"This study shows blood glucose within the normal range is associated with diseases"
Let's review a Mendelian Randomization Study together
One of the things I enjoy on Twitter is people highlighting studies that I haven't read, or sometimes haven't read for a long time. Very often, it's people highlighting these studies to disagree with something I've said. I actually think this is kind of fun, and the only way I can ever remember what studies say is reading them over and over again, so I appreciate having to go back and double check and read something.
This happened this week with someone who tweeted me a sophisticated type of study called a Mendelian Randomization Study which claims to show that blood glucose within the normal range is associated with diseases including retinopathy, neuropathy and arterial disease. I was tweeting about the use of continuous glucose monitors in people without diabetes, and making my usual points about the fact that we don't have any data suggesting blood glucose excursions in people with normal glucose tolerance are anything to worry about.
So I got tweeted this study, and I thought it would be a cool thing to write about, because it seems to be the kind of high quality study - that on face value - does in fact seem to show that blood glucose within the normal range is positively associated with some negative outcomes. But when we dig into the weeds, all is not what it seems.
Let me explain why I think this study has important limitations.
What are Mendelian Randomization (MR) Studies and why do people think they’re awesome?
They allow us to study the relationship between an exposure (could be a behavioural exposure or biochemical) and an outcome based on genes associated with that exposure. The idea is that by doing it this way, we can control for confounding variables.
The strength of a study design that controls for confounding variables is particularly clear when we consider lifestyle behaviours. For example, if we try and look at the relationship between physical activity and heart disease in observational studies, the people who do lots of physical activity tend to have better jobs, be more economically secure, live in a nicer area, are able to afford/prepare/consume healthier foods and etc. So it's hard to be confident that it was the physical activity alone that helped prevent heart disease. But by using the genetic variants (which are randomly inherited) associated with physical activity*, we can examine the independent relationship between our variable of interest (physical activity) and heart disease. (Note: We couldn’t do this if the same genes associated with physical activity are also associated with having a good job, and living in a nice area and eating a healthier diet).
You might be thinking “well, this makes sense for behaviours but surely all we need to do for biochemical risk factors like glucose is measure it??”. Examining the relationship between a biochemical risk factor like glucose and an outcome should be straightforward – we take tens of thousands of people and measure their blood glucose over time and health outcomes 10 or 20 years later, look at the relationship between the two and we have our answer. However, it is possible that an unknown factor, let's call it Factor X, is leading to both elevated glucose and heart disease; it just looks like high blood glucose is causing heart disease:
The idea behind MR is that if we can look at the genetic variants associated with high blood glucose (again, this only works if the genetic variants associated with glucose are not associated with other risk factors for heart disease) then we can look at the independent relationship between glucose and heart disease. It’s like an RCT, except the randomization comes from randomly inheriting alleles (alleles are different versions of the same gene) vs randomizing people to a treatment/exposure arm.
The problem is that many genes code for proteins which have broad effects.
And this is one of the problems with the study I got tweeted.
As a reminder the study was using genetic variants associated with elevated glucose to “evaluate whether high glucose levels in the normoglycemic range and higher have a causal genetic effect on risk of retinopathy, neuropathy, nephropathy, chronic kidney disease (CKD), peripheral arterial disease, and myocardial infarction (MI)”. The study conclusions were that, yes, glucose levels in the normoglycemic range and higher were prospectively associated with these negative health outcomes.
The genetic variants included were: GCP62/ABCB11, GCK, DGKB, ADCY5, CDKN2A/B, and TCF7L2. Remember that genes code for proteins. For example the GCK gene provides instructions for making a protein called glucokinase which is found in the beta-cells of the pancreas. Glucokinase acts as a glucose sensor- if the glucose concentration in the blood goes up, glucokinase kicks off a cascade of molecular processes which ultimately lead to the release of insulin from the pancreas. Depending on the particular mutation in the GCK gene, blood glucose elevation can be mild or quite severe. So far, so good – if we want to examine the relationship between glucose and health outcomes by using genes associated with glucose, then GCK seems like a good one.
However, some of the others - CDKN2A/B and TCF7L2 are associated with impaired lipid metabolism, and ADCY5 might be involved in a few pathophysiological processes. I am not sure of the exact mechanism of action of the proteins coded by these genes but it looks like both CDKN2A/B and TCF7L2 are involved in hepatic lipid regulation, and people with a particular variant of TCF7L2 have elevated blood lipids.
So we can’t be sure we’re looking at just the relationship between glucose and the negative health outcomes here. Lipids could be a confounder. (And note it's not just glucose that contributes to the development of neuropathy and retinopathy, elevated lipids also do).
We also need to consider by what mechanism the genetic variants are leading to elevated glucose and could those same mechanism(s) potentially impact other risk factors for neuropathy and arterial disease?
Remember I said that one of the genetic variants selected was glucokinase which regulates insulin secretion from the pancreas? Well, so do most of the others. Why does this matter? Because if the insulin secretory response to glucose is altered, it doesn’t just affect the glucose concentration, it affects the post-prandial lipid concentration too. So again, we can’t be sure we’re looking at just the relationship between glucose and the negative health outcomes here. [Post-prandial] lipids could be a confounder.
Note that this effect is indirect (in contrast to these genes directly altering lipid metabolism). Also note that I am presuming this happens! As far as I know we don’t have data on these genetic variants and post-prandial lipids (whether triglycerides, non-esterified fatty acids etc). Genome wide association studies (which look at the relationship between genes and diseases in thousands of people) tend to collect fasting measures of things like glucose or lipids. To understand how genes affect postprandial metabolism we’d need investigators to run meal tolerance tests as well. These are time-consuming (and therefore expensive) so they are carried out much less frequently.
So, yes this study looks very sophisticated on the surface, because they've used - woo! - genetics to isolate the independent variable glucose, but if we consider closely the role of each of the genes, we can see that they are not simply looking at a rise in blood glucose, they are looking at a mixture of insulin secretory deficit with added lipid impairment.
So this study doesn’t change my view that people with normal glucose tolerance probably don’t need to worry about their glucose excursions. So enjoy your oatmeal.
*As far as I know there is good evidence that there are genes associated with physical activity but this is not my area so let me know if subsequent research has shown this to be incorrect. I just use the example here to show why MR study designs have important methodological strengths.
*
Fascinating and excellent. Thank you for this!
Thank you! I love oatmeal and eat some almost every day! :-D