Abstract

Background Sepsis and septic shock are severe complications for surgical malignancy patients. Conventional diagnostics often fail to capture the complex infectome in these populations. This study aimed to characterize the distinct microbial and resistome landscapes in cancer versus non-cancer patients using multi-site metagenomic next-generation sequencing (mNGS) to support specific antimicrobial strategies. Methods We conducted a single-center retrospective cohort study at the General Surgery ICU of Xuanwu Hospital, including 107 septic shock patients (52 cancer; 55 non-cancer). mNGS was performed on blood, bile, ascitic fluid, and bronchoalveolar lavage samples to identify pathogens and antibiotic resistance genes (ARGs). Findings were analyzed for their association with ICU length of stay and mortality. Results Cancer patients were significantly older (median 68 vs. 51 years, p  < 0.0001) with higher comorbidity scores (CCI: 7.0 vs. 4.0, p  = 0.006). However, mNGS revealed a lower pathogen detection rate in cancer patients (53.85% vs. 85.45%, p  = 0.0006) and a lower incidence of bacteremia (25.0% vs. 45.45%, p  = 0.0426). Cancer patients had shorter ICU LOS (9 vs. 13 days, p  = 0.0369) and antibiotic durations (7 vs. 11 days, p  = 0.0368). Dominant pathogens included Klebsiella pneumoniae and Enterococcus faecium , harboring diverse ARGs across beta-lactam and aminoglycoside categories. Multivariate Cox regression identified IL-6 ( p  = 0.018) was significant prognostic indicators for cancer patients. We also examined the distribution of virulence factors, despite their low detection rates. Conclusion Septic shock in cancer patients exhibits a unique resistome signature and distinct prognostic drivers. The identification of microbial targets via mNGS was associated with the implementation of targeted antimicrobial strategies and inflammation monitoring. These findings suggest that mNGS provides valuable molecular insights that may support clinical