Why the same vitamin helps one person and harms another

Paper 8E · One landscape, many inputs · Read on Zenodo

Beta-carotene, the thing carrots are full of, was supposed to prevent cancer. In the big trials it gave smokers more lung cancer. Selenium and vitamin E, expected to protect the prostate, raised the risk instead. The usual conclusion is that supplements are just weak or unreliable, and that nutrition research cannot be trusted. This page is about a different reading: that the same substance can protect one person and harm another, for a reason you can point to, and that can turn out wrong. It is a hypothesis and an invitation to test it. It is not medical advice, and it recommends no supplement or dose to anyone.

The same substance, opposite effects

Picture a race running inside every cell, all the time, between two sides. One side repairs damage and keeps the cell orderly. The other lets damage build up and pushes the cell toward uncontrolled growth. Which side wins decides whether the tissue stays healthy or slides toward cancer. Two things matter about a race like that. Once a side has won, it is very hard to turn back, like a ball rolling toward an edge: easy to nudge over, almost impossible to lift back up. And whether a given help protects or harms depends not only on what you give, but on how far the race has already run in that particular tissue. Call it the tissue's track state: how much damage, and how many early changes, are already sitting there.

That is the explanation for the supplement mess. The same substance helps when the track is low and harms when the track is high. Give folate early, while the repair side can still win, and you feed repair, which protects. Give the same folate late, when an early cancer cell is already winning, and you feed the side that is winning, because a growing cell needs building blocks too. Same substance, opposite effect, because the ground underneath it changed.

Why the trials "disagree"

When you run a big trial, you mix people with a low track and a high track together and look at the average. If the substance helps one half and harms the other, the average comes out near zero. The effect has not vanished. It lives in the subgroups, not in the average.

And that is what you see when you read the trials by subgroup instead of by headline: folate protects those low in folate and harms those high in it; selenium does the same; beta-carotene harms smokers' already-damaged lungs and leaves non-smokers alone. The same pattern recurs across about nine chemically unrelated substances. That recurrence, across things that have nothing to do with each other in the test tube, is the strongest sign that this is one shared pattern rather than nine separate disappointments.

It does not hold for everything. Vitamin C and vitamin D do not fit. That turns out to be a point in the picture's favour, because they fall outside for a reason it predicts: they do not feed the deciding race in the same way. The pattern holds when, and only when, the substance supplies something the race actually depends on.

Alcohol, the mirror

The cleanest test of the idea is something that pushes the landscape the other way. Alcohol does that: it drains the body of folate. So if the picture holds, alcohol should harm in the mirror image of a folate supplement, hitting hardest in the people who are already low in folate.

It does. In women low in folate, heavy drinking roughly doubles breast-cancer risk; in women high in folate, that effect nearly disappears. Alcohol does not so much "give" breast cancer as lower the landscape into the state where the repair side is already losing.

This resolves something the field has been genuinely confused by. Ordinary studies clearly link alcohol to breast cancer. But when you use genetic methods to measure the causal link, you find almost nothing. How can both be true? Because the genetic method averages over everyone, and washes out an effect that lives only in the folate-low. The same washout as the supplements. A real effect that depends on the landscape is invisible to a method that adds the landscape up.

The genes set the slope, not the verdict

Some genes work as brakes on growth. Weaken one of those brakes and growth is less restrained, and that shows up in more than one place: in how the brain develops, and in the body as a raised cancer risk. These are the genes that set the balance between growth and restraint, and that balance is exactly what the race is about. It is not a coincidence that the same genes turn up in both places.

The point that follows is the heart of the idea: carrying such a variant is not a verdict. The gene shifts the landscape, tilting the ground toward growth, but whether the ball actually rolls over the edge into cancer depends on the rest of the landscape: the track state and the diet environment (folate, alcohol). The strongest variants push so hard that the race commits almost whatever the diet does, which is why they carry very high risk. The common, weak variants push only a little, below the threshold, so across a whole population their effect washes out, and can even reverse. Carrying more risk variants therefore need not mean more cancer. The variant moves the terrain; the environment decides whether the ground is ever crossed.

Genes set the terrain, the environment decides

Now the two threads are one. The genes set the terrain, how steeply it tilts toward growth. Diet and lifestyle, through folate, alcohol and the rest, decide whether the ball actually crosses the edge. This dissolves the old genes-or-lifestyle question: the genes set the landscape, the environment sets the decision, and neither means much without the other.

What is new here, and what is not

Most of the biology on this page is borrowed and thoroughly studied. Folate's double role, the interplay of gene and diet, the gene overlap between brain development and cancer are all known. The two nearest existing ideas exist already too: supplement the people who are short, not those with plenty (baseline status), and hormesis (a low dose helps, a high dose harms). We take care not to claim any of that as ours. What earns the paper its keep is two sharp, untested predictions that neither neighbouring idea makes.

The first: there are two forms of folate, the ordinary "folic acid" in supplements, and the active form the body otherwise makes for itself. For people with one common gene variant, the active form should be safer than folic acid, because folic acid builds up unconverted in them. This has never been tested for cancer, and neither hormesis nor baseline status says anything about which form is used. If it holds, there is something here; if the active form is no safer, the prediction is wrong, and we will know. It is a research prediction, not a suggestion to take anything.

The second is the union of the two threads, and it is the real new thing: take the genes that shape both brain development and cancer, and compare them with a matched set of brain-development genes that are not cancer genes. The prediction is that only the first set has its cancer risk tilted by the diet landscape (folate, alcohol), and more strongly than the second. If the two sets behave the same, the whole "genes set the terrain, diet decides" idea falls. It is a concrete, falsifiable test of the union itself.

(There are two further predictions, that the harm shows up with a delay, and that the pattern extends to new systems such as methionine and liver cancer. We are honest that these fit the picture without being unique to it, since an ordinary multi-step cancer model predicts some of the same. The two above are the ones that really discriminate.)

Where the idea stops

A few things the idea does not say. The pattern does not cover everything: vitamin C and D fall outside, and that is a bound the picture predicts, not an exception swept away. The gene part is about landscape, not diagnosis: carrying one of these variants moves the terrain, it does not decide the outcome, and the diet environment does the rest. And the direct link between the genes and cancer is clearest for colorectal cancer, not breast; the breast part is the weakest link in the chain. That is why the page leads with the mechanism, folate, alcohol and track state, and not with any diagnosis. Nothing here is medical advice or a recommendation to take a supplement.

Where this stands

This is a well-grounded hypothesis with a real pattern behind it, assembled from already-published data, plus four predictions that can show themselves wrong. The next step is the decisive studies at the level of the single person: following people over time to see whether the harm really shows up late and depends on their own track and diet landscape. Until then it stands as what it is, a lens that makes three separate puzzles look like one, and that says plainly how it could be wrong.

Read the paper

The full article is freely available on Zenodo (concept DOI 10.5281/zenodo.21470791):

Read on Zenodo → · Technical version · Dansk version

Related on this site:

• What is a race? — the water pictures behind races, tracks and the landscape this paper runs on.
• Paper 1 (Friction Theory) — the framework of competing routes under bounded resources.

The family of preventive papers (same frame, different angles):

• Paper 8d — Treating the base, not the top: cofactor support matched to a measured profile, below the diagnostic threshold.
• Paper 8b — Compound race pathology: how several small pushes across biological scales add up to disease.
• Paper 8 — One mechanism behind many diagnoses: pressure, tracks and the principle of treating at the base.