Impact of Cell-Free Next-Generation Sequencing on Management of Pediatric Complicated Pneumonia Free

We conducted a retrospective, single-site study to examine the health records of children aged 3 months to 18 years admitted to the Pediatric Hospital Medicine service at a tertiary children’s hospital between January 24, 2018, and December 31, 2020. These children were diagnosed with cCAP and had a cfNGS diagnostic workup. Criteria for the diagnosis of cCAP were defined by the following Pediatric Infectious Disease Society and Infectious Diseases Society of America guidelines: pneumonia complicated by a parapneumonic effusion, empyema, lung abscesses, or necrotizing pneumonia.⁵  Our institution does not have standardized criteria for ordering cfNGS. cfNGS is currently limited to the pediatric infectious diseases consulting physicians who collaborate in the care of these children with cCAP. Patients were identified in the electronic medical record (EMR) by (1) cfNGS sample tested while the patient was hospitalized and (2) the hospital problem list containing any of the following International Classification of Diseases, 10th Revision, codes: J16.8 (pneumonia due to other specific organisms), J18.8 (pleuropneumonia), J18.9 (pneumonia, CAP), J85.0 (necrotizing pneumonia), J85.2 (lung abscess), J90 (pleural effusion, parapneumonic effusion), J86 (empyema), or J91.8 (pleural effusion in conditions classified elsewhere). Our institutional review board approved this study, and a waiver of informed consent was granted.

We performed an EMR chart review for all children with cCAP to ascertain and validate clinical characteristics at admission that met the diagnosis of cCAP. The pleural effusion size was defined as small if it was less than one-fourth of a hemithorax, moderate if greater than one-fourth but less than one-half of a hemithorax, and large if greater than one-half of a hemithorax.⁵  Children were excluded if they had nosocomial pneumonia, ventilator-associated pneumonia, an oncologic diagnosis undergoing active treatment, a nononcologic immunocompromising diagnosis, or underlying abnormal lung pathology. Children with active tuberculosis or coccidiomycosis were also excluded. Charts of children who may have been given antibiotics in other medical facilities before hospitalization were not always available; thus, we relied on parent report as documented in the EMR for this information. Data were extracted from the EMR and transferred to REDCap for analysis, including patient demographics, laboratory results, microbiology results, respiratory support, radiographic results, thoracostomy placement, and antibiotic therapy. There were 2 reviewers per patient, and in any case where there was a discrepancy in interpretation of results or outcomes, a third reviewer (J.S.B.) adjudicated the result.

The results of cfNGS testing were compared with conventional diagnostic test results obtained during the hospitalization. Conventional diagnostic testing included blood cultures, respiratory aspirate cultures (sputum samples, tracheal aspirates, and bronchoalveolar lavage [BAL]), body fluid cultures (pleural fluid and pleural abscess), respiratory virus multiplex PCR panel (RV-PCR) studies, and Mycoplasma serologies. Blood culture results were considered positive if there was growth of recognized pathogens. Blood cultures showing growth of organisms not characteristic of cCAP (eg, coagulase-negative staphylococci) were not considered to yield positive results. Respiratory tract bacterial and fungal cultures with <1+ growth of a commensal organism were considered as yielding negative results. No viral cultures were obtained. We report positive nasopharyngeal PCR multiplex panel (NP-PCR) results for viruses but did not compare these with cfNGS results. Chart review was done to determine if and how cfNGS results changed management at the time they were reported through our hospital’s laboratory. Change in management was defined as changing therapy to a different antibiotic (narrowed spectrum of antibiotic or increased spectrum of antibiotic) and/or change in treatment duration (shorter or longer course). Change in management was restricted to the documented impression and plan in the EMR by the clinical team caring for the child at the time the cfNGS results were returned. We were not able to analyze time from initial antibiotic administration to the time of thoracotomy/chest tube placement with culture of pleural fluid because this information was not readily available for all patients. We did not review charts of children with cCAP who were not tested by cfNGS and, therefore, were not able to compare length of stay with children for whom cfNGS testing was performed.

Plasma samples were sent to the Karius company laboratory in Redwood City, California, for NGS. Karius was not involved in the decision to test, was not involved in the data analysis for this project, and did not provide funding for this study.

Basic descriptive statistics were used to characterize the sample. Patient characteristics and other categorical variables are reported using frequencies. Continuous variables were reported using median and interquartile range (IQR).

During the study period, 86 children were admitted to our hospital for cCAP and underwent cfNGS testing. Forty-six met our inclusion criteria (Fig 1). Clinical characteristics are described in Table 1. Sixteen children (35%) received antibiotics before admission, and 38 (83%) required oxygen during admission. The median duration of supplemental oxygen was 6 days (IQR 3–10), and of those who required oxygen, 12 (32%) required mechanical ventilation. Most children had either moderate (17 of 46; 37%) or large (23 of 46; 50%) pleural effusions, and 37 (80%) required thoracentesis. The median total antibiotic duration was 26 days (IQR 22–30), which included both the duration administered while hospitalized and the duration prescribed after discharge. The median total days of intravenous antibiotics was 11 days (IQR 8–16). Twenty-four (52%) of the 46 patients required admission to the ICU, with a median ICU length of stay of 6 days (IQR 4–11) (Table 1). There were no deaths.

Twelve (26%) of the 46 children had positive culture results from blood, respiratory, and/or pleural fluid or abscess specimens. Blood cultures were performed in 34 (74%) of the 46 children, with 1 (3%) yielding positive results. Twelve of these 34 children received antibiotics before admission, and thus, these blood culture results may have been altered. Of the 14 respiratory cultures (10 of which were endotracheal aspirates in intubated children, 3 from sputum samples, and 1 from BAL) from 12 children, 1 (8%) yielded positive results, although this child had respiratory cultures positive for the same organism from both an endotracheal aspirate and a BAL sample. Forty-two pleural fluid or abscess cultures were performed for 37 children, 14 of which from 10 children (27%) had positive results. RV-PCR were performed for 32 children (70%) and yielded positive results in 14 (44%) (Table 2). Mycoplasma serologies (IgM) were performed for 2 children (4%), both of which yielded positive results; 1 was an immunoblot test (Meridian Bioscience) and the other an immunoassay (Quest Diagnostics). Fungal cultures were performed for 13 children (28%), 12 of which were from pleural fluid and 1 from BAL. No fungal cultures yielded positive results.

cfNGS results were positive in 45 (98%) of 46 children and deemed likely to be pathogenic by the treating team in 41 (89%). The most common pathogens detected were S pneumoniae in 27 children (60%), S aureus in 5 (11%) and Fusobacterium nucleatum in 5 (11%) (Table 2). Full microbiologic test results are available in Supplemental Table 3. In 4 of the 6 children in whom testing detected oral flora (F nucleatum and/or Streptococcus intermedius) with no recognized typical CAP pathogen detected, the clinical history identified a possible aspiration event before diagnosis of cCAP.

Diagnosis of the likely causative organism was made by cfNGS alone in 32 (70%) of 46 children, by cfNGS and a conventional diagnostic test (cfNGS and conventional test yielded the same result) in 9 (20%), and by a conventional diagnostic test alone in 3 (6.5%). In the 3 children diagnosed by conventional testing alone, culture from pleural fluid was positive for S intermedius in 1, pleural fluid culture was positive for Streptococcus pyogenes in 1, and tracheal aspirate and BAL cultures were positive for α-hemolytic Streptococcus in 1 (not further speciated but correlated with metagenomic NGS testing from the same source; IDbyDNA, Salt Lake City, UT). In 2 children (4%), neither cfNGS nor conventional diagnostic testing were positive for a likely pathogen (Table 2).

The results of cfNGS changed management in 36 (78%) of 46 children with the use of narrower-spectrum antibiotics in 29 (81%). In some cases, cfNGS affected antibiotic management by confirming a culture isolate that was not initially believed to be the cause of infection or by confirming a persisting pathogen in a child with ongoing or worsening signs and symptoms of infection and concerns that additional bacterial pathogens may have been present, requiring an increase in the number of antibiotics used to treat the child. In 6 children (17%), antibiotics were changed based on pathogen identification, and in 1 of these children, the duration of antibiotic treatment was also shortened because cfNGS testing confirmed a diagnosis of Mycoplasma pneumoniae. This patient was thus treated with a 5-day course of azithromycin. In another child, cfNGS results led to broadening antibiotic therapy on the basis of the diagnosis of aspiration pneumonia with the identification of F nucleatum and S intermedius.

In this retrospective chart review of 46 children admitted to a tertiary children’s hospital with cCAP, cfNGS results were interpreted as positive for a causative organism in 41 (89%) and were the only means of pathogen identification in 32 (70%). The results of cfNGS changed management in 36 children (78%), leading to an antimicrobial regimen with a narrower spectrum of activity in 29 (81%). We believe that this report represents the largest number of children hospitalized with cCAP where both standard diagnostic testing and cfNGS techniques were used in the diagnosis of the causative pathogen for infection. These results, with a high diagnostic yield allowing for change in management in most children, are promising for the use of cfNGS in this patient population.

Children admitted with cCAP often receive broad-spectrum and prolonged antimicrobial therapy that may not be required if specific pathogens could be identified. Although published data indicate that S pneumoniae remains the most common bacterial cause of CAP, the high risk of incurred morbidity with cCAP often compels clinicians toward more conservative, broad-spectrum treatment strategies. Identification of a pathogen can help clinicians to narrow antimicrobial coverage, but the utility of conventional culture methods is limited. Blood cultures infrequently yield positive results (∼2.5%–10% of children hospitalized for CAP),^5,14,15  which we saw in our study (3%). Some of the differences in isolation rates may be explained by a decreased ability to isolate pathogens from blood cultures obtained after antibiotics have been given, as 35% of the children in our study were treated with antibiotics before hospitalization and collection of samples for blood culture. Although the rate of positive results from pleural fluid/empyema cultures is higher than that for blood cultures, not all patients undergo thoracentesis if the amount of fluid is small.^5,16–18  If pleural fluid is obtained, cultures are typically only positive in approximately one-third of cases,^6–8  similar to our results. NP-PCR may be helpful in identifying viral/bacterial coinfection,⁵  as bacterial infections may develop in association with a viral infection. Compared with rates of pathogen identification using conventional testing, which range from 34% to 43%, our study revealed a higher rate of pathogen detection.^6–8  Despite us showing promise in the utility of cfNGS in pediatric cCAP, a limitation of cfNGS is that it does not reveal susceptibility, thus limiting how much antimicrobial coverage can be narrowed, particularly for strains of S aureus.

cfNGS is a culture-independent, noninvasive diagnostic technique that allows for sensitive and broad sampling for >2000 organisms with a single plasma sample that may return results within 48 hours.⁹  Although some studies have demonstrated clinical use in heterogeneous adult and pediatric populations,^9,19,20  others have described lower yield compared with conventional diagnostic testing in heterogenous cohorts without well-defined test indications in specific patient populations.¹³  The authors of a recent retrospective review of cfNGS use at a pediatric tertiary facility described a change in management in 32.4% of cases, with the cohort heavily weighted toward patients with immunocompromise.²⁰  We believe that the differences in the patient population being studied may account for the differences in the rate of pathogen detection between our study and other studies evaluating the use of cfNGS and may also account for the number of cfNGS test results with >1 organism (our cCAP cohort had relatively fewer cfNGS results with multiple pathogens compared with other populations). Studies are needed to document cfNGS test characteristics for patient populations, such as children with immunocompromise, whose blood samples may yield cfNGS signals of varying strengths from multiple organisms. Patients with immunocompromise were excluded from our study, and we only examined children with the diagnosis of cCAP. Indeed, of the 10 patients who had >1 pathogen, 2 had >2 pathogens, and in most children in whom an organism was identified by cfNGS, the organism was a recognized known cause of pediatric CAP.

Studies examining the time course of detectable plasma microbial DNA after antibiotic treatment are generally lacking; however, available data suggest that cell-free DNA fragments persist longer than live bacteria in bloodstream infections²¹  and may persist during the course of bacterial death. We have previously shown that a higher yield of cfNGS in cCAP compared with blood culture results is in keeping with these findings.¹²  Therefore, it is likely that treatment with antibiotics for a short duration before cfNGS testing should not significantly affect the results because the test analyzes nucleic acid fragments, which should still be present. If a patient had received appropriate treatment of a longer period, then it is possible that the microbial nucleic acid load will have decreased to an undetectable level.

Although to our knowledge, this study is the largest of children with cCAP diagnostic testing by cfNGS, it is limited by retrospective patient selection and lack of standardized criteria for cfNGS testing. Despite stringent application of inclusion criteria for a group believed to have the highest likelihood of cfNGS results positive for a significant pathogen load, approximately one-third of the children had received antibiotic therapy before hospitalization, which may have biased the findings toward an overly favorable impression of cfNGS for all children with cCAP. Interpretation of the cfNGS results was based on the documented impression of the clinical team in the EMR and is therefore still somewhat subjective. Furthermore, whether the change in antibiotics was solely due to the cfNGS results is subjective, especially in children who had positive conventional test findings and cfNGS results that were concordant; however, to our knowledge, there were no other results or clinical factors that influenced the medical decision-making based on chart review. Orthogonal confirmation of the cfNGS result with positive culture results or other molecular testing was not available for most of the positive findings. Other potential limitations of the cfNGS test include cost, and we did not perform a cost analysis. The cfNGS test used is commercially available, but at this time, it is only performed at the manufacturer’s reference laboratory. The turnaround time may therefore be a limitation while only available at reference laboratories.

For children with serious presumed bacterial cCAP disease, particularly those who have received empirical antibiotics before hospitalization and in those who do not undergo invasive pulmonary or thoracic procedures for cultures, cfNGS may reveal the bacterial etiology of infection to assist in patient management. We believe that our data reveal the concept that cfNGS is helpful in the care of children with cCAP and support efforts toward enhanced availability of this technology.

In conclusion, in this retrospective series of 46 children with cCAP, we show that cfNGS testing identified a pathogen in 41 (89%) and affected management in 36 (78%). Although these uncontrolled data are promising, well-designed prospective studies are needed to best characterize the clinical use of cfNGS in children with complicated pneumonia as well as other infectious diseases.

The authors thank Jennifer Foley and Andrew Richardson for their generous support during this study.