[Home](/)

What we know

What a comparison like this can show

[How do plant-based and animal-based diets compare for health?](/what-we-know/health-comparisons/)

1What a comparison like this can show

[2Why a mechanism is not an answer](/what-we-know/health-comparisons/why-a-mechanism-is-not-an-answer/)

[3Protein: how much, and from where](/what-we-know/health-comparisons/protein/)

[4Calcium and bone: the milk question](/what-we-know/health-comparisons/calcium-and-bone/)

[5Fats: which one replaces which](/what-we-know/health-comparisons/fats/)

[6Heart disease and stroke](/what-we-know/health-comparisons/heart-disease-and-stroke/)

[7Death from any cause](/what-we-know/health-comparisons/death-from-any-cause/)

[8The reference diet, and what it projects](/what-we-know/health-comparisons/the-reference-diet-and-what-it-projects/)

What we know · Part 1 of 8

# What a comparison like this can show

Entry checked 20 September 2026 · 4 sentences, each with its source · no person approved these sentences · [the entry's full limits](/what-we-know/health-comparisons/#limits)

Almost nothing on this page is an experiment in the ordinary sense. You cannot randomly assign people to eat one way for twenty years, so most of the evidence comes from large groups who were asked what they ate and then followed for years, and from trials that changed one thing in the diet and measured what happened next. The first kind finds associations. The second kind can show cause, but only for the narrow question it asked.

Three things decide how much a number is worth: what it was compared against, what the analysis adjusted for, and whether the thing measured is a marker or an outcome. A marker is something you can read in a blood sample. An outcome is something that happens to a person. They do not have to agree, and on this page there are several places where they did not.

A 2025 workshop of more than 20 international protein researchers rated most widely repeated propositions about dietary protein as needing more research, and found the evidence base often insufficient in rigour, sample size, duration, or the use of surrogate markers. [[1]](#source-1)

The researchers themselves are not agreed on how much of the protein story is settled. That is the first thing worth knowing about a field where the headlines are confident.

The authors of the Swedish milk cohort state that their own findings should be read cautiously, because an observational design carries the possibility of residual confounding and reverse causation. [[2]](#source-2)

The calcium-and-fracture review found signs of publication bias in the trial literature, reporting that funnel plot inspection and Egger's regression suggested bias toward calcium supplements in the published data. [[3]](#source-3)

The last caution is about scoring foods one at a time. Most of the numbers used to compare protein sources come from a method that measures a single food in isolation and names the one amino acid it lacks.

A review of the whole-diet approach to protein quality proposes that scoring single foods systematically underestimates protein adequacy in real diets, and reports corrected values 18% to 30% higher when digestible amino acids are pooled across a day's intake. [[4]](#source-4) [[4]](#source-4)

## What this section does not show

Where the numbers come from

Almost nothing here is a randomised trial. The cohorts followed people who chose their own diets, so an association is not a cause, and the page says so at each result rather than once at the top.

The comparator decides the meaning

A finding is only readable with what it was compared against. "Vegetarian" in the largest cohort means vegetarians and vegans together, and the fat findings are substitution findings rather than statements about eating less.

Not a statement about you

These are averages over large groups. Nothing here knows your age, your medication, your test results or your history, and a population association does not transfer to one person.

What we did not check

The full 2025 EAT-Lancet report sits behind a publisher's paywall and was not read, so the entry quotes the Commission's own summary. Paywalled full texts of the cohort papers were not read either, so each sentence rests on the abstract retrieved on 20 September 2026.

All twelve elements

The entry's full limits block covers the dates, the study designs, who was and was not studied, the comparators, the adjustment sets, the uncertainty, the counter-signals and what we did not check. [Read it on the entry's first page.](/what-we-know/health-comparisons/#limits)

This page is general information about published research. It is not medical advice, and it is not a substitute for advice from a qualified health professional. No clinician-patient relationship exists, and we do not know your diet, your medications, your test results, your medical conditions or your history. Talk to a health professional before changing your diet if you are pregnant or breastfeeding, feeding an infant or child, over 65, taking medication, living with a medical condition, or if you have been diagnosed with a nutrient deficiency or have a history of disordered eating. The evidence described is about groups and diets, not about you, and not about any product listed on this site. If you are unwell, contact a health professional; in an emergency, call 911. No clinician reviewed this entry. It is a compilation of published sources by this site.

[Back toThe whole entry](/what-we-know/health-comparisons/) [NextWhy a mechanism is not an answer](/what-we-know/health-comparisons/why-a-mechanism-is-not-an-answer/)

## Sources for this section

Every numbered sentence above was checked against the source with that number on the date shown. The number links to the entry in this list, and the entry's first page lists every source it cites.

Examining widely held propositions on human dietary protein needs and benefits: a critical review of the science that shapes both the data and our understanding of an essential macronutrient. Critical Reviews in Food Science and Nutrition. critical review. [Open the source ↗](https://doi.org/10.1080/10408398.2026.2658728) Checked 19 September 2026. Backs 1 sentence in this section.

Milk intake and risk of mortality and fractures in women and men: cohort studies. The BMJ. prospective cohort studies. [Open the source ↗](https://doi.org/10.1136/bmj.g6015) Checked 19 September 2026. Backs 1 sentence in this section.

Calcium intake and risk of fracture: systematic review. The BMJ. systematic review. [Open the source ↗](https://doi.org/10.1136/bmj.h4580) Checked 19 September 2026. Backs 1 sentence in this section.

From Single Foods to Whole Diets: A Critical Framework for Protein Quality Assessment Accounting for Complementarity and Matrix Effects. Comprehensive Reviews in Food Science and Food Safety. critical review. [Open the source ↗](https://doi.org/10.1111/1541-4337.70571) Checked 19 September 2026. Backs 1 sentence in this section.

How this was written

## Method and care

Every factual sentence here was checked against a retrieved copy of the source it names: one sentence, one retrieved source, one verbatim quote, one checked date. No person reviewed or approved these sentences before publication.

The journal sources were retrieved as their Europe PMC records, which is the quotable text that registry serves; the EAT-Lancet material was retrieved as the text of the Commission's own summary report. The entry records, for each source, what was retrieved and what was not.

Reviewed by: — no named reviewer has reviewed this entry, and no clinician has. It is a compilation of published sources by this site.

Written by: Find My Next Bite's writing session, 20 September 2026, working from sources retrieved the same day. No person reviewed or approved these sentences before publication, and no clinician has read them.

Funding: this section is unfunded. It takes no sponsorship, and the rule we adopted in advance says no funder may ever be a company whose products the entries or the directory describe.

Corrections: if a source here has changed, moved or been retracted, tell us and we will suspend the sentence and record the change.

Every change this site makes to a sentence it has published is recorded, dated, at [Corrections and revisions](/what-we-know/corrections/) — what it said before, what it says now, and which kind of change it is. Nothing here is rewritten quietly.

[Read the entry from the start, with its question, its position and its full limits.](/what-we-know/health-comparisons/)

Request a correction ↗

[Read our approach](/#how-it-works)

[Explore the food →](/foods/)