A drug that did exactly what it was supposed to do, and nothing happened

A drug that did exactly what it was supposed to do, and nothing happened

On Saturday, November 7, 2026, cardiologist Steven Nissen of the Cleveland Clinic stood up at the American Heart Association meeting in Chicago and reported a result that very few people in his field expected. For more than a decade, lipoprotein(a), usually written Lp(a) and pronounced “L-P-little-a,” has been the most promising unsolved problem in heart disease. Roughly 1 in 5 people worldwide carries an elevated level; it runs in families, it does not budge with diet or exercise, and the people who have it suffer heart attacks and strokes even when they do everything else right.

The Lp(a)HORIZON trial gave 8,350 patients with prior heart disease either a placebo or pelacarsen, an injectable drug designed to switch off the liver’s production of Lp(a). The drug worked beautifully as a chemical. Lp(a) levels fell by about 78%, from a median of 236 nmol/L to 64 nmol/L, and stayed down for nearly five years. Yet the rate of heart attacks, strokes, cardiac deaths, and urgent procedures was 12.3% on the drug and 12.1% on placebo. The hazard ratio was 1.02, with a confidence interval of 0.90-1.15. In plain language, the trial did not merely fail to find a benefit. It ruled out any benefit larger than about 10%, and it found the same nothing in the subgroup with the highest levels, in patients who stayed on the drug faithfully, and in every secondary analysis the investigators ran (Nissen et al., 2026).

Nissen himself called it “very disappointing, terribly disappointing,” and then said something more interesting. Perhaps, he suggested, high Lp(a) does not actually cause heart attacks and strokes. Perhaps it is a marker for something else that does.

That sentence deserves to be taken seriously, not brushed aside. In this article, I want to do what scientists call “steelmanning” the idea: build the strongest possible version of the argument that Lp(a) is a messenger rather than the culprit, then see how well it holds up. I think you will find, as I did, that the honest answer is more nuanced than either “Lp(a) causes heart disease” or “Lp(a) is innocent,” and that the nuance matters a great deal for what you should do if your own Lp(a) is high.

What a “marker” would actually have to be

Before we line up suspects, we need to be clear about what “marker” means here, because the bar is higher than most people realize.

The reason doctors were so confident Lp(a) was a cause, not just a bystander, is genetics. Your Lp(a) level is set almost entirely by a single gene, called LPA, that you inherited from your parents. People who happen to be born with gene variants that produce more Lp(a) have more heart attacks, in a neat dose-dependent line, across every population studied. In a landmark Danish study, each doubling of genetically determined Lp(a) raised heart attack risk by about 22% (Kamstrup et al., 2009). A large British consortium found the same thing and called the dose-response relationship compelling evidence for a causal role (Clarke et al., 2009). This kind of analysis, called Mendelian randomization, is powerful because your genes are assigned at conception. They cannot be confounded by diet, smoking, wealth, or the other factors that muddy ordinary observational studies.

So for Lp(a) to be a mere marker, the “something else” it points to would have to pass three tests at once.

First, it would have to be tied to the LPA gene. If it were simply a lifestyle factor that travels with high Lp(a), the genetic studies would not have found anything.

Second, it would have to actually cause heart attacks and strokes on its own.

Third, and this is the test that eliminates most candidates, pelacarsen would have to leave it untouched. If the drug removed the real culprit along with the Lp(a), the trial should have worked.

Keep those three tests in mind. We will apply them to each suspect in turn.

Suspect 1: The oxidized fats riding on the particle

This is the most sophisticated version of the marker idea, and for a while it looked like the right one.

An Lp(a) particle is essentially an LDL (“bad cholesterol”) particle with an extra protein, called apolipoprotein(a), stapled to its outside. That extra protein has a peculiar talent. It grabs and carries oxidized phospholipids, fats chemically damaged by free radicals. In fact, most oxidized phospholipids circulating in your blood ride on Lp(a) rather than ordinary LDL. These damaged fats are nasty. They inflame the lining of arteries, train white blood cells to become more aggressive, and push aortic valve cells toward forming bone-like calcium deposits (Tsimikas, 2017). A leading theory has held that Lp(a) is really just the delivery truck, and the oxidized fats are the cargo doing the damage.

That theory passes the first two tests easily. The cargo travels with the LPA gene because it binds to the apo(a) protein, and the cargo plausibly causes disease. The question is the third test. Does pelacarsen leave the cargo behind?

It does not. In the earlier dose-finding study of pelacarsen, the oxidized phospholipids attached to apoB fell by 37-88% depending on the dose, right alongside Lp(a) (Tsimikas et al., 2020). The same thing happens with olpasiran, a different Lp(a)-lowering drug from Amgen. In its dose-ranging trial, oxidized phospholipids fell by 52-92% and tracked Lp(a) reduction almost perfectly (O’Donoghue et al., 2022). If you take the truck off the road, the cargo goes with it.

So the simple version of this suspect is cleared. Removing circulating oxidized phospholipids for five years did not prevent heart attacks either. What survives is a narrower and more unsettling possibility: that the damage from decades of oxidized-fat deposits is already baked into the artery wall and the valve by the time someone reaches their late fifties, and no amount of clearing the bloodstream now can undo it. That is not really a marker argument. It is a timing argument, and we will come back to it.

What to test: Oxidized LDL (oxLDL) is the test for this suspect. The research assay that measures oxidized phospholipids specifically on apoB particles is not on any commercial lab menu, but oxLDL is, and it measures the same kind of damage: LDL particles whose fats and protein have been chemically altered by free radicals. The optimal level is below 60 U/L. Since most of the oxidized phospholipids in your blood are carried on Lp(a), and Lp(a) is an LDL particle at its core, a person with a high Lp(a) and an oxLDL at or above 60 U/L is carrying the full “toxic cargo” that this suspect describes. A person with a high Lp(a) and an oxLDL below 60 U/L is not, and in that person the cargo theory has nothing to work with. I measure oxLDL routinely in my metabolic panels for exactly this reason. It also answers a question Lp(a) alone cannot: whether the oxidative stress that damages these particles is being kept in check by diet, exercise, sleep, and antioxidant status, which are the levers you actually control.

Suspect 2: Inflammation, with Lp(a) as the smoke alarm

Here is a fact that surprises many physicians. Your Lp(a) level is not entirely fixed by your genes. The LPA gene has a switch in its control region that responds to interleukin-6, one of the body’s main inflammatory signals. When inflammation rises, the liver makes more Lp(a). This is not just theory. In patients with rheumatoid arthritis treated with tocilizumab, a drug that blocks the interleukin-6 receptor, Lp(a) fell markedly in a German study and by half in an Argentine one, with no change in the gene itself (Müller et al., 2015; Pierini et al., 2021).

So could Lp(a) be a smoke alarm rather than the fire? Could high Lp(a) simply tell us that a person has chronic low-grade inflammation, and that inflammation damages arteries?

This fails the first test. Inflammation can nudge your Lp(a) up or down by a third or so, but the bulk of the variation between people, the difference between a level of 10 and a level of 200, is written in the LPA gene. The genetic studies that implicated Lp(a) were comparing people by their inherited gene variants, and those variants do not predict who will be inflamed. A smoke alarm whose loudness is mostly set at the factory is not a reliable readout of the fire.

It also fails a second check. If Lp(a) were a major driver of inflammation, lowering it should cool the fire. In the olpasiran trial, markers of inflammation such as C-reactive protein and interleukin-6 did not change at all even as Lp(a) fell by over 90% (O’Donoghue et al., 2022). Lp(a) and inflammation appear to be two separate things that happen to influence each other at the margins.

One version of this idea survives, though, and it is worth stating carefully. The HORIZON patients were remarkably well treated. Their median C-reactive protein was 0.8 mg/L, which is low. If Lp(a) does its damage mainly in an inflamed environment, acting as an accelerant rather than a spark, then a trial population with almost no inflammation would be the wrong place to look for a benefit. That is a plausible “conditional” hypothesis, and it is untested.

What to test: This is the easiest suspect to check. High-sensitivity C-reactive protein (hsCRP) is the workhorse, and the optimal level is below 0.5 mg/L. A value under 1 mg/L, as in the HORIZON patients, already means there is little inflammation for Lp(a) to amplify; below 0.5 means there is almost none. Most reference labs can measure interleukin-6 directly, and it’s a more specific signal because the IL-6 pathway drives the LPA gene. Fibrinogen, ferritin, and the neutrophil-to-lymphocyte ratio from an ordinary complete blood count round out the picture. There is also a simple experiment built into this suspect: if your Lp(a) was drawn during an infection, a flare of an inflammatory condition, or in the weeks after surgery, repeat it once things have settled. A fall of 20-30% suggests part of the number was inflammation; a number that doesn’t budge suggests it’s your genes.

Suspect 3: A clotting problem in disguise

Apolipoprotein (a) has an odd family resemblance. It looks a great deal like plasminogen, the protein your body uses to dissolve blood clots. The resemblance is so close that apo(a) can sit in plasminogen’s parking spots on a forming clot and refuse to move, which slows clot breakdown in the laboratory. From this came the idea that Lp(a) is really a marker of a sluggish clot-dissolving system, and that what kills people with high Lp(a) is not plaque but the clot that forms on top of it.

This suspect passes the first test easily, since the plasminogen-like structure is built into the LPA gene. It struggles badly with the third test, and in an instructive way.

If Lp(a) caused heart attacks by stabilizing clots, lowering it should help quickly. Clots form in minutes, not decades. A drug that removes 78% of clot-stabilizing particles should have shown benefit within the first year or two of the trial, as blood thinners do. Instead, nearly five years of treatment produced nothing, including in the patients who stayed on the drug without interruption (Nissen et al., 2026).

Other reasons make me doubt that clotting is the main story. If Lp(a) were a general clotting hazard, you would expect people with high levels to suffer more clots in their leg veins and lungs, and large population studies have found that association to be weak or absent. High Lp(a) is a disease of arteries, which points toward plaque rather than toward the clotting system (Kronenberg and Utermann, 2013). And nearly everyone in HORIZON was already taking aspirin or a similar antiplatelet drug, which would blunt any clot-related benefit anyway.

So the clotting hypothesis, far from rescuing the marker idea, is one of the suspects the trial most clearly cleared.

What to test: D-dimer and fibrinogen activity. Fibrinogen activity measures how much functional clotting protein your blood has on hand, which is the raw material every clot is built from and a well-established risk marker for heart attack and stroke in its own right; the optimal level is below 300 mg/dL. D-dimer is a fragment released only when a clot has formed and is being broken down, so it tells you whether that machinery is active right now and whether clots are forming where they shouldn’t; the optimal level is below 0.5 mg/L. Together, the two bracket the clot-dissolving system from both ends: high fibrinogen means more clot can form, and high D-dimer means clots are forming and being cleared. If fibrinogen activity is below 300 mg/dL and D-dimer is below 0.5 mg/L in a person with a high Lp(a), there is no evidence that a prothrombotic state is riding alongside the Lp(a), and this suspect can be set aside. Both tests are inexpensive and available at any laboratory.

Suspect 4: Insulin resistance and the diabetes paradox

As a metabolic medicine physician, this is the suspect I most wanted to interrogate, because insulin resistance sits behind so much of modern heart disease. Could high Lp(a) simply be flagging a metabolically unhealthy person?

The evidence says no, and it says so in a way I find fascinating. The relationship runs backward. People with low Lp(a) have more type 2 diabetes, not less. When insulin runs high, it appears to suppress the liver’s Lp(a) production, so people with insulin resistance and hyperinsulinemia tend to have lower levels (Kamstrup and Nordestgaard, 2013). A 2024 genetic analysis confirmed that genetically higher insulin pushes Lp(a) down (Lejawa et al., 2024).

This means a high Lp(a) is, if anything, a weak sign of a metabolically healthier person. The patient with an Lp(a) of 200 is less likely, not more likely, to be carrying hidden insulin resistance than her neighbor with an Lp(a) of 10. So Lp(a) cannot be a marker for metabolic syndrome. The arrow points the wrong direction.

The diabetes paradox does teach us something else, though, that will matter later. The Danish group found that the extra diabetes risk was tied specifically to the large-sized versions of the apo(a) protein, not to the Lp(a) concentration as such. This is one of several hints that apo(a) size has its own biology, separate from how much of it is in the blood. Hold that thought.

What to test: Fasting insulin is the single most informative test, and it is the one most often left off a standard panel; the optimal range is 2-5 µIU/mL. Pair it with fasting glucose (optimal 75-85 mg/dL) to calculate HOMA-IR (optimal < 1.0), and add hemoglobin A1c (optimal 4.8-5.2%), C-peptide, and the triglyceride-to-HDL ratio, a reliable, low-cost marker of insulin resistance. I also add 1,5-anhydroglucitol (1,5-AG, also known as GlycoMark), which captures what A1c misses: brief post-meal blood sugar spikes. 1,5-AG is a sugar-like molecule the kidneys normally reabsorb, but whenever glucose climbs above the kidney’s threshold, even for an hour or two after a meal, glucose crowds it out, and 1,5-AG is lost in the urine. The optimal level is above 20 µg/mL. A low 1,5-AG with a normal A1c therefore means the average looks fine, but the peaks do not, which is an early footprint of insulin resistance. Given what we know about the relationship, a high Lp(a) alongside a high fasting insulin is an unusual combination, and it means the two are coming from separate sources and both deserve attention.

Suspect 5: The kidneys, liver, and thyroid

The liver makes Lp(a) and, at least in part, the kidneys clear it. So it is no surprise that levels climb in chronic kidney disease and in nephrotic syndrome, and fall when the liver is failing. Kidney disease is itself a powerful driver of heart attacks and strokes. Could high Lp(a) mark kidney trouble?

In an individual patient, a rising Lp(a) can indeed be a clue to look at the kidneys, and I do. But as an explanation for the population-wide genetic signal, this suspect fails the first test outright. The LPA gene variants that raise Lp(a) do not cause kidney disease, and the Danish and British genetic studies were comparing people by those inherited variants. Whatever damage a failing kidney does to the heart, it is not what the genetics were detecting.

I include this suspect mainly because it illustrates an important distinction. Several things can push your Lp(a) around a little: kidney function, liver function, thyroid hormone, menopause, inflammation. None of them can explain why one person is born with a level of 15 and another with 250. The marker hypothesis has to explain the genetics, and acquired influences cannot.

What to test. For the kidneys, the urine albumin-to-creatinine ratio (uACR) matters most. Chronic kidney disease has two faces: a kidney that filters too slowly, and a kidney that leaks protein, and it is the leaking kind that drives Lp(a) up, with nephrotic syndrome raising it most dramatically of all. Protein in the urine also appears years before filtration falls in the diabetic and hypertensive kidney disease that most of my patients are at risk for, so uACR catches early kidney damage that a creatinine-based GFR will miss entirely. Because albumin excretion varies with exercise, fever, infection, and even posture, confirm an abnormal result on a first-morning sample before acting on it. The optimal uACR is below 30 mg/g. Pair uACR with serum albumin, since a low albumin alongside heavy urine protein is the signature of nephrotic syndrome. Then add creatinine with estimated GFR to assess filtration, and cystatin C (optimal below 0.90 mg/L) when you need a more accurate filtration marker, as in older adults, people with low muscle mass, and those eating a plant-based diet whose creatinine runs low for reasons unrelated to their kidneys. For the liver: ALT, AST, GGT, bilirubin, and albumin. Because an underactive thyroid also raises Lp(a) modestly, add TSH (optimal 0.4 to 4.0 mIU/L), free T4, free T3, and reverse T3. Free T3 is the active hormone tissues use, and reverse T3 is the inactive form the body makes under stress, inflammation, or undernutrition; low free T3 with high reverse T3 points to sluggish thyroid function at the tissue level even when TSH looks normal. The free T3-to-reverse T3 ratio captures that balance in a single number, and the optimal range is 20 to 30. If urine protein, serum albumin, filtration, liver enzymes, and thyroid are all normal, an acquired cause of a high Lp(a) is effectively excluded, and the number is yours by inheritance.

Suspect 6: The vitamin C repair-patch hypothesis

Many of my readers will have heard of this one, because it carries the name of a two-time Nobel laureate. In 1990, Linus Pauling and his collaborator Matthias Rath proposed that Lp(a) is a repair particle. Humans, unlike most animals, cannot make their own vitamin C, and vitamin C is essential for building the collagen that keeps artery walls strong. Pauling and Rath suggested that Lp(a) evolved as a stopgap: when vitamin C runs short and the artery wall weakens, Lp(a) sticks to the damaged spots and patches them. In this view, high Lp(a) marks a body trying to repair itself, and the real culprit is a chronic vitamin C shortage (Rath and Pauling, 1990).

I want to be fair to this idea, because it contains a genuine insight. Lp(a) really does accumulate at sites of arterial injury, and the apo(a) protein really does bind to the exposed scaffolding of a damaged vessel wall. In a sense, Lp(a) deposition is a response to injury before it becomes part of the problem. That is a design one can admire even while recognizing that, in a long-lived modern human eating a modern diet, the patch can become the plaque.

But as a scientific hypothesis, the vitamin C story has not held up. The original paper proposed the idea without new experiments. The supporting animal work relied on antibody cross-reactions in guinea pigs, whose own apo(a) status is unclear. The pattern across species does not fit: some animals that cannot make vitamin C have no Lp(a) at all, while the hedgehog carries an apo(a)-like protein despite making its own vitamin C perfectly well. Most importantly, no randomized human trial has ever shown that vitamin C, lysine, or any supplement meaningfully lowers Lp(a) or prevents heart attacks in people who have it.

And now HORIZON adds a final blow. If removing 78% of Lp(a) with a drug did not change outcomes, then any small change a supplement might produce in Lp(a) is, for practical purposes, irrelevant to whether you have a heart attack. Whatever vitamin C does for you, and I believe it does a great deal, it is not working through this pathway.

What to test: A fasting plasma vitamin C (ascorbic acid) level is commercially available, and this is the one blood test that bears directly on the Pauling-Rath idea. Levels below about 23 µmol/L (0.4 mg/dL) indicate depletion, and below 11 µmol/L (0.2 mg/dL) indicate frank deficiency. After your blood is drawn, it should be protected from light and frozen promptly, because ascorbate degrades quickly in the tube, so a low result from a sample that sat on a counter is not trustworthy. If your vitamin C is replete and your Lp(a) is still 200, the repair-patch hypothesis has nothing to work with in your case. If your vitamin C is low, correcting it is worthwhile for many reasons, but I wouldn’t expect your Lp(a) to move much, and the trials show that even if it did, your heart wouldn’t notice.

Suspect 7: What the drug leaves behind

There is one more place to look, and it is inside the drug’s own mechanism.

Pelacarsen works by blocking the liver’s instructions for making the apo(a) protein. No apo(a), no Lp(a). But remember that an Lp(a) particle is an LDL particle with apo(a) attached. The LDL half of the particle, the part carrying a protein called apoB, is still made. It simply leaves the liver as ordinary LDL instead of as Lp(a). In HORIZON, measured LDL cholesterol fell by about 13% on the drug, but a good part of that drop is just accounting: the cholesterol inside Lp(a) is normally counted as LDL cholesterol on a standard lab panel, and once the Lp(a) is gone, so is that cholesterol from the LDL line (Nissen et al., 2026).

So one could argue that the total number of apoB-carrying particles in the blood barely changed. The passengers simply moved from one bus to another. If the harm from Lp(a) were nothing more than the harm of an ordinary LDL particle, the trial would have been expected to show very little.

The trouble with this suspect is that it contradicts the genetics in the other direction. Genetic analyses estimate that an Lp(a) particle is about six times more dangerous than an LDL particle of the same size, which is exactly why cardiologists were so excited about targeting it (Björnson et al., 2024). If that were true, then converting 170 nmol/L of Lp(a) particles into plain LDL particles should have removed a great deal of risk. It removed none. Either the six-fold estimate is wrong, or the danger of an Lp(a) particle does not live in the particle count at all.

What to test: Apolipoprotein B is the test that settles this suspect, because every LDL particle and every Lp(a) particle carries exactly one apoB molecule. ApoB therefore counts the total number of artery-damaging particles regardless of which bus they are riding, and it is not fooled by the accounting problem that distorts LDL cholesterol in people with high Lp(a). Measure Lp(a) in nmol/L alongside, and you can see what fraction of your particles are Lp(a): a level of 236 nmol/L corresponds to roughly 12 mg/dL of apoB, so in a person with a total apoB of 90 mg/dL, about 1 particle in 8 is Lp(a). If your apoB is already low (under 70 mg/dL, or under 60 in secondary prevention) and your Lp(a) is high, the remaining risk is concentrated in those Lp(a) particles, and the question of what they are doing becomes the whole question.

That last possibility brings us to the suspect that, in my judgment, actually survives.

Where the steelman actually lands

Seven suspects, and most of them cleared. The oxidized fats leave with the drug. Inflammation and kidney disease cannot explain the genetics. Insulin resistance points the wrong way. Clotting should have shown a quick benefit and did not. Vitamin C deficiency has no human evidence. The “passengers moved buses” argument contradicts the very genetics that made Lp(a) a target.

What is left is subtler than “marker” and more troubling than “cause.” Here is the strongest version I can build.

The genetic studies measure the effect of a lifetime of high Lp(a), starting before birth. They cannot tell us what happens when you lower it at age 60. And there are two reasons to think the gap between those might be enormous for Lp(a) in particular.

The first is timing. Lp(a) seems to do much of its work early: initiating plaques in young arteries, and driving the slow calcification of the aortic valve over decades. By the time a person has already had a heart attack, plaques have formed, the valve is stiffening, and oxidized fats are embedded in the artery wall rather than floating in the blood. Clearing the bloodstream at that point may be like turning off the tap after the basement has already flooded. This would make Lp(a) fully causal and yet, for a secondary-prevention trial in older adults, nearly irrelevant. It is also exactly what Nissen and others have begun to say out loud.

The second is protein size. The apo(a) protein comes in different sizes, set by how many repeated segments your LPA gene contains. Small versions are made more efficiently, so people with small apo(a) have both higher levels and, it turns out, higher risk. A large study in Pakistan found that small protein size raised heart attack risk even after accounting for the concentration, and that concentration raised risk even after accounting for size. The two were separate (Saleheen et al., 2017). An Icelandic study disagreed and concluded concentration explained everything (Gudbjartsson et al., 2019), so the question is open. But if some genetic risk is carried by protein size rather than amount, then a drug that lowers the amount while leaving every remaining particle just as small and just as sticky would leave part of the risk untouched. No current drug changes protein size.

Put those together, and you get the honest conclusion. Lp(a) is almost certainly not a bystander. The genetics are too clean for that; I would put the odds that it is a pure, innocent marker at perhaps 15-20%. But it may be a cause whose damage is cumulative, front-loaded in life, and partly carried by a property of the particle that concentration alone does not capture. For a 60-year-old with established heart disease, that is a distinction without a difference. For a 30-year-old who has just learned her Lp(a) is 180, it may be the whole ballgame. The trials now underway with olpasiran and lepodisiran, which include people without prior heart disease, will tell us which it is.

The bigger picture: One process behind most early deaths

Step back for a moment from Lp(a) and look at what Suspects 1 through 6 have in common. Oxidation, inflammation, a clotting tendency, insulin resistance, strained kidneys and liver, and depleted antioxidant defenses are not six unrelated problems that happen to share a page. They are facets of a single process, and that process is the reason I practice the way I do.

Cardiometabolic risk is usually filed as one disease category among many, sitting beside cancer, dementia, and kidney disease on the list of things that can go wrong. That filing is a mistake. Insulin resistance, hyperinsulinemia, visceral fat, chronic low-grade inflammation, dyslipidemia, and a dysfunctional endothelium form one interlocking physiology, and that physiology sits upstream of nearly every major cause of premature death.

Look at the CDC’s ten leading causes of death in the United States. Heart disease is number one, and it has held that spot for a century. But the same cardiometabolic process does not stop there. It raises the risk of cancer at number two, stroke at number four, Alzheimer’s disease at number six, diabetes at number seven, kidney disease at number eight, and chronic liver disease at number nine. That is seven of the ten leading causes of death traceable to one shared physiology, and if you count suicide at number ten alongside the mood disorders I will come to in a moment, the reach is wider still. No other single, measurable, modifiable process touches so much of that list. These look like different diseases, and in their final form they are, but they share so much upstream biology that it is more accurate to think of them as travelers on the same road who simply get off at different exits. One person’s exit is a heart attack at 58, another’s is a cancer diagnosis at 63, another’s is a failing memory at 72, but the road that brought each of them there—insulin resistance, inflammation, and the damage they do to blood vessels—is the same road. Here is where each one gets off.

For heart disease and stroke, the pathway is direct: apoB particles build the plaque, inflammation destabilizes it, and insulin resistance accelerates both. Together they account for roughly a third of all deaths worldwide and remain the largest single cause of premature death.

For cancer, hyperinsulinemia and the elevated IGF-1 that travels with it are growth signals that tumors read as permission; chronic inflammation drives mutation and helps cancer cells evade the immune system; and excess body fat is now linked to at least 13 different cancers. Insulin resistance also worsens the prognosis once a cancer has been diagnosed, which is why I check for it in every oncology patient I see.

Type 2 diabetes is not a separate disease that this process causes; it is the same process caught at its endpoint. Alzheimer’s disease is increasingly understood as insulin resistance of the brain, with vascular disease and inflammation as co-drivers.

Depression and anxiety belong on this list as well, and this surprises people. The brain is an insulin-sensitive organ, and when insulin signaling falters, so do the energy supply and the neurotransmitter balance of the circuits that regulate mood. Chronic low-grade inflammation does the rest: the same cytokines that destabilize an arterial plaque cross into the brain, dampen motivation and pleasure, and produce the fatigue, withdrawal, and sleep disturbance that we recognize as depression. Psychiatry now has a name for this, immunometabolic depression, and it describes a large share of patients who do not respond well to antidepressants alone.

The relationship runs in both directions. Insulin resistance in young adulthood predicts major depression years later; people with anxiety disorders are more likely to have insulin resistance and abdominal fat, and depression in turn drives the inactivity, poor sleep, and comfort eating that deepen the metabolic problem. When I see a patient whose mood has darkened alongside a rising fasting insulin and hsCRP, I no longer treat those as two separate problems.

Chronic kidney disease and fatty liver disease are end-organ expressions of the same high blood pressure, high glucose, and metabolic fat deposition. Even deaths from infection and respiratory illness trace back here, because metabolic dysfunction impairs immune function, a lesson COVID taught at terrible cost.

The reason screening matters so much is that this process runs silently for years, and in many people for more than a decade, before any of those diagnoses appear. Fasting insulin, apoB, hsCRP, and uric acid can all be abnormal while glucose, LDL cholesterol, and body weight still look acceptable on a routine physical. By the time a diagnosis arrives, the window for cheap, reversible intervention has mostly closed.

The practical implication is leverage. A single fasting blood draw can identify the one modifiable process that, corrected early through diet, muscle, sleep, and targeted therapy where needed, lowers the risk of nearly every major cause of premature death at the same time. No other screening strategy in medicine has that reach. HORIZON, read this way, is not a story about a failed drug. It reminds us that what’s worth measuring and treating was never a single molecule. It was the terrain the molecule travels through.

What this means for you if your Lp(a) is high

If you have been told your Lp(a) is elevated, or if you have never had it checked, here is how I would translate all of this into practice.

Get it measured once, and ask for the result in nmol/L. Nothing in HORIZON changed the fact that a high Lp(a) predicts higher lifetime risk. What changed is why you know it. Your Lp(a) is no longer a number to be treated directly. It tells you how hard to work on everything else.

Make apoB and LDL your main lever. The clearest lesson from HORIZON is that the patients in the trial, with LDL cholesterol around 65 mg/dL and blood pressure around 125, had already squeezed most of the preventable risk out of their lives before the drug was ever injected. If your Lp(a) is high, that is your target too. A high Lp(a) justifies a lower LDL goal than your neighbor needs, and in secondary prevention I aim below 55 mg/dL.

Then go after Suspects 1 through 6. Here is the practical heart of this article. You cannot change your Lp(a), and HORIZON tells us that even a drug that can does not help once disease is established. But every one of the first six suspects is something you can measure and, with one partial exception, change. If Lp(a) is an accelerant that needs a fire to work on, your job is to put out the fires. Think of it as stewardship of the body you were given, not a fight against it.

Suspect 1: Oxidation. Your target is low oxidized LDL. The things that lower it are not exotic: stop smoking entirely, eat a whole-food, plant-based diet built around deeply colored vegetables, berries, legumes, and whole grains, which supply the polyphenols and fiber that protect LDL from oxidation, and remove the fried, ultra-processed, and sugar-laden foods that generate the free radicals doing the damage. Regular aerobic exercise and seven to eight hours of sleep both measurably lower oxidative stress. Recheck oxLDL in three to six months; it responds.

Suspect 2: Inflammation. Your target is an hsCRP under 1 mg/L, which is where the HORIZON patients sat. Excess abdominal fat is the most common source of chronic low-grade inflammation, and losing it is the most reliable fix. After that, look for hidden sources: untreated gum disease, poor sleep or sleep apnea, a chronically irritated gut, and an unaddressed autoimmune condition. For patients with established heart disease whose hsCRP stays elevated despite all of that, low-dose colchicine has randomized trial evidence for preventing further events, and it is worth a conversation with your cardiologist.

Suspect 3: Clotting. Your targets are a normal fibrinogen and a normal D-dimer. Fibrinogen falls with smoking cessation, weight loss, regular exercise, and good glucose control; it rises with inactivity, dehydration, and inflammation, so Suspect 2 and Suspect 3 overlap. A persistently elevated D-dimer should prompt a search for its cause, not a supplement to lower it. Whether you personally should take low-dose aspirin is a decision for you and your physician that weighs bleeding risk against clotting risk; a high Lp(a) tilts that balance somewhat, but it does not decide it.

Suspect 4: Insulin resistance. Your targets are a fasting insulin in the single digits, a HOMA-IR under 1.5, a triglyceride-to-HDL ratio under 2, and a normal 1,5-AG. This is the most controllable suspect. A low-glycemic, whole-food, plant-based diet, resistance training two or three times a week, a short walk after meals, and protecting your sleep will move every one of those numbers. Of all the levers in this article, this is the one I have watched change the most lives.

Suspect 5: Kidneys, liver, and thyroid. The goal here is simply to keep each organ healthy. Protect your kidneys by keeping blood pressure under 125/80, avoiding routine use of anti-inflammatory painkillers such as ibuprofen and naproxen, and staying hydrated. Protect your liver by treating fatty liver the same way you treat insulin resistance, since they are the same disease wearing two coats. If your free T3 is low and your reverse T3 is high, find out why before reaching for thyroid hormone; stress, undereating, and inflammation are common causes that resolve when addressed.

Suspect 6: Vitamin C. If your plasma level is low, correct it first through food (citrus, bell peppers, strawberries, kiwi, broccoli, and Brussels sprouts are all rich sources) and with a modest supplement if diet alone does not get you there. Do this for your immune system, collagen, blood vessels, and antioxidant defenses. Don’t do it expecting your Lp(a) to fall; it almost certainly will not, and HORIZON has shown it wouldn’t matter if it did.

Suspect 7 is the one you cannot act on directly, except by doing everything above. If the risk of a high Lp(a) lives in the particle itself, then the only strategy available today is to make sure those particles travel through arteries that are not inflamed, not oxidized, not insulin resistant, and not primed to clot.

Do not take niacin to lower Lp(a). It does lower it modestly, but two large trials found no benefit and some harm, and HORIZON has now shown that even a much larger Lp(a) reduction does not help.

Be skeptical of anyone selling you a supplement to lower Lp(a). If a 78% drop with a drug did nothing to outcomes, a 10% drop from a capsule cannot be expected to matter.

And if you are young, pay attention. The one scenario in which Lp(a) lowering may yet prove its worth is early, before the damage accumulates. The trials that will answer that question are enrolling now, and their results are expected over the next few years.

I will close where I began. A drug that worked perfectly as a chemical and failed completely as a medicine is humbling, and I think the right response is the one Nissen gave: we have to be humble about what this trial showed and what it did not. Lp(a) was given to you, and what it does in your body was designed into it long before any of us understood it. We are still learning to read it. In the meantime, the things proven to protect your heart are the same things they have always been, and high Lp(a) is simply a reason to pursue them with more resolve.

References

A drug that did exactly what it was supposed to do, and nothing happened

Dr. Daniel Thomas, DO, MS