Nothing will replace the fact that clinical trials must happen in actual patients
Every medicine that reaches a patient has survived a decade of structured scrutiny designed not to slow progress, but to protect life. In Puerto Rico and across Latin America, experts like pediatrician Dr. Nicolás Rosario are explaining how a four-phase clinical trial system—now augmented by artificial intelligence—remains the only trustworthy path from promising molecule to proven treatment. The process is long because human biology is complex, and the stakes of getting it wrong are measured not in time lost, but in lives.
- A single drug approval demands roughly ten years and four escalating phases of human trials, from a handful of volunteers to thousands of international participants—a timeline that reflects necessity, not inefficiency.
- Artificial intelligence is accelerating the earliest stages of drug discovery, scanning molecular data at speeds no human team could match, yet scientists warn that no algorithm can substitute for observing what a drug actually does inside a living person.
- Promising new therapies for cholesterol, lupus, rheumatoid arthritis, and weight management are advancing through the pipeline, alongside a dengue vaccine candidate designed to neutralize all four viral strains in a single dose.
- A persistent structural gap shadows the entire system: Hispanic and Latin American populations remain underrepresented in global trials, meaning treatments validated elsewhere may not behave identically in the bodies of those most affected by diseases like dengue.
Every pill on a pharmacy shelf has already survived roughly a decade of methodical testing before a regulator approved it. Dr. Nicolás Rosario, a pediatrician and president of the Clinical Research Investigator Group, explains why that timeline is not bureaucratic excess—it is the only reliable mechanism humanity has for knowing whether a drug is safe and whether it actually works.
The journey begins in the laboratory, where researchers identify chemical compounds with therapeutic potential. Only after extensive preliminary work do these candidates enter human testing, structured across four distinct phases. Phase one asks a small group of volunteers to help determine safety and correct dosage. Phase two expands the group and begins measuring effectiveness. Phase three scales to thousands of participants across multiple countries, stress-testing results against the full complexity of real human diversity. A fourth phase continues after approval, monitoring long-term effects in the broader population.
Artificial intelligence is beginning to reshape the earliest part of this process. AI systems can scan enormous molecular datasets, flag promising compounds, and reduce both the time and cost of reaching human trials. But Rosario draws a firm line: technology accelerates the search for candidates; it cannot replace the act of giving a drug to a person and observing what follows. The clinical trial remains the gold standard precisely because no model can fully simulate human biology.
Several treatments are currently advancing through this pipeline—oral cholesterol medications, therapies targeting autoimmune conditions like lupus and rheumatoid arthritis, and diabetes-derived compounds showing unexpected promise for weight management. A dengue vaccine designed to protect against all four serotypes simultaneously represents a particularly significant development for a region where the disease remains endemic.
Rosario also named a structural flaw the system has yet to correct: Hispanic and Latin American populations are consistently underrepresented in international trials. Because genetic variation can alter how a drug behaves, treatments validated in one population may not translate reliably to another. The science is sound, but it remains incomplete until the full diversity of humanity is reflected in the research meant to serve it.
Every medication that reaches a pharmacy shelf has spent roughly a decade in the laboratory and in human trials before a regulator ever signed off on it. The journey from a promising compound to a pill in your hand is methodical, expensive, and designed to catch problems before they harm patients. Dr. Nicolás Rosario, a pediatrician and president of the Clinical Research Investigator Group, walks through how this system actually works—and why it takes so long.
The process begins not with patients but with molecules. Researchers identify chemical compounds in the lab that show potential to treat disease. Only after extensive preliminary work do these candidates move into human testing. That human phase, Rosario explained, typically stretches across a decade. It is not bureaucratic delay. It is the only reliable way to know if something is safe and whether it actually works.
The clinical trial system is built in four distinct phases, each with a specific job. Phase one takes a small group of volunteers and focuses on safety—does this drug hurt people, and what is the right dose? Phase two expands slightly and begins measuring whether the treatment actually does what it is supposed to do. Phase three is the big one: thousands of participants across multiple countries, all receiving the same treatment, all being monitored to confirm that the results hold up in the real world. After all three phases succeed, a fourth phase continues after approval, tracking long-term effects in the general population.
Rosario was clear about why this matters. "The only way to be certain a new medication is safe and effective is through clinical trials," he said. There is no shortcut. No computer model, no matter how sophisticated, can replace the act of giving a drug to a human being and watching what happens.
That said, artificial intelligence is beginning to reshape the early stages of drug discovery. AI systems can sift through vast datasets to identify promising molecular candidates, predict which compounds might cause problems, and cut both the time and cost of getting to human trials. But Rosario was careful to note the limits: AI will speed up the hunt for new molecules, but it will never replace the need for people to actually take the drug and report what they experience. The technology is a tool, not a replacement.
Several new treatments are moving through the pipeline. Researchers are developing oral medications to lower cholesterol, exploring therapies for autoimmune diseases like lupus and rheumatoid arthritis, and studying compounds originally designed for diabetes that have shown promise for weight management. One recent candidate is a dengue vaccine designed to protect against all four strains of the virus with a single dose—a significant advance in a region where dengue remains endemic and dangerous.
Rosario also raised a structural problem in clinical research: Hispanic and Latin American populations remain underrepresented in most international trials. This matters because a drug that works well in one genetic population may behave differently in another. Broadening participation in these studies would make the resulting treatments more reliable for everyone. The system is working as designed, but it is incomplete until it includes the full diversity of humanity it is meant to serve.
Notable Quotes
The only way to be certain a new medication is safe and effective is through clinical trials.— Dr. Nicolás Rosario, pediatrician and president of the Clinical Research Investigator Group
Artificial intelligence will reduce the time needed to discover new molecules, but nothing will replace the need for clinical studies in actual patients.— Dr. Nicolás Rosario
The Hearth Conversation Another angle on the story
Why does it really take ten years? Is that time necessary, or is some of it just bureaucracy?
Most of it is necessary. You have to test in small groups first to make sure the drug doesn't poison people. Then you test in larger groups to see if it actually works. Then you test in thousands across different countries to make sure the results are real and not a fluke. That takes time because you have to wait and watch what happens to people.
But AI is getting better at predicting outcomes. Can't computers do some of this work?
They can do some of it, yes—they can help you find the right molecules to test in the first place, which saves years. But a computer cannot tell you how a human body will respond. You need actual people taking the drug and reporting what they feel, what changes, what goes wrong.
So AI speeds up the beginning but not the end.
Exactly. The bottleneck is always the human trials. You cannot rush that without risking harm.
You mentioned that Latin American populations are underrepresented in these trials. What does that mean practically?
It means a drug might be tested mostly on European and North American volunteers, then approved and given to people in the Caribbean or Central America who have different genetics, different diets, different health profiles. The drug might work differently for them, but nobody tested it to find out.
So the drug could be less safe or less effective for those populations.
Or more effective, or have different side effects. The point is you do not know. And that is a gap in the science.