Episode 3 - Fiber - How much do you 'Need'?

Ajouté : 2026-05-22

[00:00:00]Welcome back to the channel. Today's episode is on fiber and whether or not it's good for you as part of a longevity program. Recall I'm Dr. Vanipi Prasad. I read four longevity books and hundreds of the underlying scientific papers to put together this series. And after reading about fiber, I feel bloated already. So, let's get into it.

[00:00:19]Spoiler alert, every single book and every single research article on this topic, would you believe? Universally, every single one believes that eating fiber is good for you. And you need more fiber. The average American's not eating enough fiber. Now, fiber containing foods are fruits, vegetables, nuts, legumes, and guess what? There's no doubt about it. Those are good foods for you. Those are good to eat. They improve blood cholesterol, blood glucose. They improve satiety. They may help you lose weight. All these wonderful things. Some people like to say they feed the microbiome which is so good for you. All these things have been reported. Let's take a look at the underlying evidence.

[00:00:54]It's all nicely summarized in this table which shows that fiber well it's not all the same thing. You've got fiber that can hang on to water that can't hang on to water and among that can hang on to water in the gut. Some actually make a gel-like substance which is thought to slow absorption in the illium which thought to increase satiety. And in prospective randomized studies of participants against placebo controls, some types of fiber like psyllium actually can lower your cholesterol, improve your short-term glycemic control, increase satiety, perhaps even increase weight loss. It's all shown nicely in this table. But that's not true for wheat brand or wheat dextrin or inulin, which are different types of fiber. So fiber can bulk your stools, but some fibers can improve glycemic control. This is a bit of nuance that's left out of the books. And finally, whether or not actually eating plant-based material versus having fiber supplements do the same thing, that's a complete black box. That's something that's not been adequately studied in this literature. Now, a couple of the books go with the recommendation that's made by professional associations that Americans need 30 to 35 gram, but we get only 14 grams of fiber. So, when I read words like need, I want to know where does that need come from? Where is need?

[00:02:07]It's need. you need it. Well, you need it because presumably that's the level beyond which you're going to have great outcomes. So, let's take a look at that data. So, this is called the association between dietary fiber intake and stroke among US adults. So, this paper is primarily looking at stroke. Why?

[00:02:25]Because they're publishing in the journal stroke. And if you're publishing in the journal stroke, you better focus on stroke. And this is from a NHANES which is a national food frequency questionnaire that we surveyed people about what they eat for many years. We follow them up many years later to see their health outcomes. And this does something that the older papers don't do, which is mandelian randomization. Oh boy, what is that? Does it make sense?

[00:02:47]What do they find here? I'm going to walk you through it. They take a whole bunch of nhane cycles where they survey people about what they eat and you can get a sense of how many blueberries and how many nuts you're eating and those sorts of things. And they follow these people out into the future so they know how they do. They've measured a bunch of co-variants such as how old they are, whether they smoke, um what is their poverty to income ratio or so their income ratio beyond the poverty threshold and they put people in two groups. Those are the people who didn't experience a stroke or those who eventually experienced a stroke. This paper is also very concerned with if you've had a stroke, should you continue to hit that fiber really hard? And for the purposes of this video series, I'm not going to focus on that because I've already stipulated at the outset video one. This is going to be a video series about healthy people. So, we're going to focus on the people who have not yet had stroke. We're looking at incident stroke, the rate of having stroke. And then there's this DAG at the bottom, which is a directed asyclic graph, which shows you something that I'm going to get to in a little bit. All right. So, what do they do here? They do a food frequency questionnaire. dietary intake information was calculating the type and amount of food or drink ingested in 24 hours before the interview to the calorie intake, nutrients, and other components. Now, this is based on interviews and self-reporting. So, I just want to start off by saying, boy, all the usual caveats. Are people accurate about what they ate or drank?

[00:04:10]Are they reporting it faithfully? Are the people coding that reporting actually capturing how many grams they are actually ingesting? And I think you would be fair to conclude that there's error in every one of these things. And so when somebody says that he ate 14 grams, he probably ate something between 10 and 25 grams. They got no clue exactly how many grams he ate. However, this is an error, a measurement error that occurs on average in every single individual and occurring in an unbiased way. Sometimes too high, sometimes too low. And if you still see a relationship at the end of that, that is a relationship that's only a reduced or diminished relationship than the real relationship. In other words, mclassification errors that occur at random bias towards the null. They bias towards not seeing a signal when a signal exists in terms of dietary fiber and health outcomes. And they do not exaggerate because it's really at random. I I find it hard to believe that there's more error among people who eat less fiber versus more fiber, among people who have coorbidities, etc. So, that's worth pointing out. Here's table two. Table two, it should be table one.

[00:05:15]It's a it's the traits of the study population sorted by dietary fiber intake. It's a big table. I want to walk you through something I think is very interesting. So, they take all the people and they calculate based on their approximations how much fiber they think they eat per day and they put them in tiles. So the bottom third, the middle third, the top third, the the ones that really love to hit that fiber hard. And here they're showing you the difference in their demographics. So let's look at 13,000 12,000 people in each category.

[00:05:43]Let's look at BMI. So if you're in the lowest turtle one, your BMI is 29 on average. If you're in turtle 3, it's 28.

[00:05:51]And that p value is ultra significant because there's a whole bunch of people in this study. So it tells you that people who are eating more fiber are of lower BMI than people who eat a little bit less fiber, but it's not a huge difference. It's a modest difference across this population. Now, is this evidence the study is biased? Well, no, because we think that the fiber is the reason or at least part of the reason why they might have lower BMI. So, you would expect the people who are eating more fiber to have a lower BMI because it's creating satiety and they're not eating a whole bunch of other garbage as I discussed in the prior videos. What about hypertension? Look at hypertension. they have a lot less hypertension if they eat more fiber.

[00:06:26]Well, is that a bias in the data set?

[00:06:28]No. Again, no. Because we think that the fiber is good and it lowers your blood pressure and prevents you from eating foods that increase blood pressure. So, it would be natural that 45% of people have high blood pressure in the lowest tertile and that's only 40% in the highest tertile. So, those are natural are things that we'd expect. They're on the causal pathway between eating the fiber and the health outcomes we care about. We think the fiber causes these things but there are a whole bunch of things that bit harder to explain. So turns out men are eating more fiber are among people in the highest tertile they're more likely to be men than the lowest tertile. 40% of people in the lowest tertile are men 60% in the highest tertile. Okay. Education level.

[00:07:10]Hm. People in the lowest tertile only 55% got more than high school but 65% in the highest tile. Marriage status. Wow.

[00:07:19]Fiber keeps you married. You're more likely to be married to living with your partner if you're in the highest tertile of fiber by about 12 percentage points.

[00:07:26]Smoking status, you're less likely to smoke. Now, these are not on the causal pathway from eating the fiber to the health outcome. These are confounders.

[00:07:35]Turns out, spoiler alert, people eat a lot of carrots and Brussels sprouts and blueberries and nuts. They're different than the people who don't. And this study is really mixing a whole bunch of things. It's mixing all of these confounders that go handinhand with high fiber intake with the fiber intake. So, while I'm willing to to say for the sake of argument that I find it incredibly plausible that fiber may lower your BMI, fiber may keep your hypertension at bay and so that those coariant imbalances do not represent bias. These coariant imbalances do represent bias. They represent confounding and you can adjust for them. That's what they're going to say. You can adjust for them. Well, guess what? You can only adjust for them in so far as you measure them. You can't adjust for them in so far as you don't.

[00:08:20]And these are all crude measures of a broader philosophical concept of living a healthy life. These are crude measures. So I'm guarantee you this study's got residual confounding. It's got persistent bias in this healthy person bias among people who eat a lot of fiber even after adjusting for these things. And if you look at this table too, you see these are just different different people. And that's even after giving them the benefit of the doubt. So boy, what are you going to learn from this?

[00:08:50]You don't need any study. You know, you don't even need to go past elementary school to know that people who eat fruits and vegetables are probably going to be otherwise healthier than people who don't eat any of that stuff. Okay?

[00:09:01]You don't need really any education at all. This doesn't really add to knowledge I think when you have this degree of bias in terms of people who are the high-fiber diet users and indeed they present this figure because I think this figure is really key. This is fiber intake grams per day and then the odds ratio of having a stroke. This is the incident stroke. And what they show you is, boy, every single gram of fiber you're eating is just lowering your risk of stroke against the baseline that that that schmuck who's literally not eating any fiber at all. Zero fiber. That person has the highest risk. And then as you eat fiber, you're getting a lower and lower and lower risk. And look at that. There's just no end in sight. You can just fill up with that 100 grams a day of fiber. And although you're spending probably half your morning on the toilet, you are really reduced to risk of stroke to as low as it goes.

[00:09:51]Okay. What are my problems with this graph? First of all, boy, it's ridiculous. Um, the tertiles, so those three turtiles, you know, they are under 10 g, 10 to 18 g, and then over 18 grams is the top tile. They don't tell me exactly, but how many people are actually eating 50 grams a day? How many people eating 75 grams a day? I've got to imagine it's fleetingly few people in this data set. This data set is really, come on, pushing the limits. Why are you plotting out people who are eating ridiculous amounts of fiber when they're probably so few of those people? These are incredibly volatile estimates of stroke risk. Okay?

[00:10:31]And they're got to be completely different than most mortals who can only stomach so much fiber a day. And so I would cut I would truncate this graph. I throw out all this stuff at the end. And then, you know, of course, I I have no doubt that fiber intake goes handin-hand with health-seeking behavior. So, you see huge reductions. You see 50% reductions in the risk of stroke from from just modest fiber intake. In fact, the average American would be in tertile too. And even the average American doing pretty good with that 14. Um, but is this a real effect or how much of this is sporious? Okay, now the plot thickens. Mandelian randomization. What the hell? So, I think the authors of this paper and other authors of these so-called mandelian randomization studies, they understand that, wow, this is a mess. People eat a lot of fiber, people who exercise a lot, people who are Nordic skiers, people who are runners, people who play tennis four times a day, they're different than other people, not just that they play tennis four times a day or four times a week, they're different than other people because they're otherwise healthconscious. So, there's all these other confounders. So we really have difficulty isolating the benefit of the fiber itself or the tennis or the yoga etc. So that's shown here in this directed as cyclic graph that they're the confounders that affect both the fiber intake and the outcome of stroke.

[00:11:51]But is there a way using genotypes and genomics to get an unbiased estimate of fiber intake and stroke and that enters the instrumental variable the snip. Wow.

[00:12:06]single nucleide polymorphism. So here's what they're doing in this study.

[00:12:09]They're saying, "Listen, we know these people who are eating a lot of fiber are different than people who don't in ways beyond the fiber eating. But what if there were some genotype signature? Some things in your genome that gave you a propensity to eat more or less fiber?

[00:12:23]Some people with certain genetic signatures are more likely to eat fiber than those without those signatures.

[00:12:29]Let's say hypothetically these are single nucleotide polymorphisms. So that they were undergoing reassortment and distribution in the general population really without any rhyme or reason that they were really being randomly assigned by God or mother nature or how whatever you whatever you think they're being randomly given out these snips. Now, of course, everybody with the eat a lot of fiber snip is not going to eat a lot of fiber. And everyone with the don't eat any fiber at all snip is not going to some of them will also eat fiber. But on average, people with those snips that predispose you to eat more fiber are going to eat more fiber on average. And the people with the snips that predispose you to eat less fiber, they're going to eat less fiber on average. If we use the snip as the exposure, that genotypic signature and not the fiber itself, maybe we will bypass all the confounding like that healthy country club person with a low snip for um for uh fiber intake is going to be balanced by a healthy country club person with a snip for high fiber intake. And so by looking at the snips and then and similar in the same category of snip, there'll be some lower income people. the lower- income person who happens to have a who relishes kale and the lower-inccome person who relishes Twinkies. You know, these snips will be distributed across the confounders. And so, we'll be able to get an unbiased estimate of the snip for fiber intake and the outcome. Now, of course, it's going to be a little bit diluted just like the measurement error issue. It's going to be diluted because not everyone with the snip for high fiber is going to eat fiber. But on average, there should still be some differences. Okay. Boy, the moment I hear some idea like that, I think to myself, poppycck. Now, I think to myself, that's such a crazy idea. The genes for eating fiber, you know, do they exist? And I also just want to draw the readers, the viewers attention to the fact that mandelian randomization doesn't always mean the same thing. You know in the hematology oncology literature we look at mandelian randomized studies for bone marrow transplant but that means whether or not you have an HLA matched sibling available for transplant or not available for transplant not whether or not you got the transplant or not but that HLA matched signature is something very different than a snip. Okay? So it's not always the same type of genotypic signature. But anyway, long story short, or put another way, what they're saying is instead of looking at the fiber intake per se on the questionnaire, we're going to look at something that you're born with that was distributed at random, your genotype signature for fiber intake. That's what we're going to do. And we think that because it's distributed at random by God or by mother nature, then it's not going to be linked to all those confounders like your wealth or your marital status, etc., etc. Okay, that's the premise of what they're doing. Now they claim that they have indeed found many such snips and they find a whole bunch of these snips and that they are linked to fiber intake. You know this is something that with a few hours we could delve into but I would admit at the outset I'm a little bit incredulous that they would find such snips but they claim to find such snips and they claim to study those snips on the outcome of stroke and here's what they find. Let me just say one thing at the outset. You can believe they found the snips or you didn't find the snips. If you believe that they couldn't find the snips, then you think this whole type of analysis is stupid and shouldn't be done. I tend to fall in that category. But if you believe they did find the snips and they themselves believe they found the snips, I mean, that's what the authors believe.

[00:15:51]That's how they're selling the paper.

[00:15:53]Then you have to accept the results that you find. Let's look at those results.

[00:15:56]This is what I think is so interesting.

[00:15:58]Those results are broadly a total wash.

[00:16:02]Look at this. Mandelian randomized results. Lacuner stroke odds ratio 987.

[00:16:07]So one means no relationship at all. The conference interval cross is one. Look at these p values. 08.4.96.95.

[00:16:13]This is complete null. Down the list.

[00:16:16]Lacuner stroke no relationship in this data set. That data set. Stroke no relationship. Eskeemic stroke no relationship. No relationship. No relationship. Large art stroke no relationship. Cardiamolic stroke no relationship. And yet somehow in the abstract they spin it a different way.

[00:16:29]They write quote the two sample mandelian randomized analysis showed that genetic prediliction for fiber intake supported a causal relationship between dietary fiber intake and reduced risk of small vessel stroke with a p value that almost touches one. What they do is they run a million I mean not not a million but like a few dozen analyses.

[00:16:49]They throw away all the negative ones and they pick like the one positive one and they throw that in their abstract.

[00:16:55]Completely ridiculous. They've completely misframed their analysis. If you believe their Mandelian randomized study, they claim they believe it. Then you should conclude, oh boy, maybe we're completely wrong about this fiber stroke relationship because the genes that predict eating more fiber when we've completely cut out all these confounders have no correlation with stroke at all.

[00:17:15]We see this tiny signal for small vessel disease. That's probably just noise because the totality of the evidence is stone cold negative. But that's not how they do it. Of course, of course they misrepresent the small finding and draw big attention to it. And of course that error is propagated in the abstract which is propagated in the books which is propagated in everyone who talks about the books because no one's checking the fine print here. Okay, that's no one's checking the fine print.

[00:17:39]But this is a null. This is a completely null mandelian randomization. All right, what's my take-home points here? I think I've I've made my point here. Um what's my take-home points for this whole field? Fiber. Listen, do I believe eating fruits and vegetables and nuts and legumes are good for you? Yeah, I I believe that. Okay. Um, I believe they're good for you. I believe that some types of fiber are specifically linked to lower cholesterol and better blood glucose. Psyllium husk, for example, is linked to those things in randomized studies, not just observational studies, but it's not linked in randomized studies to improvements in all cause mortality because those studies have never been done. Okay. Do I believe fiber can provide satiety and help reduce calorie intake and thus help you meet your weight loss goals? Absolutely. Do I believe supplemental fiber is as good as getting your fiber from broccoli and cauliflower and leafy greens? I'm not sure about that. You know, I have some intrinsic skepticism about that and I've seen no study that can allow me to tease that difference apart. Do I believe that if I eat a little bit more fiber, you know, for each gram of fiber I'm eating, my risk of stroke is plummeting through the floor? No, I don't. I think that the benefits with respect to stroke are likely overstated due to residual confounding and just really a poor methodologic design here. Really poor methods. So overall, I support fiber, but I have to say it's one of those fields where whether you follow the bro science, and by the way, increasingly I see the influencers bro Science say that once you hit your protein goals and you're lifting weight, if you really want to get further shredded, you should increase your fiber. That's what did it for me. That's the bro science view versus the provost of and professor view where you need to eat your dietary target because of this enhances metaanalysis or study. I would say that both are at the same level of evidence.

[00:19:26]I mean I think that's my that's what I think is the most interesting conclusion. They agree on the facts.

[00:19:31]They have different philosophical frameworks for how they get there and both are operating at the same level of science. I think it would be a mistake to put more stock in this epidemiology.

[00:19:41]And the epidemiology, ironically, when they get a result that doesn't go along with their stated conclusion, they suppress it. And so that's why I think you'll never see ever see a study that shows eating blueberries is bad for you.

[00:19:52]It'll never happen. Maybe it'll happen for some really weird end point, you know, but it won't happen for cardiovascular health. The world has decided it's good for you. Studies will either reach that conclusion. They may exaggerate the ex the exact amount it's good for you. uh and if they get the opposite result, well, they won't put that in the abstract. So, those are my conclusions. Fiber, good for you. The overall theme of this series is to separate things that reduce untimely death from things that increase longevity. So, I'm saying instead of dying of a heart attack at 60, you live till 80. That's the untimely death question. Instead of dying at 88, you live to 128. That's the longevity question. I put fiber firmly in the untimely death bucket. You're avoiding untimely death. That's very important.

[00:20:36]That's really good. If people ate more fruits and vegetables and less of the other crap, our whole population life expectancy would get better. But we're not getting into the longevity bucket.

[00:20:46]And the exact amount it's going to get better is likely deeply overstated from this. And then the final thing, adherence, which is something I love to talk about, is very little literature on what happens to healthy people if you advise different regiments of fiber and whether or not they can get there and stick with that for year after year after year. Very little evidence in that space. And I think that's the that's the the missing piece of the puzzle that doing another observational study on stroke or cardiovascular disease or heart failure and fiber is a complete waste of time. All of these studies should of course be defunded by federal government. And that funding should be used to say there are four people here with different programs for how you can put fiber in your diet. And we're going to randomly assign people to those four programs and see how many stick with it 8 months later, 16 months later, 32 months later. The stick with it question, I think, is the question that we now face. So you're listening to medicine unpacked. All right, that was a whirlwind of fibros one paper, but I think we actually got into it and got to the bottom of that paper. And in further episodes, I'm going to get into caloric restriction and longevity. I think it's a really important topic. We're going to talk about many other things. And uh stay tuned for more episodes of Medicine Unpacked. Final plug, if you're in the Bay Area or you are somebody who edits videos, reach out to me at my contact link. I'll put it at the bottom. I need some help making these into short video clips. All right, until next