FormalPara Keywords

severe pneumonia, cardiac surgery, metagenomic next-generation sequencing, bronchoalveolar lavage fluid, pathogen.

Introduction

Cardiac surgical patients commonly face elevated infection risks due to conditions such as pulmonary congestion, ischemia-reperfusion injury after cardiopulmonary bypass, cardiac surgery injury, hypothermia, blood loss and malnutrition1,2. These specific risks facilitated the progression of postoperative pulmonary complications. The ensuing hospital-acquired pneumonia, with its profound influence on patient outcomes, can trigger respiratory failure, multi-organ dysfunction, prompting extended intensive care unit (ICU) stays, ventilator time, and overall hospitalization, posing formidable challenges for medical management3,4,5,6,7. Regrettably, diagnosis of these patients’ infections can be complex, as clinical and lab indicators of inflammation result not only from infection but also systemic inflammatory response syndrome (SIRS) related to factors such as post cardiopulmonary bypass ischemia reperfusion injury or endotoxin displacement8,9.

Moreover, a significant reason hampering empirical treatment of pneumonia is the pathogen uncertain and mixed infection. Early diagnosis of etiology coupled with effective treatment significantly impacts patient prognosis. Precise identification of these pathogens is the cornerstone of successful anti-infection and treatment10.

Despite conventional microbiological test (CMT) serving as the gold standard for clinical microbe detection, it exhibits limitations including extended detection time, low sensitivity and some pathogenic bacteria could not be cultured8,11. These limitations often render CMTs, such as bacterial and fungal smears and cultures, insufficient for the clinical diagnosis and management of critical patients12,13. In recent years, metagenomic next-generation sequencing (mNGS) has distinguished itself in pathogen identification. This technology can detect multiple pathogens simultaneously including bacteria, viruses, and fungi with high sensitivity and specificity. Previous research has shown that this method offers a more comprehensive and accurate detection of pathogenic microorganisms compared to CMTs, thereby supporting therapeutic decisions and infection management10,14. Furthermore, an increasing number of rare pathogens are being identified through mNGS, providing critical data for the diagnosis and management of complex, severe cases15,16. Additionally, mNGS provides information for the rapid identification of potential antimicrobial resistance by detecting sequences associated with resistance genes17,18,19.

It is regrettable that despite the extensive literature comparing the diagnostic performance of mNGS with CMTs, there were few real-world clinical studies examining the impact of mNGS of bronchoalveolar lavage fluid (BALF) on clinical outcomes in patients with severe pneumonia. This is particularly true for specific subgroups such as adult patients with severe pneumonia after cardiac surgery. This study was conducted to compare a comprehensive treatment strategy based on mNGS with a standard treatment based on CMTs, evaluating the impact of mNGS of BALF on the clinical outcomes of patients with severe pneumonia after cardiac surgery.

Materials and methods

Population and study criteria

A retrospective analysis was conducted on adult patients with severe pneumonia after cardiac surgery who were admitted to the surgical intensive care unit (SICU) of Sichuan Provincial People’s Hospital between January 2019 and March 2024. The protocol used in this retrospective study was reviewed and approved by the Ethical Review Committee of Sichuan Provincial People’s Hospital (No: 2024-30). The study was carried out in accordance with the Declaration of Helsinki and all data were anonymous before analysis. After being diagnosed with severe pneumonia after cardiac surgery, patients who provided informed consent for mNGS testing, either by themselves or through their authorized representatives, were included in the mNGS group and underwent both mNGS and traditional pathogen detection. Patients who did not consent to mNGS testing were assigned to the CMT group, and their BALF samples underwent only traditional pathogen detection, including at least bacterial and fungal smears and cultures. The diagnosis of severe pneumonia, hospital-acquired pneumonia (HAP), and ventilator-associated pneumonia (VAP) was based on the criteria established by the American Thoracic Society and the Infectious Diseases Society of America (ATS/IDSA)20,21. Assess the efficacy of treatment by analyzing symptoms, physical examination findings, and laboratory test results20,21.

The inclusion criteria were as follows: (1) Adult patients (aged > 18 years) admitted to the SICU with severe pneumonia following cardiac surgery, meeting ATS/IDSA diagnostic criteria. (2) Bronchoalveolar lavage fluid (BALF) obtained via bronchoscopy within 24 h of severe pneumonia diagnosis. (3) Complete clinical data available.

The exclusion criteria were as follows: (1) Death or discharge within 48 h of ICU admission. (2) mNGS performed DNA sequencing only and could only detect DNA viruses. (3) Pulmonary lesions caused by other factors. (4) BALF samples failing mNGS quality control, such as human-origin sequences exceeding 99%. (5) Incomplete required clinical data.

Clinical data and treatment

The clinical data of each participant were collected from hospital electronic medical records, which included: (1) age, gender, past medical history and body mass index (BMI); (2) the surgical method, duration of surgery, cardiopulmonary bypass (CPB) time, aortic cross-clamp time; (3) comorbidities, Acute Physiology and Chronic Health Evaluation II (APACHE II) score, Sequential Organ Failure Assessment (SOFA) score, laboratory results. (4) the duration in which the sample resides from submission to etiology results feedback, in the NGS and CMT groups respectively. (5) prognostic information on antimicrobial use, prognostic information on mechanical ventilation, ICU hospitalization, new-onset infection, and 28-day mortality were recorded after surgery.

Before the causative pathogens were identified, clinicians formulated routine diagnoses and treatments based on established guidelines and consensus21,22,23.

we adjusted the antibiotic regimen according to the following principles:

  1. (i)

    Upon initial antibiotic therapy, if the patient’s inflammatory markers (procalcitonin, C-reactive protein, and white blood cell count) decrease, lung imaging improves, and mNGS monitoring shows negative results, then antibiotic regimen can be de-escalated or simplified. De-escalation refers to changing an empiric broad-spectrum antibiotic regimen to a narrower antibiotic regimen by changing the antimicrobial agent or changing from combination therapy to monotherapy20.

  2. (ii)

    If the patient’s clinical and laboratory indicators improve and the initial antibiotics cover the pathogen, stick to the basic antibiotic regimen.

  3. (iii)

    If the patient’s clinical and laboratory indicators did not improve and pathogens not covered by the initial anti-infective regimen were detected, we adjusted the antibiotic therapy.

  4. (iv)

    If the patient’s clinical and laboratory indicators did not improve and the microorganisms detected by mNGS may be resistant, antibiotics were adjusted based on the pharmacokinetics/pharmacodynamics and characteristics of bacterial resistance in our center. The determination of potential pathogen resistance is based on clinical symptom improvement, mNGS results for resistance genes, and the antimicrobial susceptibility patterns specific to our research center.

Sample processing and time of sample submission to results feedback

Bedside bronchoscopy was administered post informed consent by respiratory therapists (RTs). Employing standard methods, RTs harvested > 5 mL protected BALF specimens from patients and placed each sample in a sterile sputum container24,25,26,27.

The time of sample submission to results feedback was recorded based on the timestamp when the BALF samples were logged into the microbiology lab system until the validated microbiological report was issued. This duration included the time required for bacterial and fungal cultures to confirm negative results and was obtained from the hospital information system. For the CMT group, the time of sample submission to results feedback refers to the final report time of the CMTs.

Conventional microbiological test

These were dispatched independently to the microbiology laboratory for routine evaluation, such as bacterial and fungal smears and cultures, acid-fast staining, and polymerase chain reaction for specimens from patients exhibiting suspected viral infection.

CMTs encompass bacterial and fungal smear and culture, acid-fast staining, Gomori methenamine silver staining for Pneumocystis, India Ink Staining Smear Microscopy for Cryptococcus, mycobacterial culture and identification, Gen-Xpert, as well as serological tests of blood including the (1–3)-β-D-Glucan Test (G test), Galactomannan antigen test (GM test), cryptococcal capsule antigen detection, and Mycoplasma pneumoniae antibody detection. Additionally, nasopharyngeal swab and bronchoalveolar lavage fluid PCR are used to detect respiratory pathogens, such as Klebsiella pneumoniae, Acinetobacter baumannii, Influenza A and B viruses, Respiratory Syncytial Virus, and Parainfluenza viruses (Supplementary Information 2, Tables S2.1-S2.2).

Metagenomic next-generation sequencing procedure

Perform wall breaking and centrifugation on BALF samples according to standard procedures. Subsequently, the 600 µL sample was separated into new 1.5-mL microcentrifuge tubes, and the DNA/RNA was extracted using the nucleic acid extraction or purification reagent kit (Genskey, Tianjin) following the manufacturer’s instructions. DNA libraries were constructed through DNA fragmentation, end repair, adapter ligation, and PCR amplification. The Agilent 2100 system (Agilent Technologies, Santa Clara, USA) was used for quality control of the DNA libraries. The quality qualified libraries were sequenced on the MGI-200/2000 platform, with a sequencing length of SE50 and a sequencing data volume of no less than 20 M Reads.

Bioinformatics analysis

The raw data were split using bcl2fastq software, and the connector sequences and low-quality base sequences were removed using fastp software to obtain high-quality and effective data. Sequences from the human genome were removed using the bowtie2 calibration software. Eventually, sequences that could not be mapped to the human genome were retained and the rest of the sequences were aligned to the microbial genome database, which was constructed using the sequences of bacteria, fungi, archaea, and viruses screened in the NCBI database. The microbial genome database contains 12,895 bacteria, 11,120 viruses, 1,582 fungi, 312 parasites, 184 mycoplasmas/chlamydia, and 177 mycobacteria.

Statistical analysis

All statistical analyses were performed using GraphPad Prism 9.0.0(121) (GraphPad Software, LaJolla, CA, USA). The normally distributed data are expressed as mean ± standard deviation (SD) and were compared between groups by t-testing. Nonnormally distributed data are presented as median and interquartile range (IQR) and were compared by Mann Whitney test. P values of < 0.05 were considered to denote the statistical significance.

Results

Demographic characteristics

From January 2019 to March 2024, our institution conducted 4,064 cardiac surgeries, as illustrated in Fig. 1. Among these surgeries, we identified 162 adult patients with severe pneumonia for potential inclusion in the study. Following the application of our exclusion criteria, 49 patients were deemed ineligible (Fig. 1). The reasons for exclusion included: 17 patients either passed away or were discharged within the first 48 h post-admission to the SICU; 30 patients underwent mNGS that was limited to DNA sequencing, thus only capable of detecting DNA viruses; and 2 cases where the pulmonary lesions were determined to be due to alternative etiologies (Fig. 1). Finally, this study included a total of 53 patients in the mNGS group, with an additional 60 patients allocated to the CMT group (Fig. 1). Patient demographics, characteristic baselines, and ICU special treatments in the mNGS and CMT groups were shown in Table 1 and Supplementary Information 3. Table S3.1. There were no significant differences in age, gender, BMI, types of cardiac surgery, duration of surgery, CPB time, aortic cross-clamp time between both groups (P > 0.05). It is worth noting that the mNGS group had a lower oxygenation index and higher APACHE II, SOFA, and vasoactive inotropic scores (VIS) compared to the CMT group before treatment (P<0.05).

Comparison of mNGS and CMT results

shows the distribution of pathogens detected by mNGS and CMT in this study. The most commonly detected pathogenic bacteria were Klebsiella pneumoniae in both groups, Acinetobacter baumannii complex being ranked second. Among the 53 patients in the mNGS group, a total of 53 mNGS tests were conducted, with 52 (98%) testing positive for mNGS, with only one patient having a false-negative result (Fig. 2). Conversely, in the CMT group, only 35 out of 60 patients (58%) had positive results (Table 2, P < 0.0001). Additionally, the total number of pathogens identified in the mNGS group was significantly higher compared to the CMT group (P < 0.0001). Furthermore, Pneumocystis jirovecii, Aspergillus flavus, Mycoplasma salivarium, Mycobacteroides abscessus, Human betaherpesvirus 7 and Human alphaherpesvirus 1 were detected only in the mNGS group. (Additional information: The distribution and resistance patterns of traditional cultured pathogens for all ICU patients during the study period are detailed in Supplementary Information 1.)

Treatment and prognosis

The etiology results and treatments of patients in the mNGS and CMT groups are compared in Table 2. The median time from sample submission to results feedback was significantly shorter in the mNGS group (22 h, IQR 18.0-24.5) compared to the CMT group (67 h, IQR 56.5–83.0; P < 0.0001). Antibiotic adjustments were more frequently done in the mNGS group with 62% versus 33% in the CMT group (P = 0.0021). De-escalation or simplification of antibiotic therapy was performed in 15% of cases in the mNGS group as against 3% in the CMT group, which was statistically significant (P = 0.0280). Furthermore, new antibiotics were added or the current antibiotics were upgraded in 47% of the mNGS group compared to 25% of the CMT group, also showing a significant difference (P = 0.0139).

After adjusting the treatment plan based on microbial testing results, the mNGS group showed an improvement in the APACHE II score (P = 0.0161) and SOFA score (P = 0.0076) on the 7th day after admission to the SICU (Table 3). This suggests a better physiological status and organ function at 7 days post-intervention in the mNGS cohort. Additionally, the mNGS group showed an improvement in ventilator-free days within 28 days (Table 3; P = 0.0475), with a shorter duration of invasive ventilation compared to the CMT group (Table 3; P = 0.0208).

In this study, the mNGS group showed signs of having a positive impact on length of stay in ICU (9 vs. 10, P = 0.0538), length of stay in Hospital (20 vs. 25, P = 0.1558), mortality in 28 days (19% vs. 20%, P = 0.8794), in-hospital mortality (19% vs. 22%, P = 0.7123); however, statistical analysis did not confirm these differences to be significant (Table 3).

Discussion

The pathogenic spectrum of bronchoalveolar lavage fluid

In this study, mNGS was employed to identify the causative pathogens of pneumonia following cardiac surgery. For patients who have undergone cardiac surgery, our study revealed that Gram-negative bacteria, particularly Klebsiella pneumoniae, Acinetobacter baumannii complex, Pseudomonas aeruginosa, Haemophilus influenzae and Escherichia coli are the major causes of HAP (Fig. 2), which is consistent with previous studies that utilized traditional culture methods28,29. This indicates that Gram-negative bacteria may be the first pathogen to be considered after cardiac surgery. It is noteworthy that the application of mNGS technology has provided a deeper insight into the microbial landscape of post-cardiac surgery pneumonia, uncovering rare novel pathogens that might elude detection through traditional culture methods, such as Pneumocystis jirovecii, Cryptococcus neoformans, and viruses (Fig. 2). This advancement suggests that mNGS could serve as a valuable complement to conventional diagnostic approaches, potentially leading to more accurate and timely interventions.

The impact on clinical outcomes

This study, the first of its kind, was conducted to compare a comprehensive treatment strategy based on mNGS with a standard treatment based on CMTs, evaluating the impact of mNGS of BALF on the clinical outcomes of patients with severe pneumonia after cardiac surgery. Our study has provided evidence of clinical improvement effect of mNGS with a BALF sample in adult patients with severe pneumonia after cardiac surgery.

First, mNGS was faster. The median time from sample submission to result feedback was significantly shorter in the mNGS group (22 h, IQR 18.0-24.5) compared to the CMT group (67 h, IQR 56.5–83.0; P < 0.0001). In addition, mNGS demonstrated excellent sensitivity for pathogen detection in adult patients with severe pneumonia after cardiac surgery. The significant difference in positive detection rates between mNGS and CMT methods underscores the potential of mNGS to revolutionize the diagnostic process for infectious diseases in the clinical setting. Adjusting the treatment plan based on microbial testing results may be associated with improved early clinical outcomes compared to CMT, as evidenced by statistically significant results in critical early markers, such as APACHE II and SOFA scores at 7 days post-intervention, and respiratory support requirements. However, these benefits did not translate into statistically significant differences in mortality rates or the overall length of hospital stay. This finding is consistent with the results of Yang et al. 2022, who reported a similar improvement in severe hospital-acquired pneumonia but no improvement in medium- or long-term survival, especially in patients with severe comorbidities30. Additionally, a recent multicenter randomized controlled trial (RCT) investigating the impact of mNGS on clinical outcomes in ICU patients with severe community-acquired pneumonia (SCAP) demonstrated that mNGS combined with CMTs significantly reduced the time to clinical improvement within 28 days compared to CMTs alone31. These results suggest that mNGS can enhance clinical management and outcomes in critically ill patients by enabling more accurate and timely adjustments to antimicrobial therapy. However, our findings appear to contrast with those reported by Zhang et al. 2020 and Xie et al.2019, who found that the 28-day mortality rate of the NGS group was significantly lower than that of the CMT group (P < 0.05)32,33.

Several factors might account for the discrepancies observed between these studies. First, given that this is a retrospective study, the baseline values of the mNGS group were associated with higher APACHE II and SOFA scores, as well as increased respiratory support efforts. Previous studies have shown that the higher the APACHE II and SOFA scores of sepsis patients, the worse the prognosis, which may potentially mitigate the benefits derived from mNGS in this study34,35,36. Second, when discussing the current research landscape in this field, it is crucial to acknowledge that existing studies are predominantly retrospective in nature. They are characterized by small sample sizes and significant heterogeneity among study populations. This heterogeneity presents a particular challenge in drawing broad conclusions. Specifically, our study focused on patients who developed severe pneumonia after cardiac surgery, in contrast to other studies which have included all patients with severe pneumonia regardless of the underlying cause. This distinction is critical as it suggests that findings from our research might reflect unique aspects of postoperative complications and management, differentiating it from the broader category of severe pneumonia. Further studies with larger patient populations and different clinical settings are necessary to validate these findings and assess their generalizability.

Pathogenic determination and clinical decision-making with mNGS

The mNGS was previously recognized as too sensitive for identifying microbes, and thus could lead to overuse of antibiotics. On the other hand, mNGS does not allow us to precisely differentiate between microbiome, colonizers, flora and pathogenic overgrowth31. However, mNGS results must be interpreted alongside the relative abundance of detected organisms, epidemiological, and clinical characteristics to accurately identify the causative microorganism.

Among the patients tested, Human betaherpesvirus 7 was detected in 13.3% of cases, Epstein-Barr virus in 10.6%, and Human betaherpesvirus 5 in 8.8%. However, interpreting the detection of Human betaherpesvirus 7, Epstein-Barr virus, and Human betaherpesvirus 5 requires caution to prevent overtreatment. Our study employed a combination of clinical correlation, quantitative analysis, and immune status assessment methods to distinguish between active infections and bystander activation37.

In this study, we found that in the mNGS group, 47% of patients required the addition or escalation of antibiotics. This high percentage is primarily attributed to the detection of various potentially resistant pathogens and rare microorganisms. Specifically, newly identified pathogens not covered by the initial antibiotic spectrum included Aspergillus fumigatus (4 cases), Enterococcus faecium (2 cases), Corynebacterium striatum (1 case), Aspergillus flavus (1 case), Cryptococcus neoformans (1 case), Legionella pneumophila (1 case), Mycoplasma pneumoniae (1 case), and Pneumocystis jirovecii (1 case). Additionally, potential antimicrobial resistance genes were noted in pathogens such as Klebsiella pneumoniae (4 cases), Acinetobacter baumannii complex (3 cases), Stenotrophomonas maltophilia (3 cases), and Pseudomonas aeruginosa (3 cases). The high sensitivity and specificity of mNGS facilitated the rapid detection of these pathogens, which might have been delayed or missed by CMT. This rapid optimization of anti-infective treatment resulted in significant improvements in the 7-day APACHE II score, SOFA score, and 28-day ventilator-free days in the mNGS group. Similar findings have been reported in previous studies. For instance, a study on severe pneumonia following pediatric congenital heart surgery demonstrated that the number of antibiotic adjustments in the mNGS group was significantly higher than that in the CMT group (70% vs. 35%, P < 0.05), and the proportion of pulmonary infections that improved after adjustment was also significantly higher (63% vs. 28%, P < 0.05). Within this, 45% of patients in the mNGS group required the addition or escalation of antibiotics, compared to only 21% in the CMT group38. After receiving microbiological test results and adjusting for antibiotics, pulmonary infection improved in 27/43 patients (63%) in the mNGS group compared with 11/40 patients (28%) in the CMT group in the ensuing 7 days (P = 0.032)38.

The decision-making process regarding which organisms identified by mNGS are pathogenic and require treatment changes involves several key considerations. Clinical correlation with the patient’s symptoms, radiographic findings, and laboratory markers of infection is essential. Quantitative thresholds help distinguish between colonization and infection, minimizing overtreatment11. Host factors such as immunocompetence, presence of invasive devices, and recent antibiotic exposure are crucial in assessing infection risk39,40,41. Interdisciplinary consultations support comprehensive therapeutic decisions42. Future research should standardize mNGS data interpretation and validate findings through large-scale studies.

The impact on antibiotic stewardship and resistance

In our study, the implementation of a comprehensive treatment strategy based on mNGS facilitated the rapid identification of a wide range of pathogens. This approach contributed to targeted therapy and resulted in a significant increase in antibiotic adjustments within the mNGS group, including de-escalation, escalation, and the initiation of new antibiotics (Table 2). Adjustment of antibiotics based on accurate identification of pathogens may account for the improved outcomes, such as the increase in ventilator-free days within 28 days and the reduction in the duration of invasive ventilation. Another multicenter RCT on SCAP also showed that interpreting mNGS pathogen results through rigorous laboratory and informatic methods, involving expert panel discussions and adjudications, helps to distinguish between contaminants and true pathogens accurately. This leads to a higher rate of antibiotic de-escalation in the mNGS group compared to the CMT group, particularly in immunocompetent patients31. Such an approach is particularly significant in the context of antimicrobial resistance, as it allows de-escalation from broad-spectrum antibiotics to more targeted treatments, thereby reducing the likelihood of developing resistant strains.

Economic implications and health resource utilization

Multiple investigations have consistently identified postoperative complications as significant contributors to the elevated expenses associated with cardiac surgical procedures43. Among these, prolonged mechanical ventilation has been identified as the foremost factor contributing to the surge in complication-related expenditures, surpassing the economic impact of both cerebrovascular accidents and deep sternal wound infections44,45. Our study design focused on the impact of mNGS on clinical outcomes. Consequently, we did not include an economic analysis of the use of mNGS in the diagnosis and management of postoperative severe pneumonia in cardiac surgery patients. However, the potential impact of mNGS on healthcare economics warrants further discussion. The initial cost of mNGS is higher compared to CMT; however, the downstream financial benefits, such as reduced length of intensive care unit stay and lower readmission rates due to targeted treatment, could offset the upfront costs. Future research should include a comprehensive economic analysis to elucidate these potential savings.

Limitation

Our research also has certain limitations. Firstly, the retrospective design of our study introduces potential biases, especially in patient selection for mNGS testing. The higher initial testing costs associated with mNGS mean that it is more likely to be used in patients with more severe clinical presentations. This has resulted in baseline disparities between the patient groups undergoing mNGS and those receiving CMT, potentially confounding the comparison of outcomes between these groups. Second, our study was conducted at a single center and focused specifically on a relatively narrow patient population: individuals who developed severe pneumonia after cardiac surgery. This specificity, while valuable for targeted insights, results in a smaller sample size. The limited sample size poses challenges to the statistical robustness of our findings, necessitating a cautious interpretation of the data.

Conclusion

The study highlights the clinical utility of mNGS in diagnosing pathogens in adult patients with severe pneumonia following cardiac surgery. Our research suggests that mNGS could serve as a valuable complement to conventional diagnostic approaches, potentially improving diagnostic accuracy and leading to more precise and timely interventions, with significant potential to inform clinical decision-making and enhance patient outcomes. Further research should include a larger sample size, involving multi-center, prospective, and controlled studies, which will enhance our understanding of the impact of mNGS on the medium or long-term survival rate, economic implications and health resource utilization in this specific patient population.

Table 1 Patient characteristics and baseline of two groups.
Full size table
Table 2 The etiology results and treatment of patients in mNGS and CMT groups.
Full size table
Table 3 Comparison of outcomes between two groups.
Full size table
Fig. 1
figure 1

Flow diagram of patient inclusion and exclusion. mNGS, metagenomic next-generation sequencing; CMT, conventional microbiological test.

Full size image
Fig. 2
figure 2

The pathogenic spectrum of bronchoalveolar lavage fluid detected by mNGS and CMT in this study. The number on the vertical axis in the histogram represents the corresponding number of cases of the pathogen. mNGS, metagenomic next-generation sequencing; CMT, conventional microbiological test.

Full size image