The consensus is comfortable: when a promising drug fails late in trials, we shake our heads at the waste and move on. Another billion dollars lost. Another setback for patients waiting for breakthroughs. The story writes itself, and we've all read it dozens of times.
But this framing lets us avoid a harder question. If these drugs looked so promising in earlier phases, why do they so often crater when tested at scale? And more importantly: what does the frequency of these failures tell us about the fundamental models we use to understand disease?
Recent headlines have reminded us this happens routinely. A drug targeting heart disease mechanisms we thought we understood. Treatments that worked in controlled settings but stumbled when confronted with real human biology. These aren't isolated incidents. They're part of a pattern that should be making us question our confidence in how we approach drug development itself.
The standard narrative blames complexity. Human bodies are intricate. Variables multiply. Sometimes biology surprises us. All true. But this explanation is also convenient because it requires no one to change their approach. It's bad luck dressed up as humility.
Consider what these failures might actually reveal: that our disease models are too reductive. That we've become skilled at identifying molecular targets without fully understanding the systems those targets sit within. That we've optimized for finding drugs that work in isolated conditions rather than drugs that work in the messy reality of sick people.
This matters beyond the obvious human cost. It shapes how we allocate research resources, what we teach medical students to look for, and which problems we decide are worth solving. If your framework for understanding disease is incomplete, your framework for treating it probably is too.
The blood supply crisis, the spread of drug-resistant infections, the challenges facing cancer survivors seeking fertility options after treatment: these aren't separate problems. They're symptoms of a medical system that often treats disease in silos rather than as expressions of whole-person biology. A drug that fails because it addressed one mechanism while ignoring systemic effects isn't a loss of investment. It's data. The question is whether we're willing to read it differently.
Some researchers are doing this work. They're building models that account for the interactions between systems, the ways individual variation matters, the reality that a drug's performance in a petri dish tells you almost nothing about its performance in a 67-year-old with three comorbidities. But these approaches are often slower, more expensive upfront, and require admitting uncertainty earlier in the process. They don't fit neatly into our current incentive structures.
Here's what breaks if we take this seriously: the timeline expectations around drug development would need to shift. The confidence we project about how we understand disease would need to become more measured. The way we fund research, license drugs, and structure clinical trials would all face pressure to change.
That's uncomfortable. The system as it exists has worked well enough for enough people that defending it requires only noting its successes, not interrogating its failures.
But patients waiting for treatments for conditions we claim to understand deserve better than convenient explanations. They deserve a medical research establishment willing to ask not just "why did this drug fail?" but "what about our entire approach to understanding this disease needs to be reconsidered?"
That's the question our discomfort should be pointing toward.