An altered gut microbiota (inaccurately stated as “dysbiosis”) has been linked to almost every human disease or condition. However, one of microbiome research’s challenges is that of disentangling whether changes in gut microbiota composition and/or function are causal or consequential to a specific disease or condition, or, alternatively, whether microbiota changes and disease are both driven by a third factor.

In a systematic review in Cell, Jens Walter and colleagues address the challenges and limitations of establishing causality in microbiome research based on human microbiota-associated or humanized gnotobiotic rodents.

A different gut microbiota as compared to a control has been published for almost every disease or condition, ranging from gut-related conditions to obesity, cancer and neuropsychiatric diseases. However, until now the microbiome’s causal involvement only has robust evidence for Helicobacter pylori-associated peptic ulcer and gastric cancer and Clostridioides difficile infection-associated diarrhea.

The model used most frequently in microbiome research for making causal inferences is the transplantation of human fecal microbiota into germ-free mice (also known as human microbiota-associated or humanized gnotobiotic mice). Then, scientists examine disease phenotypes and identify mechanisms in germ-free mice colonized with the fecal microbiota of patients compared with mice colonized with the microbiota of healthy controls.

Those models have limitations, however, which affect data interpretation. One important limitation is that gut microbiota changes occurring in the donor host (e.g., in the context of a specific disease vs. healthy controls) will in most cases experience an ecological shift after engraftment to mice, and that does not necessarily represent the gut microbial community associated with the donor’s pathology. As a result, due to the genetic, behavioral, physiological and anatomical differences between host species, the altered microbiome patterns observed in donor human subjects may be difficult to replicate in human microbiota-associated mice.

By systematically reviewing 38 studies that utilized human microbiota-associated mice for assessing the role of the gut microbiome in the etiology of a wealth of disease states, a successful phenotype transfer was reported in 36 out of 38 studies. Nevertheless, taking into account the complex etiology of the diseases under study and the lack of specificity of human gut microbiome changes and disease states, the authors concluded that causal claims for the microbiome in the disease states under study are improbable.


The major pitfalls detected in studies using human microbiota-associated mice models include:

  • Not testing for an altered gut microbiome either in the donor (diseased vs. healthy humans) or in the recipient mice.
  • Not replicating the donor’s gut microbiome changes in recipient rodents.
  • Not identifying underlying mechanisms (i.e., the causal component of the microbiome) linking the altered microbiome with disease.
  • A lack of standardization of experimental designs coupled with inappropriate statistical analyses (e.g., the small number of donors used, which fails to capture the large inter-individual variability of the human gut microbiome). In this regard, the authors pointed out that interpretations of results should be limited to pathophysiological and behavioral mechanisms rather than clinical symptoms in humans, especially in studies reporting human phenotypes such as autism-like behavior, which do not naturally occur in mice.

Beyond causality versus association, other challenges in translating microbiome science addressed by Colin Hill during the last Gut Microbiota for Health World Summit 2020 in Madrid include the lack of precise language, numeracy and interpretation of complicated microbiome figures and analyses.


Finally, the authors make some suggestions for establishing causal relationships between an altered gut microbiome and human disease, state or condition with scientific rigor and without hype. They suggest:

  1. Determining the microbiome alterations associated with the pathology by comparing donor samples from individuals with a disease with those who are healthy.
  2. Using an appropriate number of donors to take into account the high inter-individual variation in the gut microbiome.
  3. Performing statistical analyses by using donor numbers instead of the number of recipient mice.
  4. Not pooling donor samples before inoculating rodents (the common practice of pooling donor samples increases the chance of false positive findings).
  5. Testing if microbiome engraftment occurred and whether an altered gut microbiome in terms of composition and/or functions was transferred to recipient animals.
  6. Discussing honestly the limitations of animal models and avoiding overstatements on causal claims. Microbiome science requires precision of language and numerical accuracy coupled with a healthy dose of skepticism to keep from getting carried away by hype.
  7. Exploring mechanisms and causal components of the microbiome for the progression of the microbiome field.



Walter J, Armet AM, Finlay BB, Shanahan F. Establishing or exaggerating causality for the gut microbiome: lessons learned from human microbiota-associated rodents. Cell. 2020; 180(2):221-32. doi: 10.1016/j.cell.2019.12.025.