Association Between Procalcitonin and Antibiotics in Children With Community-Acquired Pneumonia Free

This study was nested within a prospective observational cohort study of children with pneumonia presenting to the ED from December 2014 to January 2017 at 2 tertiary care children’s hospitals in the United States. For this analysis, only subjects enrolled in the larger study with blood available for PCT measurement were included. The study was approved by the institutional review boards at both institutions.

Children ≥2 months and <18 years were eligible if they presented to the ED with evidence of acute respiratory illness (coughing, fast breathing, retractions or chest indrawing, abnormal lung exam, or respiratory failure); evidence of acute infection (documented or reported abnormal clinical radiology report and confirmed by the principal investigator at the 2 study sites. For cases in which radiographic interpretation was equivocal, cases were included only if clinical documentation supported a pneumonia diagnosis. Children were excluded if they had any of the following: tracheostomy, cystic fibrosis, or immunosuppression, hospitalization for any reason within the preceding 7 days (including interhospital transfers); if they were enrolled in the study within the previous 28 days, if they were a resident of an extended care facility, or had the presence of a clear alternative diagnosis.

All enrolled children and their caregivers were interviewed to collect sociodemographic characteristics, medical history, and history of present illness. Medical records were reviewed after discharge to collect antibiotic use data and other encounter details.

Blood samples for PCT measurement were collected at the time of enrollment and frozen at −80°C, followed by analysis on the VIDAS B.R.A.H.M.S instrument according to standard protocols.¹⁶  At both hospitals, PCT was collected exclusively for research purposes, and results were not available in real-time, so clinicians were blinded to these results. This approach allowed us to determine if decisions around antibiotic use were associated with the likelihood of bacterial pneumonia as measured by PCT.

The primary outcome was antibiotic initiation, defined as administration of systemic antibiotics on the day of ED triage or the day after triage (day 0 or day 1). For our planned secondary analysis of subjects receiving antibiotics, empirical antibiotic selection was classified as either narrow-spectrum or broad-spectrum. Narrow-spectrum antibiotic use was defined as use of either ampicillin or amoxicillin with or without a macrolide. Broad-spectrum use was defined as any antibiotic use that did not fall into the narrow-spectrum category. Those receiving both narrow- and broad-spectrum antibiotics were classified as broad-spectrum.

Baseline characteristics of the population were summarized by PCT groups corresponding to varying likelihood of bacterial etiology: <0.1 ng/mL (very low), 0.1 to <0.25 ng/mL (low), 0.25 to <1.0 ng/mL (moderate), and ≥1.0 ng/mL (high). For the primary analysis, multivariable logistic regression was used to model the association between PCT as a continuous exposure variable and antibiotic initiation. The prespecified covariates were selected on the basis of hypothesized association with PCT level or antibiotic usage and included age, comorbidities, triage disposition, previous antibiotic use, and viral testing. Comorbidities were defined as chronic health conditions that either: (1) require daily medications; (2) require contact with specialty health care providers or more frequent contact with a primary health care provider; or (3) illness that limits abilities to perform age-appropriate activities (sometimes, often, or always). Previous antibiotics included antibiotics received before ED arrival as well as antibiotics given in the ED before PCT being measured. Odds ratios with 95% confidence intervals were reported.

Among those receiving empirical antibiotics, we also evaluated the relationship between PCT and antibiotic selection. For this analysis, we constructed a second multivariable logistic regression model in which empirical antibiotic selection (narrow- or broad-spectrum) was the outcome. Covariates included in this model were identical to those in the primary analysis.

We also performed descriptive analyses to identify potential opportunities for stewardship interventions. For this, we examined antibiotic initiation and selection among all children with PCT values corresponding to very low risk (<0.1 ng/mL) or low risk (0.1 to <0.25 ng/mL) for bacterial disease.^17,18  Results were stratified by the presence or absence of critical illness, defined as the need for intensive care or the presence of parapneumonic effusion requiring drainage at any point during the encounter. We reasoned that children with noncritical illness could be targeted for antibiotic-reducing interventions using PCT, whereas those with critical illness would nearly always receive empirical antibiotics regardless of PCT value. Finally, we evaluated patient outcomes, focusing on critical illness, length of stay (LOS), and postempirical antibiotic use as key outcomes to assess antibiotic appropriateness.

There were 1116 children enrolled in the main cohort study. For this analysis, 488 (44%) with provider-confirmed radiographic pneumonia and a blood sample obtained for PCT analysis were included (Fig 1). The median age was 5 years (IQR [interquartile range] 2–10, Table 1). A total of 221 (45%) children had an underlying comorbidity, with asthma being the most frequent. Nearly all children (n = 463; 95%) were hospitalized, including 125 (27%) who required intensive care. Most (n = 411; 84%) received antibiotics before having a PCT sample drawn (mean 15.8 hours after first antibiotic). Just over one-half (n = 263; 54%) had viral testing performed. Those in higher PCT groups were more likely to have testing performed and to have a positive result (Table 1).

The overall median PCT was 0.36 ng/mL (IQR 0.10–2.38, range 0.025–200). There were 115 (24%) children in the very low likelihood for bacterial etiology PCT group, 99 (20%) in the low likelihood group, 103 (21%) in the moderate likelihood group, and 171 (35%) in the high likelihood group (Table 1). Children in the lower likelihood groups were significantly older, whereas those in the higher likelihood groups were more likely to have chest retractions or indrawing.

There were 403 (83%) children with empirical antibiotic use (Fig 2). Of those receiving antibiotics, 308 (76%) received broad-spectrum therapy (median PCT, 0.46; IQR, 0.12–2.83) and 95 (24%) received narrow-spectrum therapy (median PCT, 0.24; IQR, 0.08–2.52). Eighty-five children (17%) did not receive antibiotics (median PCT, 0.32; IQR, 0.09–1.33). In multivariable analyses, PCT was not associated with antibiotic initiation (OR [odds ratio], 1.02; 95% CI [confidence interval], 0.97%–1.06%) or with the spectrum of antibiotic use (OR, 1.07; 95% CI, 0.97%–1.17%).

Among all children, 452 (93%) had no change in the empirical antibiotic treatment plan, including 92% of those receiving broad-spectrum therapy, 91% receiving narrow-spectrum therapy, and 98% receiving no empirical antibiotics. Among the 85 children receiving no empirical antibiotic therapy, 46 (54%) had PCT values >0.25; only 2 children had antibiotics later initiated. Children receiving narrow-spectrum empirical therapy later switched to broad-spectrum therapy (7 of 95; 7%). Children receiving empirical broad-spectrum antibiotics were more likely to have critical illness (34%) and longer LOS (median 64 hours) as compared to those receiving narrow-spectrum antibiotics (10%; 43 hours) or no empirical antibiotics (20%; 46 hours; P < .001 for both comparisons).

There were 167 children with noncritical illness and PCT values <0.25 ng/mL indicating low likelihood for bacterial pneumonia (Table 2). Among this group, 136 (81%) received empirical antibiotics, with 92 (68%) receiving broad-spectrum antibiotics (Table 2). Additionally, 110 of 189 (58%) children with noncritical illness and PCT values >0.25 ng/mL received broad-spectrum therapy. We considered these 246 children (69% of all children with noncritical illness) to represent potential opportunities for antibiotic-reducing interventions. Among the 132 children with critical illness, 47 (36%) had PCT values <0.25 ng/mL.

In this study, PCT concentration was not associated with empirical antibiotic initiation or selection among nearly 500 children with pneumonia. The vast majority of children received antibiotics, and broad-spectrum antibiotics were used most often. Nearly one-half of children had PCT values consistent with very low or low likelihood of bacterial etiology. Whereas many factors may influence decisions around antibiotic initiation and selection, the observed discordance between likelihood of bacterial etiology on the basis of PCT and empirical antibiotic treatment decisions suggest opportunities for stewardship.

Reducing unnecessary antibiotic use is a principal component of the “Choosing Wisely” initiative.¹⁹  Additionally, consensus national childhood pneumonia management guidelines stress the importance of limiting antibiotic exposure whenever possible.²⁰  Despite studies demonstrating a very high prevalence of viral etiologies and a low prevalence of bacterial etiologies, most children in our study received antibiotics.⁴  This is consistent with data indicating that most children with pneumonia continue to receive antibiotics in outpatient settings²¹  and that antibiotics may be unnecessarily prescribed to hospitalized patients overall.²²  Given the lack of highly sensitive bacterial diagnostics,^5,6  an objective biomarker such as PCT could enhance stewardship efforts. A previous study found that PCT had a sensitivity of 85% for typical bacterial disease at the <0.25 ng/mL (low likelihood) cut-off and a 100% sensitivity at the <0.1 ng/mL (very low likelihood) cut-off.²³ 

We found that many children at very low or low likelihood for bacterial etiology on the basis of PCT concentration received antibiotics, with a high frequency of broad-spectrum antibiotic use, a finding that conflicts with guideline recommendations promoting narrow-spectrum therapy in the majority of children.²⁰  The low rate of narrow-spectrum antibiotic use in our study also aligns with previous studies which demonstrate low overall use of narrow-spectrum therapy and substantial variability across centers for childhood pneumonia.^24–26  We noted that changes to the initial treatment approach were rare, including among those receiving no empirical antibiotics and those receiving narrow-spectrum therapy. This suggests that these approaches were safe and effective and that stewardship efforts may be most beneficial at the time initial treatment decisions are made. To that end, we found important potential opportunities for stewardship interventions, with two-thirds of noncritically ill children representing potential targets for either eliminating empirical antibiotic therapy or substituting narrow- for broad-spectrum therapy. We also noted that 35% of those with critical illness had PCT values indicating low likelihood of bacterial etiologies. Whereas it may be more difficult to limit antibiotic use in critically ill children,²⁷  these data suggest that viral etiologies remain prevalent even among those with severe disease.

A meta-analysis of 26 randomized controlled trials in adults found that PCT guidance for the initiation and duration of antibiotic use led to less days of antibiotic exposure and less antibiotic side effects.²⁸  The literature for pediatric studies is more sparse. Two trials concluded that PCT reduced antibiotic prescribing (14% to 30%) as compared to standard of care in children with lower respiratory tract infection,^17,18  whereas one-third found no difference.²⁹  In 1 of these studies, however, use of PCT was also associated with increased antibiotic use among those with nonpneumonia lower respiratory tract infection for whom antibiotics are not routinely indicated.¹⁸ 

Importantly, we noted that over one-half of children receiving no empirical antibiotics had PCT values >0.25 with only 2 children later receiving antibiotics. This underscores the poor specificity of PCT (45% at <0.25 ng/mL)²³  and is a reminder that this biomarker does not replace clinical judgement and should be used judiciously and only when antibiotic treatment decisions are equivocal. Use of PCT also contributes to increased costs, and in those who otherwise may not require a blood draw, discomfort and inconvenience. Thus, value may be low unless it results in meaningful and safe reductions in antibiotic use. A definitive, pragmatic multicenter effectiveness trial is needed to confirm the utility and value of PCT in a variety of clinical care settings. Coupling PCT with effective clinical decision support is a particularly attractive strategy that could help to minimize unnecessary antibiotic initiation and improve adherence to guideline recommendations when antibiotics are needed.^30–32 

There were several limitations to our study. Many children received antibiotics before collection of a blood sample for PCT measurement (mean 15.8 hours after first antibiotic). Although previous antibiotic use was incorporated into our models, the PCT values in our study may be artificially lower than if no antibiotics had been administered. Given the expected half-life of PCT with effective treatment of bacterial pathogens (25–30 hours),⁷  as well as the high prevalence of viral etiologies in childhood pneumonia,⁴  the impact of this limitation on our findings is likely modest. Our study was not designed to capture comprehensive etiologic data, and we were not able to correlate PCT concentrations with microbiologic etiology. Instead, we infer the relationship between PCT and bacterial etiology on the basis of current evidence and its safety in previous clinical trials.^7–10,33,34  In our analysis examining targets for potential antibiotic stewardship interventions, we acknowledge that other unmeasured factors beyond critical illness may necessitate empirical antibiotics or use of broad-spectrum therapy (eg, failed outpatient therapy, penicillin allergy) regardless of PCT values. Stewardship opportunities may also be reduced in areas in which use of broad-spectrum therapy is less prevalent. Further, a substantial fraction of children enrolled in the parent cohort study did not have blood available for PCT analysis, and few children discharged from the ED were included in our study. Although this is indicative of clinical practice with well-appearing children typically not receiving clinical blood draws, it suggests that our study may not be fully reflective of all children with pneumonia presenting for emergency care. Finally, our study was conducted at two tertiary care children’s hospitals and results may not be generalizable to general and community pediatric hospitals.

Our study found that neither empirical antibiotic initiation nor selection in children with pneumonia were associated with PCT concentration when clinicians were blinded to PCT results, suggesting a discordance between likelihood of bacterial etiology and antibiotic treatment decisions. Nearly one-half of children had PCT values indicating very low or low risk for bacterial pneumonia, although >80% received antibiotics, and antibiotic selection was often inconsistent with national guideline recommendations. Use of PCT to inform antibiotic treatment decisions may help to promote more judicious antibiotic use.

The authors acknowledge the collective efforts of the ICE-CAP Study Team for making this work possible.