The obvious answer is: probably not. But studies of the gut microbiota have an inherent high risk of arriving at such erroneous conclusions. The more outcomes we measure in complex biological systems, the more random associations we find. Journalists and researchers should therefore be extra careful about how we communicate new discoveries.
In 2012, a quite famous article showed that countries with a high consumption of chocolate had more Nobel Laureates per capita than countries with populations less fond of chocolate (1). The article was in fact meant as a joke, but many readers overlooked this, cited the article and took it quite seriously, since various explanations can easily be offered on how and why specific substances in chocolate may potentially strengthen brain function.
The lesson to be drawn from this is that even though two things are correlated, one does not necessarily cause the other. This is easy to understand for relatively simple relationships. For example, most of us would intuitively think that our hair colour does not determine our house number (or vice versa) – even if we happen to live in a street where all the redheads live in houses with prime numbers.
Correlation is not necessarily causation
In more complex biological contexts, it is often possible to propose a mechanism that explains what we have observed – such as specific substances in chocolate strengthening the brain. Even when the observation of correlation is random.
The great challenge arises when researchers measure a myriad of variables at the same time without a clear prior hypothesis. A research group in the Netherlands humorously illustrated these pitfalls in an article showing that certain gut bacteria correlate with being together with in-laws at Christmas dinner (2). Similar to the case with chocolate and Nobel Laureates, mechanisms can be found to link the in-laws, stress levels and gut bacteria – but that does not prove causation.
The more outcomes we measure, the more completely random associations we find. That is precisely the challenge for gut microbiota. Humans typically host 150–200 bacterial species in their gut. This means that mapping the microbiota of a group of people and then asking them the next day what the weather is like, would probably show that a few species out of the 200 were more plentiful in people who responded that it rained the next day. If we are not careful, we might conclude that these species attract local rain showers.
Could just as well be the opposite
But how can we then establish causal links between the gut microbiota and health and disease?
One solution to this dilemma is to replicate the study in a new group of people and check whether the correlation is present again. This will hopefully be sufficient to refute the examples with hair colour, house numbers and rainclouds.
Some circumstances may however mean that we find the same correlation over and over again, even though it is not necessarily causal. For example, many intestinal diseases change the growth conditions of gut bacteria.
We can therefore quite consistently measure that people who have these diseases differ from healthy people with respect to their composition of gut bacteria. But this does not prove that the bacterial composition causes the disease. The direction of causation may just as well be the opposite. Similarly, the abundance of ferns is not what causes heavier rain in the rainforest.
Diet affects both us and our bacteria
Another important challenge for gut microbiota research is that our diet affects both our health and our gut bacteria. The food that we eat directly affects our health, and this is not necessarily mediated by its effect on the gut bacteria.
We have considerable knowledge about which foods are healthy and unhealthy. We also know that some of the things we eat reach the gut bacteria and thus determine their growth conditions and promote certain bacterial species.
Some species are often present in higher amounts among healthy and fit people than among people with diet-related diseases. But this does not necessarily mean that these species promote health. This may just mean that both the body and the gut bacteria respond to the healthy diet independently.
Similar to the chocolate example, many substances that bacteria produce are known to be healthy. So we can often speculate that there is a causal link from bacteria to health effects, but it is hard to obtain solid proof.
The come-back microbes
What should we then do to determine with certainty whether gut bacteria have a causal effect on our health – for example, on our weight, cholesterol levels, blood pressure or inflammation level? An intervention study in which a group of people eat differently and are measured before and after their diet change often reveals changes in both their health and their microbiota – but it cannot prove that the gut bacteria directly affect health, even though the two are correlated.
If we repeat the study, the correlation will be even more certain – but still does not provide proof of causality. We can investigate relevant mechanisms in the body and suggest underlying explanations for how the bacteria affect biomarkers related to health, and this can certainly support our narrative – but we still have no solid evidence.
Is it then impossible to prove a causal effect of the microbiome?
The ultimate study would be to directly change the composition of the microbiota – without changing anything else – and then see whether this changes people’s health status. The closest we can come in practice to this is a so-called faecal microbiota transplantation, which has proven very effective in fighting intestinal infections caused by Clostridioides difficile.
However, the effect of a faecal microbiota transplantation on other diseases is not well established – and the effect on the biomarkers mentioned previously (weight, cholesterol levels, blood pressure and inflammation) are typically transient. Part of the challenge with faecal microbiota transplantation is that it is really difficult to completely replace our gut bacteria with different ones. No matter how many intestinal lavages and antibiotic treatments we get, a residue of our original microbiota will persist and grow back when conditions allow.
Germ-free mice therefore present a very exciting opportunity for researchers, since they are bred entirely without gut bacteria, so that you can construct their microbiota from scratch. Unfortunately, transferring conclusions from such experiments to humans poses many challenges (3).
Often we just forget to explore further
All the described reservations and risks for misinterpretation take some time to explain and understand. Nevertheless, such topics as diet, health and the microbiota are receiving huge attention from the public. Researchers and journalists therefore need to carefully think about how to disseminate information. In the worst case, misleading messages can negatively affect individual as well as public health.
Journalists are trained to formulate catchy headlines with clear messages, and making a clear conclusion is also more fun for researchers. But there are often many caveats to consider. The huge availability of information on the web and on social media is a gift but also poses a challenge.
Diet and health are hot topics, and almost regardless of which positive or negative health effects of eating something that you have heard about, you can search Google and find nice scientific articles that support your view. But if you search for articles that question your assumption, you typically find just as many of these. Often we just forget to explore further, once we have confirmed our own perception.
Like diet, the bacterial composition of the gut has become a hot topic, but the misinformation in this field has gradually become so extensive that researchers have now expressed a need to focus on the importance of how we communicate about this (4). So if you read an exciting article in the newspaper tomorrow that gut bacteria cause rain showers, perhaps stop and reflect a bit longer. And then remember your umbrella if the forecast says rain.