Tests in the Diagnosis
William E. Benitz, MDa,b,*
Definitive diagnosis of sepsis rests upon isolation of pathogenic bacteria in cultures of
specimens obtained from normally sterile spaces within the body. Because of the
close relationship between sepsis (as reflected in a systemic inflammatory response)
is often considered to be sine qua non for diagnosis of sepsis. Whereas such a strict
criterion is certainly appropriate for clinical research, a positive blood culture should
not be required for diagnosis of sepsis in everyday practice. Patients with sepsis
may not have any positive cultures, or cultures may be positive only from sites other
than blood. Conversely, isolation of bacteria in a blood culture may reflect asymptom-
atic bacteremia or contamination. Supplemental diagnostic tests based on evaluation
of the immune response often can help resolve ambiguities in these situations.
Diagnostic tests for neonatal sepsis have two distinct but complementary functions.
Ascertainment of infants who have serious bacterial infection, with identification of the
causative organism, allows selection of the optimal antimicrobial agent and duration of
treatment. Because tests for which results are quickly available have low sensitivity,
their utility for making decisions about initiation of treatment is limited. It is typically
necessary to rely upon the greater sensitivity (and corresponding negative predictive
value) of delayed testing (eg, serial hematological or acute-phase reactant measure-
ments) or results (eg, cultures) to guide discontinuation of antibiotic therapy initiated
empirically. Both aspects are addressed below.
This article reviews the role of conventional diagnostic tests—cultures, blood
counts, acute-phase reactants, and inflammatory mediator levels—in diagnosis of
aDivision of Neonatal and Developmental Medicine, Stanford University School of Medicine,
750 Welch Road, Suite 315, Stanford, Palo Alto, CA 94304, USA
bPackard Children’s Hospital, 725 Welch Road, Palo Alto, CA 94304, USA
* Division of Neonatal and Developmental Medicine, Stanford University School of Medicine,
750 Welch Road, Suite 315, Stanford, Palo Alto, CA 94304.
E-mail address: email@example.com
? Neonatal sepsis ? Bacterial cultures ? Blood counts
?Acute-phase reactants ?Cytokines
Clin Perinatol 37 (2010) 421–438
0095-5108/10/$ – see front matter ª 2010 Elsevier Inc. All rights reserved.
early-onset neonatal sepsis. The developing roles of molecular methods for detection
of bacterial DNA and other novel diagnostic methods such as proteomics are
addressed elsewhere in this issue.
Samples of blood for culture may be obtained by arterial or venous puncture after
preparation of the site with an antibacterial solution (alcoholic solutions of chlorhexi-
dine, iodine, or povidone-iodine are preferable to aqueous povidone-iodine).1–3Data
regarding sterilization of intravenous catheter sites indicate that cleansing for 30
seconds or two consecutive cleansings are superior to a single, brief (5–10 seconds)
disinfection.4Specimens obtained from newly placed intravenous or umbilical cathe-
ters5are also suitable. Specimens from catheters that have been in place for hours or
days may also be informative,6but isolation of an organism (particularly coagulase-
negative Staphylococci) often reflects colonization of the catheter or infusion tubing
rather than bacteremia.7
The volume of blood required to achieve optimal sensitivity of neonatal blood
cultures has not been precisely defined. Historically, sample volumes much smaller
than those considered appropriate for older patients have been used to minimize
development of anemia from large or repeated phlebotomies. Use of small volumes
was supported by data on circulating colony counts in infants with Escherichia coli
sepsis, which suggested that an inoculum as small as 0.2 to 0.5 mL may be sufficient.8
The only prospective comparison of isolation rates from small (0.2 mL) and larger
(2 mL) inocula found a sensitivity of 96% and specificity of 99%.9Other reports indi-
cate that larger volumes of blood are required to identify cases of low-level bacter-
emia, which may account for up to two-thirds of the cases of neonatal sepsis.10In
vitro studies simulating inocula with low (1–3 colony-forming units [CFU]/mL) and
ultralow (<1 CFU/mL) levels of bacteremia indicate that microorganism recovery
should be increased using volumes of 1 to 2 mL for each culture.11Therefore, inocu-
lation of at least 1 mL of blood into a single culture bottle is recommended. Because
anaerobic organisms are rarely implicated in early-onset sepsis,12the entire specimen
should be used for aerobic culture. There is little or no increase in ascertainment of
bacteremia when multiple cultures are obtained from different sites.13,14However,
recovery of identical organisms from replicate, independent specimens provides
unequivocal evidence of a true bacteremia, and multiple negative cultures may
provide stronger evidence of clearance of a previously documented bacteremia.15
Experience with manual methods, relying on development of turbidity in the culture
broth, demonstrated that 96% of the cultures obtained before antibiotic administration
were positive after 48 hours and 98% were positive at 72 hours.16Automated systems
for detection of bacterial growth substantially decreased the time to detection of posi-
tive cultures.17One such system detected 94% of true positive bacterial cultures
obtained before antibiotic therapy (excluding those yielding coagulase-negative
Staphylococci, Corynebacteria, or yeast) within 24 hours and 97% within 36 hours.
Few additional positive results developed between 36 and 72 hours.18Other investi-
gators have reported similar results.19–21For asymptomatic term infants with sus-
pected early-onset sepsis, negative cultures after 36 hours of incubation are
sufficient to exclude sepsis and permit discontinuation of empiric antibiotic therapy.
In the sick neonate, negative cultures are of less value in excluding sepsis. There are
few data on paired simultaneous blood cultures in neonates, so the sensitivity of blood
culture in detection of bacteremia is difficult to assess. Although at least one series
suggests a sensitivity of 100%,13it is important to recognize that this performance
may not be representative. Some of the most informative data is provided by Pourcyr-
ous and colleagues,6who compared yields of peripheral and umbilical blood speci-
mens with sample volumes of 1 to 2 mL. Excluding possible contaminants
(discordant organisms, Staphylococcus epidermidis), cultures obtained from umbilical
catheters and by venipuncture were concordant (ie, each positive) in only 14 of 22
paired samples, implying a sensitivity of only 78%. Among infants in whom bacterial
sepsis was confirmed at autopsy, 18% to 33% have negative premortem blood
cultures.22–24These data suggest a false-negative rate of approximately 20%. In
a sick infant, factors beyond a negative blood culture, including clinical signs and
the results of hematological and acute-phase reactant measurements, should be
considered before empiric antibiotic therapy is discontinued.
A limited evaluation, including a blood culture and observation without treatment
has been suggested for asymptomatic infants thought to be at increased but
moderate risk of bacterial sepsis.25The logic underlying this recommendation should
be questioned. A probability of a positive blood culture sufficient to justify obtaining
one should also merit timely initiation of empiric antibiotic therapy to prevent deterio-
ration of the infant’s condition due to untreated bacteremia. Conversely, development
of clinical signs of infection in an untreated infant requires re-evaluation—including
a new blood culture, which is much more likely to be diagnostic in a symptomatic
infant—and initiation of treatment. The probability that a report of a positive culture
will be the initial or sole stimulus for initiation of therapy appears to be very low.26
The role of spinal fluid cultures in neonates with suspected sepsis remains controver-
sial. Culture-proven bacterial meningitis occurs in approximately 0.25 of every 1000
live-born infants.27,28Blood cultures fail to yield the causative organism in 15% to
50% of infants with bacterial meningitis.29,30This is not an historical artifact, as blood
cultures were negative in 38% (35 of 92) of infants diagnosed between 1997 and 2002
ina large multicenter series.31Conversely, up to13% ofinfants with early-onset sepsis
also have meningitis.32–34Therefore, culture of spinal fluid obtained before or soon
after administration of antibiotics35may be essential to bacteriologic diagnosis.27,36
Because the prevalence of meningitis is low, some have advocated omission of
lumbar puncture in apparently well, term infants undergoing evaluation solely because
of maternal risk factors.37Fielkow and colleagues38observed no meningitis cases
among 284 neonates evaluated because of maternal risk factors (attack rate upper
95% confidence limit 1.1%). Among 712 infants evaluated for maternal risk factors
alone, Schwersenski and colleagues39found only 1 with meningitis (attack rate upper
95% confidence limit 0.4%). Because of the low yield and large number of infants who
would need lumbar puncture to detect one additional infant with meningitis without
bacteremia, it may be reasonable to omit lumbar puncture in such infants.
Similarly, the low rate of meningitis among preterm infants who present with signs of
respiratory distress syndrome (RDS) have led some to recommend omission of lumbar
in those infants.37The data supporting that recommendation are less compelling. In
203 neonates with RDS who had successful lumbar punctures, Eldadah and
colleagues40observed no positive cerebrospinal fluid cultures. Hendricks-Munoz
and Shapiro41reviewed the records of 1390 infants with RDS who did not have lumbar
punctures and found no evidence of missed cases of meningitis. Those studies
reported patient experiences before 1990. The clinical contexts for RDS and early-
onset sepsis have changed substantially since then, owing to use of antepartum
steroids and intrapartum antibiotic prophylaxis, respectively. RDS and bacterial
Diagnosis of Early-Onset Neonatal Sepsis
pneumonia are clinically indistinguishable,42so additional, more current data are
needed to support omission of lumbar puncture in preterm infants with signs of RDS.
In summary, although it is rare in asymptomatic term infants, meningitis is still seen
as a complication of neonatal sepsis. Evaluation of sick newborn infants should
include performance of a lumbar puncture for cerebrospinal fluid culture.
Culture of tracheal aspirate specimens is a useful supplement to blood cultures in
diagnosis of early-onset sepsis. Sherman and colleagues43identified 25 infants with
positive Gram stains of tracheal aspirates among 320 infants less than 8 hours of
age who were either not intubated or had been intubated for less than 30 minutes,
had perinatal risk factors for infection, and had respiratory signs and an abnormal
chest radiograph. All had positive cultures of the tracheal aspirate, but blood cultures
were positive in only 14 (56%) of these infants. Tracheal aspirate cultures yielded
useful diagnostic information, otherwise not available, in at least 3.4% of these infants.
In a subsequent series, cultures were positive in 106 of 733 tracheal aspirate speci-
mens obtained from infants with respiratory distress in the first 12 hours after birth.44
Blood cultures were positive in only 53 (50%) of these infants, suggesting that tracheal
aspirate culture provides unique diagnostic information in 7.2% of the infants evalu-
ated. In a more recent report, 18 of 139 (6.5%) tracheal aspirate cultures obtained
from intubated infants less than 12 hours of age yielded bacterial growth.45Nine of
the positive cultures were considered to represent contaminants, on the basis of light
growth of nonpathogenic organisms (bacillus, diphtheroids, Streptococcus sanguis,
micrococci, nonpathogenic Neisseria, Streptococcus viridans) or of multiple species
of coagulase-negative Staphylococcus. Only two of the nine infants with presumed
tracheal infection had a positive blood culture, implying an incremental diagnostic
yield of 5% (7 of 139 cultures). Therefore, culture of tracheal aspirate specimens
obtained in the first 12 hours after birth adds significant diagnostic information.
Because the trachea quickly becomes colonized after intubation,46however, such
cultures are not useful after prolonged tracheal intubation.
Cultures of urine rarely provide useful diagnostic information for infants undergoing
evaluation for early-onset neonatal sepsis. In 1979, Visser and Hall47suggested that
the risk of suprapubic aspiration was not justified by the low yield (2 of 188) of urine
cultures obtained in infants less than 72 hours of age. Subsequent data have
confirmed those conclusions. Among 280 infants greater than or equal to 35 weeks
gestation for whom urine cultures were obtained in the first 24 hours after birth, the
urine culture was positive in only 1, who was also bacteremic.48No positive cultures
were found in 349 preterm infants (<1500 g birth weight) evaluated in the first 24
hours.49Urine culture is neither necessary nor helpful in infants less than 72 hours
Cultures obtained from superficial sites such as the axilla, umbilical stump, external
ear canal, nasopharynx, and indwelling orogastric or endotracheal tubes correlate
poorly with pathogens isolated from nonpermissive, normally sterile sites.50,51They,
therefore, have low positive predictive value and may lead to erroneous assumptions
about the identity of the causative agent. Use of these cultures to guide management
should be strongly discouraged.52
Early observations that neonates with bacterial sepsis often have abnormal white
blood cell counts led to the hypothesis that hematologic studies might allow diagnosis
of sepsis before results of cultures become available. Because the imperfect sensi-
tivity of any single measurement, including the total white cell count, total neutrophil
count, neutrophil count, immature-to-total (I:T) or immature-to-mature neutrophil
ratios, soon became apparent, several strategies for combining multiple observations
have been proposed.53–55The advantages of this approach have been demonstrated
by Rodwell and colleagues55(Table 1). Combination of multiple hematological deter-
Performance of hematological findings and a hematological scoring system
in 298 neonates evaluated for sepsis during the first postnatal month55
[ I:T ratioa
Y or [ neutrophil counta
Immature:mature ratio R0.3
[ immature neutrophil counta
Y or [ white cell countb
Platelet count %150,000/mm3
3395 39 93
R2 10063 21100
R3 96 7831 99
R489 8945 99
R5 4196 52 94
R622 100 8693
Hematological score R3
Preterm infants %24 hours of
Term infants %24 hours of agef
Infants >24 hours of ageg
100 748 100
93 8365 97
aAs defined by the reference ranges of Manroe et al.53
b%5000 or R25,000, 30,000 or 21,000 per mm3at birth, 12–24 hours and day 2 or after,
cR31 for vacuolization, toxic granulation, or Dohle bodies according to the 0 to 41 scale of
Zipurksy et al.56
dAssign one point for each individual abnormal hematological finding listed above (2 points for
an abnormal total neutrophil count if no neutrophils are present on the blood smear) and sum to
determine the score.
eTen sepsis cases among 113 preterm infants.
fThree sepsis cases among 130 term infants.
gFourteen sepsis cases among 55 infants.
Data from Manroe BL, Weinberg AG, Rosenfeld CR, et al. The neonatal blood count in health and
disease. I. Reference values for neutrophilic cells. J Pediatr 1979;95(1):89–98.
Diagnosis of Early-Onset Neonatal Sepsis
improvement in specificity and positive predictive value without a reduction in sensi-
tivity. Unless the score is very high (R5), however, most infants with elevated scores
will not prove to have bacterial infection. When infants being evaluated for early-onset
sepsis are considered separately, all 13 with culture-proven sepsis had a hematolog-
ical score greater than or equal to three, providing sensitivity and negative predictive
values of 100%. Two-thirds of the preterm infants and more than 90% of the term
infants with a hematological score greater than or equal to three on the first postnatal
day did not have confirmed sepsis—giving the test a low positive predictive value.
Inclusion of infants with probable infection increased the positive predictive value of
scores greater than or equal to three in these infants to only 33% for preterm and
62% for term infants. Because the number of babies with sepsis in these subgroups
is small, the confidence intervals on these test performance characteristics are
wide. Greenberg and Yoder found that the sensitivity and negative predictive values
of hematological scores obtained at 1 to 7 hours of age are much lower than for those
obtained at 12 to 24 hours of age (Table 2),57confirming earlier observations that
hematological results are often initially reassuring in infants with proven sepsis.58,59
More recently, Ottolini and colleagues26assessed the utility of early blood counts in
evaluation of 1665 asymptomatic infants greater than or equal to 35 weeks gestational
age deemed at risk for early-onset sepsis because of maternal or intrapartum risk
factors (maternal group B streptococcal colonization, intrapartum fever R38.0?C,
rupture of membranes >18 hours before birth, or a previous infant with invasive group
B streptococcal disease, and either no intrapartum antibiotic prophylaxis or only
a single dose <4 hours before birth). Only 7 of the 454 at-risk infants who had abnormal
blood counts (white blood cell >30,000 or <5000/mm3; absolute neutrophil count
<1,500/mm3, or an I:T ratio >0.2) developed clinical signs of sepsis; none had a posi-
tive blood culture. In this group of patients, the sensitivity, specificity, and positive and
negative predictive values of an abnormal white blood cell count were 41%, 73%,
1.5%, and 99%, respectively. Of the other 17,655 term infants born at their center
during the study period, 1996 to 1999, 283 were diagnosed with clinical sepsis: 134
had clinical signs evident at the initial examination and 149 developed clinical signs
of sepsis within the first 48 hours. Blood cultures yielded pathogenic organisms in 8
of these infants, all of whom had both clinical signs and abnormal white blood cell
Performance of hematological scoring systems for diagnosis of early-onset group B
streptococcal sepsis in the first 24 hours after birth57
Initial blood counts (1–7 hours of age)
Manroe et al53
Rodwell et al55
Spector et al54
Blood counts at 12–24 hours of age
Manroe et al53
Rodwell et al55
Spector et al54
100 7373 100
Data from Greenberg DN, Yoder BA. Changes in the differential white blood cell count in screening
for group B streptococcal sepsis. Pediatr Infect Dis J 1990;9(12):886–9.
indices. These data suggest that an early complete blood count adds little diagnostic
information to clinical assessment in the evaluation of either asymptomatic or symp-
tomatic neonates with possible bacterial infection.
Normal hematological findings in infants with clinical findings consistent with sepsis
false-positive rate, so they should not be relied upon to guide decisions about initiation
of antibiotic therapy in such infants. Close monitoring for development of clinical signs
of infection should replace reliance on blood counts in infants for whom antibiotic
empiric treatment while those studies are pending.
Inflammatory stimuli of any sort, including infection, trauma, or ischemia, cause
marginalization, extravasation, and activation of granulocytes and monocytes, with
release of multiple proinflammatory cytokines, including interleukin (IL)-1b, IL-6, and
tumor necrosis factor-a (TNF-a).61These mediators stimulate production of a variety
of proteins referred to as acute-phase reactants. The time courses of these responses
in adults have been well characterized (Fig. 1)61and follow a similar pattern in
neonates. Several of these, including C-reactive protein (CRP), fibrinogen (for which
the erythrocyte sedimentation rate may be a surrogate measurement), haptoglobin,
orosomucoid, amyloid protein A, and procalcitonin (PCT) have been evaluated as
diagnostic tools in neonatal sepsis.
Fig. 1. Time course of acute-phase reactant responses to an inflammatory stimulus. (From
Gabay C, Kushner I. Acute-phase proteins and other systemic responses to inflammation.
N Engl J Med 1999;340(6):448–4; with permission.)
Diagnosis of Early-Onset Neonatal Sepsis
Erythrocyte Sedimentation Rate
The availability of a simple and inexpensive method for determination of the erythro-
cyte sedimentation rate using only few drops of capillary blood62led to consideration
of this test for early diagnosis of neonatal sepsis. Adler and Denton63described
a gradual increase in sedimentation rate over the first 2 weeks after birth, with the
95th percentile increasing from 2 mm per hour at 24 hours to 6 mm per hour at 48
hours of age. Evans and colleagues64noted that all four infants less than 3 days of
age with confirmed sepsis in their series had microsedimentation rates greater than
the 95% percentile for healthy infants (6 mm/hr). A subsequent report from Ibsen
and colleagues confirmed that although an elevated initial sedimentation rate supports
a diagnosis of serious bacterial infection, a low value does not exclude the diag-
nosis.65Because this test is easily performed even where access to laboratory tech-
nology is limited, it may still have some utility in such settings. In other situations, it has
been replaced by more specific biochemical measurements, which also have been
more extensively evaluated.
IL-6 released by activated granulocytes stimulates the liver to produce CRP, which
takes its namefrom the observation that itforms an insoluble complex with the C-poly-
saccharide of Streptococcus pneumoniae. Serum levels may increase 100- to 1000-
fold in response to bacterial infection or other inflammatory conditions. In infants
with treated infections, serum levels begin to increase 6 to 8 hours after the onset
of illness and peak after 2 to 3 days (Fig. 2).66Because of this typical delayed
response, the sensitivity of CRP elevation at the time of evaluation for suspected
Fig. 2. Time course of CRP levels in neonates with group B streptococcal infection. (From
Philip AGS. Response of C-reactive protein in neonatal Group B streptococcal infection.
Pediatr Infect Dis 1985;4(2):145–8; with permission.)
sepsis is low, particularly with early-onset sepsis. In the author’s experience in over
1000 infants, only 35% of infants with culture-proven early-onset sepsis and 39% of
those with proven or probable early-onset sepsis had an elevated CRP (R10 mg/
L).67Several smaller series had previously reported sensitivity values ranging from
29% to 90%,68and similar results have been described more recently.69,70Because
many neonates with early-onset sepsis will not be identified by measurement of
a serum CRP level, it should have no role in the initial evaluation of these infants.
Serial determinations of CRP levels over the first 2 to 3 days of illness may be quite
useful in determining the duration of empiric antibiotic therapy, however.67,71Noting
that at least one of three CRP levels obtained at the time of evaluation and 12 and
24 hours later was elevated in 92% of infants with gram-negative bacteremia and in
64% of those with gram-positive bacteremia (79% of those with organisms other
than S epidermidis), Pourcyrous and colleagues71suggested that serial normal CRP
values, in the absence of other clinical findings suggestive of infection, might indicate
that it is safe to discontinue antibiotics. Benitz and colleagues67examined this ques-
tion, using a different schedule for sample collection, with specimens obtained at the
time of the initial evaluation, 8 to 24 hours after that, and again 24 hours after the
second collection. The initial measurement added no information to the other two.
Two delayed levels separated by 24 hours had a negative predictive value of 99.7%
for both proven and for proven or probable sepsis and likelihood ratios of 0.15 and
0.03 for proven and proven or probable sepsis, respectively. Extending the interval
of sampling into the second 24 hours enhances the negative predictive value of serial
normal levels. Because persistently normal CRP levels correlate strongly with the
absence of infection, they may be used to determine that antibiotics may be safely dis-
continued. This strategy may reduce the duration of empiric treatment of infants with
nonspecific clinical findings and at-risk asymptomatic neonates.
Determination of the duration of therapy by monitoring for normalization of elevated
CRP levels has also been suggested.72,73Although these preliminary observations
suggest that a return of the serum CRP level to normal may be an indication that
the duration of treatment has been adequate, the numbers of infants with proven
infection with pathogenic bacteria in these series (eight and four, respectively, all of
whom were treated for at least 5 days) are not sufficient to permit assessment of
that hypothesis. Until data from a much larger sample of infants with demonstrated
infection is available, this application of CRP levels cannot be recommended.
Many conditions other than bacterial infection may be associated with elevated
CRP levels. Elevated levels may also be seen with viral infection, traumatic or ischemic
tissue injuries, hemolysis, meconium aspiration, or chorioamnionitis without invasive
fetal or neonatal disease. One or more elevated CRP levels, in the absence of other
data indicating that the infant is infected, should not constitute a sole indication for
continuation of empiric antibiotic therapy.
Considerable attention has recently been drawn to PCT as a potential alternative to
CRP for diagnosis of neonatal sepsis. PCT is produced both in the liver and by macro-
phages, with serum levels rising 4 to 6 hours after exposure to bacterial products.74
Because this PCT response is more rapid than elevation of CRP, this marker is an
attractive alternative for detection of early-onset sepsis. Initial reports of its clinical
application appeared to support this expectation, suggesting sensitivity of an elevated
serum level might be as high as 100%.75,76This early promise has not been fully real-
ized after more extensive evaluation. Three challenges, in particular, have limited the
application of PCT levels in evaluation of newborn infants for possible sepsis: the need
Diagnosis of Early-Onset Neonatal Sepsis
for detailed age-specific normative ranges, disappointing sensitivity early in the course
of infection, and lack of specificity.
Early experience in newborn infants suggested a spontaneous increase in PCT
levels over the first day after birth.77This observation was soon confirmed and
extended by others. Data collected by Sachse and colleagues78indicated that PCT
levels in healthy infants peaked at 24 to 36 hours after birth and decreased thereafter.
Chiesa and colleagues79established normative reference ranges for PCT during the
first 48 hours after birth, which include an increase for the upper 95th percentile
from 0.7 ng/mL at birth to 20 ng/mL at 24 hours, followed by a decline to less than
2 ng/mL at 48 hours of age (Fig. 3A). Turner and colleagues80extended these obser-
vations to preterm infants, in whom serum PCT levels peak at approximately 28 hours
of age, but at levels lower than those observed in term infants. Clinical utility of PCT in
diagnosis of early-onset sepsis will depend upon establishment of normative ranges
specific to both gestational and postnatal ages. The limitations of a single cutoff value
during this interval may be exemplified by the results of Resch and colleagues,81who
found a sensitivity of 77% and specificity of 91% using a cutoff value of 6 ng/mL for
infants between 0 and 12 hours of age.
Applying age-specific reference values in a sample of 120 infants admitted to their
neonatal intensive care unit, Chiesa and colleagues82found a sensitivity of 92.6% and
a specificity of 97.5% for PCT concentrations for detection of early-onset sepsis.
Sepsis was confirmed by positive blood cultures in 14 infants and presumed on the
basis of clinical signs and ancillary laboratory studies in 14 more (all of whom were
born to women who received intrapartum antibiotics). Notably, however, the sensi-
tivity of elevated PCT levels during the first 6 hours after birth was only 50% (3 of 6
tested infants with sepsis; Fig. 3B). In a subsequent report, results were similar at
24 hours of age (sensitivity 95% and specificity 96%), but less favorable at birth (sensi-
tivity 79% and specificity 95%).
Several reports have subsequently examined the performance of PCT levels at birth,
using either cord blood specimens or samples obtained soon after birth. In a sample of
120 infants, 21 of whom were infected, Guibourdenche and colleagues83found that
Fig. 3. (A) Reference range for procalcitonin levels in healthy neonates over the first 48
hours after birth. Broken lines represent upper and lower boundaries of the 95% confidence
interval. Solid line represents the geometric mean. (B) Procalcitonin values among infants
with culture-proven or presumed sepsis in the first 48 hours after birth. (From Chiesa C, Pan-
ero A, Rossi N, et al. Reliability of procalcitonin concentrations for the diagnosis of sepsis in
critically ill neonates. Clin Infect Dis 1998;26(3):664–2; with permission.)
a PCT level greater than 2.5 ng/mL at birth had a sensitivity of 87% and specificity of
90%. Almost identical results were obtained by Joram and colleagues using a cord
blood level cutoff of 0.5 ng/mL.84Kordek and colleagues85reported that the sensitivity
and specificity of PCT levels in umbilical cord blood at an optimal cutoff value of less
than or equal to 2.6 ng/mL were only 69% and 81%, respectively. In a subsequent
series, that group found a sensitivity of 80% and a specificity of 72% using a cord
blood cutoff value of 1.22 ng/mL. In each of these series, PCT was more sensitive
than CRP. More recently, Lopez Sastre and colleagues, using a threshold PCT level
of 0.55 ng/mL, reported a sensitivity of only 75% and specificity of only 72% at birth.86
Therefore, the most appropriate threshold value of confirmation or exclusion of early-
onset sepsis remains indeterminate, and neither the sensitivity nor specificity of an
elevated PCT level at birth is sufficient to guide antibiotic therapy for an infant at
risk for serious infection.
Moderately increased sensitivity of PCT, as compared with CRP, may be offset by
lower specificity in the population of interest: infants with signs of possible sepsis who
are not infected. Elevated levels of PCT are found in infants with respiratory distress
syndrome,87hemodynamic instability,87asphyxia,88intracranial hemorrhage,88pneu-
mothorax,88or resuscitation88and in those with infection. The positive predictive value
of this measure is therefore limited by the many noninfectious causes of an elevated
PCT level in a sick neonate, so it should not be used to justify continuation of antibiotic
therapy in an infant who does not otherwise have convincing signs of sepsis.
ILS AND OTHER CYTOKINES
Because acute-phase reactants are produced in response to proinflammatory cyto-
kines, direct measurement of serum cytokine levels promised to provide an earlier
indication of infection than could be achieved by measurement of the secondary
responses. Initial assessments of several cytokines, including IL-1b,89IL-6,90–92IL-
8),92–94soluble IL-2 receptor (sIL2R),95and TNF-a92demonstrated early responses
to bacterial infection in neonates. Initial reports of high sensitivity and specificity
have not been consistently borne out by subsequent experience. Dollner and
colleagues96evaluated several inflammatory mediators in cord blood specimens
and found that IL-1b, IL-6, and IL-8 measurements were associated with the largest
areas under the receiver-operator characteristic curves (Fig. 4). The sensitivities of
these tests at the optimal cutoff values were only 74%, 84%, and 81%, respectively.
Santana Reyes and colleagues97collected blood specimens at the time of admission
of newborn infants with suspected infection. They found that sensitivity and specificity
were 80% and 92% for CRP, 60% and 87% for IL-1b, 61% and 80% for IL-6, 62% and
96% for IL-8, 54% and 92% for TNF-a, and 63% and 94% for soluble receptor of IL-2.
As an individual analyte, none of these mediators reliably identifies infected neonates
or excludes infection in those at risk.
Noting that serum levels of cytokines often return to normal as those of acute-phase
reactants rise, several groups have proposed use of combinations of measurements
to enhance the likelihood of an abnormal result in the face of uncertainty about the
stage of illness at which an infant is evaluated for suspected infection. In concept,
elevation of a cytokine level would permit detection of infants early in the course of
infection, before CRP elevation is apparent, while the delayed elevation of CRP would
identify infants later in the course, after a transient cytokine response has resolved.
Franz and colleagues98found that the combination of IL-8 and CRP on admission
Diagnosis of Early-Onset Neonatal Sepsis
detected early-onset bacterial infection with a sensitivity of 92% and specificity of
74%. Because this combination was much more specific that the combination of I:T
ratio and CRP level, they estimated a 40% reduction in the number of infants in
whom treatment might be initiated using the former rather thanlatter criterion. Santana
Reyes and colleagues97found the best combination for sepsis diagnosis to be CRP,
IL-8, and sIL2R, with a sensitivity of 85% and a specificity of 97%. Others have
confirmed enhanced performance of combined as compared with single markers,
but no panel of analytes has been prospectively shown to have both excellent sensi-
tivity and good specificity.
Timely diagnosis of neonatal bacterial infection remains one of the most common and
problematic tasks encountered in neonatal medicine. Evaluation of infants at risk for
early-onset infection must be based on the pregnancy and intrapartum history and
clinical manifestations of illness in the infant, which may be absent or nonspecific.
While careful observation for development of clinical signs remains essential to
management of these infants, the potential for reduction of morbidity and mortality
with early initiation of antibiotic therapy for bacterial infection has stimulated an
ongoing search for laboratory methods for diagnosis of infection before clinical signs
become apparent. Cultures of normally sterile body fluids, and of blood in particular,
Fig. 4. Receiver-operator characteristic curves for cord blood serum cytokine determinations
for diagnosis of early-onset neonatal sepsis in preterm infants. (From Dollner H, Vatten L,
Linnebo I, et al. Inflammatory mediators in umbilical plasma from neonates who develop
early-onset sepsis. Biol Neonate 2001;80(1):41–7; with permission.)
remain crucial to definitive diagnosis of sepsis, but there are both false-positive and
false-negative results, and results become available only after a day or two. Despite
long-established use of hematological counts for identification of infected infants,
these tests are neither sensitive nor specific, particularly at the beginning of an
episode of early-onset sepsis. Over the past two decades, several acute-phase reac-
tants, including CRP and PCT, have been evaluated. More recently, serum levels of
inflammatory mediators have been considered. Of these, IL-6 and IL-8 appear most
promising. However, because levels of these cytokines decrease quickly after the
initial response, optimal utility may be obtained in combination with CRP or PCT,
which increase more slowly but remain elevated for longer periods of time.
Ancillary diagnostic tests for early-onset infection have potential applications in
three circumstances. First, in the case of an infant who is seriously ill, with shock,
respiratory failure, lethargy, acidemia, or the like, the decision to initiate antibiotic
therapy is not dependent on laboratory test results. The role of ancillary tests in these
infants lies in retrospective ascertainment of those who were not infected and for
whom antibiotics can be safely discontinued after some interval of empiric treatment.
Negative culture results, particularly when supplemented by absence of signs of an
acute-phase response, are most useful in this situation. Only CRP has been system-
atically evaluated for this purpose.67,71,99Second, a test that could identify infants who
are septic among those with only historical risk factors (prolonged rupture of
membranes, intrapartum fever, group B streptococcal colonization, etc) or with minor
clinical signs (transient hypoglycemia, tachypnea, an oxygen requirement, etc) would
be useful to ensure treatment of infected infants while allowing nontreatment of others.
No currently available test or combination of tests has sufficient sensitivity and spec-
ificity to achieve these objectives. In these situations, the choices are either initiation of
empiric therapy, followed byreliance on cultures and acute-phase reactants to identify
infants for whom treatment was unnecessary, or close clinical observation, with initi-
ation of therapy for infants who develop clinical signs of infection. Third, acute-phase
reactants or cytokine levels may be of value for monitoring response to therapy or
prognosis in infants in whom infection is confirmed. Failure of CRP levels to return
to normal or recurrent elevation after an initial improvement may be an indication of
a significant complication, such as subdural empyema complicating bacterial
meningitis, for example. Although preliminary reports suggest that normalization of
acute-phase reactants may indicate adequacy of a course of treatment, enabling
discontinuation of therapy, experience in infants with culture-proven sepsis is not
yet sufficient to support recommendation of this practice.
1. Calfee DP, Farr BM. Comparison of four antiseptic preparations for skin in the
prevention of contamination of percutaneously drawn blood cultures: a random-
ized trial. J Clin Microbiol 2002;40(5):1660–5.
2. Mimoz O, Karim A, Mercat A, et al. Chlorhexidine compared with povidone-iodine
as skin preparation before blood culture. A randomized, controlled trial. Ann
Intern Med 1999;131(11):834–7.
3. Trautner BW, Clarridge JE, Darouiche RO. Skin antisepsis kits containing alcohol
and chlorhexidine gluconate or tincture of iodine are associated with low rates of
blood culture contamination. Infect Control Hosp Epidemiol 2002;23(7):397–401.
4. Malathi I, Millar MR, Leeming JP, et al. Skin disinfection in preterm infants. Arch
Dis Child 1993;69(3 Spec No):312–6.
Diagnosis of Early-Onset Neonatal Sepsis
5. Cowett RM, Peter G, Hakanson DO, et al. Reliability of bacterial culture of blood
obtained from an umbilical artery catheter. J Pediatr 1976;88(6):1035–6.
6. Pourcyrous M, Korones SB, Bada HS, et al. Indwelling umbilical arterial catheter:
a preferred sampling site for blood culture. Pediatrics 1988;81(6):821–5.
7. Ruderman JW, Morgan MA, Klein AH. Quantitative blood cultures in the diagnosis
of sepsis in infants with umbilical and Broviac catheters. J Pediatr 1988;112(5):
8. Dietzman DE, Fischer GW, Schoenknecht FD. Neonatal Escherichia coli septi-
cemia—bacterial counts in blood. J Pediatr 1974;85(1):128–30.
9. Solorzano-Santos F, Miranda-Novales MG, Leanos-Miranda B, et al. A blood
micro-culture system for the diagnosis of bacteremia in pediatric patients. Scand
J Infect Dis 1998;30(5):481–3.
10. Kellogg JA, Ferrentino FL, Goodstein MH, et al. Frequency of low level bacter-
emia in infants from birth to two months of age. Pediatr Infect Dis J 1997;16(4):
11. Schelonka RL, Chai MK, Yoder BA, et al. Volume of blood required to detect
common neonatal pathogens. J Pediatr 1996;129(2):275–8.
12. Noel GJ, Laufer DA, Edelson PJ. Anaerobic bacteremia in a neonatal intensive
care unit: an eighteen-year experience. Pediatr Infect Dis J 1988;7(12):858–62.
13. Sarkar S, Bhagat I, DeCristofaro JD, et al. A study of the role of multiple site blood
cultures in the evaluation of neonatal sepsis. J Perinatol 2006;26(1):18–22.
14. Wiswell TE, Hachey WE. Multiple site blood cultures in the initial evaluation for
neonatal sepsis during the first week of life. Pediatr Infect Dis J 1991;10(5):365–9.
15. Sarkar S, Bhagat I, Wiswell TE, et al. Multiple site blood cultures are mandatory to
document the clearance of bacteremia in neonates with sepsis. Pediatr Res
2003;53(4) [abstract 2848].
16. Pichichero ME, Todd JK. Detection of neonatal bacteremia. J Pediatr 1979;94(6):
17. Pauli I Jr, Shekhawat P, Kehl S, et al. Early detection of bacteremia in the neonatal
intensive care unit using the new BACTEC system. J Perinatol 1999;19(2):127–31.
18. Garcia-Prats JA, Cooper TR, Schneider VF, et al. Rapid detection of microorgan-
isms in blood cultures of newborn infants utilizing an automated blood culture
system. Pediatrics 2000;105(3 Pt 1):523–7.
19. Kumar Y, Qunibi M, Neal TJ, et al. Time to positivity of neonatal blood cultures.
Arch Dis Child Fetal Neonatal Ed 2001;85(3):F182–6.
20. Janjindamai W, Phetpisal S. Time to positivity of blood culture in newborn infants.
Southeast Asian J Trop Med Public Health 2006;37(1):171–6.
21. Jardine L, Davies MW, Faoagali J. Incubation time required for neonatal blood
cultures to become positive. J Paediatr Child Health 2006;42(12):797–802.
22. Eisenfeld L, Ermocilla R, Wirtschafter D, et al. Systemic bacterial infections in
neonatal deaths. Am J Dis Child 1983;137(7):645–9.
23. Squire E, Favara B, Todd J. Diagnosis of neonatal bacterial infection: hematologic
and pathologic findings in fatal and nonfatal cases. Pediatrics 1979;64(1):60–4.
24. Pierce JR, Merenstein GB, Stocker JT. Immediate postmortem cultures in an
intensive care nursery. Pediatr Infect Dis 1984;3(6):510–3.
25. Schrag S, Gorwitz R, Fultz-Butts K, et al. Prevention of perinatal group B strepto-
coccal disease. Revised guidelines from CDC. MMWR Recomm Rep 2002;
26. Ottolini MC, Lundgren K, Mirkinson LJ, et al. Utility of complete blood count and
blood culture screening to diagnose neonatal sepsis in the asymptomatic at risk
newborn. Pediatr Infect Dis J 2003;22(5):430–4.
27. Wiswell TE, Baumgart S, Gannon CM, et al. No lumbar puncture in the evalu-
ation for early neonatal sepsis: will meningitis be missed? Pediatrics 1995;
28. Holt DE, Halket S, de Louvois J, et al. Neonatal meningitis in England and Wales:
10 years on. Arch Dis Child Fetal Neonatal Ed 2001;84(2):F85–9.
29. Visser VE, Hall RT. Lumbar puncture in the evaluation of suspected neonatal
sepsis. J Pediatr 1980;96(6):1063–7.
30. Shattuck KE, Chonmaitree T. The changing spectrum of neonatal meningitis over
a fifteen-year period. Clin Pediatr (Phila) 1992;31(3):130–6.
31. Garges HP, Moody MA, Cotten CM, et al. Neonatal meningitis: what is the corre-
lation among cerebrospinal fluid cultures, blood cultures, and cerebrospinal fluid
parameters? Pediatrics 2006;117(4):1094–100.
32. Hamada S, Vearncombe M, McGeer A, et al. Neonatal group B streptococcal
disease: incidence, presentation, and mortality. J Matern Fetal Neonatal Med
33. Grimwood K, Darlow BA, Gosling IA, et al. Early-onset neonatal group B strepto-
coccal infections in New Zealand 1998–1999. J Paediatr Child Health 2002;38(3):
34. Neto MT. Group B streptococcal disease in Portuguese infants younger than 90
days. Arch Dis Child Fetal Neonatal Ed 2008;93(2):F90–3.
35. Kanegaye JT, Soliemanzadeh P, Bradley JS. Lumbar puncture in pediatric bacte-
rial meningitis: defining the time interval for recovery of cerebrospinal fluid path-
ogens after parenteral antibiotic pretreatment. Pediatrics 2001;108(5):1169–74.
36. Heath PT, Nik Yusoff NK, Baker CJ. Neonatal meningitis. Arch Dis Child Fetal
Neonatal Ed 2003;88(3):F173–8.
37. McIntyre P, Isaacs D. Lumbar puncture in suspected neonatal sepsis. J Paediatr
Child Health 1995;31(1):1–2.
38. Fielkow S, Reuter S, Gotoff SP. Cerebrospinal fluid examination in symptom-free
infants with risk factors for infection. J Pediatr 1991;119(6):971–3.
39. Schwersenski J, McIntyre L, Bauer CR. Lumbar puncture frequency and cerebro-
spinal fluid analysis in the neonate. Am J Dis Child 1991;145(1):54–8.
40. Eldadah M, Frenkel LD, Hiatt IM, et al. Evaluation of routine lumbar punctures in
newborn infants with respiratory distress syndrome. Pediatr Infect Dis J 1987;
41. Hendricks-Munoz KD, Shapiro DL. The role of the lumbar puncture in the admis-
sion sepsis evaluation of the premature infant. J Perinatol 1990;10(1):60–4.
42. Vollman JH, Smith WL, Ballard ET, et al. Early-onset group B streptococcal
disease: clinical, roentgenographic, and pathologic features. J Pediatr 1976;
43. Sherman MP, Goetzman BW, Ahlfors CE, et al. Tracheal aspiration and its clinical
correlates in the diagnosis of congenital pneumonia. Pediatrics 1980;65(2):
44. Sherman MP, Chance KH, Goetzman BW. Gram’s stains of tracheal secretions
predict neonatal bacteremia. Am J Dis Child 1984;138(9):848–50.
45. Booth GR, Al-Hosni M, Ali A, et al. The utility of tracheal aspirate cultures in the
immediate neonatal period. J Perinatol 2009;29(7):493–6.
46. Harris H, Wirtschafter D, Cassady G. Endotracheal intubation and its relationship
to bacterial colonization and systemic infection of newborn infants. Pediatrics
47. Visser VE, Hall RT. Urine culture in the evaluation of suspected neonatal sepsis.
J Pediatr 1979;94(4):635–8.
Diagnosis of Early-Onset Neonatal Sepsis