AICAR: The Human Trials Were Run, and the Big One Was Stopped for Futility

01What it is

Two molecules, one name, and the naming has collapsed them. What comes in the vial is acadesine, a nucleoside, meaning a sugar joined to a nitrogen containing ring, molecular weight 258.23. What your cells make from it is AICA ribonucleotide, usually written ZMP, weight 338.21, the same molecule with a phosphate stuck on. Ask PubChem for AICAR and it hands back the phosphate, not the thing in the vial. The 2026 World Anti-Doping Agency list has the same crossed wire: it prohibits AICAR by name, and acadesine appears nowhere in it.

Acadesine is carried into cells and phosphorylated to ZMP, which resembles AMP, the low fuel signal inside a cell, closely enough to switch on AMPK, AMP-activated protein kinase, the enzyme that tells a cell to stop storing energy and start burning it. ZMP is also a normal step in the pathway every human cell uses to build purines from scratch, the raw material for DNA and for the cell's own energy currency, so this is a molecule you already have. (PMID 23228986, PMID 18674809)

It was never about exercise. Gensia had it in phase 3 by the early 1990s on a different rationale: acadesine raises adenosine, a signal that widens blood vessels and calms inflammation, in tissue short of oxygen. (PMID 7574044)

02What the trials found

Heart surgery is where "studied in thousands of patients" comes from. Mangano 1997, in JAMA, made the reputation by pooling five randomised double blind trials, 4,043 people on the same 7 hour infusion into a vein, and reporting heart attacks down 27 percent. (PMID 9002496)

Then Merck built the trial to settle it. RED-CABG randomised 3,080 people out of a planned 7,500 and stopped on a futility check written into the protocol in advance. The bad outcome hit 75 of 1,493 on the dummy infusion, 5.0 percent, and 76 of 1,493 on acadesine, 5.1 percent. (PMID 22782417, NCT00872001) Roughly 7,100 randomised across the pooled programme and RED-CABG, of whom about 3,550 received acadesine. None of the trial reports I read measured endurance, body composition or muscle.

The only people I found put on a treadmill with this drug in them were twelve patients with chronic stable angina. Holdright 1994 gave each of them acadesine into a vein at 6, 12, 24 and 48 milligrams per kilogram, and a dummy infusion, in random order. The placebo-adjusted change in how long they lasted before their electrocardiogram showed strain ran minus 0.1 percent, then 11.1, then 12.9, then minus 3.2. Eight times the drug, and the numbers wander up and back down. Time to angina, total exercise time and recovery time "were not consistently altered by acadesine". (PMID 7918131)

The studies that looked at muscle are small, and every one went into a vein. Cuthbertson 2007 infused 29 healthy men and tracked 2-deoxyglucose, a tracer standing in for glucose: uptake rose about 2.1 fold on the drug against 4.7 fold on a bicycle in the same study, while AMPK "was unchanged after 20 min or 3 h of AICAR". (PMID 17513706) The follow up infused 10 or 20 milligrams per kilogram an hour for three hours into a forearm vein, in six healthy young men, eight older men and eight men with type 2 diabetes. Glucose uptake rose, and doubling the rate raised it further in the two older groups. What doubling did not move was AMPK, which the authors call "the surprising lack of stimulation of skeletal muscle AMPK isoform activity". (PMID 19190259) Boon 2008 infused ten men with type 2 diabetes at 0.75 milligrams per kilogram a minute and reported that "AMPK phosphorylation in skeletal muscle was not increased", while recording in the same sentence a rise in the phosphorylation of acetyl-CoA carboxylase, a protein AMPK acts on, so something downstream moved where the enzyme itself did not. (PMID 18709353)

Three human infusion studies, three sets of measurements in muscle, and the enzyme the whole story rests on did not move.

03The mouse study, and the word in its abstract

Narkar 2008, in Cell, is where every exercise mimetic page traces back to. The methods, verbatim: male C57BL/6J mice, eight weeks old, "treated with AICAR (500mg/kg/day, i.p.) for 4 weeks for treadmill running tests", 15 to 20 animals per group. Intraperitoneal means injected into the abdominal cavity. Those mice "ran longer (~23%) and further (~44%)" than mice given the dummy injection. (PMID 18674809)

Two things travel wrongly out of it. The famous 44 percent is distance, and time was about 23 percent. And the abstract calls it "the orally active AMPK agonist AICAR", then says so again in the introduction, while the methods say the mice were injected. A widely read consumer guide gets both halves wrong in one sentence, calling the injections subcutaneous and the 44 percent running time.

Whether it survives being swallowed was answered in people seventeen years earlier. Dixon 1991, at Gensia, double blind against placebo in healthy men, oral solution against vein at 10 to 100 milligrams per kilogram: "The drug was poorly bioavailable (less than 5%) when administered orally in solution." A review published the same year as the Cell paper treats that as the reason it could not be developed for diabetes. (PMID 2037706, PMID 18671468)

04How it compares

Enforcing a ban on something everybody already makes is hard. Doping control samples from 499 athletes averaged 2,186 nanograms per millilitre of urine with a standard deviation of 1,655, a wide natural spread to draw a line through. (PMID 20225061)

05Where the dose came from

The published doses are grams. RED-CABG gave 42 milligrams per kilogram into a vein, about 2,940 milligrams at 70 kilograms, once. The leukaemia trial infused up to 315 milligrams per kilogram in 24 people with treatment resistant chronic lymphocytic leukaemia and settled on 210 as the highest dose it could give before toxicity stopped the escalation. (PMID 23228986) What is sold is a 50 milligram vial of powder, about one fifty-ninth of that single infusion.

The dosing pages disagree. One gives 25 milligrams a day under the skin for no more than two weeks, while stating that there are "no official guidelines on its administration", and I could find no derivation of that figure on it. A second gives 1 to 3 milligrams a day for 8 to 12 weeks, roughly an eighth to a twenty-fifth of the first number over four to six times the duration, and reads its own distance from the trial ceiling as "a wide safety margin".

Scaling the mouse dose by body surface area, my arithmetic rather than a published derivation, gives about 2,840 milligrams a day at 70 kilograms, roughly 110 times the larger page's figure and between about 950 and about 2,800 times the smaller one's.

06What could go wrong

The safety file is not empty. RED-CABG posted its adverse events: among 1,442 people given the single infusion and 1,457 given salt water, serious events ran 327 against 329. (NCT00872001)

The only thing separating drug from placebo across the 1997 pooled analysis of 4,043 patients was "a transient increase in serum uric acid". The drug is broken down into uric acid, which at high levels causes gout and stresses the kidney. In the leukaemia trial a patient at the lowest dose tested, 50 milligrams per kilogram, had a rise counted as a dose limiting toxicity. Escalation carried on anyway, with everyone from the next group given allopurinol to hold uric acid down, and no further cases appeared. It stopped at 315 milligrams per kilogram, where two of three patients hit a limiting toxicity, leaving 210 as the most anyone tolerated. Four of the 24 had kidney impairment, two of them at 315. All recovered. The French trial in myelodysplastic syndromes, a group of bone marrow disorders, closed in June 2015 with five people enrolled, and the registry gives its reason in two words: renal toxicity, meaning harm to the kidneys. (NCT01813838)

What stops that being a clean bill is that the exposure studied is not the exposure sold: one infusion, or five doses over 15 days, into a vein, under supervision. I found no human repeat dose study at daily dosing for weeks, and none for injection under the skin.

07What nobody knows

Whether anything happens at the doses sold. I found no study of any design at the milligram doses that circulate, given under the skin.

Whether any athlete has been sanctioned with AICAR as the identified substance. I could not find one, and given that everybody already carries some in their urine, plus a detection window of about four and a half hours in a horse, that absence may be a fact about the test. (PMID 28407446)

What is in the vials. I could find no independent published assay of grey market AICAR.

08My take

What makes AICAR strange is that the evidence is not missing. Somebody spent thirty years and two phase 3 programmes finding out, and the answer came back no, twice, on a question about hearts. Then an exercise story arrived on the back of a mouse paper whose abstract misdescribes its own methods, and vials turned up at doses a hundred to three thousand times below the ones that were tested and failed.

One thing about who found what. The pooled analysis that made the reputation was published in January 1997, before Dennis Mangano took worldwide rights to the compound from Metabasis in November 2000, and the two year survival paper came out in 2006, after he founded the company that held those rights. (PMID 16814669, PMID 18457469) The trial that found nothing was run by Merck and the Duke Clinical Research Institute, with a different steering committee. I am not alleging anything about the earlier work. The positive findings sit in one place and the null finding does not, and a reader should be told rather than left to find it.

If you have seen AICAR for sale, I want to know which molecular weight the certificate of analysis gave, 258 or 338, and whether the page quoted that 44 percent as running time or as distance.

Frequently asked

Has AICAR been tested in humans?+

Far more than most compounds of this kind. Roughly 7,100 people were randomised across a pooled cardiac surgery programme and one large confirmatory trial, of whom about 3,550 received acadesine, always by drip into a vein over about seven hours during heart surgery. That confirmatory trial, RED-CABG, randomised 3,080 of a planned 7,500 and was stopped early when a futility check written into the protocol in advance found a benefit very unlikely, with the bad outcome occurring in 5.0 percent of the placebo group and 5.1 percent of the acadesine group. None of that work measured endurance, body composition or athletic performance.

Does AICAR activate AMPK in people?+

Three separate human studies infused it into people and measured AMPK, the enzyme that reads a cell's fuel gauge, in muscle, and in all three the enzyme itself did not move: activity and phosphorylation were unchanged in the two studies that measured both, and phosphorylation was unchanged in the third, which did record a rise in a protein AMPK acts on further down the chain. In one of those studies, riding a bicycle raised it in the same men on the same day. The authors of a follow up study call that a surprising lack of stimulation in their own discussion. AMPK activation by AICAR is well established in rodent muscle and in cells.

Can AICAR be taken by mouth?+

It was measured in healthy men in 1991, comparing an oral solution against the same doses given into a vein, and the published finding is that the drug was poorly bioavailable, less than 5 percent, when given orally in solution. A 2008 review treats that as the reason the compound could not be developed for diabetes. The 2008 mouse paper describes the compound as orally active in its abstract while its own methods say the mice were injected into the abdominal cavity.

Where does the AICAR dose people inject come from?+

It is convention on commercial pages rather than a finding, and the pages disagree: one gives 25 milligrams a day under the skin for up to two weeks, another gives 1 to 3 milligrams a day for 8 to 12 weeks, for the same named product. The published human doses are grams delivered into a vein, 42 milligrams per kilogram in the heart surgery trial and a ceiling of 210 milligrams per kilogram in the leukaemia trial. I could find no published derivation of either circulating figure, and no human study by injection under the skin.

Is AICAR safe?+

The safety record that exists is unusually large and all of it is by the wrong route. In the trial with posted results, 1,442 people given a single infusion into a vein had 327 serious adverse events against 329 in 1,457 people given salt water. The consistent signal is a rise in uric acid, because the drug is broken down into it, and in the leukaemia trial that was severe enough at the lowest dose tested that later patients were given a preventive drug. A French trial gives renal toxicity, meaning harm to the kidneys, as its registry reason for stopping. I found no human safety data for daily dosing over weeks and none for injection under the skin.

Is AICAR banned in sport?+

It has been prohibited continuously since the list that took effect on 1 January 2009, where it entered under gene doping alongside GW 1516. On the 2026 list it appears under S4.4.1 as an activator of AMPK, in the same line as BAM15 and MOTS-c, prohibited at all times both in and out of competition, and classified as a non-specified substance, which means an athlete who tests positive cannot use the reduced penalty route that exists for substances a panel accepts might plausibly have been taken for a reason unrelated to sport.

---

This content is for educational and informational purposes only and is not medical advice. Peptides discussed are research compounds and may not be approved for human use. Nothing here should be used to diagnose, treat, cure, or prevent any disease. Always consult a qualified healthcare professional before starting any peptide, supplement, or protocol. Individual responses vary. Do not self-administer compounds without proper medical supervision.