Infectious disease remains to be a global health concern. Despite significant improvements in microbiological testing techniques and medical treatment, the mortality rate remains high (Bloom and Cadarette, 2019). Rapid and accurate etiological diagnosis of pulmonary infection is the fundamental way to control infection, reduce mortality, and prevent the development of drug-resistant bacteria. At present, the primary approach for identifying the etiological pathogens of pulmonary infectious disease is microbial culture. Although it is effective in addressing certain clinical issues, conventional microbiological tests still suffer from some limitations, such as low positive rate, insufficient reliability, and long turnaround time (Li et al., 2019; Sin et al., 2014). In recent years, several rapid diagnostic techniques for identifying the causes of infectious diseases have been applied in healthcare settings. These approaches include antigen/antibody assays and polymerase chain reaction-based nucleic acid detection of specific pathogens (Subramony et al., 2016). However, these tests are typically restricted to healthcare professionals who anticipate potential disease-causing agents, and their outcomes are frequently influenced by thresholds (Dimech, 2021).

Metagenomic next-generation sequencing (mNGS) based on high-throughput sequencing technology has emerged as a solution. Initially, because of its high price and complex operating procedures, mNGS was only used in some scientific research fields (Chiu, 2013; Lloyd-Price et al., 2016). With the display of its superior performance and the reduction of detection cost (van Nimwegen et al., 2016), next-generation sequencing has quickly entered clinical practice, and its application in tumor diagnosis and individualized therapy (Coombs et al., 2018; Tsoulos et al., 2017) and prenatal diagnosis (Xu and Shi, 2014) have been reported. The first application of mNGS in the diagnosis of infectious diseases was reported in the New England Journal in 2014, in which mNGS was used to diagnose an infectious 14-year-old patient with severe combined immunodeficiency syndrome (Wilson et al., 2014). Since then, an increasing number of case reports and clinical studies have pointed out the value of mNGS in the diagnosis of infectious diseases, and research is no longer limited to central nervous system infections (Guan et al., 2016; Yao et al., 2016). Studies on the osteoarticular system (Ruppe et al., 2017), the respiratory system (Leo et al., 2017; Ruppe et al., 2016), and other systems have also been reported.

As a new detection method, mNGS has obvious advantages compared with traditional pathogenic detection methods: (1) It can detect all pathogenic microorganisms simultaneously and quickly; (2) It can quickly detect pathogens that are time-consuming or difficult in traditional pathogenic culture, such as Mycobacterium tuberculosis; (3) Diagnosis of rare and emerging pathogens. At present, there are several challenges in the clinical application of mNGS for diagnosing infectious diseases, such as the absence of standardized protocols, the complexity of interpreting test results, and relatively high expenses (Han et al., 2019). Moreover, it should be noted that the respiratory system is not a sterile environment, which adds complexity to the interpretation of the data. Numerous clinical studies have been published on the utility of mNGS in diagnosing pulmonary infections. However, most of these studies had small sample sizes and were limited to specific populations or pathogens (Lin et al., 2023; Shen et al., 2023; Sun et al., 2022; Wang et al., 2023). BALF mNGS has been used in our department for many years to assist in the diagnosis of pathogens of pulmonary infection. In actual clinical work, we have observed that BALF mNGS has indeed provided great help for the diagnosis and treatment of most patients, but we are not clear about which patients are more suitable for this examination or benefit more. There is a scarcity of studies on the direct impact of BALF mNGS on the diagnosis and treatment of patients with suspected pulmonary infections, including both mild and severe cases. This study was conducted to gain a deeper understanding of the practical application of mNGS in clinical practice.

Our study revealed a significantly higher prevalence of fever (32.1% vs. 3%, p=0.001) among patients diagnosed with pulmonary infections than among those without pulmonary infections. Patients with malignant tumors (11.6% vs. 0, p=0.032) were more prone to developing pulmonary infections than those with hypertension, cardiovascular disease, and diabetes. However, in this study, the number of patients with pulmonary infections complicated by malignant tumors is relatively small (n=26). Therefore, this finding is considered as exploratory, and further validation with larger clinical samples is needed to confirm the correlation between pulmonary infections and tumors. The likelihood of infection is significantly elevated in the presence of fever and abnormal lung infiltration. A comparative analysis of pulmonary CT images between the PI and NPI groups showed that consolidation (29% vs. 9.1%, p=0.15) was more prevalent in patients with lung infections. In contrast, bilateral involvement (63.4% VS 81.8%, p<0.001), ground-glass opacity (21% vs. 45.5%, p=0.002), and interstitial lesions (17% VS 60.6%, p<0.001) were more common in non-pulmonary infections. The comparative analysis of blood routine, biochemical and other blood test results indicated that patients in the lung infection group exhibited lower levels of Hb (118.85 g/L vs. 129.64 g/L, p=0.006) and serum albumin (36.6 g/L vs. 39.5 g/L, p=0.01), as well as higher levels of inflammatory markers, including CRP (19.35 mg/L vs. 3.13 mg/L, p<0.001), ESR (33 mm/60 min vs. 9.5 mm/60 min, p=0.004), ferritin (239.2 ng/mL vs. 134.5 ng/mL, p=0.009), SAA (31.9 mg/L vs. 4.1 mg/L, p=0.003), and FIB (4.76 g/L vs. 3.6 g/L, p<0.001). These findings demonstrate that, in our clinical practice, patients who exhibited fever, had malignant tumors and abnormalities such as consolidation on CT images, and showed elevated inflammatory markers on blood tests were more likely to have lung infections. These findings are consistent with observations in clinical practice and have been documented in published studies (Akinosoglou et al., 2013; Aliberti et al., 2021; Musher and Thorner, 2014; Valvani et al., 2019). It may also have a certain guiding significance for selecting appropriate patients with suspected pulmonary infections to perform BALF mNGS in the future.

Several published studies have shown that the positive rate and sensitivity of mNGS are superior to those of CMT, especially in the diagnosis of mixed infections, critically ill patients, and immunocompromised individuals (Deng et al., 2022; Parize et al., 2017; Zhong et al., 2023). A prospective multicenter study demonstrated that untargeted next-generation sequencing detected significantly more clinically relevant viruses and bacteria than conventional methods (36% vs 11%, p<0.001) in immunocompromised adults, with high negative predictive value (64/65, 95% CI 0.95-1), highlighting its diagnostic potential (Parize et al., 2017). Deng et al. (2022) compared the diagnostic performance of bronchoalveolar lavage fluid (BALF) mNGS versus CMT in pediatric pneumonia. Their results showed that mNGS had significantly higher overall pathogen detection rates (91.3% vs. 59.2%, p<0.001), with superior sensitivity for bacterial and viral infections. However, mNGS exhibited lower sensitivity for Mycoplasma pneumoniae (42.1% vs. 100%) compared to CMT. Additionally, mNGS identified virus-bacteria co-infections as the most prevalent polymicrobial pattern. Zhong et al. (2023) found that mNGS showed significantly higher detection rates than conventional culture (93.3% vs. 29.3%, p<0.0001) in AIDS patients with pulmonary infections. mNGS also identified polymicrobial infections in 60% of these patients, further emphasizing its superior diagnostic value. Nevertheless, Peng et al. (2021) retrospectively analyzed 60 critically ill immunocompromised patients with suspected pneumonia and compared the diagnostic performance of BALF mNGS and CMTs. Their study found that the overall diagnostic accuracy was similar between mNGS and CMTs. While mNGS was better at identifying viral pneumonia, it was less accurate for fungal infections. The authors suggested that combining mNGS and CMTs may be the optimal diagnostic strategy.

Collectively, the diagnostic effectiveness of mNGS varies depending on the specific types of pathogens, types of specimens being tested, and patient characteristics. Regardless of cost, the combination of mNGS and CMT remains the most effective diagnostic method (Qu et al., 2022).

For certain pathogens that lack rapid and effective diagnostic methods in some medical institutions, such as C. psittaci, P. jirovecii, NTM, L. pneumophila, mNGS detection methods also have unique benefits (Chen et al., 2023; Du et al., 2023; Huang et al., 2023; Tang et al., 2023), which is consistent with our findings ( Figure 2 ). This study showed that mNGS can detect a wider range of pathogens compared to CMT. It was able to identify pathogens that typically require advanced culture conditions, such as S. pneumoniae, H. influenzae, S. constellatus, Nocardia spp.

Although both mNGS and CMT showed a high positive rate in patients with suspected lung infections, mNGS showed a significantly higher rate ( Figure 3A ). There was also a relatively high level of disagreement between the two methods ( Figure 2A ), which could be due to the broad range of tests included in the CMT, such as the T-SPOT, sputum culture, BALF GM test, and blood G test. CMT positivity cannot completely confirm the corresponding pathogen infection but will be included in the positive statistical results.

This study found comparable detection rates for M. tuberculosis and Aspergillus between mNGS and CMT ( Figures 3G, H ). CMT, including sputum smear, BALF X-pert, T-SPOT, and galactomannan tests, is cost-effective for routine cases. But mNGS excels in complex scenarios, detecting M. tuberculosis in atypical or CMT-negative cases and identifying Aspergillus species for targeted therapy. mNGS also detects co-infections, enhancing diagnostic scope. While CMT is preferred for routine tuberculosis screening, mNGS may be more economical when invasive procedures are needed. mNGS is not recommended for routine use but adds value in challenging cases. However, subgroup analyses are often underpowered ( Supplementary Table 3 ), with most sample sizes below those required for 80% power, except for immunocompetent and pulmonary infection subgroups. Thus, subgroup findings are exploratory, requiring larger studies for validation. Unlike previous studies, we conducted an analysis to assess the impact of the mNGS test on the final diagnosis and adjustment of the treatment regimen for each patient to gain a deeper understanding of its impact on actual clinical diagnosis and therapy. The mNGS diagnosis impact (D1–D7) and treatment impact (T1–T6) classifications were enhanced by evaluating inter-rater reliability, yielding Cohen’s kappa values of 0.866 (95% CI: 0.796–0.937) for diagnosis ( Supplementary Table 4 ) and 0.868 (95% CI: 0.783–0.953) for treatment ( Supplementary Table 5 ), indicating strong agreement. Our results show that BALF mNGS has several beneficial effects on diagnosis and treatment. Specifically, mNGS assists in the diagnosis of mixed infections and identification of infectious pathogens. Moreover, it aided in initiating appropriate antibiotic treatment and confirmed the effectiveness of empirical treatment. Further subgroup analysis showed that BALF mNGS in immunocompromised patients suspected of having pulmonary infection could have more positive effects on diagnosis (42.79% vs. 62.86%, p= 0.0267) and treatment (58.56% vs. 71.43%, p= 0.148).

Liang and colleagues (Liang et al., 2022) conducted a study to analyze the effect of BALF mNGS on the treatment of patients with suspected lower respiratory tract infections. The results showed that 3.6% of patients were degraded by antibiotics (reducing the antibacterial spectrum or reducing the types of antibiotics), 23.6% were upgraded (increasing the types of antibiotics or changing to broad-spectrum antibiotics), 60.7% remained unchanged, and 12.1% were transferred to a lung hospital. Studies have shown that the majority of patients do not adjust their treatment plan following mNGS testing; however, the mNGS results verify the accuracy of empirical treatment, which is consistent with our research results ( Figure 4 ). We are confident that the results of mNGS will have a positive impact on this particular group of patients. Xiao and colleagues (Jin et al., 2022) also found that the ultimate diagnosis and treatment adjustment of cases primarily relied on the combined findings of mNGS, CT, other tests, and clinical features. Modifications and adjustments were made solely based on the results of mNGS in 32 (32/246, 13.01%) and 23 (23/246, 9.35%) cases, respectively. These modifications have beneficial effects on the disease progression and prognosis of these patients. Most of these patients were infected with M. tuberculosis, NTM, or atypical pathogens.

Han and his colleagues (Han et al., 2023) discovered that the results of plasma mNGS had a positive effect on 83 patients (57.1%). This was achieved by accurately diagnosing or excluding infections and initiating targeted therapies. However, only 32.4% (11/34) of the negative mNGS tests showed a positive impact, suggesting that plasma mNGS testing alone may not be a powerful tool for ruling out infections in clinical settings.

Although the tested samples and classification criteria for the impact on diagnosis and treatment may vary, our study classified a negative result from the NGS test as having no effect on the clinical diagnosis. However, our results showed that for suspected pneumonia patients with normal immunity, approximately half of the mNGS results had a positive impact on the diagnosis and treatment. More clinical trials and studies are required to gain a deeper understanding of the appropriate individuals to test, the optimal timing for testing, and how to optimize the diagnostic efficiency and socioeconomic benefits of mNGS for pulmonary infectious diseases. Furthermore, this manual classification system prioritizes diagnostic context over automated interpretation. Retrospective design allows granular analysis but limits real-time clinical generalizability. Current framework requires specialized expertise, making it less suitable for emergency settings where rapid decision-making predominates.

In clinical practice, we frequently encounter patients presenting with either radiological pulmonary abnormalities and/or respiratory symptoms, which necessitates comprehensive clinical evaluation to differentiate true pulmonary infections from other pathologies and decide on antimicrobial therapy initiation. Our study specifically addressed this diagnostic challenge through rigorous case selection criteria (detailed in “Study design and patient population”), where the cohort potentially included non-infectious pulmonary conditions such as interstitial lung disease, pulmonary edema, and neoplastic lesions. Critical analysis revealed that patients demonstrating specific clinical features (e.g., febrile presentation), comorbid conditions (particularly malignancies), characteristic CT findings (consolidation or ground-glass opacities), and laboratory abnormalities (elevated inflammatory markers) showed significantly higher probabilities of confirmed pulmonary infections. These findings systematically validate our clinical experience through three key dimensions: 1. Diagnostic Triangulation: Integration of microbiological, radiological, and laboratory evidence. 2. Risk Stratification: Identification of high-yield clinical predictors for infection likelihood. 3. Diagnostic Stewardship: Guidance for judicious antimicrobial use in ambiguous cases. This evidence-based alignment between empirical clinical judgment and systematic research outcomes reinforces the necessity of multidimensional assessment in pulmonary infection diagnosis. The concordance observed provides scientific validation for current diagnostic protocols.

Compared to CMTs, mNGS has a higher initial cost. This is due to the expensive sequencer, sophisticated reagents, and significant computational resources required for data analysis. As a result, the cost of a single mNGS test is often several times higher than that of traditional methods. However, mNGS provides a comprehensive view of the microbial community in a sample within a relatively short time, typically within 24–48 hours. This rapid turnaround time can potentially reduce the overall cost of care by enabling more targeted and timely treatment, thereby avoiding prolonged and inappropriate empirical therapy (Miller and Chiu, 2021). In contrast, CMTs have limitations in detecting fastidious or unculturable pathogens. For example, many viruses and some intracellular bacteria are difficult to identify using conventional culture methods, which can lead to misdiagnosis or delayed diagnosis. This, in turn, affects the effectiveness of treatment and patient outcomes. In contrast, mNGS can detect a wide range of pathogens, including novel and emerging ones, without prior knowledge of the causative agent. This enhanced diagnostic capability can lead to more effective treatment strategies and improved patient prognosis. In summary, while the initial costs of implementing NGS technologies in infectious disease diagnostics are high, the potential for improved diagnostic accuracy, reduced hospital stays, and more effective patient management can justify these expenses. The continued development and refinement of NGS technologies, alongside advances in bioinformatics and data analysis, are expected to further enhance their cost-effectiveness and clinical utility in the future (Liu and Ma, 2024; Haslam, 2021).

This study has several limitations. First, most patients underwent only mNGS DNA testing, potentially missing RNA viruses. Second, as a single-center study with few immunocompromised patients, generalizability is limited. Third, the retrospective design introduced selection bias, as mNGS was often used for complex cases or patients with better socioeconomic status, restricting applicability to broader populations. Fourth, missing data in routine tests or CMTs, particularly in milder cases, may have affected precision. A complete case analysis ( Supplementary Table 6 ), including only patients with complete data, confirmed robust results. Fifth, the lack of standardized mNGS interpretation criteria and unblinded diagnostic adjudication may have introduced bias, though mitigated by expert consensus and predefined thresholds. Finally, cost-effectiveness was not evaluated. Future prospective, multicenter studies with larger cohorts, standardized criteria, blinded adjudication, and cost analyses are needed to validate our findings.