Probiotics for weight loss: what the trials actually show

For weight and body composition, not one sex-stratified result in men is positive, anywhere. CGMCC1.3724: P=0.53 in men versus P=0.02 in women. HA4597: post-hoc male subset p=0.12. SBT2055, BNR17 and the Akkermansia trials published no sex breakdown at all. Class-level pooled body weight runs −0.26 to −0.60 kg with two of five estimates crossing zero, and pooled fat mass is not significant where reported. Do not call that a ceiling: a 47-RCT network meta-analysis reaches −2.83 kg (95% CI −4.33 to −1.32) for specific multi-strain combinations at low-to-moderate certainty, with no strain designators reported at the "Mix" level, so it names nothing you can buy. The one prespecified body-weight primary endpoint met anywhere in that literature belongs to pasteurised A. muciniphila MucT (1.2 ± 0.7 kg regain versus 3.2 ± 0.4 kg, P=0.012, n=90; PMID 42120725 (opens in a new tab)) — and three caveats are mandatory: it is weight-loss maintenance, not weight loss; it is one positive trial against two that missed their endpoints; and no sex-stratified result exists for any of the three. In the trial that tested both preparations, the live arm (n=9) showed no significant metabolic benefit — which matters, because the product on sale is live, under the house code WB-STR-0001, which returns zero PubMed hits and has no public accession.

Weight, body composition and cholesterol, in the two paragraphs this site published on them, with the three rows they hold in the 22-row table reproduced below.

Weight, body composition and cholesterol — the short versions

For cholesterol, the one strain with two met LDL-C primary endpoints cannot be bought. L. reuteri NCIMB 30242 met a registered LDL-C primary in two 2012 trials (n=114 and n=127 completers) and has never been independently replicated in 14 years; every human efficacy study traces to the same McGill/Micropharma group. Meanwhile two completed LDL-C-primary trials of that strain have never reported resultsNCT01630668 (opens in a new tab) (150 enrolled, completed June 2014, sponsor Micropharma) and NCT02734706 (opens in a new tab) (111 enrolled, completed February 2019, sponsor UAS Labs). The unpublished base (261 enrolled) is roughly as large as the published base (281 enrolled, 241 completers). NCIMB 30242 is in none of the 27 teardowns and in none of the 166 SKUs whose labels we read.

Where these outcomes sit in the 22-row table

The grade is assigned on the hub’s 22-row table and argued above. The row s are reproduced here unchanged, so the grade a reader quotes from this page is the same grade the table carries. What a grade measures is set out under how the grades work: it is a statement about the strength of the evidence for a named strain, never a statement that a product does anything.

The rows of the 22-row outcome table this page is the long version of. Default order: as numbered in the full table. Sortable columns are marked; sorting needs JavaScript, the order above does not.
# Outcome Best grade at strain level The strain, and the trial behind it Buyable at the dose that was studied?
4 Weight-regain prevention after dieting Grade C for Weight-regain prevention after dieting — one positive trial with a material design defect Pasteurised Akkermansia muciniphila MucT (PMID 42120725 (opens in a new tab); NCT05417360 (opens in a new tab)) No — the product on sale is live, under a house code with zero published literature, and the live arm was null
5 Visceral fat Grade C for Visceral fat — one positive trial with a material design defect Lactobacillus gasseri BNR17 at 10¹⁰ CFU/day, −21.6 cm² vs placebo, P=0.012 (PMID 29688793 (opens in a new tab)) No — absent from the teardown set
6 LDL cholesterol Grade C for LDL cholesterol — one positive trial with a material design defect Limosilactobacillus reuteri NCIMB 30242 (PMID 22067612 (opens in a new tab); PMID 22990854 (opens in a new tab)) No — absent from the market universe entirely

The strains named above, and the labels in this audit that print them

Every row below is a line on a printed supplement panel, read off that product’s own teardown. A panel printing a designator is a fact about the panel. It is not evidence that the product was tested, that it delivers the amount that was studied, or that the strain does anything — several of the paragraphs above say in terms that it does not, and the dose is unknowable wherever the panel gives no per-organism quantity.

L. rhamnosus CGMCC1.3724

Not printed on any panel this audit has transcribed.

L. gasseri BNR17

Not printed on any panel this audit has transcribed.

L. gasseri SBT2055

Not printed on any panel this audit has transcribed.

Pasteurised A. muciniphila MucT, and the live house code WB-STR-0001
  • Pendulum Akkermansia Panel prints “Akkermansia muciniphila WB-STR-0001” at 100 million AFU.
  • Pendulum Glucose Control Panel prints “Akkermansia muciniphila WB-STR-0001” with no per-organism quantity — the amount is inside a blend total.
  • Pendulum Metabolic Daily Panel prints “Akkermansia muciniphila” with no per-organism quantity — the amount is inside a blend total · this record carries a verification flag.
L. reuteri NCIMB 30242

Not printed on any panel this audit has transcribed.

Scope, stated because an absence here is a published claim: this searches the organism slots transcribed from 156 of the 166 labels in the audit. A panel that is not fully transcribed can hide a line, so “not printed on any panel this audit has transcribed” is exactly that, and never “not sold anywhere”. Each product links to its own label audit, where the panel, the designator resolution and the arithmetic are set out in full.

Two paragraphs of § “The short versions”, which was published on the home page under #short-versions and is split here by paragraph, never by sentence. Both paragraphs are complete and unaltered, in their published order.