3 rejections or delays on Hunter syndrome treatments in under 8 months. Advocates staging a funeral at FDA headquarters. Parents saying the agency kills children every time it refuses accelerated approval. I understand the rage. I also think the FDA's new Plausible Mechanism Framework, which lets drugs for ultra-rare diseases win approval based on biomarkers, target engagement, and natural history controls instead of randomized controlled trials, is a scientifically premature concession that will hurt exactly the patients it's designed to help.
Let me grant the strongest version of the opposing case: demanding a traditional RCT from a population of 200 or 500 patients is often logistically impossible. Kai Brighton and the advocacy community are right about that constraint. It's real. But acknowledging a limitation of RCTs is not the same as validating whatever replaces them, and the replacement on offer has a track record that should make anyone cautious.
Biomarkers Are Hypotheses, Not Outcomes
The framework allows approval when a drug demonstrates a "plausible mechanism of action" plus biomarker changes that predict clinical benefit. The word "predict" is doing catastrophic work in that sentence. Biomarkers fail as surrogates constantly. A 2022 analysis in JAMA Internal Medicine (n=253 accelerated approvals, 1992 to 2021) found that 43% of drugs approved on surrogate endpoints had no confirmatory trial completed 5 years later. That's not a rare failure mode. That's close to a coin flip.
The most famous example remains the cardiac arrhythmia saga. Encainide and flecainide suppressed premature ventricular contractions, the biomarker. They also tripled cardiac death rates. The CAST trial (NEJM, 1991, n=1,498) killed the drugs and the assumption behind them. Biomarker correction looked like clinical benefit. It wasn't.
Ultra-rare diseases involve gene therapies and cell therapies with novel mechanisms, long latency periods for adverse effects, and almost no historical safety data to anchor expectations. These are precisely the conditions where you most need clinical outcome data, not the conditions where you should waive it.
The Post-Market Problem Nobody Wants to Discuss
Approval is not the end of the story. It's the beginning of pharmacovigilance. For a drug treating 300 patients nationally, who is running the post-market surveillance? FDA's adverse event reporting system, FAERS, depends on volume. A serious signal in a population of 300 can hide for years. A 2020 Drug Safety study (n=95 orphan drugs) found that the median time to detect a post-market safety signal for orphan drugs was 4.2 years, compared to 2.8 for non-orphan drugs. Smaller populations mean slower signals.
Paul Melmeyer of the Muscular Dystrophy Association flagged this obliquely when he raised concerns about "removal of staff with the regulatory expertise" needed for consistent implementation. He's being polite. The actual problem is structural: the FDA is building an approval pathway that demands sophisticated post-market follow-up while simultaneously losing the personnel to execute it.
The asymmetry here matters to me professionally and personally. Approving a therapy that doesn't work costs a family years of false hope and, for gene therapies, potentially forecloses future treatment options because the immune system has been primed. Approving one that actively harms costs everything. The downside risk for a patient with a fatal rare disease is not zero just because their baseline prognosis is terrible.
Commissioner Makary says the agency wants to "remove barriers and exercise regulatory flexibility." Vinay Prasad calls the framework "a revolutionary advance in regulatory science." I'd feel better about both statements if either man could point to a validated surrogate endpoint framework for any of the specific diseases in the pipeline right now. They can't, because those validation studies haven't been done.
The Denali Hunter syndrome decision arrives April 5. I hope the drug works. I hope the biomarker data are right. But hope is not a regulatory standard, and the families of children with MPS II deserve a system that catches failure before it compounds into tragedy. Speed without surveillance is not compassion. It's abandonment with better branding.
