fish oil or real fish?

[lillo9546]

Hi there!
In the search of omega 3 facts, I found this https://chriskresser.com/the-fish-vs-fi ... smackdown/ which also explain monetary comparision!

What do you think about omega 3 /fish oil supplements intake?

[Anesau]

I'd choose algae oil supplements for omega 3s - contains EPA and DHA just like fish oil (algae is where fish get their EPA & DHA), rather than just ALA omega 3s like flax/walnuts/canola/etc.

Comparing supplements, it looks like cheap algae oil is 872mg DHA/EPA per dollar, while cheap fish oil is 4930mg per dollar. Algae oil might be better for environmental / ethical reasons, but it does seem to cost more. Between fish oil and just eating whole fish, fish oil is more DHA/EPA per dollar, but I'd think fish itself would be preferable - part of the increased cost is paying for all the other nutrients/calories in the fish, and (as a personal opinion) it's better to get nutrients from the regular diet where possible.

ALA omega 3s are the cheapest and easiest. They're in a bunch of stuff most people consume for unrelated reasons, like canola oil and walnuts. However, their usefulness partially depends on demographics -- young men convert only 8% of ALA to EPA and <=4% to DHA, while women convert 21% and 9% respectively [Oregon State]. EPA and DHA are only conditionally essential nutrients; except for stress/aging/other extenuating circumstances, most people and especially most women don't need to worry about supplementing EHA/DPA as long as they're eating a normal diet with sufficient ALA sources.

Since "essential fatty acid deficiency in healthy individuals in the United States is virtually nonexistent" [NIH], my main takeaway is that supplementing is probably not worth it unless you have a very compelling health reason and/or it's recommended by a doctor.

[black_son_of_gray]

Google Scholar is your friend, here. With respect to omega-3 fatty acids and (cardiovascular) health, I would start with the publications of William Harris. For context of the importance of omega 3's in the diet (and how much/what kind you might want to consume), a wonderful place to start might be his paper, "The Omega-3 Index: a new risk factor for death from coronary heart disease?" which shows up on the first page of results (for me at least). Check out the PDF, look at figure 5, then pick your jaw up off the floor if you need to. Anyway, that's a good place to start.

[Sclass]

Figure 5 in the reference above was an eye opener.

I eat the capsules. Mostly for the taste. I like the fishy flavor of the oil. I get this brand that has a strong mackerel flavor.

I eat a lot of pickled mackerel. Not for health. I just like it. Even with fatty Atlantic mackerel I don’t think I can get as much oil as with a daily capsule. The capsules are a big dose.

[Anesau]

Interesting! I looked around at the lit reviews for DHA/EPA in general; I thought I remembered a null / not statistically significant result last time I looked, but maybe I was over-indexed to the Chochrane review. It definitely looks like DHA/EPA are more important than I previously thought.

I'll admit to having been a bit unfairly suspicious of William Harris at the start given the replication crisis & his findings seeming to make much stronger claims than others', but after digging deeper it looks like most of the literature agrees with the direction he's moving.

Thank you for adding more info!

[IlliniDave]

I take a fish oil supplement and an additional DHA supplement along with it. The only way I could reach my target intake is eating fatty fish 5-6 days per week, which is too much. For me it's not so much about cardiovascular risk as it is about other potential benefits. Seems to blunt a lot of my environmental allergies is the most notable one day-to-day.

[Anesau]

There's a lot of weird interactions with vitamins like that. My mother gets told she probably has cancer about every other year (though she never has actually had cancer); for some reason, taking extra vitamin E makes it stop looking like she has cancer. For chronic/intractable health problems, "experiment with safe supplements" seems like a decent strategy where regular options fail.

I have lower-hanging fruit if I get concerned about my heart, but maybe I'll join the EPA/DHA train for vaguer might-as-well reasons. I get supplements for free as Christmas and birthday gifts anyways, because me & my family are exciting like that.

[black_son_of_gray]

I'll admit to having been a bit unfairly suspicious of William Harris at the start given the replication crisis & his findings seeming to make much stronger claims than others', but after digging deeper it looks like most of the literature agrees with the direction he's moving.

I totally understand the suspicion.

With respect to replication problems, here's the thing: a lot of seemingly well-designed studies simply aren't.

As an example, I'll frame this in terms of "risk of sudden cardiac death", so that you can look at the figure 5 I've referenced in my previous post to "play along". Let's say that you want to test whether omega 3's play an important role in sudden cardiac death. What a typical study might do is gather a bunch of participants, sort them into two statistically matched groups (e.g. age, sex, etc.), give half a placebo, give the other half a dosage of omega 3's, then tally up the sudden cardiac deaths in both groups after some amount of treatment time has passed and finally compare them to see if the omega 3 group had a different amount of deaths. Bonus points if the treatment groups were randomly assigned, if the control group was a true placebo (vs. some other treatment), and if both the treatment groups and the experimenters were 'blinded', meaning that when pills were dished out, they didn't know who was getting what, eliminating a kind of bias.

The above description is considered the gold standard. The above description, however, isn't a very good test of the question. Why?

Well, it's worth mentioning that the above protocol is actually good for drugs. Why? Because drugs are compounds that aren't naturally in the body to begin with, and so when you give a person a new trial drug, you know that the body doesn't have any to start with. The placebo is a true null situation, and the treatment dosage is all that the participant will see.

Omega 3's are already in the body. Different people have different amounts. Different people eat different amounts. So all treatment and placebo groups are already potentially very heterogeneous for the very thing you are trying to manipulate*. The problem with many omega 3 studies is that 1) they don't assess what level people have before they get the treatment/placebo, and 2) they don't assess what level people have after treatment/placebo. All they know is that they gave one group pills with X mg of omega 3's. So if both the treatment and placebo groups were already quite high in Omega 3 levels, would you expect to see an effect (even if omega's were actually effective)? If the treatment dosage that you gave out wasn't enough to move the needle on a person's Omega 3 levels (but you didn't measure what those levels were to confirm), would you expect to see an effect? Perhaps you can see how that might make it difficult to know whether omega 3's are effective or not? Perhaps you can see how a meta-analysis of these kinds of studies would also conclude ... "meh"?

The reason why I pointed people to William Harris's work is because he understood that problem, and went looking for a way to reliably measure what a person's omega 3 status actually was. The method he came up with, which is to look at the percentage of fatty acids that are in red blood cell membranes that are omega 3s, he calls the Omega 3 Index. It's a somewhat niche kind of lab test to do, so he has also set up a company that will do it for a fee for retail consumers if you want to know your own. (It is a little unfortunate that he's kind of the only big player in town with respect to the Omega 3 Index, and consumer testing, but that does appear to be changing as the idea gets more traction).

The Omega 3 Index (or something like it) is extremely important to answering questions like "do omega 3's play a role in ...?" because it allows you to 1) actually compare people with higher or lower omega 3 status (vs merely what pill they took), and 2) actually know if a supplementation regimen raised the omega 3 status of the participant.

If we look at the figures in the paper I cited in my previous post, Figure 1 shows the dose-response relationship between how much omega 3's a person takes and how much it changes their Omega 3 Index. Figure 4 shows, effectively, the range of Omega 3 Index values that humans generally occupy (~3-10%). Figure 5 shows how incredibly predictive the Omega 3 Index is with respect to sudden cardiac death. It is simply much better than the other factors that people hyperventilate about in their blood panels (LDL, HDL, triglycerides, etc.). If your Omega 3 Index is in the top quartile, your risk of sudden cardiac death is maybe 10% of those whose Omega 3 Index is in the bottom quartile. This is a relatively older paper, and omega 3's are studied in relation to a huge number of conditions, rather that just sudden cardiac death, so I would recommend also looking at the newer research out there - just keep in mind what the researchers are measuring. Are they measuring the actual variable they care about, or just some non-validated proxy like dosage of a pill?

*A lot of research that studies nutrition supplements (e.g. vit D, magnesium, etc.) have this problem.

[jacob]

Fig 5 seems to be paywalled for non-academics. Alternate reference, please?

[Anesau]

I didn't expect to come back into this conversation, but I'm obviously hooked (no pun intended?), so I did some further digging. I'm still not fully understanding how the literature supports the benefits of omega 3s to the extent that William Harris is claiming. I realize medical statistics is not my specialty, so I'm just going to post my thoughts out into the void and hope it nerd-snipes someone into correcting me (@black_son_of_gray, your thoughts would be very appreciated if you were willing to chime in again!).

--

As a starting question to your previous post: I can see how the Omega 3 Index would be useful to picking apart the mechanism of action. However, I still don't understand why I (as a layperson) would ever prefer it over just seeing the effects of people taking omega 3s in a randomized, double-blind, placebo-controlled study.

Let's say I see great evidence that omega 3s raise the Omega 3 Index. Then I see great evidence that the Omega 3 Index predicts outcomes I care about (ex. all-cause mortality, cardiovascular events, etc.). However, all of the studies that look at whether omega 3s predict those outcomes turns out to be ~null. Why would I privilege the longer chain of studies (going through a new index) over the direct effects of giving people omega 3s?

Studies going through the index (does supplementation raise the index? then, separately, does the index predict events?) seem way more susceptible to statistical / data manipulation than just looking at whether supplements get a particular outcome. Plus while it lets you take apart some factors (compariong omega 3 status instead of supplement effects), you get others (what if a person's baseline omega 3 status is all that matters, and people that take supplements increase their omega 3 index but don't have any better outcomes (because they still have the negative effects of whatever caused their low baseline)?).

As someone looking at whether to take supplements, the primary concern seems to just be effects of supplements on outcomes, not the intermediate steps.

I understand people have different starting & ending levels over the course of supplementation. However, that's the point of randomized experimental/control groups & a large study size; a sufficiently powered study should have a low enough chance of being due to random variations in group members (hence <0.05 p-value, for <5% chance of the outcome being due to random chance, correct?)

--

In trying to figure out the state of the evidence, I came across a meta-review of the effects of omega 3s (EPA & DHA) supplementation co-authored by William Harris. In it, we're walked through the list of intervention trials since the 80s, including the methodology and outcomes. Below I'll give a quick summary of every included trial as I understand it, then my overall thoughts. Again, I'm aware I could go wrong at any step of the chain here, so anyone stepping in to correct me would be appreciated.

--

The trials:

1980s

DART Trial - just being advised to consume >=2 servings of fatty fish weekly caused a 29% dip in all-cause mortality; subgroup given fish oil capsules got same result

1990s

GISSI-Prevention - one fish oil capsule per day = 41% reduction of risk in all-cause mortality at 3 months

2000s

DART-2 - same as DART in advising people to eat oily fish, but this increased the risk of cardiac death compared to nonspecific advice; turns out the increased deaths came from the subgroup given fish oil (study dismissed for methodological issues)

JELIS - first (& only?) study to be primarily female; risk of sudden cardiac death ~same between control and treatment (study dismissed for lack of power to detect significance)

GISSI-HF - "first large, randomized, double-blind, placebo-controlled trial" and found 9% reduction in all-cause mortality, 8% reduction in cardiovascular hospitalizations

2010s

Alpha Omega - "double-blind, placebo-controlled, randomized, secondary prevention study" with daily EPA & DHA consumption for 40 months; no reduction in cardiovascular events, all secondary outcomes null (dismissed for reasons I don't fully understand?)

OMEGA - randomized, double blind, compared daily omega 3 capsules to olive oil controls; no difference in cardiac death or all-cause mortality (dismissed for low power, because the study authors over easimated the effects of omega 3s & had less total health events than expected)

SU.FOL.OM3 - "double-blind, randomized, placebo-controlled, 2 × 2 factorial trial" checking for positive health effects of B vitamins and omega 3s. Null results for omega 3s. (dismissed for being underpowered due to lower than expected event rate, and omega 3 supplementation being too long-term(?))

ORIGIN - looked for effects of omega 3s in those with both cardiovascular disease & diseases causing irregular blood sugar; double-blind, randomized, compared omega 3s against olive oil controls; no statistically significant decrease in death/disease (dose may be too low?)

Risk and Prevention - compared omega 3s vs olive oil control for those at high cardivascular risk who've not yet had a heart attack; randomized, double-blind; no significant decrease in measureable endpoints for the main groups; a couple of subgroup analyses showed mild beneficial effects, including specifically for women (lower than expected rates of hard cardiovascular endpoints; olive oil could an inappropriate control)

--

Some thoughts:

• There is one study showing a negative effect of omega 3s, six ~null effects, and three positive effects.
• Of the null/negative effects, the authors had reasons to dismiss almost all of them on grounds of methodological issues. But somehow there was not much discussion of methodological issues in the positive effect studies, despite them being some of the oldest? It's certainly possible that only the null studies had methodological issues, but I'd need to look into each study deeper; for now I'm just vaguely suspicious
• Some of the arguments given for why null studies were dismissed seems at odds with Harris's ongoing hypothesis
  • The OMEGA study has a methodological issue of the authors overestimating the effect of omega 3s, leading to underpowered research... but that seems very odd, given the large effect Harris seems to be claiming that Omega 3s have?
  • The Alpha-Omega placebo might be invalid because the placebo contains ALA (vs the EPA/DHA omega 3s that were the treatment being tested). This seems like a valid concern scientifically, but as a random layperson, if ALA is doing just as well that seems like a totally valid strike against strong claims in favor of supplementing EPA/DHA?
  • Olive oil as a placebo was also put under suspicion because it could have its own heart benefits. But again, as a layperson, if olive oil helps the heart as much as EPA/DHA, I feel like the correct update is toward "EPA/DHA only helps insofar as all unsaturated fats help", which again points away from the usefulness of supplementation
• The only study called out specifically for enrolling at least 50% women found null effect in the Omega 3s group. This is could easily be a statistical aberration, but given that women process & convert Omega 3s differently (and have different cardiovascular disease symptoms), the studies seem mostly limited to the effects of omega 3s on men
• One of the biggest positive effects for omega 3s was in the DART trial where they got a 29% dip in all-cause mortality just by suggesting people eat fatty fish(!!!). That seems like a huge effect for basically zero treatment. Again, it's possible, it just seems unlikely. If the effects are this large from small interventions, I'm surprised the majority of trials ended up seeing ~no effect?
• Without knowing the number of subgroups they looked at in Risk and Prevention it could just be the jellybean effect, but it's on me to dig deeper into the original study before dismissing the outcomes.

Given the reproducibility issues in medicine - here's a study talking about how 51-89% of medical studies fail to replicate... (the irony) - I'm just baffled by how massive of an effect Harris seems to claim for EPA/DHA. My best guess is that it has a real positive effect, but probably something more like a <= 5% reduction in all-cause mortality.

Again, apologies if this seems too one-sided in my suspicion. I genuinely don't understand the large effects claimed for EPA/DHA and am hoping to learn a lot from this thread!

[Anesau]

Fig 5 seems to be paywalled for non-academics. Alternate reference, please?

Here's an open access link. There are multiple copies of the study floating around, only some of which are paywalled.

ETA: apologies, I think bsog actually meant this study. The names are very similar, but the figure in this one makes more sense given the context

[Ego]

Is it this one?



I got to the full study by clicking the PDF option next to the first study listed in the google scholar results. Full PDF here

[black_son_of_gray]

The figure 5 I was referring to was, as Anesau has linked, from this paper (PDF link)

I'm still not fully understanding how the literature supports the benefits of omega 3s to the extent that William Harris is claiming. I realize medical statistics is not my specialty

S'all good. I'm certainly not an expert in medical statistics either (or fatty acid biology/chemistry), nor do I want to put Harris up on some pedestal. He's clearly an expert, though, and his Omega 3 Index is a useful framework for studying the topic.

I won't be able to address everything you've brought up, but I'll try to add helpful comments where I think I can.

As a starting question to your previous post: I can see how the Omega 3 Index would be useful to picking apart the mechanism of action. However, I still don't understand why I (as a layperson) would ever prefer it over just seeing the effects of people taking omega 3s in a randomized, double-blind, placebo-controlled study.

Hmm. Let me reframe this a little. Consider the following two questions:
1. "Are higher EPA/DHA levels in the body protective against X disease/condition?"
2. "Does taking a pill containing EPA/DHA protect against X disease/condition?"

They sound similar, but #1 is the better first question to ask, because #2 has many more questions that could influence the answer, like "If I give everyone the same dosage of pill, does everyone absorb the same amount?" (As I understand it, the answer is "no") "Are some formulations of EPA/DHA more readily absorbed?" (the answer is "yes", and not all studies use the same formulation) "Are some dosage amounts simply too small to meaningfully move the needle for many study participants?" (the answer is "yes") And there are many other questions. Why does this matter for the random person on the street? Because a "no" answer to question #2 does not necessarily mean that the answer to #1 is also "no". If you get a "yes" answer to #1, and a "no" answer to #2, then that means you're probably dealing with one of those complicating issues (which may be a simple fix, but you need to know where to narrow things down). If you get a definitive answer of "no" to #1, then you can stop wasting money on answering #2 because it is "no" by default. (Yeah, I'm overgeneralizing here, but bear with me.)

The ideal scientific manipulation to answer #1 is a study that goes something like this: We hypothesize that higher EPA/DHA levels are protective against X disease/condition. To test this, we will take a population of people we know are low in EPA/DHA (and therefore presumably at greater risk), we will treat half of them with whatever dosage of EPA/DHA is required to boost them to a higher status, then we will see over time if this boosted group fairs better. See also: prospective study. In this case, it is clear what the independent variable is (EPA/DHA status) and what the dependent variable is (disease). If you just did that by giving all participants the same dosage of pill, you will end up with very poor control of the independent variable (or even unknown control of it, if you don't test before and after levels).

Granted, I see your point of just wanting to know whether the pills will work or not, but there are more fundamental questions that have to be hammered down (i.e. #1) before (#2) becomes the useful focus. (For what it's worth, I think the state of the science is getting towards #2 decently quickly)

I understand people have different starting & ending levels over the course of supplementation. However, that's the point of randomized experimental/control groups & a large study size; a sufficiently powered study should have a low enough chance of being due to random variations in group members (hence <0.05 p-value, for <5% chance of the outcome being due to random chance, correct?)

If you lump people with high and low status together in your treatment groups, you're artificially minimizing the effect size of treatment by dilution. Effect size, power, and sample size would all be negatively effected by lumping, no? Why is that better?

[Regarding the studies listed, I apologize, I simply don't have time for that kind of deep dive right now.]

Without knowing the number of subgroups they looked at in Risk and Prevention it could just be the jellybean effect, but it's on me to dig deeper into the original study before dismissing the outcomes.

I can't speak for every study, particularly older ones, but my professional understanding is that no reputable journal these days is going to let you publish a study without some form of correction for multiple comparisons. It's absolutely standard (at least, in the field I was in).

I'm just baffled by how massive of an effect Harris seems to claim for EPA/DHA. My best guess is that it has a real positive effect, but probably something more like a <= 5% reduction in all-cause mortality.

Again, apologies if this seems too one-sided in my suspicion. I genuinely don't understand the large effects claimed for EPA/DHA and am hoping to learn a lot from this thread!

Maybe some of your confusion of effect sizes has to do with the nuances of study "end points"? For example, the paper I referred to showed a massive effect for "sudden cardiac death", which I understand to mean "your heart stops beating suddenly and you die". Note that clinical literature has an insane array of similarly worded but not synonymous conditions. Sudden cardiac death doesn't necessarily equate to "heart attack" or "myocardial infarction" or "coronary artery disease" or "cardiovascular disease" or "arteriosclerosis" or "cerebrovascular disease". Some of these can be defined very specifically and some can be defined broadly (i.e. "all-cause mortality). So depending on how a study defines its endpoints, you may end up with dramatic differences in apparent effect size. From what I gather, there is something about EPA/DHA that is particularly effective in preventing "sudden cardiac death", even if they are less effective at say, "preventing plaque deposits from building up in the arteries".

That being said, for super broad clinical endpoints like "all-cause mortality", seemingly small effect sizes, like say 5-10%, are actually kind of good, right? Are there other treatments/supplements/drugs (besides exercise, which seemingly blows everything away) that do considerably better? I'm genuinely curious.

I know we have some medical doctors floating around this forum somewhere - I'd be interested in them pointing out my ignorance on these matters as well.

[IlliniDave]

Not an MD by any stretch. I look at supplementation as an art more than a science. With different starting points, different epigenomic states, and different genetics, one-size-fits-all answers just aren't available. I tend to look at the evidence filtered through expert interpretation, and if it seems like it might promote better health without much in the way of potential adverse effects, I'm inclined to err on the side of trying it. I've heard it takes something on the order of 30 years for new understandings to make it into everyday mainstream medical practice and I'm too old to wait around for medical field-wide pronouncements of efficacy. Fish oil is one of those.

[lillo9546]

Is it this one?



I got to the full study by clicking the PDF option next to the first study listed in the google scholar results. Full PDF here

I've tried my best to understand the image, without positive results. I am italian and I do speak english, but sometimes It's hard to decode information.
Could anyone explain this figure 5 meaning? Thanks!!

[mathiverse]

On the graph, lower on the vertical axis is better. Going from the left to the right, you have increasing weekly seafood consumption per set of mothers during pregnancy (see the amounts on the lower axis). You can see that increased seafood consumption correlates with (?) a lower (therefore better) outcome when subsequently born children have their verbal IQ and fine motor skills tested.

[lillo9546]

On the graph, lower on the vertical axis is better. Going from the left to the right, you have increasing weekly seafood consumption per set of mothers during pregnancy (see the amounts on the lower axis). You can see that increased seafood consumption correlates with (?) a lower (therefore better) outcome when subsequently born children have their verbal IQ and fine motor skills tested.

Thanks! So it's omega 3 positive

[mathiverse]

Thanks! So it's omega 3 positive

No problem. And, yes.