ABSTRACT
Metagenomic next-generation sequencing (mNGS) has shown promise in the diagnosis of infectious diseases in adults, while its efficacy in pediatric infections remains uncertain. We performed a retrospective analysis of 1,493 mNGS samples from pediatric patients with blood, central nervous system, and lower respiratory tract infections. The positive percent agreement (PPA) and the negative percent agreement (NPA) of mNGS were compared to conventional microbiological tests (CMT) based on clinical diagnosis. The agreement of mNGS compared to CMT, as well as the clinical impact of mNGS, were valuated. Using the clinical diagnosis as a reference, mNGS demonstrated a significantly higher overall PPA compared to CMT (53.1% [95% CI = 49.7 to 56.6%] versus 25.8% [95% CI = 22.8 to 28.9%]), while maintaining a comparable overall NPA (93.2% [95% CI = 91.3 to 95.1%] versus 97.2% [95% CI = 95.9 to 98.4%]). In septic patients under 6 years of age or with immunosuppressive status, mNGS showed a higher PPA and a comparable NPA compared to CMT. The overall PPA and NPA of mNGS compared to CMT were 75.3 and 75.0%, respectively. The majority of cases of Streptococcus pneumoniae, Streptococcus agalactiae, Mycobacterium tuberculosis complex, and Pneumocystis jirovecii infections were identified by mNGS. A positive clinical impact of 14.0% (206/1,473), a negative impact of 0.8% (11/1,473), a nonimpact of 84.7% (1,248/1,473), and an unknown impact of 0.5% (8/1,473) were observed in the mNGS results. Notably, the positive impact was greater among immunosuppressed patients than among nonimmunosuppressed individuals (67/247, 27.1% versus 139/1,226, 11.3%; P < 0.001). mNGS is valuable for pathogen detection, diagnosis, and clinical management of infections among pediatric patients. mNGS was thus effective for the diagnosis of pediatric infections, which may guide clinical management. Patients with immunosuppressive conditions benefited more from mNGS testing.
KEYWORDS: diagnosis, clinical application, pediatric, infection, mNGS
INTRODUCTION
Infectious diseases are a major cause of morbidity and mortality in children worldwide (1, 2). Early identification of pathogen is crucial for the diagnosis and treatment of infections, such as sepsis, central nervous system (CNS) infection, and pneumonia, in children (3–8). Conventional microbiological tests (CMT), including culture, smear, serology tests, and PCR, are effective for pathogen detection in routine clinical practice. However, the limited sensitivity and narrow detection range of CMT make it challenging to identify all pathogens (9, 10).
Metagenomic next-generation sequencing (mNGS) has emerged as a promising approach for the detection of common, rare, and emerging microorganisms (9, 11, 12), providing advantages in infection diagnosis, pathogen detection, and clinical treatment guidance (12–16). Despite its potential benefits, mNGS is not yet widely used in routine clinical practice.
Previous studies have primarily focused on the performance of mNGS in adults, either alone or in combination with pediatric populations. However, comprehensive assessments of diagnostic performance and the clinical impacts of mNGS among large pediatric cohorts are limited. We therefore conducted an assessment of diagnostic performance and clinical impact of mNGS among children admitted to a tertiary children’s hospital in China to improve the application of mNGS in this population.
MATERIALS AND METHODS
Study design and data collection.
We retrospectively collected 1,493 samples from patients with suspected or diagnosed infections who were admitted to the Children’s Hospital of Fudan University in Shanghai, China, between 1 September 2019 and 31 May 2021. Samples were collected if they were (i) from patients with suspected or diagnosed sepsis, CNS infection, or lower respiratory tract infection (LRTI); (ii) from patients aged 0 to 18 years; (iii) from blood, bronchoalveolar lavage fluid (BALF), cerebrospinal fluid (CSF), or sputum; and (iv) tested by mNGS and culture with or without other CMT within 3 days using the same sample type. Samples from patients without sufficient information in the electronic medical records were excluded. The diagnostic performance of mNGS and CMT, the contribution of mNGS to the microorganism identification and clinical management were assessed. Informed consent was obtained from the parents of each patient.
For each sample, clinical information, mNGS and CMT results were independently reviewed from electronic medical records by two clinicians (Y.Z. and M.G.). The indication for mNGS was decided upon at the clinician’s discretion for diagnosis of infections, for monitoring infections, or to rule out infections in patients at high risk. DNA microorganisms detected by mNGS or CMT were analyzed.
Metagenomic next-generation sequencing.
Different DNA extraction methods were used for the four sample types. DNA samples were then processed with KAPA Hyper Prep Kits (Roche, Basel, Switzerland) for library construction using the manufacturer’s instructions. DNA sequencing was performed on an Illumina NextSeq 550 platform using a 75-cycle single-end sequencing kit. Negative and nontemplate controls were processed from DNA extraction to sequencing in parallel for each batch of samples.
Raw reads were deduplicated, trimmed and quality filtered. Trimmed reads were then aligned to the human reference genome to remove human reads. Microorganism taxonomic classification for the remaining reads was performed using Centrifuge (v1.0.3). For each microorganism, the number of reads was compared to the number in the negative control to exclude possible contaminants (see supplemental methods S1 and Table S1 in the supplemental material).
Conventional microbiological tests.
Culture, PCR, serology tests, 1,3-β-d-glucan test, or galactomannan test were ordered based on the clinician’s judgment (see Table S2).
Definition of infectious samples.
The final diagnoses of sepsis (3, 17), CNS infection (18–21), and LRTI (6, 7) were made by supervising clinicians based on existing guidelines or diagnostic standards (see supplemental methods S2). Samples were classified as infectious (blood samples from patients with sepsis, CSF samples from patients with CNS infection, or BALF/sputum samples from patients with LRTI), noninfectious (blood samples from patients without sepsis, CSF samples from patients without CNS infection, or BALF/sputum samples from patients without LRTI), or unknown based on the sample type and the final clinical diagnoses (see Fig. S1A in the supplemental material).
Clinical adjudication of mNGS results.
mNGS results were adjudicated at microorganism and sample levels. First, microorganism(s) detection by mNGS was judged as true positive (TP), false positive (FP), true negative (TN), or false negative (FN) based on whether identical microorganism(s) were found in CMT after excluding clinically common contaminating microorganisms, whether patients improved after antimicrobial treatment, and other likely explanation for infections (see Fig. S1B). Second, samples were adjudicated as TP, FP, TN, and FN based on the adjudication of microorganism(s). Among infectious samples, those with TP results for one or more microorganism(s) by mNGS were judged as TP samples, while they were judged as FN samples if no TP results were detected. Among noninfectious samples, those with FP results identified by mNGS were judged as FP samples and as TN samples if no microorganisms were detected (see Fig. S1C).
Evaluation of the clinical impact of mNGS.
The clinical impact of mNGS was assessed as a positive impact (new diagnosis or early microbiological diagnosis by the mNGS result, medication added [antibiotics, antivirals, antifungal, and/or supportive therapy], and/or medication changed [escalated, de-escalated, and/or discontinued]), negative impact (unnecessary additional tests or therapy), nonimpact (no changes on diagnosis or treatment), and unknown (undetermined clinical impact) according to the mNGS results.
Statistical analysis.
We performed the statistical analysis with Stata 15 (Stata Corp LLC). Continuous variables with skewed distributions are reported as medians and interquartile ranges. Categorical variables are expressed as frequencies and percentages. Pearson’s chi-square test or Fisher exact test was performed for comparative analysis. The positive percent agreement (PPA), the negative percent agreement (NPA), and the positive and negative predictive values of mNGS compared to CMT were calculated. All analyses were two tailed. Results with a type I error of 0.05 were considered statistically significant.
Ethical statement.
The study was approved by the Ethics Committee of the Children’s Hospital of Fudan University (2019-300).
RESULTS
Sample characteristics.
Of 1,493 samples collected, 1,473 had a confirmed infectious status (798 infectious samples and 675 noninfectious samples), while 20 had unknown status (Fig. 1). In all, 1,015 of 1,473 samples (68.9%) were from infants, and 247 of 1,473 samples (16.8%) were from immunosuppressed patients. Of the 1,473 samples, 1,017 (69.0%) were ordered for diagnosis, 352 (23.9%) for monitoring, and 104 (7.1%) to rule out possible infections (Table 1; see also Table S3); 1,185 (80.4%) of samples were analyzed using culture and at least one other CMT (667 [45.3%] with one other CMT, 383 [26.0%] with two other CMT, and 135 [9.2%] with three other CMT), and 288 (19.6%) using culture only.
FIG 1.
Flowchart of the characteristics of 1,493 samples. Of 1,493 samples collected from blood, CSF, BALF and sputum, 798 were infectious samples, 675 were noninfectious samples, and 20 had unknown status. BALF, bronchoalveolar lavage fluid; CSF, cerebrospinal fluid.
TABLE 1.
Characteristics of 1,473 samples from pediatric patients with suspected or diagnosed infections
| Characteristicsa | No. (%) |
||
|---|---|---|---|
| Total samples (n = 1,473) | Infectious samples (n = 798) | Noninfectious samples (n = 675) | |
| Age | |||
| Median age (mo)b | 2.0 (0.6–24.0) | 3.0 (0.9–24.0) | 2.0 (0.3–24.0) |
| Distribution | |||
| 0 to 28 days | 495 (33.6) | 203 (25.4) | 292 (43.3) |
| 29 days to 11 mos | 520 (35.3) | 326 (40.9) | 194 (28.7) |
| 1 to 5 yrs | 233 (15.8) | 153 (19.2) | 80 (11.8) |
| 6 to 18 yrs | 225 (15.3) | 116 (14.5) | 109 (16.2) |
| Sex | |||
| Male | 918 (62.3) | 495 (62.0) | 423 (62.7) |
| Female | 555 (37.7) | 303 (38.0) | 252 (37.3) |
| Immune condition | |||
| Immunosuppressed | 247 (16.8) | 170 (21.3) | 77 (11.4) |
| Immunodeficiency | 157 (63.6) | 109 (64.1) | 48 (62.3) |
| Transplant | 36 (14.6) | 31 (18.2) | 5 (6.5) |
| Long-term steroids | 23 (9.3) | 10 (5.9) | 13 (16.9) |
| Chemotherapy | 21 (8.5) | 13 (7.7) | 8 (10.4) |
| Hematological malignancy | 9 (3.6) | 6 (3.5) | 3 (3.9) |
| Long-term steroids plus chemotherapy | 1 (0.4) | 1 (0.6) | 0 |
| Nonimmunosuppressed | 1,226 (83.2) | 628 (78.7) | 598 (88.6) |
| Sample type | |||
| Blood | 615 (41.8) | 356 (44.6) | 259 (32.5) |
| CSF | 576 (39.1) | 173 (21.7) | 403 (50.5) |
| BALF | 196 (13.3) | 188 (23.6) | 8 (1.0) |
| Sputum | 86 (5.8) | 81 (10.1) | 5 (0.6) |
| Indication for testing | |||
| Diagnosis | 1,017 (69.0) | 446 (55.9) | 571 (84.6) |
| Monitoring | 352 (23.9) | 341 (42.7) | 11 (1.6) |
| Rule-out | 104 (7.1) | 11 (1.4) | 93 (13.8) |
| CMT | |||
| Positive | 250 (17.0) | 232 (29.1) | 18 (2.7) |
| Negative | 1223 (83.0) | 566 (71.0) | 657 (97.3) |
| mNGS | |||
| Positive | 494 (33.5) | 448 (56.1) | 46 (6.8) |
| Negative | 979 (66.5) | 350 (43.9) | 629 (93.2) |
BALF, bronchoalveolar lavage fluid; CMT, conventional microbiological tests; CSF, cerebrospinal fluid; mNGS, metagenomic next-generation sequencing.
The interquartile range is indicated in parentheses.
Diagnostic performance of mNGS and CMT.
In comparison with the final clinical diagnoses, the overall PPA of mNGS in 1,473 samples was significantly higher than that of CMT (53.1% [95% CI, 49.7 to 56.6%] versus 25.8% [95% CI, 22.8 to 28.9%]; P < 0.001), while the overall NPA of mNGS was slightly lower than that of CMT (93.2% [95% CI, 91.3 to 95.1%] versus 97.2% [95% CI, 95.9 to 98.4%]; P = 0.001), where 95% CI represents the 95% confidence interval. Of different sample types, the PPA of mNGS and CMT was higher in BALF and sputum samples than that in CSF and blood samples, but showing the highest difference of 37.0% observed in CSF samples. However, the NPA of mNGS showed slightly lower or no differences in different sample types (Table 2; see also Table S4, and Fig. S2).
TABLE 2.
Diagnostic performance of mNGS and CMT in analysis of 1,473 samples
| Sample type | Positive % agreement (95% CI)a |
Negative % agreement (95% CI)a |
||||
|---|---|---|---|---|---|---|
| mNGS | CMT | P | mNGS | CMT | P | |
| Blood | 44.9 (39.8–50.1) | 23.0 (18.7–27.4) | <0.001 | 92.3 (89.0–95.5) | 95.8 (93.3–98.2) | 0.095 |
| CSF | 44.5 (37.1–51.9) | 7.5 (3.6–11.4) | <0.001 | 94.0 (91.7–96.4) | 98.5 (97.3–99.7) | 0.001 |
| BALF | 67.6 (60.9–74.2) | 41.5 (29.8–44.9) | <0.001 | 87.5 (64.6–110.4) | 87.5 (64.6–110.4) | >0.99 |
| Sputum | 74.1 (64.5–83.6) | 40.7 (30.0–51.4) | <0.001 | 80.0 (44.9–115.1) | 100 | >0.99 |
| Total | 53.1 (49.7–56.6) | 25.8 (22.8–28.9) | <0.001 | 93.2 (91.3–95.1) | 97.2 (95.9–98.2) | 0.001 |
Positive and negative percent agreements of mNGS and CMT were compared to the final clinical diagnosis. BALF, bronchoalveolar lavage fluid; CMT, conventional microbiological tests; CSF, cerebrospinal fluid; mNGS, metagenomic next-generation sequencing.
In the analysis of mNGS performance in blood samples by age, mNGS had a higher PPA than CMT in samples from all subgroups, while the NPA showed no statistical differences among samples from patients under 6 years of age (see Table S5). Of samples from patients with different immune status, mNGS also showed a higher PPA and a similar NPA in blood samples from immunosuppressed patients compared to those from individuals without immunosuppression. No similar tendency was found for CMT in blood samples (see Table S6).
Agreement of mNGS compared to CMT.
Of the 1,473 samples, the positive rate of mNGS was higher than that of CMT for all-microorganism identification (33.5% [494/1,473] versus 17.0% [251/1,473], P < 0.001). mNGS and CMT results were both positive in 189 samples (12.8%) and both negative in 917 (62.3%). Compared to CMT, the PPA and NPA values of mNGS were 75.3% (95% CI, 70.0 to 80.6%) and 75.0% (95% CI, 72.6 to 77.5%) and the positive and negative predictive value of mNGS were 38.3% (95% CI, 34.0 to 42.5%) and 93.7% (95% CI, 92.1 to 95.2%), respectively. Of the 189 positive samples, 151 (79.9%) were in accordance (80 [42.3%] totally matched, and 71 [37.6%] partially matched), while 38 (20.1%) were discordant (Fig. 2A). Of the 305 samples (20.7%) with positivity in mNGS only, 240 (90.6%) were adjudicated as TP, 21 as FN, and 44 as FP. Among the 62 samples with positivity in CMT only (4.2%), 32 were identified with TP microorganisms in infectious samples, while the remaining 30 samples showed inconsistent results (FN) regarding the microorganisms from 14 infectious samples and FP microorganisms from 16 noninfectious samples. The PPA for BALF and sputum results was higher than that for blood and CSF results (90.1 and 91.7% versus 63.5 and 52.6%), while the NPA for blood and CSF results was conversely higher than that for BALF and sputum results (76.4 and 82.8% versus 47.0 and 40.0%) (Fig. 2B to E).
FIG 2.
Agreement of mNGS compared to CMT in 1,473 samples. (A) Accordance of microorganisms detected via mNGS or CMT. (B to E) Positive and negative agreement percentages for mNGS and CMT in the analysis of blood, CSF, BALF, and sputum samples. BALF, bronchoalveolar lavage fluid; CMT, conventional microbiological tests; CSF, cerebrospinal fluid; mNGS, metagenomic next-generation sequencing; NPA, positive percent agreement; NPV, negative predictive value; PPA, positive percent agreement; PPV, positive predictive value.
Extra microorganism detection by mNGS.
mNGS yielded additional pathogen detection that was interpreted by clinical adjudicator to be significant in different sample types, with 105 samples performed qPCR validation (Supplemental Results). In the 424 of 459 TP samples identified with pathogens by mNGS with or without CMT results, approximately half of BALF and sputum samples had multiple microorganisms, while most blood and CSF samples had a single microorganism identified (see Table S7 and Fig. S3). Of the 459 TP samples, 151 (32.9%) were adjudicated as TP due to identical microorganism(s) detected by both mNGS and CMT, 235 (51.2%) achieved improvement after treatment, and 5 (1.1%) had no other likely explanation but consistent with clinical condition (see Table S8). Of pathogens identified in the 459 samples, the majority of Klebsiella pneumoniae (12/14, 85.7%), Escherichia coli (10/11, 90.9%), Mycobacterium tuberculosis complex (10/10, 100%), Enterococcus faecium (8/10, 80.0%), Streptococcus pneumoniae (5/5, 100%), and Pneumocystis jirovecii (3/3, 100%) infections in blood samples, all Streptococcus agalactiae (15/15, 100%), most human betaherpesvirus 5 (8/10, 80.0%), E. coli (8/12, 66.7%), and E. faecium (4/5, 80.0%) infections in CSF samples, and the majority of S. pneumoniae (18/20, 90.0%), human betaherpesvirus 5 (11/12, 91.7%), P. jirovecii (10/10, 100%), and Stenotrophomonas maltophilia (7/14, 50.0%) infections in BALF samples, as well as S. pneumoniae (12/13, 92.3%), human betaherpesvirus 5 (10/10, 100%), and Staphylococcus aureus (3/4, 75.0%) infections in sputum samples, were detected by mNGS only. In all, approximately two-thirds of these pathogens were detected by mNGS only, with the highest percentage of 84.9% (73/86) in CSF samples (Fig. 3; see also Table S9 and Table S10).
FIG 3.
Pathogens detected by mNGS and/or CMT in 459 true positive samples by sample type. (A) Bacteria, fungi, and viruses identified by mNGS, CMT, or both. Pathogens are listed in order of the contribution of mNGS in blood samples. Blue bars indicate microorganisms detected by only mNGS, red bars by only CMT, and green bars by both mNGS and CMT. (B) Percentages of pathogens identified by mNGS, CMT or both. The percentages of pathogens detected by only mNGS (blue), only CMT (red), and both methods (green) in all pathogens. BALF, bronchoalveolar lavage fluid; CMT, conventional microbiological tests; CSF, cerebrospinal fluid; mNGS, metagenomic next-generation sequencing; MTBC, Mycobacterium tuberculosis complex.
Clinical impact of mNGS.
Of the 1,473 cases, mNGS had a positive impact in the analysis of 206 cases (14.0%). mNGS resulted in a new diagnosis in 160 cases (10.9%), and new medications were added in 104 cases (7.1%), including half that were antibiotics. In addition, 51 (3.5 %) cases involved medication changes, including 15 with escalated antimicrobial coverage, 12 with de-escalated coverage, and 24 with discontinued antimicrobials because of negative mNGS results. However, 11 cases (0.8%) had a negative impact. 10 cases were ordered additional tests due to mNGS results and confirmed to be negative. Another patient was prescribed antibiotics but was later considered as colonization. The remaining 1,248 cases (84.7%) had no impact on testing or clinical treatment, and 8 cases were associated with an unknown impact.
The analyses of samples from immunosuppressed patients were associated with more positive impacts than those from patients without immunosuppression (67/247, 27.1% versus 139/1,226, 11.3%; P < 0.001), while there was no statistical significance in the negative impact among these two populations (3/247, 1.2% versus 8/1,226, 0.7%; P = 0.28) (Table 3). Samples from septic and LRTI patients achieved more positive clinical impacts compared to those with CNS infection (81/282 [28.7%] versus 41/576 [7.1%], P < 0.001; 84/615 [13.7%] versus 41/576 [7.1%], P < 0.001) (see Table S11).
TABLE 3.
Clinical impact based on mNGS results in 1,473 samples
| Clinical impacta | No. (%) of samples |
||
|---|---|---|---|
| Immunosuppression (n = 247) | Nonimmunosuppression (n = 1,226) | Total (n = 1,473) | |
| Positive impact | 67 (27.1) | 139 (11.3) | 206 (14.0) |
| New diagnosis | 56 (22.7) | 104 (8.5) | 160 (10.9) |
| Medication added | 42 (17.0) | 62 (5.1) | 104 (7.1) |
| Antibiotics | 19 (7.7) | 32 (2.6) | 51 (3.5) |
| Antivirals | 19 (7.7) | 22 (1.8) | 41 (2.8) |
| Antifungal | 0 | 0 | 0 |
| Supportive therapy | 4 (1.6) | 7 (0.6) | 11 (0.8) |
| Medication changed | 14 (5.7) | 37 (3.0) | 51 (3.5) |
| Escalated | 5 (2.0) | 10 (0.8) | 15 (1.0) |
| De-escalated | 4 (1.6) | 8 (0.6) | 12 (0.8) |
| Discontinued | 5 (2.0) | 19 (1.6) | 24 (1.6) |
| Negative impact | 3 (1.2) | 8 (0.7) | 11 (0.8) |
| Unnecessary testing | 3 (1.2) | 7 (0.6) | 10 (0.7) |
| Unnecessary therapy | 0 | 1 (0.1) | 1 (0.1) |
| No impact | 175 (70.9) | 1,073 (87.5) | 1,248 (84.7) |
| Unknown | 2 (0.8) | 6 (0.5) | 8 (0.5) |
Samples with more than one impact were calculated separately.
DISCUSSION
In this study, we comprehensively assessed the diagnostic performance and clinical impact of mNGS analysis of 1,493 samples from blood, CSF, BALF, and sputum among pediatric patients aged from newborn to 18 years. Our findings demonstrated mNGS was effective for pathogen detection and antimicrobial treatment for better clinical management.
Culture is a commonly used reference for the assessment of the diagnostic performance of other tests. However, its low sensitivity makes it inappropriate in real clinical practice. Recently, more studies have chosen to use the results of comprehensive judgment processes, such as clinical diagnosis as reference, which better reflect the infection status (5, 13, 16, 22–24). Considering clinical diagnoses as the reference, previous studies revealed that the PPA of mNGS ranged from 50.7 to 92.9% (16, 19, 22, 24), with an NPA of 59.0 to 98.59% (16, 19, 20, 22, 24) in the evaluation of various infections. Our PPA of mNGS (53.1%) was similar to the study by Miao et al. (22) but lower than that observed by Tao et al. (16), while our NPA of 93.2% was higher than that observed by Miao et al. (22) and Tao et al. (16) but lower than that observed in by Zhang et al. (20). This discrepancy may be attributed to differences in patient populations, since previous studies included both adult and pediatric patients with approximately half having LRTI, whereas we focused on a younger population with sepsis or CNS infections.
To comprehensively assess the diagnostic performance of mNGS, we analyzed the PPA and NPA of mNGS by sample type, age, and immune status. In analyses of blood, CSF, BALF, and sputum samples, mNGS had a higher PPA than CMT, while the NPA showed no significant differences. Extra microorganisms identified by mNGS need to be confirmed by further clinical judgment by clinicians. Microorganisms were difficult to identify by CMT in CSF samples. However, mNGS increased the PPA the most in CSF samples. Although mNGS showed a statistically lower NPA than CMT in the analysis of CSF samples, the NPA of 94.0% was considered clinically high. Therefore, mNGS generated a significantly higher PPA but a comparable NPA in the analysis of our CSF samples.
Pathogen identification using CMT in young children with sepsis is challenging. mNGS had a significantly higher PPA than CMT in the analysis of blood samples among all age subgroups, while no differences were observed in the NPA among those under 6 years of age. mNGS therefore presented high value in this population. In addition, the PPA of mNGS in septic patients with immunosuppression was 2-fold higher than that in nonimmunosuppressed patients because mNGS identified opportunistic pathogens in immunosuppressed individuals. However, no similar differences were observed in CNS or LRTI samples between these two populations. mNGS was beneficial for pathogen detection in immunosuppressed patients with sepsis.
mNGS was advantageous for pathogen identification that were missed by CMT, accounting for at least two-thirds of all pathogens in this study. S. pneumoniae is a common fastidious bacterium that is difficult to detect via culture (25). In our study, S. pneumoniae was mostly detected via mNGS in all sample types. S. agalactiae was prevalent in newborns and infants and was difficult to detect via CMT. All cases of S. agalactiae infection detected in CSF and BALF samples and half in blood samples were detected via mNGS only. The Mycobacterium tuberculosis complex and opportunistic pathogen Pneumocystis jirovecii were all detected via mNGS in immunosuppressed patients because they were undetectable using current CMT. However, some pathogens were detected via CMT that were missed by mNGS. We considered the reasons to be that the identified microorganisms were discordant with clinical diagnosis or the infection was diagnosed using different sample types, the microorganisms were present in low levels, the bacteria were intracellular, which makes obtaining circulatory genomic DNA difficult, or fungi with hard-to-break cell walls were present; these explanations are consistent with previous studies (5, 20, 22). In addition, we included only CMT results that had corresponding mNGS results obtained within 3 days, and these pathogens missed by mNGS may have been identified in an unmatched sample by mNGS. Therefore, we might have underestimated the pathogen detection ability of mNGS.
mNGS guides clinical management via the identification of more pathogens, accounting for 3.4% to 14% of changes in diagnosis and/or treatment strategies (5, 12, 13). However, previous studies reported the clinical impacts in limited adult populations. In our study, mNGS showed a positive impact of 14.0%, which is similar to the findings of Lee et al. (13) but higher than those of Hogan et al. (12) and Wilson et al. (5). We also found that mNGS had ~2-fold higher levels of positive impacts in immunosuppressed patients by enabling the identification of more opportunistic pathogens, as well as in septic and LRTI patients, which presented promising advantages in these populations. However, mNGS indeed generated a 0.8% negative impact, which confused clinical judgment and resulted in extra costs for unnecessary testing or medications. The interpretation of findings of such microorganisms is still challenging for clinicians.
Currently, empirical antimicrobial treatment is widely used in clinical practice, especially in the treatment of children. Patients treated with antimicrobials before pathogen testing therefore received fewer changes in medications when mNGS was performed, which led to lower clinical impacts than clinicians’ expectations in previous studies and our study. Furthermore, one-quarter of our patients with confirmed infections underwent mNGS for monitoring; this may have led to fewer diagnosis or medication changes but the continuation of the ongoing antimicrobial treatment regimen. Our turnaround time for mNGS was within 2 days. With the shortening of the turnaround time, mNGS may contribute more to reducing unnecessary empirical antimicrobial use among infectious and noninfectious disease patients.
This study had several limitations. Patients underwent mNGS analysis based on clinicians’ decisions, which may have led to selection bias. We did not detect RNA viruses, which may have led to an underestimation of the performance of mNGS in pathogen identification. In addition, clinical adjudication of mNGS with improvement after treatment were classified regardless of the probability of self-healing of infections with inappropriate therapy or without treatment. Multicenter studies are recommended to assess the value of mNGS in different pediatric populations.
In conclusion, mNGS yielded higher PPA results and comparable NPA results compared to CMT, particularly in the analysis of patients with sepsis under 6 years of age and those with immunosuppression. The implementation of mNGS-guided diagnosis and treatment may result in better clinical management. Furthermore, immunosuppressed patients benefited more from mNGS testing than nonimmunosuppressed patients.
ACKNOWLEDGMENTS
This study was supported by the National Key Research and Development Program of China (2022YFC2703603 to B.W.), the National Key Research and Development Program of China (2021YFC2701800 and 2021YFC2701805 to G.L.), the Science and Technology Commission of Shanghai Municipality (22Y11902700 to B.W.), the China Primary Health Care Foundation (MTP2022A014 to B.W.), and the Shanghai Hospital Development Center (SHDC22020217 to M.G.).
There are no conflicts of interest for any of the authors.
Footnotes
Supplemental material is available online only.
Contributor Information
Bingbing Wu, Email: 081107271@fudan.edu.cn.
Erin McElvania, NorthShore University HealthSystem.
REFERENCES
- 1.GBD 2016 Disease and Injury Incidence and Prevalence Collaborators. 2017. Global, regional, and national incidence, prevalence, and years lived with disability for 328 diseases and injuries for 195 countries, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 390:1211–1259. doi: 10.1016/S0140-6736(17)32154-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Fleischmann-Struzek C, Goldfarb DM, Schlattmann P, Schlapbach LJ, Reinhart K, Kissoon N. 2018. The global burden of paediatric and neonatal sepsis: a systematic review. Lancet Respir Med 6:223–230. doi: 10.1016/S2213-2600(18)30063-8. [DOI] [PubMed] [Google Scholar]
- 3.Shane AL, Sánchez PJ, Stoll BJ. 2017. Neonatal sepsis. Lancet 390:1770–1780. doi: 10.1016/S0140-6736(17)31002-4. [DOI] [PubMed] [Google Scholar]
- 4.Weiss SL, Fitzgerald JC, Pappachan J, Wheeler D, Jaramillo-Bustamante JC, Salloo A, Singhi SC, Erickson S, Roy JA, Bush JL, Nadkarni VM, Thomas NJ, Sepsis Prevalence, Outcomes, and Therapies (SPROUT) Study Investigators and Pediatric Acute Lung Injury and Sepsis Investigators (PALISI) Network . 2015. Global epidemiology of pediatric severe sepsis: the sepsis prevalence, outcomes, and therapies study. Am J Respir Crit Care Med 191:1147–1157. doi: 10.1164/rccm.201412-2323OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wilson MR, Sample HA, Zorn KC, Arevalo S, Yu G, Neuhaus J, Federman S, Stryke D, Briggs B, Langelier C, Berger A, Douglas V, Josephson SA, Chow FC, Fulton BD, DeRisi JL, Gelfand JM, Naccache SN, Bender J, Dien Bard J, Murkey J, Carlson M, Vespa PM, Vijayan T, Allyn PR, Campeau S, Humphries RM, Klausner JD, Ganzon CD, Memar F, Ocampo NA, Zimmermann LL, Cohen SH, Polage CR, DeBiasi RL, Haller B, Dallas R, Maron G, Hayden R, Messacar K, Dominguez SR, Miller S, Chiu CY. 2019. Clinical metagenomic sequencing for diagnosis of meningitis and encephalitis. N Engl J Med 380:2327–2340. doi: 10.1056/NEJMoa1803396. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Haq IJ, Battersby AC, Eastham K, McKean M. 2017. Community-acquired pneumonia in children. BMJ 356:j686. doi: 10.1136/bmj.j686. [DOI] [PubMed] [Google Scholar]
- 7.Korang SK, Nava C, Mohana SP, Nygaard U, Jakobsen JC. 2021. Antibiotics for hospital-acquired pneumonia in neonates and children. Cochrane Database Syst Rev 11:Cd013864. doi: 10.1002/14651858.CD013864.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Gu W, Deng X, Lee M, Sucu YD, Arevalo S, Stryke D, Federman S, Gopez A, Reyes K, Zorn K, Sample H, Yu G, Ishpuniani G, Briggs B, Chow ED, Berger A, Wilson MR, Wang C, Hsu E, Miller S, DeRisi JL, Chiu CY. 2021. Rapid pathogen detection by metagenomic next-generation sequencing of infected body fluids. Nat Med 27:115–124. doi: 10.1038/s41591-020-1105-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Chiu CY, Miller SA. 2019. Clinical metagenomics. Nat Rev Genet 20:341–355. doi: 10.1038/s41576-019-0113-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.d’Humières C, Salmona M, Dellière S, Leo S, Rodriguez C, Angebault C, Alanio A, Fourati S, Lazarevic V, Woerther P-L, Schrenzel J, Ruppé E. 2021. The potential role of clinical metagenomics in infectious diseases: therapeutic perspectives. Drugs 81:1453–1466. doi: 10.1007/s40265-021-01572-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Gu W, Miller S, Chiu CY. 2019. Clinical metagenomic next-generation sequencing for pathogen detection. Annu Rev Pathol 14:319–338. doi: 10.1146/annurev-pathmechdis-012418-012751. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hogan CA, Yang S, Garner OB, Green DA, Gomez CA, Dien Bard J, Pinsky BA, Banaei N. 2021. Clinical impact of metagenomic next-generation sequencing of plasma cell-free DNA for the diagnosis of infectious diseases: a multicenter retrospective cohort study. Clin Infect Dis 72:239–245. doi: 10.1093/cid/ciaa035. [DOI] [PubMed] [Google Scholar]
- 13.Lee RA, Al Dhaheri F, Pollock NR, Sharma TS. 2020. Assessment of the clinical utility of plasma metagenomic next-generation sequencing in a pediatric hospital population. J Clin Microbiol 58:e00419-20. doi: 10.1128/JCM.00419-20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Niles DT, Revell PA, Ruderfer D, Marquez L, McNeil JC, Palazzi DL. 2022. Clinical impact of plasma metagenomic next-generation sequencing in a large pediatric cohort. Pediatr Infect Dis J 41:166–171. doi: 10.1097/INF.0000000000003395. [DOI] [PubMed] [Google Scholar]
- 15.Niles DT, Wijetunge DSS, Palazzi DL, Singh IR, Revell PA. 2020. Plasma metagenomic next-generation sequencing assay for identifying pathogens: a retrospective review of test utilization in a large children’s hospital. J Clin Microbiol 58:e00794-20. doi: 10.1128/JCM.00794-20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Tao Y, Yan H, Liu Y, Zhang F, Luo L, Zhou Y, An K, Yang R, Yang B, Xu T, Xie L, Ren H, Xu Z, Cao Q, Mo X. 2022. Diagnostic performance of metagenomic next-generation sequencing in pediatric patients: a retrospective study in a large children’s medical center. Clin Chem 68:1031–1041. doi: 10.1093/clinchem/hvac067. [DOI] [PubMed] [Google Scholar]
- 17.Goldstein B, Giroir B, Randolph A, International Consensus Conference on Pediatric Sepsis . 2005. International pediatric sepsis consensus conference: definitions for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med 6:2–8. doi: 10.1097/01.PCC.0000149131.72248.E6. [DOI] [PubMed] [Google Scholar]
- 18.O’Brien MP, Francis JR, Marr IM, Baird RW. 2018. Impact of cerebrospinal fluid multiplex assay on diagnosis and outcomes of central nervous system infections in children: a before and after cohort study. Pediatr Infect Dis J 37:868–871. doi: 10.1097/INF.0000000000001936. [DOI] [PubMed] [Google Scholar]
- 19.Blauwkamp TA, Thair S, Rosen MJ, Blair L, Lindner MS, Vilfan ID, Kawli T, Christians FC, Venkatasubrahmanyam S, Wall GD, Cheung A, Rogers ZN, Meshulam-Simon G, Huijse L, Balakrishnan S, Quinn JV, Hollemon D, Hong DK, Vaughn ML, Kertesz M, Bercovici S, Wilber JC, Yang S. 2019. Analytical and clinical validation of a microbial cell-free DNA sequencing test for infectious disease. Nat Microbiol 4:663–674. doi: 10.1038/s41564-018-0349-6. [DOI] [PubMed] [Google Scholar]
- 20.Zhang Y, Cui P, Zhang HC, Wu HL, Ye MZ, Zhu YM, Ai JW, Zhang WH. 2020. Clinical application and evaluation of metagenomic next-generation sequencing in suspected adult central nervous system infection. J Transl Med 18:199. doi: 10.1186/s12967-020-02360-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Ge M, Gan M, Yan K, Xiao F, Yang L, Wu B, Xiao M, Ba Y, Zhang R, Wang J, Cheng G, Wang L, Cao Y, Zhou W, Hu L. 2021. Combining metagenomic sequencing with whole exome sequencing to optimize clinical strategies in neonates with a suspected central nervous system infection. Front Cell Infect Microbiol 11:671109. doi: 10.3389/fcimb.2021.671109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Miao Q, Ma Y, Wang Q, Pan J, Zhang Y, Jin W, Yao Y, Su Y, Huang Y, Wang M, Li B, Li H, Zhou C, Li C, Ye M, Xu X, Li Y, Hu B. 2018. Microbiological diagnostic performance of metagenomic next-generation sequencing when applied to clinical practice. Clin Infect Dis 67:S231–s240. doi: 10.1093/cid/ciy693. [DOI] [PubMed] [Google Scholar]
- 23.Zhang XX, Guo LY, Liu LL, Shen A, Feng WY, Huang WH, Hu HL, Hu B, Guo X, Chen TM, Chen HY, Jiang YQ, Liu G. 2019. The diagnostic value of metagenomic next-generation sequencing for identifying Streptococcus pneumoniae in paediatric bacterial meningitis. BMC Infect Dis 19:495. doi: 10.1186/s12879-019-4132-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Chen H, Yin Y, Gao H, Guo Y, Dong Z, Wang X, Zhang Y, Yang S, Peng Q, Liu Y, Wang H. 2020. Clinical utility of in-house metagenomic next-generation sequencing for the diagnosis of lower respiratory tract infections and analysis of the host immune response. Clin Infect Dis 71:S416–S426. doi: 10.1093/cid/ciaa1516. [DOI] [PubMed] [Google Scholar]
- 25.Koeth LK, DiFranco-Fisher JM, Hardy DJ, Palavecino EL, Carretto E, Windau A. 2022. Multilaboratory comparison of omadacycline MIC test strip to broth microdilution MIC against Gram-negative, Gram-positive, and fastidious bacteria. J Clin Microbiol 60:e01410-21. doi: 10.1128/JCM.01410-21. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental text, Tables S1 to S11, and Fig. S1 to S3. Download jcm.00115-23-s0001.pdf, PDF file, 2.7 MB (2.7MB, pdf)