 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
* Department of Pediatrics, Division of Neonatology and Pediatric Critical Care, University of Ulm, Ulm, Germany
Department of Pediatrics, Free University of Berlin, University Hospital Benjamin Franklin, Berlin, Germany
Department of Pediatrics, Landeskrankenhaus Villach, Villach, Austria
|| Department of Microbiology and Infectious Disease, Royal Women's Hospital, Women's and Children's Health, Carlton, Victoria, Australia
¶ Department of Pediatrics, Royal Women's Hospital, Women's and Children's Health, Parkville, Victoria, Australia
# Department of Pediatrics, Kärnsjukhuset, Skövde, Sweden
** Department of Pediatrics, Länssjukhuset Ryhov, Jönköping, Sweden
Department of Woman and Child Health, Division of Neonatology, Karolinska Institutet and Hospital, Stockholm, Sweden
Department of Neonatology, Academic Hospital, Brussels Free University, Brussels, Belgium
|| Department of Biometry and Medical Documentation, University of Ulm, Ulm, Germany
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Methods. Term and preterm infants who were <72 hours of age and had clinical signs or obstetric risk factors suggesting neonatal bacterial infection but stable enough to wait for results of diagnostic tests were enrolled into the study. A total of 1291 infants were randomly assigned to receive antibiotic therapy according to the guidelines of each center (standard group) or to receive antibiotic therapy when IL-8 was >70 pg/mL and/or CRP was >10 mg/L (IL-8 group). The primary outcome variables were 1) the number of infants treated with antibiotics and 2) the number of infants with infections missed at the initial evaluation.
Results. In the IL-8 group, fewer infants received antibiotic therapy than in the standard group (36.1% [237 of 656] vs 49.6% [315 of 635]). In the IL-8 group, 24 (14.5%) of 165 infants with infection were not detected at the initial evaluation, compared with 28 (17.3%) of 162 in the standard group.
Conclusions. The number of newborn infants who received postnatal antibiotic therapy can be reduced with a diagnostic algorithm that includes measurements of IL-8 and CRP. This diagnostic strategy seemed to be safe.
Key Words: cytokines • C-reactive protein • newborn infant • interleukin-8 • sepsis
Abbreviations: CRP, C-reactive protein • IL-8, interleukin 8 • CI, confidence interval
Bacterial infections contribute significantly to morbidity and mortality in newborn infants.1,2 Successful treatment depends on early initiation of appropriate antibiotic therapy, but early diagnosis of neonatal bacterial infections is difficult because clinical signs are nonspecific and may initially be subtle.1,2 Therefore, a frequently adopted strategy is to start antibiotics in all infants with clinical signs or obstetric risk factors suggesting infection. Consequently, on retrospective reevaluations, the majority of infants received antibiotics "unnecessarily."1–5
In light of the ever-increasing emergence of resistant bacteria, restrictive use of antibiotic therapy is indispensable6–9 and efforts to ensure a more cautious use of antibiotic therapy in newborn infants are important. The use of laboratory tests to rule out infection was suggested to reduce unnecessary antibiotic therapy.4,5 C-reactive protein (CRP), an excellent marker for established neonatal bacterial infections, is not useful for early diagnosis.10–12 In contrast, proinflammatory cytokines increase early in neonatal bacterial infection before a rise in CRP can be documented.13–21
Interleukin 8 (IL-8), a proinflammatory cytokine predominantly produced by monocytes, macrophages, and endothelial cells,22 rises early in the course of neonatal bacterial infections.16,17,23–27 Preliminary, nonrandomized, single-center studies suggested that measurements of IL-8 and CRP may help to reduce unnecessary antibiotic therapy,17,25 but the feasibility and consequences of integrating cytokine measurements into clinical decision making have not been studied adequately. Therefore, this multicenter, randomized, controlled trial was designed to test the hypotheses that a diagnostic algorithm that includes measurements of plasma IL-8 and CRP 1) reduces antibiotic therapy and 2) does not result in more initially missed infections when compared with standard management.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Selection Criteria
Patients were eligible for the trial when they fulfilled all of the following 3 criteria: 1) at least 1 predefined clinical sign suggesting neonatal bacterial infection, 2) age <72 hours, and 3) clinical condition stable enough to wait for results of diagnostic tests before initiating antibiotic therapy. The predefined clinical signs were pallor; gray skin color; capillary refill time >2 seconds; dyspnea, tachypnea, and need of respiratory support or supplemental oxygen; thermal instability (not explained by prematurity); unexplained hyperglycemia; feeding difficulties; increasing frequency of apnea and/or bradycardia; and lethargy, irritability, and increased or decreased muscle tone.1,2,28
In addition, patients who were admitted to the nursery within the first 24 hours of life were eligible when at least 1 of the following obstetric risk factors suggested amniotic infection: rupture of membranes before the onset of labor, rupture of membranes >18 hours before delivery, foul-smelling amniotic fluid, ineffective tocolysis before 35 weeks of gestation, fetal tachycardia (>160 beats/min), and maternal rectal temperature >38.5°C.1
Exclusion criteria were septic shock or rapidly deteriorating clinical condition (in these patients, antibiotic therapy was initiated immediately), treatment already commenced (eg, during transfer), no written informed parental consent, birth weight below a center-defined threshold (eg, <1500 g), history of severe perinatal asphyxia, surgery before study entry, congenital malformations, chromosomal anomalies, congenital metabolic defects, and family history of an immunodeficiency syndrome.
Treatment Assignments
A total of 1291 patients were randomly assigned to 1 of 2 diagnostic algorithms using sealed opaque envelopes. Block randomization with a block size of 8 and stratification for gestational age (>36 + 6/7 weeks, 30–36 + 6/7 weeks, and <30 weeks of gestation) was used at each center. In all neonates, blood was collected for IL-8 and CRP measurements at the time of the initial evaluation.
In the IL-8 group, antibiotic therapy was initiated when IL-8 was >70 pg/mL17 and or CRP was >10 mg/L.10–12,17,29,30 The cutoff value for IL-8 of 70 pg/mL was established by receiver operating characteristic curve analysis in a previous study of infants with suspected early-onset neonatal bacterial infections.17 In that study, the combination of IL-8 and CRP at the selected cutoff values detected all infants with culture-proven infections, and the sensitivity and specificity to detect clinical infections at the initial evaluation seemed to be acceptable for the selected study population at mild to moderate risk for bacterial infection. Furthermore, repeated determinations of IL-8 and CRP at the selected cutoff values achieved a sensitivity and a negative predictive value of 100%.17
The standard group was evaluated and treated according to the standard guidelines for evaluation of newborn infants with suspected bacterial infection of each center. IL-8 measurements were not included in the standard evaluation. Instead, the decision to start antibiotic therapy was based on the combination of clinical signs, obstetric risk factors, and laboratory values that included CRP measurements in 7 of 8 centers.
In both groups, the same empiric antibiotic therapy was administered according to the resistance profiles encountered at each center. Physicians were allowed to start antibiotic therapy in infants of both groups at any time if they believed that the infant's condition deteriorated rapidly. The turnaround time for the availability of test results of IL-8 and CRP was <2 hours.
Follow-up Evaluation of Patients
Patients were clinically reevaluated at least twice during the first 48 hours after the initial evaluation and were followed for at least 7 days. Furthermore, clinical and laboratory evaluations were repeated whenever new clinical signs of infection developed or clinical signs persisted in initially untreated infants, and antibiotic therapy was initiated as indicated by the assigned diagnostic algorithm. Any antibiotic therapy during these first 7 days after study entry was recorded.
Classification of Patients at the End of Follow-up
For evaluating the initial treatment decision, all infants were classified at the end of follow-up as having blood culture–proven infection, clinical infection, or no infection. Blood culture–proven infection was defined by at least 1 clinical sign and a positive blood culture and an increase of CRP to >10 mg/L within 12 to 60 hours after the initial evaluation. The increase in CRP in addition to a positive culture result was required as an effort to exclude false-positive culture results because of contamination. Clinical infection was defined by at least 1 clinical sign and an increase of CRP to >10 mg/L within 12 to 60 hours after the initial evaluation. All patients who did not meet these definitions of blood culture–proven or clinical infection were considered as having no infection. Furthermore, 20 infants in the IL-8 group and 26 infants in the standard group with mild signs on initial clinical evaluation and with maximum CRP values >10 mg/L (11–36 mg/L) on repeated laboratory evaluations were also classified as having no infection because they did not receive antibiotic therapy and recovered without treatment. Infants who did not receive antibiotic therapy immediately after the initial evaluation but who were classified retrospectively as having a blood culture–proven or a clinical infection were classified as "initially missed infections."
Biochemical Determinations
CRP was measured by rate nephelometry in lithium-heparin plasma or serum samples.10–12,29 IL-8 was measured in 50 µL of lithium-heparin plasma by a random-access chemiluminescence immunoassay (Immulite; DPC, Los Angeles, CA).17 The IL-8 assay had a lower limit of detection of 5 pg/mL and was calibrated up to 7500 pg/mL. The coefficient of variation was <5% at 95 pg/mL.
Statistical Analysis
The primary outcome variables were 1) the proportion of infants treated with antibiotics within 7 days after study entry of all infants enrolled for suspected infection (referred to as the proportion of "infants treated with antibiotics") and 2) the proportion of infected infants who did not receive antibiotic therapy immediately after initial evaluation of all infected infants (referred to as the proportion of "initially missed infections"). For differences of these proportions between groups, 1-sided 95% confidence intervals (CIs) were calculated. The corresponding hypotheses were tested as a priori–ordered hypotheses at a significance level of 
= .05; therefore, no correction for multiple testing was necessary.
The first study hypothesis—that the proportion of infants who were treated with antibiotics is lower in the IL-8 group than in the standard group—was evaluated with a 
2 test, and the "number needed to treat" with a 95% CI was calculated. The second study hypothesis—that the proportion of initially missed infections in the IL-8 group is at the most 3% higher than in the standard group—was evaluated by a 1-sided test of equivalence. The data were evaluated on an intention-to-treat basis.
Sample size calculations were based on a retrospective evaluation of 297 newborn infants who had been admitted to the Division of Neonatology at Ulm in 1997 for suspected early-onset neonatal bacterial infection. Assuming a significance level = .05 and a power ß = .80, a proportion of infants who were treated with antibiotics of 39% in the IL-8 group and of 66% in the standard group, 49 patients were necessary in each group to demonstrate that the proportion of infants who were treated with antibiotics is lower in the IL-8 group. For testing the second hypothesis, the sample size calculation was based on the following assumptions: a significance level = .05, a power ß = .80, a proportion of initially missed infections of 4% in the IL-8 group and of 9% in the standard group, and an equivalence limit of 3%. On the basis of these assumptions, a sample size of 207 patients with infection in each group was required to demonstrate 1-sided equivalence of the proportions of initially missed infections. Assuming a rate of bacterial infection of 18% in the study population, a total of 1150 patients had to be enrolled into the study.
A predefined adaptive interim analysis31 (with 0 = .3) was performed after the first year of the trial when 356 patients had been enrolled. At that time, the proportion of infants who were treated with antibiotics was 69 of 175 in the IL-8 group and 111 of 181 in the standard group (P < .0001); consequently, the first study hypothesis was confirmed at this point. The proportion of initially missed infections was 8 of 45 in the IL-8 group and 14 of 52 in the standard group (P1 = .074); consequently, the second study hypothesis was not confirmed at this analysis and the study had to be continued. On the basis of these results, a new sample size calculation revealed that another 353 patients were required in each group to prove 1-sided equivalence of the proportion of initially missed infections in both groups. In July 2000, it became apparent that the required number of patients (1062 total) would be enrolled by the end of the year. The end of the study therefore was set at December 31, 2000. For evaluating the diagnostic accuracy of IL-8 and CRP, the sensitivity and the specificity with corresponding exact 95% CI were calculated.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
|
|
|
|
Values of sensitivity and specificity of IL-8 and CRP in predicting culture-proven or clinical infection are shown in Table 4. IL-8 had a higher sensitivity in infants who were tested within the first 12 hours of life than in the whole group of infants, whereas CRP had a low sensitivity at this early age.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
|
In 1981, Philip4 suggested using laboratory markers to reduce unnecessary antibiotic therapy in newborn infants. Since then, numerous studies have shown that CRP and proinflammatory cytokines such as IL-8 or IL-6 are increased in infants with sepsis.10–20,23,25–27,29,30 Our preliminary nonrandomized, single-center pilot studies suggested that a diagnostic strategy that is based on measurements of IL-8 and CRP in addition to clinical judgment may reduce unnecessary antibiotic therapy without increasing the risk for adverse outcome.17,25 This multicenter, randomized, controlled trial indeed proved that a reduction of antibiotic therapy is possible in comparison with the standard diagnostic algorithm of each center, without increasing the number of infants missed on initial evaluation. We speculate that applying this diagnostic strategy will change the clinical treatment of many infants who receive unnecessary antibiotic therapy and may contribute to a reduction of the development and spread of antibiotic-resistant bacteria.
This study confirms that IL-8 and CRP should always be measured together to optimize the sensitivity. Although IL-8 has a higher sensitivity when measured in the first 12 hours of life and CRP has a higher sensitivity later, the sensitivity for bacterial infection in either age group is greater when both parameters are combined (Table 4). Furthermore, 4 of the 13 infants with blood culture–proven infection had normal CRP values and 3 of the 13 had normal IL-8 values on initial blood sampling (Table 2), whereas all 13 infants were detected by combining both markers. This is in agreement with previously published data,16,17,23–27 and this pattern is probably a result of the short half-life of raised IL-8 in plasma early in the course of neonatal bacterial infections16,17 and the delayed increase of CRP 12 to 48 hours after the onset of infection.10–12,17
Studies of laboratory markers for the diagnosis of early-onset neonatal bacterial infections are frequently limited by a low rate of blood culture–proven infections in the study population ranging between 2.0% and 10.2%.10,12–17,20,27 In our study, only 13 of 1291 infants with suspected bacterial infection had a blood culture–proven infection. This low rate of blood culture–proven infection may have several reasons: 1) the volume of blood available for the blood culture was too small to detect low-grade bacteremia35; 2) 21% of the mothers in this study received antenatal antibiotic therapy, which may have inhibited bacterial growth in the infant's blood cultures; and 3) the study design specifically excluded infants with marked clinical signs so that antibiotic therapy could be initiated immediately in these very ill infants; this excluded the group of newborn infants with the highest probability of yielding positive blood culture results from the study.
All infants with blood culture–proven infection were detected by the combination of IL-8 and CRP at the initial evaluation. Because many clinically ill infants with laboratory evidence of infection and clinical response to antibiotic therapy do not have positive blood cultures,1,10,13–15,17,19,25 clinical infections were defined in this study as a combination of clinical and laboratory evidence, as has been done previously.10,13–15,17,21,25 To include all infected infants with certainty so that withholding antibiotics in infants who did not meet the criteria would be safe, we chose a broad definition of infection with a low threshold. This broad definition is probably also the reason for the relative high proportion of infants classified as missed on initial evaluation.
There were differences between the participating centers both in the patient population and in the threshold of the individual physicians to start antibiotic therapy immediately without waiting for results of diagnostic tests. Although all centers used similar approaches in the evaluation of infants with suspected infection—including clinical signs, obstetric risk factors, and laboratory values—the standard diagnostic algorithm was not exactly identical in all centers, but within each center, a similar number of patients were assigned to the standard and the IL-8 group. That the proposed diagnostic algorithm that includes measurements of IL-8 and CRP reduced unnecessary antibiotic therapy compared with several slightly differing standard diagnostic algorithms and despite different clinical settings at the participating centers indicates that this new approach is feasible and effective in clinical practice.
|
CONCLUSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
The enrollment criteria of this study excluded infants who were in septic shock and in rapidly deteriorating clinical condition, ie, infants who were thought to be at the greatest risk of having bacterial infections. It therefore is important to note that the suggested diagnostic algorithm is not intended for and may not be indiscriminately applied to those who are high-risk, very sick infants. In contrast, in those high-risk infants, immediate antibiotic therapy is indicated.
|
ACKNOWLEDGMENTS |
|---|
This study was supported in part by grant P.575 from the Center for Applied Clinical Studies of the University of Ulm, Germany. DPC (Los Angeles, CS) provided the Immulite automated analyzers and the kits for determination of IL-8 and sponsored the initial meeting of the investigators. However, DPC did not have any involvement in the study design, data collection, data interpretation, or writing of this report. Mikael Norman was supported by the Swedish Research Council.
We thank Sabine Schmid and Barbara Köper for maintaining the study database.
|
FOOTNOTES |
|---|
Reprint requests to (A.F.) Department of Pediatrics, Division of Neonatology and Pediatric Critical Care, University Children's Hospital, Prittwitzstrasse 43, 89075 Ulm, Germany. E-mail: axel.franz{at}ukb.uni-bonn.de
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  M. R. Bentlin, L. M. S. de Souza Rugolo, A. R. Junior, M. Hashimoto, and J. C. Lyra Is Urine Interleukin-8 Level a Reliable Laboratory Test for Diagnosing Late Onset Sepsis in Premature Infants? J Trop Pediatr, December 1, 2007; 53(6): 403 - 408. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. C.P. Franco, D. M.J. Christofalo, A. L. Sawaya, S. A. Ajzen, and R. Sesso Effects of Low Birth Weight in 8- to 13-Year-Old Children: Implications in Endothelial Function and Uric Acid Levels Hypertension, July 1, 2006; 48(1): 45 - 50. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. B. Lande, N. L. Carson, J. Roy, and C. C. Meagher Effects of Childhood Primary Hypertension on Carotid Intima Media Thickness: A Matched Controlled Study Hypertension, July 1, 2006; 48(1): 40 - 44. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 U K Mishra, S E Jacobs, L W Doyle, and S M Garland Newer approaches to the diagnosis of early onset neonatal sepsis. Arch. Dis. Child. Fetal Neonatal Ed., May 1, 2006; 91(3): F208 - F212. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J.-H. Weitkamp and J. L. Aschner Diagnostic Use of C-Reactive Protein (CRP) in Assessment of Neonatal Sepsis NeoReviews, November 1, 2005; 6(11): e508 - e515. [Full Text] [PDF]
|
|
|
|
 |
|
 K. Skogstrand, P. Thorsen, B. Norgaard-Pedersen, D. E. Schendel, L. C. Sorensen, and D. M. Hougaard Simultaneous Measurement of 25 Inflammatory Markers and Neurotrophins in Neonatal Dried Blood Spots by Immunoassay with xMAP Technology Clin. Chem., October 1, 2005; 51(10): 1854 - 1866. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2004 American Academy of Pediatrics. All rights reserved. |
||
|
||
