Effect of Viral Illness on Procalcitonin as a Predictor of Bacterial Infection in Febrile Infants

We conducted an institutional research board-approved retrospective cohort study examining infants 8 to 60 days old presenting with fever to a tertiary children’s hospital between January 1, 2018 and December 31, 2020. Patients were identified via International Classification of Diseases, Tenth Revision codes for fever or BI (Supplemental Table 5). Identified charts were then reviewed for documentation of temperature ≥38°C, either via guardian report of home measurement or charted during their hospital stay.

Initial exclusion criteria were modeled after our hospital’s clinical pathway for febrile infants, consisting of prematurity (<37 weeks), previous admission for infection, previous admission to intensive care (neonatal or pediatric), antibiotic administration before the investigated hospital visit (unless given for the febrile illness under investigation at a transferring hospital), or complex care needs (known immunocompromised condition, congenital or chromosomal abnormalities, medically fragile conditions). Additional exclusions included patients with an obvious source of infection, such as pneumonia or cellulitis, in which a culture could not reasonably be obtained, and those for whom blood cultures, urine cultures, and/or PCT levels were not obtained. Finally, patients were excluded if viral testing via a respiratory virus polymerase chain reaction (PCR) panel was not obtained or results were unknown unless rapid respiratory viral testing was performed and positive. Those with only a negative rapid respiratory test were excluded because of the limited number of viral pathogens evaluated. Of note, our hospital’s clinical pathway recommends obtaining a PCT level for all febrile infants on initial presentation, regardless of clinical appearance, and a respiratory viral PCR panel for any patient with respiratory symptoms and/or during peak respiratory viral season.

Data collected from the electronic medical record included PCT levels, bacterial cultures (cerebrospinal fluid, urine, and blood), and viral test results. If PCT testing did not occur, a chart review was undertaken to determine the reason for omission. Viruses tested in the respiratory PCR panel included adenovirus, parainfluenza 1 to 4, influenza A and B, respiratory syncytial virus, metapneumovirus, rhinovirus, enterovirus, coronavirus, and severe acute respiratory syndrome coronavirus 2 (added in late 2020). The viral pathogen PCR used at the study institution (Biofire Respiratory 2.1 Panel) has a reported sensitivity of 97.1% and specificity of 99.3%, and results were available within 1 to 2 hours. Patients were categorized into groups depending on the presence or absence of a viral infection and/or BI.

Patients were identified as having a viral infection if either their viral respiratory panel or a rapid viral test result were positive within 3 days before or after their presentation. Patients were considered to have a BI if they had a positive blood, urine, and/or cerebrospinal fluid culture, if the treating clinicians documented the culture as true infection and not contamination, and if the infant received appropriate antibiotic treatment of the presumed infection (bacteremia, UTI, or meningitis). Positive bacterial culture results in infants who did not receive a full treatment course of antibiotics or those reported to be a contaminant were not considered a BI. For patients with growth on urine culture, the presence of urinary white blood cells ≥3 cells/HPF, positive leukocyte esterase, or positive nitrites, was also required to be considered a BI. The urinary white blood cell cut-off differed from the AAP practice guideline because of the hospital laboratory policy of reporting urinary white blood cells in a range of 3 to 5 cells/HPF.¹⁸ 

For comparison, patients were grouped according to BI and CVI diagnoses: positive for both BI and CVI (BI+/CVI+), positive BI but negative CVI (BI+/CVI−), negative BI and positive CVI (BI−/CVI+), and negative for both BI and CVI (BI−/CVI−). PCT values were compared between pairs of patient groups by using the Wilcoxon rank test in the entire cohort, as well as stratified by age: 8 to 28 days and 29 to 60 days. Because of the small number of patients in the 22- to 28-day-old age range, this group was combined with the 8- to 21-day-old age range.

Logistic regression and area under the curve (AUC) analysis were used to assess the odds of febrile infants having a BI in the setting of an increased PCT value, both with and without a CVI. Measures of predictive ability (sensitivity, specificity, positive predictive value, and negative predictive value) were estimated at the AAP clinical practice guideline recommended cutoff PCT value of 0.5ng/dL.³ 

In this study, 1601 patients were identified and underwent subsequent chart review. A total of 938 (59%) patients were excluded. Of these, 626 (67%) patients met initial exclusion criteria, had an obvious source of infection for which cultures could not be obtained, or had no blood and/or urine cultures obtained or resulted. An additional 218 (23%) were excluded because of inadequate viral testing. Finally, 94 (10%) were excluded because of a lack of PCT testing or unknown results. Of these, 55 (59%) were transferred from outside institutions in which laboratory work and antibiotic administration occurred before the transfer, 10 (11%) had a PCT ordered but no results due to laboratory error or inappropriate sampling, and 29 (31%) had no PCT ordered due to clinician decision.

In total, 663 (41%) patients remained for analysis. Of these, 404 (61%) had a confirmed viral illness alone, 73 (11%) had BI alone, 48 (7%) had both bacterial and viral infection, and 138 (21%) had no infection. Demographics between the groups revealed a higher proportion of 29- to 60-day-old infants in all groups, except for the BI−only category (Table 1). Also, there was a higher proportion of boys in all groups (Table 1).

Of the 73 patients with BI alone, 6 (8%) had a UTI plus an invasive BI (bacteremia and/or meningitis), 61 (84%) had UTI only, and 6 (8%) had only an invasive BI. Of the 48 patients with viral illness plus BI, 45 (94%) had a comorbid UTI, 1 (2%) had a comorbid UTI and invasive BI, and 2 (4%) had a comorbid invasive BI (Table 2).

Compared with infants without any infection, the presence of a bacterial illness was associated with a significantly higher median PCT level overall and when stratified by age (BI+/CVI− vs BI−/CVI−, P < .01; BI+/CVI+ vs BI−/CVI−, P < .01; Table 3). A significant difference was also noted when comparing median PCT levels in infants with a confirmed viral plus bacterial illness to those with viral illness alone (BI+/CVI+ vs BI−/CVI+, P ≤ .01). Finally, median PCT levels were significantly higher for infants with a bacterial illness alone when compared with those with viral illness alone (BI+/CVI− vs BI−/CVI+, P < .01).

In the absence of bacterial illness, median PCT levels were not significantly different between infants with and without confirmed viral illness (BI−/CVI+ vs BI−/CVI−, P = .74; Table 3). However, in the presence of a bacterial illness, comorbid confirmed viral illness significantly reduced median PCT levels (BI+/CVI+ vs BI+/CVI−, P < .01). This was true in the overall comparison, but statistical significance was not detected when evaluated separately by age subgroups.

Logistic regression analysis of rising PCT levels revealed significantly increased odds of a BI when a viral infection was present when compared with patients without a viral illness. (2.98 vs 1.15; P < .01; Table 4, Fig 1). When adjusted for both age and sex, the change in odds ratio estimates was negligible and continued to demonstrate a significant difference between those with versus those without a viral illness (2.95 vs 1.13; P < .01).

AUC analysis, using the PCT cutoff value of 0.5 ng/mL, revealed a statistically significant decrease in sensitivity for detecting a BI when a confirmed comorbid viral infection was present (68% vs 44%; P < .01). Specificity revealed no significant change in detection of a BI, irrespective of the presence of a comorbid viral infection (93% vs 96%; P = .17). Finally, the positive predictive value of PCT revealed a significant decrease in the detection of a BI when a comorbid viral infection was present (85% vs 58%; P < .01), whereas the negative predictive value significantly increased (85% vs 94%; P < .01).

In our cohort of otherwise healthy febrile infants aged 8 to 60 days, PCT levels were significantly lower for infants with a BI and a comorbid confirmed viral illness when compared with those with a BI alone. This further reduced the sensitivity of this inflammatory marker, supporting previous observations cautioning the use of PCT as a single marker to guide clinical decision making when the level is low or in the normal range. However, the PCT levels in infants with comorbid bacterial and viral infection remain significantly higher than those in infants without a BI. Additionally, the elevation of PCT maintains its specificity in the identification of BI in febrile infants regardless of whether viral illness is present and consistently indicates cause for heightened concern. This is further supported by the significantly increased odds of a BI being present when the PCT level is elevated and a comorbid viral infection is present.

The use of PCT as a predictor of BI in children has gained favor in recent years. Studies in children, ranging from 8 days of age to 3 years, reveal that PCT is a more accurate predictor of BIs than CRP, white blood cell count, and absolute neutrophil count.^11,13  Studies also reveal that PCT is a more specific predictor of BI than CRP early in the infectious process.¹²  However, the authors of reports note that PCT is not reliable as a sole predictor of BI because of the relatively low sensitivity of the test. One study reveals that PCT values >0.5 ng/mL have a sensitivity of 44% when used to identify infants <21 of age days at risk for BI, whereas another reports the sensitivity of this PCT cutoff to be 73% in patients aged 7 days to 36 months.^12,17  Of note, these studies focused on children with bacteremia and meningitis (invasive BIs) without including those with UTIs. Despite the inclusion of infants with UTI, our study supports these previous results, revealing a high specificity of PCT in detecting infants with a BI alone (93%) but a low sensitivity (68%) when utilizing the AAP recommended test cutoff of 0.5 ng/dL.³ 

The authors of recent studies that include PCT in the development of a tiered approach to the evaluation of the febrile infant do not address whether a viral illness is present, making the impact of viral illness on this marker in the infant population unclear.^2,9  Viruses are theorized to induce cytokines, such as interferon-δ, that may inhibit PCT production.¹⁰  Our data support this theory by revealing an associated decrease in PCT levels in infants with a BI and concurrent viral illness when compared with patients with a bacterial illness alone. Despite this decrease, we found that the presence of a CVI did not change the specificity of the test in detecting a bacterial illness. Although we acknowledge that most patients in our study with a bacterial illness had a less-invasive form (UTI), the clear association of a comorbid viral illness with decreased PCT levels leads to concern that this alteration may extend to patients with more invasive BI.

The detection of a CVI in an infant with fever has been associated with a decreased risk of, and therefore a concern for, BI. One previous study reveals a statistically significant reduction in the risk of bacteremia or meningitis in infants aged 29 to 60 days in the presence of a respiratory virus.¹⁴  Another reveals that infants identified as high-risk via Rochester criteria with a CVI have a significantly lower risk of UTI, bacteremia, or meningitis than infants classified as high-risk without a CVI.¹  Finally, the authors of a meta-analysis of 11 studies examined the rate of UTI, bacteremia, and meningitis in patients with confirmed respiratory syncytial virus or bronchiolitis, demonstrating that only UTIs occurred with significant frequency.¹⁶  Although these studies reassure the practitioner of a lower risk of serious illness when a virus is present, consideration must be given to the influence of the virus on the laboratory markers used to assess the infant. Importantly, our study reveals that an elevated PCT level in the setting of a viral illness remains highly specific, and reveals increased odds, for a BI. This confirms that practitioners consider workup and antibiotic therapy in patients with an elevated PCT, even when a viral infection is present. Practitioners also should remember that the sensitivity of a normal PCT level is further reduced and should rely on clinical appearance and other laboratory markers in this patient population.

Our study has several limitations to consider. This is a single-site, retrospective chart review and is subject to a lack of data inclusion and the incorrect coding of diagnoses. We had a preponderance of males in all groups, which raised a question about population validity. However, through a review of the available literature, we found this trend to be consistent with other studies of febrile infants.^5,6,11,15  We acknowledge that negative test results on the PCR panel do not exclude infection due to viruses not included on the panel. Also, our study had a high exclusion rate, which can raise concern about a lack of generalizability or selection bias. We acknowledge that well-appearing infants may not have had a PCT obtained, resulting in selection bias. However, this impact is limited by our institution’s protocol, which encourages obtaining PCT levels in all febrile infants, resulting in a low number of patients (29 out of 938 [3%]) excluded because of clinician decision not to obtain this test. In addition, we had a higher proportion of patients with a bacterial illness (18%) than rates reported in previous studies. One explanation for this may be our institution’s position as a large referral center, in which most patients in the region with abnormal laboratory results and positive cultures are transferred for further care.

Our study is also limited by the relatively small number of patients with BI and comorbid CVI. This limitation likely influenced the lack of statistical significance in the age subgroup comparison of infants with bacterial illness, with and without confirmed viral illness. This also limited our ability to perform a subanalysis focusing on comorbid invasive BIs (bacteremia and/or meningitis). Further studies with larger patient populations are needed to determine if the alteration in PCT level and accuracy seen in this study remains consistent when evaluating patients with a comorbid viral illness and invasive bacterial disease.

Our study highlights the association of a CVI with the reduced sensitivity and unchanged specificity of PCT as a marker used to assess for BI in febrile infants aged 8 to 60 days. Our study also highlights the increased odds of a BI being present in the setting of an elevated PCT and confirmed viral illness. The clinical implication is important because it supports that patients with PCT levels elevated >0.5 ng/mL warrant further evaluation for a BI, as currently recommended by the AAP guidelines, even in the presence of a CVI. Additional studies are underway to specifically address the impact of comorbid viral illness on PCT levels in infants with invasive BI.