Sunday, 15 November 2026Nutrition, with receipts
Plate & Proof
A prospective cohort of 17,705 middle-aged and older Chinese adults tracked across cardiovascular-kidney-metabolic stages 0 to 3. After full adjustment, each one-unit rise in the composite index was associated with a 111 per cent increase in all-cause mortality. Observational, one country, one index, no mechanism.
Latest
Thymosin beta 4 prevented cardiac dysfunction and fibrosis in mice 28 days after ischaemia and reperfusion, and the effect vanished when the ErbB2 pathway was blocked. In a randomised, double-blind, placebo-controlled trial of 96 people with a heart attack, infarct size was significantly reduced in the subgroup dosed within 8 hours of angioplasty — and the difference across the whole trial was not significant.
A phase 2 trial of 338 adults reported a mean weight change of −24.2% at 48 weeks in the highest retatrutide group against −2.1% on placebo. A 2025 substudy with DXA scans found total fat mass down by more than a quarter in the higher groups, with a body-composition substudy of 103 people scanned twice. Neither paper says anything about a product sold under a research name.
A 2018 review notes that many multi-ingredient pre-workouts hide their amounts behind a proprietary blend, and that a supplement can carry banned substances as deliberate ingredients or as contaminants. A second review explains the regulatory gap: a supplement manufacturer is not required to submit evidence of safety and efficacy before selling. Neither paper can tell you what is in the vial you are holding; they tell you what a document cannot tell you either.
A network meta-analysis of 28 randomised trials in 23,622 people with type 2 diabetes puts tirzepatide ahead of semaglutide on both HbA1c and weight. An independent review of 22 network meta-analyses, published a year later, says the tirzepatide numbers are less impressive than some meta-analyses claim, because several amounts and time points get merged. Both can be right.
Silymarin pools 26 randomised trials in 2,375 patients with fatty liver disease and moves liver enzymes by a wide margin; N-acetylcysteine has a meta-analysis too, and every one of its thirteen studies is an animal study. Two meta-analyses on coffee disagree with each other about the same question, which is the most useful thing in this file.
A 2004 review of oxandrolone in wasting and catabolic disorders found significant gains in body composition, strength and nutritional status across most well-designed trials — and simultaneously noted that it has never been studied in sarcopenia. The drug is also the only orally active anabolic steroid approved in the US, which tells you something about the alternatives.
Most read
The EU reference value is 0.83 g per kilogram of body weight per day — far less than gym culture assumes, and easier to reach than most people think.
A nutrition label tells you what is in the jar. A certificate of analysis tells you what was measured, in which batch, by which method. We read the federal release rules and the validation guidance for analytical procedures, then set out the five lines worth finding first - and why a perfect-looking PDF is often the wrong sign.
Three-dimensional MRI puts the strain in a healthy free Achilles tendon at about one per cent. Paired-leg training studies show stiffness rising while cross-sectional area often does not. The gap between loading a tendon and changing its structure is where most of the confusion lives.
Claims we checkedSee all →
A 2022 review of the whole field reaches a blunt conclusion: many drugs targeting myostatin and activin signalling failed in clinical trials, and more often than not because they could not improve muscle function despite preserving muscle mass. A 2023 sarcopenia review reaches the same line about the entire drug category.
Across 3 132 supplements analysed in a 2022 review, 875 contained substances that were not on the label, and a Dutch market survey found 38 percent of 66 high-risk sports supplements carrying undeclared stimulants and steroids. Independent testing exists, works, and settles far less than the word certificate implies.
Three US cohorts followed for up to 32 years, 5.54 million person-years and 14,806 cardiovascular events. One egg a day or more against less than one a month: hazard ratio 0.93, interval 0.82 to 1.05. The meta-analysis of 33 estimates and 1.7 million people lands at 0.98.
Two 2023 and 2024 meta-analyses of randomized trials agree on memory and disagree about the rest. Sixteen RCTs, 492 adults: memory SMD 0.31, attention time and processing speed improved, overall cognition and executive function did not. Ten RCTs in healthy people: overall memory SMD 0.29, but SMD 0.88 in adults aged 66 to 76 against 0.03 in adults aged 11 to 31.
More from the archive
A 2024 meta-analysis of nine randomised trials pooled 11 641 people on once-weekly semaglutide and 10 479 on placebo. Body weight fell 11.49 percent against placebo, waist 9.06 centimetres, and nausea was four times more likely. Serious adverse events were not significantly different, but stopping for adverse events was about twice as likely (odds ratio 2.22).
A 2024 meta-analysis of 11 trials and 349 people with hypertension found systolic blood pressure down 5.31 mmHg — and a certainty of evidence rated low. A 2017 analysis found the effect only in clinic measurements, with nothing in 24-hour ambulatory monitoring. That is the whole story in two papers.
Ninety randomised trials, 72,598 participants. The relationship is not one curve: triglycerides and non-HDL cholesterol fall roughly linearly with intake, while LDL and HDL describe J-shaped curves. A meta-analysis that averages those four together tells you nothing you can use.
The class marketed as gut-friendly is the one class the gastroenterology guidelines treat as a trigger. Polyols are a FODMAP category in their own right, low-FODMAP trials score 22.8 mm of global symptoms against 44.9 mm on the comparator diet, and a 2023 cohort links high erythritol to a 1.80 hazard ratio for major cardiovascular events. The dental benefit, meanwhile, is 13 per cent in low-quality evidence.
Just under 135 million person-years from 185 prospective studies and 58 trials: the highest fibre consumers had 15 to 30 per cent lower all-cause and cardiovascular mortality, with the largest risk reduction between 25 and 29 grams a day. An umbrella review of 45 meta-analyses graded the inverse association with fibre as convincing. The glycaemic index was graded low to very low.
Twenty-five thousand eight hundred and seventy-one people, 1,991 fractures, and no effect: VITAL found a hazard ratio of 0.98 for total fractures on 2,000 IU of vitamin D a day. A JAMA meta-analysis of 33 trials and 51,145 people reached the same null. What the trials did find is a real change in bone density and a real rise in kidney stones.
One randomized trial replacing saturated fat with safflower oil found higher mortality; the 2025 analysis of 65,179 adults with 20,571 deaths found the opposite for polyunsaturated fat. Both are real. The difference is what was substituted, what was compared with what, and what was actually bottled.
One 52-week randomized trial found 7.5 kilograms lost on non-sugar-sweetened drinks against 6.1 on water — and called the 1.4 kilogram difference not clinically significant. The WHO review of 56 studies found a BMI difference of minus 0.6. The two literatures do not agree, and the reason is design.
Ten trials, 475 people, both diets hypocaloric. Low-carb produced no better body composition, lowered triglycerides and raised LDL, and very-low-carbohydrate diets measurably reduced flow-mediated dilation, which is the opposite of what the marketing implies.
