Changes in Ascitic Fluid Polymorphonuclear Cell Count After Antibiotics Are Associated With Mortality in Spontaneous Bacterial Peritonitis

Spontaneous bacterial peritonitis (SBP) is a feared complication of ascites that affects 10%–30% of hospitalized patients with cirrhosis with an associated mortality rate of approximately 20%.^1–3 Although efforts have been undertaken to encourage prompt evaluation and treatment of SBP, outcomes have generally remained dismal.³ There is significant interest in identifying factors that can reliably predict mortality among individuals with SBP.

Recent studies have demonstrated that increased age, nosocomial infections, high model for end-stage liver disease (MELD) scores, positive ascitic fluid cultures, and the initial use of broad-spectrum antibiotics are associated with increased in-hospital or 30-day mortality in SBP.^4,5 However, a rise in the ascitic fluid polymorphonuclear (PMN) cell count after 2 days of treatment was also shown to predict in-hospital death, and early studies demonstrated that, with successful antibiotic therapy, ascitic fluid PMN cell count is expected to decrease by 70%–75% at 48 hours.^5–7 Although guidelines recommend reassessing the ascitic fluid PMN cell count approximately 2 days after starting antimicrobial therapy, this continues to be a controversial practice.⁸

We aimed to evaluate the association between changes in PMN cell count after antibiotic initiation and mortality in patients with SBP. We hypothesized that individuals without a marked decrease in ascitic fluid PMN cell count after 2 days would have increased odds for death after adjusting for liver disease severity and other relevant factors.

Electronic medical records at 3 tertiary centers (Yale-New Haven Hospital, New York-Presbyterian Hospital/Columbia University, and Northwestern Memorial Hospital) were queried between 2006 and 2016 to identify hospitalized adults with cirrhosis who had diagnoses of SBP based on an ascitic fluid PMN cell count ≥250 cells/mL, and had follow-up paracenteses after 2 days of antimicrobial therapy. Patients with suspected secondary bacterial peritonitis were excluded as were patients who were being treated in the intensive care unit at the time of diagnosis.

Our primary outcome was in-hospital mortality. Receiver operating characteristic curve analysis was used to determine the optimal cutoff point for change in PMN cell count at 2 days. Logistic regression analysis was performed to assess the impact of relevant factors, including age, gender, nosocomial infection, cirrhosis etiology, prior diagnosis of SBP, white blood cell count, MELD-sodium (MELD-Na) score, albumin, initial absolute PMN cell count, relative change in PMN cell count, positive ascitic fluid cultures, and choice of initial antibiotics. Covariates with P < .10 in univariable models were included in the multivariable model, and our analysis was stratified based on institution (Yale/Columbia vs Northwestern). All analyses were performed in R (R Core Team, 2019). Additional methodology is provided in the Supplementary Methods.

Our study was comprised of 426 patients; the cohort had markers of advanced cirrhosis (median MELD-Na, 28; interquartile range, 11) and organ dysfunction (median creatinine, 1.5; interquartile range, 1.6). Approximately 40% had alcohol-related cirrhosis, and 15% had a prior episode of SBP. Ascitic fluid cultures were positive in 30% of cases, and 42% initially received broad-spectrum antibiotics (Table 1).

Table 1.

Patient Characteristics

	Yale			Columbia			Northwestern			Combined
	n, median	IQR	%	n, median	IQR	%	n, median	IQR	%	n, median	IQR	%
Demographics
Number of patients	96			115			215			426
Age, y	54	13		57	13		57	14		57	14
Sex (male)	75		78	79		69	135		63	289		68
Nosocomial	28		29	28		24	33		15	89		21
Length of stay (days)	15	21		13	14		9	9		10	13
Serum laboratory values
Sodium, mmol/L	131	8		132	5		132	7		132	7
Creatinine, μg/d	1.6	2.0		1.6	1.7		1.5	1.4		1.5	1.6
AST, U/L	64	60		55	63		60	51		60	57
ALT, U/L	33	39		28	26		33	30		32	32
Bilirubin, mg/dL	7.0	9.4		4.1	8.4		5.0	8.0		5.4	8.3
INR	1.8	0.8		1.7	0.9		1.7	0.8		1.7	0.8
Albumin, g/dL	2.7	0.8		2.9	1.0		2.5	0.8		2.7	0.8
WBC, 109 cells/L	9.6	9.2		9.0	9.0		9.3	7.9		9.4	8.4
Cirrhosis etiology and severity
Alcohol-related	55		57	50		43	65		30	170		40
Prior SBP	15		16	25		22	23		11	63		15
MELD-Na	29	11		28	11		28	11		28	11
Ascitic fluid studies
Initial PMN count, cells/μL	985	2554		1070	2830		1343	2828		1132	2831
Day 2 PMN count, cells/μL	340	1092		378	1126		231	820		295	940
Positive culture	35		36	32		28	61		28	128		30
Antibiotics
Broad-spectrum	51		53	61		53	65		30	177		42

ALT, alanine aminotransferase; AST, aspartate aminotransferase; INR, international normalized ratio; IQR, interquartile range; MELD-Na, model for end-stage liver disease–sodium; PMN, polymorphonuclear; SBP, spontaneous bacterial peritonitis; WBC, white blood cells.

An association between changes in PMN cell count after 2 days and in-hospital mortality was observed (Supplementary Table 1); a cutoff value of approximately 80% reduction in PMN cell count was most predictive of survival. All factors associated with in-hospital mortality on univariable logistic regression analysis were subsequently included in a multivariable model. A reduction in PMN cell count >80% was associated with improved survival (adjusted odds ratio, 0.32; 95% confidence interval, 0.17–0.58; P < .001), whereas increasing age, higher MELD-Na scores, positive ascitic fluid cultures, and broad-spectrum antibiotics were associated with increased mortality (Supplementary Table 2).

The combination of MELD-Na scores and 2-day changes in PMN cell count using a cutoff of 80% accurately risk-stratified patients in our cohort. High-risk patients (MELD-Na score >30 without a reduction in PMN cell count >80%) had a mortality rate of 43% (43/101). In contrast, low-risk patients (MELD-Na score ≤30 with a reduction in PMN cell count >80%) had a mortality rate of less than 6% (7/123), representing more than a 7-fold reduction. Intermediate-risk patients (MELD-Na score >30 with a reduction in PMN cell count >80% or MELD-Na score ≤30 without a reduction in PMN cell count >80%) had a mortality rate of 17%–20%.

Similar to recent studies, our analysis suggests that, among predominantly noncritically-ill hospitalized patients with SBP, those with older age, higher MELD-Na scores, positive ascitic fluid cultures, and initial broad-spectrum antibiotic use have increased in-hospital mortality.^4–5 However, changes in the PMN cell count over the first 2 days also carry important prognostic value, and patients who experience reductions >80% have improved survival.

Hospitalized patients should be risk-stratified based on their MELD-Na scores and the interval change in PMN cell count. For intermediate- and high-risk patients, additional studies are required to determine whether changes in antimicrobial therapy or other treatment considerations can improve outcomes. For remaining patients, prognosis is expected to be favorable, and standard treatment should be maintained.

Major strengths of our study are its multicenter design and relatively large sample size. Limitations include variability in clinical practice among institutions, especially relating to antibiotic choices and the decision to obtain follow-up paracentesis, and a lack of other potentially important parameters that could be of prognostic value. Furthermore, although our study identifies a novel strategy for risk-stratifying patients with SBP, our algorithm depends on paracentesis findings obtained 2 days after treatment initiation rather than findings immediately available at the time of diagnosis, potentially delaying changes in management.

In conclusion, we highlight that repeating paracentesis 2 days after the initiation of antibiotics can inform the prognosis for hospitalized patients with SBP. Future studies should seek to identify modifiable factors that predict response early in the treatment course and explore management strategies that improve survival for intermediate- and high-risk patients.