What actually causes a heart attack
Nobody dies of a cholesterol reading. People die of what that reading turns into inside the artery wall. This lesson goes to where it actually happens — and along the way it dismantles two stories: the one that only ever looks at the lab report, and the one that insists the lab report doesn't matter.
In the cholesterol lesson we worked through every lever that brings your LDL down. This lesson answers the question that logically comes first, and that almost nobody bothers to ask: why do we want that in the first place?
Your cholesterol number doesn't kill you. What kills you is a plaque in a coronary artery that tears open, throws a clot, and cuts off a piece of heart muscle from its blood supply. The blood test is a forecast. The artery wall is the place.
Once you look at that place, the picture gets more precise.
ApoB is the brick. Inflammation is the bricklayer. Without particles lodging in the artery wall, no plaque forms — however inflamed you happen to be. But how fast that wall goes up, and whether it collapses, is decided by the artery itself. Both axes are causal. They just sit at different points in the chain. And each has its own levers.
Part 1: What happens inside the wall
The chain, step by step
Your arteries are lined with a single layer of cells: the endothelium. It isn't passive plumbing — it's an active organ. It decides what gets through, it produces nitric oxide to open the vessel, and it stops platelets and immune cells from sticking.
Beneath it, in the intima, atherosclerosis begins. Like this:
- ApoB particles get in. LDL particles are small enough to slip between endothelial cells into the wall. This happens in everyone, constantly. The more particles circulating, the more get in.
- They get stuck. The ApoB on their surface binds to proteoglycans — sugar chains in the connective tissue of the wall. This is the decisive step. A particle that comes back out does nothing. One that stays is the beginning.
- They're modified. Inside the wall — not in the blood — the trapped LDL gets oxidised and altered. That turns it into foreign material as far as the immune system is concerned.
- Macrophages move in, eat the modified particles, and become foam cells. They can't leave. They die there.
- A necrotic core forms from dead cells and fat, and the body lays a fibrous cap of connective tissue and smooth muscle over the top — like a plaster.
- If the cap tears, blood meets that highly thrombogenic core. A clot forms within minutes. The vessel closes. That's the heart attack.
The proof that step 2 — getting stuck — is the one that matters comes from an unusually elegant experiment. Mice were bred whose LDL was genetically altered so that it could no longer bind to the artery wall — while their blood cholesterol stayed just as high. The result: substantially less atherosclerosis.
It isn't the amount in the blood that makes the plaque. It's the lodging in the wall. The amount in the blood only determines how many particles get the chance.
Skålén K, Borén J et al. (2002). Nature 417:750–754. DOI: 10.1038/nature00804 · Causal animal experiment. Mechanistic framework: Tabas I, Williams KJ, Borén J (2007). Circulation 116:1832–1844. DOI · The mechanism is shown here in animals; the human confirmation comes from genetics and drug trials (below).
How ApoB and inflammation relate
Here's where the popular conversation reliably takes a wrong turn.
A common claim runs: "Cholesterol isn't the problem. Inflammation is the problem." It is half right — and the other half leads to false conclusions.
Because these two things don't sit side by side. They sit one behind the other:
ApoB is the necessary condition. Inflammation is the amplifier. No bricks, no wall — however fast the bricklayer works. But give him bricks, and he decides how fast it goes up. Ignore ApoB and you leave the cause standing. Ignore inflammation and you leave the second lever on the table.
The test case: the Tsimane
If inflammation were the real driver, you'd expect to find a population that is heavily inflamed and has the arteries to show for it. That population exists. And the answer isn't the one the story predicts.
Among the Tsimane, an indigenous people of the Bolivian Amazon, 705 adults over 40 were scanned by CT. 85% had no coronary calcium at all. Among those over 75, it was still 65%. That is the lowest coronary plaque burden ever measured in a population — five times lower than in a US comparison cohort.
And now the part that matters: 51% of them had elevated hsCRP — a chronic inflammatory load from parasites and infection. Their LDL sat at 91 mg/dL. Not especially low. Moderate.
High inflammation. Moderate ApoB. Almost no plaque.
Kaplan H et al. (2017). Lancet 389(10080):1730–1739. DOI · Cross-sectional study, 705 adults. Observational, not causal — the Tsimane differ from us in many ways (activity, diet, smoking). The finding doesn't disprove a causal role for inflammation; it disproves a ranking. Inflammation alone isn't enough.
A dampener I won't leave out: atherosclerosis is not purely a disease of civilisation. Whole-body CT scans of 137 mummies from four pre-industrial cultures found atherosclerosis in 34% — including hunter-gatherers (Thompson RC et al., Lancet 2013;381:1211–1222, DOI). Anyone using the Tsimane as proof that "lifestyle fixes everything" is overreaching too.
Part 2: The second axis — inflammation is real
So there's no misunderstanding: the inflammation axis is not a sideshow. It's proven — with the strongest study design there is.
CANTOS — 10,061 patients after a heart attack, all with raised hsCRP. They were given canakinumab, an antibody that blocks the inflammatory molecule IL-1β. It cut hsCRP by 37% — and left their blood lipids completely untouched. Verbatim from the paper: "Canakinumab did not reduce lipid levels from baseline."
The result: 15% fewer cardiovascular events (HR 0.85; 95% CI 0.74–0.98).
That's the proof. Blocking inflammation prevents heart attacks — without lowering LDL by a single mg/dL. There is a second axis.
Ridker PM et al. (2017), CANTOS. NEJM 377(12):1119–1131. DOI · RCT. Genetically confirmed: variants in the IL-6 receptor lower coronary risk (IL6R MR Analysis Consortium, Lancet 2012;379:1214–1224, DOI · Mendelian randomisation). The caveats: canakinumab didn't reduce all-cause mortality, it caused more fatal infections — and the FDA refused to license it for this indication.
But the axis is called IL-6. It isn't called "CRP".
This is where the reasoning commonly goes wrong. If inflammation is causal, and hsCRP measures inflammation, then surely CRP down means fewer heart attacks. It sounds coherent.
It's wrong. And it's wrong three different ways:
| The evidence | What happened |
|---|---|
| The genetics | People with genetically lifelong elevated CRP have no increased risk: RR 1.00 (0.90–1.13). Nothing. If CRP were the driver, that couldn't be true. |
| The control test | Methotrexate is a powerful anti-inflammatory. In the CIRT trial (4,786 patients) it did not lower IL-1β, IL-6 or CRP — and it prevented nothing (HR 0.96). It hit the wrong axis. |
| The largest trial, 2025 | CLEAR SYNERGY, the largest colchicine trial (7,062 patients, 3 years): CRP down by 1.28 mg/L. Events: HR 0.99. Exactly nothing. |
Genetics: CRP CHD Genetics Collaboration (2011). BMJ 342:d548. DOI · Mendelian randomisation, n = 194,418. — CIRT: Ridker PM et al. (2019). NEJM 380(8):752–762. DOI · RCT. — CLEAR SYNERGY: Jolly SS et al. (2025). NEJM 392(7):633–642. DOI · RCT, n = 7,062. Earlier, smaller colchicine trials (COLCOT, LoDoCo2) were positive — the biggest and longest one isn't.
The read: hsCRP is the thermometer. IL-6 is the fire. And you don't put out a fire by turning down the thermometer. That does not make hsCRP useless — as a risk indicator it's excellent. It makes it a poor treatment target.
Just how good an indicator it is comes out in a striking number: across more than 31,000 patients already taking a statin, hsCRP predicted the future better than LDL did. Highest versus lowest hsCRP quartile: cardiovascular death HR 2.68. For LDL in the same model: HR 1.27. (Ridker PM et al. 2023, Lancet 401:1293–1301, DOI.)
And that number gets misused constantly. It does not mean "LDL is overrated". It means: in people whose LDL has already been lowered by a statin, the risk that remains is more inflammation-driven. The LDL gradient is flat because LDL has already been treated — that is the statins' success, not their failure. Ridker's own conclusion: you need both.
Part 3: Not every plaque is the same
Vulnerable versus stable
Two people can carry the same amount of plaque and face completely different risks. What counts is the construction:
| Vulnerable plaque | Stable plaque | |
|---|---|---|
| Fibrous cap | thin (under 65 µm) | thick |
| Lipid core | large, necrotic | small |
| Inflammatory cells | many macrophages | few |
| Calcification | spotty, sparse | dense, extensive |
| Does it narrow the vessel? | often barely (under 50%) | often more |
⚠️ The last two rows are counterintuitive, which is exactly why they matter: the dangerous plaque is often the one that barely narrows the artery. It grows outwards (positive remodelling) and stays invisible to standard testing for years. Source: Virmani R et al. (2006). JACC 47(8 Suppl):C13–C18. DOI · autopsy pathology. Cap thickness in ruptured plaques: 23 ± 19 µm, 95% under 64 µm (Burke AP et al. 1997, NEJM 336:1276–1282, DOI).
And here it gets decisive for our question. The same autopsy study of 113 men who died suddenly of cardiac causes also examined their blood work:
Cholesterol makes the plaque unstable. Smoking makes it thrombogenic.
The total-cholesterol-to-HDL ratio in men with ruptured plaques was 8.5 — against 5.5 in those with stable plaques. In multivariate analysis, a raised cholesterol/HDL ratio was associated with the presence of vulnerable plaques. Smoking, by contrast, was the factor that predisposed to thrombosis.
Burke AP, Virmani R et al. (1997). NEJM 336(18):1276–1282. DOI · Autopsy, 113 men, sudden cardiac death. Observational, in the deceased — not an intervention trial, but a very direct look at the scene.
What modern autopsy data show
The classic account runs: plaque tears, a clot forms, the patient dies. A large modern autopsy series shows a different picture.
600 autopsied sudden coronary deaths in northern Finland. What did they find?
- Rupture or erosion: only 24%
- Bleeding inside the plaque: 24%
- Stable plaque, no acute complication at all: 52%
The authors, verbatim: "Less than half of sudden deaths due to CAD had evidence of acute plaque complication, an observation which is contrary to historical perceptions." In more than half, the heart muscle itself was the problem — 78% had hypertrophy, 93% had myocardial fibrosis.
Holmström L et al. (2022), Fingesture study. European Heart Journal 43(47):4923–4930. DOI · Autopsy series, 600 cases drawn from 5,869 consecutive sudden cardiac deaths. What this does and doesn't mean: it doesn't invalidate the plaque story. It shows that in the statin era, other mechanisms carry more weight — and that blood pressure and heart-muscle health matter in their own right.
Is calcium good or bad? Both.
This is frequently flattened out, and bad advice follows from it.
- Spotty, fine calcification is bad. Microcalcifications in the cap create local stress peaks and make rupture more likely. In one CT study, 63% of people with acute coronary syndrome had spotty calcification — against 21% with stable angina.
- Dense, coarse calcification is closer to protective. In the same study, large dense calcification appeared in 55% of stable cases but only 22% of acute ones. And in the MESA cohort, for the same calcium volume, a higher calcium density carried lower risk (HR 0.73 per standard deviation).
And a point that spares a lot of people needless alarm: statins make plaques smaller and more calcified at the same time. If your calcium score rises on a statin, that isn't treatment failure — it can be stabilisation.
Spotty calcification: Motoyama S et al. (2007). JACC 50(4):319–326. DOI · case-control CT. — Calcium density: Criqui MH et al. (2014), MESA. JAMA 311(3):271–278. DOI · prospective cohort. — Statins and calcification: Puri R, Nissen SE et al. (2015). JACC 65(13):1273–1282. DOI · pooled trial imaging. — The coronary calcium score remains an excellent risk stratifier (CAC above 300: RR 9.67 versus CAC = 0; Detrano R et al. 2008, NEJM 358:1336–1345, DOI) — it simply measures the scars of past activity, not what's happening now.
Part 4: Plaques can be stabilised — and we've watched it happen
This is one of the most important findings of the last twenty years, and it isn't theoretical. You can look inside a coronary artery — with ultrasound (IVUS) and optical coherence tomography (OCT) — and see what changes.
Under aggressive LDL lowering, three things happen at once, all directly measured:
- The plaque shrinks. In GLAGOV (evolocumab plus statin, LDL down to 36.6 mg/dL): plaque volume −0.95%, versus +0.05% on placebo. 64% showed regression, against 47% on placebo.
- The lipid core shrinks. In PACMAN-AMI the lipid core index fell more than twice as far as on placebo.
- The cap gets thicker. This is the one that matters: in PACMAN-AMI the minimum cap thickness grew by +62.7 µm, against +33.2 µm on placebo (difference +29.7 µm, p = 0.001). HUYGENS confirms it (+42.7 vs +21.5 µm), and there the macrophage index fell as well.
That is plaque stabilisation. Seen directly, not inferred. A thin-capped, inflamed, lipid-rich plaque turns into a thick-capped, quieter one.
GLAGOV: Nicholls SJ, Nissen SE et al. (2016). JAMA 316(22):2373–2384. DOI · RCT (n = 968). — PACMAN-AMI: Räber L et al. (2022). JAMA 327(18):1771–1781. DOI · RCT (n = 300). — HUYGENS: Nicholls SJ et al. (2022). JACC Cardiovasc Imaging 15(7):1308–1321. DOI · RCT (n = 161). These are RCTs with an imaging endpoint — not with heart attacks as the endpoint. The clinical benefit is inferred from the LDL rule.
And the limit of it: even at an LDL of 36.6 mg/dL, plaque continued to progress in 36% of patients. LDL lowering is the strongest lever we have. It is not a switch.
What lifestyle does to plaque — the thin truth: the only modern CT trial on this (DISCO-CT, DASH diet plus medication versus medication alone, n = 92) missed its primary endpoint — no difference in plaque volume. Only a secondary finding was positive: less non-calcified plaque. And the famous Ornish study ended with 35 evaluable participants, measured the lumen rather than the plaque, and bundled five interventions at once — including a large LDL reduction that can't be separated from the rest. That lifestyle shifts plaque composition favourably is plausible. That it shrinks plaque volume is not established. (Henzel J et al. 2021, JACC Cardiovasc Imaging 14(6):1192–1202, DOI; Ornish D et al. 1998, JAMA 280(23):2001–2007, DOI.)
Part 5: Antioxidants — what the trials show
This is where the wider view of the artery wall often leads people astray. The reasoning runs:
"Inside the wall, LDL gets oxidised. Oxidised LDL is the real poison. So I'll take antioxidants — and protect my artery wall without touching my cholesterol at all."
The hypothesis is plausible and mechanistically well grounded. It has been tested in large randomised trials — and it did not hold up.
| Trial | Substance | Result |
|---|---|---|
| Heart Protection Study n = 20,536 | Vitamin E + C + beta-carotene | Blood vitamin levels doubled to quadrupled → major vascular events: 1.00 (0.94–1.06). Exactly nothing. |
| HOPE / HOPE-TOO n = 9,541 | Vitamin E, 400 IU | No benefit — and more heart failure (RR 1.13; 1.01–1.26) |
| Physicians' Health Study II n = 14,641 | Vitamin E + C | No benefit — and haemorrhagic stroke HR 1.74 (1.04–2.91) |
| ATBC n = 29,133 smokers | Beta-carotene | 🔴 Lung cancer +18%, all-cause mortality +8% |
| CARET n = 18,314 | Beta-carotene + retinol | 🔴 Stopped early for harm. Lung cancer RR 1.28, all-cause mortality RR 1.17 |
| Cochrane meta-analysis 78 RCTs, n = 296,707 | Antioxidants overall | Verbatim: "Beta-carotene and vitamin E seem to increase mortality." |
HPS: Lancet 2002;360:23–33. DOI. — HOPE-TOO: Lonn E et al. (2005). JAMA 293(11):1338–1347. DOI. — PHS II: Sesso HD et al. (2008). JAMA 300(18):2123–2133. DOI. — ATBC: NEJM 1994;330:1029–1035. DOI. — CARET: Omenn GS et al. (1996). NEJM 334:1150–1155. DOI. — Cochrane: Bjelakovic G et al. (2012). CD007176. DOI. The US Preventive Services Task Force has recommended against beta-carotene and vitamin E for cardiovascular prevention since 2022 (grade D).
Why the supplements do not work
Healthy people were given vitamin E in escalating doses — up to 2,000 IU a day, five times the usual trial dose — for eight weeks. Then three validated markers were used to measure whether lipid peroxidation in the living human being actually came down.
No effect. At any dose.
The blood levels rose dutifully, dose by dose. The process they were meant to inhibit didn't budge.
Meagher EA et al. (2001). JAMA 285(9):1178–1182. DOI · RCT, mechanism study using GC/MS measurement of lipid peroxidation markers.
The reason: in a test tube vitamin E is an antioxidant. In a living body it doesn't act as one. Your own enzymes (superoxide dismutase, catalase, glutathione peroxidase) react with free radicals orders of magnitude faster. And the oxidation that matters happens inside the artery wall — not in the plasma. That's where the capsule never functionally arrives.
The same trial, the same patients, two interventions
The Heart Protection Study tested both — in the same trial, in the same 20,536 patients:
| Intervention | What it achieved | Result |
|---|---|---|
| Simvastatin | LDL down 1.0 mmol/L | −24% major vascular events |
| Antioxidant cocktail | Vitamin levels 2–4× higher | ± 0% |
Same trial, same patients, same years. One intervention lowers the particle count and works. The other hits its blood-level target perfectly — and changes nothing. Source: Heart Protection Study Collaborative Group (2002). Lancet 360:7–22 (simvastatin) and 360:23–33 (antioxidants).
Why oxidised LDL is not an independent marker
Across two large cohorts (over 50,000 people), oxidised LDL predicted heart attacks — but only until you adjusted for ApoB. Put ApoB and oxLDL into the model together, and only ApoB survived.
The reason is as simple as it is inescapable: you can only oxidise the LDL you have. Fewer particles means less substrate to oxidise.
The most effective "antioxidant lever" for your artery wall is not an antioxidant. It's less ApoB. The oxidation hypothesis is not disproven as a mechanism — oxidised LDL genuinely sits in the plaque. What's disproven is the therapeutic conclusion: that you can fight it from a capsule.
Wu T et al. (2006). JACC 48(5):973–979. DOI · Health Professionals Follow-up Study plus Nurses' Health Study, 501 events. Verbatim: oxLDL is "not an independent overall predictor of CHD … less predictive … than apoB". Observational — but two large, independent cohorts.
⚠️ If you train, this part matters — and it's rarely told: high-dose antioxidants don't just fail to help, they can blunt your training adaptation. In a randomised trial, vitamin C (1,000 mg) plus vitamin E (400 IU) completely blocked the improvement in insulin sensitivity that training produced — the gain appeared only in the group without the vitamins. The upregulation of the body's own defence enzymes and of mitochondrial biogenesis (PGC-1α) was suppressed too. The oxidative stress of training isn't the damage. It's the signal. (Ristow M et al. 2009, PNAS 106(21):8665–8670, DOI · RCT, n = 39. Confirmed for endurance training: Paulsen G et al. 2014, J Physiol 592(8):1887–1901, DOI.)
Part 6: What actually protects the artery wall
If the capsule can't do it — what can? Sorted by how strong the evidence is.
Here we have randomised trials with heart attacks as the endpoint — not lab values.
Lower your blood pressure the lever with the strongest endpoint evidence−20% events+
High blood pressure is mechanical stress on the artery wall — shear forces, stretch, endothelial damage. And lowering it is causally established across 123 randomised trials covering 613,815 people:
For every 10 mmHg of systolic reduction: 20% fewer major cardiovascular events (RR 0.80), 27% fewer strokes (RR 0.73), 13% fewer deaths.
And exercise delivers exactly that: in people with hypertension, aerobic training lowers systolic pressure by 8.3 mmHg (95% CI 6.0–10.7). On the arithmetic, that maps to roughly 17% fewer major events.
Two caveats that belong with it: the causal data come from drug trials. That exercise-induced blood-pressure lowering delivers the same is plausible but unproven. And in people with normal blood pressure, aerobic training doesn't lower systolic pressure significantly — this path is only open to you if your pressure is raised.
Causality: Ettehad D et al. (2016). Lancet 387(10022):957–967. DOI · 123 RCTs. — Exercise effect: Cornelissen VA, Smart NA (2013). JAHA 2(1):e004473. DOI · 93 RCTs, 5,223 participants.
Exercise after heart disease heart attacks as the endpoint, GRADE high−28% heart attacks+
This is the strongest evidence that exercise prevents heart attacks, and it comes from secondary prevention: 85 randomised trials, 23,430 people with coronary artery disease.
Heart attack: RR 0.72 (0.55–0.93). GRADE: high. That's the top evidence rating Cochrane awards. Long term (beyond three years): cardiovascular mortality RR 0.58, heart attacks RR 0.67.
What it does not show: all-cause mortality was not significantly reduced (RR 0.87; 0.73–1.04). And these trials never identified the mechanism — nowhere do they show it runs through endothelial function.
Dibben G et al. (2021). Cochrane Database Syst Rev 11:CD001800.pub4. DOI · 85 RCTs, 23,430 patients. NNT 75 for one heart attack prevented.
Stop smoking it makes the plaque less thrombogeniclarge+
We saw it above: in the autopsy study, cholesterol was the factor that made plaques unstable — and smoking was the factor that predisposed to thrombosis. Two different steps in the same chain.
Smoking damages the endothelium directly, raises clotting tendency, drives oxidation and inflammation. And as you saw in the cholesterol lesson: the enormous benefit of quitting runs almost entirely past your blood lipids — LDL stays essentially where it was. It's the cleanest illustration that protecting your arteries and lowering your cholesterol are two different jobs.
Burke AP, Virmani R et al. (1997). NEJM 336(18):1276–1282. DOI · Autopsy: smoking predisposes to thrombosis, cholesterol to plaque vulnerability.
Here we know exactly why it works — but nobody has shown that improving these markers prevents heart attacks.
Shear stress: blood flow scours the artery lining the causal demonstration in humanscausal+
Why does exercise improve the endothelium? Not "because muscles work". The answer is: because blood flows faster.
Flowing blood drags against the artery lining — shear stress. That force activates a signalling cascade (PI3K/Akt) in the endothelium which switches on the enzyme eNOS. eNOS produces nitric oxide. NO opens the vessel, inhibits platelets, blocks immune cells from sticking, and restrains smooth muscle cells from proliferating. It is the artery wall's central protective molecule.
And now the experiment that proves it: healthy men trained both forearms identically for eight weeks. On one arm, a blood-pressure cuff damped the blood flow — the muscular work stayed exactly the same in both.
The result: strength and forearm size grew equally in both arms. Endothelial function improved only in the arm with full blood flow. In the damped arm: nothing.
It isn't the muscle work that improves the endothelium. It's the blood flow. That is a causal demonstration in living humans — and the most precise argument for exercise that physiology offers.
Tinken TM et al. (2010). Hypertension 55(2):312–318. DOI · The cuff experiment, 8 weeks. Molecular mechanism: Dimmeler S et al. (1999). Nature 399:601–605. DOI · Akt phosphorylates eNOS at serine 1177.
Endothelial function (FMD) exercise improves it — but read the small print+2.8 points+
Flow-mediated dilation measures how well your artery widens when blood flow rises. Aerobic training improves it by 2.79 percentage points, resistance training by 2.52.
Three caveats that almost never travel with that number:
- In healthy people the gain is temporary. In one study FMD climbed from 5.9% to 9.1% by week 2 — and was back at 6.9% by week 8. Not because training stopped working, but because the artery adapts structurally. Function normalises on top of the new, better build.
- As a risk marker in healthy people it's weak. In people with disease it predicts events well (RR 0.87 per percentage point); in asymptomatic people the confidence interval touches 1.00.
- 🔴 Not one study shows that improving FMD prevents heart attacks. That is the weakest point in the entire artery-wall argument, and it belongs on the page.
Ashor AW et al. (2015). Sports Med 45(2):279–296. DOI · meta-analysis. — Transience: Tinken TM et al. (2008). J Physiol 586(20):5003–5012. DOI. — Prognostic value: Ras RT et al. (2013). Int J Cardiol 168(1):344–351. DOI.
Arterial stiffness better validated than endothelial function−0.4 m/s+
Pulse wave velocity measures how fast the pressure wave travels down your aorta. Stiff pipe, fast wave. It's the better validated of the two artery-wall markers: for every 1 m/s increase, cardiovascular risk rises 14% — and it improves risk prediction on top of the classic risk factors.
Aerobic training lowers it by −0.39 m/s (gold-standard measurement). In people whose arteries are already stiff: −1.0 m/s — twice the effect.
And if you lift, this matters: the circulating worry that heavy resistance training stiffens your arteries doesn't survive scrutiny. It rests on 8 studies with 193 young people whose arteries were already supple to begin with — and even there, middle-aged participants showed no effect. The larger, newer meta-analyses find exactly nothing (−0.013 m/s across 310 healthy adults). Resistance training even improves endothelial function.
Prognostic value: Vlachopoulos C et al. (2010). JACC 55(13):1318–1327. DOI · 17 studies, 15,877 people. Confirmed with individual data: Ben-Shlomo Y et al. (2014). JACC 63(7):636–646. DOI. — Exercise: Ashor AW et al. (2014). PLoS One 9(10):e110034. DOI · 42 RCTs. — Resistance training: the worry originates with Miyachi M (2013), Br J Sports Med 47(6):393–396, DOI (8 studies, n = 193, I² = 89%); contradicted by Ceciliato J et al. (2020), Curr Hypertens Rep 22(8):51, DOI (10 RCTs, n = 310: null).
Probably on one measure, and less clearly than it is usually presented.
HIIT versus moderate continuous training the evidence in detailmixed+
Endothelial function: HIIT improved FMD by 4.31 percentage points, moderate training by 2.15 — a 2.26-point difference in HIIT's favour. Sounds decisive. But: that's 7 studies with 182 patients. The authors themselves note that the small samples limit the finding. A 2025 network meta-analysis found moderate training came out best. The evidence is a coin toss.
Arterial stiffness: no difference between HIIT and moderate training.
VO₂max: HIIT wins — by 1.2 ml/kg/min. Real, but modest.
Hard endpoints: the only trial that ever tested HIIT against real events — Generation 100, 1,567 older adults over 5 years — missed its primary endpoint. Exercise did not reduce all-cause mortality versus control. HIIT versus control: HR 0.63 (0.33–1.20) — not significant. The trend is there. The proof isn't.
My take as a coach: HIIT is an excellent tool for VO₂max, and a time-efficient one. But anyone claiming it is demonstrably better for your arteries is leaning on 182 patients across seven small trials. The dose you'll actually keep doing beats the dose that looks optimal on paper.
FMD: Ramos JS et al. (2015). Sports Med 45(5):679–692. DOI · 7 RCTs, 182 patients. — Arterial stiffness: Way KL et al. (2019). J Sci Med Sport 22(4):385–391. DOI · no difference. — VO₂max: Milanović Z et al. (2015). Sports Med 45(10):1469–1481. DOI. — Endpoints: Stensvold D et al. (2020), Generation 100. BMJ 371:m3485. DOI · primary endpoint missed.
What is not established
The lesson could stop here, at "exercise protects your arteries". That would be incomplete. There are three places where the evidence is thinner than it usually gets presented.
What isn't established — and what nobody tells you:
1. No large trial shows that exercise prevents heart attacks in primary prevention. The best attempt, Look AHEAD, ran 5,145 people for 9.6 years: more exercise, more weight lost, better HbA1c, better fitness, better risk factors — and HR 0.95 (0.83–1.09). Stopped for futility.
2. No study shows that a better FMD or a suppler artery prevents events. We know these markers indicate risk. We do not know that improving them lowers it.
3. And the chain "lifestyle → less inflammation → fewer heart attacks" has never been demonstrated end to end. That training lowers hsCRP is a small effect and runs largely through fat loss. That lowering CRP prevents events has been refuted outright.
Look AHEAD: The Look AHEAD Research Group (2013). NEJM 369(2):145–154. DOI · RCT, n = 5,145, median 9.6 years, stopped for futility. — Exercise and hsCRP: Fedewa MV et al. (2017). Br J Sports Med 51(8):670–676. DOI · effect size 0.19 without weight loss (small). The two RCTs built specifically to test this found no effect in the absence of weight loss (Church TS et al. 2010, Med Sci Sports Exerc 42(4):708–716, DOI).
Why I still don't read this as an argument against exercise: Look AHEAD tested a combined intervention in people with diabetes, in an era when the control group was already excellently medicated — leaving little room for an added effect. And secondary prevention delivers, at GRADE-high strength, exactly what Look AHEAD couldn't show: fewer heart attacks from exercise.
But the reasoning changes. I no longer tell my clients "exercise lowers your cholesterol" — it barely does. I tell them: exercise lowers your blood pressure, keeps your endothelium responsive, strips visceral fat, improves your insulin sensitivity and your fitness — and fitness is the strongest observed correlate of life expectancy we have. That's more than enough. It doesn't need an inflated cholesterol story bolted on.
The two-axis model
| Axis 1: ApoB | Axis 2: the artery wall | |
|---|---|---|
| Role | The brick. Necessary condition. | The bricklayer. Sets the pace. |
| Without it… | …no plaque forms | …plaque grows more slowly and stays more stable |
| How you lower it | Fat swap · oats · psyllium · nuts · legumes · plant sterols · weight · medication if needed | Blood pressure · exercise (shear stress) · quitting smoking · visceral fat · sleep and stress (indirectly) |
| Strength of evidence | Very strong genetics + RCTs + imaging | Strong for blood pressure and rehab weaker for the markers |
| What doesn't work | Leaving it alone and hoping | 🔴 Antioxidant capsules. 300,000 participants, zero benefit, some harm |
"Cholesterol isn't the problem, inflammation is" is not progress — it's trading one error for another. Ignore ApoB and you leave the cause standing. Ignore the artery wall and you leave the second lever on the table. Pull both. They cost you the same week.
Where to start tomorrow
1. Work axis one. The five habits from the cholesterol lesson: oats, nuts, legumes, olive oil instead of butter, a paper filter in your coffee. That's the causal layer. 2. Get your blood pressure measured — and take it seriously. It's the strongest artery-wall lever there is, and the only one with hard endpoint evidence from hundreds of thousands of people. 3. Think of exercise as flow, not calories. What repairs your endothelium is the blood stream — so anything that reliably lifts your circulation counts. Brisk walking counts. Stairs count. An interval counts. 4. Add strength work. It doesn't stiffen your arteries — that's been refuted — and it builds the muscle your metabolism runs on. 5. And leave the capsule alone. No antioxidant has ever prevented a heart attack. Some have caused harm. And in training, they actively interfere with adaptation.
- Lowering cholesterol naturally (Lesson 34) — axis one in full: every lever, every effect size, with a lever calculator and recipe anchors.
- VO₂max and life expectancy — why cardiorespiratory fitness is the strongest observed correlate of lifespan we have.
- Losing visceral fat — the fat that raises inflammatory load and worsens insulin sensitivity.
- Supplement Compass — what belongs in your stack and what you can skip.
New lessons every two weeks
Grounded, every number sourced, and always with what it means for your actual week. Sign up and you'll get them first.
Join the newsletterSources
The causal chain: ApoB → plaque
Skålén K, Borén J et al. (2002). Nature 417:750–754. DOI — binding-deficient LDL produces less atherosclerosis despite the same hyperlipidaemia. The key evidence.
Tabas I, Williams KJ, Borén J (2007). Circulation 116(16):1832–1844. DOI — response-to-retention: the mechanistic framework.
Borén J, Chapman MJ, Krauss RM, Ference BA et al. (2020), EAS consensus. European Heart Journal 41(24):2313–2330. DOI — LDL causes atherosclerosis.
Ference BA et al. (2012). JACC 60(25):2631–2639. DOI — Mendelian randomisation (n = 312,321): 39 mg/dL less LDL for life = 54.5% less coronary disease. Cumulative exposure is what counts.
Cohen JC, Hobbs HH et al. (2006). NEJM 354(12):1264–1272. DOI — PCSK9 loss-of-function: 28% lower LDL for life → 88% less coronary disease.
Kaplan H et al. (2017), Tsimane. Lancet 389(10080):1730–1739. DOI — 85% with no coronary calcium despite a high inflammatory load (observational).
Thompson RC et al. (2013), Horus study. Lancet 381(9873):1211–1222. DOI — atherosclerosis in pre-industrial populations too (34% of 137 mummies).
The inflammation axis
Ridker PM et al. (2017), CANTOS. NEJM 377(12):1119–1131. DOI — blocking IL-1β cuts events by 15% with no lipid lowering. The proof of a second axis.
IL6R Genetics Consortium (2012). Lancet 379(9822):1214–1224. DOI — Mendelian randomisation: the IL-6 pathway is causal.
CRP CHD Genetics Collaboration (2011). BMJ 342:d548. DOI — CRP itself is NOT causal (RR 1.00; 0.90–1.13). Marker, not lever.
Ridker PM et al. (2019), CIRT. NEJM 380(8):752–762. DOI — methotrexate missed the IL-1β/IL-6 axis and did nothing (HR 0.96). The control test.
Jolly SS et al. (2025), CLEAR SYNERGY. NEJM 392(7):633–642. DOI — the largest colchicine trial (n = 7,062): CRP lowered, events HR 0.99. Nothing.
Ridker PM et al. (2023). Lancet 401(10384):1293–1301. DOI — in 31,245 statin-treated patients, hsCRP predicts the future better than LDL (CV death HR 2.68 vs 1.27).
Plaque biology, type and stabilisation
Burke AP, Virmani R et al. (1997). NEJM 336(18):1276–1282. DOI — cholesterol makes plaque vulnerable, smoking makes it thrombogenic (autopsy, 113 men).
Virmani R et al. (2006). JACC 47(8 Suppl):C13–C18. DOI — defining the thin-cap plaque.
Farb A, Virmani R et al. (1996). Circulation 93(7):1354–1363. DOI · Jia H, Jang IK et al. (2013). JACC 62(19):1748–1758. DOI — rupture versus erosion.
Holmström L et al. (2022), Fingesture. European Heart Journal 43(47):4923–4930. DOI — in 52% of sudden coronary deaths the plaque was stable.
Motoyama S et al. (2007). JACC 50(4):319–326. DOI — spotty calcification is dangerous, dense calcification closer to stable.
Criqui MH et al. (2014), MESA. JAMA 311(3):271–278. DOI — higher calcium density = lower risk (HR 0.73 per SD).
Puri R, Nissen SE et al. (2015). JACC 65(13):1273–1282. DOI — statins make plaque smaller and more calcified.
Detrano R et al. (2008), MESA. NEJM 358(13):1336–1345. DOI — the coronary calcium score as a risk stratifier.
Nicholls SJ, Nissen SE et al. (2016), GLAGOV. JAMA 316(22):2373–2384. DOI · Räber L et al. (2022), PACMAN-AMI. JAMA 327(18):1771–1781. DOI · Nicholls SJ et al. (2022), HUYGENS. JACC Cardiovasc Imaging 15(7):1308–1321. DOI — plaque shrinks, lipid core shrinks, cap thickens.
Henzel J et al. (2021), DISCO-CT. JACC Cardiovasc Imaging 14(6):1192–1202. DOI — primary endpoint missed; lifestyle and plaque volume: the data are thin.
Antioxidants and cardiovascular endpoints
Heart Protection Study Collaborative Group (2002). Lancet 360(9326):23–33. DOI — n = 20,536: vitamin levels multiplied, events 1.00 (0.94–1.06). The simvastatin arm of the same trial: Lancet 360(9326):7–22 — LDL down 1.0 mmol/L, events down 24%.
Lonn E et al. (2005), HOPE-TOO. JAMA 293(11):1338–1347. DOI — vitamin E: more heart failure (RR 1.13).
Sesso HD et al. (2008), PHS II. JAMA 300(18):2123–2133. DOI — vitamin E: haemorrhagic stroke HR 1.74.
ATBC Study Group (1994). NEJM 330(15):1029–1035. DOI — beta-carotene: lung cancer +18%, all-cause mortality +8%.
Omenn GS et al. (1996), CARET. NEJM 334(18):1150–1155. DOI — stopped early for harm.
Bjelakovic G et al. (2012), Cochrane. CD007176. DOI — 78 RCTs, 296,707 participants: "Beta-carotene and vitamin E seem to increase mortality."
Meagher EA et al. (2001). JAMA 285(9):1178–1182. DOI — vitamin E up to 2,000 IU does not lower lipid peroxidation in humans. The pill misses its own target.
Wu T et al. (2006). JACC 48(5):973–979. DOI — oxidised LDL loses all independent predictive power once you adjust for ApoB.
Ristow M et al. (2009). PNAS 106(21):8665–8670. DOI · Paulsen G et al. (2014). J Physiol 592(8):1887–1901. DOI — high-dose antioxidants blunt training adaptation.
US Preventive Services Task Force (2022). JAMA 327(23):2326–2333. DOI — recommends against beta-carotene and vitamin E for cardiovascular prevention (grade D).
Protecting the wall: blood pressure, shear stress, exercise
Ettehad D et al. (2016). Lancet 387(10022):957–967. DOI — 123 RCTs, 613,815 people: per 10 mmHg systolic, 20% fewer major events.
Cornelissen VA, Smart NA (2013). JAHA 2(1):e004473. DOI — exercise lowers systolic pressure 8.3 mmHg in hypertensives; not significantly in normotensives.
Dibben G et al. (2021), Cochrane. CD001800.pub4. DOI — 85 RCTs, 23,430 patients: exercise-based rehab reduces heart attacks, RR 0.72, GRADE high.
Tinken TM et al. (2010). Hypertension 55(2):312–318. DOI — the cuff experiment: blood flow improves the endothelium, not muscular work.
Dimmeler S et al. (1999). Nature 399(6736):601–605. DOI — shear stress → Akt → eNOS → NO.
Ashor AW et al. (2015). Sports Med 45(2):279–296. DOI (exercise → FMD) · Tinken TM et al. (2008). J Physiol 586(20):5003–5012. DOI (the FMD gain is transient in healthy people) · Ras RT et al. (2013). Int J Cardiol 168(1):344–351. DOI (FMD prognosis).
Vlachopoulos C et al. (2010). JACC 55(13):1318–1327. DOI · Ben-Shlomo Y et al. (2014). JACC 63(7):636–646. DOI — pulse wave velocity as a predictor.
Ashor AW et al. (2014). PLoS One 9(10):e110034. DOI — exercise lowers arterial stiffness.
Ceciliato J et al. (2020). Curr Hypertens Rep 22(8):51. DOI — resistance training and arterial stiffness: no effect (10 RCTs, n = 310). The opposing view (Miyachi M 2013, Br J Sports Med 47(6):393–396, DOI) rests on 8 studies with 193 young adults.
Ramos JS et al. (2015). Sports Med 45(5):679–692. DOI (HIIT vs moderate, FMD) · Way KL et al. (2019). J Sci Med Sport 22(4):385–391. DOI (arterial stiffness: no difference) · Milanović Z et al. (2015). Sports Med 45(10):1469–1481. DOI (VO₂max: +1.2 ml/kg/min).
Stensvold D et al. (2020), Generation 100. BMJ 371:m3485. DOI — primary endpoint missed.
The Look AHEAD Research Group (2013). NEJM 369(2):145–154. DOI — intensive lifestyle intervention, 9.6 years: HR 0.95, stopped for futility.
Fedewa MV et al. (2017). Br J Sports Med 51(8):670–676. DOI · Church TS et al. (2010), INFLAME. Med Sci Sports Exerc 42(4):708–716. DOI — exercise lowers hsCRP only slightly, and in an RCT, not at all without weight loss.
A note on interpretation: this lesson keeps three things strictly apart — mechanism (what happens in the cell and in the artery wall), surrogate evidence (what happens to markers like FMD, pulse wave velocity, hsCRP or plaque volume), and endpoint evidence (what happens to heart attacks and deaths). Only the last of these proves benefit. Where I rely on mechanism or surrogate, I say so. Autopsy and cohort studies show associations, not causation; the causal claims here rest on randomised trials, Mendelian randomisation and controlled experiments. This lesson is education, not medical advice. Decisions about medication and the interpretation of your own results belong with your doctor.