Response for: International Investigation of the Gut-Lung Axis in Systemic Sclerosis-Interstitial Lung Disease

Dear Editor,

We are grateful for Pagnoni and colleagues’ interest in our study. Indeed, this exploratory investigation is to our knowledge, the first multi-center study linking bacterial species to systemic sclerosis-related interstitial lung disease (SSc-ILD). In addition, it is one of few studies in the field of rheumatology, which adjusted for known confounders in analyses of the intestinal microbiota.

We acknowledge the methodological issues raised by Pagnoni et. al., including suggestions on how to probe generalizability across study centers, how to synchronize biospecimen collection and on statistical methodology. Great technical advancements have been made during the last decade in this field, and we appreciate these comments for our future work. We also note other novel approaches on how to integrate whole genome shotgun sequencing with metabolomics and clinical data [1].

We recognize Pagnoni and colleagues’ concern regarding multiple comparisons in our species-level association analyses. As this is the first study of its kind, we employed a discovery-based approach emphasizing effect size estimation and hypothesis generation as compared to hypothesis testing. In the context of microbiome research, wherein large sample sizes are required to detect effects, especially under conditions of multiple hypothesis testing, the results suggest potential targets for future hypothesis-driven analyses. In addition, Cohen’s d was provided to facilitate sample size estimations needed to guide future investigations in this emerging area of research.

To deal with generalizability of effects, we specified a general linear model with a random effect for site. This model accounts for the variability in outcomes due to differences in site. The estimate for group is the weighted average of the within-site and between site components of the group differences. Although this model provides the statistical basis for generalization of results across sites, we agree that the leave-one-out-sensitivity analyses is an important tool for assessing the influence of each site on the results and the robustness of statistical models. We acknowledge that there may be important site-specific differences or associations with the microbiome due diversity of diets and environmental exposures between international sites that the random effects model does not capture. Ultimately, replication in well-powered independent datasets is required to confirm the validity and generalizability of the findings.

In agreement with Pagnoni and colleagues, we welcome future integrative “multi-axis” approaches, possibly incorporating disease phenotyping based on circulating biomarkers, immune profiling of leukocytes and echocardiological assessments. In this study, we used quantitative image analysis of high-resolution computed tomography [2], a valid approach for approximating the burden of pulmonary fibrosis that is more sensitive than visual assessment. Lung ultrasound, while useful as a bedside adjunct and highly sensitive for subpleural changes, does not capture deep or whole-lung volumetric disease and lacks standardization; therefore, it was not used for primary ILD burden quantification in this study. We further concur that KL-6 is a promising biomarker for SSc-ILD and future studies on this biomarker, in combination with immune profiling of leukocytes, could be of interest in microbiota research. The presence of a gut-lung axis does not diminish the importance of cardiopulmonary dimensions in SSc-ILD.

In summary, we sincerely appreciate the interest raised in our exploratory investigation. Our international group welcomes participation of other sites and is grateful for the feedback on how to build upon this research in our future studies.