A Guide to Diet-Microbiome Study Design

Abstract

Intense recent interest in understanding how the human gut microbiome influences health has kindled a concomitant interest in linking dietary choices to microbiome variation. Diet is known to be a driver of microbiome variation, and yet the precise mechanisms by which certain dietary components modulate the microbiome, and by which the microbiome produces byproducts and secondary metabolites from dietary components, are not well-understood. Interestingly, despite the influence of diet on the gut microbiome, the majority of microbiome studies published to date contain little or no analysis of dietary intake. Although an increasing number of microbiome studies are now collecting some form of dietary data or even performing diet interventions, there are no clear standards in the microbiome field for how to collect diet data or how to design a diet-microbiome study. In this article, we review the current practices in diet-microbiome analysis and study design and make several recommendations for best practices to provoke broader discussion in the field. We recommend that microbiome studies include multiple consecutive microbiome samples per study timepoint or phase and multiple days of dietary history prior to each microbiome sample whenever feasible. We find evidence that direct effects of diet on the microbiome are likely to be observable within days, while the length of an intervention required for observing microbiome-mediated effects on the host phenotype or host biomarkers, depending on the outcome, may be much longer, on the order of weeks or months. Finally, recent studies demonstrating that diet-microbiome interactions are personalized suggest that diet-microbiome studies should either include longitudinal sampling within individuals to identify personalized responses, or should include an adequate number of participants spanning a range of microbiome types to identify generalized responses.

Keywords: diet; dietary intake; methodology; microbiome; personalized nutrition; study design.

Figure 1

Diet and the gut microbiome interact to influence host health. Dietary intake and the gastrointestinal microbiome interact to affect host physiology and disease. These interactions are complex and likely depend on interplay between the host's genetics and immune system and the composition and function of the gut microbiome. Consumption of foods begins the digestive process, leading to the production of host-derived metabolites primarily absorbed through the stomach and small intestine. Host genetics impacts this metabolism and can also be influenced by diet with some diet-gene interactions altering host gene expression. Food and dietary products that reach the large intestine can shape the microbiome directly. These products are also acted upon by microbes to produce microbially derived metabolites. Host-derived metabolites that have previously entered host-circulation may also pass into the large intestine via biliary enterohepatic recirculation or by transport across the gut epithelium resulting in additional opportunities for the host and microbes to jointly act upon dietary derived metabolites.

Figure 2

Considerations for participant enrollment and data collection. When planning diet-microbiome studies researchers should take into consideration numerous participant features when determining inclusion and exclusion criteria. Demographic information should be collected with a focus on features and exposures known to impact the microbiome. Information about medication use, including commonly consumed over-the-counter medications should be recorded. Depending on the research question, researchers may choose to exclude participants who consume supplements, prebiotics, or probiotics. Alternatively, researchers may ask participants to maintain consistent use of nutritional products throughout the study period. Recent infection and antibiotic exposures should be considered as part of the inclusion and exclusion criteria as should alcohol consumption. In longitudinal studies, normal cycles including sleep cycles, menstrual cycles, and meal timing or fed/fasted cycles should be collected as covariates. The metabolic status of participants can be an important covariate and, when possible, exercise level and hydration status should be recorded and held constant. For research with metabolic outcomes, exclusion of participants who have recently experienced weight loss or weight gain may be important. Typically, we exclude pregnant and lactating participants from our dietary intervention studies, but in some instances it may be appropriate to include these individuals. Finally, bowel habits should be considered at enrollment and during study participation. When collecting multiple consecutive stool samples from participants the frequency of stooling can impact study retention and complicate study timelines so it can be helpful to interview participants about their usual bowel habits before enrollment.

Figure 3

Example diet and microbiome sampling timeline for a longitudinal cross-over dietary intervention study. We recommend that diet-microbiome studies longitudinally follow individual participants and include multiple consecutive microbiome samples per study timepoint or period. As an example, a cross-over study design using repeated longitudinal sampling would include a total of 9–15 microbiome samples per person over three sampling time points. Here we show the mid-range of 4 samples per time point (labeled with M). Optimally, for each microbiome sample collected, 2–3 days of dietary records will be collected from each time point for a total of 15–18 dietary records over the study. In this example 2 days of records are collected prior to each sample (labeled with D). These records overlap, resulting in a total of 5 days of dietary records at each time point.