Collagen, Seed Oils, Sweeteners and Creatine: What the Evidence Actually Says
A three-hour evidence review that takes apart five headline topics: the collagen meta-analysis that supposedly debunked collagen, ketone production as liver protection, butyrate against colon cancer, the seed oil panic, and the creatine study that claimed creatine only adds water weight. The recurring lesson is that a headline conclusion and the underlying data table often tell different stories.
Overview
This podcast episode works through the topics the host covered over the past year, but with the commentary and pushback the short videos never had room for. It opens with the 2025 collagen meta-analysis that split its randomised trials by funding source and reported no significant effect in the independently funded group — yet six of those seven studies individually showed a benefit, and the null result reflects low statistical power rather than absence of effect.
The discussion then covers the fair critique that collagen has never been compared gram-for-gram against protein, and the receptor-signalling mechanism that could explain a unique collagen effect. Next comes a study in which mice unable to produce ketones developed far worse liver fat and fibrosis, reframing high ketones as a release valve rather than a warning sign. A large section is devoted to butyrate, which accumulates inside glucose-dependent colon cancer cells and shifts gene accessibility away from growth and towards cell death — with fibre, not butter, as the sensible source.
The seed oil segment reviews roughly 150 to 200 studies showing neutral-to-slightly-positive effects for omega-6, notes that the omega-6 to omega-3 ratio only tracks with risk far above typical intake, and flags repeatedly reheated oil as the one weak negative signal. The episode closes with the creatine study whose clamped training volume removed creatine's main mechanism, and with non-nutritive sweeteners, where an eight-year prospective association on cognitive decline meets a near-total absence of randomised trials on brain outcomes.
Key quotes
5Collagen peptides seem to be effective for improving your skin. It's a mild effect. Don't expect anything completely transformative.
The scientific literature can sometimes be pretty far behind the anecdotes that come out.
Ultimately the clinical data is always going to trump any sort of mechanistic research.
People are so focused on seed oils that they're kind of missing the forest for the trees.
We are deeming non-nutritive sweeteners as safe and fine for cognitive health, but we haven't actually tested it.
Key ideas
9The funding-source split that made headlines
A 2025 meta-analysis of around 28 randomised trials divided the studies by industry versus independent funding and by risk-of-bias grade. Industry-funded trials showed a benefit of collagen peptides on skin; the independently funded subgroup reported no statistically significant effect, which is where the debunking story came from.
Statistical significance is not the same as no effect
Opening the study table shows six of the seven independently funded trials individually favoured collagen, with only one showing nothing. Shrinking the pool from 28 studies to 7 collapses the certainty of the estimate, so the honest reading is not enough data rather than no effect.
The comparison nobody has run
None of the included trials compared collagen peptides against an equal amount of ordinary protein, which is the critique the host calls excellent. One burn-recovery trial favoured collagen over protein and one cell study showed a unique collagen effect, but that is far too thin to settle the question.
Signalling, not building blocks
The usual objection is that cells break peptides into amino acids anyway, so protein should do the same job. The counter-argument is that collagen peptide receptors sit on the cell surface, so the binding itself may tell fibroblasts to lay down mature collagen fibres — plausible, but not yet proven.
Ketone production as a liver release valve
Blood ketones track with worse liver health, which looks alarming until you see the mouse work. Animals engineered so they cannot produce ketones developed markedly more liver fat, damage and fibrosis, suggesting ketogenesis is the liver's way of exporting fat rather than the cause of the problem.
Why butyrate strangles colon cancer cells
Colon cancer cells preferentially run on glucose and cannot burn butyrate, so this short-chain fat builds up inside them. It then interacts with the enzymes that tag DNA packaging, closing access to pro-growth genes while opening the ones that drive cell death, all while sparing healthy colon cells.
Fibre beats butter as a butyrate source
Butter does contain butyrate, but it arrives packaged with saturated fats linked to less favourable outcomes. Fermentable fibre produces butyrate and other beneficial short-chain fats in the gut, and fibre itself carries a large, consistent association with lower cancer incidence.
Seed oils: the clinical data outweighs the mechanism
The oxidation and inflammation mechanisms are real talking points, but across roughly 150 to 200 studies omega-6 intake shows a neutral to slightly favourable relationship with cardiovascular outcomes. Many of the trials pointing the other way are decades old and would not survive modern peer review.
The sweetener question is an evidence gap
A prospective study over about eight years associated higher non-nutritive sweetener intake with faster cognitive decline, with erythritol and xylitol carrying the strongest signal and the risk concentrated in insulin-resistant participants. The real surprise was how few randomised trials have ever tested these sweeteners on cognitive outcomes at all.
Practical takeaways
7- 1
Read the table, not just the abstract 12:00
When a summary says no significant effect, open the study list and check how many individual trials pointed which way. A small subgroup can produce a null result purely because there is not enough data to detect a real signal.
- 2
Set realistic expectations for collagen 21:00
The signal points to modest improvements in skin hydration, elasticity and wrinkling over 8 to 16 weeks. Treat it as a small compounding advantage, not a reversal of ageing.
- 3
Get butyrate from fibre 1:01:00
Fermentable fibrous foods let your gut microbes make butyrate and related short-chain fats without the extra molecules that come with fatty sources. It is the option with the deepest supporting literature.
- 4
Raise omega-3 rather than fearing omega-6 1:32:00
Risk only climbs when the omega-6 to omega-3 ratio sits far above typical intake. Nudging omega-3 up moves you into the favourable part of the curve without micromanaging every cooking oil.
- 5
Always ask what you are replacing it with 1:46:00
Cutting one fat only helps if what replaces it is better; swapping seed oils for trans fats is a clear step backwards. The substitution matters more than the ingredient in isolation.
- 6
Creatine needs room to work 2:01:00
The study that found no lean mass benefit kept training volume identical between groups, which removes creatine's main mechanism of allowing more work. If you supplement, let your sessions actually get harder.
- 7
Decide your own sweetener tolerance 2:19:00
The strongest associations sit with erythritol and xylitol, and the evidence on brain outcomes is thin in both directions. Small shifts — lemon in water instead of a sweetener — cost nothing if the uncertainty bothers you.
Topics & chapters
15Why this podcast exists
An unhurried, once-a-year format with no slides and no time pressure. The plan is a handful of topics plus the best community pushback.
The collagen meta-analysis
Around 28 randomised trials split by funding source and risk-of-bias grade. Industry-funded studies favoured collagen; independently funded ones reported no significant effect.
Power, not absence of effect
Six of seven independently funded trials individually showed benefit. The null comes from a shrunken data pool rather than from collagen failing.
Collagen versus protein
The strongest critique is the missing gram-for-gram protein comparison. One burn study and one cell experiment hint at a unique collagen effect.
Receptors on the cell surface
Collagen peptide receptors could mean the signal, not the amino acids, drives new collagen fibres. The host is explicit that this is not yet settled.
Community comments on collagen
Anecdotes on knee recovery and long-horizon effects. Evidence often lags experience, but anecdote alone cannot carry a conclusion.
Ketones and liver health
Mice unable to produce ketones developed worse liver fat and fibrosis. High ketones look like compensation for liver stress, not its cause.
Reader stories on low carb and the liver
Two listeners reversed fatty liver disease with ketogenic approaches and weight loss. Impressive, but one route working does not rule out others.
Butyrate against colon cancer
Butyrate accumulates in glucose-dependent cancer cells and shifts which genes stay accessible. Growth genes close, cell-death genes open, healthy cells are spared.
Where butyrate should come from
Butter delivers butyrate alongside less desirable saturated fats. Fermentable fibre is the cleaner route and carries its own strong cancer literature.
The ketogenic cancer debate
A long, detailed reply to a listener on the Warburg effect, beta-hydroxybutyrate and glutamine restriction. The conclusion is a hybrid model, not a single metabolic cure.
Seed oils: mechanism versus clinic
Oxidation and inflammation pathways are plausible but the clinical data dominates. Across roughly 150 to 200 studies omega-6 looks neutral to slightly favourable.
The omega-6 to omega-3 ratio
Risk rises only at ratios far beyond typical intake, measured in blood and tissue. Ordinary consumption sits squarely in the neutral zone.
Canola oil and repeated heating
Nine of ten trials showed neutral or favourable inflammation results. The one exception involved oil reheated roughly twenty times.
The creatine study and the sweetener question
Clamped training volume removed creatine's main mechanism, so the null result is a caution rather than a debunking. The episode ends on sweeteners, cognitive decline and the missing trials.