Detection of pathogenic microorganisms using metagenomic next generation sequencing for patients with suspected infection presenting in the emergency department

From 4 to 19% of patients admitted to an emergency medicine department (ED) present with clinical signs of infections [1,2]. The standard diagnostic approach includes blood cultures (BC), regarded as the gold standard for pathogen detection. However, true positive BC is found in only 2.6–6% of patients, highlighting the limitations of traditional culture-based methods [3,4]. Metagenomic next-generation sequencing (mNGS) has emerged as a promising diagnostic tool that may improve pathogen detection rates and facilitate targeted antimicrobial therapies. This study aimed to evaluate the sensitivity and specificity of mNGS compared to BC in ED patients with suspected infections, describe the microorganisms detected, and assess the potential impact of mNGS results on antimicrobial therapy. This retrospective study included patients admitted to the hospital from the ED between March 2022 and February 2023. Among 28 208 patients seen during the study period, 7501 were admitted to the hospital. Of these, 196 mNGS samples were obtained. Following the exclusion of 16 patients due to missing corresponding BC or loss to follow-up, a final cohort of 170 patients was analyzed. This cohort was selected based on the availability of both BC and mNGS diagnostics, representing a convenience sample from the broader patient population. All included patients were adults with suspected infections, and data on laboratory results, medical history, and clinical course were extracted from medical records. Ethical approval was obtained from the Medical Faculty of the University of Tübingen, Germany (No. 796/2022BO2). Fragments of freely circulating DNA are found in almost all body fluids [5]. Nucleic acid isolation, library preparation and sequencing, and bioinformatic analysis for mNGS from blood were performed using the DISQVER pipeline (Noscendo GmbH, Germany) [6–8]. 1.134 bacteria, 149 fungi, 78 DNA viruses, and 71 parasites could be distinguished. Results were reviewed for plausibility and potential impact on antimicrobial therapy by an expert panel. Plausible mNGS results were analyzed for potential antibiotic treatment adjustments (no change, escalation, deescalation, and stop). Diagnostic performance metrics were calculated, including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Statistical analyses were performed using established methodologies. The final cohort consisted of 170 patients. Demographic and clinical characteristics were similar between mNGS-positive (n = 83) and mNGS-negative (n = 87) groups. No specific selection criteria beyond the availability of both BC and mNGS diagnostics were applied. Pathogens were detected in 48.8% (83/170) of patients by mNGS, compared to 16.5% (28/170) by BC. Among the mNGS-positive cases, 63 had negative BC results, highlighting the added diagnostic value of mNGS. Conversely, eight cases had positive BC results but negative mNGS results. The diagnostic performance of mNGS was assessed. Sensitivity was calculated as 71.43% [95% confidence interval (CI): 51.33–86.78], and specificity was 55.6% (95% CI: 47.07–63.96; Fig. 1a). To assess the performance of mNGS diagnostics while incorporating additional clinical data as a composite reference standard, 13 implausible results were excluded (eight positive mNGS results and five negative mNGS results). Diagnostic performance metrics were calculated based on the remaining cases. The positive percent agreement was 90.48% (95% CI: 82.09–95.80), and the negative percent agreement was 94.05% (95% CI: 86.65–98.04). The PPV was 93.83% (95% CI: 86.63–98.04), and the NPV was 90.8% (95% CI: 83.60–95.03). Overall diagnostic accuracy was 92.6% (95% CI: 87.13–95.82; Fig. 1b). Additionally, the area under the curve for mNGS diagnostics and plausibility analysis was 92.4% (95% CI: 88.4–96.4%; Fig. 1c). mNGS detected a mean of 3.2 microorganisms per sample, compared to 1.1 microorganisms per BC. Pathogens identified by mNGS included 31 Gram-positive bacteria, 39 Enterobacterales, 12 nonfermenting Gram-negative bacteria, and 75 anaerobes. BC detected only a fraction of these, with no detection of anaerobes or nonfermenters. Expert panel analysis suggested that mNGS results could have led to therapy modifications in 61 cases (35.9%). These included 24 cases of deescalation, seven cases of escalation, and 30 cases where therapy could have been stopped. The study showed a 5.6-fold increase in pathogen detection rate per sample by mNGS (bacteria only). The failure to detect bacteria in BC was based on the limitations of culture-based detection methods, such as low sensitivity, prolonged processing times for slow-growing, fastidious, and anaerobic microorganisms, and overgrowth of the culture plate by the fastest-growing microorganism. Further, technical limitations of BC include contaminations from skin flora, difficulties maintaining an anaerobic environment, and ongoing antibiotic exposure [9–12]. Overall, 20.5% (17 out of 83) patients with positive mNGS results had an ongoing antibiotic therapy, compared to 7% (2 out of 28) patients with positive BC. Thus, an ongoing antibiotic suppressed positive BC results, but not mNGS. In 19 of 20 cases with both, positive mNGS and BC results, all bacteria detected by BC were found by mNGS diagnostics as well. However, eight BC were positive, but mNGS remained negative. Contaminations might have led to false positive BC results. The decision as to whether a positive BC is considered contaminated, or reflecting true bloodstream infection is often difficult. After the expert review, 12 cases from both groups were considered implausible and excluded. Very low cfDNA signals might have been reported as negative mNGS results due to low read counts that remained below the empirically set statistical threshold limit [7]. This burden accounts especially for blood with a high human host DNA background [13,14]. Technical analysis algorithms for mNGS diagnostics need to be constantly adapted based on a growing clinical database.Fig. 1: (a) Contingency table comparing positive and negative test results of mNGS vs. BC. (b) Contingency table comparing positive and negative test results of mNGS vs. the composite reference standard (CR) after antibiotic stewardship panel evaluation. (c) AUC curves of the NGS results and their corresponding plausibility. Positive percent agreement (PPA), negative percent agreement (NPA), positive, and negative predictive values (PPV and NPV). AUC, area under the curve; BC, blood cultures; mNGS, metagenomic next-generation sequencing; NPV, negative predictive value.The study demonstrated a substantial improvement in pathogen detection rates using mNGS compared to BC in ED patients with suspected infections, particularly for anaerobic and fastidious organisms. Sensitivity, specificity, PPV, and NPV metrics underscore its superior diagnostic performance. Importantly, mNGS results could have influenced antimicrobial therapy in 35.9% of cases. While mNGS represents a promising diagnostic advancement, expert interpretation remains essential, particularly in cases with low cfDNA signals or high human DNA background. Limitations include the inability of mNGS to provide antimicrobial susceptibility data, underscoring the continued necessity of culture-based methods. Integration into clinical workflows requires further evaluation. Future studies should address cost-effectiveness and real-time implementation challenges. Acknowledgements Conflicts of interest There are no conflicts of interest.