Sample appraisal
Magnesium threonate for memory and sleep — the trial announces “7.5 years of brain rejuvenation” from a difference of 2.24 points
ReliabilityLimited DesignRandomized clinical trial (RCT)
Magnesium beat placebo on the global cognition score, the primary outcome, by a narrow margin: +8.4 versus +5.6 points, p = 0.043 — which, redoing the calculation, is on the usual yardstick. Two thirds of the gain also came with placebo, the pattern of retaking a test. The 7.5 years of brain rejuvenation are not a measurement: they are 2.24 points divided by a population decline rate, in a calculation the methods call exploratory. On the objective sleep measures, nothing.
The effects of magnesium L-threonate (Magtein®) on cognitive performance and sleep quality in adults: a randomised, double-blind, placebo-controlled trial · Frontiers in Nutrition · 2025
What the appraisal found
- Placebo produced two-thirds of the cognitive gainIn the main outcome, the magnesium group gained 8.40 points and placebo 5.60 — leaving a difference of 2.80 points on a scale calibrated with a mean of 100 and a standard deviation of 15. The two groups did the same tests twice, six weeks apart.
- The primary passes, by a narrow marginThe gain over placebo on the global cognition score was 2.80 points, at p = 0.043. The methods declare a one-sided yardstick, but, redoing the calculation, that p is on the usual yardstick. With about thirty outcomes tested without correction, it is a signal to be confirmed.
- The 7.5 years are a conversion, not a measureThe raw difference between the groups was 2.24 points. The 7.5 years come from dividing this value by a population decline rate of 0.3 points per year. The methods themselves classify the calculation as exploratory.
The decisive problem
Selection of what was reported: about thirty outcomes tested without any correction, and four of the findings the abstract highlights — heart rate, heart rate variability, “cognitive age” and the slice by more severe complaint — are not among the registered outcomes. The primary result, which was registered, passes the usual yardstick by a narrow margin.
This is a sample: an excerpt from a real appraisal in our base, published here to show how we read a study. The full appraisal, with risk of bias domain by domain and the passage of the paper behind every sentence, comes with the plan.