Wouldn’t it be great if there was a cheap, non-proprietary, readily available treatment for patients with heart attacks (acute myocardial infarction - AMI)? That’s what doctors wanted to believe, so when they saw the early results of magnesium therapy, they did exactly that. Magnesium therapy for AMI has been labelled a “lesson in medical humility”, but I see it as another example of the pervasive bias amongst researchers, doctors and the public that leads them to overestimate the effectiveness of medical therapies. Put simply, it was another case of ‘believing is seeing’.
The story starts in the lab where, as expected, magnesium was shown to have just about every positive effect possible on the heart, including anti-platelet effects, anti-arrhythmia effects, blood vessel dilation effects and a reduction in the death of cells in heart attacks. Something this good in the lab must be good in real life, right?
The early studies
Early studies on patients showed that those given magnesium for AMI were less likely to die. Based on these studies, magnesium was widely recommended and used. Many of the studies were small, but there was one large study (with over 2,000 patients) which showed a statistically significant (p=0.03), 2 to 3% difference (reduction) in the rate of death after AMI in those given magnesium (here and here).
The larger studies
Later, a study with over 58,000 patients showed no benefit from magnesium therapy – not just no significant benefit, no benefit at all (here and editorial here). Due to concerns over that study, a further study with over 6,000 patients was performed (here), which also showed no survival benefit from magnesium. Soon after this, in the early 2000’s, magnesium as a therapy for AMI was dead; magnesium had lost its mojo.
Why did the benefit from magnesium taper off over time (the ‘Decline Effect’), from outstanding lab and animal studies, and somewhat beneficial early human studies, to a complete absence of effectiveness?
It is possible that real differences between studies explain the differences. For example:
1. Animal studies are often not replicated in humans due to biological differences
2. Early studies may have been performed in patients with higher risk (therefore having greater benefit), and
3. The later studies may have included other, better treatments for AMI patients, stealing magnesium’s thunder
These are all reasonable explanations, but so are these:
1. Animal studies are often poor quality, and biased towards showing and effect that might not exist (see previous blog post here)
2. The smaller, earlier studies are more likely to be biased towards showing an effect due to poorer methods (here and here, for example)
3. Publication bias may explain why studies showing no benefit (or harm) from magnesium therapy were not published (as shown here)
The bottom line
Early results showed a small and barely statistically significant benefit, a benefit that everyone was expecting to see from the lab and animal studies. The expectation of a result is a strong driver towards seeing a result, so one possible explanation of the magnesium story is that biased studies, and biased interpretations of those studies, overestimated the benefit from magnesium.