Podcast episode 122 (in German)

Microbiome tests – one hundred three percent difference, zero point eight from the genes

October 30, 2026

Eat the same meal, and blood fat levels diverge between people by 103 percent — the individual response is real. Genes contribute 0.8 percent to predicting it, and the labs that turn this into diet plans read different things from the same sample.

What it is about

A microbiome test combines a stool sample for the bacteria with a genetic analysis, and a model calculates scores for individual foods from them. The episode separates the research such offers cite from the product in the box. The foundational study in 1,002 adults shows cleanly that people respond very differently to the same meal and that the microbiome accounts for a measurable share of this. A randomized trial of the associated program, however, tested the overall package, not the personalization. Then come 2 independent evaluations of the providers, in which the same sample led to contradictory findings — between companies and even within one company. The bottom line: the foundation holds, the product does not.

Key points

In the foundational study with 1,002 adults, the response to the same test meal varied by 103 percent for blood fat, 68 percent for blood sugar and 59 percent for insulin.
The microbiome explained 7.1 percent of the variation in blood fat, the nutrients of the meal 3.6; for blood sugar it was the other way round, 6.0 versus 15.4 percent.
Genes contributed 9.5 percent to predicting blood sugar, 0.8 percent for blood fat and 0.2 percent for C-peptide.
A randomized trial in 347 people over 18 weeks found nothing for LDL and 0.13 millimoles per liter for triglycerides; the comparison group received only a leaflet and a video.
In the evaluation by the US standards institute, one provider reported 20 species from identical material, another more than 1,400; of 1,208 genera found, only 3 appeared in every sample.
Of the 7 providers tested, 3 reported the germ Clostridioides difficile as present and 4 as absent — from the same sample.
One provider delivered “healthy” twice and “not healthy” once from the same material; in the third report, 7 of 10 functional areas were below average.

Where the data stop

The variation between people is real and cleanly measured; that diet acts individually is a finding. The prediction model hits blood sugar decently at 0.77, and blood fat at 0.47 better than guessing but far from accurate. Several authors advise the company that sells the test or work for it. And the randomized trial tested the program, not the scores — the authors themselves write that exactly this comparison is still outstanding.

On the product side, reproducibility is missing. In Europe, 21 experts from 8 countries discussed the reports of 6 providers on a single stool sample: once severely altered, once average, once anti-inflammatory, once pro-inflammatory, and none named its reference group. In 2024 an international panel stated that evidence of clinical benefit is scarce; the German professional society for gastroenterology has advised since 2018 against deriving dietary recommendations from such tests. According to Stiftung Warentest, entry-level tests cost from 150 euros; the real damage does not appear on the invoice: reduced quality of life and, in the worst case, an overly one-sided diet.

Listen (in German)

Listen on Spotify Listen on Apple Podcasts Listen on Amazon Music

Published on October 30, 2026.

The topic in the database

Send in a stool sample, get a personalized diet plan back.Microbiome tests with diet plan

All episodes in the app (German)

Information only, not medical advice and not a usage or dosage recommendation. Prescription-only and unapproved substances belong in the hands of a physician. Last updated: 2026-10-09.