All ingredients

Vegetables

Early research

Jerusalem artichoke

In a double-blind trial of 45 people in three groups, bifidobacteria rose 1.2 log10 over 21 days. But the short-chain fatty acids that would carry that effect did not change, and flatulence rose significantly.

Oils & fats

Early research

Perilla oil

In a double-blind trial following 52 people for 12 months, a bone resorption marker fell. But what was measured was not bone mineral density — it was a heel ultrasound — and there were 52 participants.

Vegetables

Early research

Watercress

In a single-blind pilot trial, 26 people took dried watercress capsules for 28 days and LDL fell significantly. But there was no consistent effect on other lipids, and the paper states it is exploratory and needs confirmation in larger trials.

Grains & legumes

Early research

Amaranth

A 16-week trial in Kenya. The porridge with amaranth carried 5.6 times the iron of the control. Neither hemoglobin nor iron status moved. In the same trial, a micronutrient powder giving far less iron did reduce anemia.

Fruits

Early research

Goji berry

There is a trial in which eating goji berries raised macular pigment optical density. It had 27 participants, no masking, and calls itself a pilot. And the zeaxanthin said to explain the result is not listed in this food's composition table at all.

Vegetables

Early research

Oyster mushroom

Two double-blind trials have tested this mushroom. The 2022 one set postprandial glucose as its primary endpoint and it did not move. The 2026 one set LDL cholesterol and it did not move either. What moved in both cases were secondary outcomes.

Vegetables

Early research

Celery

In a 2022 trial, systolic pressure fell from 141 to 130 mmHg, measured by 24-hour ambulatory monitoring — a good method. But what participants took was not celery stalks. It was 1.34 g a day of extract from the seed.

Grains & legumes

Early research

Cheonggukjang

In the 2025 menopause trial, symptom scores fell in all three groups. That sounds like good news until you see that all three groups ate cheonggukjang. Without a placebo arm, there is no way to separate the improvement from time and expectation.

Fruits

Early research

Mango

A 2026 trial in JAHA improved endothelial function. But the intervention was one avocado and one cup of mango together, and the comparator was low-fat, low-fibre food. How much of it belongs to mango cannot be read from this design.

Seafood

Early research

Cod

In a 2007 trial, cod protein improved insulin sensitivity compared with other animal proteins, measured by the gold-standard method. But it held 19 people, and a 2020 follow-up using a cod protein supplement moved nothing at all — not glucose, insulin, lipids or body composition.

Fruits

Early research

Figs

Two randomised trials have actually fed people figs, and both improved bowel measures. But in the four-month trial the severity of abdominal pain and CRP did not move, and in the pregnancy trial adding walnut was no better than fig alone.

Nuts & seeds

Early research

Macadamia nuts

A 2008 controlled-feeding trial lowered total and LDL cholesterol significantly. A 2023 free-living trial moved nothing significantly. The decisive difference between them was not the macadamias but what they displaced.

Nuts & seeds

Early research

Pine nuts

The claim that pine nuts suppress appetite rests on one 2008 trial. But what that trial gave people was not pine nuts — it was 3 g of fatty acids extracted from them, in a capsule — and it never measured how much anyone actually ate.

Fruits

Early research

Dates

Several findings suggest eating dates in late pregnancy shortens labour and reduces the need for induction. But the same paper states that "the overall evidence quality was unacceptable, and risk of bias was high in most studies". And 66.5 g of every 100 g is sugar.

Grains & legumes

Early research

Doenjang

A 12-week double-blind randomised trial found visceral fat fell — only in people carrying the T allele of PPAR-γ2. And in the same trial, insulin and HOMA-IR rose. Meanwhile, the biggest number on doenjang's label is sodium.

Nuts & seeds

Early research

Hazelnut

A network meta-analysis of 34 randomised trials in 1,677 people ranked pistachio first for triglycerides, LDL and total cholesterol, with hazelnut losing to pistachio on triglycerides and total cholesterol and to walnut on triglycerides. The authors noted the low quality of the evidence. Hazelnut's real strength is not lipids but 15 mg of vitamin E and 6.18 mg of manganese.

Vegetables

Early research

Lettuce

A 2025 randomised trial in Korean adults found total sleep time up by about 35 minutes. But what was given was two capsules a day of an extract from one specific cultivar, not lettuce leaves. The 15 kcal and 370 µg of vitamin A per 100 g are real.

Grains & legumes

Early research

Soy milk

In a 2026 network meta-analysis, replacing 500 mL of cow milk with soy drink lowered systolic pressure by 8.23 mmHg — about what a single blood-pressure tablet does. But the same paper's overall verdict is that "few differences were observed", and that number is graded low certainty.

Herbs & spices

Early research

Vinegar

A 2025 meta-analysis of seven trials reported fasting glucose down 21.9 mg/dL and HbA1c down 1.53%. That is what metformin achieves. Coming from seven small trials, it is implausibly large. The direction is consistent; the size is not something to accept as printed.

Fruits

Early research

Elderberry

A 2021 systematic review that GRADE-assessed five randomised trials concluded it may not reduce the risk of catching a cold. It may shorten duration and severity, but the evidence is uncertain. And raw, it will make you ill.

Fruits

Early research

Mulberry

A meta-analysis of 13 randomised trials found post-meal glucose down at 30, 60 and 90 minutes. But fasting glucose, HbA1c, HOMA-IR and all four lipid measures showed no difference. And what those trials mostly used was the leaf, not the fruit.

Tea & drinks

Early research

Peppermint

Across 10 randomised trials in 1,030 people, peppermint oil beat placebo. But adverse events were more frequent (RR 1.57), the authors stated the quality of evidence was very low, and what was tested was an enteric-coated capsule, not tea.

Fruits

Early research

Persimmon

A trial feeding 9–15 g a day of persimmon tannin fibre for 12 weeks lowered total cholesterol. But that was not eating persimmons — it was a fibre fraction extracted from unripe fruit, baked into bars. And there is a separate caution about astringent persimmon on an empty stomach.