37,485 People Tested the B Vitamin Theory. Nothing Happened.

This is one of the clearest worked examples in modern medicine of a plausible chain of reasoning failing at the last link, and it is worth understanding in detail because the same pattern recurs constantly in supplement marketing.

Homocysteine is an amino acid produced during methionine metabolism. Elevated levels associate with cardiovascular disease and stroke, an association established across roughly 30 prospective and retrospective studies pooled in a 2002 JAMA meta-analysis covering more than 5,000 coronary events. The relationship was real but modest: a 25% lower usual homocysteine level — about 3 µmol/L — was associated with 11% lower ischaemic heart disease risk and 19% lower stroke risk. The authors themselves described homocysteine as "at most a modest independent predictor."

Meanwhile, folic acid and B vitamins reliably lower homocysteine. The chain looked complete: lower the marker, lower the risk.

It was tested properly. A 2010 meta-analysis in Archives of Internal Medicine pooled eight randomised, double-blind, placebo-controlled trials — 37,485 participants, median treatment duration five years, in people at elevated cardiovascular risk.

The supplements performed exactly as predicted biochemically, reducing homocysteine by an average of 25%, with the largest reductions in those starting highest. That is precisely the magnitude the observational data suggested should translate into meaningful benefit.

The outcomes did not move. Major vascular events came in at a relative risk of 1.01. Major coronary events 1.03. Stroke 0.96. All-cause mortality 1.02. Cancer incidence 1.05. These are not small benefits; they are null results with confidence intervals tight enough to exclude any clinically meaningful effect.

The lesson generalises. A biomarker that moves with a disease is not automatically a lever that controls it. Homocysteine may be a downstream consequence of the same processes that damage arteries — it also rises with impaired kidney function, with age, and with B12 or folate deficiency, all of which independently predict poor outcomes. Correcting the number without addressing what produced it changed nothing.

Two practical notes. High-dose folic acid can mask B12 deficiency, delaying diagnosis of a condition with neurological consequences. And the cancer incidence figure, while not statistically significant, was numerically higher in the treated group — enough that benefit cannot be assumed harmless. None of this applies to rare inherited severe hyperhomocysteinaemia, where treatment genuinely matters.

Why this matters for tracking: an elevated homocysteine is still worth knowing about, but for what it points at rather than what it is. It is a reasonable prompt to check B12, folate and renal function, all of which have clearer interpretations and clearer actions. The mistake is treating the number itself as the target — which is exactly the experiment that 37,485 people already ran.