reach approximately $420 million per year. AI eliminates 99.96% of that labor. The savings are real, but the more important point is that document preparation has been the rate-limiting step between a trial producing data and a regulator seeing that data, and removing it unblocks everything downstream.

Why China Matters Here

Commissioner Makary has been unusually direct about the competitive context. China surpassed the United States in Phase 1 clinical trial starts around 2021, and the gap has widened since, with growth in Chinese trial initiations described by Makary as "exponential." Chinese regulatory authorities operate under a fundamentally different approval framework that prioritizes speed, and the question confronting American drug development is whether the existing 60-year-old clinical trial infrastructure can sustain competitive advantage against a system that processes equivalent regulatory decisions in half the time.

Real-time trials do not lower American safety standards. They compress the dead time between safety reviews, which is a distinction that sounds subtle but is mechanically profound: the scientific rigor of Phase 2 and Phase 3 data review does not change. What changes is that the FDA stops waiting six months to see data that already exists in the sponsor's database, and sponsors stop waiting four months for a response to a submission that the regulator could have engaged with in real time.

The Strongest Counterargument

Speed has historically compromised safety in drug regulation, and the evidence is not ambiguous. Vioxx, approved in 1999 through a standard review, was withdrawn in 2004 after being linked to an estimated 88,000 to 140,000 excess heart attacks in the United States alone. The FDA's accelerated approval pathway, which was designed to speed access to drugs for serious conditions, has been criticized because 21% of accelerated-approval drugs had confirmatory trials that missed FDA deadlines. Faster access to partial data could create institutional pressure to greenlight drugs prematurely, especially when a commissioner is publicly framing delay as the enemy and competition with China as the motivation. Regulatory caution is not waste. Sometimes the hiatus between Phase 2 and Phase 3 catches signals that a real-time dashboard, tuned for speed, might flatten into statistical noise.

The FDA's design partially addresses this concern. Aggregated signals, not raw patient data, preserve the analytical layer between observation and interpretation, and the FDA is positioned as a monitor rather than a co-pilot in trial execution. Whether that separation holds under political and commercial pressure to accelerate approvals is the real test, and it will not be answered by two pilot trials with cooperative sponsors. It will be answered by the first time a real-time signal suggests a safety problem and the FDA must decide whether to act on incomplete data or wait for the full picture.

What This Analysis Does Not Prove

Makary's 45% figure has no published methodology. It may include activities that serve legitimate quality-control functions reframed as "dead time" for political effect. Until the FDA publishes a detailed breakdown of what that 45% comprises, the number should be treated as a policy argument rather than an audited finding.

J&J's 700-to-15-minute claim covers document drafting, not the intellectual analysis those documents represent. A 700-hour clinical study report includes data interpretation, anomaly investigation, and regulatory strategy decisions that cannot be meaningfully compressed to 15 minutes. The AI compresses the writing, not the thinking.

China's trial-count advantage lacks critical nuance because Chinese Phase 1 trials operate under different ethical review standards, informed consent practices, and data integrity expectations. More trials does not mean better trials, and Makary's framing of the comparison as straightforward competition omits the quality dimension entirely.

Finally, two pilot trials with blue-chip sponsors (AstraZeneca and Amgen) tell us nothing about scalability. These companies have the technical infrastructure, regulatory sophistication, and financial resources to implement real-time data streaming. The 4,800 small and mid-size biotech companies that sponsor clinical trials in the United States may not, and the gap between a successful pilot and a functional system is where most regulatory technology initiatives go to die.

The Bottom Line

If you are a patient enrolled in a clinical trial or waiting for one to open, this policy change will not affect you immediately, but the signal it sends matters enormously: the FDA is publicly acknowledging that its own processes, not safety science, are the bottleneck in drug access, and that acknowledgment is prerequisite to structural reform that could shave one to three years off development timelines within the next decade.

If you work in clinical operations at a pharmaceutical company, submit a comment on the RFI before May 29, 2026, because the pilot program selections in August will define the technical standards for real-time data sharing that the rest of the industry will eventually be required to adopt. Companies that participate in the pilot will shape the framework. Companies that wait will inherit it.

If you invest in pharmaceutical companies or biotech, the metric to watch is not revenue impact from faster approvals but rather which companies build the data infrastructure to participate in real-time trials at all, because the gap between firms that can stream clinical signals in real time and firms still running SAS scripts on locked Excel spreadsheets will become a competitive moat within five years. Walsh's office has signaled that real-time capability will eventually become the default expectation, not the exception, and companies that lack it will face longer review timelines by comparison. Rebuild your diligence checklists accordingly.