Twenty-three randomised trials pooled together put berberine’s effect on body weight at 1.9 lb. Semaglutide, in the trial that made it famous, took off 33.7. Both numbers come from the best evidence available for each.
The trial behind the 2024 headlines was retracted in September 2025 — four days after a meta-analysis pooled it. Ten months on, that meta-analysis still carries no correction. What survives: a real but modest effect on post-meal blood sugar, and about four pounds in the best weight trial, most of it back within a month of stopping.
Partially hydrogenated vegetable oil genuinely harmed people: when 11 New York counties restricted trans fats, heart attacks and strokes fell an extra 6.2%. That fat is gone. The fear outlived it and reattached to the oil that's left — which is usually soybean, and which the bottle doesn't tell you.
Pooled across 28 randomised trials, eggs raise LDL by 5.55 mg/dL — so 'eggs don't affect your cholesterol' is wrong. But adjust for total dietary cholesterol and the egg association with heart disease vanishes entirely. And two meta-analyses disagree about the one thing the 'it's a wash' argument needs.
Processed meat carries an 18% higher colorectal cancer risk per 50 g a day — and two reviews covering four and six million people concluded the absolute effects are 'very small' with low-certainty evidence. Both are true. Here is why, what IARC's Group 1 actually means, and why steak is a different question.
Pooled across 72 measurements, instant oatmeal scored 75 on the glycaemic index and granola scored 56 — the breakfast with the health halo performs worse than the one people feel guilty about. The real, regulator-approved oat claim is about cholesterol, at 3 g of beta-glucan a day, and it saturates there.
Researchers measured urine pH across citrus juices: orange and grapefruit alkalised it, and lemon didn't — the ritual picked the wrong fruit. But lemon does have a real, evidence-backed use. It's kidney stones, it's for people who already get them, and orange juice may be better at it.
The famous PREDIMED trial was retracted from the New England Journal of Medicine and republished — and the nuts arm did just as well as the olive oil arm. Across 92,000 people over 28 years, olive oil beat butter and margarine handily. Against other vegetable oils, no significant difference at all.
Canola is rapeseed bred low in erucic acid — the name literally means Canadian oil, low acid. The FDA ban people cite was against mustard oil, not this. Even in the rat studies the heart lesions were judged 'not related to erucic acid per se'. And in 13 head-to-head trials, canola lowered LDL more than olive oil did.
Cut dietary linoleic acid by 90% and tissue arachidonic acid doesn't move — the inflammation mechanism fails at step one. But the two recovered BMJ trials showed no mortality benefit and one showed higher deaths, which the reassurance articles never mention. Here is the whole file, including what's wrong with both halves.