Friday, April 10, 2015

Steve: More Food and Dementia Zaniness

I had thought my previous entry on the craziness in nutrition research would allow me to put that topic away for a while.  But then Andrew and Allison Heid send me a link from the Washington Post with the eye-popping headline:  “Being fat in middle age reduces risk of developing dementia, researchers say.”  I had to check it out.  Unfortunately, I have only been able to download the abstract of the article.  Without the whole article, it’s hard to make sense of what might really be behind the findings, but like the media, I’m going to jump in without all the facts.

Just to make things clear, I do have to talk about the Body Mass Index (BMI), which you probably know about.  It’s the ratio of one’s weight divided by one’s height squared.  If you want to compute your BMI, you divide your weight in kilograms by your height in meters squared.  A “healthy” BMI is considered 18.5 to 25.  I can’t help  the fact that the USA alone in the world maintains the old English measurements, but since you found this blog, you can probably find a simple conversion on the web to translate pounds and inches to kilograms and meters.

So what does the article say:
·       It was a study of nearly 2 million people in Britain aged 40 and older, with information obtained from medical records.  The people in the study were followed on average for 9.1 years.

·       Compared to individuals with a healthy weight, people who were underweight (BMI less than 20 kg/m2) were reported to have a 34% increased risk of dementia.

·       As weight increased, risk of dementia fell, and the very obese (BMI greater than 40 kg/m2) had a 29% lower risk of dementia that people with a healthy weight.

In the Post article, the head of the study, Dr. Nawab Qizilbash, acknowledges that prior research found that having a big belly in your 40s increased your risk of dementia in your 70s.  There is fairly extensive work that shows similar findings.  Dr. Qizilbash argues, however, that his study “overshadows” prior work because of the size of the sample.

So is it time to run out to buy some ice cream and cookies? 
It’s hard to say.  Here are the problems:

·       A big sample is an advantage in some ways and a disadvantage in other ways.  The bigger and more diverse the sample, the less the findings can say anything about any given individual.  In other words, the bigger the sample, the less likely the findings will apply to YOU. 

·       A big sample is a problem in another way, because the standard in research is to accept findings that are statistically significant.  Significance is highly dependent on sample size and with nearly 2 million people, this study would find nearly everything statistically significant.

·       I could carp about the accuracy of the diagnosis of dementia in medical records, but the more important issue is what is encompassed by “dementia.”  It includes Alzheimer’s disease but also several other conditions, such as Fronto-temporal dementias, Lewy Body dementia, vascular dementias.  These various disorders differ in type of brain pathology and so they likely differ in the cause. Each of these categories has sub-types that may differ from each other in important ways.If risk factors or causes of the disease differ across the different types of dementia, as everyone suspects they do, then findings from studies that combine different disease groups together will not give a clear answer.  The findings of this study or any other study like it might depend in part on how many people with each type of dementia are included in the sample, as well as how many cases are there of mixed dementia (two or more types, which is common) .   It could be that too much weight predisposes people to one type of dementia and protects them from another type.  Most of the studies you read about group together people with diverse diagnoses and so it is not at all clear what their findings show.

·       What about the magnitude of the differences reported by this study?  The amount of increased and decreased risk sounds impressive, but as my colleague Eric Loken and his collaborator Andrew Gelman have argued, a large effect in a study does not assure that findings are not due to chance or quirks associated with the sample or how the data were handled.  One of the examples they cite is work that claimed that women at peak fertility in their menstrual cycles were more likely to wear red or pink, presumably to attract a male.  The study reported a large difference, yet it could not be replicated.  In other words, it was a chance finding.  We can begin to trust a finding when it has been found across studies.

So what to do?

Resist the hysteria.  The conflicting findings show that the issue of weight and dementia is still very much up in the air.  Too much and too little weight have lots of known health risks, so a reasonably healthy weight is a good thing, but the findings are too muddled at this point to say that it matters for dementia.

Quell the understandable worry we all have about Alzheimer’s.  I’d recommend as the perfect treatment that you go to Judy’s other blog ( and bake something delicious, like the orange danish shown at the top of this entry.  After all, a little extra weight may not be a big deal.

Here’s the link to the Washington Post story:

Here’s the link to the abstract for the article:

No comments:

Post a Comment