Why two huge anti-inflammatory heart trials failed

Two pivotal cardiovascular trials, each enrolling thousands of patients and costing hundreds of millions of dollars, reported disappointing topline results within weeks of each other. On Ground Truths, Eric Topol talks with physician-scientist Marios Georgakis about why ZEUS and HORIZON missed, and what the results do and don't say about inflammation in heart disease.

A century of evidence that inflammation is real

The link between inflammation and atherosclerosis isn't new. Early in the 20th century, pathologist Nikolai Anichkov fed rabbits high-cholesterol diets and saw what we now call foam cells, cholesterol-stuffed macrophages, in their arteries. Decades of immunochemistry later confirmed dense populations of macrophages and T cells inside human plaques, and modern single-cell sequencing has revealed a genuinely complex immune microenvironment involving neutrophils, mast cells and dendritic cells alongside them. Nobody seriously disputes that inflammation happens inside arterial plaques. The open question was always whether targeting it produces a clinical benefit.

Why Mendelian randomization matters more than correlation

Georgakis explains the key tool separating a true causal driver from a bystander marker: Mendelian randomization. Because genetic variants are fixed at birth, they can't be caused by diet or disease, so if a genetic variant that changes a mechanism also changes cardiovascular risk, that mechanism is likely causal. Run through this test, LDL cholesterol and lipoprotein(a) pass clearly: any variant that lowers them also lowers coronary disease risk. CRP fails outright: genetic variants that raise or lower CRP levels show no association with cardiovascular risk at all, despite CRP being strongly correlated with future heart attacks in standard epidemiological studies. That gap between correlation and causation is exactly what the two trials were built to test.

ZEUS: a clean, brutal null result

ZEUS enrolled over 6,300 patients with chronic kidney disease, established cardiovascular disease, and CRP above 2 mg/L, and tested an antisense drug targeting IL-6 itself. The genetic case for IL-6 signaling looked strong going in, especially a 2012 discovery that a specific variant in the IL-6 receptor gene lowers CRP and lowers coronary disease risk in a dose-dependent way, later reinforced by newer variants found directly in the IL-6 gene itself. The drug suppressed CRP by the largest margin any anti-inflammatory trial has achieved. The hazard ratio for cardiovascular events came back at 0.99, essentially a coin flip. Chronic kidney disease patients were chosen partly because they accumulate cardiovascular events faster, shortening the trial, and partly because colchicine, an older anti-inflammatory already shown to cut heart attacks and strokes in five separate trials, is used cautiously in that population, leaving less competition. Whether that population choice, rather than the IL-6 hypothesis itself, explains the failure is still being tested in two related heart-failure trials and one in acute heart attack.

HORIZON: a different kind of miss

HORIZON tested pelacarsen, a Novartis antisense drug lowering lipoprotein(a), in roughly 8,300 patients with genetically elevated Lp(a). Unlike CRP, Lp(a) levels are up to 90% genetically determined, and the genetic evidence for causality is unusually strong and consistent. The trial still missed its target. Georgakis points to a mismatch in scale: genetic studies capture a lifetime, often 50-plus years, of exposure to lower Lp(a), while a clinical trial tests a few years of drug exposure started in midlife. Modeling published in 2018 suggested that matching the risk reduction seen from a 40 mg/dL lifetime drop in LDL would require roughly a 100 mg/dL drop in Lp(a); pelacarsen lowers Lp(a) by 70 to 80%, while newer siRNA drugs in development reach 97 to 99%. A separate open question is whether Lp(a) particles vary in how much oxidized phospholipid, the likely pro-inflammatory cargo, they actually carry, meaning a high Lp(a) level on a blood test might not always translate to equally high risk.

Blood markers may be measuring the wrong thing

Both discussants agree that CRP's biggest practical problem is that it's non-specific: a cold, a minor injury or any number of unrelated conditions can raise it, and it correlates weakly with inflammation actually happening inside coronary arteries. A Spanish PET-MRI study of over 700 people found only a modest correlation, around 0.24, between CRP and imaging-confirmed arterial inflammation, and a separate CT-based fat attenuation index measurement showed a similarly weak correlation. At least one new antibody trial targeting oxidized LDL is now using coronary inflammation imaging directly as an entry criterion instead of a blood test, a shift both researchers see as a more rigorous, if more expensive, way to select patients for future trials.

The takeaway

Neither trial disproves the inflammation hypothesis in atherosclerosis; both mainly show how hard it is to translate a genetically supported target into a working drug within a short trial window, using an imperfect blood marker to select patients. LDL cholesterol remains the best-validated causal target in cardiovascular medicine precisely because decades of consistent trial data back up the genetic case. IL-6 and Lp(a) still have genetic evidence on their side, but ZEUS and HORIZON are a reminder that a causal biomarker and an effective drug at a feasible dose over a realistic timeframe are two different things to prove.