Mark M Kelly, Nicole Johnston, Joanne Kuznicki, Stacy A Johnson, Hannah Imlay
The diagnostic threshold of PMN count ≥ 250 cells/mm3 used to diagnose SBP in cirrhosis may be inappropriate for diagnosing BP-MA. We identified hypothetical diagnostic thresholds for BP-MA, although performance characteristics were suboptimal, and validation from larger studies would be required prior to clinical utilization.
BACKGROUND: Bacterial peritonitis in patients with malignant ascites (BP-MA) is an understudied, high-mortality condition. While ascitic polymorphonuclear (PMN) cell count ≥ 250 cells/mm3 defines the diagnosis of spontaneous bacterial peritonitis (SBP) in patients with cirrhosis, an ascitic fluid parameter for the diagnosis of BP-MA has not been established.
DESIGN: We performed a retrospective observational study at a single academic medical center evaluating patients with malignant ascites who underwent paracentesis with a complete laboratory evaluation between 2020 and 2025. We used uni- and multivariable logistic regression to identify patient and fluid characteristics associated with positive ascitic fluid cultures. Receiver operating characteristic (ROC) curve analysis identified hypothetical diagnostic thresholds for BP-MA and examined performance characteristics.
RESULTS: A total of 337 paracenteses in 261 patients were evaluated, with 318 samples (244 patients) having negative ascitic fluid cultures and 19 samples (17 patients) having positive ascitic fluid cultures. The best performing diagnostic thresholds for BP-MA were ascitic PMN percentage ≥ 63% (sensitivity 68.4%, specificity 90.8%) and ascitic PMN cell count > 1318 cells/mm3 (sensitivity 68.4%, specificity 93.0%). A PMN cell count ≥ 250 cells/mm3 was associated with a sensitivity of 73.7% and specificity of 72.4%.
CONCLUSION: The diagnostic threshold of PMN count ≥ 250 cells/mm3 used to diagnose SBP in cirrhosis may be inappropriate for diagnosing BP-MA. We identified hypothetical diagnostic thresholds for BP-MA, although performance characteristics were suboptimal, and validation from larger studies would be required prior to clinical utilization.