Performance of Early Therapeutic Paracentesis in Spontaneous Bacterial Peritonitis and Its Effect on Clinical Outcomes

ABSTRACT

Background and Aims

Spontaneous bacterial peritonitis (SBP) is common in decompensated cirrhosis and often accompanied by large volume ascites. However, the safety of early therapeutic paracentesis is unknown and may lead to unnecessary delay in procedural performance. We aimed to assess whether paracentesis performance characteristics in the setting of SBP influenced renal or circulatory outcomes.

Methods

A single‐center, retrospective cohort study was performed involving adult patients diagnosed with SBP based on ascites fluid studies during index admission between 2010 and 2020. Patient and procedural characteristics were collected and analyzed with respect to paracentesis timing (early vs. delayed) and volume (diagnostic only vs. non‐large volume therapeutic vs. large volume therapeutic). The primary outcome was a composite of the development of AKI and/or circulatory dysfunction. Secondary outcomes included 30‐day mortality, 90‐day mortality, and change in baseline renal function.

Results

One‐hundred forty‐four patients met inclusion criteria during the study timeframe (Child‐Pugh C 74%, mean MELD‐Na 25). Sixty‐six (46%) met ACLF criteria at the time of SBP diagnosis. One‐hundred eight (75%) patients underwent a therapeutic paracentesis, of which approximately two‐thirds were performed within 24 h of admission (early); 41% were large volume (≥ 5 L of ascites removed). There was no difference in rates of the primary outcome with respect to either paracentesis timing or volume. Short‐term mortality and change in baseline renal function were independent of procedural characteristics.

Conclusions

Therapeutic paracentesis does not appear to induce renal or circulatory dysfunction in patients with SBP.

Abbreviations

AKI

acute kidney injury

CLIF‐C ACLFs

CLIF consortium ACLF score

CLIF‐OFs

CLIF organ failure score

CNNA

culture‐negative neutrocytic ascites

ICU

intensive care unit

K‐M

Kaplan–Meier

LASSO

least absolute shrinkage and selection operator

NNBA

non‐neutrocytic bacterascites

PICD

paracentesis‐induced circulatory dysfunction

PRA

plasma renin activity

RAAS

renin‐angiotensin‐aldosterone system

SBP

spontaneous bacterial peritonitis

1. Introduction

Spontaneous bacterial peritonitis (SBP) is common in patients with cirrhosis and ascites. Annual incidence is 7%–30%, with higher rates of recurrence depending on whether secondary prophylaxis is prescribed [1, 2]. Clinical presentation is variable and up to a third of cases are incidentally diagnosed in outpatients undergoing therapeutic paracentesis [3]. Standard‐of‐care treatment includes careful selection of antibiotics and volume expansion with concentrated albumin infusion to prevent acute kidney injury (AKI) and decrease mortality [4]. Given the potential for poor outcomes, the rapidity of SBP diagnosis and initiation of therapy is critical [5].

Current guidelines recommend diagnostic paracentesis in all admitted patients and/or outpatients with cirrhosis and ascites who have signs or symptoms of infection [6, 7]. However, many patients also undergo therapeutic (i.e., larger volume) paracentesis for comfort or clinical optimization. While diagnostic studies from a therapeutic paracentesis are frequently obtained for clinical expediency, this practice is not uniform. Some clinicians elect to delay therapeutic paracentesis until SBP is ruled out or treatment has begun given concern that the procedure may cause adverse outcomes.

Despite the aforementioned practice, it is unknown whether large volume paracentesis in the setting of acute infection has negative physiologic effects with respect to hemodynamics and the local immune response. There are currently no recommendations and a paucity of data to inform optimal parameters for the performance of therapeutic paracentesis in patients presenting with cirrhosis and SBP. To address this critical literature gap, we designed a retrospective study to assess whether larger volumes of ascites removed or early performance of therapeutic paracentesis in patients with cirrhosis, ascites, and an incident diagnosis of SBP result in renal and/or circulatory dysfunction.

2. Methods

2.1. Patient Selection and Study Methodology

We conducted a single‐center, retrospective cohort study of patients with cirrhosis and admission diagnosis of SBP from January 1, 2010 to December 31, 2020. The cohort was derived by querying a cloud‐based institutional research database with the following ICD‐10 codes, CPT codes, and Boolean operators: (K70.3, alcoholic cirrhosis of liver OR K74, fibrosis and cirrhosis of liver) AND K65.2 (SBP) AND (1020907, abdominal paracentesis [diagnostic or therapeutic] OR 49083, abdominal paracentesis [diagnostic or therapeutic] with imaging guidance OR 49082, abdominal paracentesis [diagnostic or therapeutic] without imaging guidance). Inclusion criteria were age ≥ 18 years, admission to an acute care or intensive care unit (ICU) bed, a diagnosis of cirrhosis by either liver biopsy or compatible clinical criteria (i.e., typical physical exam findings, imaging findings of a nodular liver and/or portal hypertension, laboratory evidence of portal hypertension and/or synthetic dysfunction), presence of ascites, and a diagnosis of SBP. Exclusion criteria included outside facility transfer prior to SBP diagnosis, outpatient SBP diagnosis without subsequent admission within a 24‐h period, non‐cirrhotic portal hypertension and secondary bacterial peritonitis.

Patient demographics, etiology and severity of liver disease, diagnostic paracentesis study data (cell count, bacterial culture), timing and clinical setting (bedside or radiology suite) of therapeutic paracentesis with relation to diagnostic paracentesis, and amount of ascites fluid removed were captured. Infection treatment parameters including receipt of concentrated albumin infusions, antibiotic selection, duration of therapy, and rationale were also recorded. Standard laboratory values including sodium, creatinine, total bilirubin, INR, and MELD‐Na scores were obtained at several intervals including:

1. 3 months prior to admission (initial baseline),

2. Admission,

3. Diagnosis of SBP (may overlap with interval 2),

4. Time of therapeutic paracentesis (may overlap with 2 and/or 3),

5. 48‐h after diagnosis of SBP,

6. Discharge, and

7. 3 months post‐discharge (new baseline).

Clinical measures of liver disease acuity including the CLIF Organ Failure Score (CLIF C‐OFs), ACLF Grade, and CLIF Consortium ACLF Score (CLIF‐C ACLFs) were captured at diagnosis of SBP, time of therapeutic paracentesis, and 48‐h after SBP diagnosis in accordance with established methodology [8, 9]. The study was approved by the University of Virginia IRB (HSR‐IRB #21910) and performed in accordance with the Declaration of Helsinki. Informed consent was waived given the retrospective nature of this study.

2.2. Definition of Spontaneous Bacterial Peritonitis and Liver Disease Acuity Scores

Patients were defined as having SBP if they met any of the following in the absence of an overt source of peritonitis (i.e., secondary bacterial peritonitis) given clinical equivalence and identical management: [10]

1. An ascites fluid absolute neutrophil count ≥ 250 cells/μL with positive gram stain or culture (SBP), or

2. An ascites fluid absolute neutrophil count ≥ 250 cells/μL with negative gram stain or culture (culture‐negative neutrocytic ascites; CNNA), or

3. An ascites fluid absolute neutrophil count < 250 cells/μL with positive gram stain or culture deemed not to be a contaminant by the primary team or consulting hepatologist (non‐neutrocytic bacterascites; NNBA).

Cases of NNBA were included given a significant proportion of these cases progress to SBP and its presence is associated with similar infection‐related and in‐hospital mortality as SBP [11].

2.3. Characterization of Paracentesis

Given the aims of this study, patients were grouped independently by timing of therapeutic paracentesis and volume of ascites removed (Figure 1). With regard to timing, patients were divided into two groups: early therapeutic paracentesis (within 24 h of SBP diagnosis) or delayed therapeutic paracentesis (between 1 and 5 days after SBP diagnosis); patients not undergoing therapeutic paracentesis (diagnostic only) were excluded from this analysis. Grouping by volume of ascites removed was as follows: diagnostic only (< 100 mL removed), non‐large volume (100 mL up to 5 L), and large volume (5 L or more) to reflect the accepted definition of a large volume paracentesis [12].

FIGURE 1.

Flow diagram with rationale for exclusion of subjects after detailed chart review (A). A complete data set included seven different time points as described in the Section 2. The final cohort met all inclusion criteria. Two independent analyses were performed on patients with a therapeutic paracentesis: One with respect to procedural timing and another with respect to volume of ascites removed (B).

2.4. Definition of Primary and Secondary Outcomes

The primary outcome selected for this study was a composite of known complications of both SBP and therapeutic paracentesis: AKI and circulatory dysfunction. A simplified definition of the International Club of Ascites criteria for AKI was utilized such that a rise in serum creatinine by 0.3 mg/dL or greater 48‐h after the performance of paracentesis was considered an AKI [13]. Circulatory dysfunction was defined by a decrease in mean arterial pressure ≥ 10 mmHg and/or initiation of vasopressors at 48‐h post‐paracentesis [14]. While paracentesis‐induced circulatory dysfunction (PICD) has traditionally been defined as a compensatory vaso‐hormonal response with a > 50% rise in plasma renin activity (PRA) to > 4 ng/mL/h at day 6 after paracentesis [15], the retrospective nature of this study precluded use of this definition. In the United States, PRA or aldosterone levels are not routinely measured in clinical practice. Secondary outcomes of interest included 30‐day mortality, 90‐day mortality, and change in baseline renal function (defined as 3‐month post‐SBP serum creatinine minus 3‐month pre‐SBP serum creatinine).

2.5. Statistical Analysis

Quantitative variables were summarized using means and standard deviations for normally distributed data; nonparametric data was reported as median and interquartile range. Between‐group comparisons utilized appropriate independent sample t‐tests or Wilcoxon rank‐sum tests. Categorical variables were summarized using frequency and percentages and between‐group comparisons were assessed using the Fisher's exact or χ ² test, as appropriate. Covariates of interest included age, biological sex, admission Child‐Pugh classification, antibiotic prophylaxis at time of SBP diagnosis, and clinical setting of paracentesis.

Significant variables on univariate analysis were initially selected for inclusion within a multivariate regression model if p < 0.25. Least absolute shrinkage and selection operator (LASSO) regression was subsequently performed to assess whether other clinically important variables were predictive of the primary outcome. Separate multivariate logistic regression models were constructed to analyze the effect of volume of ascites removed and the effect of delaying therapeutic paracentesis on the primary outcome. To determine if the volume of ascites removed was associated with any outcome, Kernel regression was performed to assess for nonlinear relationships. This statistical technique determines local probabilities of an outcome, rather than assuming a linear relationship. Reliability of probability estimates is dependent on the local sample size (i.e., fewer data points lead to greater uncertainty around the estimated probability of an outcome).

Time‐to‐event analysis utilizing the Kaplan–Meier (K‐M) method and log‐rank test for significance was performed to evaluate the effects of time and volume on 30‐day mortality. Estimation of 90‐day mortality was performed using a competing risk model inclusive of the clinically relevant variables utilized in the logistic regression models. Patients were right‐censored if alive at the end of the 90‐day follow‐up period, with interval censoring if liver transplantation or death occurred.

Alpha was prespecified as 0.05 for all statistical tests. Statistical analysis was performed using SAS (SAS version 9.4, Cary, NC) and STATA 17 (StataCorp, College Station, TX).

3. Results

3.1. General Patient Characteristics

Initial database search yielded 199 patients potentially meeting inclusion criteria. After manual chart review, 55 patients were excluded (Figure 1A), leading to a final analysis cohort of 144 patients (Figure 1B). As seen in Table 1, the mean age was 55 ± 11 years and 37% of the cohort was female. Etiology of cirrhosis was mixed and typical of a tertiary care transplant center. Nearly three‐quarters of patients were Child‐Pugh C, with the remaining quarter Child‐Pugh B. Median MELD‐Na at time of paracentesis was 25 (20, 29) and 24 (20, 29) at 48‐h post‐procedure. Seventy‐two (48%) patients met criteria for ACLF (Grade 1: 23%, Grade 2: 14%, Grade 3: 9%) at the time of paracentesis. Forty‐one patients (28%) were on β‐blockers at the time of SBP diagnosis. Between‐group time and volume analyses stratified revealed minimal differences (Tables S1 and S2).

TABLE 1.

Patient demographics and clinical characteristics.

Parameter	Total (n = 144)
Age (years), mean (SD)	55 ± 11
Sex (female), n (%)	53 (37)
Etiology of cirrhosis, n (%)
Alcohol	40 (28)
Nonalcoholic steatohepatitis	33 (23)
Viral	37 (26)
Autoimmune	7 (5)
Cryptogenic	10 (7)
Other	17 (12)
Child‐Pugh classification, n (%)
B	38 (26)
C	106 (74)
Outpatient SBP prophylaxis, n (%)	22 (15)
Beta‐blocker use, n (%)	41 (28)
Nonselective	36 (88)
Selective	5 (12)
MELD‐Na, median (IQR)
Paracentesis	25 (20, 29)
Creatinine	1.1 (0.8, 1.9)
Total bilirubin	3.7 (1.8, 7.3)
INR	1.9 (1.5, 2.3)
48 h post‐paracentesis	24 (20, 29)
Creatinine	1.0 (0.7, 1.7)
Total bilirubin	3.0 (1.6, 6.2)
INR	1.9 (1.5, 2.5)
Albumin (g/dL), median (IQR)	2.6 (2.3, 3.0)
Mean arterial pressure (mmHg), median (IQR)
Paracentesis	82 (74, 92)
48 h post‐paracentesis	79 (73, 89)
Vasopressor requirement, n (%)
At diagnosis	6 (4)
At time of therapeutic paracentesis	7 (5)
48 h post‐paracentesis	6 (4)
Hepatorenal syndrome, n (%)	24 (17)
On AMO therapy	21 (15)
Renal replacement therapy, n (%)
Chronic (prior to admission)	5 (4)
Acute, pre‐paracentesis	2 (1)
Acute, post‐paracentesis	4 (3)
Declined/not offered	7 (5)
Liver failure score
ACLF (grade 1/2/3), n (%)
Baseline	33 (23)/20 (14)/13 (9)
48 h post‐paracentesis	37 (26)/14 (10)/13 (9)
CLIF‐C OF score, median (IQR)
Baseline	9.0 (8.0, 10.0)
48 h post‐paracentesis	8.0 (7.0, 9.0)
CLIF‐C ACLF score, median (IQR)
Baseline	54 (47, 60)
48 h post‐paracentesis	51 (45, 57)
Admission or transfer to ICU during hospitalization, n (%)	42 (29)
Septic shock	14 (10)
Other	28 (19)
Patient disposition at time of paracentesis, n (%)
Acute care	128 (89)
ICU	16 (11)
Location of paracentesis, n (%)
Bedside	81 (56)
Radiology suite	63 (44)
Volume removed, therapeutic, mean (SD)	4.3 ± 3.0
Non‐large volume (< 5 L)	2.3 ± 1.2
Large volume (≥ 5 L)	7.3 ± 2.3
Infusion of 25% albumin, n (%)
Dose	143 (99)
Sort protocol a	107 (75)
Other	36 (25)
Timing
Pre‐paracentesis only	12 (8)
Post‐paracentesis, within 6 h	41 (29)
Post‐paracentesis, after 6 h	35 (24)
Pre‐ and post‐paracentesis	55 (38)
Ascites fluid gram stain, n (%)
No growth	88 (61)
Gram‐positive cocci	27 (18)
Gram‐negative rods	25 (17)
Other	2 (1)
Mixed	2 (1)
Spontaneous peritonitis subtype, n (%)
Spontaneous bacterial peritonitis (SBP)	45 (31)
Non‐neutrocytic bacterascites (NNBA)	11 (8)
Culture‐negative neutrocytic ascites (CNNA)	88 (61)
Bacteremia, n (%)	19 (13)
Type of infection, n (%)
Community‐acquired	101 (70)
Nosocomial	43 (30)
Antibiotic therapy, n (%)
Third‐generation cephalosporin (or equivalent)	95 (66)
Third‐generation cephalosporin + vancomycin	14 (10)
Broad‐spectrum	28 (19)
Other	7 (5)
Antibiotic duration (days)
Median (IQR)	7.0 (5.0, 10.0)
Mode	5.0
Range	3.0–42.0
Consistent with AASLD guidelines (5–7 days), n (%)	84 (58) b
Extended duration (> 7 days), n (%)	60 (42) c
Mortality, n (%)
Within hospitalization	15 (10)
30 days	24 (17)
90 days	37 (26)
6 months	47 (33)
1 year	59 (41)

Note: Continuous variables represented by means ± SD or medians (IQR) as appropriate. Quantitative variables represented by counts (%).

Abbreviations: ACLF, acute‐on‐chronic liver failure; AMO, albumin‐midodrine‐octreotide; CLIF‐C ACLF, European Foundation for the Study of Chronic Liver Failure Acute‐on‐Chronic Liver Failure score; CLIF‐C OF, European Foundation for the Study of Chronic Liver Failure Organ Failure score; ICU, intensive care unit; INR, international normalized ratio; MELD‐Na, Model for End‐Stage Liver Disease‐Sodium; SBP, spontaneous bacterial peritonitis.

^a Sort protocol based on Sort et al. [16], with administration of 1.5 g/kg of albumin on Day 1 and 1 g/kg of albumin on Day 3 of SBP diagnosis.

^b One patient died after 3 days of antibiotics.

^c Four patients received > 14 days of antibiotics (values: 15, 24, 42, 42).

3.2. Characterization of Spontaneous Bacterial Peritonitis

The majority of patients (61%) had CNNA and 11 patients (8%) had NNBA; the remainder met traditional SBP criteria (31%). Of those with positive ascites cultures, gram‐positive cocci were most common (48%), followed by gram‐negative rods (45%). Nineteen patients (13%) had concurrent bacteremia. All patients except one received concentrated 25% albumin (99%; mostly post‐paracentesis [92%]) and 75% of patients received the guideline‐recommend regimen of 1.5 g/kg on Day 1 and 1 g/kg on Day 3 of diagnosis [6, 16]. Antibiotics were administered for a median of 7.0 (5.0–10.0) days, with a range from 3 to 42 days and mode of 5 days. A third‐generation cephalosporin (or the equivalent) was the most common antibiotic prescribed (66%); broad‐spectrum antibiotics were utilized in 28 patients (19%). At admission, an ICU bed was required in 16 patients (11%), but an additional 26 patients (total: 42 patients, 29%) were eventually transferred to the ICU from the acute floor during their hospitalization.

3.3. Timing, Volume, and Location of Paracentesis

Three‐quarters of patients received a therapeutic paracentesis. Seventy (49%) patients had a therapeutic paracentesis performed within 24 h of admission; 38 patients (26%) had a therapeutic paracentesis performed after SBP diagnosis and later in the hospitalization. Thirty‐six patients (25%) only had diagnostic paracentesis performed; of the 108 patients undergoing therapeutic paracentesis, 44 (41%) were large volume. The mean volume removed during non‐large volume therapeutic paracentesis was 2.3 ± 1.2 L versus 7.3 ± 2.3 L for large volume paracentesis. A slight majority of patients received paracentesis at bedside (56%), with the remainder performed in the radiology suite; all procedures utilized ultrasound guidance for needle placement.

3.4. Univariate Analysis of Primary and Secondary Outcomes

AKI occurred in 14 (10%) patients; 11 cases were exclusively due to performance of the paracentesis and in 3 cases there were competing factors (Tables S3 and S4). Circulatory dysfunction was noted in 41 (28%) of patients; there was no association between the occurrence of these events (p = 0.54). The primary outcome occurred in 50 (35%) of patients. Mortality rates were 17% at 30 days and 26% at 90 days. There was no change in baseline serum creatinine post‐SBP episode across all patients within the cohort.

Univariate analysis performed on the primary (composite) outcome revealed no significant differences between groups (Table S5). Use of β‐blockers was not associated with higher rates of AKI or circulatory dysfunction, 30‐day mortality, 90‐day mortality, or change in baseline renal function (Table S6). The presence of ACLF (all grades) did not predict post‐paracentesis AKI or circulatory dysfunction but was associated with a higher rate of 30‐day mortality (23% vs. 10%, p = 0.042; Table S7), although not 90‐day mortality or change in baseline renal function. Patients in an ICU setting had higher 30‐day (44% vs. 13%, p = 0.005) and 90‐day (50% vs. 20%, p = 0.011) mortality compared to those on an acute care floor (Table S8). Patients administered broader antibiotics than third‐generation cephalosporins alone or alternative regimens (29% vs. 10%–14%, p = 0.047) were more likely to die within 30 days of SBP diagnosis (Table S9). Antibiotic duration did not affect mortality rates or long‐term renal function (Table S10).

3.5. Predictors of AKI or Circulatory Dysfunction

Separate models for volume removed (all patients, n = 144; outcome n = 53) and therapeutic paracentesis timing (excluded patients undergoing only diagnostic paracentesis, n = 108; outcome n = 40) were constructed using identical important clinical covariates, including the performing location of paracentesis, bacterial type, antibiotic type, CLIF‐OF score, and whether the infection was nosocomial. No variable was predictive of the primary outcome (Table 2). Sensitivity analyses excluding the three AKI cases with competing etiological factors did not alter these findings (Table S11).

TABLE 2.

Predictors of AKI or circulatory dysfunction in patients with SBP.

Variable	Adjusted OR (95% CI)	p
Volume analysis
Paracentesis volume
Diagnostic	Ref	Ref
Non‐large volume (< 5 L)	0.93 (0.37–2.33)	0.871
Large volume (≥ 5 L)	1.49 (0.56–3.99)	0.418
Bacterial type
No growth	Ref	Ref
Gram‐positive cocci	1.25 (0.47–3.35)	0.468
Gram‐negative rods	1.10 (0.41–2.99)	0.408
Infection type
Community‐acquired	Ref	Ref
Nosocomial	1.44 (0.62–3.40)	0.394
Antibiotic therapy
Third‐generation cephalosporin (or equivalent)	Ref	Ref
Third‐generation cephalosporin + vancomycin	1.55 (0.44–5.40)	0.494
Broad‐spectrum	0.95 (0.36–2.53)	0.921
CLIF‐OF score	0.81 (0.65–1.01)	0.064
Time analysis
Paracentesis timing
Early therapeutic	Ref	Ref
Delayed therapeutic	1.24 (0.49–3.13)	0.649
Bacterial type
No growth	Ref	Ref
Gram‐positive cocci	1.24 (0.40–3.82)	0.703
Gram‐negative rods	1.49 (0.48–4.63)	0.493
Infection type
Community‐acquired	Ref	Ref
Nosocomial	1.44 (0.52–3.98)	0.486
Antibiotic therapy
Third‐generation cephalosporin (or equivalent)	Ref	Ref
Third‐generation cephalosporin + vancomycin	1.11 (0.22–5.65)	0.899
Broad‐spectrum	0.91 (0.29–2.83)	0.875
CLIF‐C OF score	0.81 (0.62–1.04)	0.101

Note: Logistic regression analysis: volume analysis includes entire cohort (n = 144, 50 events); time analysis excludes patients who received diagnostic paracentesis alone (n = 108, 38 events).

Abbreviation: CLIF‐C OF, European Foundation for the Study of Chronic Liver Failure Organ Failure score.

3.6. Effect of Volume of Ascites Removed on Outcomes

Kernel regression demonstrated nonsignificant, nonlinear relationships between the volume of ascites removed and the primary outcome as well as 30‐day mortality and change in baseline renal function (Figure 2A–C). An inflection point around 8 L of ascites removed was noted with respect to increased probability of developing the primary outcome or 30‐day mortality.

FIGURE 2.

Relationship between the volume of ascites removed and clinical outcomes. A nonparametric Kernel regression smoother was utilized to show unadjusted associations and allows the probability to be determined locally rather than assuming a linear assumption. The blue dots represent actual data points and the red line shows the relationship between volume of ascites removed and probability of the outcome. Panel A shows the primary outcome, whereas Panels B and C demonstrate several secondary outcomes (one outlier at 15 L fluid removed is not shown).

3.7. Time‐to‐Event Analysis for Mortality

Therapeutic paracentesis timing was independent of 30‐day mortality (Figure S1A, p = 0.234) or 90‐day mortality on K‐M analysis (Figure S2A, p = 0.128). Competing risk analysis for 90‐day mortality did not demonstrate a significant effect of therapeutic paracentesis timing (Figure 3A, sHR 1.49, 95% CI 0.69–3.19, p = 0.308) but did show that nosocomial SBP increased the risk of mortality (Table S12, sHR 2.44, 95% CI 1.08–5.52, p = 0.032).

FIGURE 3.

Competing risk analysis for 90‐day mortality by timing of therapeutic paracentesis (A) and relative volume of paracentesis (B). Patients were censored at time of death, transplant, or end of follow‐up period. sHR comparing immediate therapeutic versus delayed therapeutic is 1.49 (95% CI 0.69–3.19); p‐value = 0.308 (A). sHR comparing LV therapeutic versus diagnostic is 2.40 (95% CI 0.87–6.60); p‐value = 0.089; sHR comparing NLV therapeutic versus diagnostic is 1.81 (95% CI 0.64–5.11); p‐value = 0.263 (B).

Volume of paracentesis performed did not associate with 30‐day mortality (Figure S1B, p = 0.117) or 90‐day mortality (Figure S2B, p = 0.295). When evaluating competing risks, the use of broad‐spectrum antibiotics was associated with 90‐day mortality (Table S13, sHR 2.28, 95% CI 1.05–4.93, p = 0.035). There was a nonsignificant trend of large volume paracentesis (Figure 3B, sHR 2.40, 95% CI 0.87–6.60, p = 0.089), higher CLIF‐OF score (sHR 1.17 per 1 unit increase, 95% CI 0.98–1.39, p = 0.086), and nosocomial SBP (sHR 1.93, 95% CI 0.90–4.11, p = 0.089) being associated with the outcome.

4. Discussion

There was no significant association between therapeutic paracentesis timing or volume of ascites removed on the development of AKI or circulatory dysfunction in patients with cirrhosis and SBP. The lack of association with therapeutic paracentesis and development of AKI is consistent with a recent retrospective cohort study of patients with cirrhosis with heterogeneous admitted diagnoses [17]. Given the absence of literature regarding the safety of therapeutic paracentesis specifically in the setting of active infection, our findings suggest it is reasonable to perform, if clinically indicated. Additionally, the dissociation between the development of AKI and intrinsic characteristics of the paracentesis itself supports current knowledge that AKI in SBP is unrelated to iatrogenic volume removal [18].

The pathophysiology of AKI in cirrhosis is primarily related to low effective arterial circulating volume. Splanchnic vasodilatation, with disproportionate blood pooling and reduced afferent renal arteriole inflow, leads to a compensatory upregulation of the renin‐angiotensin‐aldosterone system (RAAS). At baseline, patients with cirrhosis may still adequately perfuse their kidneys, but any perturbation in hemodynamic stability can disrupt this delicate balance [19]. While PICD could not be explicitly diagnosed in the present study given its retrospective nature and lack of routine PRA measurement, we chose to use a decrease in mean arterial pressure as an appropriate surrogate measure of PRA based on previously established correlation [20]. Our results showing no difference in AKI or circulatory dysfunction based on timing of paracentesis or volume category are consistent with prior literature noting a lack of association between PICD and these procedural variables [21]. However, it is important to note that our study population exclusively included patients with SBP, which has its own inherent increased risk of inducing AKI or circulatory dysfunction. Since PICD requires measurement of PRA on Day 6 post‐paracentesis and standard treatment of SBP includes volume expansion with albumin and antibiotics for at least 5 days, SBP treatment itself may blunt potential neurohormonal derangement.

Our findings are notable for a lack of definitive association between AKI risk and volume of ascites removed in patients with SBP. These findings contrast with prior literature, which describes a linear relationship such that each liter of ascites removed increases the risk of AKI by 24% [20]. However, the aforementioned study excluded patients with SBP and was performed in outpatients with decompensated cirrhosis. In contrast, we included critically ill patients receiving paracentesis in the ICU and patients with ACLF to broadly capture the spectrum of patients with decompensated cirrhosis who develop SBP. There is evidence that critically ill patients with tense ascites are at risk for abdominal compartment syndrome, which carries a poor prognosis and thus therapeutic paracentesis may be beneficial [22, 23]. Interestingly, we did not note higher rates of AKI or circulatory dysfunction in either the ICU or ACLF subgroups. There were higher rates of mortality in ICU patients, as well as those with ACLF, reflective of the expected prognosis and similar to published literature [24, 25]. Importantly though, mortality rates (inpatient, 30‐day, and 90‐day) were similar between paracentesis timing and volume groups as well as in procedures performed bedside or in the radiology suite. This lack of effect on short‐term outcomes is congruent with another moderate‐size retrospective cohort study [26]. Given there were no reports of nearby organ injury, it is unlikely the risk of AKI is tied to procedural complexity. Future prospective study of paracentesis in patients with SBP stratified by ICU status and ACLF grade is needed to better assess for between‐group differences.

While nearly all patients in the study received concentrated albumin infusion, the amount and timing of infusion were not standardized. However, the majority of patients (75%) received standard‐of‐care albumin dosing to prevent AKI and mortality as per the regimen published by Sort et al. Additionally, almost all patients (92%) had their infusion post‐paracentesis; a sizeable minority (38%) received albumin before and after paracentesis. Many patients had dual indications for albumin (SBP and therapeutic paracentesis) and the timing of albumin administration for each indication is potentially conflicting. Infusion for SBP is time dependent from diagnosis, whereas paracentesis‐related infusion is recommended after procedural completion although an optimal timeframe is not explicitly known. The heterogeneity of administration timing and dose in this study is likely related to its retrospective design. However, the low overall rates of AKI alone and short‐term mortality argue against differences in albumin administration being a significant confounder. Similarly, we investigated the effects of β‐blockers on outcomes in post hoc analysis given controversy surrounding their safety in patients with SBP. A prior large retrospective cohort study noted worse renal and survival outcomes in patients with SBP on β‐blockers, although the updated AASLD Practice Guidance recommends only holding them if hypotension is present [6, 27]. Patients on β‐blockers in our study were well distributed across key paracentesis subgroups and did not have higher rates of AKI or circulatory dysfunction, nor higher rates or short‐term mortality or change in baseline renal function. As the number of patients on β‐blocker therapy was only approximately a quarter of the study population, we cannot draw definitive conclusions related to their safety in SBP but these results appear to support the AASLD Practice Guidance.

Clinical care of patients in this cohort occurred before publication of the 2021 AASLD Practice Guidance which now explicitly recommends broad‐spectrum antibiotics for patients with nosocomial SBP [6]. In the 2012 AASLD Practice Guideline, which was in effect during the study timeframe, empiric nosocomial SBP‐related antibiotic therapy was recommended based on local susceptibility patterns for bacteria in patients with cirrhosis [28]. Given that institutional data did not routinely support broad‐spectrum antibiotics in these patients, some patients with nosocomial SBP still received empiric third‐generation cephalosporins (unless clinical or microbiologic data dictated broader spectrum antibiotics). Post hoc analysis demonstrated that patients with nosocomial SBP were less likely to have therapeutic paracentesis (non‐large or large) but did not have higher rates of AKI or circulatory dysfunction. Patients receiving third‐generation cephalosporins had lower rates of mortality compared to those receiving broader spectrum antibiotics, likely consistent with a lower baseline severity of illness. Antibiotic duration also did not associate with mortality or long‐term renal function. Thus, it does not appear that antibiotic characteristics played a significant confounding role in our study.

Interestingly, our study demonstrated a potential signal for increased risk of AKI or circulatory dysfunction, as well as 30‐day mortality after removal of at least 8 L of ascites. However, this finding likely represents an underpowered post hoc analysis (particularly given the local estimates were driven by relatively few occurrences). Additionally, any true relationship may simply represent a surrogate for underlying liver disease severity. When viewing the time‐to‐event curves for 90‐day mortality, there is separation starting around day 30, and competing risks analysis did not demonstrate a significant effect of paracentesis characteristics on mortality. Given that performance of therapeutic paracentesis in patients with SBP was not associated with a change in baseline serum creatinine, our findings reinforce the notion that paracentesis itself is unlikely to contribute to poor outcomes. Whether the underlying degree of portal hypertension may play a role in mortality in this patient population is unclear, as routine portal pressure measurement in this setting is not standard‐of‐care. Further study into the relationship between ascites burden, volume removed during paracentesis, and clinical outcomes in patients with SBP in a prospective fashion is warranted.

Our study should provide clinicians with additional reassurance that performance of early therapeutic paracentesis in the setting of SBP is not associated with poor clinical outcomes. Nonetheless, it is important to acknowledge several limitations. First, the retrospective and single‐center study design limits generalizability to all patients with cirrhosis. Second, whether patients receiving a diagnostic paracentesis only represented low ascites burden or clinician preference was not readily ascertainable during chart review. Third, the amount of ascites drained was at the discretion of the performing team and the procedure could have been performed prior to hepatology consultation. Fourth, there was also a lack of comprehensive objective measurements of renal function (urine output, creatinine clearance), circulatory function (PRA and aldosterone), and portal pressures. Lastly, the change in baseline renal function was impossible to assess in patients who died or were started and continued on dialysis, which reduced the sample size for this analysis and may not accurately reflect the true impact of paracentesis on long‐term renal outcomes in patients with SBP.

Concurrent performance of therapeutic paracentesis (with administration of guideline‐directed albumin) to alleviate large volume ascites in the setting of SBP does not negatively affect patient outcomes and may be performed immediately, if indicated. While prospective study is needed to confirm these results, underlying patient‐level characteristics, rather than procedural factors, may be the principal driver of poor outcomes in SBP.

Funding

The authors have nothing to report.

Ethics Statement

The study was approved by the University of Virginia Human Subjects Research IRB (HSR‐IRB #21910).

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: jgh370339‐sup‐0001‐Supinfo.docx.

Wentworth B. J., Shwetar M., Mallawaarachchi I., et al., “Performance of Early Therapeutic Paracentesis in Spontaneous Bacterial Peritonitis and Its Effect on Clinical Outcomes,” JGH Open 10, no. 2 (2026): e70339, 10.1002/jgh3.70339.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.