Same capsule, opposite result: what decides if a probiotic works

Two people buy the same probiotic. One drops a few pounds and swears by it; the other feels nothing and files the whole category under snake oil. Both are telling the truth, and a trial published in Gut Microbes in 2026 ¹ shows how. It is the plainest published answer I have seen to "why did it work for my friend and not for me".

The trial, by name

Lim and colleagues at the LOTTE R&D Center in Seoul recruited 120 overweight adults ¹, defined as a body mass index between 25 and 30 ¹, at three Korean hospitals ¹. They were randomised, double-blind, to 12 weeks ¹ of one daily sachet containing either placebo or a heat-killed preparation of Lactiplantibacillus plantarum LRCC5282 ¹, a species formerly filed under Lactobacillus. Heat-killed is the important word. The cells were inactivated at 90 °C for 30 minutes ¹, so this is a "paraprobiotic": nothing here depends on live bacteria settling into your gut. The primary endpoint, fixed before the trial began, was change in total body fat on a DXA scan ¹.

Across the whole group the answer was nothing. Body fat did not differ between arms ¹. Weight fell 1.33 kg on the paraprobiotic and 0.94 kg on placebo ¹, a gap well inside chance. No other clinical measure separated the groups ¹.

Then they split the room

The authors then went back to the stool samples taken before the first sachet and ranked every participant by how many kinds of bacteria they carried, using a composite of five α-diversity indices — observed OTUs, Shannon, Simpson, Chao1 and ACE — percentile-ranked within each treatment arm and cut at that arm's median ¹. That produced four small groups: 24 low-diversity people on the paraprobiotic, 26 low-diversity on placebo, and 25 and 26 in the two high-diversity halves ¹.

In the low-diversity half the picture changed. Weight fell 1.98 kg on the paraprobiotic against 0.95 kg on placebo ¹. BMI fell 0.74 points against 0.31 ¹. Leptin, the hormone fat tissue releases in proportion to how much fat there is, fell 2.07 ng/mL against 0.24 ¹. HDL cholesterol rose 3.21 mg/dL while placebo did not move ¹. In stool, acetate and butyrate, the short-chain fatty acids bacteria make when they ferment fibre, were higher at week 12 than in either placebo group ¹. And the community itself had moved: more Christensenellaceae, more Faecalibacterium, more Alistipes ¹, and a lower ratio of Firmicutes to Bacteroidota ¹. Within this group, the people whose Akkermansia and Eubacterium rose most were the people whose weight, fat mass and leptin fell most ¹.

In the high-diversity half, on the same sachet, only scattered differences appeared, with no coherent pattern ¹.

Why a crowded gut ignores a newcomer

The mechanism is ecology before it is biochemistry. A diverse gut community has every niche occupied: the fibre is spoken for, the mucus layer is spoken for, the bile acids are already being processed. A signal from outside, even dead bacterial cell walls, finds few open seats. A thinned-out community has room to move.

This is not a new idea, only a new test of it. In 2013, Le Chatelier and colleagues sequenced the gut bacteria of 292 Danish adults ² and found that the roughly one in four with low bacterial gene richness ² carried more body fat, more insulin resistance and more inflammation than the rest, and that the obese among them gained more weight over time ². In 2018, Zmora and colleagues gave healthy volunteers the same 11-strain probiotic ³ and found that whether it took hold in the gut lining was person-specific and predictable from the microbiome each person started with ³. The Seoul trial adds the missing piece: a placebo-controlled outcome split by that starting point.

What was not shown

Now the caveats, said out loud, because the paper is honest about them.

The primary endpoint failed. DXA body fat did not differ in the whole group or in the low-diversity subgroup ¹. Weight, BMI and leptin are secondary measures.

The diversity split was a stratified secondary analysis, not the comparison the trial was designed around ¹. Four groups of about 25 people ¹, with a weight p-value of 0.048 ¹, is a hypothesis, not a proof. The authors call for prospective trials and propose baseline diversity as a stratification variable for them ¹.

No one tracked the strain: the sequencing turned up no species-level L. plantarum signal at all, which the authors say neither confirms nor excludes biological activity ¹. Diet was assessed only qualitatively, so intake could not be quantified ¹, and the trial ends at week 12 with no follow-up reported after it ¹, so whether anything lasts is unknown.

And the money: the sponsor was LOTTE Chilsung Beverage, the strain belongs to the LOTTE culture collection, and all six authors are LOTTE employees ¹. That does not make the data wrong, but nobody without a stake has repeated it yet.

What to take from it

Do not judge a probiotic by your friend. A null average, no difference once everyone is counted together, can hide a minority who responded, and a glowing anecdote can hide a majority who felt nothing. The Seoul trial suggests the dividing line may be how diverse your gut was before you started, how many kinds of bacteria you carried, and points the field toward sorting people by that line before counting rather than averaging over it. What it does not tell you is which half you are in, or that any product on a shelf will do for you what a dead L. plantarum did for 24 overweight adults in Korea ¹. The question is now well posed. The answer is still being collected.