For years, investigators have sought a test or panel of tests able to diagnose neonatal sepsis accurately and more rapidly than is possible with the recovery of microorganisms from specimens of sterile body fluids or tissues. Although results of some studies have been encouraging, microbial isolation from blood, cerebrospinal fluid (CSF), urine, other body fluids (peritoneal, pleural, joint, middle ear), or tissues (bone marrow, liver, spleen) remains the most valid method of diagnosing bacterial sepsis. Many advances in non–culture-based methods, which may nevertheless remain microorganism specific, such as tests using polymerase chain reaction (PCR) amplification technology, are promising for more rapid diagnosis of infection. This chapter discusses nonspecific laboratory aids for the diagnosis of invasive bacterial infections. Specific microbiologic techniques are discussed in Chapter 6 and in chapters addressing specific pathogens.

Diagnostic Utility of Laboratory Tests

In establishing the usefulness of any laboratory determination, a balance must be reached between sensitivity and specificity. For a clinician needing to decide whether to institute or withhold therapy on the basis of a test result, the predictive values (and perhaps likelihood ratios [LRs]) of that test are also important. In relation to neonatal infection, these terms can be defined as follows ( Fig. 36-1 ):

  • Sensitivity: If infection is present, how often is the test result abnormal?

  • Specificity: If infection is absent, how often is the test result normal?

  • Positive predictive value: If the test result is abnormal, how often is infection present?

  • Negative predictive value: If the test result is normal, how often is infection absent?

  • LR, positive test result: If the test result is abnormal, how much does that result increase the pretest probability of disease?

  • LR, negative test result: If the test result is normal, how much does that result decrease the pretest probability of disease?

Figure 36-1, Diagnostic test characteristics. Sensitivity, specificity, positive predictive value, and negative predictive value are commonly expressed as percentages; likelihood ratios represent x-fold increases or y-fold decreases in probability.

In attempting to discover the presence of a serious illness such as neonatal bacteremia, which is life-threatening yet treatable, diagnostic tests with maximal (100%) sensitivity and negative predictive value are desirable. In other words, if infection were present, the result would always be abnormal; if the result were normal, infection would always be absent. The reduced specificity and positive predictive value that this combination may engender usually are acceptable because overtreatment with antibiotics on the basis of a false-positive result is likely to be of limited harm compared with withholding therapy on the basis of a false-negative result. Some authorities prefer the use of LRs because predictive values vary with the prevalence of a disease, whereas LRs relate only to the test performance (sensitivity, specificity). Large LRs (>10) imply that a test result would conclusively increase the probability of the disease being present, whereas small LRs (<0.1) minimize the probability of the disease being present.

In reviewing a report of a new laboratory aid for the diagnosis of neonatal sepsis, the first consideration is to determine what reference standard was used to evaluate the new test (i.e., what was the gold standard applied). In one study of infants who died with unequivocal evidence of infection at autopsy, bacteria were grown from 32 of 39 antemortem blood cultures (sensitivity of only 82%). Among 50 infants without pathologic findings of infection at autopsy, 48 had negative blood culture results (specificity of 96%). A positive blood or CSF culture result had a 94% chance of being associated with serious neonatal infection (positive predictive value of 94%), whereas a negative blood culture result indicated absence of serious infection only 87% of the time (negative predictive value of 87%). It is likely that the predictive values cited in this study already are different from the values that may be observed in practice because of the high prevalence (44%) of positive bacterial culture results in the autopsy cases reviewed. High prevalence inflates the positive predictive value and depresses the negative predictive value; low prevalence depresses the positive predictive value and inflates the negative predictive value.

The lack of perfection of the generally accepted gold standard of bacterial culture complicates the search for new laboratory aids in the diagnosis of neonatal sepsis; it may be unclear whether a new test is truly functioning better than culture, which itself may not be “perfect.” Interpretation of bacterial culture results may become even more complicated as intrapartum antibiotic prophylaxis to prevent early-onset group B streptococcal sepsis becomes more common. Yet, it may not be clinically necessary to require detection of only bacterial sepsis. Tests that yield results considered “falsely positive” in the absence of bacterial disease may still be clinically useful in assigning normal versus abnormal status if the results register positive because of serious viral disease that may require antiviral therapy (e.g., neonatal enterovirus or herpes simplex infections).

In addition, even the most useful test serving as a gold standard may function well in one population of infants, for example, very low birth weight with a greater prevalence of sepsis, and yet may function quite poorly in another population—that of older infants with larger birth weights who are growing normally and have a lesser prevalence of sepsis. Each report of a new test claiming superiority to current tests (e.g., bacterial culture) must be critically evaluated in the extended clinical setting, including target populations, with standardization within clinical laboratories and among institutions.

In Search of the Ideal Laboratory Test

Because the body’s response to an infection necessarily begins after the invasion of a pathogen, it may never be possible to diagnose an infection immediately: There may always be a lag in the physiologic response on which the diagnostic test is based. Even bacterial blood cultures performed with modern, continuously computer-monitored detection technology do not reach 100% sensitivity for the diagnosis of neonatal infection. Incubation of bacteria may take several days, and genuine bacteremia may be missed because of the small volume of blood taken from infants with very low birth weight. A set of properties of the ideal or perfect diagnostic test has been proposed. These characteristics should be kept in mind as the different laboratory tests for neonatal infection are discussed in this chapter.

First, the laboratory analyte would be biochemically stable (to ease transport requirements), easy to analyze (quick laboratory turnaround time), and obtainable from a small volume of blood. Second, the analyte would have clear diagnostic cutoffs between normal and abnormal, across various gestational ages, and across birth weights. Third, the test would be inexpensive and comparable among different laboratories so that it could be widely applied.

In addition, the ideal laboratory test for the diagnosis of neonatal infection would be maximally sensitive (no false-negative results) and highly specific (few false-positive results) and have a physiologic window of opportunity for sampling. More precisely, the test would become abnormal just as infection was present and remain abnormal for some time, to allow clinicians to use it as a diagnostic aid even if the clinical symptoms of infection were initially missed.

Finally, the ideal marker would correlate well with progress of infection, perhaps even predicting outcome. As we review each test in this chapter, it will become apparent that none of the currently available laboratory aids for the diagnosis of infection fulfills these ideal properties. Although new tests are continually being studied, it is uncertain whether any will ever achieve perfection. For a more extensive review of the older literature on laboratory aids for the diagnosis of neonatal sepsis, the reader is referred to the previous edition of this text.

Blood Cell Counts, Ratios, and Flow Cytometric Markers

Total Leukocyte Count, Differential Leukocyte Count, and Morphology

Total leukocyte counts (white blood cell [WBC] counts) are of limited value in the diagnosis of septicemia in newborns. Total leukocyte counts are particularly unreliable indicators of infection during the first several hours of early-onset sepsis because they are normal at the time of initial evaluation in more than one half of infants with proven bacteremia. Optimal interpretation of the WBC count also requires examination of the newborn’s age in hours.

Differential leukocyte counts also have not functioned well as markers for infectious disease in the newborn period. Increased percentages of lymphocytes have been described in association with pertussis and congenital syphilis, and increased monocyte counts with congenital syphilis, and perinatal listeriosis; however, either or both may also be affected by ABO incompatibility and recovery from sepsis. Eosinophilia, a common finding in premature infants, has been related to numerous factors, including low birth weight, immaturity, establishment of positive nitrogen balance, improved nutritional status, and use of total parenteral nutrition or blood transfusions, in addition to sepsis. Basophil counts tend to follow the fluctuations in eosinophil numbers in ill or healthy newborns.

Several investigators have shown that significant changes in neutrophil morphology occur in association with serious bacterial infection, with the appearance of toxic granules, Döhle bodies, and vacuolization. These features are of limited value in establishing a diagnosis; their presence has, at best, a positive predictive value for sepsis of only slightly more than 50% and, at worst, a positive predictive value of 33% to 37%.

Absolute Neutrophil Count

Recognizing the low predictive value of total leukocyte counts in serious neonatal bacterial disease, several investigators have studied the dynamics of neutrophil counts during the first month of life. These researchers and others uncovered patterns of change sufficiently constant to establish limits of normal variation ( Fig. 36-2 ) and defined noninfectious conditions involving the mother or the infant that might have significant effects on neutrophil values ( Table 36-1 and Box 36-1 ). Largely on the basis of these data, it was suggested that calculation of the absolute number of circulating neutrophils (polymorphonuclear plus immature forms) might provide a useful index of neonatal infection. Clinical experience has only partly supported this premise.

Figure 36-2, Total neutrophil counts in normal term infants and in very-low-birth-weight infants (inset).

Table 36-1
Clinical Factors Affecting Neutrophil Counts in Newborn Infants
Data from Weinberg GA, D’Angio CT: Laboratory aids for diagnosis of neonatal sepsis. In Remington JS, Klein JO, Wilson CB, et al, editors: Infectious diseases of the fetus and newborn infant, ed 7, Philadelphia, 2011, Saunders, pp 1144-1160; Xanthou M: Leucocyte blood picture in ill newborn babies, Arch Dis Child 47:741-746, 1972; Manroe BL, Weinberg AG, Rosenfeld CR, et al: The neonatal blood count in health and disease. I. Reference values for neutrophilic cells, J Pediatr 95:89-98, 1979; Gregory J, Hey E: Blood neutrophil response to bacterial infection in the first month of life, Arch Dis Child 47:747-753, 1972; and Rodwell RL, Tudehope DI, Gray PH: Hematologic scoring system in early diagnosis of sepsis in neutropenic newborns, Pediatr Infect Dis 12:372-376, 1993.
Neutrophil Counts
Factor Decrease Increase Total Immature Increase Increased I:T Ratio Approximate Duration (hr)
Maternal hypertension ++++ 0 + + 72
Maternal fever, neonate healthy 0 ++ +++ ++++ 24
≥6 hours intrapartum oxytocin administration 0 ++ ++ ++++ 120
Asphyxia (5-min Apgar score ≤ 5) + ++ ++ +++ 24-60
Meconium aspiration syndrome 0 ++++ +++ ++ 72
Pneumothorax with uncomplicated hyaline membrane disease 0 ++++ ++++ ++++ 24
Seizures: No hypoglycemia, asphyxia, or central nervous system hemorrhage 0 +++ +++ ++++ 24
Prolonged (≥4 min) crying 0 ++++ ++++ ++++ 1
Asymptomatic blood glucose ≤ 30 mg/dL 0 +++ +++ +++ 24
Hemolytic disease ++ ++ +++ ++ 7-28 days
Surgery 0 ++++ ++++ +++ 24
High altitude 0 ++++ ++++ 0 6

+, <25% of neonates affected; ++, 25%-50%; +++, 50%-75%; ++++, 75%-100%.

Ratio of immature forms to total neutrophils.

Box 36-1
Clinical Factors With No Effect on Neutrophil Counts in Newborn Infants
Data from Weinberg GA, D’Angio CT: Laboratory aids for diagnosis of neonatal sepsis. In Remington JS, Klein JO, Wilson CB, et al, editors: Infectious diseases of the fetus and newborn infant, ed 7, Philadelphia, 2011, Saunders, pp 1144-1160; Manroe BL, Weinberg AG, Rosenfeld CR, et al: The neonatal blood count in health and disease. I. Reference values for neutrophilic cells, J Pediatr 95:89-98, 1979; Xanthou M: Leucocyte blood picture in healthy full-term and premature babies during neonatal period, Arch Dis Child 45:242-249, 1970; and Gregory J, Hey E: Blood neutrophil response to bacterial infection in the first month of life, Arch Dis Child 47:747-753, 1972.

Race

Gender

Maternal diabetes

Fetal bradycardia

Mode of delivery

Total neutrophil counts in cord blood of infants delivered vaginally or by cesarean section after labor (2-14 hr) are twice those of infants delivered by cesarean section without labor.

Premature rupture of membranes, mother afebrile

Meconium staining, no lung disease

Uncomplicated hyaline membrane disease

Uncomplicated transient tachypnea of the newborn

Hyperbilirubinemia, physiologic, unexplained

Phototherapy

Diurnal variation

Most series of consecutive cases of neonatal sepsis have shown abnormal neutrophil counts at the time of onset of symptoms in only about two thirds of infants.

References .

In some series, 80% to 90% of infected infants have had abnormal values, whereas in other series, initial neutrophil counts were reduced or elevated in only one fourth to one third of infants with bacteremia, particularly when counts were determined early in the course of illness.

However, the absolute neutrophil count (ANC) may be of value in specific clinical situations. The association of neutropenia, respiratory distress, and early-onset (<48 hours) sepsis caused by group B streptococci is well documented, although the recognition that a similar association exists for early sepsis caused by other microorganisms (e.g., Haemophilus influenzae , pneumococci, and Escherichia coli ) has not been adequately emphasized. Because all infants were noted to be ill at birth or shortly thereafter, when neutrophil counts normally are increasing, a low count (0-4000 cells/mm 3 ) in this clinical setting is a highly significant finding, often reflecting a depletion of bone marrow granulocyte reserves and indicating a poor prognosis.

Total Nonsegmented Neutrophil (Band) Count

The blood smear and differential cell count during the newborn period are strikingly different from values seen at any other time of life. Immature forms are present in relatively large numbers, particularly among premature infants and during the first few days of life. The number of immature neutrophils, mostly nonsegmented (band, stab) forms, increases from a maximum normal value of 1100 cells/ mm 3 in cord blood to 1500 cells/mm 3 at 12 hours of life and gradually decreases to 600 cells/mm 3 by 60 hours of life. Between 60 and 120 hours, the maximum count decreases from 600 to 500 cells/mm 3 and remains unchanged through the first month of life. Metamyelocytes and myelocytes also are often present in significant numbers during the first 72 hours after delivery but disappear almost entirely toward the end of the first week of life. Even occasional promyelocytes and blast cells may be seen during the early days of life in healthy infants.

As neutrophils are released from the bone marrow in response to infection, an increasing number of immature cells enter the bloodstream, producing a differential cell count with a “shift to the left” even greater than that normally present in the neonate. This response is so inconstant, however, that, with few exceptions, the absolute band or immature (bands, metamyelocytes) neutrophil count has been found to be of little diagnostic value. In many infants with infection, despite an increased proportion of immature cell types in the differential leukocyte count, exhaustion of the bone marrow reserves prevents an increase in the absolute number of band neutrophils in the circulation. This is particularly common in more seriously ill patients, in whom early diagnosis is most critical.

Despite its relative insensitivity, the immature neutrophil count has been found to have good positive predictive value in some, although not all, studies. In infants with clinical evidence of sepsis and high band counts in whom culture results remain negative, follow-up cultures and investigation for a history of perinatal events that might explain the discrepancy (see Table 36-1 ) or for the possibility of infection related to other causes, such as enteroviruses, are indicated.

Neutrophil Ratios

The unreliability of absolute band counts led to the investigation of neutrophil ratios as an index of neonatal infection. Determinations have included the ratio of either bands or all immature neutrophils (e.g., bands, metamyelocytes, and myelocytes) to either segmented neutrophils (the immature-to-mature neutrophil ratio [I:M ratio]) or to all neutrophils (the immature-to-total neutrophil ratio [I:T ratio]). Despite the early enthusiasm of researchers, the clinical studies that include these determinations have failed to show a consistent correlation with the presence of serious bacterial disease. As might be expected, low band counts caused by exhaustion of marrow can produce misleadingly low ratios in the presence of serious or overwhelming infection.

References .

The I:T ratio is the best studied of the ratios. Inclusion in the numerator of all immature forms, rather than just band cells, heightens accuracy by accounting for the increase in metamyelocytes that is sometimes seen with accelerated release from the neutrophil storage pool. Use of total rather than segmented neutrophils in the denominator has the advantage of always yielding a value between 0 and 1 inclusive. The upper limit of normal for the first 24 hours is variably reported as between approximately 0.2 to 0.4. It gradually declines to about 0.12 by 60 hours of age and remains unchanged for the remainder of the first month. A normal value up to 0.2, with age unspecified, has been found in some laboratories. I:T ratios during the first 5 days of life are less than 0.2 in 96% of healthy premature infants with a gestational age of 32 weeks or less.

However, numerous clinical studies of the I:T ratio have found it too unreliable to achieve more than limited usefulness by itself, with diagnostic sensitivity of at most 60% to 90% and specificity of only 50% to 75%.

References .

Perhaps the greatest value of the I:T ratio lies in its good negative predictive value: If the I:T ratio is normal, the likelihood that infection is absent is very high.

You're Reading a Preview

Become a Clinical Tree membership for Full access and enjoy Unlimited articles

Become membership

If you are a member. Log in here