Background

Community-acquired pneumonia (CAP) is a common respiratory disease in children with significantly higher morbidity and mortality rates in developing countries than that in developed countries [1]. According to the World Health Organization (WHO) report, the incidence of CAP in children under 5 years of age in developing countries is 0.28 times/child/year, accounting for 95% of all cases of CAP in children worldwide, and the mortality rate is 1.3–2.6% [2]. In North America and Europe, the incidence rate among pre-schoolers is 36 per 1,000 children per year [3], placing a significant economic burden on families and society. Children with CAP are prone to develop pleural effusion, lung abscess, pericarditis, meningitis, and other pulmonary and external complications [4], which eventually lead to severe community-acquired pneumonia (SCAP), respiratory and circulatory failure, or even life-threatening complications. For the pathogen diagnosis of CAP, clinical work is carried out in various microorganisms, including viruses, bacteria, and atypical pathogen combination detection to avoid missed diagnosis and mixed infection, although pathogens cannot be identified or detected in 22–57% of children [5,6,7] and 13–62% of adult with pneumonia [8, 9]. Culture, the gold standard for bacterial and fungal infection diagnosis, has high accuracy but a long detection time and low detection rate and is not suitable for the identification of pathogens that are difficult to cultivate. Viruses and atypical pathogens can be detected by polymerase chain reaction (PCR) and antigen detection, but the positive rate varies greatly and suitable for detecting only the known, limited target pathogens. In addition, some unknown pathogens in nature cannot be detected. Therefore, there is an urgent need for rapid and accurate diagnostic methods to identify pathogens in children with SCAP [10].

Metagenomic next-generation sequencing (mNGS) is a method for whole nucleic acid detection using patient samples directly [11], which can simultaneously detect various microorganisms. It has the characteristics of being less affected by antibiotics, taking a wide range of samples, and does not need to set gene sequences in advance. Therefore, it greatly applicable in diagnosis of new and rare infectious diseases [10, 12,13,14]. Currently, mNGS has been applied in the diagnosis of infectious diseases of the lungs, blood system, and central nervous system in adults [15,16,17]. Sepsis, encephalitis, and bone and joint infections have also been reported in children [18,19,20]. However, no study has been conducted on bronchoalveolar lavage fluid (BALF) in children with SCAP. In this study, 84 children with SCAP were enrolled for bronchoscopic lavage. BALF was used as the sample source to evaluate the value of mNGS in the diagnosis of SCAP in children by comparing it with conventional technology (CT).

Methods

Patient Inclusions

In this retrospective analysis, 84 children diagnosed with SCAP were enrolled in Tianjin Children's Hospital from January 2019 to December 2021. After admission, a fiberoptic bronchoscopy was done within 48 h. bronchoalveolar lavage fluid (BALF) was taken for mNGS detection and CT was performed simultaneously.

Inclusion criteria were as follows: (I) all patients met the SCAP diagnostic criteria; (II) pediatric patients under 14 years of age were selected; (III) BALF obtained by bronchoscopy in patients who met the bronchoscopy indications. The following conditions will be excluded from the study: (I) mNGS was not performed; (II) incomplete clinical data; (III) loss to follow up. The study was approved by The Tianjin Children’s Hospital Ethics Committee (No. L2020-02) and was in line with the Declaration of Helsinki (as revised in 2013) (Fig. 1).

Fig. 1
figure 1

Flowcharts of enrollment and diagnosis

Full size image

SCAP definition (Includes one of the following): (I) oxygen saturation < 92%, cyanosis; (II) respiratory rate > 50 breaths/min, Infant respiratory rate > 70 breaths/min; (III) significant tachycardia for level of fever (values to define tachycardia vary with age and with temperature); (IV) prolonged central capillary refill time > 2 s; (V) difficulty in breathing; (VI) intermittent apnoea, grunting; (VII) signs of dehydration; (VIII) Infant feeding difficulties [21].

Indications for fiberoptic bronchoscopy: (I) repeated or persistent wheezing; (II) limited wheezing; (III) recurrent respiratory tract infections; (IV) hemoptysis; (V) Difficulty in withdrawing from the ventilator; (VI) Abnormal chest imaging: ① atelectasis, ② emphysema, ③ mediastinal emphysema, ④ pulmonary mass lesion, ⑤ pleural cavity lesion requiring differential diagnosis; (VII) Etiological diagnosis and treatment of pulmonary infectious diseases [22].

The study was approved by the ethics committee of Tianjin Children's Hospital, and patient data were analyzed anonymously.

Conventional technology (CT)

For hospitalized children with SCAP, blood and sputum were collected within 24 h of admission, and alveolar lavage fluid was collected within 48 h of admission. The sample source for the CT technique is mostly the BALF because it is challenging to obtain sputum from children and blood infection is quite uncommon in this study. Bacteria was detected by culture method and commercial PCR-based kits for the detection eight common respiratory bacteria (Streptococcus pneumoniae, Staphylococcus aureus, methicillin resistant Staphylococcus aureus, Klebsiella pneumoniae, Pseudomonas aeruginosa,Acinetobacter baumannii, Stenostomonas maltophilia, and Haemophilus influenzae) (Capitalbio Jinxin Bio- Technology Co., Ltd., Chengdu, China). The virus was detected by PCR (Epstein–Barr virus (EBV), cytomegalovirus (CMV), influenza A/B (Da’an Gene Technology Co., 96 Ltd., Guangzhou, China); and adenoviruses (Puruikang Biotech Co., Ltd., Shenzhen, China)). Fungi (Candida albicans, Pneumocystis vulgaris, Cryptococcus neoformans) was detected by culture method. Mycobacterium tuberculosis was detected by culture, PCR and MTB-Xpert. PCR was used to detect chlamydia and Legionella. Mycoplasma pneumoniae was detected by particle agglutination (PA) (serum MP-Ab ≥ 1:160 or MP-Ab increased or decreased by 4 times or more in recovery and acute stages), immune colloidal gold technigue (GICT) (single MP-IgM positive) [23] and PCR.

mNGS testing

Sample processing and nucleic acid extraction

The BALF sample, which is viscous and contains particles in suspension, is liquefied. Utilized a DTT reagent with a concentration of 80 mM. According to the viscosity of the sample, 1 ~ 3 times of the sample volume of DTT was added, and the sample was shaken and left at room temperature for five minutes. Non-viscous samples were directly enriched by centrifugation and the samples were stored at -20 °C. DNA was extracted from all samples using a QIAamp® UCP Pathogen DNA Kit (Qiagen) following the manufacturer’s instructions. Human DNA was removed using Benzonase (Qiagen) and Tween20 (Sigma). Total RNA was extracted with a QIAamp® Viral RNA Kit (Qiagen) and ribosomal RNA was removed by a Ribo-Zero rRNA Removal Kit (Illumina). cDNA was generated using reverse transcriptase and dNTPs (Thermo Fisher).

Library construction and sequencing

Libraries were constructed for the DNA and cDNA samples using a Nextera XT DNA Library Prep Kit (Illumina, San Diego, CA). DNA library construction includes fragmentation, end repair, ligation and purification of connectors, PCR amplification and purification, library concentration, and fragment quality monitoring. For effective libraries, the Qubit concentration was at least 1 ng/L, and the primary peak of the library fragment was between 280 and 320 bp. Library was quality assessed by Qubit dsDNA HS Assay kit followed by High Sensitivity DNA kit (Agilent) on an Agilent 2100 Bioanalyzer. Library pools were then loaded onto an Illumina Nextseq CN500 sequencer for 75 cycles of single-end sequencing to generate approximately 20 million reads for each library. For negative controls, we also prepared PBMC samples with 105 cells/mL from healthy donors in parallel with each batch, using the same protocol, and sterile deionized water was extracted alongside the specimens to serve as non-template controls (NTC).

Bioinformatics analyses

Trimmomatic was used to remove low quality reads, adapter contamination, and duplicate reads, as well as those shorter than 50 bp. Low complexity reads were removed by Kcomplexity with default parameters. Human sequence data were identified and excluded by mapping to a human reference genome (hg38) using Burrows-Wheeler Aligner software. We designed a set of criteria similar to the National Center for Biotechnology Information (NCBI) criteria for selecting representative assembly for microorganisms (bacteria, viruses, fungi, protozoa, and other multicellular eukaryotic pathogens) from the NCBI Nucleotide and Genome databases. The final database consisted of about 13,000 genomes. Microbial reads were aligned to database with SNAP v1.0beta.18. The database performance (a certain number of clinical validation samples) has a specificity of 96.2% and a sensitivity of 100%.

Interpretation of mNGS results

Currently, there is a lack of unified interpretation standards for mNGS. This study formulated the following standards based on previous literature [15, 24,25,26].

1) When the relative abundance of an organism at the species level is greater than 30% and there is supporting literature evidence of lung pathogenicity, regardless of the results of the CT method.

2) mNGS detected the same pathogenic bacteria as CT and more than 50 single microbe reads.

3) Mycobacterium tuberculosis and Legionella pneumophila are considered positive when at least one read is mapped to the species or genus level.

4) Oral and respiratory custom bacteria are not considered pathogenic bacteria, regardless of their relative abundance, unless otherwise proved or they are considered pathogenic bacteria.

CT and NGS alone could not determine whether the microorganism was infected. Finally, two senior attending physicians determined the pathogen based on clinical characteristics and reached a consensus.

Statistical analysis

Data were processed using SPSS 26.0. Continuous variables were expressed as mean ± standard deviation (SD) and assessed by independent group t tests. Categorical variables were expressed as percentage (%) and assessed by Chi-squared tests or Fisher’s exact test. A two-sided α less than 0.05 were considered as statistically significant.

Results

General characteristics of selected children

Among the 84 patients, 41 were boys, and 43 were girls, with age ranging from 2 months to 14 years and an average of 4.4 years. There were 18 (21.4%) children with basic diseases, 6 (7.1%) had congenital heart disease; 4 (4.8%) had bronchial asthma; and primary ciliary dyskinesia, hepatoblastic carcinoma, renal malformation, giant gyri malformation, diabetes, chronic granuloma, pelvic tumour, and congenital cleft palate was each reported in 1 child. The most common symptoms were cough (92.9%), fever (73.8%), and dyspnoea (35.7%), followed by poor mental health (27.4%), stridor (26.2%), and chest pain (11.9%), and the last was haemoptysis (2.4%). A total of 65 cases (77.4%) reported moist rales, and 26 (31%) had diminished respiratory sounds. In addition, 54 (62.3%) children exhibited severe lung imaging, including 36 with pleurisy (42.9%), 28 with inflammatory consolidation (33.3%), 24 with (28.6%), 20 with pleural effusion (23.8%), 20 with atelectasis (23.8%), 11 with necrosis (13.1%), and 7 with interstitial abnormalities (8.3%). The white blood cell (WBC) count (12.57 ± 6.46 × 109/L) was calculated, as was the neutrophil ratio (77.06 ± 11.96%), the lymphocyte ratio (13.98 ± 6.30%), the C-reactive protein (CRP) (77.13 ± 59.58 mg/L), and the procalcitonin (PCT) (11.03 ± 17.57 ng/mL). Before admission, 72 (85.7%) patients had been treated with antibiotics and 20 (23.8%) with glucocorticoids (Table 1).

Table 1 Clinical information of children
Full size table

The pathogen detection of SCAP

The most common pathogen detected using mNGS was Streptococcus pneumoniae (25 strains), human herpesvirus type 5 (12 strains), Haemophilus influenzae (11 strains), and Mycoplasma pneumoniae (MP) (11 strains). Pathogens, such as oral Streptococcus, Bifidobacterium brevis, Bacteroides fragilis, and Candida proximalis, were also detected. Combined with clinical manifestations, the bacteria were considered as respiratory colonising bacteria, not pathogenic ones (Fig. 2).

Fig. 2
figure 2

All microorganisms detected by Metagenomic Next-Generation Sequencing

Full size image

The most common pathogens detected using CT were MP (20 strains), S. pneumoniae (13 strains), and Staphylococcus aureus (10 strains).

Finally, 76 cases (90.5%) were diagnosed with determinate pathogen infection, and 18 (23.7%) were bacterial infections. Six patients (7.9%) were infected with a virus, and there were 10 (13.2%) cases of atypical bacterial infection. In addition, there were three (3.9%) cases of Mycobacterium tuberculosis (MTB) infection and 39 (51.3%) of mixed infection (Fig. 3A).

Fig. 3
figure 3

A Metagenomic Next-generation sequencing combined Conventional technology detected the pathogen classification. B Metagenomic Next-generation sequencing combined Conventional technology diagnose the type of polymicrobial infection

Full size image

The detection rate of mNGS and CT

The sensitivity, specificity, positive predictive value and negative predictive value of mNGS were 92.1% (70/76), 62.5% (5/8), 95.9% (70/73) and 45.5% (5/11), respectively. mNGS detection rate was 83.3% (70/84), which was higher than that of the CT at 63.1% (53/84) (P = 0.003).

The bacterial detection rate of mNGS was higher than that of the CT (80.8% [42/52] vs. 44.2% [23/52], P < 0.001) (Fig. 4A). The detection rate of Gram-negative bacteria was significantly higher in mNGS than that in the CT (90.1% [20/22] vs. 22.7% [5/22], P < 0.001), while there was no significant difference in Gram-positive bacteria (75% [30/40] vs. 57.5% [23/40], P = 0.098). S. pneumoniae was the most detected bacteria in 29 cases, in which the detection rate of mNGS was higher than that of the CT (89.2% [25/29] vs. 44.8% (13/29), P = 0.001). A total of nine cases were detected using mNGS and CT, sixteen were only positive for mNGS, and four were only positive for CT. The second most detected bacteria were H. influenzae (12 cases), in which the detection rate of mNGS was higher than that of the CT (91.7% [11/12] vs. 33.3% [4/12], P < 0.005). mNGS alone detected eight cases, CT alone detected one, and common method detected three. There were 12 cases of S. aureus, three were detected using mNGS and CT, two were only positive for mNGS, and seven were positive for CT. Among the three children with Enterococcus faecalis infection, two were detected using CT and one using mNGS. E. faecium and Micrococcus gamboge were identified by culture but not by mNGS in patient 59. There were three cases of Klebsiella pneumoniae, one in patient 11, one in patient 44, detected using mNGS, and one in patient 64, detected using CT. There were two cases of Pseudomonas aeruginosa, one of which was detected using both mNGS and CT and the other using mNGS alone. Additionally, mNGS detected Prevotella melanogenes in patient 26, Enterobacter cloacae in patients 43 and 44, Escherichia coli in patient 50, and Acinetobacter baumannii in patient 65, all of which were not detected on CT.

Fig. 4
figure 4

Comparison of pathogens detected by Metagenomic Next-generation sequencing (mNGS) and Conventional technology (CT). A Bateria detected by mNGS and CT methods. B Virus detected by mNGS and CT methods. C Fungus detected by mNGS and CT methods. D Atypical pathogen detected by mNGS and CT methods. ADV, adenovirus; CMV, human herpes virus; EBV, Epstein-Barr virus,; RSV, Respiratory syncytial virus

Full size image

For virus detection, the detection rate of mNGS was higher than that of the CT, but there was no statistical significance (76% [19/25] vs. 56% [14/25], P = 0.136) (Fig. 4B). Respiratory syncytial virus (RSV) and Epstein-Barr virus (EBV) were the most frequently detected viruses, with seven cases each. mNGS alone identified five cases of RSV, while CT alone identified one case; however in patient 79, RSV was detected using both mNGS and CT. EBV was detected by CT in six cases (patients 9,10, 13, 53, 63, and 65), and by mNGS in four cases (patients 13, 36, 63, and 65). Boca virus (HBoV) was detected in four cases, all of which were detected by mNGS. In three cases of CMV infection, both mNGS and CT were positive. Two cases of adenovirus (ADV) and one case of influenza A virus were detected using CT. However, mNGS detected bocavirus in patients 39, 42, 68, and 76, rhinovirus in patient 58, and human metapulmonary virus in patient 66.

The detection rate of fungi using mNGS was higher than that of the CT (81.8% [9/11] vs. 18.2% [2/11], P = 0.004) (Fig. 4C). Pneumocystis vulgaris (four cases) was the most frequently detected fungus, followed by C. albicans (three cases). There was one case of Cryptococcus neoformans, Aspergillus aspergillus, and Aspergillus niger. Only two instances of C. albicans were discovered using CT in patients 62 and 83. However, mNGS missed only patient 83. In patient 26, mNGS and CT were both negative, but the serum G test (1, 3-D glucan test) was positive (301 pg/mL), which matched the clinical symptoms and imaging diagnosis of fungal infection.

There was no significant difference in the detection rate of MTB between mNGS and CT (60% [3/5] vs. 80% [4/5], P > 0.005). A total of two cases were detected using mNGS and CT, one case was detected using mNGS alone, two cases were detected using CT, three were detected using MTB-Xpert in CT, and one by culture method (Table 2).

Table 2 The detection of M. tuberculosis by mNGS and CT
Full size table

Among atypical bacterial infections, mNGS detection rate was lower than that of the CT (50.0% [12/24] vs. 91.7% [22/24], P = 0.003) (Fig. 4D). CT found one case of Legionella pneumophila and one of Chlamydia trachomatis alone. The detection rate of MP using mNGS was lower than that of the CT (52.4% [11/21] vs. 95.2% [20/21], P = 0.004). In addition, 10 cases were positive for both mNGS and CT, one for mNGS alone, and 10 for CT alone. CT included MP-polymerase chain reaction (PCR) detection in 57.1% (12/21), PA detection in 52.4% (11/21), and GICT detection in 14.3% (3/21) (Table 3).

Table 3 The detection of MP by mNGS and CT
Full size table

The mixed infection of SCAP

There were 39 (51.3%) cases of mixed infection. CT alone detected 16 (21.1%) cases of mixed infection, and mNGS alone detected 20 (26.3%) cases of mixed infection. However, this difference was not statistically significant (P > 0.005).

Mixed bacterial and viral infections were the most common type, with 12 (30.8%) cases. Eight (20.5%) patients had mixed bacterial and fungal infections and five (12.8%) cases of mixed bacterial and atypical pathogens (Fig. 3B).

Diagnosis consistency between mNGS and CT

A total of 49 (58.3%) cases were positive for both mNGS and CT, 22 (26.2%) for only mNGS, 5 (6.0%) for only CT, and 8 (9.5%) were negative for both CT and mNGS.

Among 49 double-positive cases, mNGS and CT results were completely consistent in 16 (19.0%) patients and completely inconsistent in 14 (16.7%) cases. Nineteen (22.6%) cases were partially consistent, indicating that at least one detected pathogen overlapped between CT and mNGS (Fig. 5).

Fig. 5
figure 5

Consistency between Metagenomic Next-generation sequencing and Conventional technology diagnosis

Full size image

Changes after the detection of mNGS

There were 26 (31.0%) patients who changed their treatment based on mNGS results, and their symptoms were reduced. Antibiotics were changed in nine patients, antibiotics were added in five patients, antifungal drugs in nine, antivirals in seven, and antituberculosis drugs in one (Table 4).

Table 4 Changes in treatment of children with mNGS
Full size table

Discussion

Our study comprehensively evaluated the application of mNGS in children with SCAP. In this study, pathogen detection rate of mNGS was higher than that of the CT (83.3% vs. 63.1%, P = 0.003), especially for S. pneumoniae, H. influenzae, and fungi. There was no significant difference between mNGS and CT in the detection of tuberculosis and virus. However, MP detection rate was lower for mNGS than that for CT (52.4% [11/21] vs. 95.2% [20/21], P = 0.004). In addition, mNGS can detect pathogens that cannot be detected using CT, which is of great significance for the adjustment of antimicrobial therapy.

BALF was used as the sample source in this study because it was collected from infected sites and was less contaminated by oral colonisation bacteria [27, 28]. Regarding bacterial detection, our study found that mNGS had a higher detection rate for S. pneumoniae than that of the CT (89.2% vs. 44.8%, P = 0.001). In a study on children with CAP by Farnaes et al., 6.7% (1/15) of S. pneumoniae were detected by culture and cell-free plasma next-generation sequencing (CFPNGS); pneumococcus was only detected using CFPNGS in the other eight cases [29]. There was another study on invasive pneumococcal disease in 96 children; positive rate of culture method was 27.1% (n = 26), while that of mNGS test was 62.5% (n = 60) (blood, cerebrospinal fluid, and pleural effusion samples) [30]. These findings suggest that when considering S. pneumoniae infection in children with SCAP, mNGS should be actively performed to clarify the aetiology and guide anti-infection treatment. In this study, the detection of S. aureus using mNGS was lower than that of traditional methods (41.7% vs. 83.3%). However, Ren et al.'s study collected noted that mNGS (blood, cerebrospinal fluid, and BALF) was more sensitive than that of the culture method (77.8% vs. 44.4%) in adult patients with sepsis [24]. We utilised both culture and PCR, which may have enhanced the rate of S. aureus identification by conventional methods. This may also be associated with the different immune systems of children and adults.

For Gram-negative bacteria, our study found that the mNGS detection rate was significantly higher than that of the CT, especially for H. influenzae (91.7% vs. 33.3%, P < 0.005). mNGS also had a greater detection rate for P. aeruginosa, K. pneumoniae, and E. coli than that of the CT in our study. The study by Wu et al. on adult SCAP also revealed that the rate of H. influenzae identification using mNGS was higher than that of the conventional detection techniques (100% vs. 33.3%) in BALF [31]. In a study on adult sepsis, Ren et al. discovered that mNGS had a greater detection, n rate of P. aeruginosa and K. pneumoniae than that of the culture method (75.4% vs. 43.5%, 81.6% vs. 26.3%, respectively), using blood, cerebrospinal fluid, and BALF samples [24]. According to the aforementioned study, mNGS should be actively performed whenever Gram-negative bacterial infection is considered. However, a previous study revealed that after paediatric haematopoietic stem cell transplantation, the detection rate of P. aeruginosa and K. pneumoniae using conventional methods is comparable to that of the mNGS using BALF [32]. These differences may be due to the different specimen sources and immune functions.

Regarding virus detection, mNGS showed no difference compared with that of the CT in the detection of ADV, EBV, and influenza A viruses in this study. When children with SCAP caused by the above pathogens are clinically suspected, traditional methods should be considered first for pathogen detection. A previous study showed that mNGS has a low sensitivity for influenza A viruses [33]. Real-time fluorescence quantitative PCR is more sensitive than mNGS for the detection of influenza virus infection in adult SCAP [34]. In another study, mNGS (RNA-SEQ) did not detect any of the eight positive ADV samples for paediatric pneumonia but had high sensitivity to rhinovirus and RSV (100% and 96%, respectively) [35]. The potential of mNGS to identify additional viruses and subtypes that cannot be detected by conventional methods has also been confirmed in previous studies [33, 36]. In contrast, mNGS assisted in identifying cytomegalovirus (CMV) infection and not merely colonisation. In our study, mNGS and CT eventually detected three children with CMV infection, including one patient with congenital heart disease. Based on the findings of the mNGS, two cases were treated with antiviral medications, and the patients were cured. In a study, one child with leukaemia received antiviral treatment after CMV was detected using mNGS, and their symptoms improved [37]. This demonstrates that CMV should be appraised not only based on colonisation but also on the child’s underlying illness and clinical characteristics.

The fungal detection rate of mNGS was significantly higher than that of CT alone. In a study on the detection of fungal pneumonia in adults, the culture method was only used in 1/21 samples, while mNGS was used in 19/21 samples [38]. In a study of children after haematopoietic stem cell transplantation, mNGS (BALF) detected nine cases of fungal pneumonia (Pneumocystis jiroi [n = 6], Aspergillus [n = 2], and mucor [n = 1]), while none was detected using traditional methods [32]. In this study, the positive rate of fungal pneumonia detected using mNGS was 81.8% (9/11), while only 18.2% (2/11) used the traditional pathogen detection method. Among them, 54.5% (6/11) had underlying diseases, including congenital heart disease, primary ciliary dyskinesia, chronic granuloma, pelvic mass, diabetes, and congenital cleft palate. Two kinds of Aspergillus, Aspergillus aspergillus and Aspergillus niger, were also detected using mNGS, but not by traditional methods. Huang et al. also reported that mNGS detected various A. fumigatus, A. niger, A. flavus, and A. oryzae in immunosuppressed adult patients diagnosed with fungal pneumonia [39]. Therefore, mNGS should be actively applied to children with severe pneumonia and underlying diseases or immunodeficiency. In our study, three cases of C. albicans infection were confirmed, and the detection rates of mNGS and CT were similar, among which two cases had underlying diseases. Liu et al. also reported that C. albicans was one of the fungi most frequently detected in BALF cultures [40]. However, another study found that the mNGS detection rate of C. albicans was lower than that of the traditional method (70.6% vs. 82.4%) in ventilators-associated pneumonia in adults [25]. Therefore, traditional methods should be used for pathogen detection in children with C. albicans infection. Interestingly, patient 26 tested negative on mNGS and CT but positive on the 1,3-D glucan test, which combined clinical symptoms and imaging to diagnose fungal infection. Therefore, culture, mNGS, galactomannan (GM) test, and 1, 3-D glucan test should be combined in the diagnosis of fungal pneumonia to avoid missing positive samples.

The detection sensitivity of mNGS for MTB was similar to that of CT in our study. Previous studies have also confirmed that the diagnostic ability of mNGS for MTB is similar to that of the traditional detection methods [41, 42]. Therefore, when clinically considering the presence of MTB infection, traditional methods rather than mNGS should be considered first. MTB, a type of intracellular bacteria, releases less nucleic acid into the extracellular environment, and insufficient wall fracture can also affect the extraction of DNA; therefore, fewer sequences can be detected [43, 44]. mNGS combined with MTB-Xpert can be used to improve diagnostic rates [11, 43].

Surprisingly, we found that mNGS was inferior to CT in the detection of atypical pathogens (50.0% vs. 91.7%), especially for MP (52.4% vs. 95.2%). This has never been mentioned in previous studies. Only studies on children with severe unresponsive pneumonia indicated that mycoplasma detection rates of mNGS and PCR were equal (15.6% vs. 18.8%) [37]. In this study, the detection rate of MP-PCR was 57.1% (12/21), which was similar to that by mNGS. The possible reasons are the combination of PCR and two immunological methods, which increased the sensitivity of the traditional method in this study. Therefore, when considering MP infections in children with SCAP, it is not necessary to conduct mNGS examinations.

The patient received empirical anti-infective medication in the absence of knowledge of its precise aetiology. After identifying the illness' pathogenic microorganisms, 26 people (31.0%) modified their course of treatment according to the mNGS results, and their symptoms improved. In a study of adults with severe respiratory diseases, mNGS led to a change in treatment in 34.4% (11/32) of patients [13]. Combining mNGS with conventional detection technologies can improve the ability to target drug use. On the other hand, the culture process takes at least 3 days [45], while mNGS can obtain test results within 24–48 h [36]. Although PCR is quick, its reach is constrained and may not even detect the causal agent [33]. These results suggest that mNGS is important for the detection and treatment of SCAP in children.

Our study has several limitations. First, the sample size of our study was modest; hence, the number of pathogens identified was limited. Second, because of the high sensitivity of mNGS, pathogens of colonisation and contamination will also be discovered; therefore, there is no universal standard for the report's issuance and interpretation. Finally, because this was a single-centre observational study, a case selection bias may have occurred. Further prospective, multicentre, randomised controlled studies are needed to confirm the sensitivity, specificity, and clinical value of mNGS.

Conclusion

In conclusion, mNGS has its unique advantages in the detection of pathogenic agents of SCAP in children, especially Streptococcus pneumoniae, Haemophilus influenzae and fungi. In addition, mNGS can detect pathogens that cannot be detected by CT, which is of great significance for adjusting antimicrobial therapy.