Diagnosis of infection plerula

1. 914
Rapid Diagnosis of Infectious Pleural Effusions by Use of Reagent Strips
Elie Azoulay,1
Muriel Fartoukh,3
Richard Galliot,4
Fre´de´ric Baud,4
Ge´rald Simonneau,3
Jean-Roger
Le Gall,1
Benoıˆt Schlemmer,1
and Sylvie Chevret2
1
Medical Intensive Care Unit and 2
Biostatistics Department, Saint
Louis Teaching Hospital, Paris 7 University, 3
Medical and Respiratory
Intensive Care Unit, Antoine Beclere Teaching Hospital, and 4
Medical
Intensive Care Unit, Lariboisiere Teaching Hospital, Paris, France
Reagent strips have not yet been tested for use in the diagnosis of infectious pleural effusions.
A reagent strip was used to evaluate 82 patients with pleural effusions: 20 patients had tran-
sudative effusions, 35 had infectious exudative effusions (empyema in 14 and parapneumonic
effusion in 21), and 27 had noninfectious exudative effusions. Pleural fluid protein, as eval-
uated by the reagent strip, proved accurate for the detection of exudative effusions (sensitivity,
93.1%; specificity, 50%; positive predictive value, 84.3%; negative predictive value, 71.5%; odds
ratio [OR], 6.77; and 95% confidence interval [CI], 1.87–24). The reagent strip leukocyte es-
terase test effectively detected infectious exudative effusions (sensitivity, 42.8%; specificity,
91.3%; positive predictive value, 88.2%; negative predictive value, 51.2%; OR, 4.46; and 95%
CI, 1.2–16.4). Pleural pH was significantly predicted by the reagent strip but was of no as-
sistance in categorization of exudative effusions as infectious or noninfectious. Compared
with physical, laboratory, and microbiological data, the reagent strip was as accurate for
estimation of percentages of infectious and noninfectious exudative effusions. Thus, reagent
strips may be a rapid, easy-to-use, and inexpensive technique for discriminating transudative
from exudative pleural effusions and for categorizing exudative pleural effusions as infectious
or noninfectious.
A pleural effusion must be characterized as transudative or
exudative and, if exudative, as infectious or noninfectious.
These distinctions are important for choosing the appropriate
management, such as therapy for heart failure, drainage of an
infectious exudative effusion, or microbiological documenta-
tion and adequate antibiotics for empyema. Exploratory thora-
centesis is classically indicated because, in combination with
blood tests, examination of pleural fluid frequently provides the
etiologic diagnosis [1].
Use of reagent strips has been proposed for the rapid di-
agnosis of meningitis, ascites, and urinary tract infections [2–5]
and has also been validated for determination of the pH of
pleural effusions [6, 7]. However, reagent strips have not been
tested for their accuracy in the determination of the levels of
protein and leukocyte esterase in pleural fluid, 2 parameters
that are of paramount importance for distinguishing transu-
dates from exudates and for determining whether an exudate
is due to an infection [8].
We conducted a prospective 3-center study to evaluate the
performance of reagent strips in the diagnosis of pleural infec-
Received 25 October 1999; revised 16 February 2000; electronically pub-
lished 20 October 2000.
The appropriate institutional review board approved the study, and all
patients gave their informed consent.
Reprints or correspondence: Dr. Elie Azoulay, Medical ICU, Saint Louis
Teaching Hospital, 1, avenue Claude Vellefaux, 75010 Paris, France (elie
.azoulay@sls.ap-hop-paris.fr).
Clinical Infectious Diseases 2000;31:914–9
᭧ 2000 by the Infectious Diseases Society of America. All rights reserved.
1058-4838/2000/3104-0010$03.00
tion in medical intensive care unit (ICU) patients with pleural
effusions. We evaluated the value of the strips for separating
transudative from exudative effusions and then for categorizing
exudative infections as infectious or noninfectious.
Patients and Methods
Patients. From 1 December 1997 to 1 December 1998, 3 teach-
ing-hospital medical ICUs prospectively included all patients who
had a pleural effusion at or after admission to the ICU and who
were deemed by the ICU physicians to require exploratory
thoracentesis.
Pleural effusion was diagnosed clinically on the basis of absent
breath sounds at auscultation, flatness to percussion, and reduced
tactile fremitus. A chest radiograph was obtained to confirm the
diagnosis. The decision to perform exploratory thoracentesis was
based on the criteria normally used at each of the study centers.
Pleural fluid was obtained by insertion of an 8-mm needle and was
sent to the appropriate laboratories for biochemical tests (for pH
and levels of lactate dehydrogenase, protein, and glucose), micro-
biological studies (identification of pathogens in smears and cul-
tures), and cytological studies (leukocyte and lymphocyte counts).
The pleural fluid was then tested, by use of a reagent strip designed
for the testing of urine (Multistix 8SG; Bayer, Leverkusen, Ger-
many), by an investigator who was unaware of the results of the
test described above.
Data collection. The following information was collected for
each patient. Epidemiological and clinical data included sex, age,
comorbidities (chronic heart failure, chronic obstructivepulmonary
disease, immunodepression, cirrhosis, and neutropenia), the reason
for admission to the medical ICU (acute respiratory or renal failure,
shock, or coma), the severity score at admission (simplified acute
Downloaded from https://academic.oup.com/cid/article-abstract/31/4/914/376234
by guest
on 18 March 2018

2. CID 2000;31 (October) Rapid Diagnosis of Pleural Infection 915
Table 1. Pathogens isolated from patients with microbiologically
documented pneumonia and from patients with empyema.
Documented infection
No. of
patients
Pneumonia
Branhamella catarrhalis 2
Enterobacter cloacae 1
Haemophilus influenzae 4
Proteus mirabilis 2
Pseudomonas aeruginosa 7
Staphylococcus aureus 3
Streptococcus pneumoniae 3
Streptococcus species 3
Total 25
Empyema
Escherichia coli 2
Klebsiella pneumoniae 2
Mycobacterium tuberculosis 2
Peptostreptococcus species 2
Staphylococcus aureus 3
Streptococcus species 3
Total 14
physiological score II [SAPSII] [9]), and presence or absence of
clinically or microbiologically documented infection at the time of
thoracentesis. Characteristics of the pleural effusion included bio-
chemical and cytological test findings (fluid/serum [F/S] ratios for
protein and lactate dehydrogenase, glucose level, leukocyte count,
and percentage of neutrophils); findings from Gram-stained smears
and cultures for bacteria, mycobacteria, and fungi; and results from
the reagent strip (protein and leukocyte esterase measurements and
pH). The cause of the effusion was determined on the basis of
clinical findings and results of laboratory tests of the pleural fluid
and serum. Reagent strip results were read by use of a colorimeter
(Clinitek 50; Bayer [10]) and were expressed on a 6-grade scale
(grades 0–5), for the protein and leukocyte esterase levels, and as
the value for the pH (range, 6–8.50, with intervals of 0.50). All 3
reagent strip results were assigned and collected by an investigator
who was unaware of the results of the other tests.
Criteria for the etiologic diagnosis of pleural effusion. Three
categories of pleural effusion were defined [8]. Transudative effu-
sion was defined as an F/S protein ratio !0.5; infectious exudative
effusion was defined as an F/S protein ratio у0.5 and either a
positive bacteriologic culture (empyema) or clinical or microbio-
logical evidence of pneumonia (parapneumonic effusion); and non-
infectious exudative effusion was defined as an F/S protein ratio
10.5 with negative bacteriologic cultures and no evidence of pneu-
monia.
Statistical analysis. The categories of patients defined above
were considered. Transudative effusions were compared with ex-
udative effusions, and infectious exudative effusions were com-
pared with noninfectious exudative effusions. These comparisons
were made by use of the x2
test or Fisher exact test, for categorical
variables, and by use of the Wilcoxon test or the Kruskal-Wallis
test, for continuous variables. Regression splines were used to ob-
tain a nonparametric estimate of the cutoff of continuous covar-
iates influencing the risk of each effusion category. A multivariate
logistic regression model was then constructed to determine the
OR of each reagent strip result for predicting whether any effusion
was exudative and whether an exudative effusion was infectious.
The percentage of each diagnosis established on the basis of con-
ventional tests was used as the reference and is hereafter referred
to as the “observed” percentage, as opposed to the “predicted”
percentage, which was established on the basis of reagent strip
findings. Continuous variables were dichotomized according to the
spline regression results. Values for Hosmer-Lemeshow goodness
of fit were computed, as were calibration curves (observed vs. pre-
dicted rates of each event). The SAS software package (SAS, Cary,
NC) was used for all statistical evaluations.
Results
Patient characteristics. From 1 December 1997 to 1 De-
cember 1998, 82 patients admitted to the 3 medical ICUs par-
ticipating in the study underwent thoracentesis for evaluation
of a pleural effusion.
The patients were 42 men and 40 women with a median age
of 59.5 years (25th–75th percentile, 42.5–69.5 years) and a me-
dian SAPSII score of 46 (25th–75th percentile, 27.7–46.5).
Seven patients (8.5%) were HIV infected, 18 (22%) had im-
munodepression resulting from other causes, and 8 (10%) were
receiving steroid therapy (daily administration of 11 mg of
prednisone or equivalent per kg for the past 30 days). Fourteen
patients (17%) had diabetes, 29 (35%) were alcoholics, 19 (23%)
had chronic obstructive pulmonary disease, and 13 (16%) had
documented chronic heart failure. The reasons for admission
to the ICU (patients could have more than 1 reason) and the
number and percentage of patients with each reason were as
follows: acute respiratory failure (67 patients [82%]), decom-
pensation of chronic respiratory failure (10 [12%]), shock (32
[39%]), cardiogenic pulmonary edema (19 [23%]), acute renal
failure (14 [17%]), coma (11 [13.5%]), and pulmonary embolism
(7 [8.5%]). Forty-five patients (55%) required mechanical ven-
tilation (of these, 33 [40%] had a positive end-expiratory pres-
sure of 15 cm H2O), 35 (42.5%) required vasopressor agents,
and 6 (7.5%) required dialysis.
On average, admission to the ICU occurred 1 day after ad-
mission to the hospital (25th–75th percentile, 0–7 days), and
exploratory thoracentesis was done within 2 days (25th–75th
percentile, 0–7 days) after admission to the ICU. The median
duration of stay in the ICU was 11 days (25th–75th percentile,
6–19 days), and the mortality rate in the ICU was 35% (29
deaths).
On the day of thoracentesis, 47 patients (57%) were taking
antibiotics and 28 (34%) were taking diuretics. Clinical evidence
of infection was present in 55 patients (67%), including 45 with
pneumonia, 3 with digestive infection, 2 with bacteremia, 2 with
skin and soft tissue infections, 1 with meningitis, 1 with urinary
tract infection, and 1 with catheter-related infection. Microbi-
ological documentation of an infection was obtained for 32 of
these patients (39%), including 25 of those with pneumonia
(table 1). Before thoracentesis, 5 pleural effusions were con-
firmed by chest sonogram and 3 were confirmed by CT scans;
for all other effusions, thoracentesis was done on the basis of
Downloaded from https://academic.oup.com/cid/article-abstract/31/4/914/376234
by guest
on 18 March 2018

3. 916 Azoulay et al. CID 2000;31 (October)
Table 2. Results of conventional laboratory tests and of reagent strip testing of pleural fluid.
Test results
Type of pleural effusion
a
P
b
Transudative
( )n p 20
Infectious exudative
( )n p 35
Noninfectious exudative
( )n p 27
Conventional testing, median (25th–75th percentile)
pH 7.76 (7.54–8.15) 7.50 (7.39–7.94) 7.81 (7.54–8.00) .03
Fluid/serum protein ratio 0.29 (0.25–0.38) 0.58 (0.47–0.74) 0.57 (0.42–0.64) !.0001
Fluid/serum LDH ratio 0.46 (0.33–1.01) 2.41 (1.06–6.6) 0.73 (0.45–1.59) !.0001
Leukocyte count 131 (42–267) 1020 (100–2050) 250 (92–467) .005
Percentage of neutrophils 25 (11–40) 80 (63–85) 9 (1.7–20) !.0001
Reagent strip testing
c
, no. (%)
Protein grade 13 10 (50) 33 (94.3) 21 (77.7) .0006
Leukocytes grade 12 0 (0) 15 (42.8) 2 (7.4) .003
NOTE. LDH, lactate dehydrogenase.
a
According to the criteria of Light et al. [8].
b
By use of the Kruskal-Wallis test, for conventional findings, and by use of the x2
test, for reagent strip findings with the null
hypothesis that distribution of each variable was identical in the 3 groups.
c
Results not available for 4 patients with hemothorax.
clinical examination. The only adverse event related to thora-
centesis was pneumothorax, which occurred in 6 patients (7%).
Characteristics of the pleural effusion. Among the 82 ef-
fusions, 20 (24.4%) were transudates, 35 (42.7%) were infectious
exudative effusions associated with pneumonia (21 effusions)
or empyema (14 effusions) (table 1), and 27 (32.9%) were non-
infectious exudative effusions (of which 10 were malignancies,
4 were pulmonary embolisms, 4 were postoperative effusions
following upper abdominal surgery, 4 were hemothoraxes, 3
were effusions of unknown origin, and 2 were pancreatitis).
Gram-staining was positive for 6 (42.8%) of the 14 empyemas.
Table 2 reports the main findings from conventional pleural
fluid tests in the 3 patient categories as well as the results of
reagent strip testing (which was not performed for the 4 patients
with hemothorax). With the reagent strip, grade 3 was the pro-
tein level cutoff that distinguished transudative from exudative
effusions, and grade 2 was the leukocyte esterase cutoff that
distinguished infectious exudative from noninfectiousexudative
effusions (figure 1). Figure 2 displays the weak correlations
between pleural fluid pH, protein level, and leukocyte count,
either by use of the reagent strip or by use of a conventional
laboratory technique. Pleural fluid glucose levels were signifi-
cantly decreased only in the 14 patients with empyema (median,
1.7 mM/L [25th–75th percentile, 1.2–4.4 mM/L] vs. 7.9 mM/L
[25th–75th percentile, 5.5–10.1 mM/L] in patients with para-
pneumonic effusions [ ]; median, 6.4 mM/L [25th–75thP p .02
percentile, 5.1–9.2 mM/L] in patients with other effusions
[ ]). Moreover, all patients with a positive Gram stainP p .009
had a leukocyte esterase test result of grade у2. Otherwise, as
shown in table 2, pH was significantly lower in infectious ex-
udative effusions than in transudative or noninfectious exu-
dative effusions. Moreover, in patients with empyema, pH was
significantly lower than in other patients, both according to
laboratory findings (median pH, 7.46 [25th–75th percentile,
7.27–7.50] vs. 7.77 [25th–75th percentile, 7.50–8.00]; )P p .004
and according to results of reagent strips. There were 3 patients
with a pH of 7.00 and 8 patients with a pH of 7.50 among the
14 patients with empyema ( ).P p .03
Multivariate analysis. As shown in table 3, among the re-
agent strip findings, only the protein grade discriminated sig-
nificantly between exudates and transudates (sensitivity, 93.1%;
specificity, 50%; positive predictive value, 84.3%; negative pre-
dictive value, 71.5%; OR, 6.77; and 95% CI, 1.87–24). Similarly,
among the 58 patients with exudates, only a positive reagent
strip test for leukocyte esterase was significantly predictive of
infection (sensitivity, 42.8%; specificity, 91.3%; positive predic-
tive value, 88.2%; negative predictive value, 51.2%; OR, 4.46;
and 95% CI, 1.2–16.4).
Figure 3 compares the percentages of exudative effusions and
infectious effusions determined by use of conventional tests
(observed percentage) and by use of reagent strip testing (pre-
dicted percentage). Regardless of the type of effusion consid-
ered, it appears that rates of exudative or infectious effusion
observed by use of conventional laboratory tests are close to
those predicted by the results of reagent strip testing (P p
and , respectively, according to the MacNemar test)..32 P p .27
Discussion
Our 3-center prospective study compared the results of pleu-
ral fluid testing done by means of conventional methods with
those obtained by use of reagent strip testing among 82 ICU
patients with transudative, infectious exudative, or noninfec-
tious exudative pleural effusions. This classification of pleural
effusions is a valuable aid in the selection of the most appro-
priate management (e.g., treatment for heart failure or change
in antibiotic therapy and pleural drainage). Results of reagent
strip testing were not available for the 4 patients with hemo-
thorax. Data from the remaining 78 patients showed that pleu-
ral fluid protein level, as evaluated by use of the reagent strip,
was effective in discriminating between exudative and transu-
dative effusions (OR, 6.77). Moreover, in the 62 patients with
exudates, the reagent strip leukocyte esterase result, which was
available for 58 patients, accurately identified the cases of in-
fection (OR, 4.46). Among the patients with exudative effu-
sions, the percentages of infectious and noninfectious effusions
Downloaded from https://academic.oup.com/cid/article-abstract/31/4/914/376234
by guest
on 18 March 2018

4. CID 2000;31 (October) Rapid Diagnosis of Pleural Infection 917
Figure 1. Nonparametric spline estimate (thick line; [thin lines de-
note 95% CI]) of regression function f that models influence on prob-
ability of detection of exudative effusions (top) or of discrimination of
infectious from noninfectious effusions (bottom) by use of grades for
protein or leukocyte esterase measurements obtained by reagent strip
testing.
Figure 2. Comparison of pleural fluid findings from conventional
laboratory tests (fluid/serum protein ratio, neutrophil percentage, and
pH) and from reagent strip testing (protein and leukocyte esterase level
and pH). Horizontal lines in each box plot represent (bottom to top)
10th, 25th, 50th (median), 75th, and 90th percentiles. Such plots allow
for comparison of distribution of each variable (y axes) among groups
defined by results of reagent strip testing (x axes). Distributions of
conventional laboratory tests were compared with categorical results
of reagent strip testing, by use of nonparametric Wilcoxon (A and
B) or Kruskal-Wallis (C) tests. *P ! .05.
predicted on the basis of the results of reagent strip testing were
very close to those determined on the basis of conventional
laboratory tests.
To our knowledge, reagent strips have not yet been tested
for their ability to identify a pleural effusion as an exudate or
as a manifestation of infection. Reagent strips have, however,
been validated for the determination of pleural fluid pH [6, 7].
In our study, pleural pH was not an independent predictor for
infection, but it was significantly lower for infectious effusions
than for noninfectious effusions, particularly in the case of em-
pyema. As suggested by other studies, low pH values, detected
by use of either reagent strip tests or laboratory tests, could be
used as an aid when chest tube drainage is discussed [11, 12].
Urine test strips accurately estimated the pleural protein
grade: 54 (84.3%) of the 64 patients with a reagent strip pleural
protein grade 13 had an exudate, and 54 (93.1%) of the 58
patients with exudates had a reagent strip pleural protein grade
13. Nevertheless, 10 of the patients with transudates also had
a protein grade 13. Of these 10 patients, 1 had congestive heart
failure with a bilateral pleural effusion and an F/S ratio for
protein of 0.75 because of pleural amyloidosis. In 4 patients
with hemostasis abnormalities, the pleural fluid was blood-
stained and consequently produced a reagent strip proteingrade
13.
Reagent strip testing for leukocyte esterase has been found
useful in the diagnosis of infections of many body fluids but
has not yet been tested for pleural effusions [2–4]. Of the 17
patients in our study who had a reagent strip leukocyte esterase
grade у2, 15 (88.2%) had an infectious exudative effusion. Fif-
teen (42.8%) of the 35 patients with an infectious exudative
effusion had a reagent strip leukocyte esterase grade у2, ac-
counting for the same sensitivity rate as the Gram stain. The
Downloaded from https://academic.oup.com/cid/article-abstract/31/4/914/376234
by guest
on 18 March 2018

5. 918 Azoulay et al. CID 2000;31 (October)
Figure 3. Percentages of cases of exudative effusion (top) and infectious effusion (bottom) determined by results of reagent strip tests (predicted
percentage [denoted by the unshaded bars) and results of conventional laboratory tests (observed percentage [denoted by the shaded bars). L,
reagent strip leukocyte esterase grade (cutoff, grade 2); P, reagent strip protein grade (cutoff, grade 3).
grade was у2 for only 2 patients (8.7%) with noninfectious
exudative effusions and for none of the patients with transu-
dative effusions. Nevertheless, 20 patients with infectious ex-
udative effusions had a leukocyte esterase grade !2, a finding
perhaps ascribable in part to the fact that 2 patients had pleural
tuberculosis with a predominance of lymphocytes in their ef-
fusion and that most of the remaining patients were already
taking antibiotic therapy at the time of the thoracentesis.
The percentages of exudative effusions (infectious or non-
infectious) and of infectious exudative effusions predicted on
the basis of protein and leukocyte esterase findings from reagent
strip tests were very close to those determined on the basis of
conventional tests. This finding suggests that, in patients with
exudates, reagent strip leukocyte esterase testing may be val-
uable for providing rapid diagnostic orientation and for iden-
tifying pleural fluid specimens that require particular attention
from laboratory technicians because of an increased likelihood
of disease. Moreover, all patients with a positive result of Gram
staining of pleural fluid had results of grade у2 with the leu-
kocyte esterase test, and the sensitivities of these 2 tests are
equal [13]. Therefore, results of reagent strips could help ICU
physicians to maintain a high index of suspicion regarding in-
fectious pleural effusion and to alert the laboratories to shorten
the turnaround time for conventional laboratory tests to allow
for an adequate antibiotic treatment on the basis of the Gram
stain. Furthermore, reagent strips may provide sound infor-
mation on which to base immediate treatment decisions (e.g.,
whether to use antibiotics) in places where technical and/or
economic constraints reduce the availability of laboratory test-
ing, or before the results of conventional tests are available.
The epidemiology of pleural effusions in patients in medical
ICUs remains poorly known. Of the 82 patients included in
our 3-center prospective study, only 20 (24.4%) had transu-
dative effusions. In contrast, in a study that did not use a need
for thoracentesis as an inclusion criterion, Mattison et al. [14]
found that fluid overload was by far the most common cause
of pleural effusion. In our patients, infection was the leading
cause, and the bacteriologic profile was that usually observed
in community- or hospital-acquired infections in patients in
ICUs. Moreover, although many pleural fluid components have
been suggested as useful for the differentiation of exudates from
transudates, the tools widely used to determine the cause of a
pleural effusion remain the protein level, the F/S protein ratio,
and the leukocyte count [15, 16]. Our results suggest that re-
agent strips accurately predict the results of these conventional
tests and may deserve to be recommended as an ancillary tool
in the diagnostic assessment of potentially infectious pleural
effusions.
Valuable insights into the significance of pleural effusions in
patients in medical ICUs (i.e., as a symptom or syndrome) could
be obtained by performing exploratory thoracentesis, a pro-
cedure that has been proved safe [17], followed by examination
Downloaded from https://academic.oup.com/cid/article-abstract/31/4/914/376234
by guest
on 18 March 2018

6. CID 2000;31 (October) Rapid Diagnosis of Pleural Infection 919
Table 3. Multivariate analysis of exudative effusions and infection
by use of 2 models.
Model and reagent strip result
No. of
patients OR (95% CI) P
Prediction of presence of exudate in 78 pa-
tients with available reagent strip results
Protein grade 13 64 6.77 (1.87–24) .003
Leukocyte grade 12 17 6.48 (0.69–60) .10
Glucose grade !2 39 0.72 (0.17–2.9) .64
pH !7.50 10 0.98 (0.3–3.2) .98
Prediction of presence of pleural infection in
58 patients with exudative effusions
Protein grade 13 54 0.38 (0.06–2.4) .32
Leukocyte grade 12 17 4.46 (1.2–16.4) .02
Glucose grade !2 30 0.93 (0.21–4) .92
pH !7.50 9 0.79 (0.25–2.4) .67
NOTE. , by the Hosmer-Lemeshow test.2
P 1 .05 x
of the fluid at the bedside by use of reagent strips. Further
studies in a large number of patients are needed to determine
the value of this rapid, easy-to-use, and inexpensive tool for
identification of pleural fluids that do not require more expen-
sive tests, notably the impact of its use on patient outcome.
References
1. Light RW. Diagnostic principles in pleural disease. Eur Respir J 1997;10:
476–81.
2. Moosa AA, Quortum HA, Ibrahim MD. Rapid diagnosis of bacterial men-
ingitis with reagent strips. Lancet 1995;345:1290–1.
3. DeLozier JS, Auerbach PS. The leukocyte esterase test for detection of cere-
brospinal fluid leukocytosis and bacterial meningitis. Ann Emerg Med
1989;18:1191–8.
4. Smalley DL, Doyle VR, Duckworth JK. Correlation of leukocyte esterase
detection and the presence of leukocytes in body fluids. Am J Med Technol
1982;48:135–7.
5. Levy M, Tournot F, Muller C, Carbon C, Yeni P. Evaluation of screening
tests for urinary infection in hospital patients. Lancet 1989;2:384–5.
6. Cheng DS, Rodriguez RM, Rogers J, Wagster M, Starnes DL, Light RW.
Comparison of pleural fluid pH values obtained using blood gas machine,
pH meter, and pH indicator strip. Chest 1998;114:1368–72.
7. Byrd RP Jr, Roy TM. Pleural fluid pH determination. Chest1998;113:1426–7.
8. Light RW, MacGregor I, Luchsinger PC, et al. Pleural effusion: the diagnostic
separation of transudates and exudates. Ann Intern Med 1972;77:507–13.
9. Le Gall JR, Lemeshow S, Saulnier F. A new simplified acute physiology score
(SAPS II) based on a European–North American multicenter study.
JAMA 1993;270:2957–63.
10. Pugia MJ, Lott JA, Luke KE, Shihabi ZK, Wians FH Jr, Phillips L. Com-
parison of instrument-read dipsticks for albumin and creatinine in urine
with visual results and quantitative methods. J Clin Lab Anal 1998;12:
280–4.
11. Heffner JE, Brown LK, Barbieri C, DeLeo JM. Pleural fluid chemical analysis
in parapneumonic effusions: a meta-analysis. Am J Respir Crit Care Med
1995;151:1700–8.
12. Light RW. Parapneumonic effusions and empyema. Clin Chest Med 1985;6:
55–62.
13. Ferrer A, Osset J, Alegre J, et al. Prospective clinical and microbiological
study of pleural effusions. Eur J Clin Microbiol Infect Dis 1999;18:237–41.
14. Mattison LE, Coppage L, Alderman DF, Herlong JO, Sahn SA. Pleural
effusions in the medical ICU: prevalence, causes, and clinical implications.
Chest 1997;111:1018–23.
15. Kinasewitz GT. Transudative effusions. Eur Respir J 1997;10:714–8.
16. Heffner JE, Brown LK, Barbieri CA. Diagnostic value of tests that discrim-
inate between exudative and transudative effusions. Chest 1997;111:
970–80.
17. Colt HG, Brewer N, Barbur E. Evaluation of patient-related and procedure-
related factors contributing to pneumothorax following thoracentesis.
Chest 1999;116:134–8.
Downloaded from https://academic.oup.com/cid/article-abstract/31/4/914/376234
by guest
on 18 March 2018