Canakinumab and CANTOS: What Happened When Someone Finally Tested “Inflammaging” Head-On

I get asked about canakinumab more than almost any other biologic, and it’s always framed the same way: “Tony, is this the anti-aging drug that actually works?” The honest answer is that CANTOS is one of the most important trials of the last decade and also one of the most misread. It proved something real about inflammation. It did not produce a drug anyone should be chasing for longevity. Both of those things are true at once, and the gap between them is where all the interesting information lives.

What canakinumab actually is

Canakinumab (brand name Ilaris) is a fully human monoclonal antibody that binds interleukin-1β with high affinity and blocks it from reaching its receptor. It doesn’t touch IL-1α, doesn’t touch the receptor itself, doesn’t broadly suppress the immune system the way a steroid does. It removes one specific cytokine from circulation.

It has been approved in the US since 2009, originally for cryopyrin-associated periodic syndromes, and later for TRAPS, HIDS/mevalonate kinase deficiency, familial Mediterranean fever, systemic juvenile idiopathic arthritis, adult-onset Still’s disease, and — since August 2023 — gout flares in adults who can’t use NSAIDs or colchicine. Every one of those is a rare or specialist indication administered under physician supervision. None of them is “aging.”

The mechanism: NLRP3, IL-1β, and hsCRP

Here’s the pathway, and it’s worth understanding because it’s the whole reason CANTOS was designed.

Innate immune cells — macrophages, monocytes — carry a sensor complex called the NLRP3 inflammasome. It gets triggered by danger signals: cholesterol crystals in an arterial plaque, uric acid crystals, cellular debris, oxidized lipids, mitochondrial damage. Once assembled, NLRP3 activates caspase-1, which cleaves pro-IL-1β into active IL-1β.

IL-1β is the upstream master switch. It drives IL-6 production, and IL-6 drives hepatic production of C-reactive protein. That’s why hsCRP works as a downstream readout: it’s a cheap blood test that reflects how loud the IL-1β→IL-6 axis is running. If you want to understand which inflammatory markers are worth tracking and which are noise, I’ve covered that separately in reading bloodwork markers that matter.

The “inflammaging” thesis — the idea that a slow, sterile, chronic inflammatory drift is a common upstream driver of cardiovascular disease, cancer, neurodegeneration, and frailty — leans heavily on this pathway. Senescent cells secreting SASP factors feed into it, which is why the senolytics conversation and the IL-1β conversation keep bumping into each other.

But a thesis is not evidence. Nobody had ever taken a drug that only blocks inflammation, given it to tens of thousands of person-years of patients, and asked whether hard outcomes moved. That’s what CANTOS did.

What CANTOS actually showed

CANTOS enrolled 10,061 patients with a prior myocardial infarction and residual inflammation defined as hsCRP ≥2 mg/L. Mean age 61, 26% female, 40% diabetic, median LDL 82 mg/dL — so these were people already well-treated on statins. Anyone with chronic or recurrent infection, TB or HIV risk, immunocompromise, or a cancer history was excluded. Median follow-up was 3.7 years.

Trial dosing was canakinumab 50 mg, 150 mg, or 300 mg subcutaneously every three months, versus placebo. I’m reporting that as trial fact, not as a protocol.

Endpoint	Placebo	150 mg	Comment
CV death, nonfatal MI, nonfatal stroke	4.50 / 100 person-yr	3.86 / 100 person-yr	~15% relative reduction, p=0.021
Total CV events (incl. revascularization)	10.4 / 100 person-yr	8.3 / 100 person-yr	p=0.002
hsCRP vs placebo	—	37% greater reduction	Dose-dependent (26% / 37% / 41%)
LDL cholesterol	—	No change	Benefit independent of lipids
Fatal infection or sepsis	0.18 / 100 person-yr	0.31 / 100 person-yr (pooled canakinumab), p=0.02

Put the primary endpoint in absolute terms, because relative risk reduction flatters everything. The difference is about 0.64 events per 100 person-years. Run that over the trial’s 3.7-year median and you’re looking at roughly one major cardiovascular event avoided per 40-ish patients treated for nearly four years. That’s a real effect. It is not a transformation.

Two more things the headlines usually skip. First, only the 150 mg dose cleared the multiplicity-adjusted significance threshold for the primary endpoint; 50 mg and 300 mg did not. Second, there was no significant reduction in cardiovascular or all-cause mortality. The benefit was driven largely by fewer myocardial infarctions.

What CANTOS did prove is the thing it was built to prove: you can lower inflammation without touching cholesterol and still reduce cardiovascular events. That’s a genuine mechanistic win, and it’s why the field moved.

The lung cancer signal

This is the part that made headlines and, in my view, caused the most damage to clear thinking.

In an exploratory analysis published in The Lancet, incident lung cancer (129 cases) was significantly less frequent with canakinumab: HR 0.61 (95% CI 0.39–0.97) at 150 mg and HR 0.33 (95% CI 0.18–0.59) at 300 mg. Total cancer mortality was significantly lower than placebo in the 300 mg group, HR 0.49 (95% CI 0.31–0.75).

Those are striking numbers. They are also, by the authors’ own description, hypothesis-generating. Lung cancer was not a prespecified primary endpoint. The investigators explicitly acknowledged chance as a plausible explanation and flagged the potential for differential surveillance and delayed detection to bias the result. Note also that the biggest cancer effect appeared at 300 mg — the dose that failed the cardiovascular primary endpoint.

Novartis launched four trials off that signal. Here’s how they went:

• CANOPY-2 (canakinumab + docetaxel, 2nd/3rd line advanced NSCLC, n=237): missed OS and PFS. Numerically higher infections, including fatal infections.
• CANOPY-1 (canakinumab + pembrolizumab + platinum chemo, 1st line): missed OS and PFS. Grade 3–4 toxicity 64.1% vs 59.3% placebo.
• CANOPY-A (adjuvant, 1,382 patients with completely resected stage II–IIIB NSCLC, 200 mg SC every 3 weeks): missed the primary endpoint of disease-free survival. Announced August 2022.

CANOPY-A was the closest analogue to the CANTOS population — early-stage disease, prevention-adjacent — and it still failed. Lythgoe and Prasad wrote a sharp postmortem in the British Journal of Cancer arguing the whole program was a case study in over-reading a secondary endpoint and skipping validation. I think that’s the correct read.

So: the lung cancer finding was real as a statistical observation in one trial and has not replicated in the settings tested since.

The harm side

You cannot remove IL-1β from a human being without cost. IL-1β is a core antimicrobial signal.

In CANTOS, fatal infection or sepsis occurred at 0.31 per 100 person-years on canakinumab versus 0.18 on placebo (p=0.02). That’s roughly one additional fatal infection per 200 people treated for the trial duration — and remember, patients with chronic or recurrent infection, TB risk, or immunocompromise were excluded at enrollment. In a real-world population the signal would plausibly be worse. There were also significant reductions in neutrophils and platelets on drug.

Set that next to the benefit: about one event prevented per ~40 treated, about one fatal infection caused per ~200 treated, and no mortality benefit. Robert Harrington’s summary to TCTMD is the fairest one I’ve read — a modest effect, balanced by some harm, with no effect on mortality.

Why the FDA said no

Novartis did not seek approval based on the overall trial. It sought an indication for “responders” — patients who dropped their hsCRP below 2.0 mg/L on treatment, a subgroup with a much larger apparent benefit including reduced mortality.

That’s a post-randomization variable. Once you select a subgroup by how someone responded to the drug, you are no longer comparing randomized groups, and the analysis is prone to confounding. Cardiologists interviewed at the time called the odds of approval “vanishingly small.” The FDA issued a complete response letter in late 2018, stating the data did not support the proposed responder-targeted indication. In Europe, the CHMP raised major objections — that CANTOS was not robust enough to show canakinumab works in all post-MI patients — and Novartis withdrew the application in December 2018 rather than answer them.

The cost problem

Even if it had been approved, the economics were brutal. Canakinumab is priced as an orphan drug — estimates around $200,000 per year at monthly dosing, roughly $16,000 per injection. Quarterly dosing as used in CANTOS is cheaper but nowhere near affordable at population scale.

A cost-effectiveness analysis in JAMA Cardiology put the incremental cost-effectiveness ratio at approximately $6.4 million per quality-adjusted life-year at market price, against a typical willingness-to-pay threshold near $100,000. Even modeling the responder subgroup at a heavily discounted $73,000 per year, it came out around $819,000 per QALY. To make a CV indication commercially sensible, the company would have had to cut the price by something like 98% and give up a reliable orphan revenue stream. It didn’t.

What this teaches about inflammation as a longevity target

Here’s my honest read, and it’s not the exciting version.

The hypothesis survived. The drug didn’t. CANTOS showed inflammation is causal in atherosclerosis, not just correlated. That’s a genuine advance and it validates hsCRP as something worth watching in your bloodwork panel. But the effect size from wiping out one cytokine with a $200k biologic was 15% relative, no mortality benefit, and a fatal infection signal.

Cheap doesn’t automatically mean it works either. Low-dose colchicine looked like the affordable answer: COLCOT showed a 23% reduction post-MI, LoDoCo2 a 31% reduction in chronic coronary disease. Then CLEAR SYNERGY (OASIS-9), the largest colchicine CV trial ever run with about 7,000 patients and published in the New England Journal of Medicine in 2025, found that the primary composite occurred in 9.1% on colchicine versus 9.3% on placebo over a median 3.5 years — flat, despite CRP dropping. Low-dose methotrexate in CIRT also failed, though it never moved IL-1β, IL-6, or hsCRP. Lowering a marker is not the same as lowering risk.

The unglamorous inputs still carry the most evidence. Nothing in CANTOS argues against the boring stuff — not smoking, controlling visceral fat, sleeping properly, training consistently, managing metabolic health. Those move systemic inflammation and they don’t have a sepsis signal. I’ve written about the recovery side of that in the recovery protocol, and about how the pieces fit together in the overall protocol framework.

The real lesson from CANTOS isn’t “target inflammation.” It’s that a single, elegant, mechanistically perfect intervention against one cytokine bought a modest reduction in one disease and cost lives to infection along the way. Aging is not one cytokine. Anyone selling you a single-pathway answer — including me, if I ever do — should be met with the same skepticism the FDA showed here.

Canakinumab is a specialist-administered prescription biologic. It is not something to source, self-administer, or add to a stack, and I’m not going to pretend otherwise because the mechanism is interesting.

FAQ

Is canakinumab approved for preventing heart attacks?

No. The FDA issued a complete response letter in late 2018 rejecting the proposed cardiovascular indication, and Novartis withdrew its European application in December 2018. Canakinumab remains approved only for certain rare autoinflammatory diseases, Still’s disease, and gout flares.

How big was the cardiovascular benefit in CANTOS?

At the 150 mg dose, the composite of cardiovascular death, nonfatal MI, and nonfatal stroke fell from 4.50 to 3.86 events per 100 person-years, about a 15% relative reduction. There was no significant reduction in cardiovascular or all-cause mortality.

Did canakinumab really prevent lung cancer?

An exploratory analysis of CANTOS found substantially lower incident lung cancer on canakinumab, with a hazard ratio of 0.33 at the 300 mg dose. Lung cancer was not a prespecified primary endpoint, and the follow-up CANOPY-1, CANOPY-2, and CANOPY-A trials in non-small cell lung cancer all failed to meet their primary endpoints.

What were the main harms?

Fatal infection or sepsis increased from 0.18 to 0.31 per 100 person-years compared with placebo, roughly one additional fatal infection per 200 people treated over the trial period. Neutrophil and platelet counts also fell on treatment.

Does CANTOS mean I should try to lower my hsCRP?

CANTOS supports the idea that inflammation is causal in cardiovascular disease, and hsCRP is a reasonable marker to discuss with your doctor. But lowering a marker is not automatically the same as lowering risk. CLEAR SYNERGY found colchicine reduced CRP without improving outcomes after myocardial infarction, and methotrexate failed in CIRT.

Medical disclaimer: This article is for informational and educational purposes only and is not medical advice; canakinumab is a prescription biologic and any decision about it belongs with a qualified physician.