Clinical Trials Brief | May 4, 2026
This week's five signals cover FDA's launch of real-time clinical trial monitoring with AstraZeneca and Amgen, a proposed drug withdrawal after the agency discovered pivotal trial data was manipulated, a $12 billion clinical AI platform pulling out of Europe over the EU AI Act, a Nature-published trial showing how to cut checkpoint combination toxicity by 60% through delivery route engineering, and a global first for in vivo CRISPR gene editing in a Phase 3 trial.
FDA just changed the fundamental relationship between regulators and running clinical trials
What changed
On April 28, 2026, FDA announced that AstraZeneca's Phase 2 TRAVERSE trial and Amgen's Phase 1b STREAM-SCLC trial are already reporting safety and efficacy signals to the agency in real time through Paradigm Health's platform. The agency also released an RFI for a broader AI-enabled pilot program for early-phase trials, with comments due May 29, selection criteria in July, and pilot selections in August.
Why it matters
This proposes to fundamentally restructure how FDA receives clinical trial data. Instead of waiting months for sponsors to process and submit results, agency scientists will watch signals as they emerge. Commissioner Makary noted the traditional process takes 10-12 years with nearly half that time passing with no trial running. The technical feasibility is already proven.
Takeaway
This is the most significant proposed change to clinical trial infrastructure since the phased trial model was established. If it scales, it eliminates the stop-and-start pattern between phases and could cut years from the development timeline.
Sources
This item is grounded in FDA's April 28, 2026 press announcement, RAPS reporting, and Paradigm Health's collaboration announcement.
FDA proposes to withdraw Tavneos after discovering its pivotal trial data was manipulated
What changed
On April 27, 2026, FDA proposed withdrawing Tavneos, citing that ChemoCentryx personnel selected nine ADVOCATE study patients for readjudication after unblinding, including five avacopan patients whose outcomes would be changed from 'not in sustained remission' to 'sustained remission,' thereby flipping the primary endpoint result. The original non-significant analysis was never submitted to FDA. Post-marketing data also identified 76 serious liver injury cases including fatal vanishing bile duct syndrome.
Why it matters
This is not a routine safety withdrawal. The entire basis for approval, a single pivotal trial, has been declared unreliable due to deliberate data manipulation discovered through a whistleblower report more than three years after approval. FDA used unusually direct language: 'untrue statements of material fact.' The broader signal is about the integrity of the clinical trial system itself.
Takeaway
The immediate question is what happens to patients currently on Tavneos. The deeper question is systemic: how does the system detect this kind of post-database-lock manipulation earlier, and what does it mean for approvals based on single pivotal trials?
Sources
This item is grounded in FDA's April 27, 2026 CDER statement and the Federal Register notice published April 30, 2026.
OpenEvidence just pulled out of the EU and UK, and the reason is the EU AI Act
What changed
On April 30, 2026, OpenEvidence withdrew from the EU and UK markets entirely. The company, valued at $12 billion after raising $250 million in January 2026, handles 18 million clinical consultations per month in the US across 10,000 hospitals. It cited the EU AI Act's classification of clinical decision support as 'high-risk,' requiring algorithmic transparency, local clinical validation, real-time post-market surveillance, and direct manufacturer liability. The UK's lack of an equivalent statutory framework created parallel ambiguity that also drove the exit.
Why it matters
This is not a small startup retreating. A company with massive capital, clinical validation, and physician adoption has concluded that Europe's regulatory environment is too uncertain to operate in. The contrast with this week's FDA real-time trial launch is stark: while FDA actively launches AI-powered clinical trial monitoring, the EU AI Act is causing one of the world's most widely used clinical AI platforms to leave Europe. European physicians are now cut off from a tool their American counterparts use daily.
Takeaway
The policy question is now concrete: is the EU AI Act inadvertently creating a clinical AI gap where European healthcare loses access to tools available elsewhere? The answer depends on whether the framework can be made workable before the August 2 full enforcement date, or whether more companies reach the same conclusion.
Sources
This item is grounded in multiple reports from April 30 - May 1, 2026 covering OpenEvidence's withdrawal from the EU and UK markets.
A European-led trial just showed how to keep checkpoint combination efficacy while cutting its worst toxicity by 60%
What changed
The NIVIPIT trial randomized 61 patients with untreated metastatic melanoma to intratumoral versus intravenous anti-CTLA4, both combined with intravenous anti-PD1. The intratumoral arm showed grade 3-4 treatment-related adverse events in only 22.6% of patients versus 57.1% in the intravenous arm, equivalent to anti-PD1 monotherapy. Response rates on injected lesions were 65.7% including 31.4% complete responses, with concordant responses in uninjected distant lesions. Median follow-up was 55.5 months.
Why it matters
Checkpoint combinations are among the most effective cancer immunotherapies but severe toxicity limits their use. This trial suggests the toxicity problem is solvable through delivery route engineering rather than mechanism abandonment. By concentrating anti-CTLA4 locally while keeping systemic anti-PD1, comparable efficacy was achieved with dramatically less systemic toxicity. This could expand the eligible patient population for combination immunotherapy.
Takeaway
The constructive safety signal here is that class-level toxicity problems may sometimes be solved by how a drug is delivered, not whether it should be used at all. This kind of evidence could reshape how regulators think about the risk-benefit balance for combination immunotherapy.
Sources
This item is grounded in the NIVIPIT trial published in Nature in 2026, a randomized study conducted at French oncology centers including Institut Gustave Roussy.
Intellia reports the first positive Phase 3 for in vivo CRISPR gene editing, and files for approval
What changed
On April 27, 2026, Intellia Therapeutics announced positive Phase 3 HAELO results for lonvo-z in hereditary angioedema. A one-time 50mg infusion reduced attacks by 87% versus placebo (mean monthly rate 0.26 vs 2.10, p<0.0001), with 62% of patients entirely attack-free and therapy-free versus 11% on placebo. All treated patients remained off long-term prophylaxis. Safety showed only mild-to-moderate adverse events with no serious events in the treatment arm. Intellia initiated a rolling BLA and is preparing additional regulatory filings for a potential US launch in H1 2027.
Why it matters
This is a historical milestone: the first time any in vivo CRISPR gene editing therapy has demonstrated Phase 3 efficacy. While ex vivo gene editing already has approved products, in vivo editing, where genes are corrected directly inside a living patient, has never before reached this level of clinical proof. The single-dose, outpatient-setting design and the durability of the effect validate the concept of permanent genetic correction through a one-time infusion.
Takeaway
The clinical development of in vivo gene editing just moved from promise to Phase 3 proof. Regulatory, commercial, and reimbursement infrastructure for one-time genetic interventions is now being built in real time.
Sources
This item is grounded in Intellia Therapeutics' April 27, 2026 press release announcing positive Phase 3 HAELO topline results and rolling BLA initiation.