Clinical impact of bronchoalveolar lavage fluid metagenomic next-generation sequencing in immunocompromised patients with severe community-acquired pneumonia in ICU: a multicenter retrospective study

Abstract

Background

An increasing number of critically ill patients are immunocompromised. These patients are at high risk of intensive care unit (ICU) admission because of numerous complications. Acute respiratory failure due to severe community-acquired pneumonia (SCAP) is one of the leading causes of admission. Early targeted antibiotic therapy is crucial for improving the prognosis of these patients. Metagenomic next-generation sequencing (mNGS) in bronchoalveolar lavage fluid (BALF) has shown significant value in pathogen detection in recent years. However, there are few studies on summarizing pathogen profiles of SCAP in immunocompromised patients.

Methods

We performed a multicenter retrospective analysis of patients with SCAP in the ICU diagnosed between May 2021 to October 2024. Bronchoalveolar lavage fluid (BALF), blood, and sputum samples were collected and subjected to mNGS and conventional microbiological tests (CMTs). The pathogen profiles detected by the two methods were compared.

Results

In our study, compared to CMTs, mNGS increased the detection rates of mixed infections in the immunocompromised group (58.82% vs 17.96%, P < 0.05) and immunocompetent group (44.58% vs 18.72%, P < 0.05), while also reducing the rate of no pathogen detected (4.90% vs 38.73%, P < 0.05; 8.37% vs 32.76%, P < 0.05). In both groups, the proportion of positive clinical impacts (diagnosis) resulting from mNGS results exceeded 90% (96.57% vs 93.84%), and the treatment effectiveness rate in the immunocompromised group was higher than in the immunocompetent group (65.69% vs 56.40%, P < 0.05). Further analysis showed that when mNGS-guided treatment was effective, the 28-day mortality rate significantly improved in both the immunocompromised group (31.34% vs 74.29%, P < 0.05) and the immunocompetent group (42.36% vs 40.68%, P < 0.05) compared to when the treatment was ineffective.

Conclusion

This study indicates that ICU patients with SCAP, particularly those who are immunocompromised, are more likely to have polymicrobial infections. mNGS in BALF provides rapid and comprehensive pathogen profiling of pulmonary infections, thereby having a positive impact on both the diagnosis, treatment and prognosis of immunocompromised patients with SCAP.

Supplementary Information

The online version contains supplementary material available at 10.1007/s15010-025-02520-0.

Introduction

The proportion of immunocompromised patients has risen in recent years to about a third of all ICU admissions [1]. Several factors have contributed to this trend, including the increasing aggressiveness and duration of cancer therapies, the more frequent use of organ and hematopoietic cell transplants, and the introduction of steroid-sparing drugs that induce specific immune deficiencies for the treatment of autoimmune and autoinflammatory diseases [2, 3]. The higher survival rates with good quality of life put them at risk for severe infections.

Severe community-acquired pneumonia (SCAP) is one of the leading causes for intensive care unit (ICU) admission in immunocompromised patients, who are at risk for hypoxemic acute respiratory failure (ARF), sepsis, and multiorgan failure [4, 5]. However, community-acquired pneumonia (CAP) is a highly heterogeneous disease caused by a wide range of pathogens [6], which can significantly vary depending on the patient’s location, immune status, and other clinical factors [7]. The in-hospital mortality of SCAP is 30.6–60% [8–10] in Europe according to the severity of hypoxemia. Rapid identification of pathogens and precise antimicrobial therapy are crucial for improving the prognosis of immunocompromised patients with SCAP [11].

Conventional microbiological tests (CMTs) are extensively used to diagnose CAP. However, CMTs can be time-consuming and less effective in detecting atypical infections. Metagenomic next-generation sequencing (mNGS), a novel pathogen detection technology, can enable the quick and unbiased identification of harmful viruses, bacteria, fungi, and parasites [7, 12–14]. It is expected to play a significant role in the future diagnosis of SCAP, potentially becoming a first-line detection method.

However, the systematic evaluation of the clinical use of mNGS for SCAP in immunocompromised patients has been insufficient. This multicenter retrospective study therefore assessed the use of mNGS in the clinical impact of these patients in ICU.

Methods

Patients

From May 2021 to October 2024, 642 patients who were diagnosed with SCAP in the ICU were reviewed at five hospitals in China. mNGS is commonly used in our hospital’s ICU for complex cases such as SCAP, but it is not part of routine clinical practice due to its high cost. The use of mNGS in this study was driven by the research protocol, and the mNGS was performed only on bronchoalveolar lavage fluid (BALF) in this study.

SCAP was defined as CAP patients who met any of the major criteria [15]: (1) requiring tracheal intubation and mechanical ventilation; (2) septic shock, and still in need of vasoactive drugs after active fluid resuscitation or ≥ 3 minor criteria: (1) respiratory rate (RR) ≥ 30 bpm; (2) oxygenation index ≤ 250 mmHg (1 mmHg = 0.133 kPa); (3) infiltrates in multiple lung lobes; (4) disturbance of consciousness and (or) disorientation; (5) blood urea nitrogen (BUN) ≥ 7.14 mmol/L; (6) systolic blood pressure (SBP) < 90 mmHg, requiring active fluid resuscitation.

The inclusion criteria were as follows: (1) meeting the diagnostic criteria for SCAP; (2) age ≥ 18 years old; (3) underwent CMTs and mNGS at the same time. The exclusion criteria included (1) incomplete clinical data; (2) age < 18 years old; (3) acquired immune deficiency syndrome (AIDS) immunocompromised individuals.

According to the inclusion/exclusion criteria, a total of 610 patients were finally included in our study. Based on the immunocompromised status, the patients were divided into the immunocompromised group (n = 204) and the immunocompetent group (n = 406) (Fig. 1).

Fig. 1.

Study flow diagram

Based on previous studies [16, 17], the definition of immunocompromised status should met ≥ 1 of the following risk factors:

1. Hematologic malignancies;

2. Hematopoietic stem cell transplantation or solid organ transplantation;

3. Neutropenia or chemotherapy for solid tumors in the past 3 months;

4. Long-term use of corticosteroids (0.3 mg/kg/day of prednisone equivalent for 3 weeks);

5. Taken antirheumatic drugs, biological immunomodulators or immune-suppressants.

Concordance evaluation

The consistency of the results between mNGS and CMTs for clinical diagnosis was comprehensively assessed by two experienced clinicians with over 10 years of ICU experience (Fang, HL and Zhang, WW), leading to a final diagnosis based on the comprehensive results. The diagnosis took into account the patients’ clinical characteristics, results of CMTs, results of mNGS, laboratory data, and other relevant factors. Additionally, chest computed tomography (CT) and other imaging modalities are also employed to assist in the overall diagnostic process.

To highlight the differences between mNGS and CMTs in detecting pathogenic microorganisms, we defined the following criteria and used a composite pie chart to visually represent the results:

1. Complete match: The pathogenic microorganisms identified by mNGS and CMTs were in complete consistent.

2. Partial match: The pathogenic microorganisms identified by mNGS and CMTs were partially consistent, the mNGS and/or CMTs identified additional pathogenic microorganisms.

3. Mismatch: The pathogenic microorganisms identified by mNGS and CMTs were completely inconsistent.

Samples collection and conventional microbiological tests

Following the exclusion of contraindications, bronchoscopy is performed by a trained physician under sterile conditions to collect BALF (both mNGS and CMTs were performed on BALF samples). Simultaneously, blood, sputum, and other relevant samples are collected for CMTs.

CMTs include BALF culture, sputum culture, blood culture, quantitative polymerase chain reaction (qPCR), T-cell spot test (T-spot), GeneXpert, (1,3)-β-D-glucan test (G test), and Galactomannan test (GM test). qPCR is primarily used for viral detection and specific pathogen identification. T-spot and GeneXpert are used for detecting Mycobacterium tuberculosis infection. G test and GM test and qPCR are used for detecting or distinguishing invasive fungal infections.

mNGS procedure

DNA/RNA extraction

BALF samples were centrifuged at 12,075 × g and 4 °C for 5 min. For each sample, 500 μL of supernatant was used to extract genomic DNA via the PathoXtract® WYXM03202S Universal Pathogen Enrichment Extraction Kit (WillingMed, Beijing, China), and RNA was extracted using PathoXtract® Virus DNA/RNA Isolation Kit (WYXM03009S, WillingMed Corp, Beijing, China) according to the manufacturer’s protocol.

Construction of DNA/RNA libraries and sequencing

For only DNA pathogen detection, an Illumina® DNA Prep (M) Tagmentation Kit (20018705; Illumina, San Diego, USA) was used to construct the DNA libraries. For DNA and RNA pathogen co-detection, DNA and RNA were mixed and then reverse transcription of the RNA to complementary DNA (cDNA) was performed by using SuperScript® Double-Stranded cDNA Synthesis Kit (11917020, Invitrogen). The quality of the libraries was assessed using a Qubit fluorescence quantification analyzer (Thermo Fisher) and an Agilent 2100 Bioanalyzer (Agilent Technologies). Sequencing was performed on a NextSeq™ 550Dx sequencer (Illumina, San Diego, USA), with each sample achieving at least 20 million sequencing reads.

Bioinformatic analysis

Sequencing data were processed automatically to produce a detection report. The FASTQ-format data obtained from sequencing were processed with Trimmomatic v0.40 [18] to eliminate low-quality or undetected sequences, splice contaminants, high-coverage repeats, and short-read-length sequences. The high-quality sequencing data were then compared to the human reference genome GRCh37 (hg19) via Bowtie2 v2.4.3 [19] to remove human host sequences. The remaining sequences were aligned to a previously constructed reference database containing more than 24,000 pathogens via Kraken2 v2.1.0 [20] for pathogen identification. The Microbial Genome Databases were downloaded from GenBank (http://ftp.ncbi.nlm.nih.gov/genomes/genbank/).

Criteria for a positive mNGS result

To exclude bacteria suspected of being colonizers, we compared the detected microbial species with an in-house background microbial database. This database contains information on common bacterial species that are typically found as part of the normal microbiota in our patient population. Any bacterial species that matched the profile of known colonizers in the database were excluded from the analysis to focus on potential pathogens [21]. The remaining bacteria (mycobacteria excluded), viruses, and parasites undergo the following filtering criteria:

1. Bacteria and fungi: the specific reads per ten million (RPTM value) ≥ 20;

2. Virus: RPTM ≥ 3 [22];

3. Special pathogens (including Mycobacterium tuberculosis (MTB), and Cryptococcus, etc.): reads ≥ 1 since their DNA is hard to extract and unlikely to be contaminated for BALF samples [23].

Evaluation of clinical impact of mNGS results

In this study, pathogen identification was conducted by two experienced clinicians with over 10 years of ICU experience. The diagnosis was based on a comprehensive assessment including clinical presentation, laboratory tests, mNGS results, imaging studies, and adjustments to patient management. Any disagreements among clinicians were resolved through further discussion and consensus.

To evaluate the clinical impact of mNGS, we categorized its clinical impact (based on the clinical impact types) as positive, negative, or no effect. The evaluation time point set at 72 h after obtaining the mNGS results.

The clinical impact types included:

Diagnosis-related:

1. New diagnosis: mNGS results led to the identification of new causative pathogens in cases where CMTs not identified.

2. Earlier diagnosis: The detected pathogens by mNGS were consistent with the CMTs, but the results obtained by mNGS was earlier than CMTs.

3. Confirmed diagnosis: The detected pathogens by mNGS were consistent with the CMTs, but the results obtained by mNGS was later than CMTs.

4. Missed diagnosis: The microorganisms reported by mNGS were not clinically considered (determined by clinicians).

5. Null diagnosis: mNGS did not detect any pathogens, and the patient was ultimately diagnosed with pulmonary infection based on CMTs or comprehensive clinical judgment.

6. True-negative diagnosis: Both mNGS and CMTs results were negative, and the patient was final diagnosis of non-pulmonary infection [24].

Therapy-related:

1. Appropriate targeted antibiotic treatment: the empirical antimicrobial agents were unable to cover the microorganisms reported by mNGS. Subsequently, the antimicrobial therapy was adjusted based on the results of mNGS (included antibiotic change or addition of corresponding antimicrobial treatments for fungal, viral, and parasitic pathogens). Additionally, when the mNGS result is negative and the CMTs result is positive or the infection is clinically confirmed, empirical antimicrobial agents do not cover the microorganisms reported by CMTs, antimicrobial treatment should be adjusted accordingly.

2. Antibiotic escalation: the existing antimicrobial agents could cover the microorganisms reported by mNGS, and there was a need to increase in the number of antimicrobial drugs or a switch from narrow-spectrum antibiotics to broad-spectrum antibiotics.

3. Antibiotic de-escalation: reducing the quantity of antimicrobial agents, discontinuing agents, or replacing broad-spectrum with narrow-spectrum antibiotics.

4. No change: The empirical antimicrobial therapy covered the microorganisms reported by mNGS, the therapy was maintained.

5. Misguided antibiotic treatment: mNGS findings that either falsely identify non-existent pathogens or fail to detect present pathogens, potentially leading to inappropriate treatment decisions (Fig. 2).

Fig. 2.

Evaluation procedures for clinical impact of positive mNGS (A) and negative mNGS results (B)

Definition of treatment effectiveness

The effective treatment is defined as the situation that the clinical condition of a patient is stabilized after therapy (24–72 h, depending on the type of infection and the response to treatment). Patients who met the following five criteria were considered clinical stability: (1) body temperature ≤ 37.8 °C; (2) heart rate ≤ 100 bpm; (3) respiratory rate ≤ 30 bpm; (4) systemic blood pressure ≥ 90 mmHg (without the use of vasopressors); (5) O₂ saturation ≥ 90% (or arterial partial pressure of O₂ ≥ 60 mmHg, while breathing air).

The ineffective treatment is defined as either of the following situations appeared in a patient: the symptoms are not improved after the treatment be adjusted, and requiring alternative antibiotics; exacerbation and disease progression after the treatment adjustment [15, 25].

Statistical analysis

Data were statistically analyzed using SPSS (version 26; IBM Corp., Armonk, NY, USA) and R software (version 4.4.1; R Foundation for Statistical Computing, Vienna, Austria). Figures were created using GraphPad Prism (version 10; GraphPad Software, San Diego, CA, USA) and OriginPro (version 2024b; OriginLab Corp., Northampton, MA, USA). Continuous variables were tested for normality using the Kolmogorov–Smirnov test and were expressed as median and interquartile range (IQR) or as mean ± SD, depending on the distribution. Data comparisons were performed using an independent samples t-test or the Mann–Whitney U test. Categorical variables were compared using the chi-square test. The Kappa consistency test was conducted to determine whether the diagnostic results of the two methods were consistent. In this study, P < 0.05 (two-tailed test) was considered statistically significant.

Results

Clinical characteristics of patients

We conducted a retrospective analysis of patients admitted to the ICU of five participating centers between May 2021 and October 2024 with suspected SCAP, and finally 610 patients were included in this study. The patients were divided into two groups based on immunocompromised status: the immunocompromised group (n = 204) and the immunocompetent group (n = 406) (Fig. 1).

Demographic and clinical characteristics of the immunocompromised group and the immunocompetent group were compared and summarized in Table 1. There were no statistically significant differences between the groups regarding age, gender distribution, and several underlying diseases such as hypertension, diabetes, cardiovascular disease, cerebrovascular disease, and lung disease (P > 0.05). In the immunocompromised group, the levels of procalcitonin (PCT), C-reactive protein (CRP), D-dimer were higher than the immunocompetent group (P < 0.05). Additionally, the platelet (PLT) counts were lower in the immunocompromised group compared to the immunocompetent group (P < 0.05).

Table 1.

Clinical characteristics of patients

	Immunocompromised group (n = 204)	Immunocompetent group (n = 406)	P value
Age, years, median (IQR)	67 (58–74)	70 (58–79)	0.547
Gender (n, %)			0.609
Male	116 (56.86%)	222 (54.68%)
Female	88 (43.41%)	184 (45.32%)
Underlying diseases (n, %)
Hypertension	49 (26.25%)	119 (29.31%)	0.061
Diabetes	27 (13.24%)	56 (13.79%)	0.535
Cardiovascular disease	14 (6.86%)	23 (5.67%)	0.832
Cerebrovascular disease	13 (6.37%)	15 (3.69%)	0.193
Lung disease	48 (23.53%)	81 (19.95%)	0.485
Clinical laboratory tests, median (IQR)
WBC, × 109/L	7.55 (4.40–11.40)	7.70 (5.58–10.43)	0.261
PCT, ng/mL	1.97 (0.50–11.35)	0.35 (0.03–1.25)	0.000
CRP, mg/L	63.59 (16.00–152.04)	46.40 (7.47–118.38)	0.017
Platelets, × 109/L	128.50 (87.00–204.50)	175.50 (114.00–237.25)	0.000
Hemoglobin, g/L	110.00 (95.00–122.00)	112.00 (94.75–127.00)	0.198
Neutrophil count, × 109/L	5.91 (2.76–9.73)	5.88 (3.77–8.88)	0.336
Bilirubin, μmol/L	10.75 (7.65–15.80)	11.25 (8.00–16.93)	0.578
Creatinine, μmol/L	65.65 (55.20–90.08)	68.05 (53.20–91.80)	0.940
d-dimer, mg/L	4.04 (2.06–9.36)	3.02 (2.02–8.63)	0.000
Lactic acid, mmol/L	1.84 (1.40–2.60)	1.9 (1.30–2.81)	0.576
Disease severity
APACHE-II, median (IQR)	21 (16–26)	14 (9–18)	0.000
CURB-65, median (IQR)	2 (2–3)	2 (1–3)	0.008
Score ≥ 3 (n, %)	71 (34.80%)	101 (24.88%)
Score < 3 (n, %)	133 (65.20%)	305 (75.12%)
SOFA, median (IQR)	13 (10–15)	12 (9–14)	0.000
ICU treatment
MV (n, %)	177 (87.76%)	323 (79.56%)	0.029
Length of MV days, median (IQR)	7 (2–13)	9 (2–17)	0.198
Vasopressor (n, %)	158 (77.45%)	269 (66.26%)	0.004

WBC white blood count; PCT procalcitonin; CRP c-reactive protein; APACHE-II acute physiology and chronic health evaluation II; MV mechanical ventilation; IQR interquartile range; ICU intensive care unit

The immunocompromised group presented with a higher APACHE-II score and SOFA score compared to the immunocompetent group (P < 0.001). For CURB-65 scores, the immunocompromised group also had a higher median score (P < 0.05), and a greater proportion of patients had a score ≥ 3 in this group (P < 0.05) than in the immunocompetent group.

Mechanical ventilation (MV) was used more frequently in the immunocompromised group (P < 0.05). Additionally, vasopressor use was also higher in the immunocompromised group (P < 0.05).

The distribution of immunocompromised patients

Among the 204 immunocompromised patients, most had neutropenia or had received chemotherapy for solid tumors in the past 3 months (156/204, 76.47%), followed by those with hematologic malignancies (41/204, 20.10%), long-term corticosteroid use (≥ 0.3 mg/kg/day prednisone equivalent for ≥ 3 weeks) (31/204, 15.20%), taken antirheumatic drugs, biological immunomodulators, or immunosuppressants (5/204, 2.45%), and those who had undergone hematopoietic stem cell or solid organ transplantation (3/204, 1.47%) (Fig. 3).

Fig. 3.

The distribution of immunocompromised patients

Evaluation of the concordance of the mNGS and CMTs results

Among the 610 patients included in this study, the Chi-square test was conducted to compare the difference of positivity rates between mNGS and CMTs in the immunocompromised group and the immunocompetent group. The positive results of mNGS were higher than CMTs in both the immunocompromised (94.11 vs 53.43%, P < 0.001) and immunocompetent (91.13% vs 58.87%, P < 0.001) groups and all patients (92.13% vs 57.05%, P < 0.001) (Fig. 4A).

Fig. 4.

A Comparison of the positivity rates between mNGS and CMTs.Microorganism identification consistency between mNGS and CMTs in B all patients, C Immunocompromised group and D immunocompetent group

Subsequently, the Kappa consistency test was performed to further validate the diagnostic consistency between the two methods. The results of the Kappa analysis indicate that the two methods show poor level of consistency (Kappa value = 0.092, P < 0.001) (Table 2).

Table 2.

Kappa analysis of concordance between mNGS and CMTs results

mNGS	CMTs		Total
	Positive ( +)	Negative (−)
Positive ( +)	333	229	562
Negative (−)	15	33	48
Total	348	252	610

P < 0.001, Kappa value = 0.092

In the total group of 610 patients, both mNGS plus CMTs yielded positive results in 54.59% (333/610) of the patients, and negative results in 5.41% (33/610) (Fig. 4B). Additionally, 37.54% (229/610) of patients tested positive with only mNGS, while 2.46% (15/610) tested positive with only CMTs. Among the 333 patients who tested positive with both methods, 22 (6.61%) were completely matched, 98 (29.43%) were mismatch, and 213 (63.96%) were partial match.

In the immunocompromised group (n = 204), both mNGS plus CMTs yielded positive results in 50.72% (105/204) of patients and negative results in 5.31% (11/204) (Fig. 4C). 42.03% (87/204) of patients tested positive with only mNGS, while 1.93% (4/204) tested positive with only CMTs. Among the 105 patients who tested positive with both methods, 9 (8.57%) were completely match, 29 (27.62%) were mismatch, and 67 (63.81%) were partial match.

In the immunocompetent group (n = 406), both methods yielded positive results in 56.16% (228/406) of patients and negative results in 6.16% (25/406) (Fig. 4D). A total of 34.98% (142/406) of patients tested positive with only mNGS, and 2.71% (11/406) tested positive with only CMTs. Among the 104 patients who tested positive with both methods, 13 (5.70%) were completely match, 69 (30.26%) were mismatch, and 146 (64.04%) were partial match.

Difference between the two groups in detecting pathogenic microorganisms by using mNGS

The distribution of pathogen types detected by mNGS between the immunocompromised group and the immunocompetent group is presented in the Fig. 5A, B. In the immunocompromised group, mixed infections were more common, accounting for 58.82% (120/204) of cases, compared to 44.58% (181/406) in the immunocompetent group. Single type infections were found in 36.27% (74/204) of immunocompromised group, slightly lower than 47.04% (191/406) in the immunocompetent group.

Fig. 5.

Categorization of microorganism infections in patients based on mNGS results. A Immunocompromised group. B Immunocompetent group

As shown in Fig. 6, the comparative diagnostic rates of microorganisms between the immunocompromised and immunocompetent groups were evaluated using both mNGS and CMTs. The most common bacteria detected in the two groups were consistent, and the detection rate of these bacteria, including Corynebacterium species (spp). (30.66% vs 1.97%, P < 0.05), Klebsiella spp. (16.39% vs 9.02%, P < 0.05), Acinetobacter spp. (11.31% vs 9.02%, P < 0.05) and Pseudomonas aeruginosa (11.80% vs 8.52%, P < 0.05), by mNGS was higher than that of CMTs.

Fig. 6.

The distribution of diagnostic rate of microorganism by mNGS or CMTs. *P < 0.05, mNGS result compared to CMTs result. ^#P < 0.05, comparison of mNGS results between the immunocompromised group and the immunocompetent group

In immunocompromised group, compared to the CMTs, the fungi with the most commonly detection rates by mNGS were Candida spp. (25.49% vs 20.59%, P < 0.05), Aspergillus spp. (24.02% vs 7.84%, P < 0.05) and Pneumocystis jirovecii (7.84%, only detected by mNGS). The difference is that, in immunocompetent group, compared to the CMTs, the fungi with the most commonly detection rates by mNGS were Candida spp. (24.14% vs 25.86%, P < 0.05), Aspergillus spp. (11.33% vs 3.69%, P < 0.05) and Penicillium spp. (4.43%, only detected by mNGS).

The virus with the most commonly detection rates by mNGS in both groups were Epstein-Barr virus (EBV) (17.87% vs 0.82%, P < 0.05), Cytomegalovirus (CMV) (13.11% vs 0.49%, P < 0.05), Human herpesvirus 7 (HHV-7) (13.44%, only detected by mNGS) compared to the CMTs.

Notably, in the immunocompromised group, the detection rates of Corynebacterium spp., Mycobacterium tuberculosis (MTB), Non-tuberculous mycobacteria (NTM), Aspergillus spp., Pneumocystis jirovecii, CMV, and HHV-7 by mNGS were higher compared to the immunocompetent group (all P < 0.05).

Clinical impact of mNGS result in the both groups for diagnosis

The clinical impacts of positive mNGS results for immunocompromised and immunocompetent group were illustrated in Fig. 7A, B. We can found that new and earlier diagnosis were the most common effects of mNGS results, account for more than 75%. Followed by confirmed diagnosis, account for about 13.73%. Compared to the immunocompetent group, mNGS provided a higher proportion of new diagnosis for the immunocompromised group (42.16% vs 33.25%, P < 0.05). However, the other diagnostic outcomes, such as earlier or confirmed diagnosis, did not demonstrate statistically significant differences (P > 0.05).

Fig. 7.

Clinical impact of mNGS result for diagnosis. Percentage of the clinical impact of positive mNGS results related to diagnosis in the in the A immunocompromised group and B immunocompetent group. Sankey diagrams of the clinical diagnosis-related impact types and microorganism types. C Diagnostic rate of clinically confirmed mNGS or CMTs for different types of microorganism in both groups. (* indicate significant difference (P < 0.05) between mNGS and CMTs). D Difference in diagnostic rate for different types of microorganism detected by mNGS in both groups

Among the types of microbial infections detected by using mNGS, mixed infections were the most common in both groups, and it with higher detected rate than CMTs (58.82% vs 17.96%, P < 0.05) in the immunocompromised group. Bacterial infections in both groups were higher than CMTs (P < 0.05). The immunocompetent group showed a higher detection rate of viral infections than CMTs (5.91% vs 0.49%, P < 0.05). (Fig. 7C). Additionally, in both groups, the rate of no pathogens was detected by mNGS was lower than that of CMTs (immunocompromised group: 4.9% vs 38.73%, P < 0.05; immunocompetent group: 8.37% vs 32.76%, P < 0.05).

Interestingly, by calculating the difference in diagnostic rates between mNGS and CMTs, the findings revealed that the detection rates of mixed infections and bacteria in the immunocompromised group were higher than in the immunocompetent group (all P < 0.05).

Clinical impact of mNGS result in the both groups for treatment

To further evaluate the clinical impact of BALF mNGS, we analyzed the therapeutic decisions and the alleviation of infection based on the mNGS reports.

Based on the positive mNGS results, the impacts associated with therapy were greater than those related only to diagnosis or negative outcomes in both groups (all P < 0.05) (Fig. 8A). When considering the negative mNGS results, for the received escalation antimicrobial therapy was higher than the immunocompetent group (58.33% vs 2.78%, P < 0.05). Notably, patients who received de-escalation antimicrobial therapy or no change was higher than the immunocompetent group (80.56% vs 33.33%, P < 0.05) (Fig. 8B).

Fig. 8.

Clinical impact of mNGS result for treatment. A Percentage of clinical impact types with positive or negative mNGS results in both groups. B Percentage of each antimicrobial strategy among the impact related to antimicrobial therapy in both groups. Sankey diagrams of the types of antimicrobial therapy and significance for infection relief in the C immunocompromised group and D immunocompetent group

In terms of therapy-related impact, “appropriate targeted antibiotic treatment” was the most commonly utilized type among the immunocompromised group and immunocompetent group (67.65% vs 54.19%, P < 0.05) (Fig. 8C, D). There was a significant difference in the proportion of patients receiving antibiotic de-escalation between the immunocompromised group and the immunocompetent group (20.59% vs 17.48%, P < 0.05), while the rate of no change in antibiotic therapy was higher in the immunocompetent group compared to the immunocompromised group (17% vs 1.47%, P < 0.05).

In addition, we also calculated the alleviation rate after therapy. The treatment efficacy (improvement or cure) after adjustments based on mNGS was higher in both groups compared to the inefficacy (poor or unknown) (Fig. 8C, D). It is noteworthy that the treatment efficacy in the immunocompromised group was higher than that in the immunocompetent group (65.69% vs 56.40%, P < 0.05).

Further analysis showed that when mNGS-guided treatment was effective, the 28-day mortality rate significantly improved in both the immunocompromised group (31.34% vs 74.29%, P < 0.05) and the immunocompetent group (42.36% vs 40.68%, P < 0.05) compared to when the treatment was ineffective (Fig. 8C, D).

Discussion

With the growing population of immunocompromised individuals, the incidence of SCAP in this group has also increased. As a result, septic shock and multiple organ dysfunction syndrome (MODS) caused by these infections have become leading causes of mortality [17]. However, CMTs have limitations concerning the time required for detection, positive identification rates, and specificity, particularly in severe infection scenarios. Previous studies have highlighted the excellent diagnostic performance of mNGS in SCAP patients and its significant clinical impact [26–28]. Given the variable response to infection among patients with varying degrees of immunosuppression, it is important to evaluate the clinical utility of mNGS in these patients. In this study, we retrospectively analyzed 610 ICU patients with SCAP from five centers, including both immunocompromised and immunocompetent patients. Samples from blood, BALF, and sputum were collected. We compared the clinical impact of mNGS and CMTs and further analyzed pathogen distribution and clinical impact between immunocompromised and immunocompetent groups.

Immunocompromised patients are at a heightened risk of developing severe infections, including sepsis and acute ARF, due to multiple immunologic defects such as neutropenia, lymphopenia, and T and B-cell impairment. The diverse and complex nature of these immunologic profiles, compounded by the concomitant use of immunosuppressive therapies (e.g., corticosteroids, cytotoxic drugs, and immunotherapy), also contributes to the poorer prognosis in this patient population [16]. In our study, immunocompromised patients accounted for approximately one-third of all SCAP cases, which is consistent with previous reports [1, 29, 30]. Compared to the immunocompetent group, the higher disease severity scores (APACHE-II, CURB-65, and SOFA, all P < 0.05), a greater proportion of patients requiring MV and vasopressor support (all P < 0.05), and a trend toward worse 28-day mortality (57.84% vs 50.00%, P = 0.067), all suggest a poorer prognosis in the immunocompromised group (Table 1).

All types of immunosuppression are risk factors for bacterial pneumonia, and 1 out of 5 patients hospitalized for CAP is immunocompromised [31]. According to statistics, bacterial pneumonia accounts for about 5–30% of ICU admissions in CAP patients [1, 32]. However, in immunocompromised patients, the symptoms are often muted [1], and the X-ray or CT findings for these patients are typically nonspecific. The use of empirical antibiotics can also reduce the sensitivity of CMTs [33]. Therefore, diagnosing bacterial infections in immunocompromised patients remains challenge. Our study results indicate that mNGS significantly outperforms CMTs in bacterial detection rates in immunocompromised patients. Compared to CMTs, the most common bacteria detected by mNGS in the immunocompromised group were Corynebacterium spp., Klebsiella spp., Pseudomonas aeruginosa, Staphylococcus spp., and MTB (all P < 0.05). In addition, the detection rates of Corynebacterium spp. and MTB was higher than immunocompetent group (all P < 0.05) (Fig. 6). As well, the detection rates of Nocardia spp., Streptococcus spp., and Enterococcus spp. were also higher (all P < 0.05). These results are consistent with findings reported in previous studies [31, 34]. We noticed in the immunocompromised group that the detection rates of Corynebacterium spp., NTM and MTB was higher than immunocompetent group (all P < 0.05). Consistent with previous studies [32, 35, 36], this result of our study indicated that immunocompromised patients may be more susceptible to infections caused by such bacteria. mNGS helps identify infections caused by these bacteria, enabling early antibiotic adjustments and ultimately improving patient prognosis.

Although the epidemiology of fungal diseases has greatly changed over the last few decades, these infections continue to predominantly affect immunocompromised patients [37]. Various pathogenic fungi, such as Aspergillus spp., Pneumocystis jirovecii, Candida spp., and Mucor, are frequently implicated in severe infections in this population. However, diagnosing these infections remains challenging due to the nonspecific symptoms and signs they present. In our study, mNGS demonstrated excellent performance in detecting fungal infections when compared to CMTs. The most commonly identified fungi were Candida spp., Aspergillus spp., and Pneumocystis jirovecii. In this study, although the detection rate of Candida spp. is the highest among fungi (Fig. 6), it is often considered non-pathogenic in clinical practice. A careful decision should be made based on a comprehensive assessment, taking into account clinical symptoms, laboratory tests, imaging findings, and other factors. Furthermore, the detection rates of Aspergillus spp. and Pneumocystis jirovecii were higher in the immunocompromised group compared to the immunocompetent group (all P < 0.05), with Pneumocystis jirovecii not detected by CMTs (Fig. 6). These findings underscore mNGS’s ability to accurately identify fungal species and particularly those infected with Aspergillus spp. and Pneumocystis jirovecii. Early diagnosis and prompt antifungal therapy can assist in guiding appropriate antimicrobial therapy in immunocompromised patients (Figs. 7A, B, 8C, D).

Viral infections can be secondary either to community or in-hospital acquisition, or to the reactivation of latent viruses [1]. Reactivation of herpes viruses in respiratory secretions is common in immunocompromised patients [38–40]. It may reactivate later in adults with more prominent and serious presentations such as central nervous system involvement. In our study, the viruses with the higher detection rates were EBV, CMV, HSV-1, and HHV-7. Among them, the detection rates of CMV and HHV-7 were higher in the immunocompromised group compared to the immunocompetent group (all P < 0.05) (Fig. 6). This is consistent with the previous study [38, 39, 41], which indicated that CMV and HHV-7 is more commonly encountered in immunocompromised patients. Unlike infections caused by other pathogens, viral reactivation in immunocompromised individuals often follows immunosuppressive treatments (e.g., prolonged corticosteroid use, chemotherapy) [42, 43], potentially leading to severe clinical manifestations. In these patients, viral reactivation is a crucial clinical consideration. Even if a viral test is positive, it is essential to evaluate the patient’s clinical symptoms, medical history, and laboratory findings to determine whether it is a reactivation of latent viruses.

The immunocompromised host-CAP guidelines proposed by Ramirez et al. [17] highlight that the immunocompromised patients are at risk for infections with multiple opportunistic pathogens. Under the condition of an experienced team, a short delay in initiating antibiotic therapy may be acceptable, provided that better culture results can be obtained through bronchoscopy combined with bronchoalveolar lavage (BAL). Therefore, in such patients, the use of bronchoscopy with BAL can provide potential benefits. In this study, we observed that, regardless of the patient’s immunosuppressive status, mNGS results were associated with a favorable clinical response. Our results found that the proportion of positive clinical impact (diagnosis) in both groups were exceeded 90% (Fig. 7A, B), similarly with results reported by Yin et al. [44] Their research confirmed that the percentages of positive impact due to mNGS results in the varying degrees of immunosuppression groups all exceeded 70%. In comparison, the antibiotic treatment plans adjusted based on mNGS results in both groups were more likely to be de-escalated rather than escalated (Fig. 8B). Further analysis revealed that when mNGS results were positive, the positive clinical impacts of mNGS (either antimicrobial therapy-related or diagnosis-related) higher than the negative impacts in both groups (P < 0.05) (Fig. 8A). However, when mNGS results were negative, the impact on antibiotic adjustment strategies in the two groups was strikingly different. Unlike the immunocompetent group, the immunocompromised group more frequently underwent antibiotic escalation (Fig. 8A). This could be related to the higher rate of mixed infections (58.82% vs 44.58%, P = 0.001) (Fig. 7D) and more frequent rate of appropriate targeted antibiotic treatment (67.65% vs 54.19%, P = 0.001) (Fig. 8B) in the immunocompromised group than in the immunocompetent group, for these patients, antibiotic escalation should still be performed when an infection is confirmed through CMTs or clinical judgment. Furthermore, our study also found that the treatment effectiveness rate in the immunocompromised group was higher than that in the immunocompetent group (65.69% vs 56.40%, P < 0.05) (Fig. 8C, D). Further analysis showed that when mNGS-guided treatment was effective, the 28-day mortality rate significantly improved in both the immunocompromised group (31.34% vs 74.29%, P < 0.05) and the immunocompetent group (42.36% vs 40.68%, P < 0.05) compared to when the treatment was ineffective. These findings suggest that mNGS can play a crucial role in improving patient outcomes, especially in immunocompromised patients when the results are used effectively. However, when mNGS-guided treatment was ineffective (such as in cases of false-negative results or misdiagnosis from mNGS), the survival rate in the immunocompromised group significantly decreased (25.71% vs 68.66%, P < 0.05), indicating that treatment adjustments need to be made based on CMTs results at that point.

Limitation

First, there is the potential variability in sample collection, processing, and handling across different centers, which could impact the accuracy and consistency of mNGS and conventional microbiological test results. Furthermore, mNGS was performed only on BALF samples, while other types of samples, such as blood, sputum, and cerebrospinal fluid, were not included. Moreover, the interpretation of mNGS results relies on clinicians’ judgment of colonizing versus pathogenic microbes. While mNGS shows great promise, its availability for routine clinical practice may still be limited in certain healthcare settings. The widespread adoption of mNGS as a first-line detection method for SCAP may depend on factors such as institutional resources, infrastructure, and regional availability, which may vary across hospitals. Finally, in this study, CMTs were selected based on the clinical condition of the patients, relying on the clinical judgment of the physicians. This approach may have introduced bias and affected the diagnostic positivity rate of CMTs, potentially impacting the comparative analysis with mNGS. To further assess the application of the mNGS in immunocompromised patients with SCAP in the ICU, future research with larger sample sizes, use standardized CMTs and more multicenter prospective studies is needed.

Conclusion

mNGS in BALF significantly improved the diagnostic rate of infections, particularly in immunocompromised patients with SCAP, thereby positively influencing the antimicrobial treatment and clinical outcomes for these patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualized: Zhao, JJ. Methodology: Zhao, JJ. Software: Zhao, JJ. Formal analysis: Zhao, JJ and Zhu Ge, RX. Resources: Fang, HL; Sun, Y; Hu, BC; Wang, YS; Wang, XX; Zhang, Y; Yuan, LM; Qiu, CH; Yan, YQ; Hua, ZD; Wang, KY; Qiu, LY; Luo, J and Zhang, WW. Data curation: Sun, Y, Tang, J and Guo, K. Original draft: Zhao, JJ. Supervision: Zhao, JJ, Fang, HL, Zhang, XJ, and Zhuge, JC. All authors read and reviewed the manuscript.

Funding

This work was supported in part by grants from the Project of Zhejiang Provincial Department of Health (No. 2023KY1296) and Quzhou Bureau of Science and Technology (2022K71).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Conflict of interests

The authors declare no competing interests.

Ethics approval and consent to participate

The retrospective studies involving humans were approved by the ethics committee of Quzhou People’s Hospital (Quzhou People’s Hospital, Quzhou, China: Number B 2024–150). The studies were conducted in accordance with local legislation and institutional requirements. Written informed consent for participation was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and institutional requirements.

Consent for publication

Not applicable.