Since the early 1970s, group B streptococci (GBS) have been the leading causes of infection in the immediate newborn period in the United States.¹ The clinical impact of GBS sepsis has led to the development of different strategies to improve overall neonatal survival. In the absence of effective vaccination against GBS, intrapartum maternal chemoprophylaxis is the preferred method for preventing neonatal and maternal morbidity.² Lumbar puncture (LP) is often performed immediately after delivery in asymptomatic newborns after chemoprophylaxis as part of a sepsis evaluation to diagnose bacterial meningitis. However, the studies evaluating the clinical usefulness of LP have been done primarily in preterm infants⁵ or in neonatal intensive care units with both full-term and preterm infants.⁸ It may not be appropriate to generalize the findings from high-risk populations to asymptomatic full-term infants cared for in general nurseries. In addition to the increased cost from unnecessary tests, one also runs the risk of a significant number of the cerebrospinal fluid (CSF) cultures being contaminants. Such contaminants result in a prolongation of the infant's length of stay while awaiting the results of further diagnostic tests.

The three goals of this study were: (1) to determine whether the LP is useful in asymptomatic full-term newborns at risk of sepsis; (2) to see whether the neonatal white blood cell (WBC) count or immature-to-total (I:T) neutrophil ratio predicts bacteremia; and (3) to see whether the use of gentamicin in addition to ampicillin in asymptomatic infants is justified.

METHODS

Procedures

Infants with symptoms of sepsis (respiratory distress, poor vascular perfusion, temperature instability, bloody stools, lethargy, and recurrent hypoglycemia) had a sepsis evaluation performed. The evaluation consisted of an LP, blood culture, and a WBC count with differential. Symptomatic infants received 7- to 10-day courses of ampicillin and gentamicin intravenously.

Starting in 1987, all pregnant women presenting in the second trimester had cultures of the rectum and lower vagina taken for GBS. Because approximately 10% of all women at our hospital do not present for prenatal care, our hospital policy was established to allow intrapartum antibiotic treatment even if vaginal and rectal cultures had not been done. Ampicillin was given during labor if two or more risk factors for GBS sepsis were present. These risk factors were: (1) maternal colonization with GBS, (2) maternal fever, (3) prolonged rupture of fetal membranes at more than 18 hours, (4) foul-smelling amniotic fluids, (5) unexplained fetal tachycardia, and (6) elevated maternal WBC count. Immediately after delivery, the infant had a sepsis evaluation completed. The infant then received ampicillin and gentamicin intravenously for 48 to 72 hours; drugs were discontinued if blood and CSF cultures were negative for pathogens. For the purpose of assessing the usefulness of WBC counts and I:T neutrophil ratios, we used a comparison group, which consisted of infants in whom contaminants grew in blood or CSF but who were not infected.

Definitions

Laboratory Methods

CSF specimens were obtained and sent to the microbiology laboratory, where the specimens were plated on Trypticase soy agar with 5% sheep blood, MacConkey II agar, and chocolate II agar and also placed in thioglycolate broth. The specimens were incubated for 5 days at 37°C in a carbon dioxide incubator. Organisms were identified by standard laboratory methods.

Statistical Analysis

RESULTS

We did assess compliance with the postpartum evaluations of the infants. A review of 100 consecutive charts of infants born from October 1 through 16, 1994, was done to check compliance of the pediatric house staff with the protocol for sepsis evaluations. There were 10 symptomatic infants, and all were evaluated and received antibiotics pending culture, except for 1 who did not have an LP attempted and 2 who had unsuccessful LPs. There were 5 asymptomatic infants with two or more risk factors, and all 5 received complete sepsis evaluations. No unnecessary evaluations were done.

Of the 3423 asymptomatic infants, 17 (0.5%) had positive blood cultures, compared with 55 (3.2%) of 1712 symptomatic infants (P < .001). There were 72 positive blood cultures in 7 years (Table 1). Infants who had organisms isolated from the blood included 32 with GBS, 10 with Escherichia coli, 7 with Streptococcus faecalis, 6 with Streptococcus pneumoniae, (pneumococcus), 3 with Bacteroides sp, 1 with Neisseria meningitidis, 1 with Haemophilus influenzae, 2 with Staphylococcus aureus, 3 with S epidermidis, and 7 with Streptococcus sp. GBS accounted for half of the cases of sepsis despite the early recognition of risk factors and administration of antibiotic prophylaxis. For comparison, we identified all cases of full-term newborns with bacteremia and/or meningitis from the 7 years preceding institution of intrapartum prophylaxis (1980 through 1986). Table 2 shows that the rate of GBS infection per year was 2.3 per 1000 full-term births during the earlier period and 1.3 per 1000 in the latter period, which was a 43% decline (P = .018, ² test).

Table 1. Bacterial Pathogens and Contaminants in Blood of Symptomatic and Asymptomatic Full-term Newborns Within 7 Days of Birth* [View Table]

Table 2. Group B Streptococcal (GBS) Bacteremia and/or Meningitis in the Preprophylaxis Period (1981 Through 1986) and With Antepartum Ampicillin Prophylaxis in High-risk Colonized Mothers (1987 Through 1993) [View Table]

CSF specimens were obtained before initiation of postnatal antibiotic therapy in 21% of all full-term births (Fig 1). None of the specimens taken in asymptomatic infants yielded pathogens. Of the 1712 symptomatic infants, 11 infants had CSF cultures positive for pathogens (Table 3). GBS accounts for more than half of the pathogens isolated from CSF. Table 1 shows that E coli was commonly isolated from blood, yet it was not isolated from the CSF. No Gram-negative enteric organism caused meningitis in more than 24 000 deliveries. Ten of 11 infants had the same pathogens in the blood as in the CSF. No asymptomatic infant had meningitis, of 3423 evaluated by LP. The upper 95th percentile for this confidence limit is 0.0008, which means that, at most, 8 of 10 000 LPs in asymptomatic infants would yield pathogens.

Table 3. Cerebrospinal Fluid Pathogens and Contaminants Within 7 Days of Birth in Full-term Newborns* [View Table]

We next examined the sensitivity of the organisms to ampicillin as a single agent to see whether gentamicin improved empiric coverage. All isolates of GBS, enterococcus, and pneumococcus were sensitive to ampicillin. Of the 72 blood isolates, 7 were resistant to ampicillin. Two of these 7 blood isolates were ampicillin-resistant E coli susceptible to gentamicin, 1 asymptomatic and 1 with symptoms, and the other 5 were Gram-positive organisms resistant to gentamicin. Therefore, of 3423 asymptomatic infants who received ampicillin and gentamicin for 2 to 3 days, only one infant benefitted from the addition of gentamicin as a second antibiotic.

To assess the value of the WBC count and differential in predicting bacteremia, we compared bacteremic infants with a cohort having these tests done but having only contaminants isolated. The total WBC count and the I:T neutrophil ratio were significantly different between the bacteremic and comparison groups. In the bacteremic group, the total WBC count was less than that in the control group (13 500 ± 9400 vs 17 400 ± 7000; P = .007). In addition, more bacteremic patients were leukopenic (WBCs <5000/mm³) than control infants (12 of 67 vs 0 of 66; P < .001). As a diagnostic test, leukopenia has a sensitivity of 18% (95% CI, 10% to 29%) and a specificity of 100% (95% CI, 95% to 100%). The I:T neutrophil ratio was also greater in the bacteremic group (0.35 ± 0.27 vs 0.13 ± 0.13; P < .001).

In evaluating an ROC curve, a sharp break point in the upper left corner would represent an excellent value for defining an abnormal test result, whereas a diagonal line through the origin would signify that the diagnostic test has an ill-defined cutoff point. Our results (Fig 2) do not show a sharp break point. On review of the ROC curve for the I:T neutrophil ratio, the best cutoff point occurred at 0.4, which resulted in a sensitivity of 37% and a specificity of 94%. Hence, almost 63% of the cases with bacteremia would not be detected with this cutoff value.

During the 7-year study period, 3423 LPs were performed in asymptomatic infants, none of whom had meningitis but 35 of whom had contaminants. At our hospital an LP costs approximately $294, including physician and laboratory charges. With 3700 deliveries and 520 sepsis evaluations in asymptomatic full-term newborns per year, approximately $150 000 per year would be saved by omitting the LP. In addition, about five infants per year are detained an additional 48 to 72 hours awaiting results of second LPs after the first grows a contaminant.

DISCUSSION

GBS sepsis has a high incidence of coexistent meningitis; therefore, an LP to exclude the possibility of meningitis has traditionally been included in the evaluation of newborns born to GBS-colonized women who have fever or prolonged rupture of membranes, although previous studies have been in mixed preterm and full-term infants,⁵ whereas ours was exclusively in full-term newborns. In our study, 3423 LPs were done in asymptomatic full-term infants, of which none revealed meningitis, but 35 had contaminants (0.96%). Eleven cases of meningitis were identified, and all were in symptomatic infants. Neonates with meningitis most commonly presented with nonneurologic complications, such as respiratory distress, perfusion problems, and temperature instability, although many also had neurologic symptoms of lethargy or irritability, and we think that all such neonates should be classified as symptomatic.

Results similar to ours were found in four previous studies. Fielkow et al⁸ reviewed 1073 consecutive CSF cultures in 10 years in full-term and preterm infants younger than 7 days. As in our study, none of the 284 asymptomatic infants in their study had meningitis, although 1.8% had contaminants. The rate of meningitis in symptomatic infants was approximately 2%, similar to our rate of 0.6%. They concluded that LP is not indicated in asymptomatic full-term or preterm infants born to high-risk mothers. In the study of Weiss et al,⁵ 2156 premature infants with respiratory distress had 1495 successful LPs at admission. A rate of early-onset meningitis of 0.27% was found, and a contamination rate of 1.3% was found. They did not perform LP in asymptomatic infants. Blood cultures were positive in three of the four cases of meningitis. They concluded, as did Eldadah et al⁶ in 1987, that LP should be reserved for infants with positive blood cultures, central nervous system symptoms, or positive urine GBS antigen test results. Schwersenski et al¹⁰ showed similar findings, but also noted that the yield of LP increased fivefold when performed in symptomatic infants after the first week of life. Even in the one study⁷ that separately analyzed newborns weighing less than 1500 g, only one case of meningitis was found in 773 infants on day 1 of life. They concluded that "LP should only be carried out during the newborn period when there are clinical signs and symptoms of severe sepsis and not as a routine on all low birth weight babies."

In contrast to these four studies, in 1995 Wiswell et al⁹ concluded from a review of 169 000 births in US Army hospitals that asymptomatic neonates frequently have meningitis. They identified 43 cases of meningitis in the first 72 hours, including 7 in infants born at 36 or less weeks and 36 in infants born at more than 36 weeks. Seven (16%) of 43 infants with meningitis were asymptomatic, which would suggest that early meningitis may often be asymptomatic. Even more worrisome, 4 of the 7 asymptomatic infants with meningitis had negative blood cultures. In view of the results of our study and the four studies cited above, this raises the question of whether some of these organisms were CSF contaminants. The authors do not provide data on the contamination rate, nor do they state whether second LPs were performed in any of the asymptomatic infants. The data on whether CSF pleocytosis or abnormal glucose or protein levels were present are also not provided. Given that we found a ratio of CSF contaminants to pathogens of 3:1, and Weiss et al⁵ found a ratio of 4.7:1 in premature infants, it is possible that some of the isolates of Wiswell et al⁹ were contaminants. Organisms that can be pathogenic were shown to be contaminants by second LP, including Enterobacter, Streptococcus mitis, and S aureus. Wiswell et al⁹ do not identify the organisms that grew in cases of "asymptomatic meningitis," or state whether second LPs were performed.

Recent guidelines from the Centers for Disease Control and Prevention do not include the use of any antibiotics in infants whose mothers received ampicillin more that 4 hours before delivery¹; however, at the time of our study both ampicillin and gentamicin were used at our hospital and many others while awaiting culture results. Because gentamicin may cause ototoxicity, we examined whether ampicillin alone provided adequate empiric coverage of sepsis in the full-term infant. All isolates of GBS, enterococci, and pneumococci were sensitive to ampicillin, but 2 of 10 strains of E coli were ampicillin-resistant and susceptible to gentamicin. Addition of gentamicin would have increased antibiotic coverage from 90% to 93% of infants, but at a cost of exposing more than 3400 asymptomatic newborns to gentamicin. Because the incidence of ampicillin-resistance may vary among hospitals, there may be centers where empiric gentamicin use is warranted. Broad-spectrum antibiotic coverage is indicated in infants who are symptomatic and in infants born at 34 weeks or earlier while awaiting culture results.¹ Gentamicin should also be added when blood or CSF cultures grow enterococcus for synergism, but ampicillin is adequate initial therapy.

An ideal diagnostic test to establish neonatal sepsis should have high sensitivity, so no case of sepsis is missed, and a high negative predictive value, to exclude sepsis when the test is negative. In our study, a control group was compared with bacteremic infants, and when leukopenia (WBCs <5000) was analyzed, it had a sensitivity of 18% and a specificity of 100%. In 1984, Engle and Rosenfeld¹⁵ showed that factors other than sepsis could cause leukopenia in infants with sepsis. For example, infants with asphyxia whose mothers had pregnancy-induced hypertension had higher rates of leukopenia than infants with sepsis. The I:T neutrophil ratio, on the other hand, was elevated in 61% of septic infants but in only 12% of infants whose mothers had pregnancy-induced hypertension and 22% of asphyxiated infants. A review article¹⁶ on premature infants reported that an I:T neutrophil ratio of greater than 0.2 had a wide range of sensitivity (58% to 90%) and specificity (31% to 78%). Only when the I:T ratio exceeded 0.4 in our study was the specificity high enough to be clinically useful. Thus, our study showed limited usefulness of the I:T ratio as a diagnostic test to establish sepsis.

In conclusion, we have shown that asymptomatic full-term infants whose mothers are treated during birth with ampicillin are at extremely low risk of having meningitis within 7 days of birth. In more than 24 000 consecutive full-term deliveries, no infant had Gram-negative enteric meningitis. Ninety percent of all pathogens isolated from blood or CSF were susceptible to ampicillin as a single agent. The new guidelines from the Centers for Disease Control and Prevention¹ recommend either universal maternal screening for GBS colonization or a policy of using intrapartum risk factors to decide which mothers require intrapartum ampicillin. In hospitals such as ours, which have used antepartum risk factors in GBS-colonized mothers as indications for a sepsis evaluation and 48 hours of broad-spectrum antibiotics, the new guidelines would reduce the number of newborns evaluated because, they specify that only those treated less than 4 hours before delivery need be evaluated. The guidelines leave the issue of LP to the clinical judgment of the physician. Our data indicate that LP in an asymptomatic full-term newborn has a very low yield, even though we know that not all of our at-risk mothers with GBS received ampicillin more than 4 hours before delivery. A subsequent study would be desirable to determine: (1) the incidence of sepsis and meningitis in full-term newborns treated with intrapartum antibiotics, but not treated after birth, and (2) the yield of LPs in asymptomatic full-term newborns whose mothers receive ampicillin less than 4 hours before delivery.