#623

Btw - as we just looked at a lot of MR omega 3 data, I thought I’d share the below with people to help out those in context in this specific case:

I quickly discussed whether MR instruments for omega 3 are good with 4o, and this seems like a case where the instruments are not that good and hence MR may not be that valuable, for example outputs from 4o:

Several genome-wide association studies (GWAS), such as those from the UK Biobank and CHARGE consortium, have identified SNPs associated with plasma or red blood cell levels of EPA and DHA. However, these SNPs typically explain only a small proportion of variance in fatty acid levels (e.g., ~1-5%).

Pleiotropy Risk: Variants in FADS1/FADS2 influence both omega-3 and omega-6 pathways, complicating causal inference.

In cases where the instruments are great (clear, explain large portion of the variance) MR analysis can be the most amazing sets of evidence. Here it may bot be the case that instruments are reliable though.

And especially not for DHA

Instruments tend to be stronger for EPA than for DHA, likely due to more consistent GWAS associations.

Btw - I haven’t looked at the studies, but does seem like the inflammation benefits still is picked up:

Some MR studies show modest protective associations with inflammation markers (e.g., CRP), but these need replication.

Her is 4o’s conclusion

Mendelian randomization instruments for EPA and DHA are valid but modest in strength, especially compared to other traits like LDL or BMI. They’re useful for triangulating evidence, but limited power and pleiotropy must be carefully considered. For targeted questions (e.g., DHA and cognitive aging), stronger instruments or well-powered GWAS may still be needed.

2 Likes
How much omega-3 do you need?
#624

For those in US

not sure if it’s temporary or not, but goodrx has

big discount on this right now (more than 3/4ths off)

IMG_0359
IMG_03591179×1089 86.5 KB

IMG_0360
IMG_03601179×2379 144 KB

https://www.goodrx.com/vascepa?label_override=vascepa&form=capsule&dosage=1g&quantity=120&drugId=36935

2 Likes
#625

I agree with you. MR studies are only as good as the proxies they use.
The other issue are the confounders which can be very strong especially for something like EPA/DHA which are very popular supplements as well as found in the diet.
Weak instruments and strong confounders will produce meaningless outcomes.

1 Like
#626

Alzheimer Prevention Talks: A Conversation with an Alzheimer’s Expert

Virtual Event: Wednesday, Apr. 16 7:00–7:30pm ET / 4:00–4:30pm PT

Don’t miss the inaugural Alzheimer Prevention Talks: APT Webstudy Webinar featuring world-renowned Alzheimer’s expert and Alzheimer’s Therapeutic Research Institute Director, Dr. Paul Aisen, in conversation with a clinical trial participant spokesperson discussing the latest changes to the APT Webstudy, updates on the state of Alzheimer’s research, and much more. This webinar series is designed to keep APT Webstudy participants informed about how their participation is advancing the field of Alzheimer’s research.

If you have any friends or family that may be interested in learning about the importance of participation in Alzheimer’s research, please feel free to share this invitation with them.

What:

  • Updates on the APT Webstudy, including current progress and upcoming changes affecting participation;
  • Why your ongoing participation is vital;
  • Other new developments in Alzheimer’s research that are of interest; and more.

Who:

  • Guest Presenter: Dr. Paul Aisen—Alzheimer’s Therapeutic Research Institute; Director
  • Guest Host: Lupe Morales—APT Webstudy; Co-Chair of the ACTC Research Partnership Alliance

When:

Wednesday, Apr. 16, 2025
7:00–7:30pm ET / 4:00–4:30pm PT

Register here

3 Likes
#627

While this all may be true about MRs, when the RCTs are also in the same direction (negative), hard to ignore the preponderance of evidence here.

1 Like
#628

Thanks. Really good point. I asked to ChatGPT 4.5: ChatGPT - MR Studies on EPA DHA

Most interesting bit:

  1. Direction of bias
    When SNPs explain little variance, there’s a greater risk of Type II error (missing a real effect), not Type I (finding a false effect). So if a strong signal does appear despite low variance explained, it’s often more impressive — though still needs caution.

And the conclusion:

Yes — a significant MR finding can be strong even with SNPs that explain only a small fraction of variance, but only if the instruments are valid and the assumptions are met. In fact, finding significance despite limited variance can indicate a real and potentially important biological effect. Still, MR is best interpreted as part of a triangulation approach with RCTs, biology, and observational data.

So when MR studies find that DHA is a risk factor for colorectal cancer, lung cancer and cardiovascular disease while EPA is protective against depression, coronary heart disease and myocardial infarction: these are HUGE findings, as they were found despite a low variance.

On the other hand, for MR that find no associations (for examples EPA and DHA with dementia) one has to look at the details (F-statistics, effect size, confidence intervals, consistency across methods, etc.) and whether this aligns with RCTs. That’s the case.

2 Likes
Omega-3 fatty acids are a risk factor for colorectal cancer
How much omega-3 do you need?
#629

Sure, and I must admit that @adssx arguments against DHA look more and more convincing.

I was just making a point about MR in general because a lot of people think they are even better than RCTs while they are clearly not. It’s just another tool with its pro and cons like RCTs, epidemiological studies, etc.

There is a specific issue with MR though. Any grad student (anybody even) can use the MR tools and generate a “study” about anything in a few hours and at no cost. Most of them are meaningless and should be ignored but it takes time and effort to rule them out.

That said there are also truly insightful MR studies. It’s just that being an MR study should not be perceived as the highest level of evidence. Same for RCTs and the rest obviously.

2 Likes
#630

That’s helpful re variance being small.

—

There was another big type of issue with Omega 3 MR instruments though:

Pleiotropy Risk : Variants in FADS1/FADS2 influence both omega-3 and omega-6 pathways, complicating causal inference.

What are your and your AIs thoughts on that?

1 Like
Omega-3 fatty acids are a risk factor for colorectal cancer
How much omega-3 do you need?
#631

I do think you want to throw out the baby with the bathwater

For something things where the instruments are clean, clear and explain a lot of the variance they can be hugely important and perhaps more valuable or as valuable as short term RCT in sick people only when you are trying to understand the long term effect and in a population that is healthy or not like that specific clinical trial.

1 Like
#632

My thoughts: I don’t know enough.

ChatGPT 4.5: ChatGPT - MR Studies on EPA DHA

tl;dr: “Cautious conclusions: The presence of pleiotropy means MR results based solely on FADS variants require careful interpretation and additional verification (e.g., using sensitivity methods or other SNPs).”

So did the MR only used FADS variants or other variants as well? I don’t know, for instance this paper does not even mention “FADS”: Medicine

1 Like
Omega-3 fatty acids are a risk factor for colorectal cancer
How much omega-3 do you need?
#633

Unfortunately, the more I dig, the more I’m convinced that DHA is detrimental. I emailed or tweeted my questions to Yassine Hussein (PI behind PreventE4, here), Matt Kaerberlein, Brad Stanfield, Kellyann Niotis, Thomas Dayspring, Rhonda Patrick, and OmegaQuant. Let’s see if one of them gets back to me… Hopefully, someone will prove me wrong but so far it doesn’t look really good for DHA supplementation…

6 Likes
#634

No. I did wrote:

But not all of them.

1 Like
#635

This looks like horizontal pleiotropy according to this paper (SNP’s can affect Omega-3 and Omega-6 to a specific outcome, or maybe one SNP to two factors before the outcome), here it’s education, intelligence, and schizophrenia:

So would need to in that case do Multivariable MR on a possible confounding factor and account for it (in this case Omega-6 if it seems relevant). Is this correct do you think? If so you would check (1) what factor you believe is relevant in horizontal pleiotropy (2) if the study authors did a multivariable MR analysis with that factor, if not, why.

A multivariable Mendelian randomization to appraise the pleiotropy between intelligence, education, and bipolar disorder in relation to schizophrenia 2020

2 Likes
#636

Although we only used 1–3 SNPs as instrumental variables for each PUFA, the SNPs explained a relatively large variation in PUFA levels and they fulfilled the criterion as not being weak instrumental variables (F statistic > 10)

Study: Polyunsaturated fatty acids and risk of Alzheimer’s disease: a Mendelian randomization study 2019

Regarding instrument relevance, we have selected independent SNPs strongly associated with circulating PUFA concentration, which explained from 4.8 to 7.9% of phenotypic variance (mean F statistics 109–201) among the UK Biobank participants (discovery sample). In addition, we replicated these associations in an independent dataset [12] using the same NMR metabolomics platform as the one used in the UK Biobank participants (median sample size 13,516), where SNPs explained 3.1 to 6.9% of phenotypic variance in circulating PUFA. This indicates that bias due to weak instruments is unlikely to be substantial in our analyses, even though bias due to winner’s curse (due to using the UK Biobank to select SNPs and estimate their effect on PUFA) could affect the magnitude of effect estimates.

Study: Role of circulating polyunsaturated fatty acids on cardiovascular diseases risk: analysis using Mendelian randomization and fatty acid genetic association data from over 114,000 UK Biobank participants 2022

Mean F-statistics were 120- 8572 for UKBB fatty acid exposures and 15–6315 for FADS analyses, suggesting these analyses were unlikely to be substantially biased by weak instruments (Table S4). Mean F statistics for EPA analyses were 9, leading to possible weak instrument bias.

Study: Omega-3 fatty acids and major depression: a Mendelian randomization study 2024

–
I checked three MR papers and they seemed good with regards to weak instruments based on what they said.

Do people even publicize if they don’t have a good F-statistic (statistical power)/weak instrument? Might not pass peer review?

As a general rule, an F-statistic >10 indicates that the level of weak instrument bias is likely to be small. F-statistics should not be used to select IVs to avoid overfitting the estimation model. For example, an F-statistic of <10 does not indicate that an IV should not be used but, instead, it should be noted in the analysis that weak instrument bias should be a considered limitation.

https://mr-dictionary.mrcieu.ac.uk/term/f-statistic/

1 Like
How much omega-3 do you need?
#637

add Nicolas Verhoeven, PhD - Physionic - he’s a pure science guy, he doesn’t preach just dissects studies

2 Likes
#638

https://www.physionic.org

His contact page says that he’s only interested in business enquiries! :man_shrugging:

#639

maybe there is a subscriber here on this forum that can comment about it, he responds to his subscribers

2 Likes
#640
2 Likes
#641

It depends on what you are most worried about (based on evidence) … for me I’m not worried about vascular disease as I both monitor for that, and optimize my ApoB and everything else. I appreciate the mouse data on phospholipid DHA, however, I would point out that ApoE4 carriers probably respond clinically and have protection with eating fish, but not taking standard DHA supplements. The only thing that is different is that #1 You are having all the other things in fish - not just the oil, #2 It is in phospholipid form.
So I’m more worried about my brain as I have no vascular disease and can monitor for developing it - so have more fish and I’m going with the phospholipid forms of DHA, but in a mix that still has plenty of EPA.

This is simply individualized risk/benefit. Obviously, my opinion would be different for someone with heart disease and no ApoE4.

I’ll wait for more data on the phospholipid forms … and it might prove useless in the end.

3 Likes
#642

I agree that eating fish is the best way.

But what about the CVD and cancer risk in Mendelian randomization studies? Vitamin O (Omega 3) for athletes - #4 by adssx

1 Like