Application of Metagenomic Next‐Generation Sequencing in the Diagnosis of Pneumonia in Patients With Cancer

ABSTRACT

Background

With the development of new sequencing technologies, metagenomic next‐generation sequencing (mNGS) has become a diagnostic tool for respiratory tract infections. Patients with cancer may develop pneumonia caused by infections or antitumor therapy. Therefore, pneumonia in patients with cancer is more complex than that in healthy individuals. Currently, few reports are available on the use of mNGS for diagnosing pneumonia in patients with cancer.

Methods

In this retrospective study, 14 patients with cancer diagnosed with pneumonia in March 2023 were enrolled from the Emergency Department of the Chinese Academy of Medical Sciences Cancer Hospital. Sputum samples from the patients were examined using conventional tests and mNGS to identify pathogens. The mNGS and conventional test results were compared to assess the diagnostic yield in patients with cancer.

Results

The overall pathogen detection rate of mNGS was 64.29% (9/14), with corresponding diagnostic sensitivity, specificity, false‐negative rate and false‐positive rate of 90.00%, 25.00%, 10.00% and 75.00%, respectively. Among 13 paired sputum specimens, mNGS exhibited a numerically higher pathogen detection rate (61.54%, 8/13) than conventional diagnostic assays (38.46%, 5/13). McNemar's paired chi‐square test demonstrated no statistically significant difference between the two detection methods (p = 0.37), and Kappa concordance analysis generated a coefficient of 0.27 (p = 0.23), suggesting poor inter‐method consistency. Compared with conventional tests, mNGS detected additional pathogens in 8 specimens and identified a greater number of pathogens in 9/14 (64%) samples. Moreover, mNGS results led to diagnostic revisions and subsequent antimicrobial therapy adjustments in 64% (9/14) of enrolled patients. Additionally, mNGS detected antibiotic resistance genes in five patients, which provided guidance for antibiotic selection.

Conclusions

Metagenomic next‐generation sequencing (mNGS) showed potential value in pathogen detection, as it appeared to identify pathogens more rapidly and comprehensively than conventional methods. It may provide auxiliary support for the diagnosis and treatment of pneumonia in this vulnerable population.

1. Introduction

Patients with cancer have an increased risk of pneumonia and poor prognosis due to systemic immunosuppression from the malignancy and cancer treatments, such as chemotherapy and surgery [1]. In patients with cancer, the incidence of pneumonia is further following immunotherapy or radiotherapy [2, 3]. Approximately 10% of hospital admissions of patients with cancer are due to or complicated by pneumonia, particularly in patients with hematologic malignancies, in which the risk of pneumonia during treatment is estimated to be over 30% [4, 5, 6]. Pneumonia in patients with cancer is relatively more complicated and is more likely to involve mixed infections with multiple pathogens and the emergence of uncommon drug‐resistant organisms [7, 8]. Consequently, the incidence of severe pneumonia and the pneumonia case fatality rates are higher in patients with cancer than in other patients with pneumonia. In addition, the occurrence of immunotherapy‐ and radiotherapy‐associated pneumonia increases the difficulty in diagnosing infectious pneumonia in patients with cancer, affecting the choice of antimicrobial agents and prognosis.

Pneumonia can be caused by a variety of pathogens, including bacteria, viruses, mycoplasma, and fungi, which are difficult to differentiate clinically. Traditional pathogen detection methods, such as bacterial and fungal smears and cultures, PCR, and antigen detection, are time‐consuming and inefficient. The cause of community‐acquired pneumonia remains undetermined in up to 62% of cases using a combination of traditional diagnostic tests [9]. When traditional testing methods show negative results, patients are often administered empirical antibiotics, which can lead to exacerbation of the infection and misuse of broad‐spectrum antibiotics. Early and targeted antimicrobial treatment can reduce mortality from pneumonia [10].

Metagenomic next‐generation sequencing (mNGS) is a new tool that may overcome the shortcomings of traditional diagnostic methods [11, 12]. mNGS directly sequences all nucleic acid fragments in samples to simultaneously identify all potentially infectious microorganisms. In addition to pathogen identification, mNGS provides genomic information necessary for airway microbiome analysis, human host response analysis, and drug resistance prediction [13]. mNGS also plays a critical role in the diagnosis of pneumonia caused by difficult‐to‐identify pathogens and pneumonia caused by multiple pathogens [14]. In this study, we aimed to assess the value of mNGS in the diagnosis of pneumonia in patients with cancer and differentiation between infectious pneumonia and pneumonia associated with anticancer therapy.

2. Methods

2.1. Study Design and Patient Cohort

This was a retrospective study of 14 patients with cancer and pneumonia, who were admitted to the Emergency Department of the Cancer Hospital of the Chinese Academy of Medical Sciences during March 2023. Given the small sample size and lack of a comparator group, the study design is best characterized as a case series. The cancer type and stage, and history of antitumor therapy were recorded for each patient. Pneumonia was diagnosed based on the clinical presentation, blood tests, microbiological tests, and chest computed tomography (CT). The study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Medical Ethical Committee of the Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College (reference number: 24/351‐4631). All participants were informed of the purpose of the study and provided signed informed consent. All methods were performed in accordance with relevant guidelines and regulations.

2.1.1. Clinical Presentation

The presentation of pneumonia was characterized by acute onset of lower respiratory symptoms, such as fever, cough, pleurisy, dyspnea, and increased sputum production, with consistent radiographic imaging findings. All patients had cough and sputum production.

2.1.2. Imaging

Radiographic imaging features of lung parenchymal involvement are the gold standard for the diagnosis of pneumonia. All patients underwent chest CT imaging.

2.1.3. Blood Tests

All patients underwent routine blood tests and measurement of their blood C‐reactive protein and procalcitonin levels.

2.1.4. Microbiological Testing

Specimens for conventional tests and mNGS were collected prior to initiating antimicrobial therapy. The conventional tests included sputum bacterial and fungal smear and culture, and PCR tests for respiratory pathogens, including influenza A, influenza B, respiratory syncytial virus, parainfluenza virus, rhinovirus, metapneumovirus, adenovirus, bocavirus, Mycoplasma pneumoniae , and Chlamydia pneumoniae . All patient samples underwent sputum bacterial and fungal smear and culture. Due to the limited clinical resources and the retrospective nature of the study, PCR testing was only performed for patients with suspected viral infection based on clinical manifestations and laboratory indicators, which led to only 2 patients receiving PCR testing.

For mNGS analysis, 1–4 mL of sputum sample was liquefied using 0.1% dithiothreitol (DTT) at room temperature for 30 min. After receiving the samples, the laboratory performed nucleic acid extraction, library construction, high‐throughput sequencing, bioinformatics analysis, and pathogen data interpretation based on previous studies [15]. DNA was extracted from the samples using the QIAamp DNA Microbiome Kit (Cat#51704, Qiagen, Hilden, Germany), and RNA was extracted using the QIAamp Viral RNA Mini Kit (Cat#52904, Qiagen). The extracted RNA was reverse transcribed using random primers, and cDNA was pooled with DNA from the same clinical sample for sequencing library preparation. The pooled nucleic acid was enzymatically fragmented to a size of 200–300 bp, and sequencing libraries were constructed through end repair, adapter ligation, and PCR amplification. Sequencing templates were prepared using the OneTouch2 System (Life Technologies, Carlsbad, CA, USA) and sequenced using a BioelectronSeq 4000 sequencer (CapitalBio Corporation, Beijing, China) after quality control. A negative control water sample was used in each run to monitor for potential contamination. The original sequencing data were subjected to quality control, and reads with lengths of less than 50 bp, low quality, or low complexity were removed. The remaining high‐quality sequencing data were mapped to the human reference genome grch38 to deplete the human host sequences using Bowtie2 software. Subsequently, the non‐human sequences were classified by simultaneous alignment to the genomic sequence databases downloaded from the US National Center for Biotechnology Information (NCBI) and Pathosystems Resource Integration Center (PATRIC) databases, which contained data of 13,992 bacterial species, 1659 fungal species, 13,000 virus species, and 287 parasite pathogens. To identify the suspected pathogens in clinical samples, the data of different types of samples from healthy people were reviewed and the relevant reference values were calculated, including the number of reads and coverage of all bacteria, fungi, viruses, and parasites detected. Pathogens detected in the negative control samples were excluded from the results of clinical samples. The final pathogen detection results included a list of suspected pathogens, the number of reads, and genome‐level coverage statistics.

2.1.5. Definition of Pathogenic Microorganisms

A microorganism was defined as pathogenic if any one of the following criteria was fulfilled:

1. Typical pathogens (e.g., Streptococcus pneumoniae , Haemophilus influenzae , Klebsiella pneumoniae ).

2. High read counts and high relative abundance.

3. Consistent clinical, imaging, and inflammatory findings.

4. Favorable response to targeted antimicrobial therapy.

5. Exclusion of other pathogens.

Otherwise, it was regarded as colonization or contamination.

2.1.6. Clinical Treatment

All patients were initially treated with empirical antimicrobial therapy according to the Chinese Adult Community‐Acquired Pneumonia Diagnosis and Treatment Guide [16]. In patients with suspected radiation pneumonitis, the initial treatment was based on the Chinese expert consensus on the diagnosis and treatment of radiation pneumonitis [17]. The antimicrobial treatment regimen was adjusted according to the pathogens detected by conventional tests and mNGS. The assessment of effectiveness was based on the improvement in clinical manifestations and was assessed as effective or ineffective.

2.1.7. Clinical Outcome Definition

The clinical outcome of pneumonia was defined as “improved” based on a comprehensive evaluation of three criteria:

1. Clinical symptoms: resolution or significant relief of fever, cough, dyspnea, or chest pain;

2. Laboratory markers: normalization or ≥ 50% reduction in white blood cell count, C‐reactive protein, or procalcitonin levels compared with baseline;

3. Chest imaging: improvement or resolution of pulmonary infiltration, consolidation, or ground‐glass opacity on follow‐up chest computed tomography.

2.2. Statistical Analysis

SPSS 26.0 was used for data analysis (categorical data as n, %; α = 0.05, p < 0.05 was considered statistically significant). The analysis included two parts: ① Independent diagnostic efficacy of sputum mNGS: clinical comprehensive diagnosis, which combined clinical manifestations, imaging results, laboratory tests, therapeutic response, and follow‐up data, was used as the gold standard. A 2 × 2 contingency table was constructed to calculate true positive (TP), false positive (FP), false negative (FN), true negative (TN), and relevant efficacy indices, with 95% CI, calculated by the exact probability method for small sample size. ② Comparison of detection rates: a paired 2 × 2 contingency table was built for 13 paired samples; McNemar's test compared detection rate differences, and Kappa test evaluated consistency (Kappa ≥ 0.75 = good, 0.40 ≤ Kappa < 0.75 = moderate, Kappa < 0.40 = poor).

3. Results

3.1. Patient Characteristics

The 14 patients included 9 males and 5 females, with a mean age of 65 years (range, 49–84 years). Among the 14 patients, 10 had lung cancer, 2 had esophageal cancer, 1 had gastric cancer, and 1 had lymphoma (Table 1). Eleven patients had received antitumor therapy such as surgery, chemotherapy, radiotherapy, targeted therapy, and immunotherapy before the onset of pneumonia. The clinical manifestations, blood test results, and chest CT findings of the patients are shown in Table 1.

TABLE 1.

Clinical characteristics of the patients enrolled in this study.

Patient	Age	Gender	Cancer type	Stage of cancer	Anticancer treatment before enrollment	Time interval from the last anticancer therapy to the onset of pneumonia symptoms, days	Body temperature	Phlegm color	CRP	PCT	WBC	N%	CT
#1	58	M	Esophageal cancer	III	Radiotherapy	45	Fever	yellow	high	normal	high	high	patchy
#2	67	M	Gastric cancer	IV	None	—	Fever	white	high	high	high	high	patchy
#3	55	M	Lung cancer	III	None	—	Fever	white	normal	normal	high	high	patchy
#4	69	F	Esophageal cancer	III	Radiotherapy	During radiotherapy	Fever	white	high	normal	high	high	patchy
#5	64	M	Lung cancer	I	Lung surgery	7	Fever	white	high	high	low	normal	patchy
#6	67	M	Lung cancer	IV	Chemotherapy and immunotherapy	10	Fever	white	high	normal	low	high	patchy
#7	74	M	Lung cancer	IV	None	—	Fever	yellow	high	high	high	high	patchy
#8	67	F	Lymphoma	IV	Chemotherapy and targeted therapy	7	Fever	yellow	high	normal	low	low	patchy
#9	66	F	Lung cancer	II	Chemotherapy and immunotherapy	19	Normal	white	high	normal	high	high	patchy
#10	59	M	Lung cancer	III	Chemotherapy and immunotherapy	15	Normal	white	high	normal	low	normal	patchy
#11	84	M	Lung cancer	II	Radiotherapy	55	Normal	white	high	normal	high	high	patchy
#12	49	F	Lung cancer	IV	Chemotherapy and targeted therapy	105	Normal	white	high	high	high	high	patchy
#13	57	M	Lung cancer	III	Radiotherapy	15	Normal	white	normal	normal	normal	normal	patchy
#14	75	F	Lung cancer	IV	Chemotherapy and immunotherapy	4	Normal	white	normal	normal	normal	normal	patchy

Note: Normal range: CRP 0.0–0.6 mg/dL; PCT: < 0.5 ng/mL; WBC: 3.5–9.5 × 10^/L; N%: 40.0.

Abbreviations: CRP, C‐reactive protein; CT, computed Tomography; N%, neutrophil percentage; PCT, procalcitonin; WBC, White blood cell.

3.2. Pathogens Detected by mNGS and Conventional Methods

A total of 14 cancer patients with clinically confirmed pneumonia were initially enrolled. The pathogens detected by different assays in the 14 patients are summarized in Table 2.

TABLE 2.

Pathogens detected in patient samples by different methods.

Patient	Sputum culture		PCR	mNGS
	Bacterium	Fungus	Virus	Bacterium	Fungus	Virus
#1	Klebsiella pneumoniae	Candida albicans	Not tested	Bilophila wadsworthia	Negative	Human Herpesvirus 4 (EB virus) Human Herpesvirus 1
#2	Negative	Candida albicans	Not tested	Staphylococcus epidermidis Bacteroides heparinolyticus	Candida albicans Candida glabrata	Negative
#3	Negative	Mold genus	Influenza A	Streptococcus pneumoniae	Saccharomyces cerevisiae	Influenza A Human Herpesvirus 4 (EB virus)
#4	Negative	Negative	Not tested	Negative	Negative	Negative
#5	Chryseobacterium indologenes	Negative	Not tested	Negative	Negative	PBV virus
#6	Negative	Negative	Not tested	Streptococcus pneumoniae	Negative	Human Herpesvirus 4
#7	Stenotrophomonas maltophilia, Staphylococcus haemolyticus	Negative	Not tested	Klebsiella pneumoniae	Negative	Papillomavirus type 8
#8	Negative	Candida albicans	Influenza A	Gordona bronchialis	Candida albicans Candida parapsilosis	Influenza A virus
#9	Negative	Candida albicans	Not tested	Negative	Candida albicans	Human Herpesvirus 4 (EB virus)
#10	Negative	Negative	Not tested	Negative	Negative	Molluscum contangiosum virus, PBV virus
#11	Negative	Negative	Not tested	Negative	Negative	Negative
#12	Not tested	Not tested	Not tested	Sphingomonas paucimobilis, Acinetobacter baumannii, Loprene gordense, Lactobacillus rhamnosus	Candida albicans	Influenza A virus, Human Herpesvirus 4 (EB virus)
#13	Negative	Negative	Not tested	Streptococcus pneumoniae, Staphylococcus epidermidis	Negative	Negative
#14	Pseudomonas aeruginosa	Negative	Not tested	Streptococcus pneumoniae, Pseudomonas aeruginosa	Negative	Negative

3.3. Diagnostic Efficacy of Sputum mNGS

With clinical comprehensive diagnosis as the gold standard for determining pneumonia‐causing pathogens, a 2 × 2 contingency table for sputum mNGS detection results was constructed, with the detailed results shown in Table 3.

TABLE 3.

Diagnostic efficacy of sputum mNGS (n).

	Infection positive	Infection negative	Total
mNGS Positive	9 (TP)	3 (FP)	12
mNGS Negative	1 (FN)	1 (TN)	2
Total	10	4	14

Abbreviations: Negative, no pathogenic bacteria detected, or only colonizing/polluting bacteria detected; Positive, clinical pathogenic pathogens detected.

Based on the Table 3, the diagnostic efficacy indicators and their corresponding 95% CI, of sputum mNGS were calculated as follows: the pathogen detection rate was 64.29% (9/14); the diagnostic sensitivity was 90.00% (9/10, 95% CI: 55.47%–99.75%); the diagnostic specificity was 25.00% (1/4, 95% CI: 0.63%–80.59%); the false negative rate was 10.00% (1/10, 95% CI: 0.25%–44.53%); and the false positive rate was 75.00% (3/4, 95% CI: 19.41%–99.37%). These results suggested that sputum mNGS had high sensitivity with a low probability of pathogen missed detection. However, its relatively low specificity made it susceptible to interference from respiratory tract colonizing flora, leading to false positive results. Due to the small sample size of this study, the 95% CI, of some indicators was wide; expanding the sample size in subsequent studies could further narrow the interval and improve the stability of the results.

3.4. Comparison of Pathogen Detection Rates Between Sputum mNGS and Conventional Methods

One patient was excluded from the paired comparative analysis due to lack of conventional microbiological testing, and 13 patients were finally included for statistical analysis. The paired 2 × 2 contingency table was summarized in Table 4.

TABLE 4.

Paired comparison of pathogen detection results between conventional methods and sputum mNGS (n).

	Conventional tests positive	Conventional tests negative	Total
mNGS Positive	4	4	8
mNGS Negative	1	4	5
Total	5	8	13

Note: One patient without conventional testing excluded (analyzed n = 13).

Among the 13 paired samples, the pathogen detection rate of sputum mNGS was 61.54% (8/13), which was significantly higher than that of conventional detection methods (38.46%, 5/13). McNemar's paired chi‐square test revealed no statistically significant difference in pathogen detection rates between the two assays (p = 0.37). The Kappa consistency test showed a Kappa value of 0.27 (p = 0.23), indicating poor consistency between the two detection methods with no statistical significance and a strong complementary relationship between them.

3.5. Comparison of Pathogens Detected by mNGS and Conventional Methods

The mNGS assay detected more pathogens, including several fastidious and occult pathogens such as Bilophila wadsworthia , Gordonia species, Sphingomonas paucimobilis , Ralstonia mannitolilytica , opportunistic Gram‐negative bacilli, fungi, and latent herpesviruses that were rarely detected by conventional assays (Figure 1).

FIGURE 1.

Comparison of pathogens detected by mNGS and conventional tests in all 14 patients.

The sensitivity of mNGS and conventional tests for detecting co‐infecting pathogens was compared. Conventional tests detected four cases of single pathogens, four cases of two pathogens, and no cases of three or more pathogens (Figure 2A). In contrast, mNGS detected one case of a single pathogen, six cases of two pathogens, and five cases of three or more pathogens (Figure 2B).

FIGURE 2.

(A) Number of patients with different numbers of pathogen types detected by conventional tests (SC or PCR). (B) Number of patients with different numbers of pathogen types detected by mNGS.

We analyzed the consistency of the mNGS results with those of pathogens identified using conventional microbiological methods (sputum culture and PCR). When mNGS identified the same pathogens as conventional tests, the results were considered to match. When mNGS identified more pathogens than conventional tests, the results were considered inconclusive. When the pathogens identified using the two methods were completely different, the results were considered mismatched. One patient was excluded due to the absence of conventional microbiological tests, and 13 patients were finally included for comparative analysis. The pathogens identified were matched, inconclusive, and mismatched in two (15.4%), eight (61.5%), and three (23.1%) patients, respectively (Figure 3).

FIGURE 3.

Consistency of mNGS results with conventional test results.

As mentioned above, the pathogens detected by mNGS and conventional assays were completely discrepant in 3 patients.

For Patient #1, sputum culture yielded Klebsiella pneumoniae and Candida albicans, whereas mNGS detected Bilophila wadsworthia and was negative for fungi, but additionally identified Human Herpesvirus 1 and Human Herpesvirus 4 (Epstein–Barr virus, EBV).

For Patient #5, sputum culture was positive for Chryseobacterium indologenes , while mNGS was negative for bacteria and fungi but positive for PBV virus.

For Patient #7, sputum culture recovered Stenotrophomonas maltophilia and Staphylococcus haemolyticus , whereas mNGS identified Klebsiella pneumoniae and Papillomavirus type 8, with no fungal pathogens detected.

3.6. Results of Antibiotic Resistance Prediction by mNGS and Consistency Analysis With Sputum Culture for Drug Resistance

Traditionally, antibiotic resistance prediction has relied on culture phenotyping and molecular testing [18]. mNGS detects antibiotic resistance genes (ARGs). In this study, mNGS detected seven ARGs in five patients, including more than one ARG in Patients #2 and #14 (Table 5).

TABLE 5.

mNGS‐based antibiotic resistance prediction and its consistency analysis with sputum culture phenotypic testing.

Patient	Pathogens detected by mNGS	ARGs detected by mNGS	DRA	Sputum culture results	AST results
#2	Staphylococcus epidermidis	qacA msr(A) blaR1 dfrC	Quaternary ammonium compounds	Candida albicans (colonizing organism)	NA
			Macrolides
			B‐lactam
			Trimethoprim
#3	Streptococcus pneumoniae	msr(D)	Macrolides	Negative	NA
#6	Streptococcus pneumoniae	msr(D)	Macrolides	Negative	NA
#13	Streptococcus pneumoniae	msr(D)	Macrolides	Negative	NA
#14	Streptococcus pneumoniae, Pseudomonas aeruginosa	msr(D) tet(M) catB7	MacrolidesTetracylines Chloramphenicol	Pseudomonas aeruginosa	Susceptible to all tested antimicrobials; no resistance detected

Abbreviations: ARGs, antibiotic resistance genes; AST, antimicrobial susceptibility testing; DRAs, drug‐resistant antibiotic; NA, not applicable/not tested.

As shown in Table 5, mNGS detected a variety of resistance genes and predicted the corresponding classes of resistant antimicrobials in 5 samples, yet phenotypic resistance consistency verification was limited by conventional test outcomes: 3 cases had negative sputum cultures with no available antimicrobial susceptibility testing (AST) data; 1 case had Candida albicans (colonizing bacteria) identified by culture, for which AST was not performed per clinical routine; only 1 case had valid AST results. In this patient, mNGS identified both Streptococcus pneumoniae and Pseudomonas aeruginosa together with their associated resistance genes and corresponding resistant drug classes, while sputum culture only detected Pseudomonas aeruginosa —AST showed this strain was susceptible to all routinely tested antibiotics with no resistance observed, demonstrating that mNGS offers a more comprehensive detection scope and overcomes the limitations of conventional AST in antimicrobial panels, and provide more evidence for the rational clinical selection of antibiotics. Since Streptococcus pneumoniae identified by mNGS was not confirmed by culture, a consistency comparison of drug resistance for the same pathogen could not be conducted.

3.7. Outcomes of Pneumonia in Patients With Cancer

In 9 of the 14 patients (64%), the diagnosis of the cause of the pneumonia was changed, and the antimicrobial drug regimen was adjusted based on the mNGS results (Table 6).

TABLE 6.

The adjustment of antimicrobial drug regimen based on the mNGS results.

Patient	Treatment before mNGS	Treatment after mNGS
#1	Meropenem	Meropenem and Metronidazole
#2	none	Ceftriaxone
#3	none	Oseltamivir
#7	Meropenem	Sulbactam and Cefoperazone
#8	Meropenem and Fluconazole	Meropenem, Vancomycin, Voriconazole and Oseltamivir
#9	Glucocorticoid	Antifungal drug (specific unknown) in other hospitals
#11	Ceftriaxone	Glucocorticoid
#12	Moxifloxacin	Sulbactam and Cefoperazone, Fluconazole, Oseltamivir and Allicin
#14	Levofloxacin	Piperacillin and Tazobactam

Among these 9 patients, other antimicrobial drugs were added or switched in 4 patients; the dose of antimicrobial drugs was reduced and steroid was added in 1 patient owing to a negative mNGS result and a diagnosis of radiation pneumonitis; antifungal drugs were added in 1 patient; antiviral drugs were added in 1 patient; and antibacterial, antifungal, and antiviral drugs were added in 2 patients (Table 4 and Figure 4A,B).

FIGURE 4.

(A) Number of patients adjusted for antimicrobial agents based on mNGS results. (B) Percentage of antimicrobial modification types based on mNGS results.

As shown in Tables 2 and 6, the specific adjustment strategies for each patient were as follows: 1. Patient #1: Bilophila wadsworthia was detected by mNGS but missed by conventional methods, and metronidazole was added to the original meropenem regimen; 2. Patient #2: fungi were detected by sputum culture, and mixed bacterial and fungal infection was identified by mNGS. Combined with the elevated white blood cell count and procalcitonin level in the patient, bacterial infection was considered the primary cause, and ceftriaxone was initiated as the initial anti‐infective regimen; 3. Patient #3: influenza A virus infection was confirmed by mNGS, and oseltamivir was initiated for targeted antiviral therapy; 4. Patient #7: conventional methods detected Stenotrophomonas maltophilia and Staphylococcus haemolyticus , whereas mNGS identified the truly pathogenic Klebsiella pneumoniae , and the original meropenem was replaced with sulbactam and cefoperazone; 5. Patient #8: mixed bacterial/fungal/viral infection was detected by mNGS, and vancomycin and oseltamivir were added to the original regimen of meropenem plus fluconazole, with fluconazole switched to voriconazole, thus upgrading the combined anti‐infective scheme; 6. Patient #9: immune‐related pneumonia (last dose of immunotherapy 19 days prior) was initially suspected and glucocorticoid therapy was administered. After Candida albicans was detected by both sputum culture and mNGS, glucocorticoids were discontinued and antifungal therapy was initiated instead; 7. Patient #11: negative mNGS results ruled out infectious etiologies and supported the diagnosis of radiation pneumonitis (55 days after radiotherapy). Thus the original ceftriaxone was discontinued, and the treatment was changed to glucocorticoids monotherapy; 8. Patient #12: full‐spectrum bacterial/fungal/viral infection was detected by mNGS, and sulbactam and cefoperazone, fluconazole, oseltamivir and allicin were added to the original moxifloxacin regimen; 9. Patient #14: conventional methods only detected Pseudomonas aeruginosa , whereas mNGS identified a mixed infection of Streptococcus pneumoniae and Pseudomonas aeruginosa . The antimicrobial regimen was escalated from levofloxacin to piperacillin‐tazobactam.

Eight of the nine patients showed improvement in their pneumonia, except for Patient #9, who showed no improvement after antifungal therapy, but subsequently achieved remission following hormonal therapy for the diagnosed immune‐related pneumonitis. All 14 patients’ pneumonia eventually improved, but two patients with advanced tumors subsequently died due to tumor progression.

4. Discussion

This retrospective study investigated the diagnostic performance of mNGS in cancer patients with pneumonia, in comparison with conventional detection methods. Compared with traditional assays, mNGS is an unbiased method for detecting all potentially infectious pathogens in a sample [13]. Previous studies have shown that mNGS has adequate accuracy and a significantly higher sensitivity for detecting pathogens [19, 20]. Moreover, mNGS is less affected by prior antibiotic exposure [15]. In addition to pathogen identification, mNGS provides clinical microbiome analysis, human host response analysis, and drug resistance prediction [13]. Therefore, it is valuable in the identification of pathogens causing pneumonia, particularly in cases of unexplained or mixed infection [14]. Immunocompromised patients with cancer are more susceptible to severe pneumonia, mixed infection, and pneumonia caused by pathogens that are difficult to detect using conventional tests [1, 4, 5]. Our findings support that mNGS can serve as an important complementary approach for etiological diagnosis of pneumonia in this vulnerable population.

In the present study, mNGS showed multiple distinct advantages over conventional assays, including higher sensitivity for difficult‐to‐culture and slow‐growing microorganisms, faster turnaround time (≤ 30 h vs. 3–5 days for sputum culture), and superior detection of coinfecting pathogens. In terms of the pathogenic profile, mNGS successfully detected a variety of rare pathogens that could not be identified by conventional methods, encompassing bacteria, viruses, fungi, and other microorganisms. While the common pathogenic bacteria identified overlapped with those in non‐oncologic pneumonia populations, the broad and distinctive pathogenic spectrum detected by mNGS underscores the complexity and heterogeneity of pneumonia etiologies in cancer patients. Additionally, mNGS allows for reliable prediction of antibiotic resistance phenotypes, thereby furnishing evidence to inform clinical anti‐infective management.

Clinically, of 14 enrolled patients, 9 had antimicrobial regimens adjusted based on mNGS results: adding targeted agents for newly detected pathogens, replacing ineffective antibiotics, escalating antifungal or antiviral therapy for mixed infections, and discontinuing antibiotics in mNGS‐negative patients to facilitate diagnosis of non‐infectious lung injury. By overcoming the limitations of conventional methods, mNGS provided a reliable etiological basis for individualized treatment and reduced unnecessary antibiotic exposure in immunotherapy/radiotherapy patients.

Three tumor patients with pneumonia showed complete discrepancies in pathogen detection between mNGS and conventional methods, illustrating both the advantages and limitations of mNGS: in Patient #1, mNGS identified the anaerobic pathogen Bilophila wadsworthia (missed by conventional culture) and ruled out non‐pathogenic Candida albicans , guiding effective metronidazole addition; in Patient #5, mNGS had a false negative for the true bacterial pathogen (verified by clinical improvement with meropenem), with the detected PBV virus having no pathogenic significance, reminding that mNGS cannot replace conventional culture; in Patient #7, mNGS corrected the misidentification of colonizing Stenotrophomonas maltophilia and Staphylococcus haemolyticus by sputum culture, identified the true pathogen Klebsiella pneumoniae to guide effective treatment adjustment. Collectively, these cases confirm that the rational combination of mNGS and conventional culture maximizes etiological diagnosis accuracy and optimizes antimicrobial strategies in immunocompromised tumor patients with pneumonia.

Patients with cancer are susceptible to opportunistic pathogens [21, 22, 23], but differentiating infection from colonization remains a major challenge, especially given the high abundance of commensal flora in the respiratory tract. In our study, clinical improvement after targeted antimicrobial therapy confirmed the pathogenic role of several mNGS‐identified organisms. Conversely, Epstein–Barr virus was considered non‐pathogenic in asymptomatic patients due to its high latent infection rate in adults [24], and Candida albicans isolated from a patient with immune checkpoint inhibitor‐associated pneumonia (19 days after immunotherapy) showed no response to antifungal therapy but was effectively managed with glucocorticoid treatment, thus it was identified as a colonizing organism. Although quantitative models have been developed to discriminate infection from colonization [25], further studies on mNGS are required to better differentiate between infection and colonization. We suggest combining the patient's clinical presentation, blood test results, imaging findings, and empirical treatment outcomes to determine whether an opportunistic pathogen is pathogenic. Clinicians should also be aware of the possibility of fungal infections after long‐term steroid therapy.

In this study, mNGS detected five ARGs and enabled rapid, comprehensive resistance prediction to guide clinical therapy. Compared with sputum culture, mNGS can detect drug resistance more rapidly and comprehensively. Consistent with previous studies [19, 26], mNGS was able to predict drug resistance in slow‐growing or unculturable pathogens and detect inactive pathogens after antibiotic treatment. Patients with hospital‐onset lower respiratory tract infections (LRTI) have a higher burden of ARGs in their respiratory microbiome than patients with community‐onset LRTIs [27]. By identifying ARGs, mNGS may help in the early identification of future secondary lung infections [28]. The mNGS test is particularly indicated for tumor patients with severe or complicated pneumonia, as well as for tumor patients whose etiology is unknown on conventional examination and for whom empirical treatment has not been effective [29, 30]. However, we could not verify the resistance consistency between mNGS and AST for identical pathogens, mainly due to low culture positivity, frequent sputum contamination, limited AST panel coverage, and inherent differences in detection range between the two methods.

In summary, based on this small retrospective study, mNGS appears to exhibit superior diagnostic performance over conventional methods for pneumonia in patients with cancer. Immunocompromised cancer patients present with a complex etiological spectrum of pneumonia, as they are prone to infections caused by opportunistic, fastidious, and rare pathogens, accompanied by a high incidence of mixed infections. Conventional assays are prone to missed detections, whereas mNGS enables sensitive and comprehensive pathogen detection to fill critical gaps in routine etiological testing. For patients receiving chemotherapy, radiotherapy, or immunotherapy, mNGS also helps rule out infectious etiologies via negative results, facilitating accurate differentiation between infectious pneumonia and treatment‐related lung injury (immune‐related or radiation pneumonitis) and cutting unnecessary antibiotic use. With rapid turnaround, polymicrobial co‐infection identification, and antimicrobial resistance prediction, mNGS may serve as a valuable supportive tool to guide individualized management of pneumonia and refine clinical treatment strategies in patients with cancer.

This study has some limitations. First, this was a single‐center retrospective case series with a small sample size; our conclusions are only preliminary and exploratory, and thus cannot be readily generalized to wider populations. Second, all mNGS testing was performed using sputum specimens, which are prone to contamination by upper respiratory tract flora, potentially compromising the accuracy of pathogen identification. Sputum samples were selected in this study due to their non‐invasiveness, easy accessibility, and routinely collected in our clinical practice during acute infection; thus, bronchoscopy‐guided bronchoalveolar lavage (BAL) samples were not utilized in the present research. Third, the mNGS specimens were sent to a commercial laboratory rather than a hospital microbiology laboratory, which may have reduced the sensitivity because the increased turnaround time reduced pathogen viability. Fourth, mNGS is expensive and not currently covered by health insurance in China. This may have contributed to the bias in patient selection in this study. Additionally the coverage of viral PCR testing was limited, with only 2 patients undergoing this test. As a retrospective study, viral PCR was only performed selectively for patients with high clinical suspicion of viral infection rather than routinely for all enrolled participants, which may result in missed detection of mild or occult viral infections, representing a minor limitation of this study. Finally, this preliminary study lacks long‐term follow‐up data to assess the long‐term clinical benefits of mNGS‐guided treatment.

Based on the above shortcomings, large‐sample, prospective, multi‐center clinical studies are urgently needed in the future to further verify the clinical value of mNGS. To address the limitation related to drug resistance verification mentioned above, subsequent studies will expand the sample size, optimize the specimen collection and processing procedures to improve the efficacy of traditional culture, and implement synchronous targeted detection of mNGS and culture/AST. Additionally, subsequent large‐scale studies can use BAL fluid samples for tumor patients with acceptable physical conditions to further improve the accuracy of pathogen detection.

5. Conclusions

In this small, retrospective case series of 14 patients, mNGS showed potential value in identifying pneumonia pathogens in cancer patients. Within the limitations of the study, it appeared to improve diagnostic yield over conventional methods, differentiating infectious from treatment‐related lung injury, and predicting antimicrobial resistance profiles to support individualized therapy. However, given the small sample and retrospective design, these findings remain exploratory and hypothesis‐generating. Future prospective, multi‐center studies with larger samples are needed to validate mNGS's clinical utility and standardize its application in this population.

Author Contributions

Rong Qin: writing – original draft, data curation, writing – review and editing, software, formal analysis, project administration, validation. Le Tian: writing – review and editing. Chao Wang: data curation, investigation. Minghua Cong: writing – review and editing. Ning Li: writing – review and editing, conceptualization, methodology, supervision, resources, funding acquisition.

Funding

The research was supported by the Beijing Vlove Charity Foundation (Grant No. JYKY2024‐0050409022).

Ethics Statement

The study was planned and conducted in accordance with the principles of the Declaration of Helsinki, and was approved by the Medical Ethical Committee of the Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College (reference number: 24/351‐4631). All participants were informed of the purpose of this research and signed informed consent. All methods were carried out in accordance with relevant guidelines and regulations.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.