To evaluate the diagnostic performance of tNGS in pneumonia of cancer patients, a retrospective study was conducted, enrolling 237 cancer patients with suspected pneumonia who had undergone tNGS testing at the Henan Cancer Hospital from April 2022 to April 2024( Figure 1 ). Clinical information and biochemical indicators were collected through the hospital’s electronic medical record system, and ultimately, 148 patients with complete clinical information and diagnostic outcomes were included in the analysis. Patients were eligible for enrolment if they (1) were at least 18 years of age; (2) were undergoing cancer treatment; (3) had suspected pneumonia. Patients with suspected pneumonia met both of the following criteria: (1) at least one compatible symptom, such as new-onset fever, cough, or dyspnea; (2) new-onset radiological findings on chest images.

This study received approval from the Ethics Committees to ensure compliance with ethical guidelines. Patients’ identification remained anonymous throughout the study, and informed consent was waived due to the retrospective and observational nature of the research. All samples were aliquoted into 1ml portions and stored at -80°C after clinical collection.

All CMTs were performed on every sample, except for viral PCRs, which were performed at the clinician’s discretion. A total of 148 patients underwent BALF sampling, from which a portion was aliquoted for testing by tNGS. Additionally, this BALF was subjected to CMTs alongside other relevant samples, such as blood and urine. The CMTs encompassed a range of diagnostic methods including culture, antigen detection, multiplex PCR, and Xpert PCR assays.

tNGS was performed once the sample was obtained. As previous reported (Wei et al., 2024; Zhang et al., 2024), BALF samples with high viscosity were diluted 1:1 with 0.1 M dithiothreitol before nucleic acid extraction. A volume of 400μL of BALF, lysis buffer, protease K mixture, binding buffer, and 1.2 g glass bead were agitated vigorously at 4500 rpm for a total of 130 s by FastPrep-24™ 5G Instrument (MP Biomedical, CA, USA). Nucleic acid extraction was performed using the VAMNE Magnetic Pathogen DNA/RNA kit (Vazyme, Nanjing, China). Then, the RNA was reverse-transcribed into cDNA using Hieff NGS^® ds-cDNA Synthesis Kit (Yeasen, Shanghai, China). Nucleic acids were quantified using a Qubit 3.0 fluorometer. A549 human cells (GenePlus, Suzhou, China) were used as negative controls (NTC) to detect contamination, and A549 human cells spiked with Staphylococcus aureus (BeNa Culture Collection, Beijing, China) were used as positive controls (PTC). Following extraction, cDNA synthesis and library preparation were performed with the HieffNGS^®C37P4 One PotcDNA&gDNA Library Prep Kit (Yeasen, Shanghai, China). Then, the cDNA was taken through library enrichment with NadPrep^® NanoBlockers (Nanodigmbio, Nanjing, China) reagents to generate the product for targeted sequencing. Eight uniquely barcoded libraries were pooled to hybridize and capture by specific biotinylated probes for 4 hours using Geneplus design probes. Products were quantified with a Qubit 3.0 instrument using DNA HS Assay Kit (Vazyme, Nanjing, Jiangsu, China). Products were stored at −20°C until Sequencing. After library construction, sequencing was performed on the Gene+seq 100 platform(GenePlus, Suzhou, China). The sequencing read length was set to 100 base pairs (bp), with a preset data of 5 million reads. Clean reads were obtained by removing sequencing adapters, low-quality reads, or reads below 35 bp using fastp (version 0.23.1). The remaining reads were aligned to the human reference (hg38) using Burrow-Wheeler Aligner (version 0.7.12-r1039) and human reads were filtered. The filtered reads were compared with the self-built pathogenic microorganism database, and the retained results were annotated. Microbial reads within the target range were normalized to reads per million (RPM), and only microorganisms above a predefined threshold were initial reported in this study. The threshold was set at RPM ≥6 for common pathogens (excluding mycobacteria) and ≥0.5 for fungi and mycobacteria. A manual review is conducted. The common colonizing microorganisms within the respiratory tract target range was reported separately from other microorganisms in different sections to avoid interfering with clinical judgment (Li et al., 2023; Zhang et al., 2024).

Two physicians with extensive clinical experience, each having worked in the respiratory department for over 5 years, independently reviewed all inpatient medical information and microbiological test results of the patients, including tNGS. Initially, they determined whether the patients had infectious diseases. Subsequently, they identified the pathogenic microorganisms based on clinical manifestations, laboratory tests, imaging examinations, microbiological test results (including CMT and tNGS), and treatment responses. Discrepancies between the two reviewers were first resolved through in-depth discussions; if consensus could not be reached, another senior reviewer was consulted.

For continuous variables, report as the median and interquartile range (IQR). Categorical variables are represented by frequency and percentage. Inter-group comparisons are made using the unpaired t-test or the Mann-Whitney U test. For comparisons between groups of categorical variables, the chi-square test is used. In evaluating diagnostic performance, sample consistency is judged by referring to previous studies (Blauwkamp et al., 2019); partial pathogen consistency is considered as consistent. In the calculation of diagnostic performance, sensitivity, specificity, and accuracy are computed using the standard proportion formula, and the 95% confidence intervals for these proportions are determined using the Wilson method. To compare the differences in diagnostic performance and analyze diagnostic accuracy between two groups, a t-test can be employed.

All figures were drawn using GraphPad Prism version 9.5.0 (GraphPad Software LLC., San Diego, CA, USA). All analyses were performed with SPSS version 26.0 (SPSS Inc., Chicago, Illinois, USA) (Blauwkamp et al., 2019). A p-value less than 0.05 was considered statistically significant.

A total of 237 patients with suspected pneumonia were considered for inclusion. Of these, 89 patients were excluded, some because they were ultimately diagnosed as non-cancer patients, and others due to insufficient data for further analysis. One hundred and forty-eight patients met the criteria and were included in the final analysis. This group comprised 60.14% male patients. The patients were primarily composed of three types of cancer patients, with lung cancer accounting for 69.59%, hematological malignancies accounting for 18.24%, and gastrointestinal cancer accounting for 8.78%. Additionally, there were 5 patients mainly with ovarian, bladder, and breast cancer ( Table 1 ). When comparing pneumonia and non-infection patients, no significant differences in characteristics were observed between the two groups. However, the proportion of hematological malignancies was higher in the pneumonia patient group.

In the clinical diagnosis, 130 patients were confirmed to have infections, while 18 patients were non-infectious. tNGS identified 110 infected patients and found no microorganisms in 17/18 non-infectious patients, ultimately showing a sensitivity of 84.62% and a specificity of 94.44%, with a diagnostic accuracy of 85.81% ( Table 2 ). The sensitivity of CMT in diagnosing was 56.92%, with a specificity of 100%, and a diagnostic accuracy of 62.16%. The diagnostic accuracy of tNGS was significantly higher than that of the CMT method. This was mainly because tNGS had significantly higher sensitivity for fungal and bacterial infections compared to the CMT method (p<0.0001). However, due to the excessive sensitivity of tNGS, there was no significant difference in accuracy between tNGS and CMT for bacterial and fungal infections. In terms of viral infections, PCR, as the diagnostic standard, had a higher diagnostic accuracy than the tNGS method. Nevertheless, the broader coverage of viruses by tNGS provided better sensitivity. Therefore, tNGS was a more sensitive method than CMT overall.

The study further explored the diagnostic accuracy of the targeted Next-Generation Sequencing (tNGS) method for various pathogens. Ultimately, 168 pathogens were identified in these patients. Of these, tNGS identified 147(87.50%), while CMT identified 96(57.14%). The tNGS method detected a significantly higher number of pathogens compared to CMT (p<0.0001), including bacteria, fungi, and viruses ( Figure 2A ). A further comparison of the pathogen distribution profiles revealed that tNGS had higher sensitivity for detecting Pneumocystis jirovecii among fungi ( Figure 2B ). For viruses, tNGS was able to identify additional RNA viruses not covered by multiplex PCR, such as Human respiratory viruses, Human metapneumovirus, and adenovirus. In terms of bacterial detection, tNGS covered difficult-to-culture or unexpected species such as Brucella melitensis, non-tuberculous mycobacteria, and Legionella pneumophila. However, tNGS was less sensitive in detecting Aspergillus spp compared to the CMT. Further evaluation of the diagnostic performance of pathogens with a frequency count exceeding five times (Mycobacterium tuberculosis:4) showed higher diagnostic accuracy for Pneumocystis jirovecii, and Legionella pneumophila, but lower accuracy for Aspergillus spp, Streptococcus pneumoniae, and rhinovirus compared to CMT( Table 3 ). Additionally, in the diagnosis of Epstein-Barr virus (EBV), tNGS was able to detect more EBV in samples, resulting in significantly lower diagnostic accuracy than PCR. We conducted a comparative analysis of the RPM values between pathogens consistently detected by both methods and those detected exclusively by tNGS. The study findings indicate that the RPM values of Pneumocystis jirovecii detected exclusively by tNGS are comparatively lower, a phenomenon also observed in the analysis of viruses and bacteria ( Supplementary Figure 1 ). This may be attributed to the higher sensitivity of the tNGS method.

The performance of the tNGS method in the etiological diagnosis of pneumonia in different cancer patients was further explored. In patients with lung cancer, tNGS showed 80.68% of sensitivity and 93.33% of specificity ( Table 4 ). In patients with gastrointestinal cancer and hematological malignancies, tNGS demonstrates a sensitivity exceeding 90% of tNGS. Whether in patients with lung cancer, gastrointestinal cancer, or hematological malignancies, the diagnostic accuracy of tNGS was higher than CMT. Compared with the overall patients, there was no significant difference in the diagnostic accuracy of tNGS among different cancer types (p>0.05). The distribution of pathogens among different cancer types was observed ( Figure 2C ). Most of the pathogens were identified in patients with lung cancer, accounting for about 60%, with 73.42% being bacterial pathogens. Patients with hematological malignancies accounted for 25% of identified pathogens, but only 11.39% were bacterial. In gastrointestinal cancer cases, the proportion of identified pathogens was 10.12%, with fungi being relatively rare at only 5.41%. Pneumocystis jirovecii and Aspergillus spp were more commonly found in patients with lung cancer and hematological malignancies, while common pathogens such as Pseudomonas aeruginosa, Mycoplasma pneumoniae, Streptococcus pneumoniae, and others were more prevalent in patients with lung cancer ( Figure 2D ). No differences were observed in the presence of Klebsiella pneumoniae, Acinetobacter baumannii, and Mycobacterium tuberculosis in patients with different types of cancer. Interestingly, a higher number of severe acute respiratory syndrome coronavirus 2 infections were found in patients with hematological malignancies, but no differences were observed in the presence of other viruses among patients with different types of cancer.

A retrospective review of the treatment history of patients was conducted to further analyze the impact of tNGS on clinical treatment. It was found that tNGS results directly influenced antibiotic treatment decisions in many patients.

Case 1: A patient with lung cancer presented with symptoms of coughing and sputum production and was diagnosed with pneumonia at our hospital. The patient was treated with anti-infective agents, but no improvement was observed. A tNGS test for respiratory pathogens was performed, revealing the presence of Rhinovirus C and Pneumocystis jirovecii. Given the patient’s deteriorating pulmonary condition and the tNGS results, it was clinically considered that the infection was caused by both viral and fungal pathogens. Consequently, the patient was administered Levofloxacin and Piperacillin for anti-infective treatment, along with a combination of Sulfamethoxazole and Trimethoprim tablets for antifungal treatment, the patient showed signs of improvement.

Case 2: A patient with follicular lymphoma presented with fever and subsequently underwent a CT scan, which indicated symptoms of pulmonary infection. A tNGS test for respiratory pathogens detected Pneumocystis jirovecii. Symptomatic treatment with Sulfamethoxazole and Trimethoprim was administered. The patient has not experienced further fever and has shown symptomatic improvement following continued anti-infective treatment.

Case 3: A patient, post-debulking surgery for a malignant ovarian cancer, presented with symptoms of coughing and sputum production. Upon admission, biochemical tests and CT sacn indicated a pulmonary infection. tNGS test identified a specific pathogen: Mycobacterium tuberculosis complex. The patient was commenced on anti-tuberculosis therapy, which led to symptomatic relief. The patient was subsequently discharged following improvement.

These cases emphasize the important role of tNGS in guiding the diagnosis of infections in cancer patients and targeted antibiotic therapy, ultimately improving patient prognosis.