CAD: presumed mechanism and apoB was best predictor of CAD, by mendelian randomization

 

Will review my understanding of the pathogenesis of coronary artery disease. Then a review of a mendelian randomization (MR) study on lipids, finding that measurements of apolipoprotein B was the most useful one to do (see cad risk lipids mendelrandom pLos2020  in dropbox, or https://journals.plos.org/plosmedicine/article/file?id=10.1371/journal.pmed.1003062&type=printable 

 

Details: 

-- 141,016 participants from the UK Biobank who had genome-wide association studies (GWAS) to find genetic variants reliably associated with lipid and apolipoprotein concentrations 

-- mean age 57, 54% women 

-- mean nonfasting lipid levels: LDL 138 (3.57 mmol/L), HDL 56 (1.45 mmol/L), triglycerides 133 (1.50 mmol/L)

-- mean apolipoprotein B (apoB) 1.03 g/L, apolipoprotein A-1 (apoA1) 1.54 g/L 

-- they performed univariate mendelian randomization (MR), the traditional one, as well as multivariate MR, a recently developed extension that allows for an assessment for SNPs of multiple risk factors simultaneously, permitting an estimate of the direct effect of each SNP on the clinical outcomes (in this case, for example, disentangling the SNPs for apoB, LDL, and triglycerides in order to see which of these were in fact responsible for atherogenesis/clinical CHD)

-- coronary heart disease (CHD) was defined as the development of MI, acute coronary syndrome, chronic stable angina, or coronary stenosis of at least 50% 

 

Primary outcomes: 

-- to identify those SNPs in the different lipid moieties that were associated with CHD outcomes by traditional univariate MR

-- to assess multivariate MR to determine the lipid moieties that independently were associated with CHD outcomes

 

Results: 

-- SNPs found (these reflected hundreds of hitherto unknown variants: 56% to 93% of SNPs identified for each lipid trait not having been previously reported in large-scale GWASs): 

        -- LDL cholesterol: 220 genetic variants

        -- apolipoprotein B: 255 genetic variants

        -- triglycerides: 440 genetic variants

        -- HDL cholesterol: 535 genetic variants

        -- apolipoprotein A-1: 440 genetic variants

    -- 42% of these SNP’s were associated with more than one lipid-related trait

 

-- univariate analysis for higher risk of CHD (odds ratio per one standard deviation higher lipid trait): 

    -- LDL cholesterol, odds ratio 1.66 (1.49-1.86), p<0.001 

    -- triglycerides, OR 1.34 (1.25-1.44) , p<0.001 

    -- apolipoprotein B, OR 1.73 (1.56-1.91), p<0.001 

    -- HDL cholesterol, OR 0.80 (0.75-0.86) , p<0.001 (ie protective)

    -- apolipoprotein A-1, OR 0.83 (0.77-0.89), p<0.001 (ie protective)

 

    -- multivariate mathematical analysis:

        -- apolipoprotein B was the only robust lipid moiety associated with CHD, with a two-fold increased risk, OR 1.92 (1.31- 2.81), p<0.001 . after controlling for apolipoprotein B, all of the other evaluations markedly attenuated to being nonsignificant 

 

 Commentary:

-- First, as a background to the above (ie, what is the mechanism for atherosclerosis):

    -- every observational study I have seen has found that apolipoprotein B, when measured, to be more predictive of clinical cardiovascular disease than LDL concentrations

    -- the definition of LDL versus HDL is quite a crude one: electrophoretic patterns with a pretty arbitrary boundary, dividing several different lipoprotein particles into the broad categories of low vs high density

    -- a couple of large epidemiologic studies have suggested that small dense LDLs are much more atherogenic (3-fold in a Québec study) than larger LDLs

    -- these small dense LDL particles are associated with higher levels of total serum apolipoprotein B: one apoB is present on each LDL particle, both for its structural support of the LDL but also as the binder to LDL receptors. Therefore, a patient with a high apolipoprotein B level for the same LDL concentration as others will generally have small dense LDL particles

    -- it is these small dense LDL particles that are much more easily oxidized, and it is the oxidized LDL that is taken up by macrophages: the macrophage receptors bind preferentially to oxidized LDL over non-oxidized LDL, similarly with oxidized lipoprotein(a) over nonoxidized. And these macrophages then progress to becoming foam cells, ultimately leading to atherosclerotic plaques and clinical CHD outcomes. for a recent review of small dense LDLs, including their high CHD risk even in those with seemingly "normal" LDL levels, see https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9025822/

    -- the European cardiovascular guidelines from 2019 recommended that all persons should have an apolipoprotein B level measured, if available; the 2018 ACC/AHA guidelines in the US do not recommend routine testing

    -- given that we in the United States do not routinely measure apolipoprotein B levels, it turns out that there are still a few cues to high apolipoprotein B based on other factors:

       -- patients with certain high-risk factors for atherosclerosis (eg those who are overweight/obese, have metabolic syndrome or diabetes) are much more likely to have small dense LDLs. This supports the American Diabetes Association guidelines to prescribe statins to all patients with diabetes, given their high risk for atherosclerotic disease independent of their LDL concentrations (sometimes their untreated LDL levels might seem to be already at the LDL target)

        -- though there are some mixed results in the medical literature, it also seems likely that smoking is associated with small dense LDLs (see https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8532056/ )

        -- and, by the way, an impressive observational systematic review found that meds leading to LDL levels as low as 21 mg/dL in those with advanced CAD are associated with increased cardiovascular protection, without evident harm: see http://gmodestmedblogs.blogspot.com/2018/08/very-low-ldl-levels-benefit-without-harm.html, as well as a specific analysis of the FOURIER study: http://gmodestmedblogs.blogspot.com/2021/12/ldl-less-than-40-in-high-risk-patients.html

        -- it turns out that small dense LDLs are also associated with high triglycerides and low HDL levels. The directionality of the relationship between triglycerides and small LDLs is unclear, though the relationship has been confirmed in many studies, for a more recent study see https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8509760/ 

        -- the Québec study also found that those with triglycerides <132 were very unlikely to have small dense LDLs, whereas those with triglycerides >176 were much more likely

        -- a study by Hanak et al in 2004 (Am J Cardiol. 2004 Jul 15;94(2):219-22. doi: 10.1016/j.amjcard.2004.03.069) found that 79% of those with a triglycerides/HDL ratio >3.8 had LDL phenotype B (small dense LDLs) versus 81% of those with a ratio >3.8 had phenotype A (large LDLs)

    -- so, it still makes sense to measure apoB levels directly. but in the absence of this, the above seem to be helpful (but imperfect) substitutes

 

-- And there are similar issues with HDL, also with a variety of different HDL particles lumped together in this broad category

    -- there clearly are some HDL particles that are atherogenic, and do not have the usually expected protective effect

        -- some have a different apolipoprotein, apoC3, which is pro-inflammatory and is associated with atherosclerosis (as opposed to the typical and more usual apoA1, which seems to be more cardioprotective than the HDL levels, similar to the relationship be LDL and apoB)

            -- it is notable that previous clinical trials of CETP inhibitors, which actually do lead to striking increases in HDL levels, are not cardioprotective and some have actually been atherogenic.  But, these HDLs have apoC3!!! (see lipids CETP inhibitors increase ApoC3 AtheroThrombVasc2022 in dropbox or DOI: 10.1161/ATVBAHA.121.317181 )

        -- there are other dysfunctional HDL particles that have been found

    -- As a related issue, there also can be ApoC3 variants on triglycerides that are associated with increased coronary artery disease (see https://lipidworld.biomedcentral.com/articles/10.1186/s12944-021-01531-8 )

 

 

--so, what does this all mean? A lot of this information has been available for more than decade arguing that LDL and HDL are not such great surrogate markers for atherogenic risk: eg, see http://gmodestmedblogs.blogspot.com/2014/08/current-controversies-in-lipidsldl-hdl.html 

    -- based on many observational studies (and leading to the European recommendation to check apoB levels), it does seem that we should be measuring routine apolipoprotein B levels, assuming the assays we have are accurate and reproducible.

-- But, I am still a bit concerned about dismissing the potential benefits of HDL or apoA1 levels: so many studies have shown a relationship between HDL and lower levels of cardiovascular risk, and it has been incorporated in many of our cardiovascular risk predictive models. Several studies in the past have found that those on statins who had a lower total ratio of cholesterol:HDL levels had better clinical outcomes. And the Treating to New Targets trial, although they did not track apolipoproteins, did find that there was a clear relationship between LDL and HDL associated with cardiovascular events: an LDL >100 with a high HDL had the same 5-year clinical cardiovascular risk as an LDL <70 with a low HDL:

 

    -- perhaps the complexity of the composition and function of HDLs in one study might conceivably distort the large population study showing benefit of HDL (ie, was the UK Biobank sample not really representative of the population elsewhere? did they have more people with apoC3 on their HDL particles, though that was not tested in the above MR analyses??). i would be nice to have GWAS analyses from a large international database, and also checking for the apoC3 levels as well as apoA1

    -- and, perhaps we should be measuring apoA1 levels instead of HDL???? (needs to be tested in studies, of course)

 

-- this segues into the MR analyses above:

    -- there was a very impressive linkage in the traditional (univariate) MR assessment of the lipid moieties and the expected cardiovascular outcomes: SNPs associated with LDL and apoB levels had worse clinical outcomes, and those with HDL and apoA1 had better ones 

    -- multivariate MR found that the only thing that mattered was the apoB levels, which certainly reinforces that we should be checking these...

    -- there are still some lingering concerns with even the multivariate MR, though it does allay some of my prior concerns about the interplay of different SNPs as per prior blog: http://gmodestmedblogs.blogspot.com/2023/04/mendelian-randomization-alcohol-does.html (this blog also explains mendelian randomization in more detail). some of my continued concerns (though would appreciate any comments from those more savvy on this issue):

        -- the current analysis only includes the interaction between the different lipid moieties. we do know that there are other interactions with lipids leading to clinical heart disease, including smoking, drinking alcohol, hypertension, diabetes, psychological stress, anxiety, depression, etc etc. Do SNPs associated with these important cardiovascular risk factors have any interaction with the lipid ones? (i will be sending out an MR report on many of these risk factors in the next, and last for awhile, MR blog).  is there a role that epigenetics plays affecting cardiovascular risk? (some studies on these issues are noted in the alcohol blog above. epigenetics, as opposed to SNPs, reflect the effect of environmental factors, such as stress etc listed above, on the genome, epigenetic changes can turn on or off the function of segments of the genome, sometimes can be passed from one generation to the next, and bypass the egg/sperm and are missed by conventional DNA sequencing)

        -- and, of course, these cardiovascular risk factors found by genetic analysis are not determinant. for example, a study a few years ago found that especially in patients who had high genetic predisposition to weight gain, a healthier diet was actually more effective in their weight management than in those with a lower genetic risk (see http://gmodestmedblogs.blogspot.com/2018/01/dietary-effect-strong-when-high-genetic.html ). And, for most of the common diseases, there is typically a significant interplay between genetics and the environment

        -- i also have some concerns about the statistical validity of a study like the multivariate MR one above:

            -- there are huge numbers of SNPs with many having complex interrelations of the different lipid moieties (of the 846 involved, 352 (42%) were connected with at least one other of the lipid moieties). How well does the mathematical process of multivariate MR actually disentangle this web? do the estimated associations with real clinical events later truly reflect reality?  we do know, for instance, as in the alcohol blog noted above, that some SNPs are more predictive of clinical outcomes than others. do we know which of the hundreds of SNPs are the most related to clinical disease and is that incorporated into the model? do we know if the most related ones for apoB, for example, are associated directly with ones that are marginally associated with LDL? It seems likely that the large increase in SNPs identified is a good thing for our understanding, but is it too much information and introduces too much "noise" in the analysis??

            -- it is also pretty striking that this study found hundreds of hither-to unknown SNPs. a gigantic increase. will studies later, perhaps using more sensitive technology, usurp the results of this one by finding more SNPs that are more tightly linked with clinical outcomes (and perhaps add on those associated with weight, diet, exercise, smoking, stress, anxiety, depression......)? See the next MR blog of a review of the known cardiovascular risk factor associations with clinical disease....

            -- is apoB the best marker of atherosclerosis? we know that inflammation plays a strong role in CHD events. and we know there is a lot of variability in different patients as to how much of an inflammatory response they develop when in a proinflammatory state. and we know there is a complex relationship between triglycerides and oxidized LDL, with some patients at the same triglyceride levels having different LDL sizes. is there some complex interplay between LDL, triglycerides, and proneness to more dramatic inflammation or oxidation part of the picture?

    -- so, I suspect there may be decreased potential for statistical validity of the results in this study since there were just so many SNPs analyzed, so many are interconnected with others and not unique, so many of these SNPs were newly discovered, the intensity of the mathematical analysis for controlling for so many of these SNPs (for example we know that controlling for more and more items in traditional non-MR multivariate analysis tends to lead to decreased association with clinical events...), and not including the non-lipid SNPs that we know affect clinical outcomes, all might  decrease the statistical validity of the findings in this study.

 

So, why is it important in primary care to discuss mendelian randomization studies?

-- They do provide an important methodology to understand highly likely causality of associations when RCTs cannot be done

    -- for example, we could not do a randomized controlled trial where half the population drank different amounts of alcohol for years and we looked at outcomes. Or consuming different amounts of coffee, or chocolate. Or if we can’t randomize people, such as those with high LDL or high HDL levels.

-- to the extent we control for all of the potential confounders (some known, others not) through multi-variate analysis, MR provides significant insight into causality (large traditional observational epidemiologic studies can only provide associations, not causality)

-- MR studies may reveal linkages between genetic factors that had not been suspected, improving our understanding of diseases and potentially changing our clinical approaches

-- however, for the vast majority of common genetic diseases (overweight, diabetes, hypertension, hyperlipidemia etc.), there is a complex interplay between the environment and genetics. And sometimes, as in the obesity study above, environment plays a leading role over genetics. This all means that we should not be swayed into genetic determinism. And my concern here is that the increasing use of these and other genetic analyses will push patients as well as us folks into taking a fatalistic approach. Environmental changes are important. And they may also very well affect genetics and epigenetics, and significantly change a person’s proclivity to many important diseases

-- and this study does highlight that, at this point, we probably should be checking apoB levels in patients: that would be an important basis for confirming if there is a real association with cardiac events, if lowering apoB levels is cardioprotective and at what level, and perhaps even targeting new meds to decrease apoB levels..... we do have some information so far: apoB levels can be reduced through diet and most meds (eg statins, fibrates, ezetimibe, PCSK9s)

 

that all being said, I do plan to do one more MR blog assessing studies evaluating the multitude of cardiovascular risk factors. And I promise it will be shorter than this one (but I did think it was important to review the atherosclerosis pathogenesis….)

 

geoff

-----------------------------------

If you would like to be on the regular email list for upcoming blogs, please contact me at gmodest@bidmc.harvard.edu

  

to get access to all of the blogs (2 options):

 

1. go to http://gmodestmedblogs.blogspot.com/ to see the blogs in reverse chronological order

  -- click on 3 parallel lines top left, if you want to see blogs by category, then click on "labels" and choose a category​

  -- or you can just click on the magnifying glass on top right, then type in a name in the search box and get all the blogs with that name in them

 

2. go to https://bucommunitymed.wpengine.com/, a website from the Community Medicine section at Boston Medical Center.  This site does have a very searchable and accessible list of my blogs and is the easiest to view blogs and displays more at a time.

 

if you would like to see the article, please email me. 

 

please feel free to circulate this to others. also, if you send me their emails (gmodest@bidmc.harvard.edu), i can add them to the list