THE CRY FOR MORE DRUGS, MORE QUICKLY COMES FROM both sides of the political aisle. It sped the bipartisan passage of the 21st Century Cures Act in 2016—sweeping legislation that, among other things, aimed to streamline drug and device approval. And in his first address to Congress, President Donald Trump spoke of the “slow and burdensome” approval process of the Food and Drug Administration and recommended “we slash the restraints.”

How that might be done, and whether it should be done, is a matter of much debate. One of Trump’s early candidates for commissioner of the FDA was Jim O’Neill, a former official at the U.S. Department of Health & Human Services, who called for releasing drugs to the public after they satisfied only basic safety requirements. Rather than miring new drugs in expensive clinical trials, he said, which can last years and cost millions of dollars, “let’s prove efficacy after they’ve been legalized.”

That idea is controversial, but not without notable adherents. Most drugs currently go through a three-phase approval process, with phase 1 focusing primarily on safety using a small number of people. Phase 2 tests safety and effectiveness, generally through trials in which some patients get the drug and others receive the current treatment or a placebo. Phase 3 also uses control groups and measures how well the drug works in larger populations. But as long as a drug appears safe in a phase 2 study, the argument goes, why not release it then? Instead of going through phase 3, drug companies can monitor patient data—real-world evidence, or RWE—to see if the drugs work.

The NEWDIGS (for NEW Drug Development ParadIGmS) “think and do” tank at the MIT Center for Biomedical Innovation has been putting meat on the bones of this idea. “We think we can do better than the current approach,” says the center’s executive director and physician Gigi Hirsch. “It may be possible to speed up the pace in which patients can benefit from the science.”

In the NEWDIGS model, certain drugs that meet a critical need—novel antibiotics, promising cancer drugs, or therapies for Alzheimer’s disease, for instance—could be candidates for “adaptive licensing.” They would become available in stages, through what Hirsch calls a progressive reduction of uncertainty. The manufacturer would supply preliminary research, such as a phase 2 study, in its application for this path to approval. If the drug were selected for adaptive licensing, only the sickest patients would be eligible to take it, at least initially. The manufacturer would be required to track these patients and collect RWE from electronic health records and other sources, which could be used to apply for approval to expand the drug’s use to a wider category of patients.

Some cancer drugs, for instance, are first tailored to a small number of patients with a specific genetic mutation and later tested on a bigger group of patients with similar pathologies. Such a drug might make an ideal candidate for this model, according to Hans-Georg Eichler, senior medical officer of the European Medicines Agency. Eichler was instrumental in initiating a recently ended European pilot program that allowed some investigational drugs to propose an adaptive pathway. Many submissions were rejected, but at least one drug, a form of gene therapy for the blood disorder beta-thalassemia—under development by bluebird bio, a biotech company—is making its way toward approval.

The approach is not without its critics. Some point out the estimate that roughly half of the drugs that pass muster in phase 2 trials fail in phase 3, suggesting that drugs approved through adaptive licensing could have a 50% chance of being ineffective, unsafe or both. Joel Lexchin, a professor emeritus of health policy and management at York University in Toronto, coauthored a critique in BMJ of adaptive licensing, casting doubt on how helpful patient registries, EHRs and other RWE would be in vetting the drug’s performance. Real-world data, for instance, suggested that hormone replacement therapy prevents heart attacks in women. That finding was later disproved in a large, tightly monitored randomized control trial—exactly the kind of testing ground that adaptive licensing would cut out of the process. “Using real-world evidence to determine efficacy just doesn’t work,” Lexchin says.

The EMA’s Hans-Georg Eichler believes that desperate patients may view expedited access to new drugs very differently. “The willingness to tolerate risk and uncertainty is in direct proportion to the unmet need,” he says. But the approach does mark a sea change. In adaptive licensing, he says, patients must become more active partners in negotiating the trade-offs between uncertainty and the potential benefits of timely access. Eichler says regulators have avoided patient participation in such decision-making in the past.

“We thought that involving patients in such discussions would be inappropriate, and that if we did involve them, the public would lose trust in our regulatory decisions,” he says. “But maybe that was wrong.”