Clinical Trials Brief | May 11, 2026
This week's five signals cover FDA's parallel overhaul of both trial monitoring and internal review infrastructure through AI, a rare cancer drug reconsideration that tests where the evidence bar is moving, new detail on the practical fallout of OpenEvidence's EU exit, a joint FDA-industry rethink of how safety data is collected in trials, and a UK-led colorectal cancer trial reporting zero relapses at nearly three years with immunotherapy before surgery.
FDA is rebuilding both how it receives trial data and how it reviews it, at the same time
What changed
On April 28, 2026, FDA announced that two proof-of-concept real-time clinical trials are already running: AstraZeneca's Phase 2 TRAVERSE in mantle cell lymphoma and Amgen's Phase 1b STREAM-SCLC in small cell lung carcinoma, with real-time safety and efficacy signals flowing to the agency through Paradigm Health's platform. FDA estimates the approach could reduce trial duration by 20-40% and save at least $120 million annually. Then on May 6, FDA launched Elsa 4.0, an upgraded internal AI tool with custom agents, document generation, quantitative analysis, and OCR, available to all FDA staff. The same day, the agency completed consolidation of more than 40 disparate data systems into HALO (Harmonized AI & Lifecycle Operations for Data), which integrates directly with Elsa so reviewers can query submission data without manually uploading documents.
Why it matters
These are not two separate stories. They are the external and internal halves of the same transformation. Real-time trial monitoring changes how data flows from sponsors and sites to the agency. Elsa and HALO change how the agency processes, queries, and acts on that data once it arrives. Together, they amount to the most comprehensive modernization of FDA's operating infrastructure in decades. Commissioner Makary noted the traditional process takes 10-12 years with nearly half that time idle. Chief AI Officer Jeremy Walsh framed the goal directly: the integration of Elsa and HALO will let FDA 'rapidly advance regulatory science and deliver more cures and meaningful treatments to patients faster.'
Takeaway
FDA is not experimenting with AI at the edges. It is replacing core infrastructure on both the trial-monitoring and the review side simultaneously. Sponsors should watch the real-time trial RFI (comments due May 29) and prepare for a regulatory environment where the agency can see your data, query it, and act on it faster than ever before.
Sources
This item is grounded in FDA's April 28, 2026 real-time clinical trials press announcement, the May 6, 2026 Elsa 4.0 and HALO press announcement, and RAPS reporting.
FDA agrees to reconsider a rare cancer drug it rejected in January, testing where the evidence bar is moving
What changed
On May 7, 2026, STAT News reported that FDA agreed to reconsider its January rejection of Ebvallo, a treatment for rare EBV-associated cancers. The reversal came after the manufacturer and the agency reached agreement that the single-arm clinical trial data submitted could be sufficient to support review and potential approval. The original rejection had been a surprise to the rare disease community given the severity of the condition and the limited treatment options available.
Why it matters
This matters beyond one drug because it tests a broader 2026 pattern: FDA has been signaling greater flexibility on evidence standards through single-trial approval pathways, mechanism-based reviews for ultra-rare diseases, and Bayesian methods. The Ebvallo reconsideration puts that flexibility to a concrete test. If the agency approves a rare cancer drug on single-arm data after initially rejecting it, the signal to sponsors developing rare disease therapies is clear: bring strong single-arm evidence and be prepared to engage the agency on why it should be sufficient.
Takeaway
The evidence bar in rare disease is not being lowered. It is being reshaped around what kind of evidence is genuinely decision-relevant when patient populations are small and alternatives are absent.
Sources
This item is grounded in STAT News reporting from May 7, 2026 on FDA's decision to reconsider Ebvallo.
The practical fallout from OpenEvidence's EU exit is already visible, and August 2 is approaching
What changed
Since OpenEvidence withdrew from the EU and UK on April 30, 2026, citing the EU AI Act's high-risk classification, reports have surfaced of European physicians resorting to VPNs to maintain access to the platform. This raises data governance and compliance concerns in hospital and clinical research settings. Meanwhile, the EU AI Act becomes fully enforceable on August 2, 2026, meaning all clinical decision support tools operating in Europe must comply with transparency, validation, and liability requirements by that date. The PRAC met May 4-7 with no new safety referrals started or concluded.
Why it matters
The OpenEvidence story is no longer just about one company leaving. It is about what happens to the clinical AI ecosystem in Europe as the August 2 deadline approaches. If a $12 billion platform backed by the NEJM and AMA cannot justify operating under the EU AI Act, smaller companies with fewer resources will face the same calculation. The VPN workarounds show that physician demand for these tools has not disappeared; it has simply moved underground, creating a regulatory gap that is harder to govern than the one the AI Act was designed to close.
Takeaway
The next three months will determine whether the EU AI Act's approach to clinical AI drives better governance or simply drives clinical AI tools out of Europe. The August 2 deadline makes this a live regulatory countdown, not an abstract policy debate.
Sources
This item builds on OpenEvidence's April 30 withdrawal and subsequent reporting on physician workarounds, combined with the approaching August 2, 2026 EU AI Act full enforcement date.
FDA and 20+ pharma companies are rethinking what safety data actually needs to be collected in trials
What changed
On May 5, 2026, TransCelerate BioPharma and FDA published a summary report from a joint tabletop exercise conducted in late 2025. The exercise brought together more than 20 pharmaceutical R&D leaders and 15+ FDA representatives from CDER and the Center for Clinical Trial Innovation to test selective safety data collection (SSDC) approaches across hypothetical scenarios in obesity, cardiovascular, pulmonology, and dermatology. SSDC is a risk-based approach aligned with ICH guidance that allows trials to focus safety data collection on the risks that matter most rather than capturing everything by default.
Why it matters
The default in clinical trials has been to collect comprehensive safety data on every patient, regardless of whether the data is decision-relevant. This drives up trial cost, site burden, and patient burden without necessarily improving safety decisions. If FDA and industry can agree on when selective safety data collection is appropriate, it removes one of the biggest practical barriers to pragmatic trial designs that enroll broader patient populations in more real-world settings. The fact that this was a joint exercise with 20+ companies and 15+ FDA staff signals that both sides are ready to move past theoretical discussion.
Takeaway
This is a safety story about safety methodology itself. If trials can be designed to collect the safety data that actually matters rather than everything that can possibly be recorded, the downstream effects touch cost, feasibility, patient access, and the scalability of pragmatic trial designs.
Sources
This item is grounded in TransCelerate's May 5, 2026 press release and the published summary report from the joint FDA tabletop exercise.
A UK-led colorectal cancer trial reports zero relapses at nearly three years with immunotherapy before surgery
What changed
Results published May 5, 2026 from the UK-led NEOPRISM-CRC trial showed that patients with stage 2-3 mismatch-repair-deficient colorectal cancer who received nine weeks of pembrolizumab before surgery had zero relapses at nearly three years of follow-up. Under the current standard of care, approximately 25% of similar patients relapse after surgery plus adjuvant chemotherapy. The trial also developed personalized circulating tumor DNA blood tests that could predict treatment response, potentially allowing clinicians to identify which patients are most likely to benefit from the neoadjuvant immunotherapy approach.
Why it matters
Zero relapses in a cancer trial with nearly three years of follow-up is an exceptional result. Colorectal cancer is the third most common cancer worldwide, so even a finding in a biomarker-defined subset has enormous population-level implications. The neoadjuvant design, giving immunotherapy before surgery instead of chemotherapy after, could fundamentally change the treatment sequence for this patient group. The companion ctDNA biomarker work adds a precision medicine layer that makes the approach more practically deployable.
Takeaway
This is the kind of trial result that can reset a standard of care. If validated in larger trials, neoadjuvant immunotherapy could replace post-surgical chemotherapy for mismatch-repair-deficient colorectal cancer, eliminating both the toxicity of chemo and the recurrence it fails to prevent in one in four patients.
Sources
This item is grounded in ScienceDaily's May 5, 2026 report on the NEOPRISM-CRC trial results.