PDA-MAPS - Pancreatic Ductal Adenocarcinoma - Microbiome As Predictor of Subtypes - Fistula

Postoperative pancreatic fistula (POPF) is a common and potentially severe complication following pancreas surgery. To date, there is no objective predictor anticipating the occurrence of POPF. Here, we aim to prospectively evaluate the orointestinal microbiome as predictor for POPF. In total 105 patients were recruited prospectively before elective pancreas surgery. To this end, buccal, and rectal swabs were collected (in total 203 samples). After excluding patients with delayed surgery (> 1 week after recruitment) and pancreatectomy, as well as all samples that had not reached the desired sequencing depth, 75 buccal and 70 rectal samples (n= 75) were included in the subsequent analysis. Microbial composition was determined by Oxford Nanopore Technologies. For buccal samples 16S rRNA sequencing, and for rectal swabs metagenomic sequencing was performed. Out of 75 buccal samples 19 were collected from patients with type B or C fistula (rPOPF), and 56 showed biochemical leak (BL) or no fistula (nPOPF). Out of 70 rectal samples, 17 were collected from patients belonging to rPOPF. Significant differences between groups were detected for β-diversity (unweighted UniFrac p-value = 0.003) and α-diversity (observed species, p-value 0.004) in buccal but not rectal swabs. Notably, even after incorporation of 67 potentially confounding variables, the results remain significant (p-value = 0.002). Interestingly, the body-mass index (BMI) was the most obvious confounding factor. Four species were found to be differentially abundant in the relevant fistula group: Streptococcus gorondii, Gemella haemolysans, Haemophilus parainfluenzae, Gemella sanguinis. Building a model with these species and the BMI resulted in an AUROC of 94.2 %. This combined model outperforms BMI alone (86.5 %) and species alone (78.6 %) Our results show that POPF can be predicted by the oral microbiome in combination with BMI.