Food and Behaviour Research

Donate Log In

19 February 2015 - The Conversation - How a simple vitamin B prescription could help people with Alzheimer's

Peter Garrard, Reader in Neurology at St George's, University of London


A recent 'meta-analysis' - pooling results from a number of randomised controlled treatment trials - concluded that B vitamin supplementation does not reduce the risk of cognitive decline and dementia.
Extraordinarily, however, this much-quoted meta-analysis does not even include the major clinical trial carried out at Oxford University (the VITACOG study) - which was specifically designed to invesitgate this question, and found significant benefits from B vitamin supplementation. Instead, it includes other trials that were simply not even capable of addressing the question.

This article - by an expert in the field - explains how publication of this meta-analysis effectively prevented a consortium of leading UK researchers from obtaining funding for studies to follow up the very promising VITACOG findings.

For further information about the VITACOG trial, and the OPTIMA project of which it formed a part, see:

The VITACOG trial, a preliminary clinical trial in subjects with high plasma homocysteine levels, showed that the brains of those who received B-vitamins shrank significantly less rapidly than those of the placebo group, particularly in areas that are associated with early pathological changes in Alzheimer's.

Such a striking result seemed to indicate the need for a nationwide trial to test whether the outcome would translate into a clinically important disease-modifying effect on the rate of progression in mild cognitive impairment.

The arguments for conducting the trial were overwhelming and, with the assistance of a national network of experts in dementia and clinical trials, I prepared the scientific and economic case for funding. Opposition to the idea, however, appeared from an unexpected quarter – a meta-analysis of cognitive outcome data taken from completed trials of B vitamins for stroke and heart attack prevention. Somehow, a statistical miscellany of recycled results was rapidly elevated to a status little short of definitive scientific proof.

Meta-analysis can be a powerful way of drawing robust conclusions from the results of an experiment that has been conducted multiple times on small populations.

In the B vitamin case, the numbers included in the meta-analysis were impressive. Yet numbers mean nothing if the data is neither uniform nor directly relevant to the question. Closer scrutiny revealed that few of the trials focused on dementia, that the ages of patients who took part were well below those associated with the development of Alzheimer’s disease, and that inclusion did not require the presence of mild cognitive impairment. And astonishingly, the results from the VITACOG trial were not included.

Unsurprisingly, the outcome of the pooled analysis was anodyne: neither the treatment nor the placebo group showed any meaningful change on any measure of cognitive status during follow-up. In other words – to quote one of the members of the original VITACOG study team – the analysis merely demonstrated that: “taking B vitamins won’t prevent cognitive decline in those who overall, do not show cognitive decline anyway.”

Yet the absence of any difference between the two treatment arms has been erroneously, widely, and without qualification interpreted as evidence against the benefits of B vitamins in Alzheimer’s disease.

In recently published letters to the journal, I and many colleagues from around the world have pointed out the flaws in the meta-analysis and its harmful misinterpretation: harmful for medical research, for the cause of dementia prevention and, most of all, for the thousands of individuals who could have benefited from a safe and simple intervention.

We remain convinced that the clinical benefits of B vitamins in groups at high risk of Alzheimer’s disease should be allowed to be rigorously tested, but the damage caused by the public misinterpretation of a null study will take a while to unpick.