The challenge of risk stratification in well-appearing, febrile infants has plagued clinicians for decades.1–4 With the low prevalence of conditions such as bacterial meningitis (0.1% to 0.5%),5 yet a high potential for morbidity and mortality in missed cases, considerable effort has been put toward identifying infants at low risk for invasive bacterial infections (IBIs, bacteremia and bacterial meningitis). Many available clinical prediction rules for risk stratification to identify infants at low risk for having an IBI have moderately high sensitivities (90% to 95%) and, thus, high negative predictive values (97% to 99%).2,3,6,7 Given the uncommon occurrence of IBIs, the high negative predictive value of these tools is often used by clinicians as reassurance in safely doing fewer interventions in a subset of well-appearing febrile infants.
In 2021, the American Academy of Pediatrics (AAP) released updated guidelines to help inform clinical decision making in the care of well-appearing febrile infants 8 to 60 days of age.8 Within the guideline, there is an emphasis on the use of inflammatory markers, including procalcitonin (PCT), C-reactive protein, absolute neutrophil count, and height of fever to help identify low-risk infants. However, although risk stratification centers on identifying infants at low risk for having a bacterial source for infection, viral infections are the most common cause of fever in the well-appearing febrile infant population.9 The authors of studies have documented that >50% of infants have a virus identified during work-up as a possible etiology for their fever,10 and even this figure likely underestimates the percentage of viral-positive infants, given unstandardized practices around screening.
With the high prevalence of viral etiologies within well-appearing febrile infants and the new AAP Clinical Practice Guideline recommendations for obtaining inflammatory markers, Kusma et al investigated the impact of a confirmed viral infection on PCT levels in well-appearing febrile infants both with and without concurrent bacterial infection.11 Previously healthy, term infants 8 to 60 days old without localizing signs of infection presenting with fever were included in the study. Viral infections were confirmed by PCR respiratory viral panels. Bacterial infections were included if a positive urine, blood, or CSF culture result was documented and the infant received an appropriate course of antibiotics for a bacterial infection.
The final analysis included 663 infants: 404 (61%) had a confirmed viral illness alone, 73 (11%) had bacterial infection alone, 48 (7%) had both bacterial and viral infection, and 138 (21%) had no identified source infection. They compared cohorts on the basis of infection type and stratified by age groups of 8 to 28 days and 29 to 60 days.
As expected, the results revealed that infants with a bacterial illness had significantly higher PCT levels when compared with infants without any infection (P < .01) and when compared with infants with only a viral infection (P < .01). Infants who had concurrent viral and bacterial infections had significantly higher PCT levels than infants with a viral infection alone (P ≤ 0.01) but significantly lower PCT levels than infants with a bacterial infection alone (P < .01).
Looking at the predictive ability when using a cut-off value of 0.5 ng/dL, the results revealed that PCT had significantly decreased sensitivity (68% vs 44%; P < .01) but similar specificity (93% vs 96%; P = .17) in detecting bacterial infection in the presence of a comorbid viral infection. The positive predictive value decreased (85% vs 58%; P < .01), whereas the negative predictive value increased (85% vs 94%; P < .01).
Test Characteristics and Implications for Procalcitonin
Kusma et al’s study raises new questions about how to interpret a PCT in a febrile infant, particularly in the setting of viral coinfection. Explicitly, the authors propose the interpretation of PCT as a highly specific marker for bacterial infection. Although it is true that the AAP guideline recommends further work-up if the PCT is elevated, is this because, (1) PCT is a sensitive marker, and we cannot rule out bacterial disease when elevated, or (2), as Kusma et al suggest, PCT is specific for a bacterial infection when elevated, and so an elevation raises heightened concern that a bacterial infection is actually present, thus ruling in disease?
To understand the theory of how PCT acts as a biomarker, understanding its kinetics is an important starting point. PCT is the prohormone of the hormone calcitonin. Calcitonin is exclusively produced by C-cells of the thyroid gland in response to hormonal stimuli. In contrast, after bacterial insult, PCT gets induced in many organs via stimulation of the inflammatory cascade and is subsequently released into circulation.12 In non-disease states, physiologic levels of PCT are low and can increase 100 to 1000-fold because of circulating endotoxins and cytokines induced by bacteria, including interleukin-6, tumor necrosis factor-α, and interleukin-1b.13 In contrast, cytokines released after viral infection, such as interferon-γ, lead to the downregulation of PCT.13 In addition, PCT starts to rise after 2 hours of exposure to bacteria and peaks within 12 to 24 hours.12 The in vitro properties of PCT should, in theory, be great for ruling in bacterial infections. The in vivo studies measuring PCT during acute infections, however, reveal a less clear direct correlation.
Previous studies comparing PCT to CRP in the febrile infant population have revealed that PCT performs with a similar sensitivity but better specificity for detecting bacterial infections. Woelker et al report a sensitivity of 92% and specificity of 64% for procalcitonin when using a cutoff value of 0.26 ng/mL to detect a serious bacterial infection (urinary tract infection, bacteremia, or bacterial meningitis).14 In the study by Milicet et al, a PCT ≥0.5 ng/mL had a sensitivity and specificity of 85% for detecting an IBI compared with 75% when using a CRP >20 mg/L.15 Conversely, the data reveal a higher sensitivity for bacterial infections in patients with lower respiratory tract infections. A small study of patients with viral bronchiolitis with superimposed bacterial PNA (confirmed via endotracheal aspirate culture positivity) did not reveal any difference in PCT levels between intubated infants with and without bacterial pneumonia.16 Similarly, Matha et al did not detect any difference in PCT levels between children with suspected sepsis between viral and bacterial etiologies.16 Interestingly, in this study, they did have a subsection of children (15: 10 with bronchoalveolar lavage positive cultures, 3 with focal bacterial infections, and 2 with bacteremia) that had concurrent viral infections. Again, no difference in PCT levels was noted. In addition, in the children with confirmed bacteremia, the area under the receiver operating characteristics curve was moderate at 0.65.17
To understand and interpret the numbers, we have to recall the basics of the definitions of sensitivity and specificity and the acronyms of SnOUT and SpIN. A test with high sensitivity, if the result is negative, is good at ruling out disease, meaning we can be confident that a negative test result truly indicates that the disease is not present. Whereas a test with high specificity, if the result is positive, is good at ruling in disease, allowing us to be confident that the condition is truly present in the setting of a positive test result. In addition, sensitivity and specificity are test characteristics. They are elucidating the usefulness of a test, in this case, PCT on the basis of its performance against a gold standard (culture results). So, in the context of well-appearing, febrile infants, a test with high sensitivity would correctly identify infants with a bacterial infection from among those with positive culture results. This results in a low false negative rate, thus helping to rule out infants who do not have bacterial infections. Alternatively, a test with a high specificity in this context would correctly identify infants who do not have a bacterial infection from among those with negative culture results. This results in a low false positive rate, which helps to rule in infants who do have bacterial infections.
Given that bacterial infections in febrile infants can be serious, at times are largely asymptomatic, aside from fever, and treatment is recommended early in the course of the disease to prevent progression, ideally a test with high specificity could direct what infants need more invasive work-up (like a lumbar puncture). A test with a high sensitivity could identify which infants would not need the more thorough evaluation.
If the goal in the use of PCT is to assist in ruling out a bacterial infection in a well-appearing febrile infant, then Kusma et al’s study continues to underscore that for infants with viral infection alone, a negative PCT is associated with correctly identifying infants that do not have a bacterial infection. If the goal in the use of PCT is to assist in ruling in a bacterial infection in a well-appearing febrile infant, then Kusma et al’s study reveals an interaction in the predictive ability in infants with bacterial infections when they have a concurrent viral infection. Further data on a larger, more representative population is needed to better explore these findings.
Currently, the current AAP Clinical Practice Guideline recommendations are to limit further work-up in the setting of a negative (<0.5 ng/mL) PCT result, consistent with test characteristics of high sensitivity. PCT, however, is a screening, not a diagnostic biomarker, and is inherently imperfect, ironically, much like the evaluation and risk stratification of well-appearing febrile infants.
COMPANION PAPER: A companion to this article can be found online at www.hosppeds.org/cgi/doi/10.1542/hpeds.2022-007070.
Drs Shine and McDaniel conceptualized and drafted the initial manuscript and revised and reviewed the final manuscript; and both authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
FUNDING: Dr McDaniel received support from the AAP for the national QI collaborative, REducing Variability in Sepsis Evaluation (REVISE II) for well-appearing febrile infants.
CONFLICT OF INTEREST DISCLOSURES: The authors have indicated they have no potential conflicts of interest relevant to this article to disclose.
Comments