Here we describe the clinical accuracy validation and practical clinical application of metagenomic next-generation sequencing. Key advances in our research include (1) The accuracy of quantitative microbial cfDNA sequencing tests validated that platform's analysis and interpretation of sequencing date. (2) Detect a wide range of sample types include Sputum, Bronchoalveolar lvge fluid, Cerebrospinal fluid, Blood et al. (3) The mNGS tests were an in-house microbiology laboratory, which increases the accuracy of results because of preserves the vitality of the microbes due to reduced turnaround time from bedside to bench. (4) All samples combined both RNA and DNA sequence, most of other studies using DNA sequence only. This approach allowed for simultaneous host transcriptional profiling, enriched for actively transcribing microbes (versus nonviable taxa or latent)(17).
In this study, we systematically compared mNGS identification and culture in a pairwise manner. The founding that mNGS detection rate was significantly higher than culture is in contrast to that recently reported in California where mNGS are not more sensitive than culture when detected Klebsiella pneumoniae(17). The differences between studies may have been driven by discrepancy in the overall level of microbiology laboratory service. Results from different laboratories may not be universally applicable. In addition, culture often reported a single microorganism or negative, which may be due to the competition between the different species of microorganisms or to the administration empirical antibiotics to patients prior to obtaining samples(18). However, mNGS is less affected by prior antibiotic exposure because it does not depend on pathogen survival(19, 20), and for most BALF and sputum samples, mNGS reports up to 5 + microorganisms. The high detection rate in the present study was closer to that reported in a retrospective study where mNGS-positive samples was significantly higher than that of culture-positive samples(19).
Not surprisingly, mNGS exhibited better diagnostic performance than conventional microbiological culture for detecting bacteria and fungus. The mNGS test was able to detect a possible pathogen in the vast majority of cases. This was consistent with previous research. But among these “mNGS false-negative” cases, mNGS testing failed to detect fifteen cases of M. tuberculosis, a higher number than other false-negative bacteria. This may be due to the difficulty of breaking M. tuberculosis during nucleic acid extraction. Although bead tapping can increase the detection rate of certain bacteria and fungi that contain rigid cell walls, it can simultaneously decrease detection sensitivity by increasing host background release of human DNA(7). Excluding blood samples, the sensitivity of mNGS testing were highly in other samples (ranging from 88.89 to 93.79%), which may be due to the low pathogen burden in blood compared with other types of specimens. We showed here that, there was a detected 195-fold higher microbial reads in the paired other samples than in blood from the same patient. Higher level of pathogen load in samples can increase the credibility of mNGS results. The mNGS of sputum and bronchoalveolar lavage fluid outperformed that for other types of samples in the sensitivity of detection of pathogens, the number of reads, the ratio of microbial reads and the rate of genome coverage.
However, multiple probable pathogens were identified in many cases whose clinical symptoms were not concordance with pathogens. The additional detections complicate the interpretation of microbiological reports. Our accuracy validation tests ruled out contamination as the source of these reads, we speculate that the additional pathogens detected were primarily components of the human microbiome and possibly derived from mucosal barrier, cfDNA residue from previous infections and transient bacteremia caused by colonizing bacteria. Dead microorganisms don’t cause disease, but they may remain to secret detectable small nucleic acid fragments. Therefore, it is recommended that “probable” pathogens should be comprehensively judged by professionals who have certain bioinformatics knowledge and are engaged in clinical infection or clinical microbiology, combined with the reading length of the detected pathogenic microorganisms as well as the differences in the sample types, the phylogeny and underlying molecular biology of the pathogens, the clinical background of patients, imaging data and other laboratory examination results. Considering antimycobacterial, antifungal and antiviral therapies can be toxic and costly, clinicians must be comprehensively judged before initiating antimicrobials therapy for pathogens with few sequences detected, such as M. tuberculosis, Aspergillus and Nocardia. Without correct interpretation, blind use of antimicrobials based on mNGS results will inevitably lead to unnecessarily broaden or prolong antibiotic therapy.
The finding that 84 of 200 (42.0%) blood mNGS results, 89 of 154 (57.8%) bronchoalveolar lavage fluid mNGS results lead to positive impact, which indicated alveolar lavage fluid mNGS had a higher positive impact than blood mNGS. Our result is different from recently published in a multicenter in USA where 6/82 (7.3%) plasma mNGS results were considered positive impact(9). In our study, clinical impact based on the decision made by the treatment team after interpretating the mNGS results and whether leading to a favorable clinical outcome. Another retrospective study on pediatric patients was performed at USA showed mNGS added little diagnostic value when conducted simultaneously with conventional testing, there was no change in management when additional organisms were identified by mNGS in the majority of cases(21), but this study was only assessed the use of mNGS for pediatric patients. In short, the reason why we are inconsistent with the above studies may be due to differences in sample types, patient populations, testing indication and the proportion of critically ill patients. In the past, mNGS has been used as the last test resort for critically ill patients due to price as high as 500 US dollars(22). However, for critically ill patients, the clinical outcome may be poor, although mNGS results may be useful in guiding antibiotic management. In our study, we tend to order mNGS as a first-line test for critically ill patients and patients with suspected atypical pathogen infection. This may explain why this study has higher positive clinical impact versus previous study. Nevertheless, 16 of 518 (3.1%) cases included condition were very critically and clinical impact was difficult to ascertain.
The limitations of this study include the following. First, the number of some sample types like sputum and peritoneal is low. Second, this study is a single-central study with limited representativeness, which conclusion cannot be generalized to a broader scale of laboratory. Third, some patients were given empirical antimicrobic treatment before sampling, we did not conduct a longitudinal assessment of the effects of antimicrobial administration before and after on mNGS and culture results.
In summary, here we present a retrospective study to fully evaluate the utility of mNGS for the diagnosis of suspected infectious diseases and the real-world clinical impact. More data are needed to determine the clinical impact of mNGS and to instruct which types of samples and patient populations are more favorable for mNGS testing. Nonetheless, this study suggests that mNGS testing may be practical in these scenarios: (1) as first-line detection tool for critically ill patients and immunosuppression patients who need to identify pathogens as soon as possible, (2) for identification of unknown, rare and atypical pathogens. (3) as a complementary test for culture-negative pathogens.