sample sizes and a high resolution of clinical phenotypes and medication are required, while accounting for variables known to affect the gut microbiome. Finally, drug effects are often dose-dependent, yet dosage is rarely considered in microbiome studies.To overcome these limitations, we propose a general framework for separating disease from treatment associations in multi-omics cross-sectional studies and apply it to gut metagenomic, host clinical and metabolomic measurements of 2,173 European residents from the multicentre MetaCardis cohort. The MetaCardis cohort includes patients with metabolic syndrome, severe and morbid obesity, T2D, acute and chronic coronary artery disease and heart failure, and healthy control individuals. Considering cardiometabolic disease (CMD) and herein frequently prescribed medications, we investigated drug-hostmicrobiome associations for eight major indications (antidiabetic,

Aims Recent technical developments have allowed the study of the human microbiome to accelerate at an unprecedented pace. Methodological differences may have considerable impact on the results obtained. Thus, we investigated how different storage, isolation, and DNA extraction methods can influence the characterization of the intestinal microbiome, compared to the impact of true biological signals such as intraindividual variability, nutrition, health, and demographics. Methods and results An observative cohort study in 27 healthy subjects was performed. Participants were instructed to collect stool samples twice spaced by a week, using six different methods (naive and Zymo DNA/RNA Shield on dry ice, OMNIgene GUT, RNALater, 95% ethanol, Zymo DNA/RNA Shield at room temperature). DNA extraction from all samples was performed comparatively using QIAamp Power Fecal and ZymoBIOMICS DNA Kits. 16S rRNA sequencing of the gut microbiota as well as qPCRs were performed on the isolated DNA. Metrics included alpha diversity as well as multivariate and univariate comparisons of samples, controlling for covariate patterns computationally. Interindividual differences explained 7.4% of overall microbiome variability, whereas the choice of DNA extraction method explained a further 5.7%. At phylum level, the tested kits differed in their recovery of Gram-positive bacteria, which is reflected in a significantly skewed enterotype distribution. Conclusion DNA extraction methods had the highest impact on observed microbiome variability, and were comparable to interindividual differences, thus may spuriously mimic the microbiome signatures of various health and nutrition factors. Conversely, collection methods had a relatively small influence on microbiome composition. The present study provides necessary insight into the technical variables which can lead to divergent results from seemingly similar study designs. We anticipate that these results will contribute to future efforts towards standardization of microbiome quantification procedures in clinical research.