Diagnostic value of mNGS for CNSIs in hematopoietic stem cell transplant patients
Year: 2025
Type: article
Study design: Observational Study
Abstract: To The Editor: Hematopoietic stem cell transplantation (HSCT) is a potential restorative treatment for a range of conditions, including cancers, bone marrow failure syndromes, and various genetic disorders. However, central nervous system infections (CNSIs) are a common and serious complication among HSCT patients.1–3 Owing to immune suppression, HSCT recipients are particularly vulnerable to infections caused by various pathogens. Their clinical manifestations are often nonspecific and prompt diagnosis is difficult.4 Traditional diagnostic methods, such as cerebrospinal fluid (CSF) culture, polymerase chain reaction (PCR) testing, and imaging, while possessing certain sensitivities and specificities, often fail to provide effective microbiological diagnosis in a timely manner, especially in cases of low microbial load or mixed infections, given the diverse range of potential pathogens and the slow diagnostic speed.5 In recent years, the rapid development of genomic technologies, particularly metagenomic next-generation sequencing (mNGS), has provided new perspectives on pathogen detection. mNGS allows for the comprehensive analysis of all microbial DNA/RNA in the CSF without relying on predefined primers, offering high sensitivity and a broad spectrum of pathogen detection capabilities.6–8 Compared to traditional microbiological tests, mNGS has shown distinct advantages in identifying difficult-to-culture pathogens, mixed infections, and viral or fungal pathogens.9–11 Therefore, we believe that CSF mNGS has a great potential for the diagnosis and treatment of CNSIs in patients undergoing HSCT. In this study, we diagnosed and treated 121 HSCT patients with suspected CNSIs by combining clinical manifestations, traditional methods, and CSF mNGS results. We found that CSF mNGS had greater diagnostic advantages than conventional microbiological testing (CMT) and helped to adjust the treatment of CNSIs in patients with HSCT. This retrospective study included 121 patients who underwent HSCT at the Soochow Hopes Hematonosis Hospital between October 2018 and October 2023, all of whom presented with suspected CNSIs. There were 75 male and 46 female patients with a median age of 33 years (range, 6–65 years old). Among them, there were 52 patients with acute myeloid leukemia (AML), 35 with acute lymphoblastic leukemia (ALL), 13 with myelodysplastic syndromes (MDS), 7 with severe aplastic anemia (SAA), 5 with T cell lymphoblastic lymphoma (T-LBL), 4 with chronic myeloid leukemia (CML), 2 with hemophagocytic lymphohistiocytosis (HLH), 2 with chronic myelomonocytic leukemia (CMML), and 1 with aplastic anemia-paroxysmal nocturnal hemoglobinuria (AA-PNH). In total, 114 patients underwent haplo-HSCT, and 7 patients underwent MSD-HSCT. After transplantation, 56 patients developed graft-versus-host disease (GVHD) and 26 developed myelosuppression. The specific diagnosis and treatment procedures were as follows. For patients with suspected CNSIs, computed tomography (CT) and/or magnetic resonance imaging (MRI) and lumbar puncture were usually performed within 48 hours of symptom onset, if possible. A team composed of 2 to 3 clinicians specializing in infectious diseases made the final judgment according to the Expert Consensus on Diagnosis and Treatment of Central Nervous System Infection in Neurosurgery of China (2021 edition).12 The specific clinical diagnostic criteria were as follows: fever, symptoms of intracranial hypertension, turbid or purulent CSF, leukocytosis, glucose <2.2 mmol/L, and CSF glucose/blood glucose content ≤0.4.2 Based on the clinical diagnosis, positive microbial cultures were found in the specimen smears, drainage tube heads, implants, and CSF (excluding contamination and colonization). As most patients could not tolerate biopsy and fluid G/GM tests were not routine tests, our retrospective study only considered clinical symptoms (such as fever and intracranial hypertension), abnormal brain CT/MRI, routine and biochemical abnormalities of the CSF, and pathogens detected in the CSF. Accordingly, their treatments were adjusted. A total of 69 patients (57.02%) were diagnosed with CNSIs. In the process of diagnosis, mNGS exhibited outstanding efficacy, with sensitivity of 94.20% (95% confidence interval [CI]: 85.82–98.40), specificity of 78.85% (95% CI: 65.30–88.94), positive predictive value (PPV) of 85.53% (95% CI: 75.58–92.55), negative predictive value (NPV) of 91.11% (95% CI: 78.78–97.52), and accuracy of 87.60% (95% CI: 80.38–92.89). Meanwhile, CMT had sensitivity of 5.80% (95% CI: 1.60–14.18), specificity of 100.00% (95% CI: 93.15–100), PPV of 100.00% (95% CI: 39.76–100), NPV of 44.44% (95% CI: 35.26–53.92), and an accuracy of 46.28% (95% CI: 37.17–55.57). We found that mNGS had greater diagnostic efficiency in terms of sensitivity, NPV, and accuracy than CMT (p < 0.001). We also analyzed the diagnostic efficacy of mNGS for different types of pathogens. For viral infections, mNGS achieved sensitivity of 96.23% (95% CI: 87.02–99.54), specificity of 85.29% (95% CI: 74.61–92.71), PPV of 83.61% (95% CI: 71.91–91.85), NPV of 96.67% (95% CI: 88.47–99.59), and accuracy of 90.08% (95% CI: 83.32–94.77). For bacterial infections, the sensitivity was 80.00% (95% CI: 28.36–99.49), specificity was 85.34% (95% CI: 77.58–91.22), PPV was 19.05% (95% CI: 5.45–41.91), NPV was 99.00% (95% CI: 94.55–99.97), and accuracy was 85.12% (95% CI: 77.51–90.94). For fungal infections, mNGS demonstrated sensitivity of 100.00% (95% CI: 59.04–100), specificity of 92.98% (95% CI: 86.64–96.92), PPV of 46.67% (95% CI: 21.27–73.41), NPV of 100.00% (95% CI: 96.58–100), and accuracy of 93.39% (95% CI: 87.39–97.10). These findings show that mNGS has good efficacy for the diagnosis of different pathogens. In the confirmed CNSIs cases, the distribution of pathogens was as follows: viral infections in 52 cases (52/69, 75.36%), fungal infections in 6 cases (6/69, 8.70%), bacterial infections in 5 cases (5/69, 7.25%), tuberculosis infections in 4 cases (4/69, 5.80%), toxoplasmosis in 1 case (1/69, 1.45%), and mixed fungal and viral infections in 1 case (1/69, 1.45%). Figure 1 shows the organism frequencies detected using CSF mNGS in clinically adjudicated cases. The pathogens with relatively high detection rates were Epstein-Barr Virus (EBV) (25 cases), cytomegalovirus (CMV) (11 cases), human herpes viruses (HHV-6) (10 cases), Mycobacterium tuberculosis complex (MTBC) (4 cases), and Torque Teno Virus (TTV) (4 cases). Consistent with reports of infrequent CNS detection13 and isolated observations suggesting occasional intra-CNS replication,14 we report CSF TTV only as a hypothesis-generating finding rather than an etiologic diagnosis. Based on the mNGS findings, antimicrobial regimens were modified in 32 cases (26.45%), with 26 (81.25%) treatment escalations (including 3 rituximab treatments) and 6 (18.75%) treatment de-escalations. In 89 (73.55%) patients, no modifications were made, as conventional treatment was considered adequate based on the mNGS results and clinical response. All the patients who received escalated therapy showed clinical improvements. These results demonstrate the utility of mNGS in optimizing antimicrobial therapy.Figure 1.: The frequencies of pathogens detected using mNGS in clinical diagnosis. mNGS = metagenomic next-generation sequencing.Our study evaluated mNGS for diagnosing CNSIs in patients undergoing HSCT and confirmed that it has better application potential than CMT in immunocompromised patients. For different pathogens, mNGS showed better performance in diagnosing viral and fungal pathogens, which supports the results of previous studies.15–17 The low sensitivity for bacterial infections suggests the need for further optimization to detect low-burden or mixed bacterial infections.16–19 Notably, mNGS is sensitive to uncommon fungi (eg, Rhizopus, Aspergillus fumigatus), which are often missed by conventional tests.20,21 These findings emphasize the broader pathogen detection of mNGS, especially in low-burden or complex CNSIs, distinguishing our study from previous studies that focused mainly on viral and bacterial infections. Additionally, consistent with previous studies, mNGS can assist in adjusting antimicrobial therapy.16,17,22,23 However, its specificity for mixed infections was low, suggesting that protocol optimization might be needed. Future research should refine the diagnostic criteria, improve bioinformatics tools, explore the integration of mNGS with other diagnostic methods, and improve its cost-effectiveness.24 Our study had some limitations, including its retrospective and single-center design, which may limit the generalizability of our findings. A larger multicenter cohort is needed to validate the efficacy of mNGS. Moreover, the limited availability of CNS tissue might permit misclassification between EBV CNSI and CNS-PTLD in individual cases, which has also been acknowledged as a key limitation. In summary, our study supports the use of mNGS as a promising tool for diagnosing CNSIs in immunocompromised patients after HSCT, with significant potential for improving diagnostic accuracy and patient management. Further optimization and larger studies are required to fully integrate mNGS into clinical practice. ACKNOWLEDGMENTS The authors thank all the authors for their guidance, inspiration, or discussion of the topic, ideas, perspectives, and arguments of the paper. We also appreciate the mNGS sequencing and statistical analysis provided by Dinfectome Inc., Nanjing, China.
Source: Blood Science
Authors: Yujie Ji, Ziyan Jin, Ming Song, Ying Wang, Xiong Zeng, Xiao Ma, Dan Liu
Language: English
Subfield: Clinical Biochemistry
Domain: Life Sciences
Open Access status: diamond
OpenAlex ID: W4417505946