Clinical application value of simultaneous plasma and bronchoalveolar lavage fluid metagenomic next generation sequencing in patients with pneumonia-derived sepsis

Abstract

Background

Despite the increasing use of metagenomic next-generation sequencing (mNGS) in sepsis, identifying clinically relevant pathogens remains challenging. This study was aimed to evaluate the clinical utility of simultaneous plasma and bronchoalveolar lavage fluid (BALF) detection using mNGS.

Methods

This retrospective study enrolled 95 patients with pneumonia-derived sepsis (PDS) admitted to the intensive care unit (ICU) between October 2021 and January 2023. Patients were divided into two groups: mNGS group (n = 60) and the non-mNGS group (n = 35), based on whether simultaneous plasma and BALF mNGS were conducted. All patients underwent conventional microbiological tests (CMT), including bacterial/fungal culture of peripheral blood and BALF, as well as sputum culture, detection of 1, 3-beta-D- glucan in BALF and RT-PCR testing. The clinical data of the enrolled patients were collected, and the detection performance and prognosis of plasma mNGS, BALF mNGS and CMT were compared.

Results

The mNGS group exhibited a lower mortality rate than the non-mNGS group (35.0% vs. 57.1%, P = 0.034). Simultaneous detection in dual-sample resulted in a higher proportion of microorganisms identified as definite causes of sepsis alert compared to detection in either plasma or BALF alone (55.6% vs. 20.8% vs. 18.8%, P<0.001). Acinetobacter baumannii, Stenotrophomonas maltophilia, Candida albicans, and human mastadenovirus B were the primary strains responsible for infections in PDS patients. Patients with lower white blood cells and neutrophil indices had a greater consistency in dual-sample mNGS. Patients in the mNGS group had more antibiotic adjustments compared to the non-mNGS group (85.71% vs. 33.33%, P<0.001). The percentage of neutrophils was a risk factor for mortality in PDS patients (P = 0.002).

Conclusion

Dual sample mNGS has the advantage of detecting and determining the pathogenicity of more pathogens and has the potential to improve the prognosis of patients with PDS.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-024-10292-5.

Introduction

Sepsis is a life-threatening syndrome characterized by dysregulated host response to infection, resulting in organ dysfunction. Pneumonia-derived sepsis (PDS), which accounts for about half of all sepsis cases, has the highest mortality rate in intensive care units (ICU) worldwide, with a mortality rate of up to 50% in patients with septic shock [1–3]. Timely and accurate pathogen diagnoses are of great significance for improved clinical outcomes [4]. However, the low detection rate and time-consuming process (3–5 days) of conventional microbiological tests (CMT), including culture, serological tests, and PCR, are the problems faced by clinicians [5]. Metagenomic next-generation sequencing (mNGS) is a promising diagnostic tool for PDS due to its wide range of assays, short turnaround times, and typically results within 24–48 h compared to culture methods [6]. Invasive specimens, such as bronchoalveolar lavage fluid (BALF) are commonly used to perform mNGS in pneumonia patients. In the case of most BALF samples, mNGS reports as many as 5–10 microorganisms, which makes it challenging to distinguish between colonization or infection [7]. Plasma cell-free DNA (cfDNA) sequencing technology enables non-invasive detection of pathogen nucleic acid circulation from different anatomical sites of infection. However, the extent to which microbial DNA circulates in the bloodstream of pneumonia patients remains unclear, and this knowledge could enhance our understanding of disease pathophysiology and clinical diagnosis.

Several studies have explored the efficacy of simultaneously conducting plasma and BALF samples mNGS [8–11]. In patients who test positive for both sample types, the concordance rate can reach 90% [8]. However, some studies have indicated that plasma is a less reliable indicator compared to respiratory sample. Among suspected pneumonia patients, only 26% had at least one plasma mNGS result that matched the BALF mNGS findings [10]. Importantly, the administration of antibiotics prior to sampling poses an inevitable and intricate bias for the mNGS assessment. Additionally, the disease severity can influence the detection results of pathogens. Although mNGS is not sensitive to prior antimicrobial therapy, the cfDNA of lysed microorganisms has a short half-life, which may result in low pathogen load and false-negative results [12].

Therefore, the objective of this study is to analyze the concordance of pathogen detection between plasma and BALF samples using mNGS in PDS patients and explore the clinical utility of this concordance.

Materials and methods

Study design and participants

This retrospective study involved 150 patients with sepsis admitted to the ICU of the National Hospital of Guangxi Zhuang Autonomous Region between October 2021 and January 2023, which is a tertiary academic hospital. The inclusion criteria were as follows: (1) age ≥ 18 years; (2) ICU stay > 24 h; (3) presence of typical clinical signs of pulmonary infection, such as fever, cough, expectoration, and respiratory failure; (4) the diagnosis of pulmonary infection was supported by radiological evidence, including chest X-ray or computed tomography scan results; (5) adherence to the sepsis 3.0 diagnostic criteria issued by the Society of Critical Care Medicine (SCCM) [13]; (6) simultaneous collecting blood and BALF samples from each patient. Based on the inclusion criteria, we excluded 55 cases of sepsis unrelated to pneumonia, resulting in a final selection of 95 sepsis patients from the ICU for retrospective investigation (Fig. 1).

Fig. 1.

Flow diagram of the study

Some patients, due to personal or family circumstances, declined to undergo mNGS testing, opting solely for conventional microbiological tests (CMT). The type of testing conducted allowed for the following grouping criteria: (1) mNGS group: simultaneous mNGS testing of plasma and BALF, in addition to CMT; (2) non-mNGS group: CMT performed on both blood and BALF.

The study was approved by the Scientific Research and Clinical Trials Ethics Committee of the National Hospital of Guangxi Zhuang Autonomous Region (2022-55). Conducted in accordance with the principles of the Declaration of Helsinki.

Data collection

The collected data included demographic information and the results of laboratory examinations, such as white blood cell count (WBC), hemoglobin (Hb), blood platelet count (PLT), neutrophils percentage (NE%), lymphocytes percentage (LY%), neutrophil count (NE), hematocrit (HCT%), C-reactive protein (CRP), and procalcitonin (PCT). Clinical diagnosis, symptoms, sepsis-related organ failure assessment (SOFA) scores, treatment details, and clinical outcomes were extracted from the electronic medical records of the Ethnic Hospital affiliated with Guangxi Medical University [14, 15]. All-cause mortality at discharge is the primary outcome assessed. A standardized data collection form was utilized for this purpose. Data on pre-admission antibiotic treatment, the initial antibiotic administered upon admission, and any subsequent adjustments made based on mNGS results were also recorded. CMT includes smears and cultures of bacterial, fungal of blood and BALF, concurrently with mNGS testing, and sputum smears and cultures of bacterial, fungal, and mycobacteria, including acid-fast staining of mycobacteria. Nucleic acid amplification of Mycobacterium tuberculosis in sputum. 1,3-beta-D-glucan assay and galactomannan assay in BALF and targeted detection of influenza virus A, influenza virus B, respiratory syncytial virus, human adenovirus, parainfluenza virus, coxsackievirus A, coxsackievirus B, echovirus, Chlamydia pneumoniae, Mycoplasma pneumoniae, and Legionella pneumophila using RT-PCR assay.

Metagenomic next-generation sequencing and analysis

Sample collection and nucleic acid extraction

Whole blood samples (≥ 4 mL) and BALF sample (≥ 3 mL) were collected in anticoagulation tubes and transported at 4℃ after collection. cfDNA was extracted from plasma using PathoXtract Plasma Nucleic Acid kit (WYXM03001S, WillingMed Corp, Beijing, China) following the manufacturer’s protocol. Both DNA and RNA were extracted from BALF samples to cover all the potential pathogens. DNA was extracted using PathoXtract^® Basic Pathogen Nucleic Acid Kit (WYXM03211S, WillingMed Corp, Beijing, China) and RNA was extracted using PathoXtract^® Virus DNA/RNA Isolation Kit (WYXM03009S, WillingMed Corp, Beijing, China) following the manufacturer’s protocol [16]. DNA and RNA was eluted with 50 µL of nuclease-free water. The DNA and RNA were mixed and then reverse transcription of the RNA into complementary DNA (cDNA) was performed by using SuperScript^® Double-Stranded cDNA Synthesis Kit (11917020, Invitrogen).

Library preparation and sequencing

For cfDNA libraries construction, the KAPA DNA HyperPrep Kit (KK8504, KAPA, Kapa Biosystems, Wilmington, MA, United States) was used according to the manufacturer’s protocol. Genomic DNA libraries was constructed using the Illumina^® DNA Prep, (M) Tagmentation (20018705, Illumina) according to the manufacturer’s protocol. Pooled libraries were sequenced on NextSeq™ 550Dx system using a 75 bp, single-end sequencing kit (Illumina) and at least 20 million sequencing reads were acquired for each sample.

Bioinformatic analysis

Quality control was carried out after high-quality sequencing data were obtained. High-quality data consistent with the human reference genome sequence were screened and removed using Bowtie2 v2.4.3. By comparing the remaining sequences with the existing nucleic acid sequences of microorganisms in the database, the type of microorganism can be identified [16].

For pathogen identification, a threshold value called RPTM (reads per ten million) was used. RPTM represents the number of pathogen specific reads per ten million. For viral pathogens detection, reads count ≧ 3 was used as an empirical threshold. In the case of bacteria and fungi, positive pathogens in blood samples were required to meet an RPTM threshold ≧ 8. In BALF samples, positive bacteria and fungi were required to achieve an RPTM threshold ≧ 20. Special pathogens (including Cryptococcus and Mycobacterium) with RPTM ≧ 1 was considered positive.

Clinical adjudication of mNGS results

To confirm the composite clinical diagnosis, two experienced infectious disease specialists, who were well-versed in the ICU, independently reviewed the medical records and mNGS results of each patient. The panel of clinicians categorized the possibility of mNGS as the causative pathogen of PDS into four categories: definite, probable, possible, and unlikely. This categorization was based on previous reports by Blauwkamp et al. [17]. (1) Definite: the mNGS pathogen result was concordant with a microbiological test performed within seven days of sample collection for mNGS identification. (2) Probable: the mNGS pathogen result was considered the probable cause of the sepsis alert based on clinical, radiological or laboratory findings. (3) Possible: the mNGS pathogen result had potential for pathogenicity consistent with clinical presentation but an alternative explanation for the symptoms was more likely. (4) Unlikely: the cfDNA sequencing pathogen result had potential for pathogenicity but findings were inconsistent with clinical presentation. The final classification of the cases was at the discretion of the committee chair.

Statistical analysis

All data in this study were statistically analyzed by SPSS 25.0 (USA) and plotted using GraphPad prism8 (USA). The t-test was used to determine the normal distribution and uniformity of variance. The Wilcoxon rank test was used to calculate the variance of measured data that were not normally distributed or had variance homogeneity. Chi-square test was used for discrete variables where appropriate. P-values ≦ 0.05 were considered significant, and all tests were two-tailed. Cox proportional hazards model was applied for discovering the independent risk factors for PDS. The model was built on the “survival” package in R version (v4.2.2).

Results

Patient characteristics

The study included a total of 95 PDS patients, with 60 patients in the mNGS group, and 35 patients in the non-mNGS group. The number of patients in both groups was not equal. Table 1 documented the patient characteristics and baselines information. Statistical analyses were conducted to assess factors like age, sex, SOFA score, comorbidities, and patient outcomes. All factors, except for the SOFA score, showed no significant differences, suggesting a relatively similar baseline between the two groups. Despite the longer ICU length of stay in the mNGS group compared to the non-mNGS group (20 vs. 8, P = 0.115), the mNGS group exhibited a significantly lower mortality rate (35.0% vs. 57.1%, P = 0.034) (Table 1).

Table 1.

Clinical characteristics of participants

Characteristic	mNGS (n = 60)	Non-mNGS (n = 35)	P-value
Age, median (IQR)	68 (60–77)	73 (62–85)	0.177
Sex, n (%)
Male	47 (78.3%)	26 (74.3%)	0.652
Female	13 (21.7%)	9 (25.7%)
SOFA score, median (IQR)	11 (7–13)	8 (5–11)	0.039*
Comorbidities, n (%)
Respiratory failure	42 (70.0%)	27 (77.1%)	0.451
Hypertension	39 (65.0%)	19 (54.3%)	0.302
Diabetes	23 (38.3%)	8 (22.9%)	0.121
Hypoproteinemia	14 (23.3%)	13 (37.1%)	0.150
CAD	6 (10.0%)	8 (22.9%)	0.088
MODS	30 (50.0%)	15 (42.9%)	0.501
ARDS	20 (33.3%)	17 (48.6%)	0.142
CKD	14 (23.3%)	7 (20%)	0.706
Outcomes, median (IQR)
Hospital length of stay(days)	31 (22–53)	22 (9–46)	0.111
ICU length of stay(days)	20 (11–28)	8 (5–14)	0.115
Mortality (%)	21 (35.0%)	20 (57.1%)	0.034*

IQR Interquartile range, SOFA Sepsis-related organ failure assessment, CAD Coronary atherosclerotic heart disease, MODS Multiple organ dysfunction syndrome, ARDS Acute respiratory distress syndrome, CKD Chronic kidney disease

*P-values ≦ 0.05

**P-values ≦ 0.01

***P-values ≦ 0.001

Detection performance of plasma and BALF mNGS

The performance of mNGS was compared to CMT in all patients (Fig. 2A). The pathogen positivity rate was higher in plasma and BALF mNGS compared to blood and BALF culture, respectively (61.7% vs. 6.7%, P < 0.001; 100% vs. 40%, P < 0.001). Specifically, BALF mNGS exhibited a 100% positivity rate. No significant difference was observed in the positivity rates of different CMT methods between the mNGS and non-mNGS groups.

Fig. 2.

A comparison of the detection performance between mNGS and CMT. A The positivity rate of mNGS and CMT detection. B The frequency distribution of the number of microbial identifications by mNGS and CMT. C The proportion of subjects with microbes identified by the different methods. D The consistency of positive sample detection between plasma mNGS and CMT. E The consistency of positive sample detection between BALF mNGS and CMT. F The consistency of positive sample detection between mNGS and CMT

Subsequently, we analyzed the quantity of microorganisms identified by various methods (Fig. 2B). Within the mNGS group, 8.3% of the patients exhibited a singular microorganism through plasma mNGS, whereas 55% identified more than 4 microorganisms through BALF mNGS. The CMT method identified 1 or 2 microorganisms in 78.3% of patients in the mNGS group and 65.7% of patients in the non-mNGS group (Additional file 1: Fig. S1).

Then we assessed the consistency between mNGS and CMT methods. Among the mNGS group, a total of 33 patients were concurrently detected with microorganisms in plasma mNGS, BALF mNGS, and CMT (Fig. 2C). Out of the 33 patients who tested positively for both plasma mNGS and CMT, only 33% had consistent microbial detection (Fig. 2D). Of the 52 patients who tested positive for both BALF mNGS and CMT, 75% displayed at least one consistent microbial detection (Fig. 2E). When combing plasma and BALF sample mNGS results, out of the 52 patients who tested positively for both mNGS and CMT, 79% had consistent microbial detection (Fig. 2F).

Clinical adjudication of plasma and BALF mNGS results

In the mNGS group, 37 patients had both plasma and BALF mNGS test results positive, and among them, 28 patients were identified to have at least one consistent microorganism (Fig. 3A). A total of 284 microorganisms were identified through mNGS. Out of these, 60 microorganisms were identified in the plasma samples, while 260 microorganisms were identified in BALF samples, and 36 of them were consistently identified in both sample types (Fig. 3B). Out of the 36 detections, 55.6% (20/36) was classified as definite causes of the sepsis alert (Fig. 3C). Among the microorganisms identified exclusively in either plasma or BALF mNGS, only 20.8% (5/24 in plasma) and 18.8% (42/224 in BALF) were classified as definite causes of the sepsis alert (Fig. 3C). The proportion of definite causes increased by about 35% when microorganisms were identified in both samples compared to those identified in only one sample, the proportion of microorganisms classified as “Definite + Probable” in cases where microorganisms were simultaneously detected in both samples (88.9%) increased by 9 to 26% compared to microorganisms detected in a single sample alone (62.5% in plasma and 79.9% in BALF). This means that detecting consistent microorganisms can better clarify the clinical significance of pathogens.

Fig. 3.

Clinical performance comparison of plasma and BALF mNGS. A Concordance of pathogen between plasma and BALF mNGS. B Frequency of microbial detection by plasma and BALF mNGS. C Clinical adjustment of microorganisms

Microorganisms identified in patients with PDS

A total of 26 pathogens detected by plasma and BALF mNGS were classified as definite causes of the sepsis alert, including 15 bacteria, 5 fungi, 4 viruses, and 2 special pathogens (Fig. 4). Acinetobacter baumannii, Stenotrophomonas maltophilia, Candida albicans, and Human mastadenovirus B were the most common pathogens in PDS patients. Noting the particular sensitivity of plasma in identifying special pathogens such as Orientia tsutsugamushi and Mycobacterium tuberculosis complex. A total of 43 pathogens were classified as probable causes of sepsis alert. The most frequently detected bacteria were Corynebacterium striatum, Staphylococcus haemolyticus, Enterococcus faecium, and Burkholderia cepacia, as well as Epstein-Barr virus and cytomegalovirus. Similar to the “definite” pathogens, viruses were frequently detected in both plasma and BALF, whereas fungi were exclusively detected in BALF (Additional file 2: Fig. S2). The CMT method identified a total of 19 pathogens in both the mNGS and non-mNGS groups. The most common pathogens in both the mNGS groups and non-mNGS groups were Candida app (Additional file 3: Fig. S3).

Fig. 4.

Displays the distribution of pathogens detected as “definite” in plasma mNGS and BALF mNGS

Then we analyzed the pathogen burden based on RPTM among microorganisms identified by plasma and BALF mNGS. There were no significant differences in pathogen burden among microorganisms classified as definite, probable, possible, and unlikely causes of the sepsis alert in plasma (Fig. 5A). The pathogen burden of microorganisms classified as definite causes of the sepsis alert was significantly higher than that of microorganisms classified as probable, possible, and unlikely causes of the sepsis alert in BALF (Fig. 5B). Moreover, the pathogen burden of microorganisms identified in both plasma and BALF samples was significantly higher than that of microorganisms identified only in plasma or BALF samples (Fig. 5C and D).

Fig. 5.

Compares the number of mNGS sequences. A Comparison of sequence numbers of microorganisms detected in plasma. B Comparison of sequence numbers of microorganisms detected in BALF. C Comparison of sequence numbers of microorganisms simultaneously detected in plasma. D Comparison of sequence numbers of microorganisms simultaneously detected in BALF

The associations between inflammatory markers and mNGS results

To investigate potential factors associated with the matched results between plasma and BALF mNGS findings, we compared differences in inflammatory markers, including WBC, Hb, PLT, NE%, LY%, NE, HCT%, CRP, and PCT. The inflammatory markers between the mNGS and non-mNGS groups were generally similar, except for the Hb index, which was lower in the mNGS group (Additional file 4: Table S1). Subsequently, we compared the differences between the matched or partially matched cases and unmatched cases. There were no significant differences in Hb, PLT, NE%, LY%, HCT%, CRP, and PCT between these two groups. However, the unmatched cases exhibited significantly higher WBC indices (11.03 ± 5.42 vs. 14.30 ± 6.50 10^9/L, P = 0.041) and NE indices (8.75 ± 5.27 vs. 12.17 ± 5.78 10^9/L, P = 0.021) compared to the matched or partially matched cases (Table 2).

Table 2.

Comparison of inflammatory markers between matched or partially matched cases and unmatched cases of plasma and BALF mNGS detection

Laboratory tests	Plasma and BALF mNGS matched or partially matched cases (n = 28)	Plasma and BALF mNGS unmatched cases (n = 32)	P-value
WBC (10^9/L)	11.03 ± 5.42	14.30 ± 6.50	0.041*
Hb (g/L)	90.64 ± 22.60	88.42 ± 22.95	0.707
PLT (10^9/L)	187.54 ± 120.41	242.75 ± 138.20	0.107
NE%	82.14 ± 15.64	83.44 ± 8.87	0.689
LY%	11.22 ± 12.10	11.89 ± 17.83	0.866
NE (10^9/L)	8.75 ± 5.27	12.17 ± 5.78	0.021*
HCT%	30.22 ± 22.46	27.41 ± 7.60	0.507
CRP (mg/L)	108.03 ± 88.30	116.61 ± 94.32	0.736
PCT (ng/ml)	1.32 ± 2.32	8.84 ± 22.37	0.153

WBC White blood cell, Hb Hemoglobin, PLT Blood platelet, NE% Percentage of neutrophils in blood, LY% Percentage of lymphocytes and total white blood cell, NE Neutrophil, HCT% Hematocrit, CRP C-reactive protein, PCT Procalcitonin

^*P-values ≦ 0.05

Application of mNGS for antibiotic adjustment

All hospitalized patients received antibiotic treatment. Among the patients in the mNGS group, 56 individuals received antibiotic treatment prior to undergoing mNGS testing (Table 3). Amon the 56 patients, 85.71% (48/56) had adjustments made to their antibiotic treatment, a higher percentage compared to the non-mNGS group (33.33%, 11/33) (Table 3). Based on the mNGS results, 19.64% (11/56) of the patients had a reduction in the number or spectrum of initial empiric antimicrobial drugs, while no patients in the non-mNGS group experienced de-escalation. Antibiotic escalation was observed in 42.86% (24/56) of patients in the mNGS group, primarily due to the addition of targeted antimicrobial drugs. Additionally, one patient in the mNGS group showed symptom improvement and was discharged from the hospital before the mNGS results were obtained (Table 3).

Table 3.

The adjustment of antibiotics in the mNGS group and the non-mNGS group

Antibiotic adjustments	mNGS (n = 56)	Non-mNGS (n = 33)	P value
Adjustments	48 (85.71%)	11 (33.33%)	< 0.001***
De-escalation	11 (19.64%)	0 (0.00%)	0.007**
Reduce spectrum of agents	3 (5.36%)	0 (0.00%)	0.176
Reduce number of agents	8 (14.29%)	0 (0.00%)	0.023*
Escalation	24 (42.86%)	4 (12.12%)	0.003**
Increase number of agents	20 (35.71%)	4 (12.12%)	0.015*
Increase spectrum of agents	4 (7.14%)	0 (0.00%)	0.116
Same level replacement	13 (23.21%)	7 (21.21%)	0.827
Unadjustments	7 (12.5%)	22 (66.67%)	< 0.001***
No change	7 (12.5%)	22 (66.67%)	< 0.001***
Discharged before mNGS	1(1.79%)

^*P-value ≤ 0.05

^**P-value ≤ 0.01

^***P-value ≤ 0.001

Next, we compared the antibiotic adjustment between the matched or partially matched cases and unmatched cases from dual sample mNGS tests. Among the 27 patients with matched or partially matched results, 81.48% (22/27) had adjustments in antibiotic treatment based on the mNGS results, while 89.66% (26/29) had adjustments in antibiotic treatment in the 29 patients with unmatched results (Additional file 5: Table S2). Compared to patients with unmatched results, more patients in the group with matched or partially matched results increased spectrum of agents (14.81% vs. 0.00%, P = 0.032) (Table S2).

Factors influencing prognosis in patients with PDS

Compared to the non-mNGS group, the mNGS group exhibited a prolonged length of stay in the ICU and a reduced mortality rate (Table 1). Further analysis was conducted to compare the clinical outcomes among cases with matched or partially matched plasma and BALF mNGS results and those with unmatched results. The mortality rate was lower in the plasma and BALF consistent mNGS group compared to the inconsistent group, but the difference was not statistically significant (35.71% vs. 40.61%, P = 0.696) (Table 4).

Table 4.

Clinical outcomes between matched or partially matched cases and unmatched cases of plasma and BALF mNGS detection

Outcomes	Blood and BALF mNGS matched or partially matched cases (n = 28)	Blood and BALF mNGS unmatched cases (n = 32)	P-value
Hospital length of stay (days)	31 (22–52)	30 (22–51)	0.510
ICU length of stay (days)	20 (11–28)	19 (11–28)	0.625
Mortality (%)	10 (35.71%)	13 (40.61%)	0.696

The initially univariate regression analysis identified potential factors influencing sepsis prognosis. The univariate Cox regression analysis indicated a significant association between age (adjusted HR = 0.98) and CMT positive (adjusted HR = 2.36) with adverse outcomes (P ≦ 0.05) (Table 5). To conduct a comprehensive assessment of independent risk factors and determine the impact of matched dual mNGS results on patient prognosis, we incorporated ‘dual mNGS consistency’ and variables with P < 0.2 in the univariate analysis into the multivariate analysis. The multivariate analysis revealed that NE% emerged as an independent risk factor for poor prognosis among sepsis patients (adjusted HR = 0.94, P = 0.002) (Table 5). The consistency of dual-mNGS exhibited a trend as a risk factor for mortality in sepsis patients (adjusted HR = 1.35, P = 0.451).

Table 5.

Multivariate regression analysis of predictors of clinical outcomes

Variables	Univariate analyses		Multivariate analysis
	HR (95% CI)	P-value	HR (95% CI)	P-value
Gender	0.77 (0.38–1.54)	0.457	——	——
Age	0.98 (0.97–1.00)	0.048*	0.99 (0.96–1.01)	0.256
SOFA	1.00 (0.94–1.08)	0.914	——	——
RF	0.70 (0.36–1.38)	0.307	——	——
Dual mNGS consistent	0.94 (0.49–1.82)	0.860	1.35 (0.62–2.93)	0.451
CMT positive	2.36 (1.02–5.46)	0.044*	0.54 (0.15–2.00)	0.358
WBC	1.02 (0.98–1.08)	0.332	——	——
Hb	1.01 (1.00–1.02)	0.156	1.01 (0.99–1.03)	0.194
PLT	1.00 (1.00–1.00)	0.356	——	——
NE%	0.98 (0.96–1.00)	0.075	0.94 (0.91–0.98)	0.002**
HCT	0.99 (0.97–1.01)	0.270	——	——
CRP	1.00 (1.00–1.00)	0.747	——	——
PCT	1.01 (1.00–1.02)	0.072	1.01 (1.00–1.02)	0.064

HR Hazard ratio, CI Confidence interval, SOFA Sequential organ failure assessment, CMT Conventional microbiological tests

^*P-values ≦ 0.05

^**P-values ≦ 0.01

Discussion

Our study involved conducting a dual sample mNGS test on 60 PDS patients and comparing the antibiotic adjustment and clinical outcomes of these patients to those who did not undergo mNGS testing. Overall, mNGS detected a wider range of pathogens compared to CMT methods, even when most patients have already undergone antibiotic treatment before specimen collection. The results of metagenomic NGS can be valuable even when they are concordant with results of conventional testing or confirm empirical treatment plans. In those cases, it provides reassurance of the diagnosis and potentially rules out non-infection cases [11].

The presence of consistent pathogens in both plasma and BALF samples indicates a substantial dissemination of pathogens from the lungs to the bloodstream. In our study, 46.67% (28/60) of patients demonstrated the presence of at least one matched microorganism in both plasma and BALF through mNGS. The microorganisms that were consistent between plasma and BALF samples in mNGS analysis tends are more likely to be associated with sepsis. Compared to the microorganisms exclusively identified in either plasma or BALF samples, the percentage of microorganisms that were classified as definite causes of sepsis alert increased by 35%. Interpreting mNGS results of respiratory specimens is challenging. While subjective judgment by doctors is necessary, distinguishing between colonization and infection remains a significant challenge when formulating follow-up treatment plans using mNGS results from respiratory samples only [7]. Our findings suggest that dual samples mNGS can conveniently and promptly confirm the pathogen in patients, enabling clinicians to initiate targeted treatment at an earlier stage.

Among the pathogens identified, we found that viruses were the most easily detectable in both plasma and BALF simultaneously. This finding suggests that viruses are more likely to shed from the lungs into the bloodstream than bacteria or fungi. This is similar to previous studies, and viral DNA in the blood may also reflect a lung infection, as high blood flow in the lungs may lead to increased shedding of pathogen DNA [9, 18]. Several studies have indicated that plasma mNGS can effectively monitor viral pneumonia and viral sepsis in clinical practice [9, 19]. In bacteria, Klebsiella pneumoniae, a common pathogen of PDS, was found to be more easily detected in both plasma and BALF simultaneously [20, 21]. Plasma mNGS exhibited a lower detection rate for fungi compared to BALF, consistent with previous studies’ findings [22, 23].

We conducted an analysis of clinical indicators in the mNGS group. The results indicated that individuals with lower levels of WBC and NE index were more likely to exhibit consistent findings between plasma and BALF sample mNGS results. Previous research have also suggested a potential correlation between reduced NE counts and an increased likelihood of common human microbiome entering the bloodstream [24]. However, further investigation is necessary to validate the feasibility of these findings and elucidate the underlying mechanisms. These research findings can provide valuable insights for clinicians in selecting appropriate specimens for mNGS testing.

In this study, a higher proportion of patients in the mNGS group experienced adjustments in antibiotic treatment compared to non-mNGS patients (85.71% vs. 33.33%, P < 0.001). Those whose treatment was modified based on mNGS results received not only antibiotic escalation but also antiviral drugs (e.g., ganciclovir for Epstein-Barr virus, cytomegalovirus and Herpes simplex virus−1, ribavirin for Human mastadenovirus B), antifungal drugs (e.g., caspofungin, fluconazole), and antibiotics targeting specific bacteria (e.g., vancomycin). Patients with consistent dual sample mNGS results were more likely to receive an upgrade in the spectrum of antibiotics and the addition of specific antimicrobial agents compared to those with inconsistent results. While mortality was lower in the plasma and BALF mNGS matched group than in the plasma and BALF mNGS unmatched group, the difference was not statistically significant. This may be due to the clinical change of empiric therapy to targeted therapy when definite pathogen were detected in plasma and BALF. However, when patients with pneumoniac-sepsis were associated with bloodstream infection, they tend to present with a more severe infection state, and rapid progression of the disease predicts higher mortality despite targeted antibiotic therapy [25].

Several patient-related factors, such as age, gender, SOFA, and APACHE II scores, have been recognized as independent risk factors for sepsis-related mortality, as mentioned in previous studies [26, 27]. In our study involving PDS patients, similar results were observed. In the univariate analysis, age and positive CMT were identified as independent risk factors for mortality. Furthermore, the multivariate analysis showed that NE% was a significant risk factor for mortality in PDS patients. While there was a tendency for consistent results from dual sample mNGS to be a risk factor, it did not reach statistical significance.

In our study, mNGS is expected to be a promising infection identification test due to its advantages in pathogen detection. However, the clinical application of mNGS faces more challenges in practical aspects. Patients with PDS are critically ill and culture is considered the “gold standard” for clinical diagnosis of infection, but some slow-growing organisms may take days to weeks [28]. Although mNGS has a fast turnaround time and results are usually available within 24–48 h, the unbiased detection of mNGS increases the difficulty of pathogen identification [29]. Although mNGS is broader than multiplex PCR, mNGS has lower specificity, which is due to the limitations of mNGS in clinical diagnosis [30]. In addition, the high price of mNGS will also increase the medical burden on patients [30].

Our study has several limitations. Firstly, the small sample size necessitates larger-scale clinical trials to validate our findings. Secondly, this study mainly focused on the analysis of the detection performance of plasma and BALF mNGS, and did not elaborate on the correlation between CMT and the microorganisms detected by BALF and plasma mNGS. Thirdly, although consensus was attained by two experienced senior physicians in this study, incorporating clinical manifestations and other laboratory results, subjectivity may have introduced certain bias. Finally, as a retrospective study, the sampling time of mNGS testing was different for each patient, and several studies found differences in the performance of mNGS and conventional microbiological testing at different sampling time points (at onset of sepsis, within 24 h, 1 day, 2 days, 7 days, and 14 days) [12, 31]. Therefore, the mNGS performance of each patient at different stages of PDS may introduce bias in the detection results.

Conclusion

Simultaneously conducting plasma and BALF mNGS facilitates the identification of the causative agent of sepsis, especially when the NE% of patients is lower. mNGS has the potential to improve the prognosis of patients with PDS.

Abbreviations

mNGS

Metagenomic next-generation sequencing

BALF

Bronchoalveolar lavage fluid

PDS

Pneumonia-derived sepsis

ICU

Intensive care unit

CMT

Conventional microbiological tests

WBC

White blood cell

Hb

Hemoglobin

PLT

Platelet

NE

Neutrophils

LY

Lymphocytes

NE%

Neutrophils percentage

LY%

Lymphocytes percentage

HCT%

Hematocrit percentage

CRP

C-reactive protein

PCT

Procalcitonin

SOFA

Sepsis-related organ failure assessment

cfDNA

Cell-free DNA

cDNA

Complementary DNA

RPTM

Reads per ten million

Authors’ contributions

GQ, JYL and WG had the idea for and designed the study and had full access to all data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis. JYL, DXP and YXG wrote and revised the manuscript. XLL, CD, FFX, ZNL, LHS and QHC participated in data collection. BZ, YFZ, and SHN performed the data analysis. RFQ and FFJ contributed to critical revision of the manuscript for important intellectual content.All authors read and approved the final manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

The data presented in the study are deposited in the SRA (https://www.ncbi.nlm.nih.gov/sra/) repository, accession number PRJNA1063543.

Declarations

Ethics approval and consent to participate

The study was approved by the Research Ethics Committee of National Hospital of Guangxi Zhuang Autonomous Region (2022-55), which waived the written informed consent requirement. Conducted in accordance with the principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.