General patient characteristics
A total of 541 patients were included in this study, with 271 males and a median age of 58 years (interquartile range: 43–67 years). A total of 555 BALF samples were collected (Fig. 1), with 12 patients having 2 BALF samples and 1 patient with 3, collected based on the patients’ conditions. According to the study design, 93 patients followed up for one year, and 7 patients lost to follow-up.
Flow chart of the study design and mNGS results. * Of the 541 patients, 2 BALF samples were collected from 12 patients and 3 samples were collected from 1 patient. ** 4/487 patients had samples sent for mNGS on two separate occasions, with one mNGS assay for bacteria and the other for bacterial-fungal coinfection. Of these 4 patients, 1 responded to initial antibacterial therapy, 2 died, and 1 was diagnosed with a non-infective pulmonary nodule. These 4 patients were double-counted in the bacterial and bacterial-fungal coinfection groups. BALF, bronchoalveolar lavage fluid; CAP, community-acquired pneumonia; mNGS, metagenomic next-generation sequencing; NTM, non-tuberculous mycobacteria; TB, tuberculosis
Detection of pathogen by mNGS
The positivity rate of the 555 BALF samples was 79.1% (N = 439), and the negativity rate was 20.9% (N = 116). The mNGS results were categorized into bacteria, fungi, and mycobacteria (comprising Mycobacterium tuberculosis [MTB] and non-tuberculous mycobacteria [NTM]). Bacteria, fungi, and mycobacteria were detected in 334 (60.2%), 254 (45.8%), and 37 (6.7%) BALF samples, respectively. Only one bacterium was detected in 20.2% (N = 112) of the bacteria-positive BALF samples, ≥ 2 bacteria in 9.5% (N = 53), and coinfection with other types of pathogens in 30.5% (N = 169). Only one fungus was detected in 14.2% (N = 79) of the fungi-positive BALF samples, ≥ 2 fungi in 1.8% (N = 10), and coinfection with other types of pathogens in 29.7% (N = 165). Mycobacterium alone was detected in 1.8% (N = 10) of the mycobacterium-positive samples and coinfections with other types of pathogens in 4.9% (N = 27). These 37 cases of mycobacterium comprised 33 (5.9%) MTB and 4 (0.7%) NTM (3 intracellular mycobacteria and 1 Mycobacterium kansasii). Coinfections were detected in 238 (42.9%) BALF samples, with bacterial-fungal coinfection being the most common (N = 148, 26.7%) (Fig. 2). The top three bacteria were Streptococcus pneumoniae, Pseudomonas aeruginosa, and Haemophilus influenzae, and the top three fungi were Candida albicans, Aspergillus flavus, and Candida tropicalis (Fig. 3).
Composition of pathogens detected by mNGS. MTB, Mycobacterium tuberculosis ; NTM, non-tuberculous mycobacteria
Distribution of bacteria and fungi detected by mNGS
Pathogen detection: mNGS vs. sputum culture
Of the 541 patients, 25 patients did not have sputum samples for sputum culture. For the remaining patients, pathogens were detected by both mNGS and sputum culture, and mNGS had a significantly higher positivity rate than sputum culture (79.1% vs. 11.4%, respectively, P = 0.001). The results of the two methods were consistent in 159 cases (55 double-positive and 104 double-negative), with a concordance accuracy of 30.8% (Table 1). Of the 55 double-positive samples, 72.7% (N = 40) showed complete (N = 9) or partial (N = 31) concordance between mNGS and sputum culture.
Diagnostic value of tuberculosis (TB) detection: mNGS vs. X-pert MTB/RIF
Of the 112 pathogen-negative cases, 1 was pathologically suggestive of TB and 5 were positive by X-pert MTB/RIF. Of the 429 pathogen-positive cases, 1 was pathologically suggestive of TB, 1 improved after empirical anti-tuberculosis therapy, 4 were NTM-positive by mNGS, 31 were double-positive by both mNGS and X-pert MTB/RIF, 9 were only positive by X-pert MTB/RIF, and 2 that were not sent for X-pert MTB/RIF were positive by mNGS. Overall, 50 patients with confirmed or suspected TB and 4 patients with suspected NTM infection were immediately transferred to the Wuhan Pulmonary Hospital for further diagnosis and treatment. Ultimately, after 1 year of follow-up, 50 patients were diagnosed with TB and 4 with NTM infection. Of the 50 TB cases, 48 were tested by both mNGS and X-pert MTB/RIF. The sensitivity for diagnosing TB by mNGS and X-pert MTB/RIF was 64.6% and 85.4%, respectively (P = 0.031), and the specificity was 100% for both modalities (Table 2).
Value of mNGS for diagnosis and treatment guidance in suspected CAP
Excluding the 50 patients with TB and 4 with NTM infection, we categorized the mNGS results of the remaining 487 patients as pathogen-negative, bacteria-positive (including 1 and ≥ 2 bacteria), fungi-positive (including 1 and ≥ 2 fungi), or bacterial-fungal coinfection.
Of the 487 patients included in the evaluation of treatment effect, 67.1% (N = 327) improved after initial empirical antibacterial therapy. Of the 106 pathogen-negative cases, 49.1% (N = 52) improved with initial empirical antibacterial therapy and 45.3% (N = 48) were eventually considered non-infectious diseases, resulting in a negative predictive value of 45.3%. Of the 381 pathogen-positive cases, 72.2% (N = 275) improved with initial empirical antibacterial therapy, 81.6% (N = 311) after anti-infective therapy, and 15.0% (N = 57) were ultimately considered non-infectious diseases, resulting in a positive predictive value of 81.6%. Of the 163 bacteria-positive cases, 77.9% (N = 127) improved after antibacterial therapy. Of the 85 fungi-positive cases, 60.0% (N = 51) improved after antibacterial therapy and 12.9% (N = 11) after antifungal therapy. Of the 137 bacterial-fungal coinfection cases, 86.1% (N = 118) improved after antibacterial therapy (Fig. 4).
Proportion of treatment effect in different groups of mNGS results
Of the 106 patients who were pathogen-negative, 52 (49.1%) improved after initial empirical antibacterial treatment, 2 with intensified antibacterial treatment, and 1 with antifungal treatment. The remaining 51 cases did not show significant improvement. 48 of these were ultimately diagnosed as non-infectious diseases (Figs. 1 and 4) after relevant tests (including pathology) or 1 year of follow-up, and 3 were lost to follow-up. These non-infectious cases included 17 cases of pulmonary nodules, 7 lung cancers, 6 organizing pneumonia, 5 inactive TB, 3 chronic pulmonary inflammation, 2 interstitial lung diseases, 2 pulmonary alveolar proteinosis, 1 silicosis, 1 bronchiectasis, 1 Wegener’s granulomatosis, 1 idiopathic pulmonary hemosiderosis, 1 chronic obstructive pulmonary disease, and 1 mediastinal tumor (Fig. 5).
Composition of non-infectious diseases in different groups of mNGS results. The size of the bubble area corresponds to the number of cases of non-infectious diseases. TB, tuberculosis
Bacteria were detected by mNGS in 163 patients; of these, 122 (74.8%) improved after initial empirical antibacterial treatment, and 5 improved with intensified antibacterial treatment. Of the 36 cases that did not show significant improvement, 4 died during hospitalization, and 29 were eventually diagnosed as non-infectious diseases (Figs. 1 and 4) after relevant tests or 1 year of follow-up and 3 cases were lost to follow-up. These non-infectious cases included 11 cases of pulmonary nodules, 9 lung cancers, 4 organizing pneumonia, 2 interstitial lung diseases, 1 inactive TB, 1 chronic pulmonary inflammation, and 1 bronchiectasis (Fig. 5).
Fungi were detected by mNGS in 85 patients; of these, 44 (51.8%) improved after initial empirical antibacterial treatment, 7 after intensified antibacterial treatment, and 11 after subsequent adjustment to antifungal therapy (including 5 out of 40 molds, 6 out of 40 yeasts and 0 out of 5 mold-yeast coinfections). The remaining 23 cases did not show significant improvement. Of these, 22 cases were diagnosed as non-infectious diseases (Figs. 1 and 4) after relevant tests or 1 year of follow-up, including 9 cases of pulmonary nodules, 6 lung cancers, 1 organizing pneumonia, 1 interstitial lung disease, 3 inactive TB, 1 chronic pulmonary inflammation, and 1 panbronchiolitis (Fig. 5), and 1 was lost to follow up.
Bacterial-fungal coinfection was detected by mNGS in 137 patients; 110 (80.3%) improved after initial empirical antibacterial treatment, 8 after intensive antibacterial treatment, and 5 after subsequent adjustment to antifungal therapy (including 1 out of 50 molds, 11 out of 71 yeasts and 3 out of 16 mold-yeast coinfections). The remaining 14 cases did not show significant improvement. Of these, 7 had died during hospitalization, and 7 were eventually diagnosed with non-infectious diseases (Figs. 1 and 4) after relevant tests or 1 year of follow-up. These included 3 cases of pulmonary nodules, 2 lung cancers, 1 inactive TB, and 1 bronchiectasis (Fig. 5).