Fiber

People who eat the most fiber die less often of almost everything, and the more they eat, the bigger the gap. The awkward part is that no trial has ever managed to show that adding fiber causes the benefit — and one trial came out the wrong way.

Dietary fiber is the part of plant food your own enzymes cannot digest. Some of it dissolves and forms a gel that slows digestion; some passes through largely intact; some is fermented by gut bacteria into short-chain fatty acids. The evidence that fiber-rich diets track with longer, healthier lives is among the most consistent in nutrition — and the evidence that isolated fiber does anything to hard outcomes is close to nonexistent. Holding both facts at once is the whole art of reading this literature.

What the evidence says

Strong:

Higher fiber intake is associated with roughly 15–30% lower risk (about 31% for coronary heart disease mortality, the largest single effect) of all-cause and cardiovascular mortality, coronary heart disease, stroke, type 2 diabetes, and colorectal cancer, across 185 prospective studies and roughly 135 million person-years. Certainty on GRADE — the standard system for rating how much confidence a body of evidence deserves — is moderate, the highest any carbohydrate-quality marker achieves.^[1]
A 2025 umbrella review — a review of other reviews — pooled 33 meta-analyses covering 38 outcomes and 17.2 million people, and graded each association on its own credibility ladder — a different scale from GRADE above, ranking how believable an observational association is rather than how good the underlying evidence is. Three outcomes reached its top "convincing" tier: cardiovascular-disease mortality, pancreatic cancer, and diverticular disease. All-cause mortality, cardiovascular disease, coronary heart disease, and ovarian cancer landed one rung below, at "highly suggestive."^[2] Note what this does and does not do: it is an umbrella review of observational meta-analyses, so it makes the association harder to dismiss without making it causal.
The dose-response is approximately linear with no plateau within the observed range: each extra 8 g/day is associated with 7% lower all-cause mortality.^[3]
Viscous fibers lower LDL cholesterol — the "bad" cholesterol fraction that drives arterial plaque — modestly but reliably in randomized trials. This is a genuine causal effect, but on a surrogate marker — a lab value that stands in for the disease you actually care about.^[4]

Moderate:

WHO's 2023 guideline makes a strong recommendation on moderate-certainty evidence: at least 25 g/day of naturally occurring fiber, from food.^[5]

Weak / preliminary:

Every hard-outcome number above is observational. Fiber intake is an almost perfect proxy for an overall healthy diet, more exercise, less smoking, and higher socioeconomic status.

Caution:

No randomized trial has ever shown that increasing fiber reduces mortality, cardiovascular events, or cancer. Several good ones tried and failed.^[6]
Blanket "eat more fiber" advice is actively wrong for some people with irritable bowel syndrome, where fermentable fibers are the trigger.^[7]

What the cohorts show

Evidence: Moderate — large, consistent, dose-dependent, and entirely observational.

The anchor is a WHO-commissioned series covering 185 prospective studies and 58 trials.^[8] Comparing the highest fiber eaters with the lowest, the risk of dying of anything at all runs about 15% lower, and of coronary heart disease about 31% lower. In the table, a relative risk of 0.85 means 15% lower risk; the range beside it is the 95% confidence interval — where the true value most plausibly sits. When that range crosses 1.00, the result is compatible with no effect at all.

Outcome	Relative risk (highest vs lowest intake)	GRADE
All-cause mortality	0.85 (0.79–0.91)	Moderate
Coronary heart disease — mortality	0.69 (0.60–0.81)	Moderate
Coronary heart disease — incidence	0.76 (0.69–0.83)	Moderate
Stroke incidence	0.78 (0.69–0.88)	Low
Type 2 diabetes	0.84 (0.78–0.90)	Moderate
Cancer mortality	0.87 (0.79–0.95)	Moderate
Colorectal cancer	0.84 (0.78–0.89)	Moderate

The dose-response matters more than the headline. Every additional 8 g/day is associated with roughly 7% lower all-cause mortality (RR 0.93, 95% CI 0.90–0.95), 19% lower coronary heart disease, 15% lower type 2 diabetes, and 8% lower colorectal cancer. Fitting the data with a flexible curve rather than a straight line found no plateau — the curve was still falling at the top of the observed range. A separate analysis comparing intake bands found that 25–29 g/day was the lowest band at which most outcomes improved, which is why the authors treat it as a floor rather than an optimum and note that higher intakes likely confer further benefit. The two findings fit together: 25–29 g is where the benefit becomes reliable, not where it stops growing.

Independent analyses agree. Each 7 g/day of fiber is associated with 9% lower cardiovascular disease risk.^[9] Each 10 g/day is associated with 10% lower colorectal cancer risk, pooled across nearly two million people with zero heterogeneity between studies.^[10] Whole grains — the food, not the isolated nutrient — show associations at least as strong: three servings a day tracks with 22% lower cardiovascular disease and 17% lower all-cause mortality, while refined grains and white rice show no association at all.^[11]

Two honest footnotes the enthusiastic version of this story leaves out. The coronary-mortality analysis showed detectable publication bias — the tell-tale pattern left when studies finding nothing go unpublished, which inflates whatever survives. And more than half the variation between the stroke studies was unexplained disagreement rather than chance, meaning those studies were not really measuring the same thing.^[12]

And now the problem: the trials

Evidence: Strong that the trials are null; Weak that the trials settle the question.

If fiber caused these benefits, trials should show it. They don't.

Colorectal cancer is the sharpest test, because you can randomize people with a history of adenomas — the precursor polyps — and watch whether fiber stops them coming back. Three trials did exactly that, and all reported in 2000:

The Polyp Prevention Trial put 2,079 people on a low-fat, high-fiber, high-fruit-and-vegetable diet for four years. Adenoma recurrence: 39.7% in the intervention arm versus 39.5% in controls — a relative risk of 1.00 (0.90–1.12). Not merely non-significant. Flat.^[13]
The Wheat Bran Fiber trial compared a 13.5 g/day wheat-bran supplement against 2 g/day. Recurrence ran about 12% lower on the high dose, but the result was equally compatible with no effect at all — no significant difference (odds ratio 0.88, 95% CI 0.70–1.11).^[14]
A European trial gave 3.5 g/day of ispaghula husk (psyllium) and found adenoma recurrence went up by roughly two-thirds (odds ratio 1.67, 95% CI 1.01–2.76 — just clearing the threshold for significance).^[15]

Cochrane pooled the randomized evidence — five trials, 4,798 participants, two to eight years — and found nothing — a 4% difference well inside the range expected from chance (relative risk 1.04, 95% CI 0.95–1.13).^[16]

Cardiovascular disease is worse still. The Cochrane review of fiber for preventing cardiovascular disease found 23 randomized trials — and not one of them reported a single cardiovascular event or death. Most ran under 12 weeks. The entire randomized evidence base for fiber and heart disease consists of surrogate markers, where it shows small, real effects: total cholesterol down 0.23 mmol/L, LDL down 0.14 mmol/L.^[17]

The one genuine randomized hard-endpoint test of raising fiber was the DART trial in 2,033 men recovering from a heart attack. The arm advised to increase cereal fiber had slightly higher all-cause mortality than the arm that wasn't (non-significant). Only the fish advice helped.^[18]

And the largest long-running dietary trial, the Women's Health Initiative — 48,835 postmenopausal women, mean 8.1 years on a low-fat, higher-fiber, higher-fruit-and-vegetable pattern — came back null across the board. Cardiovascular disease moved by 2%, well inside the range expected from chance (hazard ratio 0.98, 95% CI 0.92–1.05 — a hazard ratio is the same idea as the relative risks above, applied to the rate at which events occur over time)^[19]; colorectal cancer drifted 8% the wrong way, with an interval that still allowed a modest real reduction (HR 1.08, 0.90–1.29)^[20]; breast cancer came closest to a signal at 9% lower, but its interval still touched 1.0 (HR 0.91, 0.83–1.01)^[21].

How to hold both facts

Evidence: this section is argument about the evidence, not a claim of its own.

The temptation is to pick a side. Resist it, because the trials are weaker tests than they look:

Adenoma recurrence is a surrogate, not cancer, over 3–4 years — against a disease that takes decades to develop.
Confidence intervals are wide. The wheat-bran lower bound of 0.70 still permits a 30% reduction, larger than the observational effect. These trials failed to demonstrate benefit; they did not demonstrate its absence.
The dietary-pattern trials barely changed diets. The Women's Health Initiative achieved a weak between-arm contrast, so its nulls are a poor test of fiber specifically. It was never a fiber trial.
Most were supplements, not food. Wheat bran and psyllium in a sachet are not lentils.

But the cohorts are weaker than they look, too. Fiber intake is entangled with everything else that makes people healthy, and no amount of statistical adjustment fully removes that.

The position the evidence actually licenses is narrower than either camp wants: eat fiber-rich whole foods, because the association is strong, dose-dependent, and consistent — but do not believe that isolated fiber has been shown to prevent disease, because it hasn't, and in one trial it went the wrong way. WHO drew precisely this line: its 25 g/day recommendation explicitly covers "naturally occurring dietary fibre as consumed in foods," and explicitly excludes extracted, synthetic, and added fiber — supplements, capsules, and powders.^[22]

That is not an accident of drafting. It is the world's normative guideline body declining to extend its recommendation past its evidence.

Which fiber does what

Evidence: Strong for viscous fiber and cholesterol; Weak for everything downstream of fermentation.

The old soluble/insoluble split is a poor guide. What actually predicts an effect is viscosity (does it form a gel?) and fermentability (do gut bacteria eat it?) — and the two properties do different jobs.

Viscous fibers lower cholesterol. This is the one place isolated fiber has clean randomized evidence. Oat beta-glucan at a median 3.5 g/day lowers LDL by 0.19 mmol/L (0.14–0.23) — roughly 5%, pooled across 58 trials.^[23] Psyllium does more — LDL down 0.28 mmol/L and total cholesterol down 0.38 mmol/L, with a clear dose-response from 3 to 20 g/day — though this was measured in people who already had raised cholesterol, so it is a treatment effect rather than a prevention one.^[24] Both work by trapping bile acids in a gel so they're excreted rather than recycled, forcing the liver to pull cholesterol from the blood to make more. Real, modest, and worth knowing if you are managing cholesterol.

Fermentable fibers feed the microbiome. Inulin, resistant starch (starch that escapes digestion in the small intestine and reaches the colon intact), and the fibers in legumes and whole grains are broken down by colonic bacteria into short-chain fatty acids, principally butyrate — the main fuel of the cells lining your colon, and the mechanistic heart of most fiber enthusiasm. That story is covered under dysbiosis. Read it with the species gap in mind: a great deal of the butyrate literature is mouse work.

Bulking fibers move things along. Wheat bran and cellulose are neither viscous nor readily fermented; they add mass and speed transit. Useful for regularity, and — as the trials above showed — useless for preventing polyps.

So there is no single "best fiber." Beta-glucan and psyllium for lipids; a broad mix of plants for the microbiome; and for the mortality association, no isolated fiber at all — the exposure that carries it is whole food.

How much, and from what

Evidence: Strong — these are measured food composition values and national intake surveys, not inference.

Most people eat about 16 g/day — men around 18, women around 15 — against a target of 25–30 g or more.^[25] That is the gap worth closing, and it closes faster than people expect, because legumes are extraordinary and everything else is incremental.

Food	Portion	Fiber
Split peas, cooked	1 cup (~195 g)	~16 g
Lentils, cooked	1 cup (~200 g)	~16 g
Black beans, cooked	1 cup (~170 g)	~15 g
Chickpeas, cooked	1 cup (~165 g)	~12 g
Artichoke	1 medium (~120 g)	~10 g
Chia seeds	30 g (2 tbsp)	~10 g
Avocado	1 medium (~200 g)	~10 g
Raspberries	1 cup (~125 g)	~8 g
Pear, with skin	1 medium (~180 g)	~5.5 g
Oatmeal, cooked	1 cup (~235 g)	~5.5 g
Broccoli, cooked	1 cup (~155 g)	~5 g
Apple, with skin	1 medium (~180 g)	~4.4 g
Almonds	30 g	~3.5 g
Popcorn, air-popped	3 cups (~24 g)	~3.5 g
Whole-wheat bread	1 slice (~30 g)	~2 g

Fiber values are USDA standard portions; gram weights are the corresponding standard portion sizes, rounded.^[26]

The arithmetic is unsentimental: one cup of lentils is worth eight slices of whole-wheat bread. A person eating 16 g/day who adds a cup of legumes and a cup of raspberries clears 35 g without touching anything else. Chasing the target through bread, cereal, and "high-fiber" bars is the slow, expensive way to do it.

Titrate up over two to three weeks and drink more water. Going from 16 to 30 g overnight is how people conclude they "can't tolerate fiber," when what they actually did was give their microbiome a fortnight's work in one evening.

When more fiber is the wrong advice

Evidence: Moderate for IBS; Weak for the constipation reversal; Strong that the diverticulitis avoidance advice was wrong.

This section matters, because generic fiber cheerleading does real harm to a minority of readers.

Irritable bowel syndrome. The fermentable fibers — inulin, fructo-oligosaccharides, galacto-oligosaccharides — are FODMAPs (fermentable oligosaccharides, disaccharides, monosaccharides and polyols), the exact family of short-chain carbohydrates that a low-FODMAP diet strips out to control IBS. Telling someone with IBS to eat more of them is telling them to eat more of their trigger. In a randomized primary-care trial, psyllium improved IBS symptoms while wheat bran made them worse — dropout was highest in the bran arm, and the commonest reason was symptom worsening.^[27] If you have IBS: viscous, low-fermentation fiber (psyllium, oats) yes; bran and inulin, probably not.

Chronic constipation — sometimes. The reflex is to add fiber. In a study of 63 people with constipation that had no identifiable cause, those who stopped fiber entirely went from a bowel movement every 3.75 days to daily, with bloating and straining resolving completely, while those who stayed on high fiber saw no change at all.^[28] Take this with real caution — it was not randomized, and patients chose their own group, so the high-fiber arm is plausibly made up of the people fiber was never going to help. But it is enough to justify the clinical rule: if you have added fiber for constipation and it made things worse, believe yourself and stop, rather than adding more.

Diverticular disease — the myth is dead. For decades, people with diverticulosis were told to avoid nuts, seeds, and popcorn in case fragments lodged in the pouches. Following 47,228 men for 18 years, the highest consumers of popcorn had about 28% lower diverticulitis risk (HR 0.72, 95% CI 0.56–0.92), and nut eaters trended about 20% lower with an interval that just brushed no-effect (HR 0.80, 0.63–1.01). Lower, or at worst unchanged — in no analysis higher. The authors concluded the avoidance advice "should be reconsidered."^[29] It has been. Eat the nuts.

One safety note: psyllium swells on contact with water, and taking it with too little fluid is the one way a fiber supplement can genuinely hurt you — it can obstruct. Take it with a full glass.

What's overrated

Fiber supplements as disease prevention. No randomized trial shows they prevent anything that matters, one raised adenoma recurrence by roughly two-thirds, and WHO deliberately wrote them out of its recommendation. Psyllium is a legitimate tool for lipids, IBS, and bowel regularity — three surrogate jobs it does well. It is not a longevity intervention.
"Added fiber" in processed food. Chicory-root inulin sprayed into a protein bar is not what the cohorts measured. The exposure that carried the mortality association was legumes, whole grains, vegetables, fruit, nuts, and seeds — the food matrix, not a number on a label.
The exact gram target. Twenty-five grams is a floor derived from an observational curve with no plateau, not a threshold where benefit switches on. Getting from 16 to 25 matters far more than getting from 28 to 30.
Fiber as the whole story of carbohydrate quality. It is the better-evidenced marker than glycemic index — that's a claim about certainty, not effect size. See Glycemic index, where the opposing case is presented too.

Bottom line. Eat 25–30 g a day, from legumes, whole grains, vegetables, fruit, nuts, and seeds. The association with living longer is strong, dose-dependent, and about as good as observational nutrition evidence ever gets. Just be honest with yourself about what it is: a very good reason to eat plants, and not a demonstrated causal lever you can buy in a tub.