|
Background: This investigation evaluates the feasibility
of using C-reactive protein (CRP) levels as an indicator of
bacterial infection of adult patients in the Emergency Department
(ED), by comparing them with clinical signs and routine laboratory
tests.
Methods: One hundred and fifty adult atraumatic patients admitted
through the ED of Linkou Chang Gung Memorial Hospital were
consecutively enrolled. Seventy-nine patients had documented
infection, and 58 had no infection. Body temperature (BT),
white blood cell (WBC) count, CRP levels, and the presence
of systemic inflammatory response syndrome (SIRS) were compared
between the infected and uninfected groups.
Results: SIRS was the most sensitive indicator of bacterial
infection (sensitivity 84.8%), but it had a 37.9% false-positive
rate. BT and WBC count were more specific (at 89.7% and 84.5%)
but less sensitive (at 48.1% and 43.0%, respectively). Using
Youden's Index, the best cut-off value for CRP was 60 mg/l
(sensitivity 67.1%, specificity 94.8%, positive predictive
value 94.6%, and negative predictive value 67.9%). The area
under the receiver operating characteristics (ROC) curve was
highest for CRP (at 0.88), followed by BT (at 0.77) and WBC
(at 0.67) (all p<0.05).
Conclusion: CRP is a better indicator of bacterial infection
than either BT or WBC count for adult atraumatic ED patients.
A low serum CRP level cannot safely be used to exclude the
presence of infection.
(Chang Gung Med J 2002;25:437-45)
Key words: C-reactive protein, bacterial infection, emergency
department, sepsis, systemic inflammatory response syndrome.
Patients presenting with fever, tachycardia, hyperventilation,
and leukocytosis with no known stimulus such as trauma, burns,
or pancreatitis are usually presumed to have sepsis, a typical
body response to infection.(1) Sepsis is a major cause of
morbidity and mortality, and early institution of an appropriate
antimicrobial regimen is associated with improved survival.(2)
Clinical signs of infection and routine laboratory tests are,
however, not specific and can be misleading. Infection can
be present in some patients without sepsis, particularly in
the debilitated and elderly. Acutely ill patients, however,
frequently present signs of sepsis, even when no bacterial
infection can be demonstrated. The widespread use of antibiotics
for all these patients presents problems of antibiotic resistance,
drug toxicity, and increased medical costs. Identifying patients
who are likely to benefit from an antimicrobial agent is a
priority. Since sepsis response involves the release of a
wide array of inflammatory mediators,(3) it has been suggested
that some of these mediators could be used as markers of infection
or sepsis severity.(4,5) As these mediators may be increased
in other inflammatory conditions, none is specific for infection.
C-reactive protein (CRP), in addition to the white blood cell
(WBC) count, is currently the most widely used parameter to
support a diagnosis of infection. CRP is an acute-phase reactant
produced by the liver. Plasma concentrations are normally
below 10 mg/l but increase several fold after trauma, infection,
inflammation, and other stimuli involving tissue damage.(6)
Interleukin (IL)-6 is thought to be the main mediator stimulating
CRP production, but other cytokines, like IL-1 and tumor necrosis
factor, are also involved.(7,8) CRP may be useful in diagnosing
the onset of sepsis in acutely ill patients(9,10) and for
indicating successful treatment during follow-up of the clinical
course.(11) Although large increases can occur in response
to infection, no definite correlation between infection and
changes in CRP has been documented,(9) and using a low CRP
level to exclude the presence of infection remains controversial.
This investigation evaluated the feasibility of using CRP
levels, as compared to clinical signs of infection and routine
laboratory tests, for detecting bacterial infection in the
Emergency Department (ED).
METHODS
One hundred and fifty adult atraumatic patients admitted
through the ED of Linkou Chang Gung Memorial Hospital, Taiwan
from May 14 to 16, 2001 were consecutively enrolled. Those
who were death-on-arrival and those who were referred from
wards or intensive care units (ICUs) of other hospitals were
not included. All patients were examined for signs and symptoms
of infection on ED admission. Clinical and laboratory data
collected included age, gender, admission diagnosis, patient
disposition, body temperature (BT), WBC count, CRP levels,
and other available information required for the calculation
of the Acute Physiology and Chronic Health Evaluation (APACHE
II).(12) Samples were collected for cultures of blood and
other bodily fluids, depending on the clinical symptoms.
Three groups were defined based on the clinical, laboratory,
and bacteriologic findings.
Infected patients
Seventy-nine patients had a definable source of infection
and/or positive blood cultures and received antibiotic treatment.
A patient was considered to have bacteremia if he or she had
a clinical infection and a positive blood culture. A diagnosis
of urinary tract infection required the presence of symptoms
such as frequency, dysuria, costovertebral angle tenderness,
and significant growth of bacteria of 104-5 CFU/ml in urine
culture. A diagnosis of pneumonia was based on both respiratory
symptoms, such as productive cough, dyspnea, chest pain, and
a pneumonic infiltrate that disappeared during antibiotic
treatment while the patient recovered. For other foci, distinct
radiological or microbiological documentation of the foci
and recovery during antimicrobial treatment were required.
Uninfected patients
The uninfected patient group consisted of 58 patients who,
throughout their course of admission to the hospital, had
no evidence of infection clinically or in the examinations
performed, and received no antibiotic therapy.
Possibly infected patients
Thirteen patients had an uncertain diagnosis of infection.
Two patients suffered from cholecystitis, 2 from hollow organ
perforation, and 1 from appendicitis; their blood cultures
were either negative or unchecked, and none had an ascites
culture. Three female patients had asymptomatic urinary tract
infections, along with other diagnoses that became their principal
reason for admission. All 3 had negative blood cultures, and
none received antibiotic therapy. Eleven patients in this
group had a systemic inflammatory response syndrome (SIRS).
Since a diagnosis of infection in this group was uncertain,
all 13 patients were excluded from the analysis.
The American College of Chest Physicians/ Society of Critical
Care Medicine Consensus Conference definitions of sepsis(1)
were used to identify patients with sepsis, severe sepsis,
septic shock, and SIRS in the uninfected group. The development
of septic shock and patient outcomes were documented by chart
review. Serum CRP level was measured by laser immunonephelometry
(Wako Pure Chemical Industries, Osaka, Japan).
Statistical analysis
The Mann-Whitney U test was used to compare independent samples,
and the chi-square test (or Fishers exact test when appropriate)
was used to compare proportions. The best cut-off value was
chosen using Youden's Index.(13) The receiver operating characteristic
(ROC) curve and the respective areas under the curve (AUC)(14)
were calculated. Statistical calculations were done using
the Statistical Program for the Social Sciences (SPSS, Chicago,
IL, USA). All variables were expressed as the median. A value
of p<0.05 was considered significant.
RESULTS
Admission diagnoses of the 137 patients included in the analysis
are presented in Table 1. Their ages ranged from 18 to 90
(median, 65) years, and 59% (N=81) of the patients were male.
The median APACHE II score was 10. Seven patients died, giving
a crude mortality rate of 5.1%. Seventy-nine (58%) of the
137 patients were infected, and 58 (42%) were uninfected (Table
2). Fifty (63.3%) of the 79 infected patients had an identified
etiological microorganism, and 14 (17.7%) had bacteremia.
The most common site of infection was the urinary tract, followed
by the lung, soft tissue, and biliary tract (Tables 1, 2).
Ten patients had more than 1 site of infection, and 21 patients
had more than 1 kind of infecting microorganism (Table 2).
Sixteen patients (21.5%) had severe sepsis, and 7 (8.9%) developed
septic shock.
SIRS was present in 84.8% of the infected patients and in
37.9% of the uninfected patients (p<0.001). The median
BT in infected patients was 37.8oC and in uninfected patients
was 36.9oC (p<0.001), while the median WBC counts were
10,800 and 8000/ml (p<0.001), respectively (Table 3). A
BT of > 38oC or < 36oC, fulfilling the SIRS criteria
for body temperature, was present in 48.1% of infected patients
and in 10.3% of uninfected patients (p<0.001). A WBC count
of > 12,000/ml or < 4000/ml, again fulfilling the SIRS
criteria, was present in 43.0% of infected patients and 15.5%
of uninfected patients (p<0.001); the WBC count had a similar
but lower sensitivity and specificity than BT (Table 4).
Using Youden's Index, the best cut-off value for CRP was 60
mg/l. CRP levels were ?0 mg/l in 67.1% of infected patients,
and in only 5.2% of uninfected patients (p<0.001). Median
CRP concentrations in infected and noninfected patients were
100.3 and 5.5 mg/l (p<0.001), respectively (Fig. 1). There
was no relationship between CRP levels and the presence of
bacteremia or the type of infecting bacteria (Figs. 2, 3).
The positive predictive value of the CRP level was 94.6%;
however, the negative predictive value was 67.9%. In patients
with SIRS (N=89), the negative predictive value of CRP was
only 50% (Table 3).
Figure 4 shows the ROC curves for CRP, BT, and WBC count.
The area under the ROC curve was greatest for CRP, followed
by BT and then WBC (0.88, 0.77, and 0.67, respectively; p<0.05).
There were statistically significant differences in the AUC
between CRP and BT (p=0.014) and between CRP and WBC (p<0.001),
but not between BT and WBC (p=0.102).(15)
DISCUSSION
CRP, an acute-phase protein, has been widely used clinically
for many years as a diagnostic tool for infection identification.(16)
CRP may be particularly helpful when the SIRS diagnostic criteria
are less reliable, for example in the presence of other disease
processes that affect heart rate, WBC count, or BT. In contrast
to most acute-phase proteins for which there are wide plasma
level variations (which depend on synthesis, consumption,
and catabolic rates), the plasma half-life of CRP is constant
under almost all conditions.(17) Its plasma level is determined
exclusively by its rate of synthesis, which reflects the presence
and extent of disease activity.
The present investigation was comprised of a typical, heterogeneous
adult atraumatic ED population. When comparing the sensitivity
and specificity of CRP with those of clinical signs and routine
laboratory tests for the diagnosis of infection, we considered
that 3 groups of patients:(18) a) those with documented infection;
b) those without infection; and c) those with possible infection,
with this final group being eliminated from the analysis.
The presence of SIRS was the most sensitive clinical parameter
indicating the presence of infection as compared with BT and
WBC count, but the false-positive rate was high. A BT or WBC
count that fulfilled the SIRS criteria was more specific,
but was much less sensitive to infection. Although infected
patients had WBC counts that statistically differed from those
of uninfected patients in this investigation, many other studies
have found no such differences.(10,19,20) This is presumably
due to different characteristics of the patient populations
studied. The present investigation included a broad range
of unselected medical patients in the ED, and compared to
other studies, had the largest range of variability in the
presentation of patients. BT was also found to be a more sensitive
(48.1% vs. 43.0%) and specific (89.7% vs. 84.5%) indicator
of infection compared with the WBC count. Similar findings
were reported by Povoa et al.(9) These findings imply that
in febrile or hypothermic patients, the absence of leukocytosis
does not exclude the presence of infection.
Other authors have attempted to evaluate the role of CRP in
the diagnosis of sepsis in different patient populations,
and variable cut-of values for CRP levels have been reported.
Matson et al.(8) pointed to the fact that "normal"
plasma CRP levels in critically ill patients are rarely the
same as those found in a healthy population; however, a cut-off
for the "normal" concentration was not proposed.
Yentis et al.(21) investigated changes in CRP accompanying
resolution of microbiologically proven sepsis in ICU patients
and found all 32 septic episodes had plasma CRP levels above
43 mg/l on diagnosis, with a median of 223 (range, 43-368)
mg/l. Kallio et al.(22) reported that for 66 cancer patients
with suspected infection, the optimal cut-off values for CRP
using Youden's Index was 140 mg/l, with a sensitivity of 39%
and a specificity of 70%. Ugarte et al.(19) investigated 190
critically ill patients and reported the optimal cut-off level
for CRP using Youden's Index to be 79 mg/l. The present investigation
found the best cut-off value for CRP using Youden's index
to be 60 mg/l, with a comparable sensitivity (67.1% vs. 71.8%)
but a much higher specificity (94.8% vs. 66.6%) compared with
the values of Ugarte et al.(19) The difference in specificity
between the 2 studies is reasonable, since critically ill
patients generally maintained higher "normal" CRP
concentrations than did the ED patients investigated. Both
studies for instance, found a comparable median CRP level
in infected patients (100.3 vs. 121 mg/l), but a markedly
different median CRP level in uninfected patients (5.5 vs.
56 mg/l).
Huang et al.(18) investigated 123 acutely ill ED patients
with suspected infection and reported that a combination of
CRP and WBC count provided a satisfactory predictive value
for infection. In this investigation, however, only patients
with specific complaints, including fever, chills, or acute
distress such as dyspnea, chest pain, consciousness change,
abdominal pain, arthralgia, or upper gastrointestinal bleeding,
were investigated. Many patients were dropped from the investigation
because their CRP levels were unchecked, either because the
patients had no fever or seemed uninfected, or because the
diagnosis of infection was straightforward. The investigation's
results might therefore be hampered by selection bias and
sampling error, and should be interpreted with caution.
Since it is a challenge for emergency physicians to accurately
assess each patient's disposition, an indicator proving the
presence of infection would be as important as one excluding
it. The present investigation shows that using a cut-off level
of 60 mg/l for CRP provides a good positive predictive value.
The negative predictive value however was 67.9%, suggesting
that about 1/3 of infected patients will be overlooked if
the diagnosis is based solely on the presence of a high CRP
level. The negative predictive value for CRP in patients with
SIRS was only 50%. Our results show that CRP is a poor marker
for excluding infection, particularly in patients with SIRS.
SIRS should therefore be managed as sepsis, even if a patient
has a low CRP level, unless other insult, such as trauma,
burns, or pancreatitis, is documented.
In conclusion, CRP is a good indicator of bacterial infection
in adult atraumatic ED patients, being superior to both BT
and the WBC count. CRP cannot, however, be used safely as
an indicator to exclude the presence of infection.
Acknowledgments
We are indebted to the staff of the Emergency Department,
Linkou Chang Gung Memorial Hospital, Taoyuan, Taiwan. We thank
Yu-Mei Chang, Shu-Chin Tsai, Mei-Chuang Chang, and Li-Ling
Lin for their help.
|
