Alzheimer’s disease drug-development pipelinefew candidates, frequent failures

Ninety-nine point six percent. Hold that number for a second. That's the failure rate for every Alzheimer's drug candidate that entered a clinical trial between 2002 and 2012. Not a rough estimate — a number calculated from four hundred thirteen actual trials, tracking compounds from first-in-human testing all the way to the pharmacy shelf. One drug made it. One, out of two hundred forty-four compounds tested. To understand why that number matters, you need to feel the scale of what's at stake. Cummings and colleagues report that roughly forty-four million people worldwide currently live with Alzheimer's dementia. By twenty fifty, that figure is projected to exceed one hundred million — more than doubling, driven almost entirely by population aging. Against that trajectory, the approved treatment landscape has barely moved. Five drugs have ever been approved for Alzheimer's: four cholinesterase inhibitors — tacrine, donepezil, rivastigmine, and galantamine — and one N-methyl-D-aspartate receptor antagonist called memantine. The last of those approvals came in two thousand three. None of them cure the disease or slow its progression. They manage symptoms. They help patients think a little more clearly for a little longer. That's it. The underlying biology — the plaques, the tangles, the neuronal death — continues regardless. That gap between the scale of the problem and the poverty of the solution is what Cummings and colleagues set out to quantify. Their method was systematic: they combed through ClinicalTrials.gov, the public registry where researchers are required to log ongoing clinical trials, and extracted every Alzheimer's trial registered over the decade from two thousand two to two thousand twelve. What they assembled was a detailed census of the field's effort. The numbers are substantial on their face. Four hundred thirteen trials. Two hundred forty-four unique compounds. The trials broke down into one hundred twenty-four Phase 1 studies — the first tests in humans, focused on safety — two hundred six Phase 2 studies, which begin probing whether a drug actually works, and eighty-three Phase 3 trials, the large pivotal studies required before a drug can go to regulators. Pharmaceutical companies sponsored seventy-eight percent of those trials. The National Institutes of Health alone accounted for just under seven percent. The United States ran the most trials of any single country — one hundred eighty trials, or forty-seven percent of those with recorded locations. However, when you add up all the trials run outside the United States, either exclusively or in part, the international total exceeds the domestic count. Alzheimer's drug development is a global industry, and the industry is mostly private. Activity peaked in two thousand eight and two thousand nine, with sixty-one and seventy-two trials registered respectively, then settled into roughly forty-five to fifty-one per year. That plateau matters — it means the field wasn't dramatically expanding its bets even as the disease burden was growing. What were researchers actually betting on? Cummings and colleagues sorted the four hundred thirteen trials into mechanistic categories. The largest single group — about thirty-seven percent of all trials — targeted cognition symptomatically. These were drugs designed to make patients think better, not to address the underlying disease. Another thirty-five percent were disease-modifying small molecules: compounds designed to change the biology that drives Alzheimer's. And eighteen percent were disease-modifying immunotherapies — essentially, vaccines or antibodies aimed at clearing or neutralizing harmful proteins in the brain. The distinction between symptomatic and disease-modifying is critical here. A symptomatic drug treats the smoke. A disease-modifying drug goes after the fire. And the fire, in most of these trials, was amyloid-beta — the protein that aggregates into plaques in Alzheimer's brains. Of the one hundred forty-six disease-modifying compounds studied over the decade, seventy were directed at amyloid-beta. Tau, the protein that forms tangles inside neurons, was targeted by just thirteen compounds. Sixty-two compounds addressed neuroprotection more broadly. The field had placed a large, concentrated bet on amyloid. Disease-modifying trials are also dramatically more expensive and time-consuming than symptomatic ones. In Phase 2, the mean planned treatment duration for symptomatic cognitive agents was twenty weeks. For disease-modifying small molecules, it was forty weeks. For immunotherapies, it was nearly sixty-two weeks. By Phase 3, those numbers climbed further — symptomatic trials averaged thirty-four point six weeks, small molecule disease-modifying trials averaged sixty-two weeks, and immunotherapy trials stretched to one hundred thirty-nine weeks. Patient enrollment followed the same pattern: Phase 3 symptomatic trials enrolled around three hundred fourteen patients on average, while Phase 3 disease-modifying trials enrolled over one thousand. These are not minor differences. They represent years of additional time and hundreds of millions of additional dollars per trial. Now the attrition. Cummings and colleagues traced exactly how compounds moved — or failed to move — through the pipeline. Of the compounds entering Phase 1, twenty-eight percent advanced to Phase 2. That means seventy-two percent failed at the first hurdle. Of compounds in Phase 2, only eight percent advanced to Phase 3 — a ninety-two percent attrition rate. Of the fifty-four compounds that reached Phase 3, just one was approved. That's a ninety-eight percent failure rate at the final stage. Multiply those losses across all three phases and you get the overall figure: a zero point four percent success rate. Or stated the other way, ninety-nine point six percent failure. To put that in comparative context, the paper notes that oncology — itself notoriously difficult — has a success rate of roughly nineteen percent. Alzheimer's isn't just harder. It's in a different category entirely. The plain arithmetic on display — two hundred forty-four compounds, four hundred thirteen trials, one approval — doesn't require editorial comment. It speaks for itself. Part of what makes this so damaging is where the failures concentrate. Disease-modifying trials — the ones aimed at actually changing the course of the disease — are the longest, largest, and most expensive. And of the two hundred twenty-one disease-modifying compounds tested over the decade, none showed a statistically significant drug-placebo benefit on primary outcomes. Not one. The amyloid hypothesis, which dominated the mechanistic bets, produced no approved therapies. In February two thousand fourteen, when Cummings and colleagues took a snapshot of the active pipeline, the picture was small. One hundred ten ongoing trials, ninety-five unique compounds — twenty-six Phase 1 trials, fifty-four Phase 2, and thirty Phase 3. The number that concerned them most was at the top of the funnel: only twenty-two unique agents were in Phase 1. That's the intake valve for the entire pipeline. If too few compounds enter early-stage testing, there's nothing to advance later, regardless of how good the science becomes. Of the Phase 3 programs active at the time, four of the six disease-modifying candidates still targeted amyloid-beta — the same bet that had just spent a decade failing. Applying the observed attrition rates to the February two thousand fourteen pipeline, the paper projected roughly six point four of the Phase 1 agents and four point seven of the Phase 2 agents could be expected to advance to the next stage. For the Phase 3 compounds, the odds of any reaching regulatory approval were judged to be very limited. The field was not recovering. It was treading water. Cummings and colleagues argue that changing this trajectory requires systemic moves: more compounds entering the pipeline, stronger Phase 2 programs to better characterize molecules before committing to expensive Phase 3 trials, better patient selection, earlier intervention before neurodegeneration becomes irreversible, and genuine diversification of targets beyond an unvalidated amyloid hypothesis. They also call for improved trial execution — better rating strategies, rater training, and more predictive animal models. These aren't radical proposals. They're the obvious corrections when you look at where the losses are occurring. What this analysis ultimately reveals is a mismatch in proportion. Forty-four million people living with Alzheimer's today, more than one hundred million projected by mid-century, and a decade of clinical testing that produced one approval from two hundred forty-four candidates. That's not a pipeline struggling to keep pace with the disease. That's a pipeline that has, by any honest accounting, nearly stopped. The failure rate is what it is. The question the field faces is whether it can build something genuinely different before the demographic wave arrives. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.