Research by Food: Evidence Snapshot Table

The question

How much human research has been published on each food, and what does that tell you? This page reads our PubMed snapshot, taken on 2026-10-22, in plain language. The snapshot counts papers. It does not say whether a food works, is safe, or is worth eating. It is information only.

The snapshot

The table lists 35 foods and terms. For each, it counts PubMed records whose title or abstract contains the search words, and shows how many of them are tagged as randomized controlled trials and how many records there are in all. [1]

Counted 2026-10-22 using NCBI E-utilities. Counts include studies of extracts, blends and non-health topics, and a trial can find no effect. A higher count means more human research has been indexed, not that something works.

How to read the columns

• Randomized-trial records. PubMed's tag for a randomized controlled trial describes a study where treatments "are selected by a random process". [2] The tag does not say whether the trial was large, blinded or well run.
• All records. Every PubMed record matching the search words, of any type. [1]
• The search words matter. Each row is a word search in titles and abstracts: the kefir row searches only that word, while the soy row also searches soybean and tofu, the tomato row also includes lycopene, and the citrus row also includes orange juice and grapefruit. [1]

What stands out in the numbers

The figures below are read straight from the snapshot. They show how much was published, not what was found.

Three patterns are worth noticing:

1. Fermented foods are thinly studied compared with probiotic capsules. Probiotics have 3,369 randomized-trial records, against 36 for kefir and 9 for kimchi. Many probiotic trials test capsules, not food. [1] See Fermented foods vs probiotic supplements.
2. Trials make up very different shares of the record for different foods. Beets have 361 randomized-trial records among 1,675 in all (about 22%); leafy greens have 83 among 11,725 (under 1%). The snapshot does not say why. [1]
3. "Superfood" has almost no trials under its own name. The snapshot finds 6 randomized-trial records. A systematic review of human trials of foods described as superfoods found 113 trials across 17 foods, but only limited evidence for effects on metabolic syndrome measures. [3]

Where counts mislead

• Big counts can include work that is not about eating the food. Soy (2,185 randomized-trial records) and legumes (1,451) are broad searches. [1]
• Small counts do not mean a food is ineffective. They mean little has been published under that search word.
• A count has no direction. A trial can find benefit, no difference or harm; the snapshot counts them all alike.
• Forms differ. Records on juice, extract or supplement can sit inside a food's count. A result for one form does not transfer to another.
• Proper evidence not found: we looked for a source in PubMed that treats publication counts as a measure of how well a food works, and did not find one in what we read.

What to do with the table

Use it to see where research is thin, then read the food's own page for what the trials found. For how to judge a study, see Functional food research and evidence and Observational studies vs trials.