What the magnesium evidence actually says

Magnesium is among the most recommended sleep supplements in circulation. It has a plausible mechanism, it is cheap, and it is close to harmless at ordinary doses. None of that is evidence that it will do anything for you. So it is worth looking at what the trial literature actually contains.

The most-cited synthesis is a 2021 systematic review and meta-analysis of oral magnesium for insomnia in older adults [1]. It pools three randomised controlled trials, conducted in Iran, Germany, and the United States, covering 151 people in total.

The number people quote

Magnesium reduced the time taken to fall asleep by about seventeen minutes relative to placebo. The interval sits well clear of zero. As single findings go, that is a real one, and it is the number that ends up in supplement marketing.

The number people do not quote

Total sleep time went up by roughly sixteen minutes, and the interval crosses zero. The data are consistent with magnesium costing you six minutes of sleep and with it gaining you thirty-eight. Those are different enough to warrant different decisions, and this evidence base cannot tell them apart.

Notice that the two outcomes point in directions that are easy to conflate. Falling asleep faster is not the same as sleeping longer, and it is certainly not the same as sleeping better. A supplement can do the first without doing either of the others.

How much weight the evidence carries

All three pooled trials were assessed as moderate-to-high risk of bias, and the outcomes were graded low to very low quality. The review's authors are unusually direct about what follows from that:

They still note that magnesium's low cost and wide availability mean the trial evidence may support trying it. That is a reasonable position. It is also a different claim from the one on the average supplement label.

Why this is an n-of-1 problem

Suppose the pooled estimate were rock solid. It would still be an average over 151 mostly older adults with diagnosed insomnia. If you are thirty-four and sleep tolerably, that population is not you, and the estimate was never about you. Heterogeneity of treatment effect is not a footnote here — it is the entire question a person is asking when they wonder whether to keep taking something.

This is why magnesium is the worked example on our homepage. It is the archetypal case: widely taken, biologically plausible, thinly evidenced at the population level, and perfectly tractable at the individual level, because it is cheap, reversible, and easy to start and stop on a schedule. That makes it a good candidate for a personal trial. It does not make it a recommendation, and nothing here is medical advice.