Fermented foods: what the yogurt and kefir trials measured

One trial stands alone in testing fermented food as an actual diet. Thirty-six healthy adults ate that way for 17 weeks. The variety of bacteria in their guts climbed steadily and their inflammation markers fell.¹ The measure the study was built around did not budge.¹ Everything else here is kefir and yogurt.

The one trial that tested fermented food as a diet

That study split healthy adults into two groups of 18. One ate a high-fiber diet, the other a diet built around fermented foods, and both had their gut bacteria and immune systems profiled repeatedly over 17 weeks.¹

The two diets did different things. Fiber changed what the gut bacteria were equipped to digest, without changing how many kinds of them there were. Fermented food did the reverse. The variety of species rose steadily, and markers of inflammation went down.¹

The outcome the study had nominated in advance was a score summarizing immune signaling proteins. It did not change.¹ That is the right frame for everything below. Diversity and inflammation were two readouts among many in a 36-person study built for deep profiling, not for a verdict.

It is also the study behind most of what gets said about fermented foods, and there is no second one like it.

Kefir has the most trials, and most of them are weak

One review gathered every trial of fermented-milk kefir on any human health outcome and found 18 publications reporting 16 studies.² The quality check is the useful part. Safety was assessed in only 5 of the 18 publications, and 12 of the studies carried a high risk of bias.²

Its conclusions were correspondingly careful. Kefir may have promise against the mouth bacteria behind tooth decay, and as an add-on when clearing the stomach infection Helicobacter pylori. It may help with high cholesterol and high blood pressure, on very limited evidence.² The review's own summary is that better trials are needed before anyone makes a recommendation.²

What kefir moved, and what it left alone

Six trials in 314 people were combined to look at heart and metabolic markers.³ Two things moved. Fasting insulin, the amount circulating after a night without food, fell by 3.69 micro-IU per mL, with a plausible range from -6.30 to -1.07.³ A paired score of how hard the body is working to hold blood sugar down fell by 2.56.³ Cholesterol, triglycerides, HDL, LDL, fasting blood sugar, long-run blood sugar and body weight all stayed put.³

Kefir vs comparison: fasting insulin, pooled trials0 = no effect

Fasting insulin, micro-IU per mL-3.69 (-6.30 to -1.07)

A newer trial gave 65 volunteers 250 mL of lactose-free kefir a day for 6 weeks, which is a large glass.⁴ Gut symptoms eased. Serum cholesterol, creatinine and uric acid fell. Immune and inflammation markers did not improve, and the shifts in blood sugar, sleep and mental well-being pointed the hoped-for way without reaching significance.⁴ Its cholesterol result also contradicts the six-trial combination, which found none.³

A smaller study looked only at the gut. Twenty-one female soccer players drank 200 mL of kefir a day for 28 days, and the variety of species in their guts increased.⁵

The pattern is narrow. Diversity moves in two separate designs.^1,5 Insulin sensitivity moves when trials are put together.³ Almost everything else that was measured did not.

Yogurt, where the surveys and the trials disagree

Yogurt has the largest observational literature and the weakest trials. In one review, all 6 long-term follow-up studies and all 7 snapshot surveys found that yogurt eaters had lower or better body weight.⁷

The trials in that same review did not carry the finding. Six trials plus one controlled study were small and short. One found significant weight loss, but the yogurt groups were also eating more calcium. One found weight gain that was not significant. The remaining five found weight losses that were bigger in the yogurt groups and not significant either.⁷

The direction is consistent and the significance is not. That is what a small real effect looks like from the outside. It is also what confounding looks like, because people who eat yogurt tend to do other things too.

Yogurt against milk, the only fair comparison here

Comparing a fermented food with nothing tells you little. Comparing it with its unfermented parent tells you something. A network analysis of 19 trials in 1,427 people, each running at least 12 weeks, compared dairy products against one another, yogurt against milk included.⁶

Set against milk, yogurt improved waist circumference by 3.47 cm, triglycerides by 0.38 mmol/L and HDL cholesterol by 0.19 mmol/L, the first two on low-certainty evidence.⁶ More broadly, high dairy intake showed no harm to body measures, blood fats or blood pressure whatever its fat content. Both low-fat and full-fat dairy lowered the top blood pressure number, and both may have worsened blood sugar control at the same time.⁶

That is the closest thing here to a test of fermentation itself, though yogurt and milk differ in other ways too. It is still a comparison on markers, not on anything a person would feel.

What nobody measured

Three of the foods most associated with this subject, kimchi, sauerkraut and kombucha, have no trial of their own in this record. What was tested is mostly fermented dairy: kefir in most of the trials and yogurt in the observational work.^2,7 The single diet-level study used a mix of fermented foods, and its abstract does not say which.¹

Nothing here counted live organisms either. No study reports how many viable bacteria a serving actually delivered. So none of them can separate a fermented food's live cultures from everything else about it, or say whether a pasteurized version would have done the same job.

What remains unknown

The diversity finding needs a trial big enough to carry it. It rests on one arm of one 17-week study in 18 people, whose nominated outcome did not move.¹

The kefir literature needs quality rather than volume. Sixteen studies exist, most at high risk of bias, and safety was reported in a minority of the publications.² The single signal that survives combining trials is insulin sensitivity, from 6 trials in 314 people.³

And the practical question stays open. Whether the fermented version of a food beats the unfermented one at anything a person would notice. Only the yogurt-against-milk comparison even attempts it, on markers, over at least 12 weeks.⁶ No trial here compares kimchi with plain cabbage.