#619: How Much Fish Should a Healthy Diet Contain?

(upbeat music) - Hello and welcome to another episode of Sigma Nutrition Radio. My name is Danny Lennon and of course with me is Dr. Alan Flanagan. Alan, how are you doing today? - I'm very well-tanks. I'm looking forward to the end of summer, but it's been a great summer. - Excellent, I'm very glad to hear it. And today we have quite the topic for people that I think is going to be, one of the ones that for me is quite interesting because of some of the unknowns and maybe the lesser debated areas around this question that we'll hopefully get into. And it kind of sits at that intersection of a number of interesting separate questions. So as we will get to, there are a number of subquestions that start to come from the original question we'll lead off with. And whilst we are going to be talking about the health impacts of fish consumption broadly as our topic for today, this opens the door to a number of really interesting questions that relate to study design, methodology, within studies, how we evaluate evidence, and then how we go from there and interpret that appropriately to give recommendations at both the public health level as well as the individual level. So in particular, what we want to look at is the common recommendations that are given across a variety of different organizations and official guidelines from a variety of different countries that relate to fish consumption. And so as we will mention a moment, they are worded slightly differently. Some of the recommendations may be slightly different, but by and large, there is quite a lot of coherence amongst where those recommended guidelines end up coming out for fish consumption, for general health purposes at the population level. But what we want to look at is, where do those actual recommendations come from? How strong is that evidence behind them relative to what we might presume? How that relates to the specific quantity that's discussed, and then of the evidence itself, what are some of the nuances within it? Now within that, as we will get to, there's a couple of really interesting things that relate to the background diet or the baseline diet that someone has. And so if we're going to add or remove fish from the diet, what implication that has, but also related to that as ever is the comparator food, or what are we comparing that with? So if we are increasing or decreasing fish consumption within a dietary pattern, what would that be replacing? And what is the effect of it replacing different foods? And how should we think about that when it comes to our interpretation? And so just to put in a pin of a couple of ideas that we maybe will get into, one is around the differences we see when we are thinking about what is fish actually replacing in the case where we might be adding it to the diet in replace of something else. And then also a converse question, maybe that is worth considering, is that if someone doesn't currently consume fish, depending on what their diet actually looks like, there may be differences in what the resulting impact would be from adding that fish in. And so is it implicated in all those situations? And I think there's some really, really interesting nuance to get to. So before we get to any of those more deeper questions, probably our best point to start from here, Alan, is to take a look at what those recommendations actually are, get some clarity on that, where they tend to converge, and therefore off the back of that, we can start getting into maybe where that comes from. - Yeah, so across most Western countries, the recommendations are largely similar. So for example, in the UK, the scientific advisory committee on nutrition or second recommends eating at least two portions of fish per week, and a portion serving size is generally something between 120 to 140 grams. And then there's also a specific recommendation that one of those portions or servings per week be an oily fish, so are like salmon, sardines, mackerel, et cetera. European food standard agency recommendations are generally similar to that. In America, we also have a broadly similar recommendation, the Academy of Nutrition and Dietetics recommends about two 120 gram servings of seafood per week. Again, it also recommends focusing on choices that are rich in omega-3 fatty acids, EPA and DHA. It also then makes specific recommendations for keeping mercury levels low or eating fish that are not high in mercury. So those would be fish, for example, like bluefish tuna, bluefin tuna, sorry, or swordfish. Australia, similarly, two to three servings of fish and seafood per week. In Australia, the recommended serving size is slightly higher. It's 150 grams per serving, but again, it's pretty consistent across the board. Japan, which is one of the countries in the world with the highest fish consumption, has some interesting recommendations. It doesn't necessarily make specific portions serving size recommendations per week for the general population, likely because there's just high seafood generally. It does make more nuanced recommendations, which is specific to historical public health issues in Japan, which is specifically high sodium intake. So there's a sodium cap and official Japanese public health guidelines warn to limit very, very heavy salted or salting a fish or pickling a fish. And then they also, because of the prevalence of high mercury fish and the prevalence of fish consumption generally, for pregnancy, they make very specific recommendations for pregnant women to have different stratifications. So they make a category of fish where there's no limit required. And that's salmon, sardines, horse mackerel. Then there's fish with moderate restriction recommended, which is less than 160 grams a week. That would be albacore tumour and bream, for example. And then there's a strict restriction of 80 grams or less, which is all of which is due to mercury levels, which would be, again, like I said, bluefin tuna or sord fish. So broadly speaking, those guidelines are very reflective of specific population-based requirements in Japan for the wider kind of Western countries that we have mentioned in terms of the UK, Europe, North America or Australia. The recommendations broadly fall at this kind of dose recommendation of about two servings per week and generally recommend that at least one of those servings be oily omega-3 fatty acid rich fish. As you mentioned, we have this kind of general convergence around these guidelines with this one to two servings per week, typically with at least one of those being from oily fish. Those serving sizes in that range of 120 to 150 grams per serving, more or less. And so given that level of convergence, I think this is one of the reasons why people would tend to take, well, this is a very clear, quote, evidence-based recommendation that is very solid. And so when maybe now we can think about where this comes from, because there's maybe two ways to think about general recommendations, is this a, on one hand, strictly evidence-based recommendation based on data around those actual doses, or is it more of a sensible, plausible, heuristic for overall dietary patterns based on some of the epidemiology we may have, or maybe it's somewhere in between. So to start working through that, there's maybe two layers of where these come from. One, we can maybe talk about some of the broader epidemiology in a moment, but first there's some important historical context here that I know you're very familiar with. Can you maybe walk us through some of that history of where we got to the point of these types of recommendations? - Yeah, so back in the 1970s, there was some interesting work on my two Danish researchers, Bang and Dyersburg, or Dyerburg, in Greenland. And it was specifically focused on the in-ewish populations in Greenland. And it's a kind of tail where it slipped into kind of nutrition, lore, diet, lore, generally. And essentially their observation was that this was a population consuming very high fat diets, largely marine, fat-based, and seemingly absent evidence of cardiovascular disease in the population that they surveyed. And this is really the genesis of the subsequent decades long research interest into omega-3 fatty acids and marine source to omega-3 fatty acids and cardiovascular health. And the problem was it was just based basically on random kind of ecological survey, and also based on just an examination that wasn't particularly rigorous of like records. So it wasn't really direct investigative work, it wasn't anything that we would recognize in the context of more stringent epidemiological work even. And it actually turned out not to really be the finding that was necessarily that robust and subsequent research again by a Danish group. And then by a group in Canada found relatively similar coronary heart disease rates and prevalence of atherosclerosis in some of these populations compared to, for example, age match Danish controls or Canadian controls. So it was one of these kind of nutrition myths, not dissimilar to some of the kind of western A-price stuff that you hear about that tend to do, find its way into some sort of romantic notion. of what a healthy diet looks like, but did generate substantial interest in two omega-3 fatty acids and cardiovascular health in particular, and then as nutrition formalized into a kind of more rigorous science and subsequent decades in both epidemiological methods advancing, and indeed, in randomized trials, one of which will discuss specifically the dark trial. We did start to actually get more of a proper evidential picture on the relationship between fish consumption specifically. There are omega-3 contents and cardiovascular and other metabolic and disease-related outcomes. So, this may be an important point for us to note that we might circle back to later, but within this, one of the difficulties of going through all the evidence is, again, that exposure that we're actually looking at. So, being clear of, at any one point, are we looking at seafood broadly? Are we looking at oily fish? Are we talking about omega-3 fatty acids? Are we looking at specifically EPA-DHA? Because oftentimes, people can pull in evidence from all these different strands, but normally different exposures. And so, when we're thinking about some of these recommendations, obviously, they're pieced together from combinations of those things. But what is it plausibly about fish consumption in general that we might suspect might be considered within a place within diet patterns that are health-promoting, let's say? Yeah, I think it is generally the omega-3 contents that provides the kind of most plausible explanation for fish. In addition to other, obviously, food, matrix factors that are always difficult to tease apart. And in the research, this is really important because there's always generally been a conflation between fish oil as a supplement and the marina omega-3 fatty acids. And obviously, for today's purposes, we really want to isolate fish specifically as an exposure. And, you know, this is particularly important again because with the context of randomized trials, we obviously have more epidemiology looking at fish specifically, but we really only, in terms of outcomes, only have one trial, which was the original Dart trial published in 1989, that was a specific food-based intervention. And most subsequent trials, particularly after the Dart 2 trial, which was in 2003, and that mixed fish oil supplements with oily fish, didn't replicate the original findings possibly because it diluted the nature of the exposure of interest. And as a result, from then on, the field really moved more to looking at supplement-based fish oil rather than fish as an exposure or intervention of interest. So we have the two Dart trials that you mentioned, these diet and in China randomized trial, where we have these, the surface conflicting findings, or at least a lack of replication, that kind of second trial, for those reasons you mentioned, which kind of speaks to this exposure issue. What are some of the, that in conjunction with some of the other evidence that might be the main trials that might inform us around how we get to these dose recommendations? Or why specifically do we end up landing on these one to two servings of fish per week as opposed to anything else? Yeah, digging back in time, I think probably one of the first studies that would land on one's desk chronologically would be a 1985 paper in the New England Journal of Medicine from Dan Kromhev's and colleagues. And this was a prospective cohort study from the Zutphen study, which was a population-based cohort, which took a random sample of just over a thousand Dutch men in the Netherlands. Of those 872 participated with complete follow-up across, which is one of the strengths of this study, across a 20-year period. So recruited and started the cohort in 1960, and then mortality was ascertained over this 20-year follow-up period, ending in 1980, and the paper was published in '85. And what they were looking specifically at was the incidence of coronary heart disease, and in this particular paper looking at fish consumption. The mean fishing take was about 20 grams per day, two thirds of which in terms of overall fish intake was lean fish, like cod or place, and then one third was fatty fish, herring and mackerel in particular. And with this study, the outcome looked at five different kind of categories of fish intake with the highest being over 45 grams a day, and coronary heart disease mortality specifically as the endpoint. And the study overall found that the highest fish intake was associated with an over 50% reduction in coronary heart disease mortality risk, and that was evident at both the fourth highest quintile and above. The over 45 grams a day wasn't statistically significant, but likely due to a much smaller number within that. So this was one of the first studies to suggest this relationship, because with obviously you look at the kind of daily exposure average intake, overall what it corresponded to into weekly doses was around this one to two serving range. And that was at noted at the time to be kind of lower than other populations, like particularly in Japan, where the average was 100 grams a day. So you were looking on average at about 120 grams a day, minimum weekly serving average in this particular cohort. In the study, the authors actually conclude that it was justifiable to make this recommendation for one or two fish servings per week as part of dietary guidelines for coronary heart disease prevention. And then in 89, the diet and re-infarction trial or dark trial was published, burned colleagues. This was a randomized trial, really, really nicely designed trial, which was a factorial design, which was looking at three factors. One was total dietary fat intake and fat composition. The second was fatty fish intake specifically. And the third was fiber intake specifically from cereals. And so you had these three groups, participants randomly allocated to receive or not advice on each of those three. So they could have different potential combinations of which of these advice they were implementing. The fatty fish was at least two weekly portions, so aiming for 200 to 400 weekly grams of fatty fish. And ultimately then in terms of the outcomes, which was looking again at total mortality and then combined ischemic heart disease and non-fatal myocardial infarction at two years' follow-up. The reduction in mortality risk in the fish group, fatty fish group was 29 percent over the two years. And that was entirely explained by lower ischemic heart disease deaths. There was no significant reduction in either the fiber serial group or the lower low total fat group. And that effect of fish was evident across the different fat and fiber advice subgroups without any significant interaction of those two specific pieces of dietary advice. So the dark trial was one of the very few food-based interventions at the time specifically looking at the incorporation of this dose threshold of at a minimum to oily fish servings per week and seeing this significant reduction. And that was not interacted by either the serial fiber or the total fat intake. And then unfortunately, as I kind of just mentioned earlier, dark two then sought to replicate that, but by also randomizing people to receive fish oil capsules. And this was actually one of the first trials to suggest a potential as we've come to see some concerns over fish oil supplements raised over things like a fib. And there was a noted increase in that in the fish oil arm. And again, subsequent to that, we kind of moved away as a field away from some of these more food-based interventions to very much focusing on omega-3 fatty acid supplementation. But between dark one in '89 and some of the other like the crumb hat and colleagues paper and the New England Journal of Medicine in '85 and several other kind of cohort studies used the crumb hat paper because it's broadly representative of some of those findings at the time. This minimum effect of dose, so to speak, of around one serving a week of 120 grams and one to two servings per week as a kind of range did have a degree of evidential backing both from some of that epidemiology and the dark trial as a food-based intervention trial. And so it does have a genesis in the evidence and it's not necessarily just plucked as a kind of heuristic. Super interesting, and I think it raises a number of things that we wanted to get into. One paper that we also have kind of highlighted that relates to, no, this dose, but a few interesting elements was the Moan and colleagues paper that was published in JAMA Internal Medicine. Most notably because, as we'll maybe discuss, the findings kind of center in around something that could relate to this dose of around two servings a week. But it also has a couple of interesting elements of one. Maybe we might see differences in the population, whether this is in primary or secondary prevention, as was looked at here. And then also this, some Sometimes a result that pops up in the epidemiology around a potential, almost a U-shaped curve for the risk when we look at some of these doses. But again, there's some nuances in this paper that we need to be careful with. So with that particular paper, the Mohanon colleagues paper, this is a big pooled analysis that we have a lot of participant data, for you, maybe you can walk us through some of the things that jump out, but also how it relates to these different concepts we've mentioned so far around the dose, what we end up landing on here, and then some of those implications for, as they looked at in this trial, differences between individuals with or without a faster disease. This is a really interesting study to look at. It was a pooled analysis of individual participant data from four prospective studies. So you had the pure study, which is a multi-country, across 18-country, prospective cohorts. You had the on-target study, the transcend study, and the origin study, and the latter of which were all actually intervention trials with long-term follow-up. So it took the individual data from all of these, which in total was around 191.5,000 participants from 58 countries across six continents, and seeking deliberately to examine the associations between fish consumption and cardiovascular risk and mortality, and they had a distinction, as you've just noted, between generally healthy individuals without cardiovascular disease, which was basically around 140,000, and just under 52,000 had existing vascular disease, and they categorized the participants into four groups under 50 grams per month of fish, 50 grams per month to under 175 grams per week. And that was what they labeled an intake threshold, roughly corresponding to two servings per week. Then you had this 175 gram to just under 350 gram or over 350 gram a week, the latter of which would be around a four serving per week threshold. Unfortunately, type of fish, for example, oily fish, rich in omega-3 or other fish, was only actually assessed in the origin cohort, which is another kind of caveat to the overall study. But in populations without vascular disease, which was the pure study, there was no significant association between the high compared to the lowest category of fish intake and major cardiovascular risk, or total mortality, or indeed cardiovascular mortality. Now, there's all sorts of issues with pure that we've discussed before in terms of like different background diets, very different kind of populations, a preponderance of data from kind of low middle income countries, related factors that would potentially influence the kind of associations that you're observing, and not just in relation to fish, but in relation to multiple of the outcomes that come out of the pure study. It's a very interesting study, but there's kind of several of those limitations we always need to bear in mind. And then the three trials were patients with existing vascular disease on target transcended an origin, and in that, the approximate two servings per week threshold compared to under 50 grams a month of fish was associated with a 16% significant lower risk of major cardiovascular events, an 18% lower risk of total mortality, and lower risk of cardiovascular mortality as well, and in origin, which was the only study within these to actually distinguish between oily fish or not, it was high omega-3 oily fish, specifically that was more strongly associated with lower major cardiovascular disease risk, and per 5 gram increment increase, there was a 6% lower risk of major cardiovascular disease, while other types of fish that were not rich in omega-3 fatty acids showed neutral associations. So it was a pretty interesting study overall, very highly powered, of course, across of multiple continents and countries, and generally supporting that dose threshold that we've seen in previous studies, but with the specific additional caveat that the signal in the noise was much stronger for patients with vascular disease, not necessarily in general populations without cardiovascular disease, which was the pure study included in this, but overall it's still largely supported that kind of broad dietary guideline recommendations for at least two servings per week, particularly oily fish if we factor in the finding from origin, and particularly for secondary prevention of cardiovascular disease. So we do indeed have this evidence that would be supportive of these typical recommendations, as we've discussed a number of them so far, doesn't seem to be anything that is maybe particularly surprising to some people, although as we start to get into the nuanced there is still some debate within nutrition around how much of that depends on what fish may be replacing if someone does start to add in these one to two servings of fish per week, and obviously that depends on someone's baseline diet, what foods they might not be consuming without that, and we could think of that in the example we gave in previous episodes of if we're thinking of that in a trial, it's nice and easy. We can have a comparator food there, but of course most of our data here as we're discussing can be built around some of these big cohorts that give us really useful information, but maybe that comparison is something that is more implicit that we don't exactly know what that comparison actually is. And so the conversation and the bait ends up becoming, well, if we are adding in fish to someone's diet and now they're consuming that in place of processed meat versus if someone was consuming fish instead of legumes for example, would we likely see differences and how much can we actually detect that in some of this work? So when it comes to that question of what it's being compared to and how that might change the interpretation of the impact in this case fish on a health outcome, how should we think about this in the context of the work we've discussed to this point and maybe some of the studies we're about to discuss in a moment? Yeah, I think there's probably two ways to think of it. Obviously there is the general approach in epidemiology now of conducting substitution analysis and that can be food-based substitution or as is often the case when the focus is on more some of these analyses have looked at that we'll discuss the source of protein for example. And they initially those analyses would say for example we'll swap 3% of animal protein with 3% of plant protein, intuitively that looks more like a macronutrient substitution model because it's modeling a percentage change of energy, but it is actually similar to food-based substitution models in that it's introducing a bunch of additional assumptions that are important to consider. I think there's two ways that we can think of this. One is considering some of these substitution analyses themselves and then the second is to then consider maybe some of us we will the vegetarian epidemiology which often includes pescatarian diets and kind of between the two you get a sense of where the position of fish for example as part of diets that may not include me or indeed much or at all any other sources of animals of animal-sourced foods. So I think as a point of departure though like the substitution analyses are very useful for this with all of the caveats of the distinction between food-based substitution models and macro and purely macronutrient based substitutions. There's one that you know we've discussed this is a knowledge funder and colleagues part of Lucas Schringeckel's group in Germany was published in 2003 and this was a food-based substitution model of animal-based foods and plant-based foods looking at cardiometabolic outcomes. It was a systematic review and meta-analysis so it included around 37 publications from 24 cohorts most of which were in the U.S. I think 12, 7 or 8 in Europe and the remainder in Asia and they defined specified portion sizes for meats or animal fish which includes sorry animal foods which included fish which was 50 grams per day and then for certain plant-based foods for example nuts it was a dose of 7 grams a day or legumes 12 and a half. So depending on what the plant's food was there was a range of the specified portion size of between 10 to 50 grams. Now hopefully participants I our listeners have already listened to the previous episode on food-based substitution models because at this point they might intuit that there may obviously be some issues with this kind of comparison even if it's adjusted there are additional energy considerations and there are also considerations of the wider diet pattern that is being introduced when these kinds of food-based substitutions are being modeled. Nevertheless in these participants on in the meta-analysis the replacement of 50 grams a day of fish and seafood without either 7 grams a day of nuts or 12 and a half grams a day of legumes was not associated with coronary heart disease incident there was just no signal in the noise there and that was only based on two cohorts that I'd specifically looked at that. It also wasn't associated with all cause mortality. So overall low very low certainty of evidence for substitutions of fish and seafoods with plant-based foods in relation to cardiovascular disease. of Ascular Disease, with no clear benefit, but no clear harm in these associations. And the evidence for the substitution of fish and seafood in this particular study was not necessarily particularly persuasive. It wasn't suggesting that there was any particular benefit, and it wasn't suggesting that there was any particular harm. So it would leave us with the conclusions in the context of these food swaps would probably be unnecessary relative to other changes or replacements or increases or decreases of different animal and plant foods, and probably more specifically animal foods that could be made in the context of increasing the overall healthfulness of a diet. There's a couple in this same vein that look at maybe some of these substitutions that maybe lead to some different conclusions that might be useful for us to outline to people. Maybe to try and reconcile why we might see that. One of the papers, and I see this often does the rounds here because it notably came from the Harvard group who have published quite a lot of the work based on the nurse's health study and health professionals follow up study data. This is one of the papers to come from that song and colleagues in 2016, again published in German Journal Medicine. And this is one of the ones that is typically shown for that proposal of, well, if we're looking to know the true health impact of a fish, we don't want to be looking at a comparison with things like red meat or processed meat where we might likely see a benefit. If we do the comparison here with some of the plant proteins, as was done in this study from the Harvard group, we might see something different. And indeed, within this study, we start to see when some of that fish protein is replaced with an equivalent amount of energy from plant protein. They report lower risk for mortality and cardiovascular mortality. Can you maybe walk through for you, what are some of the notable things from this study that are useful pieces of nuance to mention with this publication? And then from there, we can talk about one that contrasts with some of those findings. So the North American cohorts are fairly consistent in showing, and particularly the two in this was the health professionals follow up study and the nurse's health study. And these are long-standing cohorts with very well validated dietary assessment methods and a lot of strengths, where I think we always need to be careful with interpreting associations in some of the U.S. cohorts is their potential for replication in other populations. Because we are talking about population within these cohorts with higher energy from animal protein than you might see in other cohorts, lower energy from plant protein, overall worse health and dietary behaviors than you often see a baseline in other cohorts. And this is for epidemiology, which is population-based. So you pretty typically see in these cohorts that the replacement of any sort of reduction in animal-sourced foods in these cohorts is beneficial within this population. That is precisely what this study found. It used an energetic substitution model, which was modeling the replacement of 3% of energy with plant protein, sorry from animal protein with 3% of energy from plant protein. And indeed did find that the replacement of fish protein at 3% with plant protein was associated with a 6% lower total mortality risk, a 12% lower cardiovascular disease mortality risk, and this is obviously like a study that then gets cited to suggest that actually there is a net benefit, even if fish is considered associated with lower risk of these outcomes as an individual exposure in other cohorts. This is an example of where there's still a net benefit to the substitution of plant proteins. I'm not sure we think we can say that with much confidence for the reasons outlined in terms of the cohort-specific characteristics of this particular cohort. The fact that the study we just previously discussed was based on a far larger or much more highly powered analysis that incorporated cohorts both from Europe and Asia is something that's really important to consider and we'll move on to a European context next. But then there's also the nature of this kind of analysis itself, which is both an energetic isochaloric substitution model, but not also truly necessarily achieving that in the same way that it would be for micronutrients. In the first instance, both fish and plant protein intakes are very low in this particular cohort, and then while statistically swapping it looks fairly straightforward, you've also then got, it's essentially introducing what is a composite, so it's got multiple components to it, even though the target swap is fish protein to plant protein, you've got to account for the rest of the diet within that. And often these are achieved by additional adjustment, but there's still all these additional assumptions that are then introduced into a model like this, which are often hard to account for, and with the cohort-specific demographics that you tend to get in the NHS and HPFS, I typically, if these analyses are not replicated, always tend to take some of the findings as potentially population-specific rather than necessarily a reflection of the nature of the exposure and that particular swap in particular. And as you noted a moment ago with the importance of the actual cohort population, where we might see differences when we look at other geographic regions, and you noted that we have an example from that pan-European epic paper, the 2024 paper. Could you maybe outline some of the specifics of that and how we see something a bit different to what we've just discussed? Yes, so this 2024 paper from Zeng and colleagues was published in the American Journal of Clinical Nutrition, it was a case cohort design from the pan-European epic cohort. So this was looking at incident cardiovascular disease cases and followed up with a randomly selected subcohort as a comparative group, it had about 16,000 just over incident cardiovascular disease cases, and just over 15,000 subcohort comparators from seven European countries. And what's interesting about the distinction with this is that the animal-derived total protein intake was actually a little higher than in the health professionals follow-up study and nurses' health study combined in the Song and Colleagues study, but animal-derived protein was a little lower, it was 11.5%, and plant protein was higher at 5.4%, so about 1.4% higher. So the composition of the protein at baseline in this seven-country pan-European population is already at baseline different to that of these states. There were other aspects of health-promoting behaviors that do tend to cluster similarly between Western countries, people who do consume more plant protein are typically consuming more vegetables, more legumes, lower saturated fat intake, and other health promotion correlates of some of those characteristics of their diet. And this was looking at incident cardiovascular disease again. It models a similar substitution, model as the Song and Colleagues study, which was a 3% energy intake substitution, replacing animal protein from different sources, but from fish specifically, as we're interested in, with plant-derived protein, and there was basically no signal really for any of these outcomes. Total cardiovascular disease, confidence intervals spread across 1.0, whether that was fatal or non-fatal cardiovascular disease didn't matter, both spread across the null, total ischemic heart disease, fatal or non-fatal, all spread across the null, and the same first total stroke or fatal and non-fatal stroke. So this is a very good example of the contrast with population background factors matter in a cohort. The limitations as discussed with the methodological issues in the 3% modeling, as we mentioned in the Song and Colleagues study, do still apply here. So you're mixing these caloric variables, yes, you've got your 3% protein swap, but you've got mass-based considerations within that in terms of the different food sources, and you've got your introducing kind of assumptions in relation to the wider background dietary pattern. All of those critiques still stand. But it is, I think, an interesting contrast to the Song and Colleagues paper because it does indicate the importance of the population you are looking at, and no finding in any given cohort is, of course, independent from the population sample from which those findings are derived. So I think between the Zang and Colleagues paper and the previous Noinge Funder and colleagues met an analysis and systematic review, which is 2023, I'm not sure that we can say that the finding in the Song and Colleagues paper is necessarily representative, rather than potentially population-specific. And to make this clear for people, practically, because sometimes I think there's a bit of confusion around how to connect with this idea of why the baseline diet or background diet within these cohorts might matter as much, and why they actually create different findings. Would it be conceivable for a situation where in the populations like we saw in the Song and Colleagues, where you normally have more animal, dominant background diets, generally more unhealthy background diets? perhaps. And then, therefore, any increase in plant-based foods in those contexts perceivably is going to create this type of benefit relative to another population that is already starting on, let's say, or at least slightly above that level of healthful components of that diet. How should people conceptualize why the background diet impacts directly the potential impact? So, if you've got a population that is consuming overall very, very low levels of beneficial foods, and in this case, we'll just broadly say that, you know, plant foods, legumes, and certainly sources of plant proteins are foods that are beneficial to human health. And you're relying essentially for your model. You can't. The model is a mathematical abstraction. I think when it comes to substitution models, people really need to understand that you may not actually have people in a cohort that are consuming. No one's consuming 3% necessarily a fish protein. Like, some people likely are more pescatarians possibly. But in these cohorts, you often don't have people consuming the dose thresholds that are often modeled in a food-based substitution, or indeed, energetic substitution kind of model. So, it's a mathematical abstraction that is often then having to be forced to fit your data. It's then going to be largely weighted towards the people in your cohort that do actually consume more plants, more plant protein sources, etc. And although statistical adjustment can attempt to account for that, it's never going to abolish that. So, in a cohort with a poor background diet, you would kind of expect, especially in this cohort, yes, that the mathematical swap introduces the limitations of the cohort that you're deriving your analysis from and the other characteristics of that cohort. So, it is unsurprising in a cohort like the two used in the song and colleagues analysis that almost any reduction of animal protein or animal sources of foods and concomitant increase in plant sources would be reflected in beneficial reductions in risks of the outcomes that are also in terms of the magnitude of those risk estimates. Much more greater in magnitude than we would see in any of the other studies that we could look at, particularly ones from Europe, or certainly any from Japan. And this is something that I do want us to return to a bit later on because it's something we covered in our previous episode, the Rock Paper Salmon episode around some of these comparisons where we do have this maybe misinterpretation of what could be going on. And one of these parts that, as you've alluded to, is the jump from saying, okay, someone that normally has very little plant foods in their diet, if we could even have some clarity around that they receive a benefit from increasing, let's see legumes and nuts that they're consuming, and maybe that comes at the expense of some fish protein. A benefit in that context does not necessarily extend out to someone who consumes a plant predominant diet with, let's say, one to two servings of fish per day, would receive the same level of benefit from swapping that out and having zero servings of fish. And that's a claim that we can look at a bit later on. But ultimately, this comes to the crux of what I want to ask you, and this leads us into some of those cohorts that you mentioned, is when we're thinking about the impact of fish and potential health impact, one of those questions is does fish consumption add this benefit to, let's say, an already high quality, nutritionally adequate plant-based, planet-exclusive diet, and there's two sides of this coin, right? There's one of saying, would adding fish to that plant-exclusive diet lead to a health benefit, or the reverse, the other side of the coin, would removing fish from a diet that contains it and replacing with plant proteins lead to some benefit and how we work those out. And maybe these are very, very difficult things to directly measure, but it's where maybe some of those cohort data and looking at the pescatarian and vegetarian cohorts might lead us to some way of the answer. But there's maybe a lot of things I said there at the one time that maybe we can start working through, but just wanted to put a pin in those for people that they are some things that we should consider as we go through this. I think a good point to departure. There's probably two cohorts that are instructive to look at anytime the question of vegetarian dietary patterns or the role of fish, but no meat, for example, comes up epic oxfers and some of the kind of pan-European epic analyses. And of course, the Adventists tell study too in America. Epic oxfers, there was a paper in 2019 from Tommy Tong and colleagues, which was looking at in just under 50 or 49,000 participants who were classified into three groups. They had meat eaters, they had fish eaters who ate fish, but no meat, so pescatarians. And then they had vegetarians, which included a small group of vegans around 16, just over 16,000 of those. They were looking at a scheme of heart disease and stroke, and over 18 years of follow up for that outcome of a scheme of heart disease, the pescatarians had a significant 13% lower risk of IHD compared to the meat eaters. And the vegetarians had a 22% lower rate. The confidence intervals for the vegetarian diet were probably more convincing than the pescatarian. It was a more robustest effect estimate for the pescatarians. There was no significant difference in stroke for the fish eaters, the pescatarians. And so both of these particular kind of stratifications of vegetarian diets as between the pescatarians and the fish eaters had lower risk of a scheme of heart disease than meat eaters. And that was a stronger overall strength of association for the vegetarians than it was for the fish eaters. And this appeared to be generally explained through intermediate risk factors like lower prevalence of hypertension, LDL cholesterol, and better glycemic control in those vegetarian and pescatarians compared to the meat eaters. And then if we jump to the Adventist Health Study 2, there was a 2024 American Journal of Clinical Nutrition paper from this cohort. So this is a large mostly US base and some participants in Canada cohort of just over just short of 88,000 participants who are all within the 70 Adventist Church. It's an ethnically diverse cohort, which of course the health professionals follow up study and the nurses health study are not. So about 25% black African American participants and numbers of other ethnicities in this. And in this they categorized participants into five groups based on primarily frequency of animal consumption. So they had non-vegetarian, which is essentially the omnivorous kind of reference group. They had semi-vegetarian who were people who occasionally ate meat either monthly or weekly. They had the pescatarians again, just fish only with no meat. They had lacto-ovo vegetarian, which is eggs and dairy, but no meat. And then they had vegan no animal products at all. And looking specifically at all calls and calls specific mortality. And for the all calls mortality looking at vegetarians versus non-vegetarians. And for again, for the specific vegetarian patterns, the results are I think what listeners would likely expect. For the lacto-vegetarians, there was a 12% lower risk of all calls mortality compared to the non-vegetarians. For the pesco vegetarians, they had data two time points, age 65 and age 85. At age 65, pescatarian diet associated with a significant 15% lower risk compared to the non-vegetarian or omnivores. And at age 85 again, there was still significant, but a smaller kind of magnitude of effect of 10% lower risk. And there were various call specific outcomes that were broadly kind of similar overall between the vegetarian overall comparison between the non-vegetarians and then the specific vegetarian subtypes. So slightly more granularity in terms of the characteristics or the characterization of the different vegetarian dietary patterns. And this is something that I think when we contrast it with the Oxford epic cohort specifically is interesting because again, the Oxford epic cohort, although a vegetarian cohort that the kind of background health characteristics of the participants aren't as healthy overall as you would get in the seventh day Adventist cohort where you have multiple health promoting behaviors in terms of dietary intake, but also physical activity, smoking and alcohol, often abstinence. So you have this range of health promoting behaviors that are already present at baseline against which then you're kind of looking at the impact of diet. So it's actually giving you certainly against some of the more kind of omnivorous dietary passions within the AHS2 cohort, maybe more of a kind of direct test of the impact of some of these vegetarian dietary passions. But I think when you take the two together and some of the other papers published out of these cohorts, for me what I kind of end up seeing is the broad nature of vegetarian diets, certainly, is such that it clearly, to me, is not a dietary pattern that necessarily turns in its health benefits on the presence of any particular food per say. And we see that with often relatively similar strengths of risk estimates, et cetera. It is certainly overall relatively persuasive that as compared to omnivorous regular meat features, that various of these dietary patterns are probably more conducive to long-term health and certainly lower cardiovascular risk in particular. But it does appear that often, as between, for example, a lactovo or a pescatarian diet or a general vegetarian diet versus pescatarian, as with epigots for it, you're looking at many ways to skin a cat towards a healthy diet that don't necessarily scream, this food has to be in this dietary pattern. Yeah, and I think that's one of the crucial things of any time we're looking at some of these cohorts and we might try and look at pescatarian cohort versus vegetarian or ideally if you could find a plant-exclusive vegan cohorts with much more limited in number. But very rarely are we actually getting true direct comparisons of dietary patterns that are basically the same, just simply with the inclusion or exclusion of fish. There's never really that direct test and we don't have good evidence on that test alone. And as you noted, we're end up testing something that normally can give us information, but is not just that isolated test in and of itself. Yeah, and this is really crucial to consider certainly the substitution models as well. And the question of, is it worth increasing or specifically decreasing, in this case fish, or just indeed any food, because if you're looking at those comparisons, there's an assumption of a symmetrical change. Like if I eat less meat or if I eat less fish protein or if I just eat less fish, that I've precisely replaced that with plant protein, right? That's the assumption introduced in those analysis, but in reality, someone who decides to eat more fish might eat less beef, obviously, but they might not. They might eat less refined carbohydrate, someone who's eating less fish might have replaced that with chicken, or they might have replaced it with plants. So all of these factors are really important, and this speaks to the caveat that we were offering people in the previous episode on substitution models as to why we should approach their interpretation with a little more caution and a little more granularity. But as it relates to fish, I think that clearly it tends to exist both in more omnivorous dietary patterns and as specific components of vegetarian dietary patterns as well, and clearly where it is placed in the context of those overall dietary patterns does measure, and it's very difficult sometimes, I think, with foods that are overall associated with health. And this is the caveat, of course, we raised in that previous episode. I think we need to always be careful that where two particular foods or food groups are generally associated with the benefit to human health, we get into a weird place of trying to splice whether it would be better to, or a lot of it would be better to increase or decrease in relation to that food or in replacement of that food. So I think sometimes it can help us to step back from abstract mathematical modeling and put our actual real worlds, how do people ease hat on and how do we frame advice accordingly. And maybe for people who didn't catch that episode or maybe can't remember some of the specifics, it might be useful for us to just briefly recap that example we gave that over interpretation maybe of substitution analyses or where we can get our end conclusion to go a bit too far. So one thing is to note that in some of the studies that we've outlined, we could indeed see a case where this substitution of, let's say, the fish protein for plant-based proteins and some of these studies seems to be correlated with this lower risk. However, as the example we gave in that previous episode, there are people that will then go and claim that, well, look, if we can find any evidence that there is this benefit, let's say, then therefore the appropriate conclusion is that the amount of fish that should be recommended within any of these guidelines has to be zero. And the reason, therefore, is because if you replace it with some of these things like a plant protein, if there is evidence that there is a risk reduction, then that has to mean that there is a net harm from consuming fish and therefore their amount that should be recommended has to be zero. And if someone is consuming more than that, they're causing harm relative to not consuming it in place of these plant proteins. And so you get this type of argument, which again, you can see how someone may be connected to those dots, but as you outlined is maybe not what we should be taking from a mathematical modding of some of these substitution analyses, even without considering all of the variation in the types of conclusions from different types of studies that we've noted. Yeah, I think that's a really interesting mental exercise that would take someone to get to that point of saying that actually the guidelines are incorrect and the safe or best amount is zero. I think it's difficult to arrive at based on the evidence at all. I think the reality is there's very little persuasive volume of evidence. Bar one or two individual studies that would suggest that there is this net benefit. And most studies find that when you model the swapping of fish with plants, you get no evidence of benefit or harm, i.e. they appear to be in most studies that model a swap, whether it's isochaloric or food-based, to be relatively equivocal. And so ultimately then the question becomes to the point you raised earlier. Is it necessary? Is it necessary that someone who maybe doesn't consume fish starts consuming fish to increase the overall healthfulness of their diet? I think when we look at the vegetarian epidemiology, I don't think that's a particularly persuasive case either. I don't think we could necessarily say to someone you should do this to improve your health because I think it's clear that the multiplicity of combinations that can create a vegetarian diet are all generally associated with pretty beneficial health outcomes long term, especially for cardiovascular outcomes. And does it mean that necessarily someone consuming a plant-exclusive diet needs to abandon all of their principles to consume fish, again I don't think that's a case we can make. So I think the reality is in the conclusion that I ultimately arrive at is the current public health recommendations are generally solid and there is no real reason to see them changing, but that doesn't necessarily mean that there are prerequisites and must be followed in order to achieve that, right, we make multiple public health guidelines. And again, I always think in the context of our current populations, people would be better off focusing on things like sodium and fiber, for example, as a kind of low-hanging fruit to improve overall diet quality. So I think the guidelines can stay where they are, but they don't necessarily mean that anyone not consuming fish must or anyone currently consuming fish has to access in order to be healthier. And I think there is generally within the vegetarian epidemiology, so not vegan or plant-exclusive, clearly a signal for multiple different combinations of non-meat diets or meats that are at least very low in their frequency of meat consumption. I think that's a crucial point that clearly we cannot say that fish itself is biologically necessary or is some type of indispensable food. Quite clearly, we have dietary patterns that contain no fish that are extremely helpful for humans, and therefore the consumption is not that if you don't consume that, you must in order to be healthy. In the same vein that if someone does consume those types of recommended servings that are in those typical guidelines, it does not necessarily follow that there would be a health improvement from removing them. And so I think this kind of gets to those questions where two things can be held at the same time, which again, we don't want to open the whole conversation around EPA and DHIB, which we've covered in this episode, it might be one for another day, but in a similar way there of we could say one side that fish is not an indispensable food, doesn't necessarily mean that there is no potential biological effect of EPA and DHA. What is more of the question is to what extent for everyone is are those needed, and there's again, this still ongoing debate, but I think you framed it perfectly that the conclusion isn't that everyone must get to the two servings per week that are in those guidelines in order to be healthy. And of course, as ever, everything depends on the overall dietary pattern and would that likely be health promoting. So there's that balance required, I guess, on both sides that you've just outlined. Yeah, and I can probably think of examples within a population like a general Western diet, particularly if it's constrained by income in terms of like foods, budgets and stuff like that, where a deliberate move to consume a couple of servings of oily fish a week may actually be quite a beneficial move in that kind of context. I can conceive of that, for example, and it is good that you can get foods like you know, macro and sardines for that, you know, agi-pence in Tesco, and there's no issue with the fact that it's canned, great. That's a really simple way for people on, you know, constrained budgets to be able to get omega-3s into their diet. But for people that necessarily aren't and have the means to consume is broad, health-promoting diet, and can select any number of kind of, you know, basically have no constraints on food purchasing and nutrition kind of knowledge. And yeah, I just don't think it's a prerequisite. I don't think we have to do away with our recommendations. I think they're fine where they are, like I said, but I think that there are multiple ways by which certainly a more vegetarian or plant-forward dietary pattern can be very health-promoting over the long term. And that may or may not include fish. Perfect, and hopefully within that, I think we've covered what we had hoped to cover. I don't know if there's anything we're missing here, but we will leave it there in the description box where everyone is listening right now. I'll link to any of the studies that Alan and I have discussed today, as well as more information on the episode. Also, if you are a premium member, you will get your study notes to accompany this episode. They will be linked up there as well. Hopefully all of you listening have enjoyed this particular episode. And Alan and I will be back in another episode very shortly. So hopefully you return then. And thanks for listening. (upbeat music) Before you go, I just wanted to remind you about Sigma Nutrition Premium. So what's the idea of this subscription? Essentially, it was created with the goal of allowing you to more deeply understand the material you're hearing on the podcast episodes themselves. You'll be able to retain more of that after you finished listening or reading through the notes. And then be able to easily and efficiently revise so that in the future you can remember that information to be able to create your own content or ideas using things that you have learned. And how do we go about this? Well, there's a few different ways, but at the core of the subscription is our detailed study notes. Full of all useful descriptions, background context, diagrams, charts, et cetera, to allow you to more deeply understand some of the concepts that were mentioned throughout that particular episode, as well as the segments at the end of each episode called our key ideas segment where I recap certain key ideas. You get episode transcripts, you get then a number of premium only episodes. Some of them might be ask me anything sessions where we answer your questions that you submit it directly, or they could be a variety of other episodes that you may have seen previews to in the public feed. So for full details on this, then check out the link in the description box wherever you're currently listening right now, or just go to sigmonutrition.com and you can see all the details there. And of course, your support is what keeps sigmonutrition going. We don't run ads, we don't sell supplements, anything like that. So your support is what allows me to continue to do this. So thank you for that. I hope you do come back for the next episode regardless. And until then, have a great week, stay safe, and take care.