Abstract 382P
Background
PDAC is one of the most lethal malignancies with a poor prognosis. The microbiome's influence on cancer progression and the efficacy of treatments has emerged as a pivotal area of research. Here, we present a biomarker study exploring changes of the microbiome in PDAC patients during chemotherapy and its impact on prediction of oncological outcomes.
Methods
In a German consortium (Heidelberg, Tübingen, Munich), we prospectively recruited 67 cancer patients with locally advanced or metastatic PDAC requiring chemotherapeutic treatments (mFOLFIRINOX vs. Gemcitabine-based regimens). We collected saliva and fecal samples at the onset of chemotherapy and at four-week intervals up to nine months for shotgun metagenomic sequencing to investigate microbiome composition and function. Comprehensive clinical data collection included patient demographics, tumor characteristics, chemotherapy and other drug regimens as well as monitoring chemotherapy response and toxicity. An additional validation cohort of 37 PDAC patients from the University Medical Center Hamburg-Eppendorf was included in the study.
Results
We observed no significant differences in alpha diversity between chemotherapy responder vs. non-responder patients. Composition analyses did not reveal any differences between different chemotherapy regimens, but we found associations of microbiome composition with 3-months chemotherapy response. By applying a logistic regression model to predict the 3-months response rates, we achieved an AUROC score of 0.87 indicating a high level of accuracy in outcome prediction. On a species level, we were able to uncover that the gut commensals Christensenella minuta and Collinsella spp. are the most important species predicting chemotherapy response, while also being associated with chemotherapy-related bone marrow toxicity.
Conclusions
Our study explored for the first time the intricate relationships of chemotherapy, microbiome diversity, and response outcomes in PDAC patients in a multicentric biomarker study. Notably, our data highlight the potential of predictive microbiome models in projecting treatment efficacy which has barely been accomplished in this hard-to-treat cancer entity.
Legal entity responsible for the study
Christoph Stein-Thoeringer.
Funding
DFG.
Disclosure
All authors have declared no conflicts of interest.