A metabolism supplement guide: six ingredient families, and what each has actually shown

Caffeine, in coffee, in tea, in guarana and as caffeine anhydrous, is the backbone of this category and the only ingredient in it with a dose-response signal in pooled human trials. a 2019 dose-response meta-analysis pooled 13 randomised trials in 606 people and reported that doubling intake went with roughly 22 per cent more weight change, 17 per cent more BMI change and 28 per cent more fat-mass change.

The mechanism is set out in a 2025 mechanism review: adenosine-receptor blockade, raised catecholamine signalling, and the metabolic consequences that follow. a brown-fat trial measured brown adipose tissue activation and cold-induced thermogenic capacity in people taking caffeine and tea catechins.

The ceiling is the number to hold on to. the 2017 systematic review of caffeine safety concluded that up to 400 mg a day in healthy adults is not associated with adverse cardiovascular, behavioural, reproductive or bone effects, with 300 mg for pregnancy, and the Health Canada review reached the same place. a 2023 blood-pressure meta-analysis found a measurable rise in blood pressure across the studied range, and a 2025 meta-analysis of caffeine and sleep found the sleep effect depends on dose and timing.

A label that gives no caffeine figure is therefore a label that makes a reader's own arithmetic impossible, and that is the first thing to look for in this family.

Green coffee bean extract is the third family, and its story is a good lesson in how a category's reputation gets made.

a 2023 chlorogenic acid meta-analysis is the current best summary: three randomised trials, 103 people, chlorogenic acid at 500 mg a day or more, a pooled difference of 1.30 kg against placebo, and no heterogeneity between the studies. an eight-week metabolic-syndrome trial gave 400 mg of a decaffeinated extract twice daily for eight weeks, a 12-week chlorogenic acid study ran 500 mg of a chlorogenic acid complex for twelve weeks, and an energy-restriction trial in obese women combined the extract with energy restriction.

What made the family famous was the retracted 2012 crossover study, a sixteen-person crossover study reporting eight kilograms of loss. It was retracted in 2014. a 2011 systematic review had already described the evidence base as methodologically poor.

The mechanism claims in this family are frequently overstated. an acute glucose-response study measured what chlorogenic acid does to the glucose and gut-hormone response to a single drink, and a chlorogenic acid coffee trial looked at absorption and body mass. Neither supports the sentence about blocking carbohydrate absorption that appears on a great many sales pages.

Raspberry ketone is the most common member and it is the clearest example of a compound whose reputation rests on structure rather than on evidence. It resembles capsaicin and synephrine, both of which have metabolic effects, and that resemblance is the whole argument.

the 2005 rodent study fed mice a high-fat diet containing 0.5, 1 or 2 per cent raspberry ketone for ten weeks and reported prevention of diet-induced weight gain and increased noradrenaline-driven lipolysis in rat fat cells. a 2024 mouse study added a hypothalamic-inflammation finding two decades later.

There is no human trial of raspberry ketone alone. The nearest thing is a multi-ingredient supplement trial, a study of a multi-ingredient product with diet and exercise, from which nothing can be attributed to any single component. A diet that is one to two per cent of a compound by weight is also an exposure no capsule approaches.

The lesson generalises. Any ingredient in this category whose evidence is a rodent feeding study at a dietary percentage should be read as untested in people, whatever the mechanism diagram suggests.

Six checks, in the order a shelf makes them easy to apply.

The companion piece turns those six into a four-minute method you can run on two labels in a shop, and the ingredients page applies the same method to the bottle this desk sells, including where it fails.