How Does AREDS2 Work? The Mechanism, and Where It Runs Out

Ask how AREDS2 works and you get one of two answers. A supplement website gives you a paragraph about antioxidants neutralising free radicals. A cautious clinician tells you nobody really knows.

Both are right, which is the interesting part.

Start with the problem the formula was designed against

Gram for gram, the retina is among the most oxygen-hungry tissues in the body. Sitting inside it are the photoreceptor outer segments: dense stacks of membrane discs, packed with polyunsaturated fatty acids, of which DHA is the most abundant.

Polyunsaturated fats oxidise readily. Now put them in the one tissue in the body whose whole job is to be hit by light, and keep the oxygen supply running at full rate.

The system copes by rebuilding constantly. Outer segments grow from the base and shed from the tip, turning over roughly every ten days, and the retinal pigment epithelium underneath digests the discarded material. Every day, for eighty years, without a break.

The hypothesis behind the formula is that this arrangement runs down. Oxidative damage accumulates faster than the clean-up, debris builds under the retina as drusen, and the tissue degrades. Load the system with antioxidants and you might slow the accumulation.

That is the theory. It is plausible, it is decades old, and the entire supplement category descends from it. Whether it is correct is a separate question, and one neither trial was built to answer.

What each of the six is supposed to be doing

Vitamin C and vitamin E: two compartments, two mops

The 500 mg of vitamin C is water-soluble, so it works in the watery spaces inside and between cells. The 400 IU of vitamin E is fat-soluble, so it embeds in membranes, which is precisely where the vulnerable polyunsaturated fats are stacked.

Between them they cover both compartments. That symmetry is a large part of why the pair has been in the formula since 2001.

Lutein and zeaxanthin: the only two that arrive at the address

This is the mechanistically satisfying part of the formula. Lutein and zeaxanthin are dietary carotenoids that the body cannot make, and the macula concentrates them from the bloodstream into a visible yellow layer called the macular pigment. Nowhere else in the body does this.

Once there they do two things. They absorb short-wavelength light before it reaches the photoreceptors, and they quench reactive oxygen species locally. An antioxidant that accumulates at the exact site of the disease is not a common thing to have.

Zinc: the eye's own enzymes need it

The retina and the retinal pigment epithelium hold some of the highest zinc concentrations found anywhere in the body. Zinc is a cofactor for antioxidant enzymes, including the ones that dismantle superoxide and peroxide, so the tissue that most needs that machinery keeps a large stock of the metal it runs on.

In AREDS, zinc alone came close to significance on its own, at an odds ratio of 0.75. It is arguably the least glamorous ingredient and one of the more defensible ones.

The dose is a separate argument. The trial used 80 mg, AREDS2 tested 25 mg against it, and the difference was nothing at all. We went through why the amounts look so high in more detail.

Copper: the one that is not for your eyes

Copper does nothing for the macula in this formula, and it is not there to.

Zinc and copper compete for the same absorption pathway in the gut. Take 80 mg of zinc every day for years and you can produce a genuine copper deficiency, which causes anaemia and neurological problems. The trial designers put 2 mg of copper in to prevent a harm their own high zinc dose would otherwise have caused.

Which is why a high-zinc eye supplement sold without copper has not simplified the formula. It has removed the safety component and kept the thing the safety component was there for.

Now the part the ingredient lists leave out

Everything above is a story about chemistry. AREDS was not a study of chemistry. It was a study of photographs.

Participants were randomised, took tablets for six years, and had their retinas photographed. Graders compared the images and counted how many eyes progressed to advanced disease. Nobody measured oxidative damage in a living human macula, before or after, and then showed that the change in that damage was what produced the change in the photographs.

So the mechanism is an inference. A reasonable one, and the reason anyone thought the trial worth running. It is not among the trial's findings.

The best mechanistic story in the trial belonged to the ingredient that failed

Consider the case for omega-3 as it stood in 2006, when AREDS2 was designed.

• DHA is the most abundant polyunsaturated fatty acid in photoreceptor outer segment membranes. The tissue is structurally made of the stuff.
• Outer segments are rebuilt every ten days, so the supply has to be continuous.
• Observational studies had repeatedly linked higher dietary omega-3 intake with lower rates of advanced AMD.
• There was a plausible anti-inflammatory route on top of the structural one.

If mechanism predicted outcome, this was the ingredient that should have worked. AREDS2 gave participants 350 mg of DHA and 650 mg of EPA daily for a median of five years.

Five-year progression to advanced AMD: 31% with omega-3. 31% without. Hazard ratio 0.97.

Nothing. Not a smaller effect than hoped. Nothing.

And lutein and zeaxanthin, the ingredients that physically accumulate in the target tissue, did not clear statistical significance in the primary analysis either. They were adopted because beta-carotene had a lung cancer signal in former smokers, not because they beat it. The full set of numbers is in does AREDS2 actually work.

What that should change about how you read a label

The standard pitch on an eye supplement is mechanistic. This contains astaxanthin, a potent antioxidant. This contains bilberry, traditionally used for vision. This contains omega-3 to support macular health.

Every one of those sentences is the same kind of reasoning that predicted omega-3 would work in AREDS2. That reasoning was tested inside a 4,203-person randomised trial and it lost.

So the practical answer to how AREDS2 works is unsatisfying and worth having anyway: the package, at the amounts used, in people who already have intermediate disease, shifted progression rates. Take the package apart and reason about the pieces, and you are doing the thing the trial showed does not predict outcomes.

Which makes the panel on the back the only part that matters. Ours is a checker that compares it against the tested formulation, ingredient by ingredient.