Last week, a highly anticipated Phase 3 trial failed despite producing a dramatic improvement in the biomarker it was designed to target. The result has important implications for how the FDA evaluates new drugs.   LDL cholesterol is a biomarker. There is a strong relationship between LDL and health outcomes. The higher the LDL, the  greater the cardiovascular risk. But no one really cares about LDL levels as an end; they care about risk of heart attack, stroke, etc.   Statins are an example of a drug that lowers LDL and also lower the risk of cardiovascular events. They are among the most impactful drugs in modern medicine.   LP(a) is another cardiovascular biomarker. I'm no doctor but ChatGPT calls it a "specialized, more genetically determined cousin of LDL." There were strong reasons to believe that lowering LP(a) would have similar effects to lowering LDL, particularly reducing heart attack and stroke deaths.   Quick sidetrack... there is a big debate about how much regulators should rely on biomarkers when approving drugs. Some drugs, like statins, improve both the biomarker and clinical outcomes. Other drugs improve the biomarker but have no effect on outcomes that actually affect patients.   The appeal of approving based on biomarkers is speed. A drug's timeline for a patient may be: take the drug, time delay, biomarker improves, time delay, noticeable improvement in patient outcomes. Defining the primary outcome of a drug as improvement in the biomarker can shorten the time to regulatory approval, often by years.   There has been a push in recent years to do just that, prioritizing speed to market over confidence of efficacy. The problem is that the drug might not actually improve patient health, and might have negative side effects and/or substitute for a more effective treatment.   Back to the story... the first trial of a drug that lowered LP(a) was released last week. There was unusually strong consensus among the medical community that a drug that lowered this biomarker would decrease cardiovascular risk. Some argued the FDA should approve the drug just because it lowered LP(a) by up to 80%.   Thankfully, the FDA had not granted LP(a) as a surrogate endpoint worthy of making a drug eligible of accelerated approval. Novartis had to go through a full phase 3 trial measuring actual patient outcomes.   The trial failed. Patients saw significant reduction in LP(a) but did not improve cardiovascular outcomes.   Why is this important? There is enormous pressure from drugmakers, investors, and some patient groups to get drugs approved faster. That pressure is understandable. For patients with serious diseases, years of waiting can matter enormously.   But ineffective drugs don't help. They can have side effects. They can displace treatments that do work. They can discourage more meaningful innovation. And they usually do have significant financial cost, to patients, employers, and taxpayers.   Like most issues of public policy, the challenge is finding the right balance between competing tensions. There is good reason to get drugs to patients as quickly as possible but also to ensure that the drugs are effective. These can be at tension with each other.   The FDA is under enormous pressure from industry to accelerate access to new drugs. We @Arnold_Ventures have been one of the few voices emphasizing that speed has to be balanced against confidence in clinical efficacy.   The failed Lp(a) trial is a powerful reminder of why that balance matters. RCTs on actual patient health outcomes are still the gold standard.