Search for side effects of almost any supplement and you get a wall of vagueness: may cause stomach upset, consult your doctor, individual results vary. It reads like caution and it is actually an absence of data.
AREDS is different. It randomised 3,640 people, followed them for an average of 6.3 years, monitored serum levels, medical histories and mortality throughout, and then published what it found. The numbers exist. They are just never quoted.
Start with what the trial concluded
The safety conclusion of AREDS report no. 8 is one sentence: no statistically significant serious adverse effect was associated with any of the formulations. Four arms, six years, thousands of people, continuous monitoring.
That is a stronger safety record than almost anything else in the supplement aisle can produce, and it is the context for everything below. What follows are the smaller things, published honestly by the investigators, and they are worth knowing precisely because they are small enough to be manageable.
The zinc signal, which is the real one
Eighty milligrams of zinc a day, for years, is a pharmacological dose. It is roughly seven times the ordinary dietary intake for an adult, and the trial recorded what that did.
| Finding | On zinc | Without zinc |
|---|---|---|
| Genitourinary hospitalisation, all participants | 7.5% | 4.9% |
| Genitourinary hospitalisation, men only | 8.6% | 4.4% |
| Self-reported anaemia | 13.2% | 10.2% (p=0.004) |
| Hospitalisation for mild or moderate symptoms | 9.7% | 7.8% |
AREDS report no. 8, Arch Ophthalmol 2001. The genitourinary admissions were mostly urinary tract infection and prostatic enlargement in men, and stress incontinence in women.
The men's figure is the one to sit with: nearly double, 8.6% against 4.4%. If you are a man with an already enlarged prostate or a history of urinary infections, that is a specific thing to raise with whoever prescribed the supplement, and it is a conversation almost nobody has because almost nobody knows the number.
The anaemia finding, and why it is odd
More people on zinc reported anaemia: 13.2% against 10.2%, and that difference was statistically solid at p=0.004. Then the blood tests came back and serum haematocrit showed no difference between the groups.
So something was being reported that the bloods did not confirm. It may be reporting bias, it may be a real effect too small for haematocrit to catch, it may be something else. The investigators published both halves rather than the convenient one, which is the behaviour you want from people running a trial.
Beta-carotene, the finding that actually changed the formula
This is the only safety result in the whole story that altered what is sold, and it did not come from the first trial.
In AREDS2, lung cancers were recorded in 23 participants taking beta-carotene (2.0%) against 11 not taking it (0.9%), concentrated in former smokers. Ten years later, the odds of having had lung cancer were 1.82 times higher in the group randomised to beta-carotene, with a confidence interval of 1.06 to 3.12. Lutein and zeaxanthin, tested alongside, showed no comparable increase.
That is why the formula on the shelf now contains lutein and zeaxanthin instead. Not because they worked better, which they did not in the primary analysis, but because they did not appear to carry this. The full account is in does AREDS2 actually work.
The antioxidant arms, including one oddly cheerful result
People taking the antioxidant combination more often reported yellow skin: 8.3% against 6.0%, p=0.008. That is a carotenoid doing what carotenoids do at high intake, it is cosmetic, and it reverses.
Two other things were recorded in those arms and they point in opposite directions. Hospitalisation for mild or moderate symptoms was less frequent on antioxidants, 7.4% against 10.1%. Hospitalisation for infections was more frequent, 1.6% against 0.8%.
Neither was a prespecified outcome and neither should be over-read. We mention them because leaving out the inconvenient half of a safety table is how this category normally behaves, and because the first of them is the kind of finding that would have been trumpeted if it had gone the other way.
The dose you are not told you can choose
Every signal above that involves zinc involves 80 mg of it.
AREDS2 tested a 25 mg version directly against the 80 mg one. The hazard ratio for progression to advanced AMD was 1.04, confidence interval 0.94 to 1.14. There was no difference in effect.
So the lower dose is not a weaker compromise for people who cannot tolerate the higher one. It performed the same. Products still print 80 mg on the front as though it were a feature, and if the genitourinary numbers above concern you, that is a real and evidence-backed thing to ask your ophthalmologist about rather than a reason to stop altogether.
What the numbers do and do not license you to conclude
Supported by the trials
- No serious adverse effect over 6.3 years of follow-up
- A genitourinary signal on zinc, larger in men
- A beta-carotene lung cancer signal in former smokers
- 25 mg of zinc performed identically to 80 mg
Not supported
- That the formulation is risk-free, which no supplement is
- That the anaemia signal was confirmed on blood counts, because it was not
- That these numbers apply to products with different ingredients
- That you should stop taking something a doctor recommended, on the basis of an article
Who should have the conversation before starting
- Anyone who has ever smoked, about beta-carotene specifically, by name, on the panel.
- Men with prostate symptoms or recurrent urinary infections, about the zinc dose, with the 8.6% against 4.4% figure in hand.
- Anyone with a diagnosed anaemia, or on treatment for one.
- Anyone already taking a zinc supplement, because these amounts stack and nothing on either label will tell you that.
- Anyone taking other high-dose antioxidants, for the same reason.
None of that is a reason not to take it if you are in the group the trials studied, where progression to advanced AMD fell from 28% to 20% over five years. It is a reason to take the right version of it, which is a different thing entirely and a much shorter conversation.