Metagenomic next-generation sequencing for the clinical diagnosis and prognosis of acute respiratory distress syndrome caused by severe pneumonia: a retrospective study

Ethical approval and consent

The protocol used in this retrospective study was reviewed and approved by the Ethical Review Committee of Jiangmen Central Hospital (No: 2019-15). Patient’s informed consent was obtained from patients or their next of kin.

Study participants

A retrospective analysis was conducted on all ARDS cases resulting from severe pneumonia in patients 18 years and older, admitted to the ICU at Jiangmen Central Hospital from January 2018 to August 2019. For our study, ARDS was diagnosed according to the 2012 Berlin definition of the disease (ARDS Definition Task Force, 2012). Patients were excluded from the study if their ARDS was not caused by severe pneumonia or if they did not follow through with their treatment for any reason.

All patients were endotracheally intubated, mechanically ventilated, and underwent a fiberoptic bronchoscopy to obtain clinical specimens for microbial testing. Patients were included in the next generation sequencing (NGS) group when informed consent was provided for testing; those who were not tested by mNGS were grouped into the no-NGS group. Owing to the cost of mNGS, only DNA sequencing was performed. Samples of bronchoalveolar lavage fluid (BALF) were acquired from patients in the NGS group and sent for pathogen testing at BGI Clinical Laboratories (Shenzhen) Co., Ltd. Once the laboratory received the samples, nucleic acid extraction, library construction, high-throughput sequencing, bioinformatics analysis, and pathogen data interpretation were performed according to previous studies (Miao et al., 2018).

Microbiological testing

Both groups were tested using the same conventional method (routine culture + serum antibody + PCR). The NGS group used mNGS + routine culture + serum antibody + PCR, while the no-NGS group used routine culture + serum antibody + PCR. Pathogenic microbes that cause severe pneumonia are typically bacteria, fungi, or viruses. Restricted by inspection conditions of the hospital, serum antibody and PCR nucleic acid detection could only detect some special pathogens and viruses that were clinically difficult to culture, as a supplement to routine culture. The serum antibody included Mycoplasma pneumoniae, Chlamydia pneumoniae, Coxsackie virus, cytomegalovirus, influenza A, influenza B, respiratory syncytial virus, and parainfluenza virus. PCR nucleic acids included Legionella pneumophila, Mycoplasma pneumoniae, C. pneumoniae, Mycobacterium tuberculosis, influenza A, and influenza B.

Clinical treatment

All patients underwent empirical antimicrobial treatment according to the Chinese Adult Community-Acquired Pneumonia Diagnosis and Treatment Guide (Cao et al., 2018) and the Chinese Adult Hospital-Acquired Pneumonia and Ventilator-associated Pneumonia Diagnosis Guide (Department of Infectious Diseases, Chinese Medical Association Respiratory Branch, 2018), combined with respiratory infection indicators and imaging. All patients were treated with mechanical ventilation according to the ARDS ventilation guidelines (Bein et al., 2016; Griffiths et al., 2019). The no-NGS patients were treated with an antimicrobial regimen based on the results of conventional microbiological tests. The antimicrobial regimen of NGS patients were adjusted case-by-case according to mNGS results.

Information collection and analysis

Patient data included age, gender, basic disease, laboratory test results before treatment, ventilator parameters, conventional microbiological tests, serum biomarkers, ICU special treatment data, APACHE II, and SOFA scores. Data were collected and compared between the two groups. The primary outcome was measured by a 28-day all-cause mortality. Secondary outcomes were measured as the length of stay in the ICU, duration of mechanical ventilation, duration of extracorporeal membrane oxygenation (ECMO), duration of prone ventilation positioning, and ICU treatment costs. Patients that showed signs of immunosuppression were selected from both groups and their prognosis compared using the same aforementioned outcomes. Cox regression analysis was conducted to analyze risk factors for ARDS prognosis. The mNGS results were compared with those of conventional microbiological tests in the NGS group.

Statistical analysis

The t-test was used to determine normal distribution and uniformity of variance. The Wilcoxon rank test was used to calculate variance of measured data that were not normally distributed or had variance homogeneity. The chi-square test was used to calculate the difference between both groups. All statistical analyses were conducted using GraphPad 5.0 and R3.4.4 software. P < 0.05 was considered statistically significant.

Sample and patient characteristics

A total of 105 patients with ARDS caused by severe pneumonia were screened in this study and 10 patients were excluded based on exclusion criteria. Fourty two patients were placed into the NGS group and 53 patients in the no-NGS group. Three patients in the NGS group had two mNGS tests performed and a total of 45 BALF samples were sent for mNGS.

Patient demographics, characteristic baselines, and ICU special treatments in the NGS and no-NGS groups were shown in Tables 1–3, respectively. There were no significant differences in age, gender, basic disease, laboratory test results before treatment, ventilator parameters, APACHE II and SOFA scores before treatment, and incidences of special treatment in the ICU between both groups (P > 0.05).

Comparison of outcomes between NGS and no-NGS groups

There was a significant difference in the 28-day all-cause mortality between both groups (P = 0.006) (Table 4). The 28-day survival was significantly higher in the NGS group than in the no-NGS group (Hazard Ratio = 2.41, 95% CI: 1.21–4.17, P = 0.01) (Fig. 1). There was no significant difference in the length of stay in the ICU, duration of mechanical ventilation, ECMO, prone position ventilation, or the cost of the ICU stay between both groups (P > 0.05) (Table 4).

Prognosis of ARDS patients

Cox univariate analysis was performed on all factors and Cox multivariate analysis was performed with variates which were P < 0.2 of the Cox univariate analysis (Table S1). The NGS or no-NGS group, length of stay in ICU, and APACHE II and SOFA scores before treatment were risk factors in patients with ARDS caused by severe pneumonia. The NGS group patients had a better prognosis than that of the no-NGS group patients (P = 0.005). A shorter stay in the ICU (P = 0.037), and lower APACHE II (P = 0.016) and SOFA scores before treatment (P = 0.003) had a better prognosis (Table 5).

Comparison of mNGS results and culture results in the NGS group

The current research showed that the mNGS test can detect more pathogens than the culture method. We analyzed the consistency of pathogens identified by both techniques. The test results were considered to be consistent when the pathogens identified by mNGS were the same as the pathogens obtained from culture. The test results were also considered consistent if mNGS identified more pathogens than the culture method. The result was partially consistent when pathogens identified by both methods were partially congruent. The results were considered inconsistent when pathogens identified by both methods varied completely. Identified pathogens (31.1%) in the NGS group were consistent, 15.6% were partially consistent, and 53.3% were completely inconsistent. In the inconsistent ones, 62.5% were negative for the culture method, while 8.3% were negative for mNGS, and 29.2% were mismatched (Fig. 2).

Comparison metagenomic of NGS results and conventional microbiological tests

Some special pathogens were difficult to obtain via culture. Therefore, Legionella, Tuberculosis, Mycoplasma/Chlamydia, parasites, K. spores, etc. were defined as such. Severe pneumonia is not caused by a single pathogen and is typically accompanied by co-infections. A co-infection is defined as a non-single pathogenic infection, such as bacteria + fungi/bacteria + virus/fungi + virus/bacteria + fungi + virus.

The positive rate of mNGS virus detection was lower than that of serum antibody detection plus PCR (6.7% vs. 26.7%, P = 0.021). In this study, mNGS only performed DNA sequencing and could only detect DNA viruses, whereas viruses identified by serological antibody detection and PCR were RNA viruses, such as influenza A and influenza B. mNGS was significantly better at detecting bacteria than serological antibody testing plus PCR (24.4% vs. 0%, P = 0.001). Further, mNGS was able to detect specific pathogens better than the culture method (22.2% vs. 0%, P = 0.001) and serological antibody testing plus PCR (22.2% vs. 2.2%, P = 0.007). Additionally, mNGS was significantly better at the identification of co-infections than serological antibody tests plus PCR (26.7% vs. 0%, P < 0.001). Finally, mNGS proved to be significantly better at identifying pathogens than the culture method (91.1% vs. 62.2%, P = 0.001) and serological antibody testing plus PCR (91.1% vs. 28.9%, P < 0.001) (Table 6).

Clinical medication guidance between NGS and no-NGS groups

In the NGS group, 30 patients (71.4%) did not cover all the microbial detected by mNGS in the initial empirical antimicrobial treatment. Thus, antimicrobial regimen needs to be modified accordingly based on the mNGS results. In the no-NGS group, empirical antimicrobial treatment that could not cover the detected microbials was found in 23 patients (43.4%), according to the results of traditional microbiological testing, and they were necessary to adjust the anti-infection program (Fig. 3). Following adjustment of the anti-infective regimen, we continuously observed APACHE II and SOFA scores for both groups of patients for 7 days and found that the NGS group had a lower APACHE II score than the no-NGS group, 7 days after treatment (P = 0.041) (Fig. 4).

Immunosuppressed patients

Clinical features of immunosuppressed patients were complicated. A total of 21 immunosuppressed patients were enrolled in our study, eight were subjected to mNGS pathogen detection, and 13 did not undergo mNGS. Three cultures were positive in the NGS group, consistent with pathogens identified by mNGS, including five Pneumocystis jirovecii, one Rhizopus, one Cryptococcus, and one human herpesvirus; six were co-infections.

In the no-NGS group, nine cases were positive for culture, and two Stenotrophomonas maltophilia, two Acinetobacter baumannii, one Staphylococcus aureus, four Candida, and one Aspergillus were detected. Four cases had multi-drug resistant bacteria. There was no significant difference in the 28-day all-cause mortality between the two groups (37.5% vs. 53.8%, P = 0.659). However, there were significant differences in the length of stay in the ICU (P = 0.023), duration of mechanical ventilation (P = 0.030), and cost of the stay in the ICU (P = 0.004) between both groups of immunosuppressed patients (Fig. 5).