The evidence list
Every conclusion we have reached, with how solid it is and the rule that decided it. Including the conclusions that argue against our own shop.
We publish this because a verdict is only worth something if you can check how it was reached. The rule sits at every rung, so you can disagree with us and point at exactly where.
Strong
1 of 18
The rule: Randomised trials in people, not contradicted by other randomised trials.
Reasonable
7 of 18
The rule: Trials, systematic reviews and analyses that do not contradict each other, but which fall short of the bar above. Also where the claim is about how well something predicts rather than about the effect of doing something: nobody can be randomised to an age or to a blood value, so replication in a separate population takes the place of a trial.
Limited
8 of 18
The rule: Everything else.
Weak
2 of 18
The rule: Barely a conclusion to draw: no human research, or research that contradicts itself throughout.
Where no trial can exist
One exception, written down here so it cannot be applied when it suits us. A claim about how well a measurement predicts an outcome cannot be tested by randomising anybody: you cannot assign a person their age, their albumin or their grip strength. For those, the equivalent of a second trial is a second population, and a measure built in one cohort and still holding in another can reach Reasonable. It can never reach Strong, because the thing that would earn that rung is impossible in principle rather than merely missing.
Measurements
The rule: No rating for how solid it is, because nothing is being estimated. Somebody measured, and this is what came out.
These sit outside the rungs. The scale above says how well research supports a statement about an effect, and a laboratory analysis of what is inside a tub is not an effect but a reading. Putting a confidence rating on it would invent an uncertainty that is not there.
What is actually in a supplement?
Often not what the label says. Published analyses of commercial products keep finding contents that do not match, and undeclared pharmaceutical substances in plant products.
Laboratory analysis of products
Direct laboratory analysis of products bought off the shelf, across several categories and several countries. Not an estimate of an effect: a reading of what was in the tub.