Diagnostic and Therapeutic Impact of Metagenomic Next-Generation Sequencing in Tuberculous Osteoarticular Infections with Negative or Confounding Conventional Cultures Authors: Zhirui Xue, Hongyan Li, Xianming Wang, Yuanqing Cai, Zida Huang, Wenbo Li, Yang Su, Zhaoyang Wu, Xinyu Fang, Wenming Zhang Published: 2026 ABSTRACT PurposeDiagnosing tuberculous osteoarticular infections (TB-OAI) remains challenging due to frequent false-negative or confounding conventional culture results. This study evaluated the diagnostic and therapeutic utility of metagenomic next-generation sequencing (mNGS) for occult TB-OAI in patients presenting with negative or misleading culture outcomes.Patients and MethodsWe retrospectively analyzed 13 patients with confirmed TB-OAI, encompassing periprosthetic, fracture-related, and native joint infections. Patients were stratified by conventional culture results into strictly culture-negative (n=8) and culture-confounded (n=5; yielding non-mycobacterial organisms) groups. A composite reference standard of mNGS positivity combined with histopathological or clinical validation established the definitive diagnosis. We assessed diagnostic yield, therapeutic modifications, and clinical outcomes.ResultsConventional culture failed to identify Mycobacterium tuberculosis in all 13 cases (0% sensitivity) and yielded misleading non-mycobacterial flora in 5 cases (38.5%). Conversely, mNGS successfully identified the pathogen in 100% (13/13) of patients, corroborated by histopathology in all cases. Consequently, mNGS results changed clinical management from empirical antibiotics to targeted anti-tuberculosis therapy in all cases (100%). Postoperative erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) levels decreased significantly (P < 0.05). Over a mean follow-up of 17.5 ± 3.0 months, 12 patients achieved durable infection eradication. One patient experienced early recurrence requiring a two-stage revision, ultimately achieving successful infection control.ConclusionmNGS serves as a promising diagnostic rescue tool for occult TB-OAI when conventional cultures are negative or misleading. While limited by sample size, these preliminary findings suggest mNGS effectively guides the transition from empirical to targeted anti-tuberculosis therapy and limits diagnostic delays. Introduction Osteoarticular infections (OAI), encompassing implant-related infections, primary septic arthritis, and osteomyelitis, are devastating complications that severely compromise patient mobility, induce excruciating pain, and carry the potential for life-threatening systemic involvement. Consequently, these infections significantly diminish patients’ quality of life and impose a substantial economic burden on global healthcare systems.1,2 Among the diverse etiology of OAI, Mycobacterium tuberculosis stands out as a particularly formidable adversary. Often characterized as the “great imitator,” tuberculous osteoarticular infections (TB-OAI) present a rare yet formidable clinical challenge. Osteoarticular tuberculosis (TB-OAI) accounts for approximately 1% to 3% of all tuberculosis cases and 10% to 15% of extrapulmonary tuberculosis cases. Delayed diagnosis of TB-OAI frequently leads to severe outcomes, including irreversible joint destruction and neurological compromise.3,4 Because TB-OAI typically manifests with an insidious onset, a prolonged clinical course, and highly atypical radiographic and symptomatic presentations, early and accurate identification is exceptionally difficult, frequently resulting in diagnostic delays or misdiagnoses.5–7 Accurate and timely pathogen identification is fundamental to the successful implementation of antimicrobial stewardship and surgical interventions. Although conventional microbial culture remains the gold standard, this method has critical limitations when addressing the Mycobacterium tuberculosis complex (MTBC). First, MTBC culture methods are notoriously time-consuming, requiring 2 to 8 weeks for results, and exhibit a low positivity rate, rendering them inadequate for urgent clinical guidance.8 Furthermore, the formation of biofilms9 and the suppressive effects of prior empirical antibiotic therapy frequently result in strictly negative culture outcomes.10,11 In clinical practice, conventional culture results present two distinct diagnostic challenges for occult TB-OAI. “Culture-negative” cases yield no microbial growth, leaving clinicians without a confirmed pathogen. Conversely, “culture-confounded” cases yield non-mycobacterial organisms (such as Staphylococcus spp. or Escherichia coli) that provide a plausible alternative diagnosis, thereby masking the underlying tuberculous infection and directing management toward ineffective empirical antibiotics. This misclassification plunges the patient into a prolonged diagnostic odyssey, frequently characterized by recurrent surgical failures.12,13 In recent years, metagenomic next-generation sequencing (mNGS) has emerged as a revolutionary, culture-independent diagnostic modality. By comprehensively analyzing total nucleic acids in clinical samples, mNGS provides unbiased pathogen detection with significantly reduced turnaround times, driving its widespread application in clinical infectious diseases.14,15 Previous studies have consistently demonstrated that mNGS offers superior sensitivity compared to traditional culture methods in identifying the causative agents of complex OAI.16,17 Despite these advancements, a critical knowledge gap has persisted. To date, the majority of existing literature has focused broadly on the general diagnostic performance of mNGS. However, evidence remains limited regarding mNGS-guided management in TB-OAI cases where conventional cultures are negative or yield plausible yet misleading results. Therefore, we conducted a retrospective study to evaluate the diagnostic capability of mNGS in unmasking MTBC infections confounded by these specific conventional culture outcomes. Our secondary objective was to quantify the clinical impact of these mNGS findings, specifically assessing how they drive a definitive therapeutic pivot from ineffective antimicrobial regimens to targeted anti-tuberculosis therapy, thereby optimizing patient outcomes. Methods Study Design and Patient Enrollment This study was approved by the Ethics Committee of the First Affiliated Hospital of Fujian Medical University. This single-center, retrospective case series enrolled patients with OAI who were diagnosed with and treated for Mycobacterium tuberculosis infection between January 2020 and February 2025. All diagnoses were confirmed by a multidisciplinary team comprising two senior orthopedic surgeons, two senior microbiologists, and two senior infectious disease specialists, adhering to the diagnostic criteria established by the 2018 International Consensus Meeting (ICM). OAI was defined to encompass the following categories: (1) native joint infection; (2) periprosthetic joint infection (PJI), including persistent infection following one-stage or two-stage revision, as well as debridement, antibiotics, and implant retention (DAIR); (3) implant-associated infection excluding PJI, such as fracture-related infection; and (4) osteomyelitis. Demographic and baseline clinical data were systematically collected for all participants, encompassing history of tuberculosis, clinical symptoms, inflammatory markers preoperatively and at final follow-up, T-SPOT.TB assay results, preoperative and/or intraoperative microbial culture results, intraoperative histopathological findings, mNGS results and sequence read counts, treatment strategies, antimicrobial regimens prior to and following mNGS testing, infection control status, and diagnostic delay. Given the inherently low sensitivity of conventional microbial culture for Mycobacterium tuberculosis, a composite reference standard was employed. A definitive diagnosis was established by the detection of MTBC sequences via mNGS, concomitant with at least one of the following confirmatory criteria: (1) histopathological evidence of caseous necrosis, chronic granulomatous inflammation, or Langhans giant cells; (2) a positive T-SPOT.TB assay; or (3) substantial resolution of clinical symptoms and a significant, sustained decline in inflammatory markers (erythrocyte sedimentation rate [ESR] and C-reactive protein [CRP]) following the initiation of standard anti-tuberculosis therapy directed by mNGS results. Inclusion criteria were defined as follows: (1) confirmed OAI attributable to Mycobacterium tuberculosis; (2) adequate sampling for pathogen detection, including both conventional microbial culture and mNGS; and (3) complete medical documentation. Exclusion criteria comprised: (1) incomplete medical records; (2) sample contamination or suspected contamination; (3) concurrent infections at extra-articular or internal sites; and (4) a follow-up duration of less than one year. Postoperative inflammatory markers were extracted from the final follow-up visit. The cohort was subsequently stratified into two groups based on conventional microbial culture outcomes: a culture-confounded group (yielding non-mycobacterial organisms) and a strictly culture-negative group. Specimen Collection and Microbial Culture Diagnostic specimens were collected preoperatively or intraoperatively. To ensure diagnostic consistency, identical clinical specimens were apportioned for both testing modalities. Three sample aliquots were submitted for conventional microbial culture, and one aliquot was submitted for mNGS analysis per patient. This replicate culture strategy was employed to maximize diagnostic yield and minimize sampling error. Specimens were transferred to the microbiology laboratory for culture within 30 minutes of collection, and any remaining sample aliquots were stored at −80°C. An optimized culture protocol was implemented to maximize microbial yield. Liquid specimens (eg, synovial fluid, pus) were inoculated into Bactec Plus Aerobic/F and Anaerobic/F blood culture bottles (Becton Dickinson, Germany) or Bactec Peds Plus/F bottles (Becton Dickinson) and incubated for 14 days in a Bactec 9050 automated incubator (Becton Dickinson). Solid or tissue specimens were placed in sterile Eppendorf tubes containing 1 mL of tryptic soy broth (Qingdao Hope Bio-Technology Co., Ltd., China). Samples were processed in an automated tissue homogenizer until complete homogenization was achieved. The resulting homogenate was inoculated onto blood agar plates and incubated under both aerobic and anaerobic conditions for 14 days. Intraoperative implants (eg, prostheses, intramedullary nails) were placed in sterile containers and submerged in approximately 400 mL of sterile saline. To dislodge adherent biofilms, the containers were subjected to sonication (40 kHz for 5 minutes), followed by centrifugation of the sonicate fluid at 4000 rpm for 5 minutes. The supernatant was discarded, and the resulting pellet was resuspended in sterile phosphate-buffered saline (PBS). This suspension was inoculated into Bactec Plus/F and Bactec Lytic/10/F culture bottles and incubated for 14 days in a Bactec 9050 automated system (Becton Dickinson). mNGS Sample pretreatment for mNGS was performed as detailed in the microbiological culture section, and subsequent sequencing was conducted according to previously established protocols.18 For liquid samples, a 500 μL aliquot was utilized. Tissue specimens underwent complete homogenization prior to analysis. The mNGS procedure was executed as follows: (i) Nucleic acid extraction: Total deoxyribonucleic acid (DNA) was extracted from 500 μL of liquid specimen or tissue homogenate using the TIANamp Micro DNA Kit (DP316; Tiangen Biotech, China) in strict accordance with the manufacturer’s instructions. (ii) Library construction and sequencing: Extracted DNA was randomly fragmented into 200–300 base pair (bp) segments. Library concentrations were quantified utilizing the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, USA). Following circularization, the libraries were amplified via rolling circle amplification to generate DNA nanoballs (DNBs). The prepared DNBs were loaded onto a sequencing flow cell and sequenced on the BGISEQ-500 platform (BGI Genomics, China). (iii) Bioinformatics analysis: Raw sequencing data were filtered to remove low-quality reads and sequences shorter than 35 bp. Human host DNA was subsequently depleted by mapping the remaining reads against the human reference genome (hg19) using the Burrows-Wheeler Aligner (BWA). The retained high-quality, non-host reads were aligned against comprehensive microbial genomic databases for precise taxonomic classification into bacterial, fungal, viral, and parasitic domains. Interpretation of mNGS Results The interpretation of mNGS results was performed in strict accordance with previously established criteria.19–22 Genome coverage was defined as the cumulative length of the detected pathogen sequences divided by the total length of the reference genome. Relative abundance at the genus level was calculated as the proportion of a specific microbial genus relative to all detected genera within the same major taxonomic group (ie, bacteria, fungi, viruses, or parasites). Based on these parameters, diagnostic positivity thresholds were established as follows: (i) Background Contaminants and Quality Control: To monitor for environmental and reagent contamination, a blank control sample containing a standardized concentration of background nucleic acid (human genomic DNA) without microbial templates was processed and sequenced concurrently with each batch of clinical samples. Microorganisms commonly recognized as background flora and frequently detected across various sample types in our laboratory (eg, Burkholderia, Ralstonia, Delftia, Sphingomonas, Aspergillus, and Sodaria spp.) were classified as pathogenic only if their relative abundance at the genus level was ≥80%. (ii) Common Pathogens: For other pathogenic genera, the organism exhibiting the highest genome coverage and the highest standardized stringently mapped read number at the species level (SDSMRNS) was identified as the causative pathogenic species. (iii) MTBC Detection: Given the inherently low nucleic acid yield of mycobacteria, the Mycobacterium tuberculosis complex (MTBC) was classified as pathogenic if the standardized stringently mapped read number at the genus level (SDSMRNG) was ≥1. Outcome Measurement Treatment success was defined by the fulfillment of all the following criteria at the final follow-up visit: (1) complete absence of clinical symptoms associated with local infection; (2) normalization of serum inflammatory markers; (3) lack of radiographic evidence indicating persistent infection (eg, implant loosening, progressive bone resorption, or osteolysis); and (4) definitive clinical assessment by the attending surgeon confirming infection eradication. Conversely, treatment failure was designated if infection recurrence was identified or if one or more subsequent surgical interventions were required for infection control. The diagnostic criteria for recurrent infection were strictly based on the 2018 Musculoskeletal Infection Society (MSIS) guidelines for periprosthetic joint infection.23 Statistical Analysis Continuous variables were assessed for normality. Normally distributed continuous variables were presented as means ± standard deviations (SD), whereas non-normally distributed variables were reported as medians with interquartile ranges (IQR). Categorical variables were summarized as frequencies and percentages. To evaluate the clinical efficacy of the mNGS-guided targeted anti-tuberculosis therapy, paired Student’s t-tests were utilized to compare inflammatory marker levels (ESR and CRP) at baseline prior to treatment and at the final follow-up. All statistical analyses were conducted using IBM SPSS Statistics, version 27.0 (IBM Corp., Armonk, NY, USA). P < 0.05 was considered to indicate statistical significance. Results Demographic Characteristics Based on the inclusion and exclusion criteria, 15 patients were initially assessed for eligibility; however, two were excluded due to insufficient laboratory data, yielding a final cohort of 13 patients. A flow diagram was designed and is presented in Figure 1. The cohort comprised five PJI, one FRI, and seven native joint infections that initially presented as suspected septic arthritis but were ultimately confirmed as tuberculous arthritis via mNGS (Table 1). The study population included eight females and five males, with a mean age of 65.3 ± 15.7 years (range, 26–82 years). Anatomical sites of infection involved the hip (n=6), knee (n=5), elbow (n=1), and sacroiliac joint (n=1). None of the patients reported a prior history of tuberculosis. The median diagnostic delay from symptom onset to mNGS-confirmed Mycobacterium tuberculosis infection was 9 months (range, 1.5–18 months).Table 1Demographic Information of the 13 Patients with Tuberculous Osteoarticular InfectionsPatient IDSexAgeTypeLocationClinical SymptomsHistory of TuberculosisDuration of Diagnostic DelayPreoperative ESR(mm/h)PreoperativeCRP(mg/L)P1F69PJIKneePain, swelling, pusNo12m11365.19P2M75PJIHipPainNo2m4313.5P3F82PJIHipPain, swelling, pusNo1m425.2P4M72TAKneePain, pusNo36m120109.29P5F55TAKneePain, pusNo36m3984P6F73PJIHipPainNo12m12050.87P7F77PJIKneePain, swellingNo1m3585.64P8F81FRIHipPainNo2m9823.45P9M50TAHipPainNo24m177.3P10M74TAKneePainNo9m4455.8P11M56TAElbowPainNo1m8356P12F26TASacroiliac jointPainNo3m6526.6P13F59TAHipPainNo12m2316Abbreviations: F, female; PJI, prosthetic joint infection; m, month; M, male; TA, tuberculous arthritis; FRI, fracture-related infection. Figure 1The flow diagram of enrollment of patients.A flowchart of a retrospective review of 15 cases with tuberculous osteoarticular infections. Table 1Demographic Information of the 13 Patients with Tuberculous Osteoarticular InfectionsPatient IDSexAgeTypeLocationClinical SymptomsHistory of TuberculosisDuration of Diagnostic DelayPreoperative ESR(mm/h)PreoperativeCRP(mg/L)P1F69PJIKneePain, swelling, pusNo12m11365.19P2M75PJIHipPainNo2m4313.5P3F82PJIHipPain, swelling, pusNo1m425.2P4M72TAKneePain, pusNo36m120109.29P5F55TAKneePain, pusNo36m3984P6F73PJIHipPainNo12m12050.87P7F77PJIKneePain, swellingNo1m3585.64P8F81FRIHipPainNo2m9823.45P9M50TAHipPainNo24m177.3P10M74TAKneePainNo9m4455.8P11M56TAElbowPainNo1m8356P12F26TASacroiliac jointPainNo3m6526.6P13F59TAHipPainNo12m2316Abbreviations: F, female; PJI, prosthetic joint infection; m, month; M, male; TA, tuberculous arthritis; FRI, fracture-related infection. Demographic Information of the 13 Patients with Tuberculous Osteoarticular Infections Abbreviations: F, female; PJI, prosthetic joint infection; m, month; M, male; TA, tuberculous arthritis; FRI, fracture-related infection. The flow diagram of enrollment of patients. Comparison of mNGS and Microbial Culture results and Confirmatory Validation Conventional microbial cultures were negative in eight patients. In the remaining five patients (Patients 1–5), preoperative or intraoperative cultures isolated non-mycobacterial flora—specifically Enterococcus faecalis, Staphylococcus epidermidis, Staphylococcus hominis, Trachococcus konrei, and Proteus mirabilis, respectively—which initially prompted targeted antibacterial therapy (Table 2). Conversely, mNGS identified the MTBC in all 13 patients, with a median sequence read count of 296 (range, 52.5–1,321.5). Within the five culture-confounded specimens, mNGS concurrently detected the respective isolated bacteria in Patients 1, 3, and 4, while additionally identifying Streptococcus agalactiae in Patient 4. Synovial or granulation tissue exhibiting pronounced inflammatory changes was resected intraoperatively and submitted for histopathological examination. Chronic granulomas were present in the histological sections of all 13 patients (100%), with typical multinucleated giant cells observed in five patients (38.5%). The T-SPOT.TB assay was positive in nine patients (69.2%) and negative in four patients (30.8%). Seven patients initially presented with a severe inflammatory joint phenotype, leading to a preliminary diagnosis of septic arthritis; routine bacterial cultures were negative in five of these cases. Subsequent mNGS analysis detected MTBC across all seven cases, establishing the definitive diagnosis of tuberculous arthritis.Table 2T-SPOT.TB, Microbial Culture, mNGS and Histopathological Examination Results of 13 PatientsPatient IDMicrobial CulturemNGSReadsHistopathological ExaminationT-SPOT.TBP1E. faecalisMTBC+E. faecalis1Chronic granulomatous inflammation+MGCPositiveP2S. hominisMTBC144Chronic granulomatous inflammationPositiveP3S. epidermidisMTBC+S. epidermidis57Chronic granulomatous inflammationPositiveP4H.kunziiH.kunzii+S. dysgalactiae+MTBC4169Chronic granulomatous inflammation+MGCPositiveP5P. mirabilisMTBC296Chronic granulomatous inflammationPositiveP6NegativeMTBC48Chronic granulomatous inflammationPositiveP7NegativeMTBC1627Chronic granulomatous inflammation+MGCNegativeP8NegativeMTBC2598Chronic granulomatous inflammationNegativeP9NegativeMTBC66Chronic granulomatous inflammationNegativeP10NegativeMTBC5Chronic granulomatous inflammation+MGCPositiveP11NegativeMTBC316Chronic granulomatous inflammationNegativeP12NegativeMTBC644Chronic granulomatous inflammationPositiveP13NegativeMTBC1016Chronic granulomatous inflammation+MGCPositiveAbbreviations: mNGS, metagenomic next-generation sequencing; TB, tuberculosis; E. faecalis, Enterococcus faecalis; MTBC, Mycobacterium tuberculosis complex; MGC, multinucleated giant cell; S. hominis, Staphylococcus hominis; S. epidermidis, Staphylococcus epidermidis; H.kunzii, Helcococcus kunzii; S. dysgalactiae, Streptococcus dysgalactiae; P. mirabilis, Proteus mirabilis. Table 2T-SPOT.TB, Microbial Culture, mNGS and Histopathological Examination Results of 13 PatientsPatient IDMicrobial CulturemNGSReadsHistopathological ExaminationT-SPOT.TBP1E. faecalisMTBC+E. faecalis1Chronic granulomatous inflammation+MGCPositiveP2S. hominisMTBC144Chronic granulomatous inflammationPositiveP3S. epidermidisMTBC+S. epidermidis57Chronic granulomatous inflammationPositiveP4H.kunziiH.kunzii+S. dysgalactiae+MTBC4169Chronic granulomatous inflammation+MGCPositiveP5P. mirabilisMTBC296Chronic granulomatous inflammationPositiveP6NegativeMTBC48Chronic granulomatous inflammationPositiveP7NegativeMTBC1627Chronic granulomatous inflammation+MGCNegativeP8NegativeMTBC2598Chronic granulomatous inflammationNegativeP9NegativeMTBC66Chronic granulomatous inflammationNegativeP10NegativeMTBC5Chronic granulomatous inflammation+MGCPositiveP11NegativeMTBC316Chronic granulomatous inflammationNegativeP12NegativeMTBC644Chronic granulomatous inflammationPositiveP13NegativeMTBC1016Chronic granulomatous inflammation+MGCPositiveAbbreviations: mNGS, metagenomic next-generation sequencing; TB, tuberculosis; E. faecalis, Enterococcus faecalis; MTBC, Mycobacterium tuberculosis complex; MGC, multinucleated giant cell; S. hominis, Staphylococcus hominis; S. epidermidis, Staphylococcus epidermidis; H.kunzii, Helcococcus kunzii; S. dysgalactiae, Streptococcus dysgalactiae; P. mirabilis, Proteus mirabilis. T-SPOT.TB, Microbial Culture, mNGS and Histopathological Examination Results of 13 Patients Abbreviations: mNGS, metagenomic next-generation sequencing; TB, tuberculosis; E. faecalis, Enterococcus faecalis; MTBC, Mycobacterium tuberculosis complex; MGC, multinucleated giant cell; S. hominis, Staphylococcus hominis; S. epidermidis, Staphylococcus epidermidis; H.kunzii, Helcococcus kunzii; S. dysgalactiae, Streptococcus dysgalactiae; P. mirabilis, Proteus mirabilis. Changes in Treatment Strategies, Surgical Interventions, and Patient Outcomes For the eight strictly culture-negative patients, empirical antimicrobial therapy comprising vancomycin combined with third-generation cephalosporins was administered.24 For the five culture-confounded patients, targeted antibacterial therapy was initiated based on the isolated organisms. All 13 patients underwent surgical intervention, which included one-stage revision, 1.5-stage revision, two-stage revision, total knee arthroplasty (TKA), total hip arthroplasty (THA), debridement with implant removal and antibiotic spacer implantation, or debridement with lesion resection (Table 3). Following the receipt of intraoperative mNGS reports confirming MTBC, therapeutic regimens were adjusted in 100% of the cohort. Standardized anti-tuberculosis therapy was immediately initiated. The majority of patients commenced an initial intensive phase with standard weight-based dosages of quadruple therapy (isoniazid, rifampin, pyrazinamide, and ethambutol [H-R-Z-E]), subsequently transitioning to a continuation phase. Although the precise duration of the initial quadruple phase before stepping down varied or was not uniformly documented due to the retrospective nature of this study, all 13 patients strictly received targeted anti-tuberculosis therapy for a confirmed minimum total duration of 12 months post-discharge, adhering to established guidelines.7 The primary treatment success rate was 92.3% (12/13), as 12 patients required no further surgical interventions after the index procedure during a mean follow-up of 17.5 ± 3.0 months. One patient experienced infection recurrence one month postoperatively, necessitating a subsequent two-stage revision. At the final follow-up, all 13 patients demonstrated an absence of clinical infection signs and restored joint function, yielding an ultimate infection eradication rate of 100%. Paired t-tests indicated a significant reduction in both ESR and CRP levels at the final follow-up compared with preoperative baselines (P < 0.05). Surgical procedures utilized for specific infection types included: one-stage revision (n=2, PJI), two-stage revision (n=2, PJI), 1.5-stage revision (n=1, PJI), debridement with implant removal and spacer implantation (n=1, FRI), debridement with lesion resection (n=2, elbow and sacroiliac tuberculous arthritis), and primary TKA or THA (n=5, knee or hip tuberculous arthritis). A representative case is illustrated in Figure 2. Patient 10 presented with suspected septic arthritis of the knee. Preoperative microbial cultures were negative, prompting empirical vancomycin and cefotaxime. Intraoperatively, extensive inflammatory granulation tissue and caseous necrosis were observed. Following extensive debridement and lesion excision, a primary TKA was performed. Tissues were submitted for histopathology and mNGS. Histological analysis confirmed chronic granulomatous inflammation with typical multinucleated giant cells. Concurrently, mNGS detected MTBC with a sequence read count of 5 (Table 4). The H-R-Z-E regimen was immediately commenced. At the final follow-up, inflammatory markers were normalized, and radiographs demonstrated stable prosthesis positioning without evidence of loosening or osteolysis, indicating complete infection eradication (Figure 2).Table 3Clinical Characteristics, mNGS-Guided Antimicrobial Therapy, Surgical Interventions, and Outcomes of 13 Patients with Tuberculous Osteoarticular InfectionsPatient IDPrevious Antibiotic Treatment (Before mNGS)Antibiotic Treatment (after mNGS)TreatmentDuration of Follow-UpESR(mm/h), Last Follow-UpCRP(mg/L),last Follow-UpOutcomeP1V+CAZIREP+V+PT1.5-stage revision19m53.64CureP2V+CAZIREP+L+LZD2-stage revision12m146.2CureP3VREP+L+V1-stage revision20m20.9CureP4V+L+LZDIREP+V+LTKA22m175.8CureP5V+CAZIREP+LTKA18m182.3CureP6V+CROIRE+L1-stage revision15m205.2CureP7V+CAZIREP+L+LZD2-stage revision18m164.7Cure*P8V+LIREP+MDIA15m153.9CureP9V+CAZIREP+V+CAZTHA19m25CureP10V+CAZIREPTKA21m72.7CureP11V+CLRIREP+LDebridement+lesion resection13m122.7CureP12TECIREP+LDebridement+lesion resection17m211.1CureP13V+CAZIREP+LTHA18m135.5CureNotes: Cure*:The patient developed recurrence 1 month after 1-stage revision and subsequently underwent 2-stage revision. Infection was ultimately controlled at the final follow-up.Abbreviations: mNGS, metagenomic next-generation sequencing; V, vancomycin; CAZ, ceftazidime; I, isoniazid; R, rifampin; E, ethambutol; P, pyrazinamide; PT, piperacillin sodium and tazobactam sodium; m, month; L, levofloxacin; LZD, linezolid; TKA, total knee arthroplasty; CRO, ceftriaxone sodium; M, meropenem; DIA, debridement, implant removal, antibiotic spacer implantation; THA, total hip arthroplasty; CLR, clarithromycin; TEC, teicoplanin. Table 4mNGS Detection Results of Patient 10SpeciesReadsMycobacterium tuberculosis complex5Abbreviation: mNGS, metagenomic next-generation sequencing. Figure 2The clinical data of patient 10. (A) Preoperative radiographs. (B) Postoperative radiographs at the final follow-up. (C) Intraoperative pathological test.Composite: 2 knee X-rays, 1 tissue micrograph with purple cells on pink. Table 3Clinical Characteristics, mNGS-Guided Antimicrobial Therapy, Surgical Interventions, and Outcomes of 13 Patients with Tuberculous Osteoarticular InfectionsPatient IDPrevious Antibiotic Treatment (Before mNGS)Antibiotic Treatment (after mNGS)TreatmentDuration of Follow-UpESR(mm/h), Last Follow-UpCRP(mg/L),last Follow-UpOutcomeP1V+CAZIREP+V+PT1.5-stage revision19m53.64CureP2V+CAZIREP+L+LZD2-stage revision12m146.2CureP3VREP+L+V1-stage revision20m20.9CureP4V+L+LZDIREP+V+LTKA22m175.8CureP5V+CAZIREP+LTKA18m182.3CureP6V+CROIRE+L1-stage revision15m205.2CureP7V+CAZIREP+L+LZD2-stage revision18m164.7Cure*P8V+LIREP+MDIA15m153.9CureP9V+CAZIREP+V+CAZTHA19m25CureP10V+CAZIREPTKA21m72.7CureP11V+CLRIREP+LDebridement+lesion resection13m122.7CureP12TECIREP+LDebridement+lesion resection17m211.1CureP13V+CAZIREP+LTHA18m135.5CureNotes: Cure*:The patient developed recurrence 1 month after 1-stage revision and subsequently underwent 2-stage revision. Infection was ultimately controlled at the final follow-up.Abbreviations: mNGS, metagenomic next-generation sequencing; V, vancomycin; CAZ, ceftazidime; I, isoniazid; R, rifampin; E, ethambutol; P, pyrazinamide; PT, piperacillin sodium and tazobactam sodium; m, month; L, levofloxacin; LZD, linezolid; TKA, total knee arthroplasty; CRO, ceftriaxone sodium; M, meropenem; DIA, debridement, implant removal, antibiotic spacer implantation; THA, total hip arthroplasty; CLR, clarithromycin; TEC, teicoplanin. Clinical Characteristics, mNGS-Guided Antimicrobial Therapy, Surgical Interventions, and Outcomes of 13 Patients with Tuberculous Osteoarticular Infections Notes: Cure*:The patient developed recurrence 1 month after 1-stage revision and subsequently underwent 2-stage revision. Infection was ultimately controlled at the final follow-up. Abbreviations: mNGS, metagenomic next-generation sequencing; V, vancomycin; CAZ, ceftazidime; I, isoniazid; R, rifampin; E, ethambutol; P, pyrazinamide; PT, piperacillin sodium and tazobactam sodium; m, month; L, levofloxacin; LZD, linezolid; TKA, total knee arthroplasty; CRO, ceftriaxone sodium; M, meropenem; DIA, debridement, implant removal, antibiotic spacer implantation; THA, total hip arthroplasty; CLR, clarithromycin; TEC, teicoplanin. Table 4mNGS Detection Results of Patient 10SpeciesReadsMycobacterium tuberculosis complex5Abbreviation: mNGS, metagenomic next-generation sequencing. mNGS Detection Results of Patient 10 Abbreviation: mNGS, metagenomic next-generation sequencing. The clinical data of patient 10. (A) Preoperative radiographs. (B) Postoperative radiographs at the final follow-up. (C) Intraoperative pathological test. Discussion Main Findings of This Study OAI caused by Mycobacterium tuberculosis are rare, and the absence of specific diagnostic and therapeutic guidelines poses significant clinical challenges. To the best of our knowledge, this represents one of the few studies specifically investigating the diagnostic utility and therapeutic rescue role of mNGS in occult TB-OAI confounded by strictly negative or dangerously misleading conventional cultures. Our principal findings demonstrated that conventional microbial cultures uniformly failed to isolate M. tuberculosis across the cohort. Furthermore, a substantial proportion of these cultures yielded common non-mycobacterial pathogens, creating profound diagnostic illusions. In stark contrast, mNGS successfully identified the MTBC in all cases, with these genomic findings consistently corroborated by histopathology and T-SPOT.TB assays. Ultimately, the integration of mNGS results prompted a definitive therapeutic pivot in antimicrobial stewardship for every patient. The timely identification of the pathogen via mNGS facilitated this appropriate management, which directly contributed to successful and durable clinical outcomes, as evidenced by the complete resolution of clinical symptoms, the normalization of serum inflammatory markers, and the absence of progressive radiographic abnormalities. The Illusion of Routine Cultures and Diagnostic Dilemmas Because OAI caused by Mycobacterium tuberculosis are intrinsically paucibacillary and frequently present with atypical clinical symptoms, their diagnosis remains a formidable clinical challenge. In our cohort, conventional microbial cultures were strictly negative for a substantial subset of patients. A plausible explanation for this could be the prior administration of empirical antibiotic therapy. Furthermore, the exceptionally slow doubling time of M. tuberculosis dictates a very low bacterial load within clinical specimens, fundamentally hindering isolation via routine culture. Additionally, previous literature indicates that M. tuberculosis can form complex biofilms, which may further compromise the yield of conventional microbial techniques.9,25 More importantly, we encountered a critical diagnostic illusion: conventional cultures isolated common, fast-growing bacteria—such as Enterococcus faecalis and Staphylococcus species—in the remaining cases. We hypothesize that these organisms might represent secondary colonizers establishing themselves within sinus tracts, or alternatively, superficial contamination during sample acquisition. However, existing reports also suggest that M. tuberculosis can occasionally coexist with faster-growing microorganisms in true polymicrobial infections.26 Indeed, mNGS analysis in our study concurrently detected these identical non-mycobacterial species alongside the MTBC in several of these culture-confounded cases. This finding suggests a genuine coexistence of M. tuberculosis with these organisms in certain clinical scenarios, further obscuring the primary pathogen. Clinically, the ramifications of strict culture-negativity or misattribution to common bacterial flora are profound. Such scenarios frequently lead to premature diagnostic closure, wherein TB-OAI is erroneously managed as a conventional pyogenic arthritis, FRI, or PJI. Upon this misclassification, clinicians typically initiate empirical broad-spectrum antibiotic regimens; however, these agents afford no therapeutic coverage against MTBC. Concurrently, patients often undergo repeated surgical debridements based on this flawed premise. Because these interventions do not specifically target tuberculous pathology—such as the radical excision of caseous necrotic tissue—and are unaccompanied by an essential, standardized course of systemic anti-tuberculosis pharmacotherapy, they are inherently predisposed to failure. Ultimately, the unabated progression of the infection drives extensive osteoarticular destruction, pathological fractures, and severe structural joint damage. Left unchecked, this trajectory may culminate in the systemic dissemination of M. tuberculosis, precipitating catastrophic patient outcomes.27,28 High Sensitivity of mNGS in Diagnosing Tuberculosis, Adjustments to Treatment Regimens, and Clinical Outcomes Unlike traditional microbial culture, mNGS offers the intrinsic advantage of unbiased, broad-spectrum pathogen detection. For fastidious organisms with demanding growth requirements, such as Mycobacterium tuberculosis and Mycoplasma species, mNGS directly captures nucleic acid fragments for bioinformatics alignment, thereby significantly enhancing detection sensitivity. Furthermore, because mNGS is a nucleic acid-based assay, it maintains a robust diagnostic positivity rate even following extensive antibiotic exposure. By leveraging automated sequencing platforms and advanced bioinformatics, mNGS typically generates comprehensive pathogen reports within 24 to 48 hours, dramatically reducing the turnaround time compared with the weeks often required for mycobacterial culture.29 In the present study, the sequence read counts for MTBC were notably low. This finding aligns consistently with previous literature and can likely be attributed to several interrelated biological and technical factors. First, M. tuberculosis is a facultative intracellular pathogen that predominantly resides within host macrophages; within pathological tissues such as granulomas, it typically exists in a dormant or low-replication state, resulting in a fundamentally low absolute abundance of mycobacterial nucleic acids in clinical samples.30,31 Second, M. tuberculosis possesses a highly unique, lipid-rich cell wall heavily fortified with mycolic acids and waxes. Conventional nucleic acid extraction protocols often lack the stringency required to efficiently lyse this robust envelope, leading to suboptimal DNA release.32,33 Third, in the vast majority of clinical tissue samples, host genomic DNA exponentially exceeds microbial DNA. This overwhelming host background consumes a significant proportion of the sequencing depth, resulting in the severe dilution of pathogen-specific reads and further complicating the detection of the already sparse MTBC signal.34,35 Despite these characteristically low read counts, the detection of MTBC holds profound clinical significance. Because M. tuberculosis is not a common environmental contaminant and does not constitute normal human flora, even a minimal stringently mapped read number is highly indicative of true infection in the appropriate clinical context. To rigorously validate our mNGS findings and exclude any theoretical possibility of laboratory contamination, we systematically cross-referenced the genomic data with intraoperative histopathology. In all cases, histological examination revealed chronic granulomatous inflammation, frequently accompanied by the presence of typical multinucleated giant cells. These histological features strongly supported the mNGS-based diagnosis of TB-OAI in our clinical context. However, we acknowledge that granulomatous inflammation is not completely specific to tuberculosis and can present in other osteoarticular diseases, further highlighting the necessity of molecular confirmation via mNGS. The ultimate clinical utility of any novel diagnostic modality resides in its capacity to direct clinical decision-making. Prior to the integration of mNGS, the median diagnostic delay within our cohort was nine months. During this protracted interval, patients uniformly received empirical broad-spectrum antibiotic therapy targeting common pyogenic flora, and several underwent recurrent, ineffective surgical debridements. The deployment of mNGS effectively terminated this diagnostic odyssey. By unmasking the MTBC, mNGS catalyzed a fundamental therapeutic pivot in all cases. While the majority of the cohort was transitioned to the standard four-drug anti-tuberculosis regimen (H-R-Z-E), specific cases necessitated tailored triple therapy or the integration of adjunctive antimicrobial agents (eg, levofloxacin, vancomycin, or linezolid). Such cross-class therapeutic adjustments are paramount. Empirical broad-spectrum antibiotics lack efficacy against M. tuberculosis. Furthermore, we hypothesize that in polymicrobial scenarios where M. tuberculosis coexists with common bacteria, conventional antibiotics selectively suppress the fast-growing pyogenic flora. This selective pressure inadvertently conceals the underlying mycobacterial infection, allowing it to silently progress into protracted, treatment-refractory chronic osteoarticular lesions. The significant postoperative reduction in inflammatory markers (ESR and CRP) not only validated the efficacy of the MTBC-targeted therapy but also retrospectively corroborated MTBC as the primary causative pathogen. Ultimately, the synergistic application of targeted pharmacotherapy and radical surgical intervention yielded an initial treatment success rate of 92.3% (12/13), precluding the need for further surgeries in the vast majority of patients. Although one patient required a subsequent two-stage revision due to early recurrence, durable infection eradication was achieved by the final follow-up. This isolated case of early relapse underscores a critical clinical reality: TB-OAI is exceptionally refractory. It demonstrates that while rapid, precise pathogen identification is the cornerstone of management, the mechanical eradication of M. tuberculosis embedded within extensive osseous and soft-tissue necrosis remains a formidable surgical challenge. Limitations and Future Prospects Several limitations of the present study warrant consideration. First, the retrospective, single-center design and the small sample size constrain the statistical power. The inclusion of varying infection types (periprosthetic, fracture-related, and native joint infections) introduces clinical heterogeneity, which limits broad statistical inference. Furthermore, utilizing mNGS positivity within the composite reference standard introduces selection bias, naturally overrepresenting diagnostically complex cases where mNGS successfully identified the pathogen. Nevertheless, given the epidemiological rarity of TB-OAI, this descriptive series provides hypothesis-generating clinical evidence. Second, mycobacterium-specific diagnostic modalities, such as the Xpert MTB/RIF assay or specialized liquid culture media, were not routinely employed within our initial institutional workflow. Consequently, a direct comparative analysis between the diagnostic efficacy of mNGS and these targeted molecular assays could not be performed. Future prospective, multicenter investigations incorporating rigorous cost-effectiveness analyses are warranted to establish standardized clinical protocols for the integration of mNGS into the routine diagnostic algorithm for occult osteoarticular infections. Conclusion In summary, within this selected cohort, mNGS demonstrated high diagnostic yield and served as a promising diagnostic rescue tool for occult TB-OAI, successfully unmasking infections completely obscured by false-negative or confounding routine broad-spectrum conventional cultures. By decisively piercing these diagnostic illusions, mNGS enabled rapid, pathogen-directed therapeutic pivots. This timely redirection of treatment subsequently facilitated favorable clinical outcomes, thereby preventing catastrophic diagnostic delays and the profound morbidity of misdirected surgical interventions. Moving forward, integrating this advanced molecular modality into standardized diagnostic algorithms will be imperative to optimize antimicrobial stewardship and fundamentally improve clinical outcomes for complex, culture-refractory osteoarticular infections.