- Email: [email protected]
Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis
Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis Michel Mikhael, MD1, L. Steven Brown, MS2, and Charles R. Rosenfeld, MD1 Objective To validate established neonatal neutrophil reference ranges (RRs) and determine the utility of serial measurements of neutrophil values in the first 24 hours to predict the absence of neonatal early-onset sepsis (EOS).
Study design Retrospective study of 2073 admissions to the neonatal intensive care unit (2009-2011). Neonates were classified as blood culture-positive, proven EOS (n = 9), blood culture-negative but clinically suspect EOS (n = 292), and not infected (n = 1292). Neutrophil values from 745 not-infected neonates without perinatal complications were selected to validate RR distributions. Positive and negative predictive values were calculated; area under receiver operating characteristic curves (AUCs) were constructed to predict the presence or absence of EOS. Neutrophil value scores were established to determine whether serial neutrophil values predict the absence of EOS. Results Seventy-seven percent of admissions to the neonatal intensive care unit were evaluated for EOS: 9 (0.56%) had proven EOS with positive blood culture #37 hours; 18% had clinically suspect EOS. Neutropenia occurred in preterm neonates, and nonspecific neutrophilia was common in uninfected neonates. The distribution of neutrophil values differed significantly between study groups. The specificity for absolute total immature neutrophils and immature to total neutrophil proportions was 91% and 94%, respectively, with negative predictive value of 99% for proven and 78% for proven plus suspect EOS. Absolute total immature neutrophils and immature to total neutrophil proportions had the best predictability for EOS >6 hours postnatal with an AUC 0.8. Neutrophil value scores predicted the absence of EOS with AUC of 0.9 and 0.81 for proven and proven plus suspect EOS, respectively. Conclusion Age-dependent neutrophil RRs remain valid. Serial neutrophil values at 0, 12, and 24 hours plus blood culture and clinical evaluation can be used to discontinue antimicrobial therapy at 36-48 hours. (J Pediatr 2014;164:522-8).
T
he evaluation for early-onset sepsis (EOS) within 72 hours postnatally occurs commonly in the neonatal intensive care unit (NICU).1 Early warning signs and clinical manifestations, however, often are few, subtle, or nonspecific.2 Therefore, the decision to treat often is decided on the basis of maternal, prenatal, intrapartum, and neonatal risk factors.3 Although the origin of respiratory distress may be benign, cannot be differentiated from group B streptococcal (GBS) sepsis/pneumonia clinically or by the use of chest radiographs.4,5 Thus, the development of a rapid screening tool for EOS has been sought, including the evaluation of circulating neutrophils,5-9 C-reactive protein,10 inflammatory markers,11,12 and numerous other tests. Most of these biomarkers have relatively poor sensitivity or specificity compared with a positive blood culture and proven sepsis.13 None provides immediate results. Because EOS can be associated with considerable mortality and morbidity, delaying antibiotic therapy for the results of a screening test may be contraindicated. The approach to evaluating EOS generally includes examination of maternal/neonatal clinical factors, initiation of antibiotic therapy after obtaining a single blood culture (and cerebrospinal fluid cultures when indicated), and obtaining another diagnostic test (eg, neutrophil count or C-reactive protein).14 The results of blood culture usually are negative; nonetheless, antibiotic treatment sometimes is administered for 5-10 days for a clinically suspect infection. These infants may have unnecessarily prolonged antibiotic exposure in many cases.2 Notably, empiric antimicrobial therapy for EOS in very low birth weight neonates has been associated with an increased risk of necrotizing enterocolitis or death.15 Rather than develop a tool that delays the initiation of antimicrobial therapy, it would be preferable to identify a simple biomarker with high specificity and negative predictive value (NPV) that, together with a negative blood culture and improved clinical status, would exclude EOS, permitting the early cessation of antimicrobial therapy without endangering the neonate.
ATI ATN AUC CBC EOS GBS I:T
Absolute total immature neutrophils Absolute total neutrophils Area under the curve Complete blood count Early-onset sepsis Group B streptococcal Immature to total neutrophil proportion or ratio
NICU NPV NVS PPV PROM ROC RR
Neonatal intensive care unit Negative predictive value Neutrophil value score Positive predictive value Prolonged rupture of membranes Receiver operating characteristic Reference range
From the 1Division of Neonatal-Perinatal Medicine, Department of Pediatrics, University of Texas Southwestern Medical School; and 2Department of Health Systems Research, Parkland Health & Hospital System, Dallas, TX C.R. is supported by the George L. MacGregor Professorship. The authors declare no conflicts of interest. 0022-3476/$ - see front matter. Copyright ª 2014 Mosby Inc. All rights reserved. http://dx.doi.org/10.1016/j.jpeds.2013.10.080
522
Vol. 164, No. 3 March 2014 Neonatal neutrophil reference ranges (RRs) from birth to 28 days postnatal age were established >30 years ago.9 The RRs for absolute total neutrophils (ATN) and absolute total immature neutrophils (ATI) and the immature to total neutrophil proportion or ratio (I:T) are dynamic during the first 96 hours postnatally. Thus, a single value uncorrected for postnatal age may not accurately diagnose EOS. Several perinatal variables alter the distribution of neutrophils (eg, maternal hypertension is associated with neonatal neutropenia, and fetal asphyxia may result in grossly abnormal neutrophil values).9,16 Of importance, antibiotic exposure in the absence of infection does not alter the neutrophil values. When these factors are taken into consideration, the neutrophil RRs have value in indicating EOS.5,9,16 Mouzinho et al17 confirmed this in preterm neonates. Nonetheless, one complete blood count (CBC) uncorrected for age frequently is used, and the upper limits of the I:T RR often are elevated to increase specificity, which in turn may increase false-negative values. We initially used the neutrophil findings to screen for EOS5; we subsequently observed that serial ATN, ATI, and I:T values obtained before initiating antibiotics and 12 and 24 hours later plus blood culture were useful in determining when to stop antimicrobial therapy. This laboratory tool, however, has not been studied. The purpose of this study was to revalidate the published neutrophil RRs9 and determine whether “timed” serial neutrophil counts aid in ruling out EOS, thereby permitting the cessation of antibiotic therapy in neonates with transient clinical signs of infection, negative blood culture, and a diagnosis of “clinically suspect sepsis.” We also examined the time to a positive blood culture to determine whether antimicrobial exposure in neonates with “suspect EOS” could be shortened to #36 hours.
Methods This was a retrospective study of neonates who were admitted to the Parkland Hospital NICU, Dallas, TX, from May 2009 through April 2011. This is level-IIIC NICU with 15 000 deliveries and 1250 admissions annually. Inclusion criteria included neonates $23 weeks’ gestation undergoing a EOS evaluation #72 hours postnatal age on the basis of perinatal risk factors (eg, prolonged rupture of membranes [PROM], unexplained preterm birth, and clinical neonatal signs, including respiratory distress, seizures, and hypoglycemia). All symptomatic neonates were admitted to the NICU. The standard of care (Figure 1; available at www.jpeds.com) included 2 blood cultures collected from different sites within 2-3 minutes of each other, the drawing of serial CBCs with manual differential white blood cell counts at initiation of the EOS evaluation (0 hours) and 12 and 24 hours later, and promptly starting antimicrobial therapy. If the patient’s blood culture had no bacterial growth at 48 hours, neutrophil values were within the published RR,9 and clinical signs of infection were absent, antibiotics were stopped at 36-48 hours. Exclusion criteria included neonates
who were provided comfort care and those without an EOS evaluation #72 hours postnatally. Maternal and neonatal data included variables previously shown to affect neutrophil distribution #72 hours postnatally (eg, maternal hypertension and fever).4,9,18 Additional data included the following: (1) maternal age, race, ethnicity, gestational age at delivery, hypertension, and chorioamnionitis (maternal temperature >38 C); (2) intrapartum and delivery room variables (eg, PROM >18 hours, meconium-stained amniotic fluid, mode of delivery, intubation and/or chest compression with or without medications, Apgar scores, umbilical artery pH, base excess); and (3) neonatal birth weight, sex, intracranial hemorrhage, hemolytic disease, neutrophil values and age after birth, and result and time to positive blood culture. Data were obtained from an institutional review board–approved, validated database in existence >30 years. Additional data were obtained from chart reviews. The study was approved by the Institutional Review Board of the University of Texas Southwestern Medical School and Parkland Hospital. Neonates were categorized as proven EOS with 1-2 blood cultures growing a pathogenic organism and receiving at least 5-7 days of antimicrobial therapy; suspect EOS with negative blood culture, variable abnormal neutrophil values, clinical signs of sepsis/infection, and antimicrobial therapy 5-7 days; and noninfected neonates who were asymptomatic, had a negative blood culture, normal serial neutrophil values, and in whom antimicrobial therapy was stopped at 48 hours. To establish a control group to revalidate the neutrophil RR, we selected 745 noninfected neonates without exposure to prenatal, intrapartum, and postnatal factors known to affect neutrophil values.9,17,18 Neutrophil Value Score The neutrophil value score (NVS) was developed to assess the use of serial neutrophil values for determining the absence of EOS. It uses the sum of the ATN, ATI, and I:T values obtained at 0-6, 10-16, and 22-28 hours of life that were within the upper and lower ranges of each RR, resulting in a maximum score of 9. Neonates without a CBC for each time period were excluded. The Mann-Whitney U test was used to compare the NVS between groups. Receiver operating characteristic (ROC) curves were constructed to assess the accuracy of the NVS for determining the absence of EOS. Statistical Analyses ATN, ATI, and I:T values from noninfected neonates were used to validate the RR9 by examining their distribution outside the 10th and/or 90th percentiles by postnatal age. We similarly examined the neutrophil values for proven and suspect EOS. Comparisons were performed with Pearson c2. Demographics and clinical variables were analyzed by ANOVA and the c2 test. A Bonferroni correction was used in pair-wise comparisons of the 3 groups when ANOVA was P < .05. Sensitivity, specificity, and positive predictive values (PPVs) and NPVs were determined for neutrophil values <10th and >90th percentiles, comparing proven and proven plus suspect vs noninfected controls. ROC curves 523
THE JOURNAL OF PEDIATRICS
www.jpeds.com
were constructed for abnormal neutrophilic values at each time period when a CBC was obtained to identify neutrophil values and time periods with the best predictability for EOS. Statistical analyses were performed with IBM SPSS version 19 (SPSS Institute, Chicago, Illinois).
Results There were 2073 NICU admissions during the study (Figure 2; available at www.jpeds.com); 77% were evaluated for EOS, including 1292 noninfected neonates, 292 with suspect EOS, and 9 with blood culture proven EOS (Table I). There were 255 neonates #30 weeks’ gestational age, and they included 14% of noninfected, 24% of suspect, and 22% of blood culture–proven EOS. Neonates with proven EOS had 25 CBCs obtained, clinically suspect neonates had 860, and noninfected controls had 2132; neonates averaged 2.9 CBCs/neonate during the first 24 hours of evaluation. There were no differences in maternal variables or infant’s gestational age, birth weight, sex, or route of delivery (Table II; available at www.jpeds.com). Neonates with suspect and proven EOS, however, had a lower 5-minute Apgar score, more asphyxia (Apgar score <5 at 5 minutes), required resuscitation at birth (intubation and/or chest compressions), and were more likely delivered to women with PROM, meconiumstained amniotic fluid, and chorioamnionitis. Nine neonates had at least one positive blood culture (Table II). Four had 2, whereas 5 (56%) had only 1 of 2. Blood cultures were positive in 7 of 9 neonates within 24 hours, with an average of 18 hours. All blood cultures were positive by 37 hours. Patient 4 was clinically ill, with a diagnosis of pneumonia via chest radiograph. Distribution of Neutrophil Values Neutrophil values from uninfected control neonates (2123 counts) were plotted on appropriate RR. Forty-nine percent of ATN values were <10th or >90th percentiles; 25% were neutropenic. In contrast, only 10% of ATI and 6% of I:T
Vol. 164, No. 3 values were >90th percentile, suggesting a normal distribution (Figure 3; available at www.jpeds.com). Because neutropenia might reflect preterm neonates,17 we excluded values from 449 uninfected neonates <37 weeks’ gestation. The incidence of neutropenia decreased to 9% (Figure 3, B). Notably, 24% of control ATN values were >90th percentile, demonstrating nonspecific neutrophilia. In neonates with proven EOS, 60% of ATN values were outside the RR, and 40% were neutropenic (data not shown). In addition, 36% of ATI and 56% of I:T values were outside the RR. Thus, the distribution of neutrophil findings for neonates with proven EOS differed significantly from the RR (P < .05). Neonates with suspect EOS had 860 CBCs; 66%, 38%, and 36% of ATN, ATI, and I:T values, respectively, were outside the RR. Their values also differed significantly from the RR (P < .05), consistent with previous studies.16 When abnormal values for the 3 groups were compared at each time period, neonates with EOS were more likely than noninfected controls to have abnormal neutrophil values (P < .05, Pearson c2). Prediction of EOS To determine the predictability of each neutrophil result corrected for postnatal age for the presence and/or absence of EOS, we calculated sensitivity, specificity, PPV, and NPV for each neutrophil/band value for proven and proven plus suspect vs noninfected neonates by using all values (Table III). ATN values in proven and proven plus suspect EOS had modest sensitivity and specificity (<66%) but poor PPV. Notably, the ATN NPV was 99% and 78%, respectively. ATI and I:T alone had modest-to-poor sensitivity and PPV, but the specificity and NPV ranged from 78% to 99%. We next examined 3017 neutrophil values from the 3 study groups; 2741 (91%) were collected within the recommended time periods of 0-6, 10-16, and 22-28 hours (see Methods). We constructed ROC curves for each time period and determined the area under the curve (AUC) for ATN, ATI, and I:T from the uninfected control group vs proven and proven plus
Table I. Demographics, clinical characteristics, and distribution of neonates without and with suspect and proven EOS Maternal age, years Gestational age, weeks Birth weight, g Male sex Cesarean delivery 5-minute Apgar score* Evidence of asphyxia Resuscitation during delivery Maternal hypertension PROM Chorioamnionitis MSAF
Not infected (n = 1292)
Suspect (n = 292)
Proven (n = 9)
P value
27.2 6.4 34.9 4.1 2447 905 702 (54) 755 (58) 8 (7-9)† 111 (9)† 176 (14)† 347 (27) 191 (15)† 100 (8)z 207 (16)z
27.0 6.6 34.8 5.5 2510 1152 154 (53) 162 (56) 8 (6-9) 50 (17) 63 (22) 59 (20) 51 (18) 72 (24) 78 (27)
23.8 4.7 35.7 5.3 2799 1152 3 (33) 3 (33) 7 (6-9) 2 (22) 3 (33) 3 (33) 4 (44) 2 (22) 4 (44)
>.05 >.05 >.05 >.05 >.05 <.01 <.01 <.01 >.05 .03 <.01 <.01
MSAF, meconium-stained amniotic fluid. Data are means SD. Data are number of neonates with a presenting variable, and values in parentheses are percent of each study group except for Apgar score. *Data are medians and IQR. †P < .05 control compared with suspect. zP < .05 control compared with proven.
524
Mikhael, Brown, and Rosenfeld
ORIGINAL ARTICLES
March 2014
Table III. Prediction of EOS using individual neutrophil parameters in neonates with proven and proven plus suspect EOS Proven ATN ATI I:T Proven plus suspect ATN ATI I:T
Sensitivity, %
Specificity, %
PPV, %
NPV, %
60 36 56
51 91 94
1.4 4.5 9.8
99 99 99
66 38 37
51 91 94
36 64 72
78 78 78
suspect EOS (Figure 4). At no time was the AUC for ATN >0.58 (Table IV; available at www.jpeds.com). At 0-6 hours, the ATI and I:T AUC were generally #0.7; this improved over the ensuing 24 hours and was >0.7 for both at 22-28 hours. Because neutrophil values had high specificity and NPV (Table III), we assessed all 9 serial neutrophil values
collected over 24 hours by using the NVS to predict the absence of EOS. Seventy-six percent of neonates (n = 796) had a CBC at each time period, including 6 with proven EOS, 229 with suspect, and 561 who were noninfected. We calculated the sum of neutrophil values within the 90th and 10th percentiles of the RR, using 9 as a maximum score. The difference in NVS between proven and proven plus suspect EOS and noninfected neonates was P < .01, MannWhitney U test. The NVS had an AUC of 0.90 and 0.81 for proven and proven plus suspect, respectively, to predict no EOS (Figure 5). The specificity for proven EOS with NVP scores 6, 7, or 8 was 100% for no EOS, and proven plus suspect had values of 78%, 92%, and 97%, respectively. When we examined the NVS in the 292 neonates with clinically suspect EOS (ie, negative blood culture but antimicrobial treatment 5-7 days), 98 (43%) had a NVS $6, and the predominant characteristic of this group was preterm birth and low birth weight (data not shown). If the NVS was increased to $7, 51 neonates (22%) may have been spared from prolonged antibiotic therapy. Notably, 18 had scores of 8 or 9.
Figure 4. ROC curves for neutrophil measurements for neonates A-C, without vs proven EOS and D-F, without vs proven plus suspect EOS at A, D, 0-6 hours B, E, 10-16 hours and C, F, 22-28 hours. Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis
525
THE JOURNAL OF PEDIATRICS
www.jpeds.com
Vol. 164, No. 3
Figure 5. ROC curves representing NVS for neonates without EOS vs with A, proven and B, proven plus suspect EOS.
Discussion Neonatal EOS is associated with high morbidity and mortality and generally approached emergently. Withholding antimicrobial therapy pending results of a screening test may be contraindicated. Few investigators have examined the blood culture in conjunction with a laboratory test to rule out EOS, which may limit the duration vs onset of antimicrobial therapy in the large population of at-risk neonates. We examined the use of serial neutrophil values corrected for postnatal age as an adjunct to 2 blood cultures plus clinical signs in predicting the “absence” of EOS within 24-36 hours of initiating a sepsis evaluation, thereby decreasing the duration of antimicrobial therapy, duration of hospitalization, and cost of health care. There was very good NPV and specificity for serial neutrophil values in the absence of EOS when the established neutrophil RR was used.9 We developed the NVS to score serial neutrophil values corrected for postnatal age and observed very goodto-excellent predictability for the absence of EOS. Importantly, the neutrophil RRs were shown to remain valid when we corrected them for gestational age. Thus, the use of serial neutrophil counts in conjunction with blood culture can decrease antibiotic exposure in a large population of neonates evaluated for EOS without risking adverse outcomes by waiting. Less than 1% of live born neonates have blood culture– proven EOS,1 which may reflect increased prenatal identification and treatment of maternal GBS before birth and/or increased maternal antimicrobial use in the intrapartum period, resulting in negative neonatal blood culture.14,19 If the latter is true, it is exceedingly difficult to separate neonates with partially treated EOS caused by maternal antibiotics and those without a real risk of EOS, often resulting in prolonged antimicrobial treatment in the latter. A low blood volume collected for blood culture is also known to reduce the number of positive blood cultures and in526
crease the occurrence of false-negative blood cultures.20 Moreover, as noted previously, clinical signs often are inconclusive and undistinguishable from benign disorders such as transient tachypnea. Our study included all symptomatic neonates delivered at our institution, yet only 9, or <1%, had positive blood culture, whereas 18% of treated neonates had risk factors for EOS and negative blood culture. Notably, 67% of proven EOS had GBS sepsis and 22% Escherichia coli, and 78% had a positive blood culture reported within 24 hours. Because only 1 of 2 blood cultures were positive in 5 cases, 28% might have been incorrectly considered false negative if only a single blood culture had been collected,20,21 supporting the use of 2 blood cultures obtained from different sites in assessing EOS. Neutrophil values initially were suggested as adjuncts in the diagnoses of neonatal EOS but were not corrected for the effects of perinatal variables or postnatal age.6-8 Manroe et al5,9 subsequently established neutrophil RR, demonstrating that when perinatal variables and postnatal age were accounted for, ATN, ATI, and I:T had high sensitivity for diagnosing EOS that was primarily caused by GBS. However, the specificity was even better, that is, >90% for the absence of EOS.5,9,22 Validation of these RR has not been examined in detail in >30 years. We observed that the original neutrophil RR remain valid because no more than 10% of counts for any neutrophilic measurements were >90th or <10th percentiles after correcting for postnatal age and preterm birth.17,23 Neutrophilia, however, occurred in 24% of noninfected neonates. This finding has been suggested to be due to high altitude24-26; however, Dallas is 450 ft above sea level. These neonates were evaluated because of clinical signs of EOS; thus, they are not normal. Their neutrophilia is more likely attributable to a nonspecific stress response and possibly excessive catecholamines rather than high altitude. These neonates would not have been included in the original RR because they are not considered “normal.” This may Mikhael, Brown, and Rosenfeld
ORIGINAL ARTICLES
March 2014 explain why ATN values alone are poor predictors of EOS.24 The Centers for Disease Control and Prevention guidelines for prevention of perinatal GBS infection recommend a CBC at birth and/or 6-12 hours postnatally for neonates born to women colonized with GBS and inadequately treated.27 Neutrophil measurements at 0-6 hours of life were not diagnostic for EOS, consistent with recent observations in a larger population.28 Their reliability, however, improved with increasing postnatal age. Thus, serial neutrophil values corrected for postnatal age may be preferred to determine the risk of EOS. Little evidence exists that a single CBC has any validity, and waiting may not be safe. It is important to note that the RRs were derived from manual differentials; to date, we found no studies in which the authors compared manual and computer-generated differentials via flow cytometry. The reliability of a CBC obtained <12 hours postnatally was observed to be quite poor. Furthermore, the sensitivity and PPV were less reliable than the specificity and NPV in predicting the absence of EOS. This improved with increasing postnatal age, supporting the use of serial neutrophil values corrected for postnatal age. The NVS uses all 9 neutrophil values and determines the proportion within the published RR for each neutrophil measurement. The reliability exceeds 80% with 6 normal values and >90% with $7 normal values. Thus, the NVS can be used to stop antibiotics in neonates with clinical risk factors, but transient or mild clinical signs and negative blood culture at 24-36 hours, potentially decreasing the number receiving antibiotics >24 hours by 50%. We were unable to determine whether the amelioration of clinical signs by 6-8 hours of age in this retrospective study might serve as an independent sign of a low probability for EOS; however, we would suggest that in our population this might have reflected the initiation of antibiotics within the first 2-4 hours of life and thus clinical improvement as the result of initiation of treatment. Notably, 22%-43% of the suspect neonates who received 5-7 days of antibiotics had a NSV $6-7. When we excluded ATN measurements at 0-6 hours postnatal age, the AUC had lower specificity (0.8), reflecting the high NPV of ATN that increases the statistical weight of the NVS. Prospective studies are needed in large populations to further validate the NVS and define neonates with suspect EOS. Furthermore, the number of blood culture positive neonates was low, which can only be addressed in larger clinical studies. We have demonstrated that the use of the established neutrophil RR9 and serial neutrophil measurements obtained over 24 hours after initiating an evaluation for EOS have high specificity and NPV in a population of term and preterm neonates. In conjunction with 2 blood cultures, this could allow cessation of antibiotic therapy in a large group of neonates commonly evaluated for EOS. In a NICU that treats all neonates evaluated for EOS for 5-7 days, this will decrease the duration of antimicrobial therapy 3-4 days as well as the duration of hospitalization for neonates $36 weeks’ gesta-
tion (50%), resulting in a conservative estimate of savings of $2.5 million annually in our NICU. It also is likely to decrease fungal infections in the NICU. We encourage others to examine the algorithm noted (Figure 1) and to consider studying this approach. n Submitted for publication Jul 3, 2013; last revision received Oct 9, 2013; accepted Oct 29, 2013. Reprint requests: Charles R. Rosenfeld, MD, Department of Pediatrics, UT Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75390-9063. E-mail: [email protected]
References 1. Escobar GJ. The neonatal “sepsis work-up”: personal reflections on the development of an evidence-based approach toward newborn infections in a managed care organization. Pediatrics 1999;103:360-73. 2. Schelonka RL, McCracken G. Bacterial and fungal infections. In: Avery GB, MacDonald MG, Seshia MMK, Mullett MD, eds. Avery’s neonatology: pathophysiology & management of the newborn. 6th ed. Philadelphia (PA): Lippincott Williams & Wilkins; 2005. p. 1235-73. 3. Mukhopadhyay S, Eichenwald EC, Puopolo KM. Neonatal early-onset sepsis evaluations among well-appearing infants: projected impact of changes in CDC GBS guidelines. J Perinatol 2013;33:198-205. 4. Ablow RC, Driscoll SG, Effmann EL, Gross I, Jolles CJ, Uauy R, et al. A comparison of early-onset group B steptococcal neonatal infection and the respiratory-distress syndrome of the newborn. N Engl J Med 1976; 294:65-70. 5. Manroe BL, Rosenfeld CR, Weinberg AG, Browne R. The differential leukocyte count in the assessment and outcome of early-onset neonatal group B streptococcal disease. J Pediatr 1977;91:632-7. 6. Xanthou M. Leucocyte blood picture in ill newborn babies. Arch Dis Child 1972;47:741-6. 7. Gregory J, Hey E. Blood neutrophil response to bacterial infection in the first month of life. Arch Dis Child 1972;47:747-53. 8. Zipursky A, Palko J, Milner R, Akenzua GI. The hematology of bacterial infections in premature infants. Pediatrics 1976;57:839-53. 9. Manroe BL, Weinberg AG, Rosenfeld CR, Browne R. The neonatal blood count in health and disease. I. Reference values for neutrophilic cells. J Pediatr 1979;95:89-98. 10. Philip AG. Response of C-reactive protein in neonatal group B streptococcal infection. Pediatr Infect Dis 1985;4:145-8. 11. Messer J, Eyer D, Donato L, Gallati H, Matis J, Simeoni U. Evaluation of interleukin-6 and soluble receptors of tumor necrosis factor for early diagnosis of neonatal infection. J Pediatr 1996;129:574-80. 12. Santana C, Guindeo MC, Gonzalez G, Garcia-Munoz F, Saavedra P, Domenech E. Cord blood levels of cytokines as predictors of early neonatal sepsis. Acta Paediatr 2001;90:1176-81. 13. Malik A, Hui CP, Pennie RA, Kirpalani H. Beyond the complete blood cell count and C-reactive protein: a systematic review of modern diagnostic tests for neonatal sepsis. Arch Pediatr Adolesc Med 2003;157:511-6. 14. Polin RA. Management of neonates with suspected or proven early-onset bacterial sepsis. Pediatrics 2012;129:1006-15. 15. Cotten CM, Taylor S, Stoll B, Goldberg RN, Hansen NI, Sanchez PJ, et al. Prolonged duration of initial empirical antibiotic treatment is associated with increased rates of necrotizing enterocolitis and death for extremely low birth weight infants. Pediatrics 2009;123:58-66. 16. Engle WD, Rosenfeld CR. Neutropenia in high-risk neonates. J Pediatr 1984;105:982-6. 17. Mouzinho A, Rosenfeld CR, Sanchez PJ, Risser R. Revised reference ranges for circulating neutrophils in very-low-birth-weight neonates. Pediatrics 1994;94:76-82. 18. Jackson GL, Engle WD, Sendelbach DM, Vedro DA, Josey S, Vinson J, et al. Are complete blood cell counts useful in the evaluation of asymptomatic neonates exposed to suspected chorioamnionitis? Pediatrics 2004;113:1173-80.
Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis
527
THE JOURNAL OF PEDIATRICS
www.jpeds.com
19. Ottolini MC, Lundgren K, Mirkinson LJ, Cason S, Ottolini MG. Utility of complete blood count and blood culture screening to diagnose neonatal sepsis in the asymptomatic at risk newborn. Pediatr Infect Dis J 2003;22:430-4. 20. Schelonka RL, Chai MK, Yoder BA, Hensley D, Brockett RM, Ascher DP. Volume of blood required to detect common neonatal pathogens. J Pediatr 1996;129:275-8. 21. Pierce JR, Merenstein GB, Stocker JT. Immediate postmortem cultures in an intensive care nursery. Pediatr Infect Dis 1984;3:510-3. 22. Engle WD, Rosenfeld CR, Mouzinho A, Risser RC, Zeray F, Sanchez PJ. Circulating neutrophils in septic preterm neonates: comparison of two reference ranges. Pediatrics 1997;99:E10. 23. Faix RG, Hric JJ, Naglie RA. Neutropenia and intraventricular hemorrhage among very low birth weight (less than 1500 grams) premature infants. J Pediatr 1989;114:1035-8.
528
Vol. 164, No. 3 24. Schmutz N, Henry E, Jopling J, Christensen RD. Expected ranges for blood neutrophil concentrations of neonates: the Manroe and Mouzinho charts revisited. J Perinatol 2008;28:275-81. 25. Carballo C, Foucar K, Swanson P, Papile LA, Watterberg KL. Effect of high altitude on neutrophil counts in newborn infants. J Pediatr 1991; 119:464-6. 26. Maynard EC, Reed C, Kircher T. Neutrophil counts in newborn infants at high altitude. J Pediatr 1993;122:990-1. 27. Verani JR, McGee L, Schrag SJ. Prevention of perinatal group B streptococcal disease–revised guidelines from CDC, 2010. MMWR Recomm Rep 2010;59:1-36. 28. Newman TB, Puopolo KM, Wi S, Draper D, Escobar GJ. Interpreting complete blood counts soon after birth in newborns at risk for sepsis. Pediatrics 2010;126:903-9.
Mikhael, Brown, and Rosenfeld
ORIGINAL ARTICLES
March 2014
Table II. Blood culture results, clinical descriptors, and distribution of neutrophil values for neonates with proven EOS Patient no.
EGA, weeks
1 2 3 4 5 6 7 8 9
39 41 25 33 40 37 38 38 30
Pathogen
No. of positive blood cultures
Incubation time, hours
Escherichia coli Escherichia coli GBS CONS GBS GBS GBS GBS GBS
2/2 2/2 2/2 1/2 1/2 1/2 1/2 2/2 1/2
18 12 24 37 23 15 24 15 31
Clinical description Respiratory failure Respiratory failure HMD Pneumonia Respiratory failure Respiratory failure Hypoxia, TGA HIE, PPHN PPHN, hypotension
Abnormal neutrophil values 7/9 4/9 4/9 3/9 6/9 6/9 0/3 4/9 4/9
CONS, coagulase-negative Staphylococcus; EGA, estimated gestational age; HIE, hypoxic ischemic encephalopathy; HMD, hyaline membrane disease; PPHN, persistent pulmonary hypertension of the newborn; TGA, transposition of great arteries.
Table IV. AUC for ATN, ATI, and I:T values from neonates with proven and proven plus suspect EOS at the 3 time periods used for analysis Proven 0-6 hours 10-16 hours 22-28 hours Proven plus suspect 0-6 hours 10-16 hours 22-28 hours
ATN
ATI
I:T
0.35 0.28 0.47
0.48 0.68 0.71
0.67 0.88 0.77
0.51 0.50 0.58
0.66 0.71 0.73
0.71 0.75 0.75
Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis
528.e1
THE JOURNAL OF PEDIATRICS
www.jpeds.com
Vol. 164, No. 3
Figure 1. Algorithm for identification, initiation, and cessation of antimicrobial therapy in neonates at risk of EOS.
Figure 2. NICU admissions and distribution of neonates with and without EOS (May 2009-April 2011). 528.e2
Mikhael, Brown, and Rosenfeld
March 2014
ORIGINAL ARTICLES
Figure 3. Distribution of neutrophil values within the Manroe et al RRs9 in 745 noninfected neonates between birth and 72 hours: A, ATN C, ATI and D, I:T. B, Distribution of ATN after correcting for preterm birth <37 weeks’ gestation.
Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis
528.e3
Study design Retrospective study of 2073 admissions to the neonatal intensive care unit (2009-2011). Neonates were classified as blood culture-positive, proven EOS (n = 9), blood culture-negative but clinically suspect EOS (n = 292), and not infected (n = 1292). Neutrophil values from 745 not-infected neonates without perinatal complications were selected to validate RR distributions. Positive and negative predictive values were calculated; area under receiver operating characteristic curves (AUCs) were constructed to predict the presence or absence of EOS. Neutrophil value scores were established to determine whether serial neutrophil values predict the absence of EOS. Results Seventy-seven percent of admissions to the neonatal intensive care unit were evaluated for EOS: 9 (0.56%) had proven EOS with positive blood culture #37 hours; 18% had clinically suspect EOS. Neutropenia occurred in preterm neonates, and nonspecific neutrophilia was common in uninfected neonates. The distribution of neutrophil values differed significantly between study groups. The specificity for absolute total immature neutrophils and immature to total neutrophil proportions was 91% and 94%, respectively, with negative predictive value of 99% for proven and 78% for proven plus suspect EOS. Absolute total immature neutrophils and immature to total neutrophil proportions had the best predictability for EOS >6 hours postnatal with an AUC 0.8. Neutrophil value scores predicted the absence of EOS with AUC of 0.9 and 0.81 for proven and proven plus suspect EOS, respectively. Conclusion Age-dependent neutrophil RRs remain valid. Serial neutrophil values at 0, 12, and 24 hours plus blood culture and clinical evaluation can be used to discontinue antimicrobial therapy at 36-48 hours. (J Pediatr 2014;164:522-8).
T
he evaluation for early-onset sepsis (EOS) within 72 hours postnatally occurs commonly in the neonatal intensive care unit (NICU).1 Early warning signs and clinical manifestations, however, often are few, subtle, or nonspecific.2 Therefore, the decision to treat often is decided on the basis of maternal, prenatal, intrapartum, and neonatal risk factors.3 Although the origin of respiratory distress may be benign, cannot be differentiated from group B streptococcal (GBS) sepsis/pneumonia clinically or by the use of chest radiographs.4,5 Thus, the development of a rapid screening tool for EOS has been sought, including the evaluation of circulating neutrophils,5-9 C-reactive protein,10 inflammatory markers,11,12 and numerous other tests. Most of these biomarkers have relatively poor sensitivity or specificity compared with a positive blood culture and proven sepsis.13 None provides immediate results. Because EOS can be associated with considerable mortality and morbidity, delaying antibiotic therapy for the results of a screening test may be contraindicated. The approach to evaluating EOS generally includes examination of maternal/neonatal clinical factors, initiation of antibiotic therapy after obtaining a single blood culture (and cerebrospinal fluid cultures when indicated), and obtaining another diagnostic test (eg, neutrophil count or C-reactive protein).14 The results of blood culture usually are negative; nonetheless, antibiotic treatment sometimes is administered for 5-10 days for a clinically suspect infection. These infants may have unnecessarily prolonged antibiotic exposure in many cases.2 Notably, empiric antimicrobial therapy for EOS in very low birth weight neonates has been associated with an increased risk of necrotizing enterocolitis or death.15 Rather than develop a tool that delays the initiation of antimicrobial therapy, it would be preferable to identify a simple biomarker with high specificity and negative predictive value (NPV) that, together with a negative blood culture and improved clinical status, would exclude EOS, permitting the early cessation of antimicrobial therapy without endangering the neonate.
ATI ATN AUC CBC EOS GBS I:T
Absolute total immature neutrophils Absolute total neutrophils Area under the curve Complete blood count Early-onset sepsis Group B streptococcal Immature to total neutrophil proportion or ratio
NICU NPV NVS PPV PROM ROC RR
Neonatal intensive care unit Negative predictive value Neutrophil value score Positive predictive value Prolonged rupture of membranes Receiver operating characteristic Reference range
From the 1Division of Neonatal-Perinatal Medicine, Department of Pediatrics, University of Texas Southwestern Medical School; and 2Department of Health Systems Research, Parkland Health & Hospital System, Dallas, TX C.R. is supported by the George L. MacGregor Professorship. The authors declare no conflicts of interest. 0022-3476/$ - see front matter. Copyright ª 2014 Mosby Inc. All rights reserved. http://dx.doi.org/10.1016/j.jpeds.2013.10.080
522
Vol. 164, No. 3 March 2014 Neonatal neutrophil reference ranges (RRs) from birth to 28 days postnatal age were established >30 years ago.9 The RRs for absolute total neutrophils (ATN) and absolute total immature neutrophils (ATI) and the immature to total neutrophil proportion or ratio (I:T) are dynamic during the first 96 hours postnatally. Thus, a single value uncorrected for postnatal age may not accurately diagnose EOS. Several perinatal variables alter the distribution of neutrophils (eg, maternal hypertension is associated with neonatal neutropenia, and fetal asphyxia may result in grossly abnormal neutrophil values).9,16 Of importance, antibiotic exposure in the absence of infection does not alter the neutrophil values. When these factors are taken into consideration, the neutrophil RRs have value in indicating EOS.5,9,16 Mouzinho et al17 confirmed this in preterm neonates. Nonetheless, one complete blood count (CBC) uncorrected for age frequently is used, and the upper limits of the I:T RR often are elevated to increase specificity, which in turn may increase false-negative values. We initially used the neutrophil findings to screen for EOS5; we subsequently observed that serial ATN, ATI, and I:T values obtained before initiating antibiotics and 12 and 24 hours later plus blood culture were useful in determining when to stop antimicrobial therapy. This laboratory tool, however, has not been studied. The purpose of this study was to revalidate the published neutrophil RRs9 and determine whether “timed” serial neutrophil counts aid in ruling out EOS, thereby permitting the cessation of antibiotic therapy in neonates with transient clinical signs of infection, negative blood culture, and a diagnosis of “clinically suspect sepsis.” We also examined the time to a positive blood culture to determine whether antimicrobial exposure in neonates with “suspect EOS” could be shortened to #36 hours.
Methods This was a retrospective study of neonates who were admitted to the Parkland Hospital NICU, Dallas, TX, from May 2009 through April 2011. This is level-IIIC NICU with 15 000 deliveries and 1250 admissions annually. Inclusion criteria included neonates $23 weeks’ gestation undergoing a EOS evaluation #72 hours postnatal age on the basis of perinatal risk factors (eg, prolonged rupture of membranes [PROM], unexplained preterm birth, and clinical neonatal signs, including respiratory distress, seizures, and hypoglycemia). All symptomatic neonates were admitted to the NICU. The standard of care (Figure 1; available at www.jpeds.com) included 2 blood cultures collected from different sites within 2-3 minutes of each other, the drawing of serial CBCs with manual differential white blood cell counts at initiation of the EOS evaluation (0 hours) and 12 and 24 hours later, and promptly starting antimicrobial therapy. If the patient’s blood culture had no bacterial growth at 48 hours, neutrophil values were within the published RR,9 and clinical signs of infection were absent, antibiotics were stopped at 36-48 hours. Exclusion criteria included neonates
who were provided comfort care and those without an EOS evaluation #72 hours postnatally. Maternal and neonatal data included variables previously shown to affect neutrophil distribution #72 hours postnatally (eg, maternal hypertension and fever).4,9,18 Additional data included the following: (1) maternal age, race, ethnicity, gestational age at delivery, hypertension, and chorioamnionitis (maternal temperature >38 C); (2) intrapartum and delivery room variables (eg, PROM >18 hours, meconium-stained amniotic fluid, mode of delivery, intubation and/or chest compression with or without medications, Apgar scores, umbilical artery pH, base excess); and (3) neonatal birth weight, sex, intracranial hemorrhage, hemolytic disease, neutrophil values and age after birth, and result and time to positive blood culture. Data were obtained from an institutional review board–approved, validated database in existence >30 years. Additional data were obtained from chart reviews. The study was approved by the Institutional Review Board of the University of Texas Southwestern Medical School and Parkland Hospital. Neonates were categorized as proven EOS with 1-2 blood cultures growing a pathogenic organism and receiving at least 5-7 days of antimicrobial therapy; suspect EOS with negative blood culture, variable abnormal neutrophil values, clinical signs of sepsis/infection, and antimicrobial therapy 5-7 days; and noninfected neonates who were asymptomatic, had a negative blood culture, normal serial neutrophil values, and in whom antimicrobial therapy was stopped at 48 hours. To establish a control group to revalidate the neutrophil RR, we selected 745 noninfected neonates without exposure to prenatal, intrapartum, and postnatal factors known to affect neutrophil values.9,17,18 Neutrophil Value Score The neutrophil value score (NVS) was developed to assess the use of serial neutrophil values for determining the absence of EOS. It uses the sum of the ATN, ATI, and I:T values obtained at 0-6, 10-16, and 22-28 hours of life that were within the upper and lower ranges of each RR, resulting in a maximum score of 9. Neonates without a CBC for each time period were excluded. The Mann-Whitney U test was used to compare the NVS between groups. Receiver operating characteristic (ROC) curves were constructed to assess the accuracy of the NVS for determining the absence of EOS. Statistical Analyses ATN, ATI, and I:T values from noninfected neonates were used to validate the RR9 by examining their distribution outside the 10th and/or 90th percentiles by postnatal age. We similarly examined the neutrophil values for proven and suspect EOS. Comparisons were performed with Pearson c2. Demographics and clinical variables were analyzed by ANOVA and the c2 test. A Bonferroni correction was used in pair-wise comparisons of the 3 groups when ANOVA was P < .05. Sensitivity, specificity, and positive predictive values (PPVs) and NPVs were determined for neutrophil values <10th and >90th percentiles, comparing proven and proven plus suspect vs noninfected controls. ROC curves 523
THE JOURNAL OF PEDIATRICS
www.jpeds.com
were constructed for abnormal neutrophilic values at each time period when a CBC was obtained to identify neutrophil values and time periods with the best predictability for EOS. Statistical analyses were performed with IBM SPSS version 19 (SPSS Institute, Chicago, Illinois).
Results There were 2073 NICU admissions during the study (Figure 2; available at www.jpeds.com); 77% were evaluated for EOS, including 1292 noninfected neonates, 292 with suspect EOS, and 9 with blood culture proven EOS (Table I). There were 255 neonates #30 weeks’ gestational age, and they included 14% of noninfected, 24% of suspect, and 22% of blood culture–proven EOS. Neonates with proven EOS had 25 CBCs obtained, clinically suspect neonates had 860, and noninfected controls had 2132; neonates averaged 2.9 CBCs/neonate during the first 24 hours of evaluation. There were no differences in maternal variables or infant’s gestational age, birth weight, sex, or route of delivery (Table II; available at www.jpeds.com). Neonates with suspect and proven EOS, however, had a lower 5-minute Apgar score, more asphyxia (Apgar score <5 at 5 minutes), required resuscitation at birth (intubation and/or chest compressions), and were more likely delivered to women with PROM, meconiumstained amniotic fluid, and chorioamnionitis. Nine neonates had at least one positive blood culture (Table II). Four had 2, whereas 5 (56%) had only 1 of 2. Blood cultures were positive in 7 of 9 neonates within 24 hours, with an average of 18 hours. All blood cultures were positive by 37 hours. Patient 4 was clinically ill, with a diagnosis of pneumonia via chest radiograph. Distribution of Neutrophil Values Neutrophil values from uninfected control neonates (2123 counts) were plotted on appropriate RR. Forty-nine percent of ATN values were <10th or >90th percentiles; 25% were neutropenic. In contrast, only 10% of ATI and 6% of I:T
Vol. 164, No. 3 values were >90th percentile, suggesting a normal distribution (Figure 3; available at www.jpeds.com). Because neutropenia might reflect preterm neonates,17 we excluded values from 449 uninfected neonates <37 weeks’ gestation. The incidence of neutropenia decreased to 9% (Figure 3, B). Notably, 24% of control ATN values were >90th percentile, demonstrating nonspecific neutrophilia. In neonates with proven EOS, 60% of ATN values were outside the RR, and 40% were neutropenic (data not shown). In addition, 36% of ATI and 56% of I:T values were outside the RR. Thus, the distribution of neutrophil findings for neonates with proven EOS differed significantly from the RR (P < .05). Neonates with suspect EOS had 860 CBCs; 66%, 38%, and 36% of ATN, ATI, and I:T values, respectively, were outside the RR. Their values also differed significantly from the RR (P < .05), consistent with previous studies.16 When abnormal values for the 3 groups were compared at each time period, neonates with EOS were more likely than noninfected controls to have abnormal neutrophil values (P < .05, Pearson c2). Prediction of EOS To determine the predictability of each neutrophil result corrected for postnatal age for the presence and/or absence of EOS, we calculated sensitivity, specificity, PPV, and NPV for each neutrophil/band value for proven and proven plus suspect vs noninfected neonates by using all values (Table III). ATN values in proven and proven plus suspect EOS had modest sensitivity and specificity (<66%) but poor PPV. Notably, the ATN NPV was 99% and 78%, respectively. ATI and I:T alone had modest-to-poor sensitivity and PPV, but the specificity and NPV ranged from 78% to 99%. We next examined 3017 neutrophil values from the 3 study groups; 2741 (91%) were collected within the recommended time periods of 0-6, 10-16, and 22-28 hours (see Methods). We constructed ROC curves for each time period and determined the area under the curve (AUC) for ATN, ATI, and I:T from the uninfected control group vs proven and proven plus
Table I. Demographics, clinical characteristics, and distribution of neonates without and with suspect and proven EOS Maternal age, years Gestational age, weeks Birth weight, g Male sex Cesarean delivery 5-minute Apgar score* Evidence of asphyxia Resuscitation during delivery Maternal hypertension PROM Chorioamnionitis MSAF
Not infected (n = 1292)
Suspect (n = 292)
Proven (n = 9)
P value
27.2 6.4 34.9 4.1 2447 905 702 (54) 755 (58) 8 (7-9)† 111 (9)† 176 (14)† 347 (27) 191 (15)† 100 (8)z 207 (16)z
27.0 6.6 34.8 5.5 2510 1152 154 (53) 162 (56) 8 (6-9) 50 (17) 63 (22) 59 (20) 51 (18) 72 (24) 78 (27)
23.8 4.7 35.7 5.3 2799 1152 3 (33) 3 (33) 7 (6-9) 2 (22) 3 (33) 3 (33) 4 (44) 2 (22) 4 (44)
>.05 >.05 >.05 >.05 >.05 <.01 <.01 <.01 >.05 .03 <.01 <.01
MSAF, meconium-stained amniotic fluid. Data are means SD. Data are number of neonates with a presenting variable, and values in parentheses are percent of each study group except for Apgar score. *Data are medians and IQR. †P < .05 control compared with suspect. zP < .05 control compared with proven.
524
Mikhael, Brown, and Rosenfeld
ORIGINAL ARTICLES
March 2014
Table III. Prediction of EOS using individual neutrophil parameters in neonates with proven and proven plus suspect EOS Proven ATN ATI I:T Proven plus suspect ATN ATI I:T
Sensitivity, %
Specificity, %
PPV, %
NPV, %
60 36 56
51 91 94
1.4 4.5 9.8
99 99 99
66 38 37
51 91 94
36 64 72
78 78 78
suspect EOS (Figure 4). At no time was the AUC for ATN >0.58 (Table IV; available at www.jpeds.com). At 0-6 hours, the ATI and I:T AUC were generally #0.7; this improved over the ensuing 24 hours and was >0.7 for both at 22-28 hours. Because neutrophil values had high specificity and NPV (Table III), we assessed all 9 serial neutrophil values
collected over 24 hours by using the NVS to predict the absence of EOS. Seventy-six percent of neonates (n = 796) had a CBC at each time period, including 6 with proven EOS, 229 with suspect, and 561 who were noninfected. We calculated the sum of neutrophil values within the 90th and 10th percentiles of the RR, using 9 as a maximum score. The difference in NVS between proven and proven plus suspect EOS and noninfected neonates was P < .01, MannWhitney U test. The NVS had an AUC of 0.90 and 0.81 for proven and proven plus suspect, respectively, to predict no EOS (Figure 5). The specificity for proven EOS with NVP scores 6, 7, or 8 was 100% for no EOS, and proven plus suspect had values of 78%, 92%, and 97%, respectively. When we examined the NVS in the 292 neonates with clinically suspect EOS (ie, negative blood culture but antimicrobial treatment 5-7 days), 98 (43%) had a NVS $6, and the predominant characteristic of this group was preterm birth and low birth weight (data not shown). If the NVS was increased to $7, 51 neonates (22%) may have been spared from prolonged antibiotic therapy. Notably, 18 had scores of 8 or 9.
Figure 4. ROC curves for neutrophil measurements for neonates A-C, without vs proven EOS and D-F, without vs proven plus suspect EOS at A, D, 0-6 hours B, E, 10-16 hours and C, F, 22-28 hours. Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis
525
THE JOURNAL OF PEDIATRICS
www.jpeds.com
Vol. 164, No. 3
Figure 5. ROC curves representing NVS for neonates without EOS vs with A, proven and B, proven plus suspect EOS.
Discussion Neonatal EOS is associated with high morbidity and mortality and generally approached emergently. Withholding antimicrobial therapy pending results of a screening test may be contraindicated. Few investigators have examined the blood culture in conjunction with a laboratory test to rule out EOS, which may limit the duration vs onset of antimicrobial therapy in the large population of at-risk neonates. We examined the use of serial neutrophil values corrected for postnatal age as an adjunct to 2 blood cultures plus clinical signs in predicting the “absence” of EOS within 24-36 hours of initiating a sepsis evaluation, thereby decreasing the duration of antimicrobial therapy, duration of hospitalization, and cost of health care. There was very good NPV and specificity for serial neutrophil values in the absence of EOS when the established neutrophil RR was used.9 We developed the NVS to score serial neutrophil values corrected for postnatal age and observed very goodto-excellent predictability for the absence of EOS. Importantly, the neutrophil RRs were shown to remain valid when we corrected them for gestational age. Thus, the use of serial neutrophil counts in conjunction with blood culture can decrease antibiotic exposure in a large population of neonates evaluated for EOS without risking adverse outcomes by waiting. Less than 1% of live born neonates have blood culture– proven EOS,1 which may reflect increased prenatal identification and treatment of maternal GBS before birth and/or increased maternal antimicrobial use in the intrapartum period, resulting in negative neonatal blood culture.14,19 If the latter is true, it is exceedingly difficult to separate neonates with partially treated EOS caused by maternal antibiotics and those without a real risk of EOS, often resulting in prolonged antimicrobial treatment in the latter. A low blood volume collected for blood culture is also known to reduce the number of positive blood cultures and in526
crease the occurrence of false-negative blood cultures.20 Moreover, as noted previously, clinical signs often are inconclusive and undistinguishable from benign disorders such as transient tachypnea. Our study included all symptomatic neonates delivered at our institution, yet only 9, or <1%, had positive blood culture, whereas 18% of treated neonates had risk factors for EOS and negative blood culture. Notably, 67% of proven EOS had GBS sepsis and 22% Escherichia coli, and 78% had a positive blood culture reported within 24 hours. Because only 1 of 2 blood cultures were positive in 5 cases, 28% might have been incorrectly considered false negative if only a single blood culture had been collected,20,21 supporting the use of 2 blood cultures obtained from different sites in assessing EOS. Neutrophil values initially were suggested as adjuncts in the diagnoses of neonatal EOS but were not corrected for the effects of perinatal variables or postnatal age.6-8 Manroe et al5,9 subsequently established neutrophil RR, demonstrating that when perinatal variables and postnatal age were accounted for, ATN, ATI, and I:T had high sensitivity for diagnosing EOS that was primarily caused by GBS. However, the specificity was even better, that is, >90% for the absence of EOS.5,9,22 Validation of these RR has not been examined in detail in >30 years. We observed that the original neutrophil RR remain valid because no more than 10% of counts for any neutrophilic measurements were >90th or <10th percentiles after correcting for postnatal age and preterm birth.17,23 Neutrophilia, however, occurred in 24% of noninfected neonates. This finding has been suggested to be due to high altitude24-26; however, Dallas is 450 ft above sea level. These neonates were evaluated because of clinical signs of EOS; thus, they are not normal. Their neutrophilia is more likely attributable to a nonspecific stress response and possibly excessive catecholamines rather than high altitude. These neonates would not have been included in the original RR because they are not considered “normal.” This may Mikhael, Brown, and Rosenfeld
ORIGINAL ARTICLES
March 2014 explain why ATN values alone are poor predictors of EOS.24 The Centers for Disease Control and Prevention guidelines for prevention of perinatal GBS infection recommend a CBC at birth and/or 6-12 hours postnatally for neonates born to women colonized with GBS and inadequately treated.27 Neutrophil measurements at 0-6 hours of life were not diagnostic for EOS, consistent with recent observations in a larger population.28 Their reliability, however, improved with increasing postnatal age. Thus, serial neutrophil values corrected for postnatal age may be preferred to determine the risk of EOS. Little evidence exists that a single CBC has any validity, and waiting may not be safe. It is important to note that the RRs were derived from manual differentials; to date, we found no studies in which the authors compared manual and computer-generated differentials via flow cytometry. The reliability of a CBC obtained <12 hours postnatally was observed to be quite poor. Furthermore, the sensitivity and PPV were less reliable than the specificity and NPV in predicting the absence of EOS. This improved with increasing postnatal age, supporting the use of serial neutrophil values corrected for postnatal age. The NVS uses all 9 neutrophil values and determines the proportion within the published RR for each neutrophil measurement. The reliability exceeds 80% with 6 normal values and >90% with $7 normal values. Thus, the NVS can be used to stop antibiotics in neonates with clinical risk factors, but transient or mild clinical signs and negative blood culture at 24-36 hours, potentially decreasing the number receiving antibiotics >24 hours by 50%. We were unable to determine whether the amelioration of clinical signs by 6-8 hours of age in this retrospective study might serve as an independent sign of a low probability for EOS; however, we would suggest that in our population this might have reflected the initiation of antibiotics within the first 2-4 hours of life and thus clinical improvement as the result of initiation of treatment. Notably, 22%-43% of the suspect neonates who received 5-7 days of antibiotics had a NSV $6-7. When we excluded ATN measurements at 0-6 hours postnatal age, the AUC had lower specificity (0.8), reflecting the high NPV of ATN that increases the statistical weight of the NVS. Prospective studies are needed in large populations to further validate the NVS and define neonates with suspect EOS. Furthermore, the number of blood culture positive neonates was low, which can only be addressed in larger clinical studies. We have demonstrated that the use of the established neutrophil RR9 and serial neutrophil measurements obtained over 24 hours after initiating an evaluation for EOS have high specificity and NPV in a population of term and preterm neonates. In conjunction with 2 blood cultures, this could allow cessation of antibiotic therapy in a large group of neonates commonly evaluated for EOS. In a NICU that treats all neonates evaluated for EOS for 5-7 days, this will decrease the duration of antimicrobial therapy 3-4 days as well as the duration of hospitalization for neonates $36 weeks’ gesta-
tion (50%), resulting in a conservative estimate of savings of $2.5 million annually in our NICU. It also is likely to decrease fungal infections in the NICU. We encourage others to examine the algorithm noted (Figure 1) and to consider studying this approach. n Submitted for publication Jul 3, 2013; last revision received Oct 9, 2013; accepted Oct 29, 2013. Reprint requests: Charles R. Rosenfeld, MD, Department of Pediatrics, UT Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75390-9063. E-mail: [email protected]
References 1. Escobar GJ. The neonatal “sepsis work-up”: personal reflections on the development of an evidence-based approach toward newborn infections in a managed care organization. Pediatrics 1999;103:360-73. 2. Schelonka RL, McCracken G. Bacterial and fungal infections. In: Avery GB, MacDonald MG, Seshia MMK, Mullett MD, eds. Avery’s neonatology: pathophysiology & management of the newborn. 6th ed. Philadelphia (PA): Lippincott Williams & Wilkins; 2005. p. 1235-73. 3. Mukhopadhyay S, Eichenwald EC, Puopolo KM. Neonatal early-onset sepsis evaluations among well-appearing infants: projected impact of changes in CDC GBS guidelines. J Perinatol 2013;33:198-205. 4. Ablow RC, Driscoll SG, Effmann EL, Gross I, Jolles CJ, Uauy R, et al. A comparison of early-onset group B steptococcal neonatal infection and the respiratory-distress syndrome of the newborn. N Engl J Med 1976; 294:65-70. 5. Manroe BL, Rosenfeld CR, Weinberg AG, Browne R. The differential leukocyte count in the assessment and outcome of early-onset neonatal group B streptococcal disease. J Pediatr 1977;91:632-7. 6. Xanthou M. Leucocyte blood picture in ill newborn babies. Arch Dis Child 1972;47:741-6. 7. Gregory J, Hey E. Blood neutrophil response to bacterial infection in the first month of life. Arch Dis Child 1972;47:747-53. 8. Zipursky A, Palko J, Milner R, Akenzua GI. The hematology of bacterial infections in premature infants. Pediatrics 1976;57:839-53. 9. Manroe BL, Weinberg AG, Rosenfeld CR, Browne R. The neonatal blood count in health and disease. I. Reference values for neutrophilic cells. J Pediatr 1979;95:89-98. 10. Philip AG. Response of C-reactive protein in neonatal group B streptococcal infection. Pediatr Infect Dis 1985;4:145-8. 11. Messer J, Eyer D, Donato L, Gallati H, Matis J, Simeoni U. Evaluation of interleukin-6 and soluble receptors of tumor necrosis factor for early diagnosis of neonatal infection. J Pediatr 1996;129:574-80. 12. Santana C, Guindeo MC, Gonzalez G, Garcia-Munoz F, Saavedra P, Domenech E. Cord blood levels of cytokines as predictors of early neonatal sepsis. Acta Paediatr 2001;90:1176-81. 13. Malik A, Hui CP, Pennie RA, Kirpalani H. Beyond the complete blood cell count and C-reactive protein: a systematic review of modern diagnostic tests for neonatal sepsis. Arch Pediatr Adolesc Med 2003;157:511-6. 14. Polin RA. Management of neonates with suspected or proven early-onset bacterial sepsis. Pediatrics 2012;129:1006-15. 15. Cotten CM, Taylor S, Stoll B, Goldberg RN, Hansen NI, Sanchez PJ, et al. Prolonged duration of initial empirical antibiotic treatment is associated with increased rates of necrotizing enterocolitis and death for extremely low birth weight infants. Pediatrics 2009;123:58-66. 16. Engle WD, Rosenfeld CR. Neutropenia in high-risk neonates. J Pediatr 1984;105:982-6. 17. Mouzinho A, Rosenfeld CR, Sanchez PJ, Risser R. Revised reference ranges for circulating neutrophils in very-low-birth-weight neonates. Pediatrics 1994;94:76-82. 18. Jackson GL, Engle WD, Sendelbach DM, Vedro DA, Josey S, Vinson J, et al. Are complete blood cell counts useful in the evaluation of asymptomatic neonates exposed to suspected chorioamnionitis? Pediatrics 2004;113:1173-80.
Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis
527
THE JOURNAL OF PEDIATRICS
www.jpeds.com
19. Ottolini MC, Lundgren K, Mirkinson LJ, Cason S, Ottolini MG. Utility of complete blood count and blood culture screening to diagnose neonatal sepsis in the asymptomatic at risk newborn. Pediatr Infect Dis J 2003;22:430-4. 20. Schelonka RL, Chai MK, Yoder BA, Hensley D, Brockett RM, Ascher DP. Volume of blood required to detect common neonatal pathogens. J Pediatr 1996;129:275-8. 21. Pierce JR, Merenstein GB, Stocker JT. Immediate postmortem cultures in an intensive care nursery. Pediatr Infect Dis 1984;3:510-3. 22. Engle WD, Rosenfeld CR, Mouzinho A, Risser RC, Zeray F, Sanchez PJ. Circulating neutrophils in septic preterm neonates: comparison of two reference ranges. Pediatrics 1997;99:E10. 23. Faix RG, Hric JJ, Naglie RA. Neutropenia and intraventricular hemorrhage among very low birth weight (less than 1500 grams) premature infants. J Pediatr 1989;114:1035-8.
528
Vol. 164, No. 3 24. Schmutz N, Henry E, Jopling J, Christensen RD. Expected ranges for blood neutrophil concentrations of neonates: the Manroe and Mouzinho charts revisited. J Perinatol 2008;28:275-81. 25. Carballo C, Foucar K, Swanson P, Papile LA, Watterberg KL. Effect of high altitude on neutrophil counts in newborn infants. J Pediatr 1991; 119:464-6. 26. Maynard EC, Reed C, Kircher T. Neutrophil counts in newborn infants at high altitude. J Pediatr 1993;122:990-1. 27. Verani JR, McGee L, Schrag SJ. Prevention of perinatal group B streptococcal disease–revised guidelines from CDC, 2010. MMWR Recomm Rep 2010;59:1-36. 28. Newman TB, Puopolo KM, Wi S, Draper D, Escobar GJ. Interpreting complete blood counts soon after birth in newborns at risk for sepsis. Pediatrics 2010;126:903-9.
Mikhael, Brown, and Rosenfeld
ORIGINAL ARTICLES
March 2014
Table II. Blood culture results, clinical descriptors, and distribution of neutrophil values for neonates with proven EOS Patient no.
EGA, weeks
1 2 3 4 5 6 7 8 9
39 41 25 33 40 37 38 38 30
Pathogen
No. of positive blood cultures
Incubation time, hours
Escherichia coli Escherichia coli GBS CONS GBS GBS GBS GBS GBS
2/2 2/2 2/2 1/2 1/2 1/2 1/2 2/2 1/2
18 12 24 37 23 15 24 15 31
Clinical description Respiratory failure Respiratory failure HMD Pneumonia Respiratory failure Respiratory failure Hypoxia, TGA HIE, PPHN PPHN, hypotension
Abnormal neutrophil values 7/9 4/9 4/9 3/9 6/9 6/9 0/3 4/9 4/9
CONS, coagulase-negative Staphylococcus; EGA, estimated gestational age; HIE, hypoxic ischemic encephalopathy; HMD, hyaline membrane disease; PPHN, persistent pulmonary hypertension of the newborn; TGA, transposition of great arteries.
Table IV. AUC for ATN, ATI, and I:T values from neonates with proven and proven plus suspect EOS at the 3 time periods used for analysis Proven 0-6 hours 10-16 hours 22-28 hours Proven plus suspect 0-6 hours 10-16 hours 22-28 hours
ATN
ATI
I:T
0.35 0.28 0.47
0.48 0.68 0.71
0.67 0.88 0.77
0.51 0.50 0.58
0.66 0.71 0.73
0.71 0.75 0.75
Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis
528.e1
THE JOURNAL OF PEDIATRICS
www.jpeds.com
Vol. 164, No. 3
Figure 1. Algorithm for identification, initiation, and cessation of antimicrobial therapy in neonates at risk of EOS.
Figure 2. NICU admissions and distribution of neonates with and without EOS (May 2009-April 2011). 528.e2
Mikhael, Brown, and Rosenfeld
March 2014
ORIGINAL ARTICLES
Figure 3. Distribution of neutrophil values within the Manroe et al RRs9 in 745 noninfected neonates between birth and 72 hours: A, ATN C, ATI and D, I:T. B, Distribution of ATN after correcting for preterm birth <37 weeks’ gestation.
Serial Neutrophil Values Facilitate Predicting the Absence of Neonatal Early-Onset Sepsis
528.e3