The many pitfalls of personalized advice
In the era of AI, several companies are emerging that offer personalized dietary advice and/or food supplement suggestions based on fecal analysis results and sometimes based on a questionnaire. While we are enthusiastic about the concept of personalization, unfortunately most of these companies seem to fall into some important pitfalls.
Steering towards a certain relative abundance does not work
First, companies often simply use the relative abundance of bacterial species or genera and compare these with and attempt to steer towards the relative abundance observed in healthy cohorts. This ignores the fact that no one has been able to define healthy ranges of relative abundance per species because different microbiome compositions can be healthy (1). Instead, the gut microbiome behaves like an ecology in which the composition depends on the circumstances, such as the host’s genetic makeup, lifestyle, diet, etc. (2). Therefore, it is better to look at functional aspects of the microbiome (e.g. overall butyrate production, inhibition of pathogens, or gut barrier function). Contrary to simply using relative abundance, these functional aspects do provide targets for treatment, which why this is what Microbiome Center and Biovis use.
Most dietary recommendations are based on incorrect assumptions and weak evidence
Second, the strength of evidence behind many kinds of dietary advice is extremely weak. Much is based on observational studies, which are prone to large number of biases and confounders. Here too, the rather reductionistic approach that is often taken plays a role. E.g. in observational studies it is typically assumed that it possible to study a dietary component in solitude, but this ignores the highly interactive characteristic of diets. Moreover, this also ignores that it is known that one person may have e.g. peak blood sugar response to a cookie and not to a banana, while a second person has the opposite response (3). In other words, it is impossible to conclude that e.g. a banana is unhealthy because it causes peak blood sugar, because each individual responds differently. Thus, any dietary advice that is based on simplistically linking a certain food component to certain effects is not in line with the reality of our physiology. Therefore, we argue that a better approach is to first acknowledge that with the extreme complex and person-specific responses to diets, the only generalizable knowledge regarding diets is logic: eat foods that are close to what our physiology has evolved on. This means real foods that are minimally processed, which may include vegetables, fish, fruits, meat, nuts, seeds, water, herbs, etc. Second, it is useful for anyone to learn to recognize their personal response to various foods and food patterns, it doesn’t become more personalized than that!
Avoid supplements suggestions based on assumptions, only at species level, or without references
Third, often these companies advertise to provide personalized suggestions for probiotic supplements. Unfortunately, these suggestions in many cases are not based on direct clinical evidence for the products but instead on various kinds of assumptions about stimulating taxa that are low in the fecal analysis. Here, the first pitfall applies: health cannot be simply linked to relative abundance. In addition, the suggestions for probiotics are based on bacterial species in these products, which ignores that effects are strain-specific (different strains of the same bacterial species can have different effects). Suggesting a product based on species is like a pet shop that suggests to buy “a dog”, ignoring the differences between e.g. a bulldog or a labrador (both the same species). Moreover, this reasoning is rather reductionistic and that doesn’t work in a complex ecology such as the gut microbiome. That is why it is better to select probiotics based on direct clinical evidence with the exact strain. This circumvents assumptions about pathways and looks directly at clinical outcomes. Finally, many providers do not reveal what evidence their suggestions are based on, which makes it difficult to assess whether or not the advice is realistic.
Conclusion: use evidence-based personalized advice
We believe that any personalized microbiome treatment advice must be based on (1) direct mechanistic and clinical evidence, (2) studies using the exact strain, (3) dosage as used in studies, and (4) must be completely transparent about the underlying evidence.
Do you want to read more insights?
You can follow us on LinkedIn: https://www.linkedin.com/company/microbiome-center/
Or read previous insights flashes at: https://dev.studioboszkers.nl/microbiome-center/en/medici/kennisflits/
References:
- https://doi.org/10.1053/j.gastro.2020.09.057
- https://doi.org/10.1016/j.shpsa.2021.11.005
- https://doi.org/10.1016/j.cell.2015.11.001