Search This Blog

Monday, 2 January 2017

Scientific Fraud, in the Abstract

Someone posted this scare story from November in the Low Carb and Paleo group on Facebook.
Luckily I missed it at the time as I was busy with arguably more important things, but being on holiday now I think it warrants a little attention.
It was an abstract - that's all for now - presented at a conference of the American Heart Association, based on data from the WHI study (Womens' Health Initiative - the follow-up from the huge, long-term study where lowering saturated fat in the diets of women seemed to cancel out all the expected benefits from improving carbohydrate quality and reducing trans fat intake).

It generated headlines like 

Mostly meat, high protein diet linked to heart failure in older women (AHA)
American Heart Association Meeting Report – Presentation: 627 – Session: EP.RFO.28

NEW ORLEANS, Nov. 14, 2016 — Women over the age of 50 who follow a high-protein diet could be at higher risk for heart failure, especially if much of their protein comes from meat, according to preliminary research presented at the American Heart Association’s Scientific Sessions 2016.

Researchers evaluated the self-reported daily diets of 103,878 women between the ages of 50 and 79 years, from 1993 to 1998. A total of 1,711 women developed heart failure over the study period. The rate of heart failure for women with higher total dietary protein intake was significantly higher compared to the women who ate less protein daily or got more of their protein from vegetables.

While women who ate higher amounts of vegetable protein appeared to have less heart failure, the association was not significant when adjusted for body mass.

High-protein diet linked to heart failure in older women (CNN)

and so on.

None of these press releases (there were others) linked to any paper, but a bit of searching brought up the actual abstract in a doc.x format.

It says 
Results: Among 103,878 women in the study sample, 1711 women developed HF through 2005. Incremental biomarker calibrated dietary protein consumption was associated with an increase in the risk for Heart Failure. An inverse association was found between higher intakes of energy adjusted vegetable protein and HF although this association wasn’t statically [sic] significant if the association was adjusted to BMI and diet quality.

But what you find here is also that (a) the non-significant association mentioned above was no longer even inverse once fully adjusted, and (b) - in my words -
"An inverse association was found between higher intakes of energy adjusted animal protein and HF, although this association wasn’t statically [sic] significant, if the association was adjusted to BMI and diet quality."

Protein intake was verified with biomarkers, but these were incapable of distinguishing between animal and vegetable protein. The amount of vegetable protein in each quintile, if this study is typical, would have been about 1/3 the amount of animal protein in the corresponding quintile, with the amount of vegetable protein consumed by even the upper quintile being insufficient to support life, so a valid comparison of this sort is not possible. This difference alone is enough to explain the variance in the unadjusted results.

What about total protein? This set of results is not adjusted for BMI 

This matters because heart failure, as a disease of ageing, may be associated with poor appetite, which would tend to increase the protein percentage of the diet. The same is true of heavy drinking, a known cause of cardiomyopathy and heart failure. Overweight also increases the risk of heart failure, as shown by this Mendellian randomisation study.

The results show that an increase of one unit of BMI increases the risk of developing heart failure by an average of 20 per cent.

If that's so, then a high protein, low carb diet would reduce the risk. Mechanistically I can't think of a reason why protein would cause heart failure, and none is mentioned in the abstract. Ketone bodies, a product of protein metabolism even on a high carb diet, are the preferred fuel of cardiac muscle.

In any case, adjustment for BMI is essential, and so is adjustment for diet quality (especially the B vitamins essential for the metabolism of amino acids). Vitamin deficiencies are associated with heart failure, which makes sense mechanistically, and the AHEI 2010 score is one of the diet quality algorithms, confounded by concepts of virtue (it would be better if it focussed on essential nutrients in this case).
Do that, and there's obviously nothing to see.

But even when there's nothing to see, there are plenty happy to report it, as long as it fits the plant-based bullshit bias of the health-reporting media and the AHA.

1 comment:

honora said...

Apposite choice of song. It's good to have these reminders of how scientific fraud can occur. Thanks for your sleuthing and explanation of how it occurred.