there may be useful interventions, but they will be of little value on average — if the benefit is universal, then it will be small; if it is large and predictable, then it will be limited to the few with a particular disease; otherwise, it will be unpredictably idiosyncratic so those who need it will not know it. Thus, the metallic laws: the larger the change you expect, the less likely it is; the low-hanging fruit, having already been plucked, will not be tested; and the more rigorously you test the leftovers, the smaller the final net effects will be.

- Gwern Branwen

every mouthful of food you and I have ever taken contained many billions of kinds of complex molecules whose structure and physiological effects have never been determined – and many millions of which would be toxic or fatal in large doses… we are daily ingesting thousands of substances that are far more dangerous than saccharin – but in amounts that are safe, because they are far below the various thresholds of toxicity. At present, there are hardly any substances, except some common drugs, for which we actually know the threshold.

- Edwin Jaynes (...)

In the last century, half a revolution happened: you can now buy many thousands of substances that claim to promote health, and perhaps a couple of them do.

The promise is to fine tune your health: to prevent idiosyncratic disorders, to treat the huge murky “subclinical” area of health problems, all the minor things that make life worse. And so most UK adults take supplements (about half of those multivitamins). Several problems with this:

  1. Absence of general evidence / Evidence of harm. Many supposedly health-promoting substances have uselessly weak evidence. For instance, frequent use of multivitamins is probably somewhat harmful: they increase mortality for the average user, maybe due to overdosing you with antioxidants.

  2. Physiology is personal. Even for substances that have general warrant, the ‘heterogeneity’ in their effects and side-effects can be enormous, even for quite closely matched pairs. (For instance, some people don’t get any stimulation from caffeine for genetic reasons. Morphine, the central example of a powerful and basic drug, has a “number needed to treat” post-op pain of 2.9 - i.e. on average a high dose only works well for one in three people!)

  3. Geographical and seasonal variation: for instance, during winter, around a third of UK adults are deficient in vitamin D.

  4. Snake oil on the margin: The supplement industry is regulated (for instance, you have to apply to the European Commission if you want to make a health claim for your product), but misleading claims and inaccurate concentrations are common. (For instance, the Ayurvedic supplement bacopa has been known to contain unsafe levels of lead and other heavy metals.)

  5. Powdered supplements are often 2-4x cheaper than pills, but are fiddly and sometimes taste bad. 1

  6. What counts is latent: There are now cheap places to get blood tests (or genome hits) for particular biomarkers, which you’d think would close the gap. But blood markers are only proxies for the real target variables: mortality, productivity, mood, etc.

The missing half of the revolution is measurement. The sensible supplementer needs three kinds of data to avoid waste and unnecessary risk:

  • general clinical findings,
  • personal experiments (biochemistry before and after, control doses, measurements of actually valuable responses),
  • chemical assays of particular products.

A shame that general clinical findings are so unreliable, and that getting strong personal data remains the province of heroically nerdy people, willing to invest dozens of hours into self-experimentation (reading papers, double-blinding with self-filled capsules, data collection), including learning how to analyse the results sensibly. There’s something sad about this: that external validity is so hard in biomed (and society…) that all we can really trust is local inference, n-of-1 description.

Despite plummeting measurement costs (blood tests down by 100x, genome sequencing down by 100,000x, all the consumer QS gizmos), the money and time required for an actually-scientific supplement habit are still prohibitive. So: you take safe inexpensive things and live with the uncertainty - or, more, you rely on a prior that evolution is hard to beat on body matters, and lean against taking anything except the most convincing substances.

There are economies of scale to summarising and operationalising research, testing batches, and filling capsules. And removing gatekeeping for cheap important tests has the benefit of raising our autonomy, putting us in control of at least the minor things. So is this a gap in the market? I don’t really know, I just want it to exist. (There are already well-funded toy versions of a personalised service, but their offering is pretty superficial so far.)

Another general counterargument

There is sometimes value in mere sufficiency. Across species, across phyla, there seems to be a pair of modes for a metabolism: full-steam growth vs damage control. (This is an abstraction over thousands of metabolic processes of course.) From Ricon's excellent Longevity FAQ:

When there is an abundance of nutrients, the signal is to focus on reproduction, while when they are scarce, the cell focuses on reducing the production of, and promoting the repair of, damage.

The underlying claim is something like "metabolism is violent, so things which boost it may end up causing damage". And similarly, it's at least possible that antioxidants dampen the body's active repair mechanisms.

So aiming to close all gaps - calories, amino acids, antioxidants - may end up having bad metabolic effects!
  1. For instance the nutritional shakes used by the NHS are three times cheaper in powder form. They're still £17 each for some reason, possibly because they're verifying the contents of each batch within narrow tolerance.


Post a comment:

Comments appear after moderation. You can use html or markdown.
Page just refreshes after successful submit.