Novo Nordisk reported on Friday that its experimental heart drug ziltivekimab did exactly what it was designed to do, and it did not matter at all. In a Phase 3 trial of more than 6,300 people with cardiovascular disease, kidney disease, and elevated inflammation, the drug lowered the inflammation markers it targets, IL-6 and high-sensitivity C-reactive protein, but produced no reduction in major adverse cardiovascular events compared with placebo. The hazard ratio was 0.99, with a confidence interval straddling 1.0, which in plain terms means the drug had essentially no effect on heart attacks, strokes, or cardiovascular deaths. Novo's shares fell as much as 10%.
The result is worth more than a passing note about a failed drug, because it is one of the cleanest illustrations medicine offers of a distinction that gets blurred constantly: the difference between moving a marker and helping a patient. Ziltivekimab moved its marker beautifully and helped its patients not at all. The gap between those two facts is the entire lesson of the trial, and it is a lesson the drug-development world keeps having to relearn at enormous cost.
The surrogate moved; the outcome did not
Start with what "worked" means here, because the drug did work, on everything except the point. Ziltivekimab is an antibody that blocks IL-6, a pro-inflammatory signaling molecule, and inflammation has long been suspected of driving the process that clogs arteries and triggers cardiovascular events. The drug engaged its target and drove down the inflammation markers, IL-6 itself and hsCRP, that measure inflammatory activity in the blood. On every biological readout the drug was designed to affect, it succeeded.
Those markers are surrogates. A surrogate is a measurement that stands in for the thing you actually care about, used because it is faster and easier to measure than the real outcome. Nobody wants a drug because it lowers hsCRP; they want it because it prevents heart attacks. The entire premise of developing ziltivekimab was the assumption that lowering the inflammation markers would translate into fewer cardiovascular events, that moving the surrogate would move the outcome. ZEUS tested that assumption directly, by actually counting the heart attacks, strokes, and deaths, and the assumption failed. The surrogate dropped as intended, and the outcome did not move at all. A drug can do everything it was built to do at the level of biology and still be worthless at the level of the patient, and this trial is a textbook demonstration of exactly that.
Why this is the reason outcomes trials exist, and why they keep getting shortcut
The seduction of surrogates is that they are fast and clean. Measuring whether a drug lowers a blood marker takes weeks and a modest number of patients. Measuring whether it prevents heart attacks takes years and thousands of patients, because you have to wait for enough events to occur to tell a real effect from chance. ZEUS enrolled more than 6,300 people and ran long enough to count cardiovascular events, which is why it cost what it cost and took as long as it took. That expense and delay create constant pressure across medicine to develop, approve, and prescribe drugs on the strength of surrogate movement rather than waiting for the hard outcome data.
ZEUS is a direct rebuke to that pressure. Had anyone been willing to accept "lowers inflammation markers" as sufficient evidence that ziltivekimab helps hearts, they would have endorsed a drug that lowers a number and prevents nothing. The only thing that revealed the truth was the long, expensive, inconvenient outcomes trial that actually counted the events. The causal chain everyone assumed, inflammation markers down, therefore events down, simply broke somewhere between the marker and the outcome, and it broke invisibly. You could not have seen it in the hsCRP data, which looked great. You could only see it by measuring what actually happened to people, which is precisely what outcomes trials are for and precisely what surrogate-based shortcuts skip. Every argument for approving cardiovascular drugs on biomarker changes alone has to reckon with a hazard ratio of 0.99 sitting under a perfect-looking set of inflammation numbers.
What it means for the inflammation hypothesis, handled carefully
It is tempting to read ZEUS as a verdict that inflammation does not cause cardiovascular disease, and that reading would overstate the case. The evidence is more tangled and more interesting than a clean yes or no.
The inflammation hypothesis has real support. An earlier trial of a different anti-inflammatory drug, canakinumab, which blocks IL-1β upstream of IL-6, did reduce cardiovascular events, establishing that anti-inflammatory therapy can in principle help. The cheap old anti-inflammatory colchicine has also shown cardiovascular benefit in trials. So the idea that inflammation contributes to cardiovascular risk is not refuted by ZEUS; it had already cleared meaningful hurdles. That is what makes ziltivekimab's clean failure genuinely puzzling rather than simply confirmatory, and it opens several distinct explanations that the trial alone cannot settle.
One possibility is that the hypothesis is weaker or more conditional than the earlier trials suggested, and that reducing inflammation helps less, or in fewer patients, than hoped. A second is that IL-6 specifically is the wrong lever: lowering hsCRP by blocking IL-6 moves a marker of inflammation without disrupting the actual mechanism through which inflammation damages arteries, meaning the surrogate is correlated with the disease process but not sitting on its causal path. A third is that the population was wrong: patients with both advanced cardiovascular disease and chronic kidney disease carry enormous competing risk driven by non-inflammatory pathways, and any benefit from calming inflammation could be swamped by everything else driving their events. These are not idle distinctions, because they point to different futures for the drug and the theory, and ZEUS does not choose among them.
The tests still coming, and the safety cost
Novo is treating the failure as potentially specific rather than fatal, and is continuing two other outcomes trials of the same drug, one in heart failure and one in patients following a heart attack, both due to read out in the first half of 2027. That is a bet that ZEUS failed because of its particular population, and that ziltivekimab might still help in different cardiovascular settings where the inflammatory contribution is larger or the competing risks smaller. Those readouts will do real work in distinguishing the explanations above: if the drug also fails there, the IL-6-targeting approach looks broadly wrong; if it succeeds, ZEUS looks like a population problem rather than a mechanism problem. Until then, the honest statement is narrow and firm: in this population, blocking IL-6 lowered inflammation and did nothing for cardiovascular outcomes.
There is also a cost that makes the null result worse than merely disappointing. Serious infections were more common in the ziltivekimab group, which is an expected consequence of suppressing an immune-signaling pathway, and there was no difference in overall mortality. So the drug did not simply fail to help; it carried a real downside, added infection risk, in exchange for no cardiovascular benefit. A useless drug is bad; a useless drug with a side-effect burden is worse, because patients would be taking on genuine harm for nothing.
How to read it
The durable lesson of ZEUS is not really about ziltivekimab, and not even mainly about inflammation. It is about the permanent gap between a surrogate and an outcome, and the danger of trusting the first as a stand-in for the second. A biomarker moving in the right direction feels like progress and photographs like success, and it is neither until an outcomes trial confirms that the marker's movement actually carried through to the thing that matters. The causal chain from surrogate to outcome is an assumption, not a fact, and ZEUS is a reminder that the assumption can be wrong even when the underlying idea, that inflammation matters, has genuine support, because a true general principle does not guarantee that a specific drug hitting a specific marker in a specific population will help.
It is worth noting that Novo knows this distinction as well as any company, because its own weight-loss franchise was validated not on the surrogate of weight alone but on a hard cardiovascular-outcomes trial showing real event reduction. ZEUS is the cautionary mirror image of that success: a drug that won on the surrogate and lost on the outcome. For anyone weighing cardiovascular drugs, or any drugs whose approval rests on a biomarker, the number to keep in mind is 0.99, the hazard ratio of a therapy that did everything right except help a single patient avoid the event it was meant to prevent. That is why the long, costly, slow outcomes trial exists, and why the pressure to skip it should be resisted every time it arises. The inflammation hypothesis will get its further tests in 2027. The lesson about surrogates was delivered in full on Friday.
Primary sources
- Novo Nordisk's press release via BioSpace, GlobeNewswire, and StockTitan for the ZEUS headline results, including that ziltivekimab demonstrated target engagement and reductions in free IL-6 and hsCRP but did not reduce major adverse cardiovascular events versus placebo, hazard ratio 0.99, 95% confidence interval 0.88 to 1.11, the trial's enrollment of over 6,300 people with ASCVD, CKD, and inflammation defined by hsCRP at least 2 mg/L, the once-monthly 15 mg dosing, the continuation of the HERMES heart-failure and ARTEMIS post-heart-attack trials with first-half-2027 readouts, the Q3 2026 non-cash impairment charge, that serious infections were more common with ziltivekimab consistent with IL-6 inhibition, and that no difference in all-cause mortality was observed.
- STAT for the framing of the trial as a test of whether reducing inflammation can cut cardiovascular risk and the roughly 7% share decline.
- CNBC for the up-to-10% share drop, chief scientific officer Martin Holst Lange's statement that the result does not change the company's cardiovascular commitment, the description of the drug as added to standard treatment, and the context of Novo's Wegovy oral launch and pipeline-confidence pressures.