Baylor University Medical Center Proceedings July 2017 - 263

Statistical analysis was performed with GraphPad Prism 7
(GraphPad Software, San Diego, CA), with P < 0.05 used for
statistical signiﬁcance. Data are presented as mean ± standard error. Student's t test and Fischer's exact tests were used to compare
means and frequencies. Logistic regression was used to assess
the impact of laboratory and clinical variables on the outcome
of SBP or complications.
RESULTS
Of the 301 procedures, 185 were performed in men and
116 in women. Of the 301 ultrasound-guided paracenteses,
219 cases were associated with liver cirrhosis, 105 with cancer,
and 45 cases with both liver cirrhosis and cancer. Of the 301
paracenteses studied, 16 resulted in diagnoses of SBP (5%),
whereas 275 did not. Fifteen of the 16 cases were diagnosed
based on ANC ≥ 250 cells/mm3; the other case was diagnosed
based on a positive culture with Citrobacter freundii and had an
ANC of 123 cells/mm3. Of the remaining 15 cases, 13 had
negative cultures and 2 had positive cultures (Enterococcus
faecium and Citrobacter freundii). Among the 275 negative
cases, three had positive cultures considered contaminants (one
with Staphylococcus simulans and two with Propionibacterium
acnes).
Univariate analysis comparing positive and negative cases
demonstrated that positive cases signiﬁcantly diﬀered from negative cases in having decreased diagnosis of cirrhosis, increased
diagnosis of any type of cancer, more abdominal pain, a larger
depth of the largest ﬂuid pocket, increased rate of prior SBP,
and increased serum white blood cell count (Table). In contrast,
serum sodium, total ascites drained, the presence of fever or
confusion, gender, age, and indication for diagnostic evaluation
only did not signiﬁcantly diﬀer between groups. There was a

Table. Univariate analysis comparing cases with and without a
diagnosis of spontaneous bacterial peritonitis
No SBP
(n = 275)

SBP
(n = 16)

P
value

Abdominal pain

20%

75%

<0.001

Fever

2.5%

6.3%

0.36

Confusion

3.5%

6.3%

0.46

Cirrhosis

73%

31%

<0.001

Cancer

31%

69%

0.004

Diagnostic only

6.3%

0.61

Male gender

63%

38%

0.06

Age (years)

63.0 ± 0.7

61.9 ± 3.7

0.71

Prior SBP

19%

50%

0.008

Variable

Sodium

133.6 ± 0.7 135.8 ± 1.4

0.47

White blood cell count (cells/cc3)

6.5 ± 0.3

11.0 ± 1.5

<0.001

Depth of largest pocket (cm)

6.2 ± 0.2

7.7 ± 0.4

0.008

Total volume drained (L)

4.7 ± 0.2

4.6 ± 0.8

0.83

SBP indicates spontaneous bacterial peritonitis.

July 2017

positive correlation between the depth of the largest pocket
and total volume of ascites drained (r = 0.37, P < 0.0001).
Next, multiple logistic regression was performed to assess
the impact of the variables with signiﬁcant diﬀerences on univariate analysis (prior SBP, diagnosis of cirrhosis, diagnosis of
cancer, abdominal pain, depth of largest ﬂuid pocket, and serum
white blood cell count) on the diagnosis of SBP. Only two variables were independent predictors, speciﬁcally the depth of the
largest ﬂuid pocket (P = 0.008) and complaint of abdominal
pain (P = 0.006). A receiver-operator curve analysis of pocket
depth and SBP diagnosis had an area under the curve of 0.73
(P = 0.002); at a cutoﬀ of 5 cm, sensitivity was 100% and
speciﬁcity was 32%.
Of the 301 cases, there were two major adverse events
(0.67%), speciﬁcally hemorrhage requiring emergent laparotomy and hemorrhage resulting in death. A multiple logistic
regression was performed with all the variables assessed for SBP
plus platelet count and international normalized ratio (INR) as
input variables, and hemorrhagic complication as the outcome
variable. The resultant model was not signiﬁcant (P = 0.06)
and there were no independent predictors. The average platelet
count for cases with and without complications was 77.5 and
148.9 K/μL (P = 0.39). The average INR for cases with and
without complications was 1.2 and 1.3 (P = 0.39).
DISCUSSION
Multiple variables, including abdominal pain, cirrhosis,
cancer, prior SBP, white blood cell count, and amount of ascites, may be associated with SBP (9). The primary ﬁnding of
this study is that only abdominal pain and depth of the largest ﬂuid pocket were independent predictors of SBP. Though
abdominal pain is considered the hallmark of SBP (10), its
utility as a tool to predict a positive diagnosis of SBP is limited
by its low speciﬁcity, as accumulation of ascites alone can lead
to abdominal pain. Increased depth of ﬂuid pocket was associated with SBP. Despite the positive correlation between largest
pocket depth and total ascites volume drained, total volume
drained did not diﬀer between patients with and without SBP.
This may be because not all of the ﬂuid was removed in every
patient, or a certain threshold amount of ﬂuid is needed to
create conditions permissive of SBP. Depth of largest pocket is
an indirect estimate of the total volume of ascites and is readily
measurable with ultrasound and therefore is useful as a clinical
tool. Receiver-operator curve analysis and the ﬁnding that no
patient with a ﬂuid pocket <5 cm in depth had SBP suggest
that a 5-cm cutoﬀ may help to exclude certain patients with
low-volume ascites from the need for paracentesis, potentially
avoiding unnecessary exposure to procedure-related risks. This
patient population is of particular importance to interventional
radiologists, who are likely to be consulted in cases of smallvolume ascites necessitating ultrasound guidance.
In this study, clinical ﬁndings such as fever and confusion were not helpful in selecting patients for paracentesis.
Indeed, overall physician clinical impression carries a low
sensitivity of 76% and speciﬁcity of 34% (11). In our study,
none of the paracenteses performed for diagnostic purposes

Predictors of spontaneous bacterial peritonitis

263

Table of Contents for the Digital Edition of Baylor University Medical Center Proceedings July 2017