Journal Reading Infeksi 2

PPDS : Desak Sembah Laksmi Dewi

Pembimbing : Dr. dr I Nyoman Wande, Sp.PK(K)
Diagnostic Microbiology and Infectious Disease 107 (2023) 116025 Tanggal presentasi : Kamis, 28 Desember 2023

Contents lists available at ScienceDirect

Diagnostic Microbiology and Infectious Disease

jo ur na l h om ep ag e : w w w . el s e v ie r . c om /l oc at e / di ag m i c r o bi o

Comparative performance of microbiological methods for the detection

of tuberculous meningitis pathogens in cerebrospinal fluid
Yuling Lin, Weili Zhang, Ying Xiong, Yue Wang, Qiuju Yu, Ying Ma*, Yi Xie**
Department of Laboratory Medicine, West China Hospital, Sichuan University, Chengdu, China

A R T I C L E I N F O A B S T R A C T

Article history: The aim of this study was to comprehensively evaluate metagenomic next-generation sequencing (mNGS),
Received 2 May 2023 Acid-fast bacillus stain (AFB), MGIT960 culture, polymerase-chain-reaction (PCR), and Xpert MTB/RIF in the
Revised in revised form 11 June 2023 diagnosis of tuberculous meningitis (TBM). A cohort of 280 patients who presented with suspected TBM (ie,
Accepted 9 July 2023
headache or altered mental status with clinical signs of meningism) were analyzed. The sensitivities of the 5
Available online 15 July 2023
assays for the diagnosis of TBM ranged from 10.0% to 70.0%. The AFB had the lowest sensitivity of 10.0% (0.5-
45.9), while mNGS and PCR had the highest sensitivity, both at 70.0% (35.4-91.9). mNGS demonstrated a dis-
Keywords:
tinct advantage in identifying a wider array of pathogens, including viruses, in CSF samples. PCR was a cost-
Mycobacterium tuberculosis
Metagenomic Next-Generation Sequencing
effective option with excellent sensitivity and specificity. However, no single method was statistically signifi-
Tuberculosis meningitis cantly better than any other in the diagnosis of TBM. New diagnostic techniques are urgently needed for the
Cerebrospinal fluid independent, rapid and accurate detection of Mycobacterium tuberculosis to guide the diagnosis of TBM.
Xpert MTB/RIF © 2023 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/)

1. Introduction detection rate, making it less suitable for early diagnosis. Acid-fast
bacillus stain (AFB) of CSF is a cheap and widely available tech-
Tuberculosis (TB) is a serious infectious disease that has been the nique for rapid diagnosis of TBM, but is often insensitive [5]. In
leading cause of death from a single communicable agent until the recent years, emerging techniques have provided promising results
coronavirus (COVID-19) pandemic [1]. Among the various manifesta- for the diagnosis of TBM. One such technique is Xpert MTB/RIF, a
tions of TB, tuberculous meningitis (TBM) is one of the most devastat- rapid and sensitive nucleic acid amplification test (NAAT) that can
ing syndromes, which causes fatality or disability of nearly half of the also detect rifampicin susceptibility [6]. By targeting the 81 bp
affected people [2]. Early diagnosis and timely treatment of TBM are rifampicin resistance core region (RRDR) of the rpoB gene, Xpert
critical factors that significantly impact patient survival. However, MTB/RIF can identify mutations associated with rifampicin resis-
many patients are initially administered empirical antituberculous tance, providing valuable information for both TB diagnosis and
therapy due to insensitive and time-consuming diagnostic methods drug resistance detection. A positive Xpert result is a strong indica-
required for disease confirmation [3]. tion that the patient had M. tuberculosis infection and was resistant
TBM is the most severe form of TB, affecting all ages, but young to rifampicin. A negative result does not rule out the possibility
children and people with untreated HIV are particularly vulnerable. that the patient was not infected and follow-up TB-DNA, imaging
Although TBM accounts for about 1% of new TB cases, once and other methods are need to confirm the diagnosis [7]. Hence,
infected it causes death or disability in nearly half of all patients the limitations of current diagnostic methods, such as prolonged
[4]. The incidence of TBM is directly related to the prevalence of TAT, low detection rates, and potential empirical antituberculous
pulmonary TB [2]. Therefore, early detection of M. tuberculosis is an therapy, which can lead to antibiotic resistance and increased
important tool in the control of disease progression. Conventional healthcare costs, necessitate the development of innovative diag-
clinical microbiological testing using cerebrospinal fluid (CSF) is the nostic methods [3]. A method that can overcome the limitations of
primary approach, but it has limitations. M. tuberculosis culture is current tests, accurately diagnose TBM and guide clinical manage-
considered the gold standard for diagnosing TBM, but has a pro- ment is therefore urgently needed.
longed turn-around-time (TAT) of at least 2 weeks and a low Metagenomic next-generation sequencing (mNGS) is a promising
diagnostic technology, which has been widely used in the detection
of infectious pathogens in recent years [8]. Compared with traditional
* Corresponding authors. Tel.: 18980601863. Sanger sequencing, which requires preamplification of a known
** Corresponding author. Tel.: 189 8060 5782. sequence and low-diversity sample, mNGS can identify a wide range
E-mail addresses: [email protected] (Y. Ma), [email protected] (Y. Xie).

https://doi.org/10.1016/j.diagmicrobio.2023.116025
0732-8893/© 2023 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/)
2 Y. Lin et al. / Diagnostic Microbiology and Infectious Disease 107 (2023) 116025

of pathogenic microorganisms in a single assay without bias [9]. This 2.3. Acid-fast bacillus stain
breakthrough technology has revolutionized the field of infectious
disease diagnostics by enabling comprehensive and unbiased patho- Briefly, 2 to 4 mL of CSF was added to a disposable centrifuge tube,
gen detection [10]. A recent large-scale, multicenter, prospective the supernatant was discarded after centrifuging at 3500 rpm for 5
study with CSF samples from infectious meningitis and encephalitis minutes, and the precipitate was taken for Acid-fast bacillus stain
evaluated the performance and efficacy of mNGS in comparison with (BA4015, BaSO, Guangdong, China) and smear microscopy.
routine diagnostic testing. The results showed that conventional
microbiologic testing is often inadequate to detect all pathogens. 2.4. MGIT960 culture
Hence, mNGS of CSF may represent a potential breakthrough in the
diagnosis of meningoencephalitis, which could guide earlier or Briefly, 500 mL of CSF was added to MGIT culture tubes containing
more targeted treatment of nerve-invasive infections and prompt 800 mL of MGIT supplement. The tubes were then mixed and incu-
identification and treatment of non-infectious diseases [11]. Given its bated in the MGIT960 apparatus for 6 weeks. After the machine auto-
capabilities, mNGS holds great promise in the rapid diagnosis of TBM matically determined a positive result, the BACTEC MGIT SIRE kit
in CSF. (Becton, Dickinson) was used to test all culture-positive strains for
The current landscape of research comparing the performance of susceptibility to isoniazid, rifampin, ethambutol, and streptomycin
clinical assays for TBM is dominated by small-scale studies, leaving a according to the manufacturer’s instructions.
critical gap in the understanding of their efficacy. To address this
limitation, we conducted a comprehensive retrospective study at 2.5. PCR
West China Hospital of Sichuan University between April 2020 and
April 2021, involving 280 adult patients with clinically suspected Briefly, 800 mL of CSF and 60 mL of proteinase K were added to
TBM (headache for >3 days or altered mental status [Glasgow Coma 5 mL sample tubes. Then concert@HF 24 (CONCERT, Fujian, China)
Scale <15] with clinical signs of meningism at examination—ie, neck was used to extract nucleic acid automatically. Finally, 100 mL of
stiffness or Kernig’s sign) who were not infected with HIV [5]. The nucleic acid was collected and frozen at -20°C for storage. The PCR
aim of this study was to rigorously assess the diagnostic efficacy of premix was prepared as follows: reaction solution 37.8 mL per well,
mNGS and other traditional microbiological methodologies, including Taqase 0.2 mL and UNGase 0.06 mL. Subsequently, sample 2 mL was
AFB, MGIT960 culture, PCR and Xpert MTB/RIF in the diagnosis of added to each well. Amplification was performed using a LightCycler
TBM, using a standardized clinical case definition of TBM as the gold 480 with the following reaction conditions: 37°C 3 min 1 cycle; 94°C
standard for the study. 1 min 1 cycle; then 95°C 5 s, 60°C 30 s, 40 cycles.

2.6. Xpert-MTB/RIF
2. Materials and methods
Detection of M. tuberculosis was performed using the Xpert MTB/
2.1. Study design and participants
RIF (Cepheid, CA) assay according to manufacturer’s recommenda-
tion. According to the original volume of the CSF, 1-2 mL of 4% NaOH
We conducted a retrospective study to evaluate the performances
was added to the pretreatment tube, the sample was mixed thor-
of mNGS, AFB, MGIT960, PCR, and Xpert MTB/RIF for the diagnosis of
oughly and allowed to liquefy for 20 minutes at room temperature.
TBM. Patients aged 18 years or older with suspected TBM based on
After inverting the pretreatment tube 10-20 times, 2 mL of liquid was
clinical symptoms and signs were included in this study [12]. Patients
added with a sterile pipette to the assay cartridge. The test cartridge
with contraindications to lumbar puncture, unsigned informed con-
was loaded into the GeneXpertÒ System (Cepheid, CA) for testing and
sent, and HIV-coinfection were excluded. The study was approved by
the instrument automatically displayed the result after 2.5 hours.
the Ethics Committee of West China Hospital, Sichuan University
Xpert MTB/RIF assay results were determined by measuring the
(approval number: 2022-270).
fluorescence signal and the built-in calculation algorithm of the
instrument (invalid [QC failure]; M. tuberculosis not detected; M.
2.2. Diagnostic classification tuberculosis detected and rifampicin resistance [rifampicin resistance
detected, not detected or indeterminate]). Based on the Ct value of M.
Previous studies used different definitions of tuberculous men- tuberculosis in the sample, the results of the semiquantitative quanti-
ingitis [13−19]. To make it easier to compare the results of differ- fication of M. tuberculosis was shown as high (Ct value <16), medium
ent studies and to optimize the use of existing data, Marais et al. (Ct value of 16-22), low (Ct value of 22-28) or very low (Ct value
developed standardized diagnostic criteria for tuberculous menin- >28).
gitis [12]. According to the uniform clinical case definition, TBM
was divided into 3 categories: definite, probable and possible, 2.7. mNGS
based on the clinical, laboratory and radiological findings. The
diagnosis of TBM was considered to be definite when a patient The MGISEQ-2000 (BGI, Guangdong, China) gene sequencer was
presenting with symptoms or signs consistent with meningitis ful- used for sequencing reactions, and the pathogen sequences were
fils 1 or more of the following criteria: AFB seen in the CSF, M. compared with Mycobacterium TB (GenBank ID: GCF_001708265.1).
tuberculosis cultured from CSF, or a CSF M. tuberculosis positive Three main steps were used: (1) Sample processing: 1.5 to 3 mL of
commercial nucleic acid amplification test. If all the above criteria patient’s CSF was collected in accordance with standard sample col-
were negative, probable or possible TBM was determined accord- lection procedures. First, 7.2 mL Lyticase (RT410-TA, TIANGEN BIO-
ing to a diagnostic scoring system in the uniform clinical case defi- TECH, Beijing, China) was added to 600 mL of sample for enzyme
nition [12]. Specifically, the patient’s scores in each section (clinical breaking reaction, followed by 250 mL of 0.5 mm glass beads for
criteria, CSF criteria, cerebral imaging criteria, evidence of TB else- physical breaking, and finally the sample was mixed and shaken.
where) were added, a patient may be classified as probable TBM if Then 300 mL of sample was taken for DNA extraction according to
imaging was available and the score was 12 or more, or if imaging the procedure of TIANamp Micro DNA kit (DP316, TIANGEN BIOTECH,
was not available and the score was 10 or more. A patient may be Beijing, China). (2) Library construction and sequencing: The
classified as possible TBM if imaging was available and the score of extracted nucleic acids were first subjected to enzymatic fragmenta-
6 to 11, or if imaging was not available and the score of 6 to 9. tion, end-repair, adapter-ligation and PCR adapter-ligation for
 Y. Lin et al. / Diagnostic Microbiology and Infectious Disease 107 (2023) 116025 3

construction of DNA libraries. The DNA libraries were quantified by clinical research, a total of 82 cases were diagnosed as non-TBM,
Agilent 2100 Bioanalyzer and their concentrations were quantified while 198 cases were diagnosed as TBM. Among the TBM cases, 10
by Qubit dsDNA HS Assay kit (Thermo Fisher Scientific Inc). Finally, were classified as definite, 18 as probable, and 170 as possible,
the DNA libraries were sequenced using MGISEQ-2000 [20]. (3) Bio- according to the established rules (Fig. 1).
informatics analysis: Low-quality sequences were first removed to
generate high-quality sequencing data. Then, some of the high-qual-
ity data were excluded by comparison of Burrows−Wheeler Align- 3.2. Patient characteristics
ment [21] with human reference genome (hg19) matched. The
remaining data, consisting of bacteria, fungi, viruses, and parasites, Between April 1, 2020, and April 1, 2021, a total of 605 HIV-unin-
were classified by simultaneous alignment to the Pathogens Metage- fected individuals underwent screening. Among them, 325 did not
nomics Database (PMDB). The classification reference databases were meet the criteria for enrollment, and 280 patients were finally
downloaded from NCBI (ftp://ftp.ncbi.nlm.nih.gov/genomes/). Posi- included and underwent testing. The average age of the 280 patients
tive results for the detection of M. tuberculosis complex matching was 48.4 years; 76 patients (27.1%) were over the age of 60 years;
sequence (NC_00962.3). 63.5% of the patients were male (Table 1). A comparison of lateral
findings revealed that as the identification of TBM became more
accurate, the patients tended to be older and predominantly male
3. Results with suspected TBM. Furthermore, there was a gradual decrease in
the number of lymphocytes in the CSF, along with a lower CSF-blood
3.1. Flow chart of patient selection in this study glucose ratio. We also performed a statistical analysis of 198 TBM and
82 non-TBM using the independent samples MannWhitney test,
After considering the patient’s clinical features and relevant labo- which showed that CSF routine, cerebrospinal fluid biochemistry and
ratory tests, and applying a standardized case definition for TBM in CSF-blood glucose ratio were significantly different from each other.

Fig. 1. Flow chart.

4 Y. Lin et al. / Diagnostic Microbiology and Infectious Disease 107 (2023) 116025

Table 1
Baseline characteristics.

Characteristic Value (280 participants) 198 TBM Value (82 non-TBM) P-Value (TBM vs non-TBM)

Value (170 possible) Value (18 probable) Value (10 definite)

Age
Mean - years 48.4 48.6 51.4 54.5 46.6 0.21
Distribution - no. (%)
19-25 years 26 (9.3) 17 (10.0) 2 (11.1) 1 (10.0) 6 (7.3) 0.42
26-40 years 84 (30.0) 46 (27.1) 5 (27.8) 1 (10.0) 32 (39.0) 0.49
41-60 years 94 (33.6) 62 (36.5) 3 (16.7) 5 (50.0) 24 (29.3) 0.00
>60 years 76 (27.1) 45 (26.5) 8 (44.4) 3 (30.0) 20 (24.4) 0.13
Male sex−no. (%) 178 (63.5) 111 (65.3) 11 (61.1) 7 (70.0) 49 (59.8) 0.58
CSF routine
CSF white cell count (per mL) 211.1 (108.1−314.1) 240.7 (92.0−389.4) 86.3 (54.4−118.3) 132.0 (28.6−235.4) 187.1 (13.9−360.3) 0.00
CSF lymphocytes 32.5 (27.6−37.3) 39.8 (33.3−46.4) 60.3 (42.4−78.2) 26.2 (2.9−49.5) 12.0 (5.7−18.2) 0.00
Cerebrospinal fluid biochemistry
CSF protein (g/L) 1.5 (1.0−2.0) 1.7 (0.9−2.6) 1.5 (1.2−1.9) 2.6 (1.5−2.8) 0.8 (0.6−1.0) 0.00
CSF glucose (mmol/L) 3.5 (3.3−3.7) 3.4 (3.1−3.6) 3.0 (2.3−3.8) 3.3 (1.2−5.5) 3.9 (3.5−4.2) 0.00
CSF Cl (mmol/L) 122.5 (121.6−123.5) 121.7 (120.5−122.9) 121.5 (116.7−126.3) 119.7 (107.6−131.8) 124.8 (123.5−126.0) 0.00
CSF-blood glucose ratio 0.6 (0.5−0.6) 0.5 (0.5−0.6) 0.4 (0.3−0.5) 0.3 (0.2−0.4) 0.6 (0.6−0.7) 0.00

3.3. Comparative performance of 5 microbiological methods for the positive by all methods, indicating the variability and complementary
diagnosis of tuberculous meningitis nature of these diagnostic techniques in identifying TBM.

The diagnostic sensitivities of 5 methods for detecting TBM in the 3.5. Distribution of pathogenic microorganisms in CSF
280 participants with a definitive diagnosis were as follows: 7.6% (95%
CI 4.5-12.4) for mNGS, 0.6% (95% CI 0.0-3.7) for AFB, 5.0% (95% CI 0.9- A total of 250 pathogenic bacteria were found in CSF from 151
18.2) for MGIT960, 6.1% (95% CI 2.7-12.6) for PCR and 8.8% (95% CI 3.9- patients (Fig. 3A). Among them, the top 5 pathogens identified were
17.7) for Xpert MTB/RIF, when compared to the reference standard of human herpesvirus type 4 (32 cases), M. tuberculosis (20 cases), human
definite, probable and possible TBM. The specificity of all methods, herpesvirus type 1 (17 cases), Klebsiella pneumoniae (13 cases), and
except mNGS, was 100% (Table 2). Although the sensitivity of mNGS Candida parapsilosis (12 cases). In contrast, conventional microbiolog-
was relatively better than the other 4 methods when using definite and ical methods such as culture and smear microscopy only identified 32
probable TBM as a reference standard, it did not reach 100%. In cases of pathogenic strains in the same set of 280 CSF samples, with the high-
definite TBM, both mNGS and PCR showed the same sensitivity. est detection rate of Cryptococcus neoformans in 16 cases (Fig. 3B).
Upon comparison of the 2 sets of data, the match rate was 12.8%,
3.4. Overlap of commonly used clinical methods highlighting the higher sensitivity and broader pathogen detection
capability of mNGS compared to traditional microbiological methods.
Among the CSF samples obtained from the included patients, a
total of 8 cases were diagnosed with TBM by both mNGS and routine 4. Discussion
microbiologic testing methods, including AFB, MGIT960, PCR, and
Xpert MTB/RIF. Additionally, 12 cases were diagnosed with TBM In this study, no single test performed significantly better than the
solely by mNGS. The observation of the allocation and overlap of pos- others for the diagnosis of TBM in CSF. However, mNGS could detect a
itive CSF samples by mNGS, AFB, MGIT960, PCR, and Xpert MTB/RIF larger number of potential pathogens compared to routine microbio-
revealed that out of the 20 cases that were positive by mNGS, 6 were logic testing. While CSF bacterial culture and smear microscopy
positive by Xpert MTB/RIF, 6 were also positive by PCR, 1 by identified a total of 30 pathogens, mNGS identified a total of 85
MGIT960, and 1 by AFB (Fig. 2). Notably, none of the cases were pathogenic microorganisms. These findings highlight the enhanced

Table 2
Diagnostic performance of 5 commonly used methods against the uniform clinical case definition.

mNGS AFB MGIT960 PCR Xpert MTB/RIF

Reference standard: definite, probable, and possible tuberculous meningitis

Positive tests 15/198 1/173 2/40 7/115 7/80
Sensitivity 7.6% (4.5−12.4) 0.6% (0.0−3.7) 5.0% (0.9−18.2) 6.1% (2.7−12.6) 8.8% (3.9−17.7)
Specificity 93.9% (85.7−97.7) 100.0% (93.8−100.0) 100.0% (46.3−100.0) 100.0% (85.9−100.0) 100.0% (69.9−100.0)
PPV 75.0% (50.6−90.4) 100.0% (5.5−100.0) 100.0% (19.8−100.0) 100.0% (56.1−100.0) 100.0% (56.1−100.0)
NPV 25.0% (9.6−49.4) 29.8% (24.2−36.0) 11.6% (4.3−25.9) 21.7% (15.4−29.7) 14.1% (7.8−23.8)
Reference standard: definite, probable tuberculous meningitis
Positive tests 11/28 1/27 2/13 7/21 7/19
Sensitivity 39.3% (22.1−59.3) 3.7% (0.2−20.9) 13.3% (2.3−41.6) 33.3% (15.5−56.9) 36.8% (17.2−61.4)
Specificity 96.4% (93.1−98.2) 100.0% (97.9−100.0) 100.0% (86.7−100.0) 100.0% (96.3−100.0) 100.0% (93.8−100.0)
PPV 55.0% (32.0−76.2) 100.0% (5.5−100.0) 100.0% (19.8−100.0) 100.0% (56.1−100.0) 100.0% (56.1−100.0)
NPV 93.5% (89.6−96.0) 89.4% (84.7−92.8) 71.1% (55.5−83.2) 89.9% (83.3−94.1) 85.9% (76.2−92.2)
Reference standard: definite tuberculous meningitis
Positive tests 7/10 1/10 2/5 7/10 6/9
Sensitivity 70.0% (35.4−91.9) 10.0% (0.5−45.9) 40.0% (7.3−83.0) 70.0% (35.4−91.9) 66.7% (30.9−91.0)
Specificity 95.2% (91.7−97.3) 100.0% (98.0−100.0) 100.0% (89.1−100.0) 100.0% (96.6−100.0) 98.9% (92.5−99.9)
PPV 35.0% (16.3−59.1) 100.0% (5.5−100.0) 100.0% (19.8−100.0) 100.0% (56.1−100.0) 85.7% (42.0−99.2)
NPV 98.8% (96.4−99.7) 96.3% (92.9−98.2) 93.0% (79.9−98.2) 97.8% (93.3−99.4) 96.5% (89.3−99.1)
 Y. Lin et al. / Diagnostic Microbiology and Infectious Disease 107 (2023) 116025 5

Fig. 2. Venn diagram of positive diagnostic tests.

capability of mNGS in providing valuable clinical information beyond results showed no significant superiority of mNGS compared to tradi-
traditional testing methods. tional laboratory testing methods (mNGS vs. AFB P = 0.03, vs.
The sensitivity of mNGS in this study was found to be up to 70%, MGIT960 P = 1, vs. PCR P = 1, vs. Xpert MTB/RIF P = 1), indicating sig-
with a specificity of 96.4%. Previous studies have reported varying nificant heterogeneity across studies for the same indicator. Several
sensitivity (ranging from 27%-84%) and specificity (ranging from factors could contribute to these differences. First, different sequenc-
96%-100%) for mNGS in the diagnosis of TBM [22]. Many of these ing methods were used in each study, resulting in variations in the
studies have concluded that mNGS is superior to traditional microbi- depth of sequencing and the number of genes detected. For example,
ological methods for early diagnosis of TBM [11]. However, our study some studies used the fast-running model on the Illumina HiSeq

Fig. 3. mNGS versus other microbiological methods for the detection of pathogenic microorganisms.
6 Y. Lin et al. / Diagnostic Microbiology and Infectious Disease 107 (2023) 116025

instrument with a depth of 5-10 million single-end, 140-base-pair and mNGS, and finally compare the sensitivity and specificity of the
reads per patient, while others used the BGISEQ-500/50 platform same assay for different types of samples.
with an average depth of 20 million reads per sample [11,23]. In the- In summary, while 2 new technologies, mNGS and Xpert MTB/RIF,
ory, increasing the sequencing depth can help improve the sensitivity show promising results, they do not completely solve the problem of
and specificity of the assay, but there is an upper limit to this cover- laboratory diagnosis of TBM. Currently, there is no single method to
age. Once the sequencing depth reaches a certain level, there are independently diagnose TBM. Among the tested methods, PCR per-
enough reads that each base of the target gene to be detected has formed best when using definite TBM as a reference standard. Both
been read at least once or even more times, and increasing the mNGS and Xpert MTB/RIF have similar test performances, but a nega-
sequencing depth at this point will not significantly improve sensitiv- tive test does not exclude TBM and should be used in conjunction
ity and specificity [24]. Due to the low total amount of DNA extracted with other conventional methods for accurate diagnosis. AFB staining
from CSF, good results can be obtained with different sequencing has limited value in diagnosing TBM. Although MGIT960 is time-con-
platforms for the diagnosis of TBM when a sequencing depth of suming and moderately effective, it can provide subsequent drug sus-
100 million is achieved. Second, our study included a wider range of ceptibility results if the culture is positive. Therefore, there is an
patients compared to other studies, beyond Mycobacterium TB-con- urgent need for developing new laboratory testing technologies for
firmed patients only. While this may have made the results more TBM.
clinically relevant and instructive, stricter inclusion criteria for
patients tend to result in better assay performance. Lastly, sample
Authors’ contributions
quality, particularly the low levels of pathogenic bacteria in CSF, can
also impact test results. While the sensitivity and specificity of mNGS
Yuling Lin was involved in research design, performed the data anal-
in our study were not as high as expected, several factors may have
ysis and interpretation, drafted and revised the manuscript. Weili Zhang
contributed to these results, including differences in sequencing
was responsible for collating data. Ying Xiong, Yue Wang and Qiuju Yu
methods, patient inclusion criteria, and sample quality. Further
were responsible for doing the test and recruiting patients. Yi Xie and
research and optimization of mNGS protocols, including standardized
Ying Ma conceptualized the study, supervised the research design, data
pretreatment steps for CSF samples, validation of specificity in HIV
analysis, and interpretation; and critically reviewed and revised the
coinfection cases, and comparison with other diagnostic methods,
manuscript. All authors read and approved the final manuscript.
are warranted to improve the laboratory testing of TBM.
Strengths of this study included its large sample size, retrospec-
tive study design and prior to the initiation of anti-TB treatment. A Declaration of Competing Interest
graph evaluating the performance of mNGS with conventional meth-
ods indicates that the specificity of mNGS is not always 100%, mean- The authors declare that they have no competing interests.
ing that nondetection of the target pathogen does not necessarily
exclude disease. The sensitivity of all assays increased as the diagno- Funding
sis became more accurate, with molecular assays (mNGS, PCR, Xpert
MTB/RIF) showing a greater increase in sensitivity compared to AFB The work was supported by the Science and Technology Depart-
and MGIT960, with PCR demonstrating the greatest increase in sensi- ment Project of Sichuan, China (2020YFS0555); and the Special Foun-
tivity. Therefore, PCR may be the best diagnostic method among the dation for National Science and Technology Basic Research Program
5 methods for detecting TBM, considering sensitivity, specificity, and of China (2019FY101200).
cost of the test.
Although traditional microbiological methods may be more cost-
effective for detecting TBM, the application of mNGS is not limited to Ethics declarations
TBM alone [25]. It has the ability to detect all pathogens in a sample
simultaneously, facilitating simultaneous diagnosis of multiple infec- Ethics approval and consent to participate. This prospective study
tious diseases [26]. A comparison of pathogenic distribution between was approved by the Ethics Committee of West China Hospital, Sich-
mNGS and conventional laboratory tests revealed that herpesvirus IV, uan University (approval number: 2022-270). Each participant or
also known as EBV, was the most frequently detected pathogen in CSF. their legal representatives gave written informed consent before
This finding may be related to latent EBV infection or reactivation. enrollment.
Additionally, mNGS showed significantly better detection rates for
fungi compared to traditional methods, with improved detection rates References
for Candida parapsilosis and Aspergillus fumigatus, which can be valuable
in the diagnosis of fungal meningitis. Therefore, clinical selection of the [1] Global tuberculosis report 2022. Geneva: World Health Organization; 2022
appropriate test should be based on the patient’s condition, and mNGS Licence: CC BY-NC-SA 3.0 IGO.
[2] Thwaites GE, van Toorn R, Schoeman J. Tuberculous meningitis: more questions,
may be considered as an additional diagnostic tool when necessary. still too few answers. Lancet Neurol 2013;12:999–1010. doi: 10.1016/S1474-
In this experiment, the detection performance of Xpert Ultra was 4422(13)70168-6.
not assessed, and subsequent studies could address this research gap [3] on behalf of the Tuberculous Meningitis International Research Consortium, Wil-
kinson RJ, Rohlwink U, Misra UK, van Crevel R, Mai NTH, et al. Tuberculous men-
by using CSF. Furthermore, based on the conclusions and experimen- ingitis. Nat Rev Neurol 2017;13:581–98. doi: 10.1038/nrneurol.2017.120.
tal design of this study, 3 aspects could be optimized in future [4] Huynh J, Donovan J, Phu NH, Nghia HDT, Thuong NTT, Thwaites GE. Tuberculous
research. First, due to the low bacterial content of CSF and the suscep- meningitis: progress and remaining questions. Lancet Neurol 2022;21:450–64.
doi: 10.1016/S1474-4422(21)00435-X.
tibility of mNGS results to various factors, standardized pretreatment
[5] Cresswell FV, Tugume L, Bahr NC, Kwizera R, Bangdiwala AS, Musubire AK, et al.
steps for CSF samples should be developed to minimize the occur- Xpert MTB/RIF Ultra for the diagnosis of HIV-associated tuberculous meningitis: a
rence of false positives and false negatives. Second, the specificity prospective validation study. Lancet Infect Dis 2020;20:308–17. doi: 10.1016/
S1473-3099(19)30550-X.
estimates were primarily derived from HIV-uninfected individuals,
[6] Opota O, Mazza-Stalder J, Greub G, Jaton K. The rapid molecular test Xpert MTB/
and further data are necessary to validate the specificity of mNGS for RIF ultra: towards improved tuberculosis diagnosis and rifampicin resistance
HIV coinfection. Finally, this study only analyzed CSF, and subsequent detection. Clin Microbiol Infect 2019;25:1370–6. doi: 10.1016/j.cmi.2019.03.021.
studies could add different categories of samples, such as sputum, [7] Donovan J, Thu DDA, Phu NH, Dung VTM, Quang TP, Nghia HDT, et al. Xpert MTB/
RIF Ultra versus Xpert MTB/RIF for the diagnosis of tuberculous meningitis: a pro-
alveolar lavage fluid and puncture fluid, by testing them simulta- spective, randomised, diagnostic accuracy study. Lancet Infect Dis 2020;20:299–
neously for M. tuberculosis using AFB, MGIT960, Xpert MTB/RIF, PCR 307. doi: 10.1016/S1473-3099(19)30649-8.
 Y. Lin et al. / Diagnostic Microbiology and Infectious Disease 107 (2023) 116025 7

[8] Ramachandran PS, Wilson MR. Metagenomics for neurological infections— [17] Thwaites GE, Quy HT, Hien NQ, Lan NTN, Tuan VN, Mai PP, et al. Dexamethasone
expanding our imagination. Nat Rev Neurol 2020;16:547–56. doi: 10.1038/ for the treatment of tuberculous meningitis in adolescents and adults. N Engl J
s41582-020-0374-y. Med 2004;351:1741–51. doi: 10.1056/NEJMoa040573.
[9] Simner PJ, Miller S, Carroll KC. Understanding the promises and hurdles of meta- [18] Van Der Weert EM, Hartgers NM, Schaaf HSimon, Eley BS, Pitcher RD,
genomic next-generation sequencing as a diagnostic tool for infectious diseases. Wieselthaler NA, et al. Comparison of diagnostic criteria of tuberculous meningi-
Clin Infect Dis 2018;66:778–88. doi: 10.1093/cid/cix881. tis in human immunodeficiency virus-infected and uninfected children. Pediatr
[10] Gu W, Miller S, Chiu CY. Clinical metagenomic next-generation sequencing for Infect Dis J 2006;25:65–9. doi: 10.1097/01.inf.0000183751.75880.f8.
pathogen detection. Annu Rev Pathol Mech Dis 2019;14:319–38. doi: 10.1146/ [19] Torok ME, Chau TTH, Mai PP, Phong ND, Dung NT, Chuong LV, et al. Clinical and
annurev-pathmechdis-012418-012751. microbiological features of HIV-associated tuberculous meningitis in Vietnamese
[11] Wilson MR, Sample HA, Zorn KC, Arevalo S, Yu G, Neuhaus J, et al. Clinical metage- adults. PLoS One 2008;3:e1772. doi: 10.1371/journal.pone.0001772.
nomic sequencing for diagnosis of meningitis and encephalitis. N Engl J Med [20] Jeon YJ, Zhou Y, Li Y, Guo Q, Chen J, Quan S, et al. The feasibility study of non-inva-
2019;380:2327–40. doi: 10.1056/NEJMoa1803396. sive fetal trisomy 18 and 21 detection with semiconductor sequencing platform.
[12] Marais S, Thwaites G, Schoeman JF, To € ro
€ k ME, Misra UK, Prasad K, et al. Tubercu- PLoS One 2014;9:e110240. doi: 10.1371/journal.pone.0110240.
lous meningitis: a uniform case definition for use in clinical research. Lancet [21] Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler
Infect Dis 2010;10:803–12. doi: 10.1016/S1473-3099(10)70138-9. transform. Bioinformatics 2009;25:1754–60. doi: 10.1093/bioinformatics/btp324.
[13] Bhigjee AI, Padayachee R, Paruk H, Hallwirth-Pillay KD, Marais S, Connoly C. Diag- [22] Yu G, Zhao W, Shen Y, Zhu P, Zheng H. Metagenomic next generation sequencing
nosis of tuberculous meningitis: clinical and laboratory parameters. Int J Infect for the diagnosis of tuberculosis meningitis: a systematic review and meta-analy-
Dis 2007;11:348–54. doi: 10.1016/j.ijid.2006.07.007. sis. PLoS One 2020;15:e0243161. doi: 10.1371/journal.pone.0243161.
[14] Van Well GTJ, Paes BF, Terwee CB, Springer P, Roord JJ, Donald PR, et al. Twenty [23] Yan L, Sun W, Lu Z, Fan L. Metagenomic next-generation sequencing (mNGS) in
years of pediatric tuberculous meningitis: a retrospective cohort study in the cerebrospinal fluid for rapid diagnosis of Tuberculosis meningitis in HIV-negative
western cape of South Africa. Pediatrics 2009;123:e1–8. doi: 10.1542/peds. population. Int J Infect Dis 2020;96:270–5. doi: 10.1016/j.ijid.2020.04.048.
2008-1353. [24] Guo W, Chen X, Liu R, Liang N, Ma Q, Bao H, et al. Sensitive detection of stage I
[15] Saitoh A, Pong A, Waecker NJ, Leake JAD, Nespeca MP, Bradley JS. Prediction of lung adenocarcinoma using plasma cell-free DNA breakpoint motif profiling.
neurologic sequelae in childhood tuberculous meningitis: a review of 20 cases eBioMedicine 2022;81:104131. doi: 10.1016/j.ebiom.2022.104131.
and proposal of a novel scoring system. Pediatr Infect Dis J 2005;24:207–12. doi: [25] Guo Y, Li H, Chen H, Li Z, Ding W, Wang J, et al. Metagenomic next-generation
10.1097/01.inf.0000154321.61866.2d. sequencing to identify pathogens and cancer in lung biopsy tissue. EBioMedicine
[16] Kalita J, Misra UK, Ranjan P. Predictors of long-term neurological sequelae of 2021;73:103639. doi: 10.1016/j.ebiom.2021.103639.
tuberculous meningitis: a multivariate analysis. Eur J Neurol 2007;14:33–7. doi: [26] Miller S, Chiu C. The role of metagenomics and next-generation sequencing in infec-
10.1111/j.1468-1331.2006.01534.x. tious disease diagnosis. Clin Chem 2021;68:115–24. doi: 10.1093/clinchem/hvab173.