Bringing doctors together

or

A free, private network for verified physicians

  • Discover your community

  • Stay current

  • A profile that represents you

  • Find your people

About RoonPrivacyTerms

    Overview

    Ascites fluid PMN counts below the SBP threshold predict 30-day mortality in a national hospitalized cirrhosis cohort.

    Nov 11, 2026DOI: 10.1097/hc9.0000000000001071Journal ArticleHepatology Communications

    Written by: Scott SilveyNilang PatelJacqueline G O'LearyJasmohan S Bajaj

    INTRODUCTION

    Altered immune function and gastrointestinal bacterial translocation are hallmarks of advancing cirrhosis and are associated with poor outcomes.12 Bacterial translocation can result in spontaneous bacterial peritonitis (SBP), and ≥250 PMN/mm3 of ascites is currently considered diagnostic of SBP with or without a positive culture.3 Recognition and treatment of all patients with SBP is essential to prevent acute kidney injury (AKI), organ failures, acute-on-chronic liver failure, and death.456 Below the 250 PMN/mm3 threshold, the PMN (polymorphonuclear neutrophil) count may still reflect altered intestinal microbiome, bacterial translocation, and immune recruitment in patients with cirrhosis. However, with changes in the microbiome, demographics of cirrhosis, and antimicrobial resistance, the impact of ascites PMN/mm3 threshold and linkage to adverse outcomes other than SBP needs re-evaluation.78 Prior single-center studies have hinted that a lower PMN count could be a harbinger of worse outcomes, but larger nationally-representative studies are needed.91011 Our aim was to determine the impact of different ascites fluid PMN thresholds on 30-day mortality in a national cirrhosis cohort.

    METHODS

    The Veterans Affairs Corporate Data Warehouse (VA-CDW) is a national database of Veteran electronic health records, containing information related to patient demographics, diagnoses, medications, laboratory findings, vitals, and outcomes. We collected data from veterans admitted between 2009 and 2024 for complications of cirrhosis using validated ICD-9/10 codes (Supplemental, https://links.lww.com/HC9/C507) who received a paracentesis ≤24 hours of hospital admission, identified through validated CPT codes.12 IRB approval with waiver of informed consent was obtained from the Richmond VA IRB for this protocol before study initiation. Among those with multiple admissions within the study time period, the first (index) instance of cirrhosis with a paracentesis was evaluated. Basic comparisons were made between those who survived or died based on their admission ascites PMN/mm3 <100, 100–249, and ≥250. SBP prophylaxis (primary and secondary) before the tap was also captured.We analyzed PMN thresholds ranging from 50 to 500 with the goal of identifying the cutoff value that optimized positive predictive value (PPV) and odds ratio (OR) of 30-day mortality. Specifically, we tested this range of PMN threshold values using logistic regression models/confusion matrices over each potential cutoff and then used smooth and monotonically increasing splines to fit curves to the raw data, visually selecting inflection points where the PPV and ORs did not increase substantially.13 Covariate-adjusted odds ratios, which included additional demographics, clinical variables, and in-hospital events (liver/renal/coagulation failure defined by EASL criteria)14 that were significantly associated with 30-day mortality were also analyzed as a sensitivity analysis. Two adjusted models were fit: one that contained baseline prognostic factors only, and one that contained both baseline covariates and those that occurred during hospitalization. Other metrics [sensitivity, specificity, NPV (negative predictive value)] were also calculated at the selected cutoff value. Then, we compared predictive metrics at our selected cutoff value to the standard ≥250 PMN value traditionally used to diagnose SBP. As an additional sensitivity analysis, prior SBP history was modeled as an interaction term. Finally, as an alternative analysis, PMN counts were modeled continuously using natural cubic splines and a logistic regression framework with internal knots at the 20th, 40th, 60th, and 80th percentiles of the distribution. The model containing the spline parameterization of PMN was tested for the overall significance of the nonlinearity terms using the likelihood ratio test, and the model’s fitted values were plotted to visualize the nonlinear relationship.
    We could not extract ascites albumin, SAAG, and total protein data due to it being stored in a separate dataset within CDW. RStudio version 4.4.1 was used for all statistical analyses. All hypothesis tests were two-sided with α=0.05.

    RESULTS

    We collected data from 4111 veterans with cirrhosis who had a paracentesis ≤24 hours of admission (median age 66, 96.3% male, 31.1% alcohol-associated cirrhosis), with a 16.9% 30-day mortality. In addition, 0.4% (n=17) received a liver transplant within 30 days, 27.9% (n=1145) were readmitted within 30 days after index hospital discharge, and 18.1% (n=743) had SBP either on admission (PMN≥250/mm3) or within 30 days of discharge. In all, 458 (11.1%) had new SBP develop in the 30 days after the initial paracentesis. In the cohort, 19.1% (142) were culture-positive, 45.1% (64) of whom had an ascites PMN/mm3 <250 (bacterascites), and 54.9% (78) with an ascites PMN/mm3 ≥250. Among patients with positive cultures, 51% of organisms were gram-positive, while 49% were gram-negative. Among organisms isolated, Escherichia coli (22.5%) was the most common, followed by Streptococcus (22.5%), Staphylococcus (18.8%), and Klebsiella (11.2%) species.
    A total of 561 (70.3% of those with SBP diagnosed by ascites PMN/mm3 and 13.6% of the total patients) had PMN counts >250, while the median PMN/mm3 count among the entire cohort was 19 [IQR: 6–68], which was also significantly higher among those who died versus survived 30 days (33 vs. 18, p<0.001). Patients who died were also significantly older, with higher Charlton Comorbidity Index (CCI) and MELD-Na score (Table 1).
    Comparison between those who died/survived at 30 days after admission
    Abbreviations: CCI, Charlson Comorbidity Index; ICU, intensive care unit; INR, international normalized ratio; IQR, interquartile range; MELD-Na, Model for End-Stage Liver Disease–sodium; PMN, polymorphonuclear neutrophil; PPI, proton pump inhibitor; SBP, spontaneous bacterial peritonitis; WBC, white blood cell.
    The distribution of PMN counts was skewed, with ~80% (n=3280) of patients with PMN/mm3 <100, 6.5% with PMN/mm3 between 100 and 249, and 13.6% of patients with ≥250 PMN/mm3 (Supplemental Table S2, https://links.lww.com/HC9/C507). Patients with 100–249 PMNs had higher admission WBC counts, higher prevalence of alcohol etiology, and lower culture positivity compared with those with <100 PMNs. Furthermore, new SBP within 30 days of the initial paracentesis became more common as the initial paracentesis PMN count increased (Table 2).
    Comparisons and outcomes according to PMN count in ascites
    Abbreviations: CCI, Charlson Comorbidity Index; ICU, intensive care unit; INR, international normalized ratio; MELD-Na, Model for End-Stage Liver Disease–sodium; PMN, polymorphonuclear neutrophil; PPI, proton pump inhibitor; SBP, spontaneous bacterial peritonitis; WBC, white blood cell.
    Among the patients, 13.1% (n=538) of patients were on SBP prophylaxis (fluroquinolones or trimethoprim-sulfamethoxazole) up to 90 days before paracentesis; the distribution of PMN counts did not differ significantly between those prescribed versus not prescribed primary or secondary SBP prophylactic antibiotics (median 17 on prophylactic antibiotics vs. 19 not on prophylactic antibiotics, p=0.11). In addition, 24.0% (n=987) were prescribed any antibiotic (Supplemental Table S1, https://links.lww.com/HC9/C507) up to 30 days after the index admission, which was significantly lower in those who died (p<0.001).
    Those with ascites PMN<100 had the lowest unadjusted mortality rate, but mortality rates were similar in those with ascites PMN/mm3 of 100–249 and ≥250 PMNs; however, transplant rates were similar across all 3 PMN thresholds (Table 2). Visual analysis of PMN thresholds revealed a PMN count near 100 optimized both positive predictive value and the odds ratio of death (Figure 1), and simultaneous computational assessment of the exact elbow of each curve revealed PMN counts between 85 and 105 as the optimal value. Specifically, increasing PMN counts showed a roughly linear relationship with 30-day mortality until a value near 100 was reached, at which point a plateau occurred and diminishing returns became evident. Thus, we selected PMN >100 as our optimal cutoff value. At this cutoff (prevalence: 20.3%), the PPV of death was 25.3% (Sensitivity 30.4%, Specificity 81.8%, NPV 85.2%), and the odds ratio of death versus those below this threshold was 1.96 (95% CI: [1.63–2.35], p<0.001). Using the traditionally defined cutoff for SBP (PMN>250), the PPV of death (26.4%) and OR of death (1.97 [1.59–2.42]) were essentially unchanged, and other metrics also remained similar (Sensitivity 21.3%, Specificity 87.9%, NPV 84.6%). For those with PMN counts between 100 and 250, a 50-unit increase in PMN count corresponded to 11% further increased odds of death (OR: 1.11, 95% CI: 0.80–1.52), but this increase was not statistically significant (p=0.53). Re-calculation of the optimal threshold using covariate-adjusted logistic regression models including age, CCI, alcohol-associated cirrhosis, admission lactulose, WBC, MELD-Na, albumin, ICU transfer, and liver/renal/coagulation failure during admission attenuated the magnitude of the odds ratio at the >100 PMN/mm3 count threshold (adj. OR: 1.59 [1.29–1.97], p<0.001), but the non-significant increase in 30-day mortality odds with a rising PMN count persisted within the 100–250 PMN/mm3 group (adj. OR: 1.01 [0.70–1.45], p=0.99). In the alternative model containing only covariates collected at baseline, these results were consistent (adj. OR: 1.63 [1.33–2.00], p<0.001, adj. OR for PMN in 100–250 group: 1.06 [0.74–1.51], p=0.74). When prior SBP history was modeled as an interaction term with the PMN threshold, we found no evidence that this subgroup experienced altered odds of the proposed threshold on 30-day mortality (interaction p-value: 0.71). When modeling PMN count continuously using cubic splines, we found that there was evidence of an overall nonlinear effect (likelihood ratio test p<0.001) and that the fitted curve showed evidence of a plateau around 100 PMNs/mm3 (Supplemental Figure S1, https://links.lww.com/HC9/C507). Internal knots were selected at PMN counts of 4, 12, 31, and 103.
    Visual assessment of PMN threshold on positive predictive value (A) and odds ratio (B) of 30-day mortality. A: positive predictive value; B: odds ratio. Abbreviation: PMN, polymorphonuclear neutrophil.

    DISCUSSION

    Our data showed admission ascites paracentesis with ≥100 PMN/mm3 significantly enhanced 30-day multivariable mortality prediction independent of antibiotic use, cirrhosis severity, organ failure, and prior SBP history. This threshold for higher mortality remained steady even when baseline values were included, and using continuous cubic spline modeling.
    The local and systemic immune impairment, microbial alterations, and external influences such as alcohol and proton pump inhibitor (PPI) use modulate the microbiome and impact the progression of cirrhosis and bacterial translocation.14 While direct measurement of bacterial translocation is difficult, the use of ascites fluid PMN/mm3 as a corollary has been suggested.3 This has been especially successful for SBP diagnosis, where early diagnosis and treatment improve or prevent AKI, organ failures, and death. Therefore, we focused on short-term mortality that may be impacted by the admission that prompted the paracentesis.
    Our data show that a considerably lower PMN/mm3 of ascites (≥100) is as predictive as the established one (≥250) for mortality prediction, even after controlling for cirrhosis severity, antibiotic prophylaxis for SBP, PPI use, hepatic encephalopathy therapy, demographics, and critical illness-related parameters.15 The robustness of this cutoff was further demonstrated by analysis using baseline-only variables and modeling the PMN counts as a continuous variable. Importantly, this included patients who were culture-positive regardless of PMN/mm3 count, and those who were diagnosed with SBP due to PMN/mm3 ≥250, which makes the results generalizable to routine clinical practice. While we studied differences in culture positivity, which tended to be higher with increasing PMN/mm3, the overall culture positivity rate was low. Moreover, the bacteriology of those that were culture-positive is similar to previously published data.16 This underlines the important role PMN/mm3 of ascites plays in predicting outcomes and could add nuance toward using it as a prognostic biomarker.161718
    As expected, most patients who underwent paracentesis had PMN/mm3 counts <100. However, the PMN/mm3 count as a continuous variable was higher in those who died, and it was not significantly affected by either SBP prophylaxis or prior SBP occurrence.3 While the reasons for these similarities in prediction among thresholds are unclear, higher WBC count and alcohol-associated etiology in the 100–249 PMN/mm3 group versus the <100 PMN/mm3 group could be a factor since they are associated with greater inflammation and intestinal barrier dysfunction.19 Several other factors can determine short-term mortality in patients with cirrhosis.15 We therefore included new antibiotic therapy, comorbidities, organ failures, and critical care need as well as cirrhosis severity and WBC count in the regression models. Despite this, we found a significant impact of the >100 PMN/mm3 cutoff, which strengthens the validity of our findings. Specifically, new antibiotic use, which was highest in PMN>250/mm3 but was similar between those with <100/mm3 and 100–250/mm3, was not associated with mortality.
    We confirmed and extended prior single-center experiences into a larger national US cohort.91011 However, veterans tend to be older men with more comorbid diseases than the background US population, which may reduce generalizability and is a limitation.20 However, similar patterns of mortality at a lower PMN count in our study compared with non-VA single-center studies underline the likelihood of applicability of these results in the non-Veteran population.11 Another limitation is the reliance on ICD-10 and CPT codes; however, these are validated codes, and the occurrence of paracentesis was confirmed by the ascites PMN/mm3 results.91021 Paracentesis provision to selectively sicker individuals could also skew the data; however, the wide range of PMN/mm3 counts in our results, with the majority of patients <100 PMN/mm3, makes this unlikely. Given the sparseness of data, we do not have additional ascites fluid studies, such as albumin and protein and resultant SAAG, which is another limitation. However, the culture results are reflective of previously published bacteriology, and we controlled for both primary and secondary SBP prophylaxis, which could be a corollary of lower ascites albumin.
    The findings suggest that ascites fluid PMN count may be a harbinger of mortality even at a lower threshold than formally used for SBP diagnosis. The implications of our results could be to ensure prompt paracentesis so that the resultant ascites fluid PMN counts may additionally be used to refine prognostication.622 However, given the climate of antibiotic resistance,122223242526 this does not support initiating antibiotic prophylaxis at this cutoff. Future studies are needed to study potential non-antibiotic interventions to improve mortality.727
    We conclude that in a large national database of inpatients with cirrhosis who underwent paracentesis within 24 hours of hospitalization, an ascites PMN/mm3 count of ≥100/mm3 shows independent predictive value for 30-day mortality that is statistically similar to the SBP threshold of ≥250/mm3. Prompt paracentesis and close monitoring of patients with PMN count ≥100/mm3 could improve prognostication of 30-day outcomes.
    HCHepatology Communications

    Hepatology Communications

    Official Account
    https://journals.lww.com/hepcomm
    hepatologycommunications