Not all "good cholesterol" particles looked alike when it came to dementia risk

HDL is often called "good cholesterol", and in general higher HDL cholesterol has been linked to a lower risk of dementia. But some studies found the opposite at very high levels. A study led from the University of Minnesota suggests one reason: HDL is not one thing. In more than 23,000 older people in the UK, the smallest HDL particles were linked to a slightly lower dementia risk, and the very largest ones to a slightly higher risk. The study shows associations only.

Some background. Fats such as cholesterol do not dissolve in blood, so they travel inside particles called lipoproteins, packages of fat wrapped in protein. HDL (high-density lipoprotein) particles collect excess cholesterol from tissues and carry it back to the liver. They come in different sizes, from small to very large, and carry different mixtures of fats. The usual blood test, "HDL cholesterol", only measures how much cholesterol they carry in total. The team suspected that the different sizes might relate to dementia risk in opposite ways.

What they did

The main analysis used the UK Biobank, a large British health study that recruited people between 2006 and 2010. For part of it, blood samples had been analysed with NMR spectroscopy, a technique that uses magnetic fields to estimate how many lipoprotein particles of each size are in the blood and what fats they carry. The team included 23,787 participants aged 67 or older (average age 68) who did not have dementia. HDL was split into four size groups: small, medium, large and very large. Dementia diagnoses came from self-reports, hospital records and death records.

A second, much smaller analysis used 147 people from ARIC, a long-running US study. Here, HDL was physically separated into two fractions in the lab (one with smaller, one with larger particles) and hundreds of individual fat molecules were measured. The authors describe this part as "hypothesis-generating".

The analyses took into account age, sex, ethnicity, the APOE4 gene variant (a known genetic risk factor for Alzheimer's disease), education, body mass index, smoking, diabetes, blood pressure, and the use of blood-pressure and cholesterol-lowering drugs.

What they found

In the UK Biobank, 1,626 people (6.8%) developed dementia over an average of 12.6 years.

• Small HDL, lower risk. More small HDL particles, and more of each kind of fat in them, went with a roughly 9% to 11% lower dementia risk for each step of one standard deviation (a measure of the usual spread between people). This link was driven mainly by the number of particles.
• Very large HDL, higher risk. More very large HDL particles, and more cholesterol and other fats in them, went with a roughly 10% to 14% higher risk per standard deviation. Here both the number of particles and what they carried seemed to matter.
• Total HDL cholesterol was not linked to dementia risk in this group.
• Robust to some checks. The links barely changed when the first two to five years of follow-up were left out, which reduces the chance that early, undiagnosed dementia was changing blood fats rather than the other way round. Accounting for deaths from other causes did not change them much either.
• Cholesterol drugs. For one type of fat (triglycerides) in some HDL sizes, the links seen in people not taking cholesterol-lowering drugs disappeared or reversed in those who did.
• The small US cohort. In ARIC, some fats in both HDL fractions were linked to a lower risk, but these results were only borderline after correcting for the many comparisons.

What to keep in mind

• Observational. These are links between a single blood measurement and later dementia. They do not show that changing HDL particles would change anyone's risk.
• Small effects. A difference of about 10% per standard deviation is modest for an individual.
• Two cohorts that can't be compared directly. The authors stress that the two cohorts used different methods, the lab fractions in ARIC do not map onto the NMR size groups, and the populations differed (the ARIC group was older, more diverse and more often had diabetes and high blood pressure). The UK Biobank results, they write, still need to be repeated in an independent cohort.
• A small, non-random sample in the US. Because of lost freezer samples, the 147 ARIC participants were fewer than 10% of those eligible and not strictly a random selection.
• No function measured. The study did not test what the different HDL particles actually do, so the explanations the authors offer (for example, that small HDL may help blood vessels or reach the brain) remain untested here.
• Mostly white participants. In the UK Biobank group, 97% were white.

Why it matters

The standard cholesterol test lumps all HDL together, and that may hide opposite signals. This study suggests that, at least for dementia, the size of HDL particles may tell more than their total cholesterol. That is a lead for research into how blood fats relate to brain aging, not a reason to change anything today: there is no test or treatment based on HDL size for dementia, and the findings need to be confirmed elsewhere.