Abstract
Introduction
Early targeted antibiotic therapy is critical for improving outcomes in ICU patients with pneumonia unresponsive to initial treatment. Metagenomic next-generation sequencing (mNGS) is a unique diagnostic tool; however, its effectiveness for high-risk populations remains unclear.
Methods
This retrospective study included 642 ICU patients with pneumonia unresponsive to initial treatment, categorized into single- or repeat-test groups based on mNGS testing frequency. We analyzed the results of mNGS and conventional microbiological tests (CMTs), compared microbial detection characteristics between patients with different immune statuses, assessed the impact of repeat testing on microbial detection and treatment adjustments and evaluated its association with patient prognosis using multivariable logistic regression and propensity score matching.
Results
Among 642 patients, patient-level agreement between mNGS and CMTs was low (κ = 0.180, p < 0.001). First bronchoalveolar lavage fluid (BALF)-mNGS results showed a higher microbial detection rate in immunocompromised than immunocompetent patients. Among patients undergoing repeat BALF-mNGS testing, partial concordance between first and second tests was most common pattern. Antibiotic treatments were modified in 67.6% of cases based on mNGS results, with a higher adjustment rate in the repeat-test group. Although ICU mortality was higher in the repeat-test group, multivariate logistic regression analysis revealed no significant association between repeat testing and mortality risk (adjusted odds ratio = 1.15, 95% CI: 0.64–2.06, p = 0.630). After propensity score matching, no significant difference was observed between two groups (absolute risk difference: 1.12%, 95% CI: −7.43% to 9.67%, p = 0.798).
Conclusion
mNGS is a valuable microbial detection tool for ICU patients with pneumonia unresponsive to initial treatment and can support early antimicrobial adjustment. Repeat testing can provide information on dynamic changes in the microbial spectrum during disease but was not associated with improved patient outcomes, suggesting that repeat testing frequency should be carefully considered to avoid unnecessary testing. mNGS results should be interpreted in conjunction with CMTs, host immune status and inflammatory biomarkers to optimise its clinical value in ICU pneumonia.
Keywords: Metagenomic next-generation sequencing, initial treatment-refractory pneumonia, etiological diagnosis, bronchoalveolar lavage fluid, antimicrobial stewardship, repeat mNGS testing
1. Introduction
Pneumonia is the leading cause of adult admission to intensive care unit (ICU) worldwide [1]. However, ICU patients with pneumonia – particularly older adults – respond poorly to initial empirical antibiotic therapy due to underlying comorbidities and a high risk of infection with drug-resistant pathogens [2–3]. For these individuals, timely evaluation is essential to identify the causative pathogens. Typically, the initial use of empirical antibiotics follows the principle of ‘early administration and broad coverage’ to rapidly control infection and improve patient outcomes. However, the overuse of antibiotics is associated with inherent risks, including microbiota dysbiosis, increased resistance and secondary infections [4–5]. Therefore, rapidly identifying pathogenic microorganisms to enable an early transition from broad-spectrum antibiotic therapy to targeted treatment remains a considerable challenge [6].
Conventional microbiological tests (CMTs) – including smear microscopy, isolation culture and polymerase chain reaction (PCR) assays – are often limited by low pathogen detection rates, stringent culture requirements and long turnaround times [7–8]. The microbial detection rate of CMTs is approximately 50% [9], and in China, the causative pathogens remain unidentified in nearly half of patients with pulmonary infections [10]. This prevents clinicians from identifying the causative pathogens early and adjusting treatment in a timely manner, leading to disease progression and an increased risk of mortality.
Metagenomic next-generation sequencing (mNGS) is an innovative pathogen-detection technology widely used in clinical settings. It boasts rapid results, high sensitivity and broad pathogen coverage, making it particularly suitable for diagnosing infections by rare, emerging, or difficult-to-detect organisms as well as infections in immunocompromised individuals [11–12]. Compared with CMTs, mNGS can provide more comprehensive microbial detection information and support early adjustment of antimicrobial therapy [13–14]. However, because mNGS detects microbial nucleic acids, it cannot independently distinguish colonisation from active infection, and its clinical utility requires further evaluation.
To date, research on the clinical value of mNGS for ICU patients with pneumonia who are unresponsive to initial treatment is limited. Accordingly, we conducted a retrospective analysis to systematically evaluate the diagnostic and therapeutic impact of mNGS in these patients.
2. Materials and methods
2.1. Ethical approval and consent
The retrospective studies involving humans were approved by the Research Ethics Committee of Renmin Hospital of Wuhan University (Number WDRM2025-K198). All procedures involving human participants were performed in accordance with the Declaration of Helsinki.
2.2. Study design and participants
Clinical data from 2,155 patients diagnosed with pneumonia and admitted to the ICU of Renmin Hospital of Wuhan University between March 2022 and December 2024 were retrospectively collected.
The inclusion criteria were as follows: (1) patient age ≥18 years; (2) fulfilment of criteria for pneumonia unresponsive to initial treatment (defined below) and (3) completion of simultaneous mNGS and CMTs (sputum or blood culture).
The exclusion criteria were as follows: (1) mNGS and CMT specimens not collected and processed within the same time period and (2) specimens deemed unqualified or associated with incomplete clinical data.
Based on the inclusion and exclusion criteria, 642 patients with an initial diagnosis of pneumonia were selected for enrolment. During the study period, 999 mNGS specimens, comprising 889 bronchoalveolar lavage fluid (BALF) specimens, 86 blood specimens, 14 cerebrospinal fluid (CSF) specimens, 5 ascites specimens, 4 lung puncture tissue specimens and 1 pleural fluid specimen, were collected. In addition, 961 sputum culture and 520 blood culture samples were obtained.
Patients were divided into the single-test and repeat-test groups according to the number of mNGS tests performed. Repeat mNGS testing was not conducted according to a predefined study protocol but was determined by the treating physicians based on their clinical judgment of the patients’ clinical manifestations and response to treatment.
According to the definition of immunocompromised status [15–16], patients were classified as immunocompetent or immunocompromised. Immunocompromised status was defined by the presence of any of the following: (1) hematologic malignancy (active or in remission for <5 years); (2) solid organ transplantation or hematopoietic stem cell transplantation; (3) chemotherapy for a solid tumour within the previous 3 months or neutropenia; (4) use of immunosuppressive agents, biologic immunomodulators or antirheumatic drugs (e.g. methotrexate, cyclophosphamide or cyclosporine); (5) corticosteroid therapy at a dose of ≥20 mg/day of prednisone (or an equivalent dose of another corticosteroid) for ≥14 days or a cumulative prednisone dose of ≥700 mg (or equivalent); (6) congenital or inherited immunodeficiency or (7) HIV infection or a history of splenectomy or thymectomy (e.g. thymectomy for thymoma associated with myasthenia gravis).
2.3. Conventional microbiological tests
Pathogenic microorganisms were identified through CMTs, including blood and sputum cultures.
2.4. mNGS procedure
DNA was extracted using the Sansure DNA Extraction Kit (Sansure Biotech Inc., Hunan, China) according to the manufacturer’s instructions. DNA libraries were constructed following standard procedures and subjected to high-throughput DNA sequencing on the Nanopore MinION platform (Oxford Nanopore Technologies, UK). Sequencing data were processed using a laboratory-established bioinformatics pipeline, including removal of low-quality and duplicate reads, filtering of human-derived sequences by alignment to GRCh38 and microbial sequence alignment against the NCBI GenBank database. The reference databases included 16S rRNA gene and internal transcribed spacer (ITS) gene databases downloaded from NCBI, covering reference sequences from 19,088 bacterial species, 8,082 fungal species and 231 non-bacterial pathogens. An mNGS result was considered positive when ≥20 sequence reads were detected. This DNA-based mNGS workflow did not cover RNA viruses, and results were typically available within 2–3 days after sample receipt.
The mNGS reports provided semiquantitative information on detected microorganisms, including the number of specific sequence reads, genome coverage and relative abundance. All results were subjected to quality assessment according to laboratory quality-control procedures, including evaluation of nucleic acid quality and potential contamination.
2.5. Assessment of the clinical relevance of microbial detection
Because mNGS detects microbial nucleic acids and cannot independently distinguish colonisation from active infection, detected microorganisms were not considered causative pathogens solely based on a positive sequencing result. The clinical relevance of detected microorganisms was independently assessed by two senior physicians based on sequencing characteristics, sample type, clinical manifestations, imaging findings, inflammatory biomarkers, conventional microbiological results and clinical response to antimicrobial therapy. Any disagreements were resolved through discussion until consensus was reached.
2.6. Criteria for pneumonia unresponsive to initial treatment
Pneumonia was considered unresponsive if, after at least 72 h of empirical antibiotic therapy [17], all following criteria were met, with at least one condition satisfied per criterion [18–20]:
No improvement or worsening of clinical signs: persistent fever (>38.0 °C), no improvement or worsening of respiratory symptoms, hemodynamic instability.
No significant improvement or progression of laboratory indicators/respiratory status: C-reactive protein (CRP) or procalcitonin (PCT) levels unchanged or increased compared with baseline; declining respiratory function (PaO2/FiO2 ratio ≤ 250 mmHg or requiring increased oxygen concentration/mechanical ventilation).
No radiological improvement or progression: follow-up chest imaging revealing enlargement of pulmonary infiltrates or new lobar infiltrates, cavities, pleural effusion, or other complications.
2.7. Definitions of antibiotic de-escalation and escalation
The definitions of antibiotic de-escalation and escalation employed in this study were consistent with those established in previous research [21–22], which categorizes antibiotics into four hierarchical ranks according to antibacterial spectrum and stewardship requirements: narrow-spectrum (rank 1), broad-spectrum (rank 2), extended-spectrum (rank 3) and protected (rank 4). Antibiotic de-escalation was defined as discontinuing one or more antibiotics from combination therapy or switching from broad-spectrum to narrow-spectrum agents. Conversely, antibiotic escalation was defined as the addition of one or more antibiotics to the existing regimen or advancement to higher-ranked antimicrobial agents within the same regimen. An unchanged regimen was characterized by the absence of modifications to either the number or rank of antibiotics or by adjustments made in opposite directions that counterbalanced each other.
2.8. Evaluation of the clinical impact of mNGS on patient prognosis
Assessment of treatment response was conducted based on a previous study [2]. At least two experienced clinicians independently evaluated each case 72–96 h after the reporting of mNGS results, adjusting treatment regimens according to a comprehensive evaluation of the following indicators; any disagreements between assessors were resolved through discussion to reach a consensus:
Effective treatment (all three criteria met simultaneously): (i) resolution of clinical symptoms; (ii) reduction in infection biomarkers (CRP or PCT); (iii) stabilization or improvement in respiratory function (increase in PaO2/FiO2 ratio compared to baseline, or a reduction in respiratory support requirements).
Ineffective treatment (any one of the following): (i) progression of clinical symptoms; (ii) no reduction or an increase in infection biomarkers; (iii) requirement for increased respiratory support or evidence of radiological progression.
2.9. Statistical analysis
Data analyses were performed using SPSS 26.0 and R version 4.4.4, with graphical representations generated in GraphPad Prism 10. Categorical data are presented as frequencies and percentages; continuous variables with non-normal distributions are reported as medians with interquartile ranges (IQRs). Between-group comparisons were conducted using the Mann–Whitney U-test. For categorical variables, the chi-square test, Fisher’s exact test, or McNemar’s test was employed as appropriate, with agreement assessed using Kappa test (Kappa ≥ 0.80 indicated nearly perfect agreement; 0.60–0.80 substantial agreement; 0.40–0.60 moderate agreement; and < 0.40 poor agreement). The association between repeat mNGS testing and ICU mortality was investigated using univariate logistic regression (p < 0.10 for inclusion), followed by multivariate logistic regression incorporating group variables, screened covariates, and clinically recognized prognostic factors to control for confounders. Model goodness-of-fit was evaluated using Hosmer–Lemeshow test. To further ensure robustness, propensity score matching (1:1, caliper width 0.2) was performed, including demographic information, comorbidities, laboratory parameters, and organ support as matching variables. Post-matching balance was assessed using standardized mean differences (SMD < 0.20). Statistical significance was set at p < 0.05.
3. Results
3.1. Patient distribution and characteristics
In total, 642 patients with pneumonia who were unresponsive to initial treatment were included in this analysis; of these patients, 463 were males and 179 were females. Patients were categorized based on mNGS submission frequency into a single-test group (n = 445) and a repeat-test group (n = 197; mNGS testing ≥ 2 times). No statistically significant differences in age or sex were observed between the two groups (p > 0.05). Regarding medical history, the proportion of patients with diabetes did not differ significantly between the groups (p > 0.05); however, patients in the repeat-test group had a significantly higher incidence of hypertension and cardiovascular or cerebrovascular diseases (p < 0.05). Laboratory findings indicated that the repeat-test group demonstrated significantly elevated white blood cell count, neutrophil percentages, CRP, and PCT, as well as a notably lower lymphocyte percentage compared with the single-test group. The rate of sputum culture submission did not differ between the two groups, whereas the rate of blood culture submission was markedly higher in the repeat-test group. Moreover, the repeat-test group showed significantly greater use of mechanical ventilation and vasoactive drugs, longer hospital stays, and increased in-hospital mortality than the single-test group (Table 1).
Table 1.
General baseline characteristics of the 642 patients included in the core antibiotic decision analysis.
| Single-test group (n = 445) |
Repeat-test group (n = 197) |
p value | |
|---|---|---|---|
| Age (years), Median (IQR) | 65 (57–74) | 68 (58–76) | 0.151 |
| Male/Female, n | 316/129 | 147/50 | 0.347 |
| Comorbidity, n | |||
| Hypertension | 180 | 103 | 0.005 |
| Diabetes | 81 | 45 | 0.172 |
| Cerebro-cardiovascular disease | 101 | 81 | <0.001 |
| Clinical laboratory data, Median (IQR) | |||
| White blood cell,109/L | 9.50 (6.36–12.75) | 11.09 (7.31–14.42) | 0.009 |
| Neutrophil, % | 80.80 (72.60–87.80) | 86.60 (78.70–91.10) | <0.001 |
| Lymphocyte, % | 11.60 (6.30–18.40) | 6.90 (4.00–12.50) | <0.001 |
| C-reactive protein, mg/L | 45.78 (10.76–90.35) | 64.06 (18.18–99.56) | 0.014 |
| Procalcitonin, ng/mL | 0.16 (0.04–0.89) | 0.41 (0.11–2.20) | <0.001 |
| Sputum culture performed, n (%) | 443 (99.55%) | 196 (99.49%) | 0.921 |
| Blood culture performed, n (%) | 233 (52.36%) | 182 (92.39%) | <0.001 |
| Organ-support therapies | |||
| MV, n (%) | 215 (48.31%) | 158 (80.20%) | <0.001 |
| Vasoactive drug, n (%) | 81 (18.20%) | 97 (49.24%) | <0.001 |
| Clinical outcomes | |||
| ICU mortality, n (%) | 43 (9.66%) | 49 (24.87%) | <0.001 |
| Length of ICU stay, (day) | 19 (15–26) | 24 (17–33) | <0.001 |
MV: mechanical ventilation; ICU: intensive care unit
Bold text indicates a statistically significant difference (p < 0.05) between the two groups. Laboratory indicators are the first measured values within 24 h of admission.
3.2. Microbial detection results of mNGS and CMTs
3.2.1. Patient-level detection results of mNGS and CMTs
Among the 642 patients, kappa analysis revealed slight agreement between mNGS and CMTs for microbial detection (p < 0.001; kappa = 0.180; Table 2). Concordant positive and negative microbial detection results were observed in 52.80% and 9.03% of patients, respectively. In addition, microorganisms were detected using mNGS alone in 36.60% of patients, whereas CMTs alone detected microorganisms in 1.56% of patients.
Table 2.
Kappa analysis of patient-level agreement between mNGS and CMTs for microbial detection.
| mNGS | CMTs |
Total | |
|---|---|---|---|
| Positive (+) | Negative (−) | ||
| Positive (+) | 339 | 235 | 574 |
| Negative (−) | 10 | 58 | 68 |
| Total | 349 | 293 | 642 |
p < 0.001, Kappa value = 0.180.
3.2.2. Microbial spectra detected using mNGS and CMTs
In total, 143 microbial species, including 92 bacterial species, 50 fungal species, and 1 additional microorganism (Mycobacterium tuberculosis), were detected using mNGS and CMTs. mNGS detected 137 microbial species, whereas CMTs detected 63 species. The three most frequently detected bacterial and fungal species across different specimen types are shown in Figure 1 (complete data and findings from the remaining less frequently tested specimen types are provided in Supplementary Table S1).
Figure 1.
Summary of the top three bacterial and fungal organisms isolated from different specimen types.
3.3. BALF-mNGS detection results stratified by immune status
Because this study primarily focused on lower respiratory tract microbial characteristics in patients with pneumonia unresponsive to initial treatment, only BALF-mNGS results were included in the immune status–stratified analysis. To minimize the influence of repeat testing and subsequent treatment interventions on the detected microbial spectrum, only the first BALF-mNGS result from each patient was included in this analysis. In total, 642 patients, comprising 553 immunocompetent and 89 immunocompromised patients, were included (the specific categories and distribution of immunocompromised status are provided in Supplementary Table S2).
The first BALF-mNGS results revealed that the microbial detection rate was significantly higher in immunocompromised patients than in immunocompetent patients (94.38% vs 86.26%, p < 0.05). Among patients with positive microbial detection, mixed bacterial–fungal detection was the predominant pattern in both groups. The proportion of mixed bacterial–fungal detection was higher in immunocompromised patients than in immunocompetent patients (57.14% vs 48.64%), although the difference was not statistically significant (p > 0.05). The proportion of bacterial-only detection was significantly higher in immunocompetent patients than in immunocompromised patients (38.78% vs 26.19%, p < 0.05), whereas the proportion of fungal-only detection did not differ significantly between the two groups (16.67% vs 12.58%, p > 0.05) (Figure 2A).
Figure 2.
Microbial profiles identified using BALF-mNGS according to immune status.
(A) Microbial positivity rate and distribution of microbial detection patterns among immunocompetent and immunocompromised patients. (Right axis) black circles indicate the overall microbial detection rate and black triangles indicate the proportion of specimens without microbial detection; (Left axis) colored bars represent the distribution of different microbial detection patterns; (B) Top five bacterial and fungal organisms identified using BALF-mNGS in immunocompromised and immunocompetent groups.
The distributions of the major bacterial and fungal microorganisms detected via the first BALF-mNGS test in the two groups are shown in Figure 2B (complete data are provided in Supplementary Table S3).
3.4. Concordance analysis of repeat BALF-mNGS results
Among the 197 patients who underwent repeat mNGS testing, 155 underwent at least two BALF-mNGS tests. We primarily analyzed the concordance between the first two BALF-mNGS results. Of these 155 patients, 109 underwent two tests and 46 underwent three or more tests.
Paired analysis of the first and second BALF-mNGS results in these 155 patients showed complete concordance in 16 patients (10.32%), partial concordance in 82 (52.90%) and complete discordance in 57 (36.77%) (Figure 3A).
Figure 3.
Concordance and compositional changes between repeated BALF-mNGS tests.
(A) Overall classification of repeated BALF-mNGS test results; (B) Partial concordance group (n = 82). All patterns share at least one identical species between the two tests (partial overlap). (C) Complete discordance group (n = 57). All patterns show no species overlap between the two tests (complete discordance). (In panels B and C, arrows denote the direction of compositional shifts).
Among the 82 patients with partial concordance, the most common pattern of change in the detected microbial spectrum was a change in species composition within mixed bacterial–fungal detection, with both tests showing mixed bacterial–fungal detection and partial species overlap (n = 49). This was followed by a shift from mixed to single detection (n = 12), a shift from single to mixed detection (n = 12) and changes in bacterial species with partial species overlap (n = 9) (Figure 3B).
Among the 57 patients with complete discordance, 9 showed conversion from positive to negative detection, 11 showed newly positive detection after an initially negative result, 7 showed a complete change in species within the same detection category, 5 showed a complete change in species within mixed detection (mixed detection in both tests but with no species overlap) and 3 showed a switch between bacterial and fungal detection. A shift from single to mixed detection occurred in 12 patients, whereas a shift from mixed to single detection occurred in 10. In all these completely discordant patterns, the microorganisms detected in the first and second tests were entirely different at the species level (Figure 3C).
3.5. Clinical impact of mNGS
3.5.1. mNGS-Guided antimicrobial treatment adjustments
Among the 642 enrolled patients, antimicrobial regimens were adjusted based on mNGS results for 67.6% (434/642) of the patients, whereas no regimen adjustment was made for 32.4% (208/642) of the patients (Figure 4A). Among those with treatment adjustments (n = 434), 144 received escalated therapy, 116 underwent de-escalation and 11 experienced other alterations. Antifungal agents were administered to 163 patients, of whom 58 received them concurrently with escalation and 48 with de-escalation (Figure 4B).
Figure 4.
(A) Proportion of patients with modified antimicrobial regimens; (B) antimicrobial adjustments based on mNGS and clinical outcomes; (C) repeat-test patients: antimicrobial adjustments based on mNGS and clinical outcomes.
Among the 197 patients who underwent repeat testing, antimicrobial regimens were adjusted according to mNGS results in 87.82% (173/197) of the patients. Among these, escalated therapy was administered to 53 patients, de-escalated therapy to 58 and other alterations to 2 patients. Antifungal agents were administered to 60 patients, including 24 who received them concurrently with escalation and 25 with de-escalation (Figure 4C).
3.5.2. Impact of mNGS-Guided antimicrobial therapy adjustments on prognosis
Among the 642 patients, the ICU mortality rate among those who underwent treatment adjustments based on mNGS results was 16.82% (73/434). Evaluation of treatment responses revealed that 61.75% (268/434) of the patients responded positively to the adjusted regimens, whereas 38.25% (166/434), including 73 deceased patients, did not respond.
Among patients in the repeat-test group who underwent mNGS-guided treatment adjustments, the ICU mortality rate was 26.01% (45/173; Figure 4C). Evaluation of treatment response demonstrated that 43.93% (76/173) responded to the adjusted therapy, whereas 56.07% (97/173, including the 45 deceased patients) showed no response.
3.6. Association between repeated mNGS testing and ICU mortality
3.6.1. Logistic regression analysis
To explore the association between repeat testing and ICU mortality, a multivariate logistic regression model was constructed, adjusting for confounding variables (Table 1). Univariate analysis revealed significantly higher ICU mortality in the repeat-test group (24.87%, 49/197) than in the single-test group (9.66%, 43/445; p < 0.001; Table 3). However, after multivariate adjustment, this association was not statistically significant (aOR = 1.15, 95% CI: 0.64–2.06, p = 0.630). This indicates that after accounting for baseline disease severity and clinical characteristics, no independent association was observed between repeat testing and ICU mortality. The Hosmer–Lemeshow test for model calibration yielded a χ2 value of 6.980 with 8 degrees of freedom (p = 0.539), confirming good model fit.
Table 3.
Exploratory multivariable logistic regression analysis of factors associated with ICU mortality.
| Variables | Univariate logistic regression |
Multivariate logistic regression |
||
|---|---|---|---|---|
| OR (95% CI) | p value | aOR (95% CI) | p value | |
| Demographics | ||||
| Age (years) | 1.06 (1.04–1.08) | <0.001 | 1.04 (1.01–1.06) | 0.002 |
| Female sex (ref: Male) | 1.04(0.64–1.71) | 0.87 | 1.12(0.59–2.12) | 0.724 |
| Comorbidities | ||||
| Hypertension | 1.99 (1.27–3.12) | 0.003 | 0.87 (0.46–1.61) | 0.865 |
| Diabetes | 2.02 (1.23–3.31) | 0.005 | 0.97 (0.51–1.83) | 0.921 |
| Cerebro-cardiovascular disease | 2.89 (1.84–4.54) | <0.001 | 1.40 (0.75–2.61) | 0.295 |
| Laboratory parameters on admission | ||||
| White blood cell (109/L) | 1.04 (1.00–1.08) | 0.033 | 0.97 (0.92–1.02) | 0.175 |
| Neutrophil percentage | 1.06 (1.03–1.08) | <0.001 | 0.96 (0.91–1.00) | 0.068 |
| Lymphocyte percentage | 0.92 (0.89–0.95) | <0.001 | 0.92 (0.85–0.99) | 0.025 |
| C-reactive protein (mg/L) | 1.01 (1.00–1.01) | 0.004 | 1.00 (1.00–1.01) | 0.170 |
| Elevated procalcitonin (≥ 0.5 ng/mL) | 2.44 (1.56–3.81) | <0.001 | 0.98 (0.54–1.76) | 0.941 |
| Organ-support therapies | ||||
| Mechanical ventilation | 13.14 (5.65–30.56) | <0.001 | 1.59 (0.52–4.84) | 0.416 |
| Vasoactive drug | 30.75 (16.14–58.60) | <0.001 | 21.16(9.55–46.89) | <0.001 |
| mNGS testing strategy | ||||
| Repeat testing | 3.10 (1.97–4.86) | <0.001 | 1.15 (0.64–2.06) | 0.630 |
OR: odds ratio; aOR: adjusted odds ratio; CI: confidence interval
Bold values indicate statistically significant associations (p < 0.05). The multivariate logistic regression model included variables with p < 0.10 in univariate analysis or established clinical relevance.
3.6.2. Propensity score matching analysis
To further evaluate the association between repeat testing and ICU mortality while reducing the potential impact of baseline differences between the groups, a 1:1 propensity score matching approach (caliper width of 0.2) was employed based on 12 variables, including age, sex and disease severity. This approach yielded 179 matched pairs. Post matching, all variables achieved standardised mean differences of 0.1 (Table 4), indicating balanced baseline characteristics across groups. Within this matched cohort, the ICU mortality rate was 21.23% (38/179) for the single-test group and 22.35% (40/179) for the repeat-test group. No significant difference was observed between the two groups (absolute risk difference: 1.12%, 95% CI: −7.43% to 9.67%, p = 0.798). This further suggests that the higher crude mortality observed in the repeat-test group may be primarily attributable to greater baseline disease severity rather than repeat testing itself.
Table 4.
General baseline characteristics of the single-test group and repeat-test group before and after PSM.
| Variable | Unmatched |
p value | PSM(1:1) |
p value | SMD | ||
|---|---|---|---|---|---|---|---|
| Single-test group (n = 445) |
Repeat-test group (n = 197) |
Single-test group (n = 179) |
Repeat-test group (n = 179) |
||||
| Demographic | |||||||
| Age (years, Median(IQR)) | 65 (57–74) | 68 (58–76) | 0.151 | 67 (58–76) | 68 (58–76) | 0.903 | 0.008 |
| Gender (male, n, (%)) | 316 (71.01%) | 147 (74.62%) | 0.347 | 140 (78.21%) | 135 (75.42%) | 0.531 | 0.066 |
| Comorbidities (n, %) | |||||||
| Hypertension | 180 (40.45%) | 103 (52.28%) | 0.005 | 91 (50.84%) | 93 (51.96%) | 0.833 | 0.022 |
| Diabetes | 81 (18.20%) | 45 (22.84%) | 0.172 | 41 (22.91%) | 37 (20.67%) | 0.609 | 0.054 |
| Cerebro-cardiovascular disease | 101 (22.70%) | 81 (41.12%) | <0.001 | 65 (36.31%) | 66 (36.87%) | 0.913 | 0.012 |
| Clinical laboratory data (Median(IQR)) | |||||||
| White blood cell,109/L | 9.50 (6.36–12.75) | 11.09 (7.31–14.42) | 0.009 | 10.51 (7.38–15.09) | 10.95 (7.16–14.41) | 0.909 | 0.013 |
| Neutrophil, % | 80.80 (72.60–87.80) | 86.60 (78.70–91.10) | <0.001 | 84.00 (76.90–89.30) | 86.50 (78.30–91.10) | 0.155 | 0.064 |
| Lymphocyte, % | 11.60 (6.30–18.40) | 6.90 (4.00–12.50) | <0.001 | 8.50 (5.20–14.40) | 7.30 (3.90–13.10) | 0.093 | 0.061 |
| C-reactive protein, mg/L | 45.78 (10.76–90.35) | 64.06 (18.18–99.56) | 0.014 | 53.39 (14.14–99.54) | 58.76 (16.42–102.18) | 0.726 | 0.014 |
| Procalcitonin, ng/mL | 0.16 (0.04–0.89) | 0.41 (0.11–2.20) | <0.001 | 0.41 (0.09–2.14) | 0.41 (0.11–2.54) | 0.637 | 0.005 |
| Organ-support therapies | |||||||
| MV, n(%) | 215 (48.31%) | 158 (80.20%) | <0.001 | 136 (75.98%) | 140 (78.21%) | 0.615 | 0.053 |
| Vasoactive drug, n(%) | 81 (18.20%) | 97 (49.24%) | <0.001 | 77 (43.02%) | 79 (44.13%) | 0.831 | 0.022 |
MV: mechanical ventilation
Boldface values indicate statistically significant association (p < 0.05).
4. Discussion
Pneumonia in ICU patients remains a challenging clinical condition, and timely acquisition of reliable microbiological information is critical for its diagnosis and management. We evaluated the value of mNGS in microbiological diagnosis and antimicrobial management in this population.
Pneumonia unresponsive to initial treatment is particularly common among individuals over 65 years of age who have chronic conditions, such as diabetes, malignancies and cardiovascular or cerebrovascular diseases [20]. Poor underlying health status and prolonged exposure to the ICU environment may contribute to rapid disease progression. The patient profile in the present study is consistent with prior reports. Because these patients often have varying degrees of immune dysfunction, the detection of fungal microorganisms is also relatively common in this population [23–24]. Current guidelines recommend that, if no clinical improvement is observed after ≥5 days of empirical antimicrobial therapy, the causative microorganisms should be promptly reassessed [2].
Compared with CMT and culture methods, mNGS provides broader microbial coverage and is particularly useful for detecting rare pathogens, mixed infections, and microorganisms in samples obtained after antibiotic exposure [25–28]. In the present study, patient-level microbial detection results showed low agreement between mNGS and CMTs (κ = 0.180, p < 0.001), consistent with the findings of Liu et al. [29]. In addition, mNGS and CMTs identified 137 and 63 microbial species, respectively, suggesting that mNGS can provide more comprehensive microbial information, consistent with the findings of Yang et al. [30]. This difference may be attributable to the distinct detection principles of the two methods: mNGS primarily detects microbial nucleic acids and does not depend on microbial culture conditions. However, because mNGS alone cannot distinguish active infection from colonization or contamination, clinicians should interpret mNGS results cautiously and corroborate them with conventional microbiological findings to avoid overinterpretation [31].
A previous study [32] showed that host immune status is an important factor influencing microbial detection and mNGS results interpretation. In the present study, stratified analysis by immune status showed that immunocompromised patients had a higher microbial detection rate than immunocompetent patients (94.38% vs 86.26%, p < 0.05) and a higher proportion of mixed bacterial–fungal detection, suggesting a potentially more complex microbial detection spectrum in this population, consistent with the findings of Li et al. [23]. However, unlike the findings of Li et al. mixed bacterial–fungal detection was the predominant pattern in both immunocompromised and immunocompetent patients in our study, highlighting the complexity of microbial detection results in ICU patients with pneumonia unresponsive to initial treatment. This finding may be related to the complex underlying conditions of these patients, prolonged exposure to the ICU environment, and prior antibiotic exposure. In addition, bacterial-only detection was more frequent in immunocompetent patients, whereas fungal-only detection did not differ significantly between the two groups, suggesting that microbial detection patterns may vary according to immune status. Regarding the distribution of specific microorganisms, Acinetobacter baumannii and Pseudomonas aeruginosa were the most frequently detected bacteria in both groups, which may be related to prolonged exposure to the ICU environment. Candida albicans was the predominant fungus in the immunocompetent group, whereas Parapsilosis, Candida tropicalis, Aspergillus, and Pneumocystis jirovecii were frequently detected in the immunocompromised group in addition to Candida albicans, consistent with the findings of Liu et al. [33]. Notably, our previous study [34] showed that fungal detection via mNGS does not necessarily indicate invasive fungal infection, particularly in immunocompetent hosts, in whom such detection may represent colonization or contamination. However, Wang et al. [35] reported that immunocompetent ICU patients may also be at risk of invasive fungal infection. Therefore, in clinical practice, the clinical significance of fungal detection should not be determined solely based on a patient’s immune status.
The microbial detection spectrum in ICU patients with pneumonia unresponsive to initial treatment may also change dynamically under the influence of factors such as disease progression and antimicrobial therapy. Analysis of repeat BALF-mNGS results in this study showed that complete concordance between consecutive tests was uncommon, with most patients exhibiting either partial concordance or complete discordance, suggesting that the microbial detection spectrum in this population may change over time. Among patients with partial concordance, the most common pattern was a change in species composition within mixed bacterial–fungal detection. Even when some microorganisms remained detectable across tests, the overall composition could change under the influence of antimicrobial selection, disease progression, and host status [36]. Among patients with complete discordance, findings such as newly detected microorganisms and conversion from positive to negative detection further suggest that mNGS results are influenced by multiple factors, including disease stage and therapeutic interventions. Nevertheless, repeat testing may still provide diagnostic value in patients with a high clinical suspicion of infection despite a negative initial result or in those with a poor response to antibiotic therapy. A Chinese expert consensus [37] also recommends repeat sampling and testing when infection cannot be clinically excluded. However, complete discordance in the microorganisms detected between two consecutive tests may complicate clinical interpretation. Therefore, mNGS results should be interpreted in conjunction with the patient’s clinical manifestations, treatment response, host status, and other microbiological findings.
mNGS results can influence clinical antimicrobial treatment decisions; however, whether this translates into clinical benefit remains controversial. Zhao et al. [38] found that mNGS facilitated antimicrobial treatment adjustments and was associated with potential clinical benefits in immunocompromised patients. In contrast, Niles et al. [39] reported that when mNGS and CMTs were performed concurrently, mNGS provided limited additional value in guiding treatment. In the present study, antimicrobial regimens were adjusted based on mNGS results in 67.6% of patients, and the proportion of treatment adjustments was significantly higher in the repeat-test group than in the single-test group (87.82% vs 58.65%, p < 0.001), reflecting clinicians’ tendency to modify antimicrobial therapy based on mNGS findings. However, because this was a retrospective study and the appropriateness of individual antimicrobial treatment adjustments was not systematically evaluated, the treatment adjustment rate primarily reflects the influence of mNGS on clinical decision-making and does not directly demonstrate a prognostic benefit.
In the prognostic analysis, the unadjusted ICU mortality rate was significantly higher in the repeat-test group than in the single-test group (24.87% vs 9.66%, p < 0.001). However, after multivariable regression adjustment, repeat testing was not independently associated with ICU mortality (aOR = 1.15, 95% CI: 0.64–2.06, p = 0.630). After propensity score matching, no significant difference in ICU mortality was observed between the two groups (absolute risk difference, 1.12%; 95% CI: −7.43% to 9.67%; p = 0.798). These findings indicate that the higher crude mortality observed in the repeat-test group should be interpreted with caution. In this study, repeat mNGS testing was not performed according to a predefined protocol but was determined by treating physicians based on changes in patients’ clinical conditions and treatment responses; consequently, patients undergoing repeat testing tended to have greater baseline disease severity. The absence of an independent association between repeat testing and ICU mortality after multivariable regression and propensity score matching further suggests that the difference in crude mortality is related to greater underlying disease severity rather than repeat mNGS testing itself. Moreover, our findings do not support an improvement in patient prognosis with repeat mNGS testing. These findings indicate that microbial detection represents only one component of clinical diagnosis and management, and that the clinical value of mNGS is influenced by multiple factors, including patients’ underlying disease severity, timing of clinical decision-making and accuracy of result interpretation. Clinicians should therefore integrate mNGS findings with host immune status, antimicrobial resistance, dynamic changes in inflammatory markers, and treatment response when adjusting antimicrobial therapy. Future prospective studies are needed to further define the patient populations most likely to benefit from repeat mNGS testing and the optimal timing of testing, as well as to evaluate the appropriateness of mNGS-guided antimicrobial treatment adjustments and their impact on clinical outcomes.
This study had several limitations. (1) The specimen types used for mNGS and CMTs differed, and the lack of corresponding BALF culture data limited paired comparisons between the two methods. (2) CMTs were limited to sputum and blood cultures, whereas other conventional diagnostic methods, such as serological tests (e.g. GM assay), were not systematically evaluated. (3) Repeat mNGS testing was performed at the discretion of the treating physicians rather than according to a predefined protocol. Patients who underwent repeat testing may therefore have represented a population with greater baseline disease severity, and selection bias may have remained despite multivariable adjustment and propensity score matching. (4) Because mNGS detects microbial nucleic acids, it cannot distinguish active infection from colonization or contamination; therefore, mNGS results should be interpreted cautiously. (5) The DNA-based mNGS workflow used in this study was unable to detect RNA viruses, limiting the evaluation of viral pneumonia and viral–bacterial co-detections. (6) As this was a single-center retrospective study, residual confounding cannot be completely excluded. Future prospective multicenter studies are needed to further validate the clinical value of mNGS in guiding antimicrobial therapy adjustments.
5. Conclusion
In conclusion, in ICU patients with pneumonia unresponsive to initial treatment, mNGS provides more comprehensive microbiological information, facilitating the identification of complex microbial profiles and supporting antimicrobial treatment adjustments. Although repeat mNGS testing can provide information on dynamic changes in the microbial spectrum during disease, it was not associated with improved patient outcomes, suggesting that its clinical value should be evaluated in the context of patient selection, timing of testing, and clinical decision-making. Repeat testing may have potential value in patients with an initially negative result but a high clinical suspicion of infection or persistent disease progression. However, mNGS results should not be used as the sole basis for adjusting antimicrobial therapy; rather, they should be interpreted together with CMT findings, host immune status, and inflammatory markers to support comprehensive assessment and dynamic management and to maximize the value of mNGS in the management of ICU pneumonia.
Supplementary Material
Acknowledgements
The authors wish to thank all research staff and patients for participating in this study. CRediT: Jinwen Min, Zhili Niu and Jing Yin designed the study, analysed the data, and wrote the first draft; Sheng Xu, Shuo Zhang and Mengling Liu collected part of the data; Liangyu Li and Jieyu Mao drew some figures; Ruiyun Li and Haiyue Zhang were responsible for re-checking the data; Xiaojun Wu was involved in designing the study, interpreting the data, revising the manuscript and editing the final draft. All authors have read and reviewed the manuscript.
Funding Statement
This study was supported by National Natural Science Foundation of China (No. 82370008).
Ethics approval and consent to participate
The retrospective study involving human participants was approved by the Research Ethics Committee of Renmin Hospital of Wuhan University (Renmin Hospital of Wuhan University, Wuhan, China: Number WDRM2025-K198), which waived the written informed consent requirement. The study was conducted in accordance with the principles of the Declaration of Helsinki.
Consent for publication
Not applicable.
Disclosure statement
The authors declare that this study was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
References
- 1.Cavallazzi R, Furmanek S, Arnold FW, et al. The burden of community-acquired pneumonia requiring admission to ICU in the United States. Chest. 2020;158(3):1008–1016. doi: 10.1016/j.chest.2020.03.051. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2].Metlay JP, Waterer GW, Long AC, et al. Diagnosis and treatment of adults with community-acquired pneumonia. an official clinical practice guideline of the American thoracic society and infectious diseases society of America. Am J Respir Crit Care Med. 2019;200(7):. 45–e67. doi: 10.1164/rccm.201908-1581ST [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3].Kalil AC, Metersky ML, Klompas M, et al. Management of adults with hospital-acquired and ventilator-associated pneumonia: 2016 clinical practice guidelines by the infectious diseases society of America and the American thoracic society. Clin Infect Dis. 2016;63(5):. 61-111. doi: 10.1093/cid/ciw353 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4].Nie XM, Li YS, Yang ZW, et al. Initial empiric antibiotic therapy for community-acquired pneumonia in Chinese hospitals. Clin Microbiol Infect. 2018;24(6):. 58.e1-658–e6. doi: 10.1016/j.cmi.2017.09.014 [DOI] [PubMed] [Google Scholar]
- 5].Cillóniz C, Torres A, Niederman MS.. Management of pneumonia in critically ill patients. BMJ. 2021;375:. 065871. doi: 10.1136/bmj-2021-065871 [DOI] [PubMed] [Google Scholar]
- 6].de Nies L, Kobras CM, Stracy M.. Antibiotic-induced collateral damage to the microbiota and associated infections. Nat Rev Microbiol. 2023;21(12):. 89–804. doi: 10.1038/s41579-023-00936-9 [DOI] [PubMed] [Google Scholar]
- 7].Reyes LF, Conway Morris A, Serrano-Mayorga C, et al. Community-acquired pneumonia. Lancet. 2025;406(10517):. 371–2388. doi: 10.1016/S0140-6736(25)01493-X [DOI] [PubMed] [Google Scholar]
- 8].Zhang D, Yang D, Makam AN.. Utility of blood cultures in pneumonia. Am J Med. 2019;132(10):. 233–1238. doi: 10.1016/j.amjmed.2019.03.025 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9].Wu BG, Zinter MS.. Metagenomic sequencing for personalized treatment in pneumonia: does better detection lead to better outcomes? Chest. 2025;167(2):. 00–302. doi: 10.1016/j.chest.2024.08.031 [DOI] [PubMed] [Google Scholar]
- 10].Zhu YG, Tang XD, Lu YT, et al. Contemporary situation of community-acquired pneumonia in China: a systematic review. J Transl Int Med. 2018;6(1):. 6–31. doi: 10.2478/jtim-2018-0006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11].Gu W, Deng X, Lee M, et al. Rapid pathogen detection by metagenomic next-generation sequencing of infected body fluids. Nat Med. 2021;27(1):. 15–124. doi: 10.1038/s41591-020-1105-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12].Gu W, Miller S, Chiu CY.. Clinical metagenomic next-generation sequencing for pathogen detection. Annu Rev Pathol. 2019;14:. 19–338. doi: 10.1146/annurev-pathmechdis-012418-012751 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13].Dong Y, Chen Q, Tian B, et al. Advancing microbe detection for lower respiratory tract infection diagnosis and management with metagenomic next-generation sequencing. Infect Drug Resist. 2023;16:. 77–694. doi: 10.2147/IDR.S387134 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14].Serpa PH, Deng X, Abdelghany M, et al. Metagenomic prediction of antimicrobial resistance in critically ill patients with lower respiratory tract infections. Genome Med. 2022;14(1):. 4. doi: 10.1186/s13073-022-01072-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Evans RA, Dube S, Lu Y, et al. Impact of COVID-19 on immunocompromised populations during the Omicron era: insights from the observational population-based INFORM study. Lancet Reg Health Eur. 2023;35:100747. doi: 10.1016/j.lanepe.2023.100747. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kooshesh KA, Foy BH, Sykes DB, et al. Health consequences of thymus removal in adults. N Engl J Med. 2023;389(5):406–417. doi: 10.1056/NEJMoa2302892. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Respiratory branch of Chinese Medical Association. The guidelines for the diagnosis and treatment of community-acquired pneumonia in Chinese adults (2016 Edition). Chin J Tuberc Respir Dis. 2016;39(4):253–279. Chinese. doi: 10.3760/cma.j.issn.1001-0939.2016.04.001. [DOI] [Google Scholar]
- 18].Arancibia F, Ewig S, Martinez JA, et al. Antimicrobial treatment failures in patients with community-acquired pneumonia: causes and prognostic implications. Am J Respir Crit Care Med. 2000;162(1):. 54–160. doi: 10.1164/ajrccm.162.1.9907023 [DOI] [PubMed] [Google Scholar]
- 19].Menéndez R, Torres A, Zalacaín R, et al. Risk factors of treatment failure in community acquired pneumonia: implications for disease outcome. Thorax. 2004;59(11):. 60–965. doi: 10.1136/thx.2003.017756 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20].Mandell LA, Wunderink RG, Anzueto A, et al. Infectious diseases society of America/American Thoracic Society consensus guidelines on the management of community-acquired pneumonia in adults. Clin Infect Dis. 2007;44 (Suppl 2):. 27–S72. doi: 10.1086/511159 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21].De Waele JJ, Schouten J, Beovic B, et al. Antimicrobial de-escalation as part of antimicrobial stewardship in intensive care: no simple answers to simple questions-a viewpoint of experts. Intensive Care Med. 2020;46(2):. 36–244. doi: 10.1007/s00134-019-05871-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22].Moehring RW, Ashley ESD, Davis AE, et al. Development of an electronic definition for de-escalation of antibiotics in hospitalized patients. Clin Infect Dis. 2021;73(11):. 4507–e4514. doi: 10.1093/cid/ciaa932 [DOI] [PubMed] [Google Scholar]
- 23.Li T, Liu J, Wang X.. Diagnostic performance and clinical utility of metagenomic next-generation sequencing in suspected pulmonary infections: a comparative study stratified by immune status. Front Cell Infect Microbiol. 2026;16:1812778. doi: 10.3389/fcimb.2026.1812778. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24].Verweij PE, Rijnders BJA, Brüggemann RJM, et al. Review of influenza-associated pulmonary aspergillosis in ICU patients and proposal for a case definition: an expert opinion. Intensive Care Med. 2020;46(8):. 524–1535. doi: 10.1007/s00134-020-06091-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25].Han D, Li Z, Li R, et al. mNGS in clinical microbiology laboratories: on the road to maturity. Crit Rev Microbiol. 2019;45(5–6):. 68–685. doi: 10.1080/1040841X.2019.1681933 [DOI] [PubMed] [Google Scholar]
- 26].Diao Z, Han D, Zhang R, et al. Metagenomics next-generation sequencing tests take the stage in the diagnosis of lower respiratory tract infections. J Adv Res. 2022;38:. 01–212. doi: 10.1016/j.jare.2021.09.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27].Kreitmann L, Helms J, Martin-Loeches I, et al. ICU-acquired infections in immunocompromised patients. Intensive Care Med. 2024;50(3):. 32–349. doi: 10.1007/s00134-023-07295-2 [DOI] [PubMed] [Google Scholar]
- 28].Puerta-Alcalde P, Monzó-Gallo P, Aguilar-Guisado M, et al. Breakthrough invasive fungal infection among patients with haematologic malignancies: a national, prospective, and multicentre study. J Infect. 2023;87(1):. 6–53. doi: 10.1016/j.jinf.2023.05.005 [DOI] [PubMed] [Google Scholar]
- 29].Xin L, Jiao X, Gong X, et al. Diagnostic value of metagenomic next generation sequencing of bronchoalveolar lavage fluid in immunocompromised patients with pneumonia. Front Cell Infect Microbiol. 2025;15:. 602636. doi: 10.3389/fcimb.2025.1602636 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Huiling Y, Jinghui Z, Xinxin Y, et al. Diagnostic performance and clinical impact of metagenomic next-generation sequencing in 841 patients with suspected lower respiratory tract infections: a 4-year retrospective study from a tertiary hospital in Shenzhen, China. Int J Infect Dis. 2026;170:108885. doi: 10.1016/j.ijid.2026.108885. [DOI] [PubMed] [Google Scholar]
- 31].Azoulay E, Mokart D, Kouatchet A, et al. Acute respiratory failure in immunocompromised adults. Lancet Respir Med. 2019;7(2):. 73–186. doi: 10.1016/S2213-2600(18)30345-X [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Fang T, Yuan F, Chen Y, et al. Emerging role of metagenomic next-generation sequencing in infectious disease diagnostics: clinical integration and future directions. mLife. 2026;5(2):148–163. doi: 10.1002/mlf2.70078. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Liu Y, Wu W, Xiao Y, et al. Application of metagenomic next-generation sequencing and targeted metagenomic next-generation sequencing in diagnosing pulmonary infections in immunocompetent and immunocompromised patients. Front Cell Infect Microbiol. 2024;14:1439472. doi: 10.3389/fcimb.2024.1439472. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Liu M, Zhang H, Li L, et al. The etiological diagnostic value of metagenomic next-generation sequencing in suspected community-acquired pneumonia. BMC Infect Dis. 2024;24(1):626. doi: 10.1186/s12879-024-09507-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Wang R, Yang H, Zhang C, et al. The complexity of invasive fungal diseases in the intensive care unit: evaluation of metagenomic next-generation sequencing. Front Cell Infect Microbiol. 2026;16:1820501. doi: 10.3389/fcimb.2026.1820501. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Cheng M, Xu Y, Cui X, et al. Deep longitudinal lower respiratory tract microbiome profiling reveals genome-resolved functional and evolutionary dynamics in critical illness. Nat Commun. 2024;15(1):8361. doi: 10.1038/s41467-024-52713-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37].Editorial Board of Chinese Journal of Infectious Diseases. Clinical practice expert consensus for the application of metagenomic next-generation sequencing. Chin J Infect Dis. 2020;38:. 81–689. doi: 10.3760/cma.j.cn311365-20200731-00732 [DOI] [Google Scholar]
- 38].Zhao J, Sun Y, Tang J, et al. The clinical application of metagenomic next-generation sequencing in immunocompromised patients with severe respiratory infections in the ICU. Respir Res. 2024;25(1):. 60. doi: 10.1186/s12931-024-02991-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39].Niles DT, Wijetunge DSS, Palazzi DL, et al. Plasma metagenomic next-generation sequencing assay for identifying pathogens: a retrospective review of test utilization in a large children’s hospital. J Clin Microbiol. 2020;58(11):. 00794-20. doi: 10.1128/JCM.00794-20 [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
Data Availability Statement
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.