Computed tomography–guided precision biopsy combined with metagenomic next-generation sequencing for etiological diagnosis in patients with blood culture-negative systemic infections

Abstract

Objective

To evaluate the diagnostic efficacy of computed tomography–guided percutaneous biopsy combined with metagenomic next-generation sequencing in patients with blood culture-negative systemic infections and to assess the clinical impact of using this combined strategy for etiological confirmation and guidance of targeted antimicrobial therapy.

Methods

This single-center retrospective observational cohort study enrolled 78 patients who met the Sepsis-3 consensus criteria for suspected systemic infection and had negative conventional microbiological work-ups (at least two sets of blood cultures) between April 2022 and March 2025. All patients underwent computed tomography–guided biopsy of radiologically identified infectious foci, with specimens processed concurrently for conventional culture and metagenomic next-generation sequencing. Diagnostic performance was benchmarked against the final comprehensive clinical diagnosis, and the influence of metagenomic next-generation sequencing findings on antimicrobial therapy modification was analyzed. Sample size calculation, based on a prior study estimating an metagenomic next-generation sequencing detection rate of 85% (α = 0.05, β = 0.2), indicated a minimum of 68 cases; accordingly, 78 patients were enrolled.

Results

Computed tomography–guided biopsy was technically successful in all 78 patients (100%). The pathogen detection rate of metagenomic next-generation sequencing (91.0%, 71/78) was significantly higher than that of conventional culture (55.1%, 43/78; p < 0.001). Using the final clinical diagnosis as the reference standard, metagenomic next-generation sequencing achieved a sensitivity of 94.7% (95% confidence interval: 86.9–98.5), specificity of 100.0% (95% confidence interval: 29.2–100.0), positive predictive value of 100.0% (95% confidence interval: 94.9–100.0), and negative predictive value of 42.9% (95% confidence interval: 9.9–81.6). Among the 35 culture-negative specimens, metagenomic next-generation sequencing established a definitive microbiological diagnosis in 28 cases (80.0%) and detected polymicrobial infections in 11 cases (14.1% of the cohort). Antimicrobial therapy was rationally adjusted based on metagenomic next-generation sequencing results in 69.2% (54/78) of the patients.

Conclusions

The integration of computed tomography–guided precision biopsy with metagenomic next-generation sequencing offers a highly effective diagnostic approach for blood culture-negative systemic infections. This synergistic strategy improves etiological diagnosis by providing high-yield target specimens that enable comprehensive, unbiased pathogen screening, facilitates differentiation between infectious and non-infectious etiologies, and supplies critical evidence for guiding precision antimicrobial therapy. These findings highlight the growing role of interventional radiology in the contemporary framework of precision infectious disease management.

Introduction

Sepsis and systemic infections remain leading causes of morbidity and mortality in intensive care units (ICUs) worldwide, and the rapid and accurate identification of the causative pathogen is the cornerstone of effective clinical management. ¹ Despite advances in diagnostic microbiology, 30%–50% of sepsis cases are culture-negative because conventional methods such as blood culture often fail to identify the culprit microorganism.^2,3 This diagnostic challenge frequently compels clinicians to initiate empirical broad-spectrum antimicrobial therapy, which increases the risk of antimicrobial resistance and may delay effective targeted treatment when the true culprit pathogen is not targeted. ⁴ Immunocompromised patients pose a particular diagnostic challenge as they are susceptible to a broader and more atypical spectrum of causative pathogens. ⁵

Cross-sectional imaging modalities, primarily computed tomography (CT), are indispensable for localizing deep-seated infectious foci but cannot perform microbial identification. ⁶ Interventional radiology bridges this critical diagnostic gap; CT–guided percutaneous biopsy enables minimally invasive, image-directed sampling of core specimens directly from the site of infection. Its diagnostic utility and safety profile are well established for complex infectious diseases, including spondylodiscitis, intra-abdominal abscesses, and deep pelvic collections.^7–9 Importantly, evidence has suggested that specimens obtained directly from the infectious nidus harbor pathogen nucleic acid at concentrations several orders of magnitude higher than those in the peripheral blood, providing a more accurate representation of the causative agent. ¹⁰

Conventional microbial culture, although the gold standard for pathogen identification, has certain inherent limitations, including prolonged turnaround times; reduced sensitivity in the presence of prior antimicrobial administration; and poor recovery of fastidious organisms, anaerobes, intracellular bacteria, and viruses. ¹¹ Metagenomic next-generation sequencing (mNGS) represents a paradigm shift in infectious disease diagnosis; this culture-independent technique performs high-throughput sequencing of all nucleic acids in a clinical sample, enabling the theoretical detection of any bacteria, viruses, fungi, or parasites. ¹² With a typical turnaround time of 24–48 h and minimal interference from past antibiotic exposure, mNGS is rapidly transforming the diagnostic approach to infectious diseases. ¹³ However, the diagnostic performance of mNGS is substantially influenced by the quality of the input specimen and its microbial burden. ¹⁴

The diagnostic performance of mNGS is optimally evaluated when applied to tissue specimens obtained directly from the infectious focus, with conventional culture of the same tissue specimen serving as the most appropriate comparator. This head-to-head comparison, rather than comparison with blood cultures, forms the primary analytical framework of our study. Although mNGS has been validated for the analyses of sterile body fluids such as cerebrospinal fluid and plasma, tissue or aspirated pus obtained via direct interventional sampling is considered the optimal substrate for diagnosing systemic infections associated with discrete local lesions.^1,14,15 Therefore, this study was designed to systematically evaluate the diagnostic performance and clinical impact of a unified strategy that combines CT–guided precision interventional sampling with unbiased mNGS–based molecular diagnosis in patients with blood culture-negative systemic infections. We specifically focused on whether this combined strategy can establish a definitive etiological diagnosis and provide a rational basis for subsequent antimicrobial therapy.

In this study, “blood culture-negative systemic infections” are defined as those in which patients meet the Sepsis-3 criteria for suspected systemic infection, have negative results from at least two sets of blood cultures drawn at different time points, and have negative results from other relevant routine microbiological investigations (including sputum, urine, and cerebrospinal fluid cultures, as applicable). This definition pertains to the pre-biopsy conventional microbiological work-up and does not imply that tissue specimens obtained via biopsy would necessarily be culture-negative.

Materials and methods

Study design and patient cohort

This single-center retrospective observational cohort study was approved by the Institutional Review Board (Ethics Committee) of Sichuan Provincial Corps Hospital, Chinese People's Armed Police Force (Approval No. [2022]007). The study was conducted in accordance with the Declaration of Helsinki (1975, revised 2024). The requirement for written informed consent was waived by the ethics committee because of the retrospective nature of the study, and all patient data were fully anonymized prior to analysis. From April 2022 to March 2025, patients who met the inclusion criteria were consecutively enrolled from the hospital's interventional radiology database.

The following inclusion criteria were applied: (a) clinical diagnosis of suspected systemic infection meeting the Sepsis-3 international consensus definitions; (b) negative results from at least two sets of blood cultures drawn at different time points, and negative results from other relevant routine microbiological investigations; (c) availability of contrast-enhanced CT, magnetic resonance imaging (MRI), or positron emission tomography (PET)–CT imaging showing at least one radiologically accessible focal lesion with a minimum diameter of 1.0 cm; and (d) complete clinical, laboratory, and follow-up data for ≥1 month after the biopsy procedure. ³

The following exclusion criteria were applied: (a) any contraindication to percutaneous biopsy (e.g. uncorrectable coagulopathy, absence of a safe percutaneous needle trajectory, and severe hemodynamic instability); (b) refusal for the biopsy procedure by the patient/guardian; (c) pregnancy; (d) incomplete clinical, laboratory, or follow-up data; and (e) loss to follow-up within 1 month after the biopsy procedure.

CT–guided biopsy procedure

All CT–guided biopsy procedures were performed by board-certified interventional radiologists with ≥5 years of dedicated clinical experience in image-guided interventions. Preprocedural planning was conducted carefully, with multiplanar reformatting and three-dimensional reconstructions of the diagnostic contrast-enhanced CT used to design an optimal needle trajectory that avoided critical anatomical structures (e.g. major blood vessels, nerve bundles, and hollow viscera).

In patients with multiple radiologically identifiable lesions, the selection of the biopsy target was guided by a hierarchical consensus-based algorithm prioritizing the following factors in descending order of importance:

1. Safety and accessibility (primary consideration). The foremost criterion was a safe percutaneous needle trajectory that avoided critical anatomical structures, including major blood vessels, nerve bundles, hollow viscera, and the pleura (for pulmonary lesions). Lesions near these structures were deprioritized irrespective of other favorable characteristics as patient safety was deemed non-negotiable.

2. Metabolic activity (when available). For patients who underwent fluorine-18 fluorodeoxyglucose (¹¹⁸F-FDG) PET–CT imaging prior to the procedure—performed at the discretion of the referring clinician rather than as a mandatory protocol—the lesion with the maximum standardized uptake value (SUVmax) was preferentially selected as this typically indicates the most metabolically active region of infection and is associated with higher microbial burden and diagnostic yield. In patients without PET–CT imaging, this criterion was not applicable, and selection proceeded to the next factor.

3. Lesion size. Among lesions meeting the safety criteria, preference was given to larger lesions, specifically those with a maximum diameter of ≥2 cm as these are technically easier to target and generally yield more diagnostic material. When there was no lesion exceeding 2 cm, the largest measurable lesion was selected, provided it was ≥1.0 cm (the minimum size required for study inclusion).

In practice, when the lesion with the highest SUVmax was deemed unsafe for biopsy, the next-most metabolically active lesion that met the safety criteria was selected. The final decision was made by the interventional radiology team in consultation with the referring clinical team, with the procedural radiologist bearing ultimate responsibility for the safety assessment.

Procedures were conducted under intermittent CT–fluoroscopic guidance using a sterile technique. A coaxial needle technique with stepwise advancement was employed to ensure precise placement of the biopsy needle within the metabolically active areas of the target lesion (typically the enhancing rim or the interface between necrotic and viable tissue). An adequate volume of tissue core samples or purulent aspirate was collected for both microbiological and histopathological examination, immediately placed into sterile, nucleic acid-free containers, and promptly transported to the clinical laboratory for further processing.

Specimen processing and laboratory analyses

All specimens were promptly delivered by a dedicated laboratory courier to Guangzhou KingMed Center for Clinical Laboratory Co., Ltd, a certified clinical reference laboratory. All subsequent analytical steps were performed in strict adherence to the laboratory's accredited standard operating procedures (SOPs) and quality control protocols.

Conventional microbiological culture

Specimens were processed according to standard clinical microbiology SOPs. This included Gram staining and light microscopy for preliminary pathogen screening, followed by inoculation onto appropriate culture media (sheep blood agar, chocolate agar, MacConkey agar, Sabouraud dextrose agar, and anaerobic blood agar). For mycobacterial isolation, specimens were decontaminated and inoculated into a dedicated automated liquid culture system (Becton Dickinson Mycobacterial Growth Indicator Tube (MGIT), Becton Dickinson, USA). All inoculated media were incubated under appropriate atmospheric conditions (aerobic, anaerobic, 5%–10% CO₂–enriched) for standardized durations (48–72 h for aerobic and anaerobic bacteria and up to 6 weeks for mycobacteria and fungi). Culture results were continuously monitored and interpreted by trained clinical microbiologists, with contamination rigorously excluded based on standard microbiological criteria.

mNGS

The mNGS testing pipeline strictly followed the laboratory's validated SOP for metagenomic detection of infectious pathogens, with four core experimental stages:

1. Nucleic acid extraction. Total nucleic acid (both genomic DNA and total ribonucleic acid (RNA)) was extracted from the clinical specimen using an automated nucleic acid extraction platform (MagPure Compact Nucleic Acid Extraction Kit, Magen, China), with extraction controls included to monitor for experimental failure.

2. Library preparation. Extracted nucleic acids were fragmented using ultrasonication, subjected to end-repair and 3'-end A-tailing, followed by sequencing adapter ligation, and then amplified via limited-cycle polymerase chain reaction (PCR) to generate sequencing-ready libraries, with no-template controls included to exclude PCR contamination.

3. High-throughput sequencing. Qualified libraries were sequenced on either the BGIseq-500 or MGIseq-2000 high-throughput sequencing platforms (BGI, China) to generate high-depth, 150-bp paired-end reads, with a minimum depth of 10 million reads per sample.

4. Bioinformatics analysis. Raw sequencing data were subjected to a stringent bioinformatic pipeline; low-quality reads, adapter sequences, and reads aligning to the human reference genome (GRCh38) were removed using BWA and Bowtie2. The remaining high-quality microbial reads were aligned against a comprehensive, curated database containing genomic sequences of bacteria, viruses, fungi, and parasites (including >100,000 microbial species).

The reference laboratory's validated criteria for reporting a positive mNGS detection were as follows:

For bacteria and fungi, a minimum of ≥3 unique reads at the genus level was required. No separate threshold was applied at the species level; species-level assignment was reported when the genus-level detection met the threshold, and the bioinformatics pipeline could confidently resolve the species based on read mapping quality and genomic coverage.

For the Mycobacterium tuberculosis complex (MTBC) and viruses, a minimum of ≥1 unique read was sufficient. This lower threshold reflects two considerations:

1. These organisms are rarely environmental contaminants in tissue specimens, so even a single specific read has high clinical significance.

2. Their smaller genomes mean that a single unique read can represent a substantial proportion of the genome coverage.

For MTBC specifically, additional quality controls were applied, including confirmation of read mapping to the MTBC reference genome and exclusion of nontuberculous mycobacteria.

For all organisms, positive detections were not reported solely based on read counts. All detections underwent a mandatory dual-review process by a senior clinical microbiologist and a molecular diagnostics specialist who jointly assessed clinical plausibility (consistency with the patient's presentation, imaging, and histopathology) and laboratory plausibility (appropriate genomic coverage, expected amplicon sizes, and exclusion of known environmental or reagent contaminants using the laboratory's validated background database). Detections that fell below the read count threshold but were deemed clinically highly suspicious were flagged for additional investigation but were not reported as positive in the final clinical report.

All final mNGS reports underwent a dual-review process by a senior clinical microbiologist and a molecular diagnostics specialist before being issued to the clinical team.

Diagnostic criteria and reference standard

Final clinical microbiological diagnosis (reference standard)

This gold-standard diagnosis was established by an independent adjudication committee comprising three senior clinicians (one intensivist, one infectious disease specialist, one interventional radiologist) who were blinded to the biopsy specimen's mNGS and conventional culture results. The primary objectives of the adjudication were to confirm the presence or absence of an infectious etiology and, when possible, to identify the causative pathogen. The final diagnosis was formulated through a comprehensive, blinded review of each patient's clinical course, including detailed clinical history, response to antimicrobial or other therapy, all available laboratory and serological results, histopathological findings from the biopsy specimen, and follow-up data from ≥1 month after the biopsy.

Criteria for infectious etiology included the following: (a) a clinical syndrome consistent with the Sepsis-3 consensus criteria; (b) radiological evidence of a focal lesion consistent with infection on cross-sectional imaging; (c) sustained clinical and radiological improvement following initiation of targeted antimicrobial therapy directed at the identified pathogen; and (d) additional supportive microbiological evidence (e.g. subsequent seroconversion for specific pathogens, positive culture from another clinical site, or confirmatory PCR for the identified pathogen). ³

Criteria for non-infectious etiology included the following: (a) no clinical or radiological improvement with broad-spectrum antimicrobial therapy; (b) clinical and radiological improvement following initiation of immunosuppressive therapy or other non-antimicrobial treatment; (c) histopathological confirmation of a malignant neoplasm or non-infectious inflammatory disorder; and (d) no supportive microbiological evidence for infection after comprehensive evaluation.

mNGS true positive was defined based on the detection of one or more microorganisms meeting the laboratory's validated read count threshold, with the identified pathogen(s) deemed clinically plausible and the patient meeting the criteria for infectious etiology (as defined above).

mNGS positive was defined as the detection of one or more microorganisms with a read count meeting or exceeding the laboratory's validated reporting threshold, with the identified pathogen(s) deemed clinically plausible for the patient's clinical presentation and radiological findings.

Culture positivity was defined as the growth of a recognized pathogenic microorganism on standard microbiological culture media, with contamination rigorously excluded based on standard clinical microbiology criteria.

Statistical analyses

All data analyses were performed using the Statistical Package for Social Sciences (SPSS) software (version 26.0, IBM Corp., Armonk, NY, USA). Continuous variables with a normal distribution are presented as mean ± SD; non-normally distributed continuous data are summarized as median and interquartile range (IQR) values. Categorical variables are expressed as counts and percentages (n, %).

Results

Patient characteristics

In total, 78 patients (45 men, 57.7% and 33 women, 42.3%) with a mean age of 57.8 ± 15.2 (range: 22–85) years were included. All patients (100.0%) presented with fever (>38.0°C), and the median symptom duration prior to biopsy was 18 (IQR: 10–30) days. A substantial proportion were immunocompromised (33.3%, 26/78), with 19.2% (15/78) receiving active immunosuppressive therapy. Prior antimicrobial exposure was documented in 83.3% (65/78) of the patients. The most common biopsy sites were the lungs (35.9%, 28/78), liver (20.5%, 16/78), spine/paraspinal region (16.7%, 13/78), and deep lymph nodes (15.4%, 12/78). Complete patient characteristics are summarized in Table 1. No major procedural complications (e.g. hemorrhage requiring intervention, pneumothorax requiring chest tube placement, or inadvertent organ injury) occurred, confirming the safety of CT–guided biopsy in this cohort.

Table 1.

Baseline characteristics of the study cohort (n = 78).

Characteristic	Value (n, %)/descriptive statistics
Age, years (mean ± SD, range)	57.8 ± 15.2, 22–85
Sex (male/female)	45 (57.7%)/33 (42.3%)
Clinical Presentation
Fever (>38.0°C)	78 (100.0%)
Symptom duration prior to biopsy, days (median, IQR)	18 (10–30)
Immunocompromised status	26 (33.3%)
- Active immunosuppressive therapya	15 (19.2%)
- Post solid organ transplantation	5 (6.4%)
- Human immunodeficiency virus (HIV) infection	3 (3.8%)
- Active hematological malignancy/Post chemotherapy	3 (3.8%)
Antimicrobial exposure prior to biopsy	65 (83.3%)
Biopsy site
Lung	28 (35.9%)
Liver	16 (20.5%)
Spine/Paraspinal region	13 (16.7%)
Deep lymph nodes	12 (15.4%)
Other (spleen, kidney, musculoskeletal)	4 (5.1%)

^aIncludes long-term glucocorticoid use, disease-modifying antirheumatic drugs, and biologic immunomodulatory agents.

IQR: interquartile range.

Diagnostic performance: mNGS versus conventional microbiological culture

After blinded adjudication by the independent clinical committee, the final clinical microbiological diagnosis confirmed an infectious etiology in 75 patients (96.2%) and a non-infectious etiology in three patients (3.8%), including two cases of non-Hodgkin lymphoma and one case of autoimmune inflammatory disorder. The overall pathogen detection rate was significantly higher for mNGS than for conventional culture (91.0%, 71/78 vs. 55.1%, 43/78; p < 0.001). Relative to the final clinical microbiological diagnosis (reference standard), conventional culture showed a sensitivity of only 57.3% (43/75), specificity of 100.0% (3/3), positive predictive value (PPV) of 100.0% (43/43), and negative predictive value (NPV) of 8.6% (3/35). Table 2 details the complete diagnostic performance metrics for both methods.

Table 2.

Diagnostic performance of mNGS and conventional microbiological culture, using the final clinical microbiological diagnosis as the reference standard (n = 78).

Diagnostic method	Sensitivity, % (95% CI)	Specificity, % (95% CI)	PPV, % (95% CI)	NPV, % (95% CI)
mNGS	94.7 (86.9–98.5)	100.0 (29.2–100.0)	100.0 (94.9–100.0)	42.9 (9.9–81.6)
Conventional culture	57.3 (45.4–68.7)	100.0 (29.2–100.0)	100.0 (91.8–100.0)	8.6 (1.8–23.1)

mNGS: metagenomic next-generation sequencing; CI: confidence interval; PPV: positive predictive value; NPV: negative predictive value.

Incremental diagnostic value of mNGS

In the 43 culture-positive specimens, mNGS showed excellent concordance with conventional culture, including 90.7% (39/43) at the genus level and 74.4% (32/43) at the species level. Discrepancies between the two methods primarily arose from two factors: (a) mNGS detected additional pathogens in polymicrobial infections (n = 6) and (b) mNGS provided more precise species-level identification than culture (n = 4).

To comprehensively examine discordant cases, we prepared a supplementary analysis (Supplementary Table S1) stratifying all discordant cases into two categories:

1. Culture-positive/mNGS-negative cases (n = 4). In these instances, conventional culture detected organisms that mNGS failed to identify. Several factors may account for these discrepancies, including low microbial nucleic acid burden in the sampled tissue, degradation of microbial DNA during transport, difficulty in lysing organisms with robust cell walls (e.g. fungi), or—critically—possible contamination of the positive culture result rather than true infection (a well-recognized limitation of conventional culture). As acknowledged in the Discussion section, mNGS detects microbial nucleic acids rather than viable organisms and cannot distinguish between live pathogens, dead organisms, or environmental contaminants; this inherent limitation, coupled with the potential for culture contamination, may explain some discordant results. Notably, all four culture-positive/mNGS-negative cases were judged as true infections by the independent committee based on comprehensive clinical and radiological criteria, suggesting genuine false-negative mNGS results rather than culture contamination; however, the precise mechanisms remain speculative.

2. Culture-negative/mNGS-positive cases (n = 28). This category represents the principal diagnostic advantage of mNGS, enabling detection of fastidious organisms, anaerobes, fungi, and polymicrobial infections that are difficult or impossible to identify using conventional culture. The pathogens identified included Nocardia spp., Brucella spp., nontuberculous mycobacteria, and various fungal species—organisms notoriously challenging to culture, particularly given that 83.3% of our cohort had prior antimicrobial exposure. Importantly, all 28 cases were confirmed as true positives by the independent adjudication committee based on clinical response to pathogen-directed therapy and/or supportive microbiological evidence.

The marked predominance of culture-negative/mNGS-positive cases (28 vs. 4 cases) highlights the superior sensitivity of mNGS in this challenging population, while the existence of culture-positive/mNGS-negative cases (4 cases) underscores an important limitation. A negative mNGS result does not definitively exclude infection, particularly in cases with low microbial burden or organisms that are difficult to detect.

The principal clinical value of mNGS was demonstrated in the 35 culture-negative specimens—an especially challenging subgroup. Within this cohort, mNGS identified a causative pathogen in 28 cases (80.0%), providing a definitive etiological diagnosis where conventional methods had failed. The pathogens uncovered included Nocardia spp. (3 cases), Brucella spp. (2 cases), and various nontuberculous mycobacteria species (5 cases). Supplementary Table S2 details the specific pathogens identified using mNGS in each of the 28 culture-negative cases with corresponding read counts and clinical context. Furthermore, mNGS detected polymicrobial infections in 11 patients (14.1% of the cohort), including bacterial–fungal and bacterial–bacterial co-infections—a finding frequently missed on routine conventional culture. The spectrum of pathogens detected using mNGS was broad, encompassing Gram-positive bacteria (31.0% of pathogen-positive cases), Gram-negative bacteria (29.6%), mycobacteria (19.7%), fungi (11.3%), and other pathogens (8.4%).

Representative case illustration

A 62-year-old man with rheumatoid arthritis on long-term methotrexate therapy presented with a 4-week history of severe, refractory lower back pain and low-grade fever (body temperature: 38.2°C–38.8°C). Contrast-enhanced CT of the chest and spine revealed an expansile, osteolytic lesion in the right scapula with associated soft tissue swelling and enhancement. The primary clinical suspicion, based on imaging, was a primary bone neoplasm or metastatic malignancy; however, an indolent infectious process remained in the differential diagnosis. A CT–guided core needle biopsy of the scapular lesion was performed for histopathological examination and microbiological analysis (conventional culture + mNGS). Histopathology showed a dense, non-specific inflammatory infiltrate rich in lymphocytes and plasma cells, inconclusive for either a neoplastic or infectious etiology. In striking contrast, the mNGS report unequivocally identified a high burden of Mycobacterium avium complex (MAC) sequences (1256 unique reads) with no other pathogens detected. MAC osteomyelitis is a recognized but uncommon entity that can radiographically mimic bone tumors, particularly in immunocompromised patients. In this instance, the combination of CT–guided biopsy and mNGS was decisive. It enabled a confirmed diagnosis of an easily overlooked atypical mycobacterial infection, distinguished it from a malignant neoplasm, and provided the evidence required to initiate a prolonged, tailored antimycobacterial regimen (azithromycin, rifampicin, ethambutol). The patient achieved significant clinical and radiological improvement after 8 weeks of targeted therapy, without unnecessary oncological work-up or invasive surgery. Imaging and pathological findings are presented in Figures 1 to 4, with detailed descriptions in the figure legends.

Figure 1.

Sagittal magnetic resonance imaging sequences showing osseous abnormalities at the T11 vertebral body. (a to d) The infectious lesion is indicated by arrowheads. (a) T1-weighted image (T1WI); (b) T2-weighted image (T2WI); (c) Fat-suppressed T2-weighted image (FS-T2WI), sagittal view; (d) Fat-suppressed T2-weighted image (FS-T2WI), axial view.

Figure 4.

Histopathological section from the right-scapula lesion biopsy specimen. Arrowheads highlight proliferating lymphocytes and plasma cells within the tissue. The specimen demonstrates inflammatory cellular infiltrate.

Figure 2.

Axial and sagittal computed tomography (CT) scans (bone window) show multifocal osteolytic bone destruction. (a to d) Arrowheads indicate each osteolytic infectious lesion. (a) Lytic lesion of the right scapula; (b) destruction of the T2 vertebral body; (c) lesion of the posterior segment of the right ninth rib; (d) osteolytic destruction of the T11 vertebral body.

Figure 3.

Fused ¹⁸F-FDG PET/CT images. Arrows mark sites with elevated radiotracer uptake corresponding to the osteolytic infectious bone lesions.

CT: computed tomography; ¹⁸F-FDG: fluorine-18 fluorodeoxyglucose; PET: positron emission tomography.

Therapeutic impact of mNGS

The definitive microbiological data provided by mNGS exerted a direct and substantial impact on clinical management; antimicrobial therapy was rationally adjusted based on mNGS findings in 54 of 78 patients (69.2%), and all adjustments were subsequently validated by favorable clinical and radiological outcomes, with no associated clinical deterioration. These adjustments fell into three categories:

1. Antimicrobial de-escalation. Narrowing the antimicrobial spectrum in 23 patients (42.6% of adjusted cases) once a specific pathogen was identified, thereby reducing unnecessary broad-spectrum exposure and minimizing the risk of antimicrobial resistance.

2. Targeted antimicrobial escalation. Starting or switching to pathogen-directed therapy in 26 patients (48.1% of adjusted cases) for organisms that were either unexpected (e.g. Nocardia spp. and Brucella spp.) or had known resistance profiles.

3. Expanded antimicrobial coverage. Adding combination or broader-spectrum therapy in five patients (9.3% of adjusted cases) to target polymicrobial infections identified exclusively using mNGS.

Preliminary analysis of clinical outcomes

Although this study was designed as a retrospective diagnostic accuracy study without a contemporaneous control group, we collected short-term outcome data to provide preliminary context for the clinical impact of the combined strategy. Among the 75 patients with confirmed infectious etiology, the median follow-up after initiation of mNGS–guided therapy was 28 (IQR: 21–42) days. During this period, the overall treatment success rate, defined as complete resolution or significant improvement of clinical symptoms, radiological improvement of the infectious focus, and no need to change the regimen because of treatment failure, was 88.0% (66/75). Among the 66 successfully treated patients, none required a regimen change during the 28-day follow-up; they continued the mNGS–guided regimen initiated after biopsy results became available. Clinical and radiological improvement was observed within the 28-day window, although we recognize that for a subset—particularly those with nontuberculous mycobacterial (NTM) infections—this improvement represented the early phase of a longer expected treatment course rather than complete resolution.

For the nine patients with treatment failure, therapy was adjusted within 28 days because of documented insufficient clinical response (e.g. persistent fever, progressive radiological findings, and clinical deterioration). Importantly, the decision to adjust therapy was based on a lack of favorable trend by day 14–21, rather than failure to achieve complete resolution, which would have been unrealistic for chronic infections within this period. The 28-day endpoint was selected pragmatically to allow early assessment of treatment response while minimizing loss to follow-up. For chronic infections, the observed “success” at 28 days should be interpreted as an early favorable response rather than definitive cure, with long-term outcomes requiring extended follow-up beyond the scope of this study.

In the immunocompromised subgroup (n = 26), the treatment success rate was 80.8% (21/26), which is clinically meaningful given the higher risk of treatment failure in this population. Six deaths occurred in the cohort of 75 patients with infectious etiology (8.0%); the independent adjudication committee attributed all deaths to progression of underlying severe comorbidities (e.g. end-stage liver disease and advanced malignancy) or multi-organ failure unrelated to the primary infectious process or the biopsy procedure.

Discussion

This study provides the first systematic evaluation of an integrated diagnostic pathway combining CT–guided precision tissue sampling with tissue-based mNGS specifically in patients with blood culture-negative systemic infections. Although previous studies have separately validated the clinical utility of interventional radiological biopsy and mNGS for infectious disease diagnosis, our key advancement lies in demonstrating the synergistic diagnostic performance and clinical utility of this combined, closed-loop workflow.^8–10,13 This integrated strategy directly addresses the core clinical dilemma of managing blood culture-negative systemic infections, which is, how to rapidly obtain high-fidelity, microbially rich specimens from the epicenter of infection and perform comprehensive, unbiased pathogen identification when conventional tests yield negative results.

The 91.0% pathogen detection rate and 94.7% sensitivity achieved using mNGS on CT–guided biopsy specimens substantially exceed the reported sensitivities of plasma cell–free DNA mNGS for sepsis (typically 50%–85%).^11,14 This stark contrast underscores the critical advantage of direct tissue sampling; biopsy specimens provide exponentially higher microbial loads and lower background human nucleic acid compared with peripheral blood, directly improving the diagnostic performance of mNGS.^10,11 More importantly, this integrated strategy enabled rational, pathogen-directed antimicrobial therapy adjustments in approximately 70% of the patients, facilitating a crucial transition from empirical, population-based therapy to individualized, precision management—a shift fully aligned with antimicrobial stewardship principles and the global effort to combat antimicrobial resistance.^11,16

A distinctive and clinically valuable attribute of this integrated approach is its capacity to resolve diagnostically ambiguous cases where cross-sectional imaging and even histopathology are inconclusive. As illustrated by our representative case, when an infectious process radiologically mimics a malignant neoplasm and tissue pathology demonstrates only non-specific inflammation, mNGS can serve as a decisive diagnostic arbiter by providing definitive microbiological evidence of infection. This ability to accurately discriminate between infectious and non-infectious inflammatory or neoplastic conditions is clinically critical because it can prevent protracted diagnostic delays, avert inappropriate and potentially harmful treatments (such as unnecessary chemotherapy, radiation therapy, or surgery), and guide timely initiation of appropriate therapy.

At our institution and at the reference laboratory, targeted PCR assays are available but are not routinely performed as part of the standard diagnostic workflow for the complex, blood culture-negative systemic infections included in this study. Specifically, targeted PCRs (including those for Mycobacterium tuberculosis, Nocardia spp., Brucella spp., and certain fungi such as Aspergillus and Cryptococcus) are used selectively when there is a strong clinical or radiological suspicion for a specific pathogen. These assays are performed at the same reference laboratory where our mNGS specimens were processed, with a typical turnaround time of 24–72 h, depending on the specific pathogen and assay type. We did not include targeted PCR results in our diagnostic performance analysis for two reasons:

1. These assays are not performed systematically on all biopsy specimens, but only on a case-by-case basis when clinically indicated.

2. Targeted PCRs, by design, can only detect pre-specified pathogens and cannot identify unexpected, rare, or novel organisms—a key advantage of the untargeted mNGS approach.

Based on the results of this study and clinical experience, we recommend prioritizing the CT–guided biopsy combined with tissue mNGS diagnostic strategy for blood culture-negative patients with suspected systemic infection who exhibit the following high-risk characteristics: (a) an immunocompromised state (e.g. long-term immunosuppressive therapy, solid organ transplantation, Human immunodeficiency virus (HIV) infection, and hematological malignancy); (b) at least one radiologically accessible focal lesion ≥1.0 cm in diameter on cross-sectional imaging; (c) failure to respond to, or rapid clinical progression despite, empirical broad-spectrum antimicrobial therapy; and (d) a clinical differential diagnosis that includes malignant neoplasm or atypical infection (e.g. with mycobacterial, fungal, Nocardia, and Brucella).

The diagnostic yield and potential to avert misdirected management are likely highest in this subgroup.

Looking forward, the potential of this integrated diagnostic framework extends far beyond pathogen identification. The same high-depth mNGS dataset can be computationally interrogated for antimicrobial resistance genes, virulence factors, and strain-typing information. ¹⁷ This paves the way for a future “sample-to-answer” diagnostic paradigm where a single minimally invasive CT–guided biopsy can simultaneously identify the causative pathogen and predict its antimicrobial susceptibility profile—a potentially transformative advance for managing complex, multidrug-resistant infections, which are a major global public health threat. ¹⁸

Study limitations

Some important study limitations have been acknowledged.

First, this was a single-center, retrospective observational cohort study, which is inherently subject to selection bias and has limited generalizability; our results therefore require validation in larger, multicenter, more diverse populations (including those from different geographic regions and healthcare settings).

Second, the absence of a contemporaneous control group of patients managed without this integrated strategy precludes definitive assessment of the causal impact of CT–guided biopsy + mNGS on “hard” clinical endpoints such as all-cause mortality, ICU length of stay, hospital length of stay, and total healthcare costs.

Third, although the final clinical microbiological diagnosis was established by a blinded, multidisciplinary independent adjudication committee (representing a strength of the study), the reference standard remains partially subjective and reliant on integrated clinical judgment rather than an absolute microbiological gold standard (e.g. isolation of the pathogen from a sterile site).

Fourth, the small number of non-infectious cases (n = 3) substantially limits the robustness of our specificity and PPV estimates, as reflected by the wide 95% confidence intervals (specificity: 29.2%–100.0%; PPV: 94.9%–100.0%). This small sample precludes a reliable estimation of the true false-positive rate of tissue-based mNGS. False-positive results—whether due to contamination, colonizing organisms, or persistent nucleic acid from prior infections—could lead to the administration of unnecessary antimicrobial therapy or inappropriate treatment of non-infectious conditions. Larger prospective studies with adequate non-infectious controls are essential to definitively establish the true specificity of this approach. This limitation is discussed in conjunction with our earlier acknowledgment of mNGS's inherent inability to distinguish between viable pathogens, non-viable organisms, and environmental contaminants, which further complicates the interpretation of positive results in the absence of a robust reference standard.

Fifth, the inherent limitations of the mNGS technology must be explicitly acknowledged:

1. mNGS detects microbial nucleic acids rather than viable organisms, and thus cannot distinguish between live bacteria, dead bacteria, or environmental contaminants, necessitating cautious interpretation of low-biomass results within the clinical context. ¹¹

2. The substantial cost of mNGS testing necessitates careful consideration of cost-effectiveness in different healthcare settings, particularly in low- and middle-income countries.

3. Variability in bioinformatics pipelines and database comprehensiveness across laboratories can influence results, highlighting the need for standardization of mNGS testing. ¹⁹

Finally, the 28-day follow-up is a limitation for chronic infections (e.g. NTM and fungal infections), as significant improvement may require >8 weeks. For these indolent infections, our observed “treatment success” at 28 days should be interpreted as an early favorable response rather than definitive cure. Future studies should incorporate longer-term outcome assessment, particularly for NTM, fungal, and other slow-growing infections, to evaluate sustained success and potential late complications.

Notwithstanding these limitations, this study has several important strengths, including a well-defined cohort with strict inclusion and exclusion criteria, a statistically calculated sample size, a blinded independent adjudication committee, comprehensive assessment of both diagnostic performance and clinical impact, and rigorous adherence to standard clinical and laboratory SOPs.

Future research directions

Prospective, multicenter, randomized controlled trials are warranted to confirm our findings and rigorously quantify the full clinical, microbiological, and economic benefits of the CT–guided biopsy + mNGS integrated strategy for blood culture-negative systemic infections. Future research should also explore the following: (a) integration of complementary “omics” data (e.g. host transcriptional response signatures and metabolomics) with mNGS data to further enhance diagnostic accuracy, enable prognostic stratification, and deepen our understanding of host-pathogen interactions in systemic infections; (b) development of standardized, cost-effective mNGS pipelines for tissue specimens to improve accessibility and reduce costs; (c) clinical utility of mNGS for detecting antimicrobial resistance genes in biopsy specimens to guide early, targeted therapy for multidrug-resistant infections; and (d) application of this integrated strategy in specific high-risk populations (e.g. pediatric patients, neonates, and solid organ transplant recipients) with blood culture-negative systemic infections.^20,21

Conclusion

The synergistic combination of CT–guided precision percutaneous biopsy and mNGS constitutes a robust and highly effective diagnostic platform for managing blood culture-negative systemic infections—a common and clinically challenging condition. This integrated strategy directly addresses a major unmet clinical need by significantly improving pathogen detection rates, enabling accurate differentiation between infectious and non-infectious etiologies, and supplying the critical microbiological evidence needed to guide precision antimicrobial therapy. These findings underscore the evolving and central role of interventional radiology as a key enabler in the ongoing global transition towards precision medicine in the management of complex infectious diseases.