Every label claim traces back to a study, and the study usually says something narrower than the headline. Six questions cover most of the gap.
Start with who was studied. Effects found in people who were deficient, older, or managing a condition often shrink or disappear in healthy people. A magnesium trial in adults with low magnesium levels says little about someone whose levels are normal. Scale matters too: a 12-person trial can start a hypothesis, but it can’t settle one, and small groups make fluke results easy. The participants line is usually the fastest way to know whether a study describes someone like you.
Check what the effect was compared against. The strongest supplement studies are randomized, placebo-controlled and double-blind: participants are split by chance, the comparison group takes an identical-looking dummy, and neither participants nor researchers know who got which until the end. That design exists because expectation is a real effect — believing you’ll sleep better measurably improves sleep for a while. Without a placebo group, a study cannot separate the supplement from the expectation of the supplement. Surveys and open-label studies still have uses, but they sit lower on the evidence ladder.
Then look at what was actually measured. Many studies measure what’s easy rather than what you’d notice. A blood marker rising is easy to measure; sleeping better, lifting more, or feeling calmer is what you’re actually buying. A supplement can move a marker without moving your life. The outcome a trial reports is the claim it supports; anything past that is extrapolation.
A study supports exactly what it measured, in the people it measured, at the dose it used. Anything broader is extrapolation.
“Significant” means detectable, not big. Statistical significance says an effect probably isn’t chance. It says nothing about size. A three-minute improvement in time-to-sleep can be perfectly significant and imperceptible. The number that maps onto daily life is the effect size, and it’s often missing from the headline. Watch for relative framing doing heavy lifting: “doubled” can describe an improvement that went from tiny to slightly less tiny. Ask what changed in absolute terms — in units you would feel.
The dose has to match the label. Studies test a specific form at a specific dose for a specific duration. If the trial used 200 mg of magnesium glycinate daily for eight weeks and a product contains 50 mg inside a blend, the study doesn’t describe that product. Benefits also don’t scale with the dose: twice the studied amount is not twice the effect, though it can be twice the side effects.
Finally: who paid, and did it replicate? Industry funding doesn’t invalidate a result, but a single small sponsored trial is the weakest form of yes. One study is a data point. The same result appearing in independent labs, and surviving meta-analyses that pool every trial on the question, is what turns “promising” into “reliable.” Our reports weigh evidence the same way: replication and pooled analyses count for more than any single result.