Diagnostic value of metagenomic next-generation sequencing in patients with febrile lung cancer with negative conventional microbiological tests and without neutropenia.

1
Introduction
Lung cancer is one of the most common and lethal malignancies worldwide and accounts for a major public health challenge (Sung et al., 2021). Patients with lung cancer are highly susceptible to infectious complications due to airway obstruction, treatment-related mucosal injury, and systemic immune dysfunction, which together contribute to poor outcomes (Rosolem et al., 2012; Rolston, 2017). Although neutropenia is a well-recognized risk factor for infection (Reyna-Figueroa et al., 2015), fever is also common in patients with nonneutropenic lung cancer, and the differential diagnoses include bacterial infection and tumor fever (TF) (Shomali et al., 2012). Fever is often the earliest and sometimes only clinical indicator of infection (Akinosoglou et al., 2013). It is generally considered a protective host immune response against invading pathogens (Marik, 2000). Fever in lung cancer may also be tumor-related and diagnosed by exclusion (Zell and Chang, 2005). Insufficient early discrimination can expose nonbacterial episodes to unnecessary antibiotics with attendant toxicity and resistance, whereas the true infection risk delays definitive therapy and worsens outcomes (Zhao et al., 2018; Messacar et al., 2017).
Conventional microbiological techniques—cultures and serology—are foundational yet limited by modest sensitivity, prolonged turnaround, and narrow pathogen coverage, particularly after prior antibiotic exposure (Goldberg et al., 2015; Grumaz et al., 2016). Performance further depends on sampling timing, technique, and organism growth requirements, leading to missed fastidious or slow-growing pathogens and diagnostic delays (Gu et al., 2019). Traditional assays also cannot profile the full microbial community, reducing utility in polymicrobial or otherwise complex infections (Wang et al., 2019). These constraints frequently yield CMT-negative evaluations in nonneutropenic patients, prolonging empiric, trial-and-error management.
Metagenomic next-generation sequencing (mNGS) is an advanced diagnostic technology capable of simultaneously detecting a broad spectrum of pathogens, including bacteria, viruses, fungi, and parasites (Tang et al., 2021). Unlike conventional culture-based methods, mNGS enables the comprehensive analysis of nucleic acids present in clinical specimens, most commonly blood, using high-throughput sequencing platforms. Human-derived sequences are bioinformatically removed, and the remaining microbial reads are aligned to comprehensive reference databases for precise taxonomic classification (Wooley et al., 2010). This approach allows the simultaneous detection of diverse pathogens from a single clinical sample (Rodino and Simner, 2024). As mNGS is not dependent on organism viability, it retains value despite antecedent antibiotics and is particularly useful for mixed, fastidious, or rare infections (Diao et al., 2022; Han et al., 2019). For example, Lin et al. (2022) reported that mNGS applied to bronchoalveolar lavage fluid significantly improved pathogen detection compared with conventional methods, particularly for Pneumocystis jirovecii and co-infections in immunocompromised patients. In addition to its broad detection capability, mNGS provides semiquantitative data that can be used to estimate pathogen loads and identify antimicrobial resistance genes, offering actionable insights for precision antimicrobial therapy (Sahoo et al., 2013; Salipante et al., 2014).
Against this backdrop, the present study aimed to evaluate plasma-based mNGS in patients with febrile lung cancer who are nonneutropenic and CMT-negative, specifically to determine whether it provides incremental diagnostic yield, thereby supporting immediate, result-guided antimicrobial decisions, which are associated with earlier defervescence.

2
Materials and methods
2.1
Patients and fever episode definition
This single-center, retrospective cohort study was conducted at the Cancer Center of the Union Hospital, Tongji Medical College, Huazhong University of Science and Technology (August 2023–October 2024). Consecutive adults (age ≥ 18 years) with histologically confirmed lung cancer who developed an eligible fever episode at admission or during hospitalization were screened. In total, 212 patients met the screening criteria; after excluding 125 patients for prespecified reasons, most commonly incomplete conventional microbiology and/or plasma mNGS data, 87 patients remained for analysis (Figure 1). Of these, 34 were positive for conventional microbiological tests (CMT) and were summarized descriptively, whereas 53 were CMT-negative and constituted the primary analysis set. The latter were stratified a priori based on the mNGS result into two exploratory arms: mNGS-positive high-suspicion infectious fever (mNGS-HSIF) (n = 37) and mNGS-negative high-suspicion tumor fever (mNGS-HSTF) (n = 16), forming the basis for analyses of incremental diagnostic yield, mNGS-prompted management changes, and landmark defervescence endpoints.
An eligible fever episode was defined as an axillary temperature ≥ 38.5 °C confirmed on repeat measurement within 6 h under routine care. Episodes that resolved within 12 h without systemic signs were considered transient and were not eligible. Patients with fever clearly attributable to noninfectious triggers (e.g., infusion/transfusion reactions, colony-stimulating factor use, or immediate postprocedural/anesthesia responses) were not considered eligible. Only the first eligible episode was analyzed in patients with multiple febrile events. Nonneutropenia was required; neutropenia was an exclusion criterion defined as an absolute neutrophil count (ANC) < 0.5 × 109/L based on the closest complete blood count around fever onset.
The index time point (0) for outcome assessment was prespecified as the time the mNGS report became available in the electronic medical record. Per institutional practice during the study period, clinicians immediately adjusted or initiated targeted anti-infective therapy at t0 for mNGS-positive cases, whereas mNGS-negative cases generally continued oncological management and supportive care; conventional antipyretics were permitted at clinicians’ discretion in both groups.
The primary endpoint was time to defervescence from t0, defined as an axillary temperature < 37.3 °C sustained for ≥ 24 h; because antipyretics were permitted per routine care, defervescence should be interpreted as symptom control rather than definitive pathogen eradication.
Prespecified secondary summaries were defervescence proportions at 24/48/72/96 h post-t0.

2.2
Inclusion and exclusion criteria
2.2.1
Inclusion
Patients were eligible if they met all of the following: (1) age ≥ 18 years; (2) histologically confirmed lung cancer with complete oncological records; (3) at least one eligible fever episode as defined in Section 2.1; (4) completion of blood cultures within ≤ 24 h, pathogen serology per institutional practice, and plasma mNGS within ≤ 72 h of the first eligible fever; and (5) availability of clinical data for outcome assessment through hospital discharge or day 28, whichever occurred first.

2.2.2
Exclusion (operational windows)
Patients were excluded for the following reasons: (1) neutropenia, defined as ANC < 0.5 × 109/L within − 24 to + 24 h of fever onset; (2) prespecified noninfectious triggers, namely, infusion-related reactions to cytotoxic/immune therapies or monoclonal antibodies during infusion or within 24 h after completion; transfusion-related reactions during or within 4–24 h after blood product administration; colony-stimulating factor-related fever within 72 h of granulocyte colony-stimulating factor (G-/GM-CSF) administration; and immediate postprocedural/anesthesia fever within 48 h of an invasive procedure or anesthesia; (3) transfer before completion of the diagnostic work-up; (4) incomplete clinical information or failed sample quality control for blood cultures, serology, or mNGS; and (5) recurrent febrile episodes in the same patient (only the first eligible episode was analyzed).

2.2.3
Sampling windows and prior antimicrobial exposure
Blood cultures and pathogen serology were performed within 24 h, and plasma mNGS was performed within 72 h of fever onset.

2.3
Sample collection
Blood cultures and pathogen serology were performed within 24 h of the first eligible fever. Plasma for mNGS (5–10 mL of ethylenediaminetetraacetic acid [EDTA] whole blood) was obtained within 72 h of fever onset in routine practice. The exact sampling timestamps were recorded in electronic medical records, and the specimens were transported immediately to the laboratories according to institutional standard operating procedures.

2.4
Metagenomic next-generation sequencing
Plasma mNGS was performed using a standardized shotgun workflow (PMseq®; BGI, Shenzhen, China) on Illumina or MGISEQ platforms (Miao et al., 2018). Cell-free DNA was extracted from EDTA plasma, and human reads were bioinformatically removed before taxonomic classification against a curated microbial reference database (PMDB) (~ 17,500 pathogenic taxa, vendor-maintained) (Li and Durbin, 2009). Each sequencing batch incorporated an internal spike-in control, together with batch-matched negative and positive controls, to monitor potential contamination and process integrity. Library preparation followed clinical-grade QC gates (e.g., input concentration/purity, fragment size distribution, run yield, and base-calling quality with Q20/Q30 reporting). Per laboratory Standard Operating Procedures (SOP), we targeted a per-sample sequencing depth of ≥ 20 million reads to ensure analytical sensitivity.
Taxa were reported only when (i) prespecified species-level read count/RPM thresholds were met and (ii) statistical enrichment over batch-matched negative controls was achieved after host-read subtraction and background filtering. Adjudication further considered clinical plausibility and treatment response, as detailed previously (Miao et al., 2018). Low-level herpes virus detection was performed using sensitivity analyses (inclusive vs. strict exclusion). Turnaround time was defined as the time from laboratory receipt to report availability in the medical record and is reported as median (interquartile range [IQR]). Detailed QC metrics, database version/release data, and positivity thresholds are provided in Supplementary Methods.

2.5
Conventional microbiological testing
Blood cultures were obtained under aseptic conditions and processed in automated incubators according to the institutional SOPs. Positive bottles were subjected to Gram staining, biochemical or matrix-assisted laser desorption/ionization–time-of-flight identification, and antimicrobial susceptibility testing, as indicated, according to Clinical and Laboratory Standards Institute (CLSI) guidelines (Clinical and Laboratory Standards Institute, 2023). Pathogen serology (atypical bacteria and respiratory viruses) was performed using enzyme-linked immunosorbent assay, according to the manufacturer’s instructions, targeting Mycoplasma pneumoniae, Chlamydia pneumoniae, respiratory syncytial virus, adenovirus, and Coxsackie B virus. Results above the assay-specific cutoffs were considered positive.

2.6
Ethics
This single-center retrospective chart review complied with the Declaration of Helsinki and applicable regulations. The study protocol was reviewed and approved prior to any data extraction or analysis by the Institutional Review Board (IRB) of Union Hospital, Tongji Medical College, Huazhong University of Science and Technology (Approval No. 2025(0436); 06 May 2025). In line with institutional policy for minimal-risk retrospective studies using pre-existing records, the IRB granted a waiver of written informed consent. No study-specific interventions or additional specimen collection were undertaken; all tests—including plasma mNGS—were ordered as part of routine clinical care. All analyses were conducted on de-identified data only, and no patient contact occurred.

2.7
Statistical analyses
All analyses were performed using SPSS version 27.0 (IBM, Armonk, NY, USA), and figures were created using GraphPad Prism version 10.0 (GraphPad Software, La Jolla, CA, USA). The primary analysis set comprised CMT-negative patients classified as having mNGS-HSIF (n = 37) or mNGS-HSTF (n = 16). Continuous variables are summarized as means ± standard deviations or medians (IQRs) according to the Shapiro–Wilk assessment of normality; categorical variables are summarized as n (%). Statistical significance was set at p < 0.05.
Baseline characteristics are descriptively presented in Table 1 without formal hypothesis testing. Pathogen detection among mNGS-positive cases is summarized in Table 2 as counts based on pathogen and organism type. Laboratory indices at fever onset (Table 3, Figure 2) were compared between groups using the Mann–Whitney U test for continuous variables and χ2 or Fisher’s exact tests for categorical variables.
The primary endpoint was defervescence, defined a priori as axillary temperature < 37.3 °C sustained for ≥ 24 h. Landmark proportions of patients defervesced by 24/48/72/96 h after t0 (time of mNGS report availability) were compared with Fisher’s exact tests and exact 95% confidence intervals (CIs) (Table 4). For time-to-event analysis (Figure 3), time to defervescence was measured from t0 to the first time point at which all subsequent temperature readings over a continuous 24-h window were < 37.3 °C. Patients without defervescence were right-censored at hospital discharge or day 28, whichever occurred first. Unadjusted group differences were evaluated using the log-rank test, and univariate Cox models provided hazard ratios (HRs) with 95% CIs (Efron method for ties). No multivariate modeling or multiple testing adjustment (e.g., false discovery rate) was performed. Given the exploratory nature of this study, no multiplicity adjustment was applied, and the p-values were descriptive. Interpretation prioritizes effect sizes and 95% CIs (e.g., median differences and risk differences). Missing patterns were reviewed, and complete case analyses were performed.

3
Results
3.1
Overview of the study population
The baseline features of the 53 CMT-negative patients are summarized in Table 1. The cohort was predominantly composed of men (88.7%), with a mean age of 64.1 years ± 6.3 years. Non-small cell lung cancer was common, with adenocarcinoma accounting for 50.9% (27/53) and squamous cell carcinoma accounting for 39.6% (21/53) (combined 90.5%), whereas small cell lung cancer and other histologies accounted for 7.5% and 1.9%, respectively. Most patients had advanced disease (stages III–IV, 94.4%). Ever-smoking was reported by 68.0% of the patients, and diabetes (43.4%), hypertension (37.7%), and chronic obstructive pulmonary disease (35.8%) were frequent. Recent chemo-/radiotherapy was documented in 43.4% of the patients.

3.2
Diagnostic yield of mNGS
Among the 53 CMT-negative patients, plasma mNGS identified ≥ 1 pathogen in 37 (69.8%) patients, whereas 16 (30.2%) were mNGS-negative. Across the 37 mNGS-positive cases, 54 pathogen detections were recorded: 18 bacterial, 28 viral, and eight fungal detections. The most frequent bacteria were Mycobacterium tuberculosis (n = 4) and Nocardia spp. (n = 3), whereas the leading fungi were Candida albicans (n = 4) and Aspergillus spp. (n = 3). The viral spectrum was dominated by herpes viruses, notably the Epstein–Barr virus (n = 6), cytomegalovirus (n = 5), and adenovirus (n = 5). Full details are provided in Table 2, and the distribution of the most frequently detected pathogens is illustrated in Figure 4.

3.3
Clinical characteristics of mNGS-high-suspicion tumor fever and mNGS-HSTF
The baseline demographics and oncological variables were comparable across the groups (Table 3). At fever onset, mNGS-HSIF showed higher inflammatory indices than mNGS-HSTF, including white blood cell count (median, 10.20 vs. 8.32 × 109/L; p = 0.023), neutrophil count (7.66 vs. 5.98 × 109/L, p = 0.009), C-reactive protein (CRP) (63.67 mg/L vs. 32.91 mg/L, p = 0.018), and procalcitonin (PCT) (0.83 ng/mL vs. 0.38 ng/mL, p = 0.003) (Figures 2A, C, D). Neutrophil and lymphocyte percentages also differed (77.48% vs. 73.58%, p = 0.011 and 15.37% vs. 18.97%, p = 0.011, respectively; Figure 2B), whereas lymphocyte count (1.61 vs. 1.67 × 109/L; p = 0.670), erythrocyte sedimentation rate (35.10 mm/h vs. 27.90 mm/h, p = 0.143), and peak temperature (39.10 °C vs. 38.90 °C, p = 0.365) were similar. Full statistics are provided in Table 3.

3.4
Treatment response and clinical outcomes
In accordance with institutional practice, the receipt of an mNGS-positive report prompted immediate result-guided antimicrobial adjustments in the HSIF arm, whereas mNGS-negative cases (HSTF) generally continued oncological treatment with supportive care (e.g., antipyretics, as required). Clinical decision-making incorporated mNGS results together with clinical presentation, laboratory data, imaging, and history. By 24 h post-t0, 29.7% of patients with HSIF versus 25.0% of patients with HSTF had defervesced (p = 1.000); by 48 h, the proportions were 73.0% versus 37.5% (p = 0.029); by 72 h, 78.4% versus 56.3% (p = 0.182); and by 96 h, 89.2% versus 68.8% (p = 0.027) (Table 4). Consistent with these landmark findings, Kaplan–Meier curves favored HSIF (log-rank p = 0.025), with an unadjusted HR for defervescence of 2.08 (95% CI, 1.09–3.97). The 6-day cumulative defervescence was 91.9% in HSIF versus 68.8% in HSTF, and by day 12, HSTF reached 93.8%, whereas HSIF was not estimable because all events occurred earlier (Figure 3). Taken together, these data support the clinical utility of plasma mNGS not only for etiological clarification but also for therapeutic stratification in patients with nonneutropenic, CMT-negative febrile lung cancer.

4
Discussion
4.1
Rationale for focusing on the nonneutropenic, CMT-negative population
At our center, routine prophylactic G-CSF administration indicates that truly neutropenic cases, especially febrile neutropenia (FN), are rare. FN also differs fundamentally from the nonneutropenic setting in terms of pathogen spectrum, inflammatory kinetics, and clinical pathways (e.g., immediate empiric broad-spectrum antibiotics are standard). Therefore, we focused on nonneutropenic patients with negative CMT (CMT-negative) to enhance internal validity and interpretability. The tradeoff is limited generalizability: conclusions primarily apply to patients with nonneutropenic lung cancer with fever. In FN, thresholds and dynamics for CRP/PCT and mNGS sampling windows and attribution frameworks require independent validation.

4.2
Infectious fever in lung cancer and the role of mNGS
Infectious fever remains the most common complication in oncological practice and can progress to sepsis, organ dysfunction, and death when inadequately controlled (Bhat et al., 2021). Triggers include post-treatment immunosuppression, invasive procedures, and translocation from colonizing niches, including respiratory, gastrointestinal, and bloodstream infections (Yusuf et al., 2023). Conventional diagnostics (cultures and serology) are foundational but limited by modest sensitivity, prolonged turnaround, and narrow coverage, particularly after prior antibiotics (Zhou et al., 2019; Wright et al., 2022; Zhang et al., 2022). In our CMT-negative cohort, plasma mNGS identified at least one organism in 69.8% of the patients, delineating a spectrum that included fastidious bacteria, opportunistic fungi, and herpesviruses. These findings have clinical implications: mNGS-positive patients had higher inflammatory marker levels at presentation and frequently underwent targeted antimicrobial adjustments, whereas mNGS-negative patients often defervesced without antibiotic escalation while continuing oncological or supportive care. Taken together, these observations support the use of plasma mNGS to shorten the empirical, trial-and-error management of diagnostically unresolved fever.

4.3
Tumor fever as a diagnosis of exclusion and how mNGS separates signals from noise
TF is a diagnosis of exclusion under major oncology guidance and is difficult to distinguish from infection when early clinical cues overlap (Pasikhova et al., 2017). In this nonneutropenic, CMT-negative subgroup, mNGS functioned as a rapid adjunct to refine attribution. By 48 h after the report, 73.0% of mNGS-positive patients had decreased fever compared with 37.5% of mNGS-negative patients; by 96 h, the proportions were 89.2% and 68.8%, respectively. Although these are unadjusted fixed-time comparisons, the findings are aligned with the intended clinical use of plasma mNGS—prompt, result-guided therapy when infection is likely and avoidance of unnecessary escalation when results are negative—thereby reducing antibiotic exposure while safeguarding patients with true infection.

4.4
Positioning within the mNGS literature and implications for practice
In prospective and retrospective studies, mNGS has been associated with a higher diagnostic yield than conventional testing and antimicrobial stewardship actions, including escalation and de-escalation, when appropriate (Wang et al., 2020; Zhou et al., 2021; Xiao et al., 2023). Our study extends this evidence to a precisely defined population of patients with nonneutropenic, CMT-negative lung cancer with fever, where the clinical dilemma is greatest. By restricting the study to diagnostically unresolved cases rather than CMT-positive episodes, we targeted the zone of maximal decision uncertainty and showed that plasma mNGS could inform immediate management, while supporting earlier defervescence. As mNGS diffuses into routine pathways, hybrid strategies that pair plasma mNGS with rapid targeted assays may balance TAT with breadth and can be tailored to local epidemiology and resource environments.

4.5
Limitations
This retrospective, single-center analysis with a modest sample size may introduce selection bias and insufficient statistical power, limiting generalizability because an independent validation cohort was not available. First, group assignment is tied to mNGS availability at the report time (t0), and positive results prompted immediate therapy modification; this design introduces potential confounding by indication and guarantee-time bias in favor of faster symptom control in the mNGS-positive stratum. Second, antipyretics were permitted under routine care; thus, “defervescence” reflects symptom control rather than definitive pathogen eradication, and the confounding effect of antipyretics cannot be fully ruled out. Third, the interpretation of mNGS results relies on clinical adjudication rather than an independent “gold standard” (e.g., histopathology). mNGS carries an inherent risk of detecting incidental findings; low-level viral reads may reflect reactivation or contamination rather than active infection (Wang et al., 2024). Conversely, a single negative mNGS result does not exclude infection when the pathogen burden is low, infection is compartmentalized, or sampling is suboptimal; antibiotic de-escalation based on a negative result should be considered only in clinically stable patients and in conjunction with alternative etiological evidence (Meng et al., 2025). Fourth, mNGS provides limited susceptibility information and, therefore, cannot replace conventional microbiology for antimicrobial guidance (Liu et al., 2024). Our conventional testing panel (blood cultures plus a limited viral serology set) may have underestimated the potential yield of CMT and inflated the comparative advantages of mNGS, especially for viral detection. Finally, although cost, access, and TAT constraints remain barriers, the feasibility of mNGS lies in its targeted application. While currently too costly for routine universal screening, its deployment as a second-line diagnostic tool for clinically difficult cases may offset costs by reducing prolonged hospitalization. Assays are often centralized and may not be reimbursed in some systems, highlighting the need for health-economic evaluations in future studies.

4.6
Future directions
Broader integration of plasma mNGS for unexplained febrile illness will require prospective, multicenter studies to validate the standardized reporting criteria. Prospective, multicenter studies should start follow-up at fever onset (rather than t0), prespecify management pathways, and incorporate time-dependent exposure models that treat mNGS results as time-varying covariates. Blinded, tiered attribution (causative/probable/bystander/contaminant) and read-count thresholds should be harmonized with sensitivity analyses of low-abundance viruses. The key secondary endpoints should include antipyretic-free defervescence, fever burden (temperature area under the curve), time to appropriate therapy, downstream outcomes (intensive care unit transfer, length of stay, mortality), TAT, and cost-effectiveness (especially in resource-limited settings). Negative results should be explicitly evaluated for stewardship value (de-escalation, avoidance of unnecessary diagnostics).

5
Conclusion
Plasma mNGS serves as a complementary diagnostic tool in patients with nonneutropenic CMT-negative lung cancer with fever. By substantially increasing pathogen detection in this diagnostically challenging subgroup, mNGS supports etiological clarification and immediate result-guided treatment decisions. Positive results facilitate targeted antimicrobial therapy, whereas negative results, interpreted in a clinical context, support the continuation of oncological or supportive care without unnecessary escalation. This approach enables individualized fever management, helping minimize avoidable antibiotic exposure while preserving timely therapy for true infections. With ongoing technical refinement and prospective validation, plasma mNGS is expected to play an expanding role in diagnostic workflows for complex or uncertain febrile presentations.