The diagnostic potential of combined quantitative polymerase chain reaction and next-generation sequencing using the same primers for periprosthetic joint infection Authors: Narumi Ueda, Jun Inoue, Kazuyuki Okuda, Hirokazu Iida Published: 2026 ABSTRACT ABSTRACTNext-generation sequencing (NGS) enables the detection of specific pathogens unidentifiable by conventional cultures, but its application in orthopedics remains inconsistent due to background contamination and irreproducible findings. This study evaluated the diagnostic performance of a novel workflow combining broad-range 16S rRNA gene quantitative PCR (qPCR) screening with downstream NGS, focusing on bacterial biomass thresholds. The qPCR assay demonstrated excellent intrarater reliability, with an intraclass correlation coefficient (ICC) of 0.961 (95% confidence interval, 0.881 to 0.997). Based on serially diluted positive controls, a quantitative threshold of 10⁵ CFU/mL was established as the minimum concentration required for the consistent detection of fastidious taxa, such as Escherichia coli. When evaluated against conventional cultures using 95 sonicate fluid and 276 pre/intraoperative tissue samples, the qPCR assay achieved a sensitivity of 80% and a specificity of 72%. Subsequent NGS sequencing of 26 clinical samples and 9 controls showed concordance in 4 of 6 culture-positive infected cases with NGS taxonomy, whereas the remaining discrepancies were likely attributable to culture-based phenotypic misidentification. Notably, among the qPCR-positive cases, three were culture-negative, including two hip prosthesis loosening cases exhibiting polymicrobial profiles, and one post-traumatic osteoarthritis case harboring low-level Staphylococcus. Crucially, this post-traumatic patient developed delayed periprosthetic joint infection (PJI) 2 years post-surgery, with cultures identifying Staphylococcus previously detected by the initial NGS analysis. Integrating qPCR screening with targeted NGS effectively refines pathogen identification, filters environmental artifacts, and overcomes the diagnostic limitations of culture-negative infections in orthopedic practice.IMPORTANCENext-generation sequencing (NGS) enables the detection of specific pathogens in clinical samples that are not identifiable by conventional methods. However, NGS applications in orthopedics have not been quantitatively evaluated, and findings have been inconsistent owing to contaminants and the presence of non-credible causative organisms. These factors primarily stem from the failure to evaluate low-biomass samples and the absence of proper controls, such as negative controls or mock community DNA samples. This study demonstrates that interpreting results from low-biomass samples requires careful consideration because NGS relies on relative bacterial abundances; distinguishing likely pathogens from contaminants is particularly challenging when bacterial loads are low. We demonstrated that combining NGS with quantitative PCR (qPCR) and applying a Cq cutoff can reduce false positives. INTRODUCTION Routine diagnostic testing may miss a significant number of patients with asymptomatic infections, and these cases are often misdiagnosed as aseptic loosening (1, 2). In clinical practice, when an artificial joint loosens without the typical symptoms of a periprosthetic joint infection (PJI) and a definitive diagnosis cannot be achieved through conventional culture tests, molecular biological techniques, such as next-generation sequencing (NGS), are used to diagnose the infection (3–17). Historically, various testing methods have been explored to detect subclinical pathogens that cannot be identified through traditional culture tests (9–13). To improve the sensitivity of culture testing using sonication and address culture-negative PJIs (12, 13, 18), contemporary molecular frameworks have increasingly incorporated multiple genetic assays (12–14, 17). Although the 2018 Musculoskeletal Infection Society (MSIS) guidelines (19) mention NGS, they lack detailed, quantitative protocols, and even when NGS is utilized, it can be challenging to determine whether the results indicate a true infection. A fundamental challenge in relying on qualitative NGS methodologies, particularly when applied to clinical samples with low microbial biomass, is that they may contain bacteria below the detection limit, or no bacteria at all. Particularly, false positives have been frequently reported with NGS when bacterial loads are low in orthopedic NGS articles (4–7). Consequently, NGS methodologies in orthopedics require more quantitative evaluation to overcome inconsistent results caused by potential contaminants and the presence of non-pathogenic or non-credible causative organisms. Because reagent and laboratory contamination can critically affect sequence-based microbiome analyses (9–11), previous reports (4–7) on the use of NGS methods to diagnose PJI may have been affected by contamination-related inconsistencies. This technical limitation highlights an important methodological gap in current molecular diagnostics: the absence of quantitative assessments and documentation, and a failure to incorporate positive benchmarks. To address these critical diagnostic limitations in low-biomass environments, the primary objective of this study was to establish and validate a pragmatic, highly reproducible molecular diagnostic workflow for PJI by utilizing identical universal primers for both broad-range 16S rRNA gene quantitative PCR (qPCR) screening and downstream NGS. Specifically, we aimed to (i) determine a statistically optimized, parameter-driven qPCR Cq threshold capable of identifying the precise tipping point where biological signals become dominated by environmental contamination; (ii) evaluate the diagnostic performance of this integrated quantitative pipeline against conventional microbiological cultures in a clinical cohort; and (iii) demonstrate the clinical utility of this molecular framework in uncovering occult subclinical pathogens and preventing the underdiagnosis or misclassification of low-biomass implant infections. MATERIALS AND METHODS Study design and setting Laboratory evaluation of qPCR To evaluate the reliability of our qPCR protocol, we used Escherichia coli (ATCC 25922) as a quantitative positive control. In a preliminary validation, E. coli colonies were cultured in 0.9% sterile saline until the turbidity reached the 0.5 McFarland standard. This strain was selected due to its established reproducibility in quality control standardizations (12–14). Bacterial suspensions were adjusted to final concentrations ranging from 10⁴ to 10⁷ CFU/mL. A total of 80 samples (20 replicates for each of the four concentrations) were prepared to validate the qPCR protocol (see Fig. 1). We assessed the consistency of the quantification cycle (Cq) values and the overall reproducibility of the assays using these samples. Research method; PCR. Clinical evaluation of qPCR A total of 95 sonicated fluid samples were obtained from 56 consecutive patients (21 males and 35 females) who underwent implant removal or debridement at our institution between May 2015 and March 2017. These patients were enrolled in a prospective cohort study (see Table 1). Inclusion criteria targeted cases requiring implant removal for which culture testing was performed, and for which the residual sonication (18) fluid was properly preserved in the clinical laboratory. Samples for qPCR were taken from this residual fluid and stored in the microbiology laboratory. First, we determined the optimal Cq thresholds for identifying orthopedic implant-associated infection (OIAI) using these samples. Then, we evaluated the diagnostic performance of qPCR compared with conventional culture results. TABLE 1Clinical characteristics of the 56 patients enrolled in the qPCR studyaCharacteristicResultOIAINon-infectionP valuecNo. of patients (no. of cases)20 (20)36 (36)No. of sonicate fluid samples (n = 95)3758Median patient age, yrs (IQR)67 (57–80)67 (57–77)0.632dGender, male/female11/910/260.044fMedian follow-up, months (IQR)37 (33–40)41 (31–48)0.051dThe median time from last operation, yrs (IQR)4.7 (1.2–10.3)11.5 (2.9–19)0.163dNo. (%) of complications of infection/re-infection until final follow-up01bJoint (%) Hip15 (75)32 (89)0.164e Knee2 (10)2 (5.5)0.451e Others (elbow 1, ankle 2, spine 1, shoulder 1)3 (15)2 (5.5)0.238eType of surgery Debridement with implant retention or removal1300.000e Removed implant without debridement030.258e One-stage exchange7270.003f Primary total hip arthroplasty060.06eThe previous surgery Salvage arthroplasty240.593e Primary arthroplasty (hip, knee, elbow)10200.690c Bipolar hemi-arthroplasty220.428e Osteotomy/implant fixation (included spinal instrumentation)6100.86fSonicate-fluid sample (n = 95) Artificial joint materials23400.494f Osteosynthesis device11160.821f Cement200.149e Bone/soft tissue120.665ePresence of sinus tract, no. of cases300.041ePeri-implant purulence110<0.001eAntibiotic recipients before sampling, n (%)10 (50)0<0.001e a OIAI, orthopedic implant-associated infection; PJI, periprosthetic joint infection; IQR, interquartile range; THA, total hip arthroplasty. b One patient in the non-infection group experienced a complication categorized as infection/reinfection during the final follow-up period. c P values were calculated to compare the OIAI/PJI group and the non-infection group. d Mann-Whitney U test. e Fisher's exact test. f Pearson’s chi-squared test. Clinical characteristics of the 56 patients enrolled in the qPCR studya OIAI, orthopedic implant-associated infection; PJI, periprosthetic joint infection; IQR, interquartile range; THA, total hip arthroplasty. One patient in the non-infection group experienced a complication categorized as infection/reinfection during the final follow-up period. P values were calculated to compare the OIAI/PJI group and the non-infection group. Mann-Whitney U test. Fisher's exact test. Pearson’s chi-squared test. Clinical evaluation of NGS Of the samples used for qPCR evaluation, 26 sonicate fluid samples were obtained from 15 consecutive patients (6 males, 9 females) who underwent implant removal or debridement at our institution between May 2015 and October 2015. These patients were enrolled in this prospective cohort NGS study (see Fig. 2). The cohort comprised 13 hip and 2 knee cases, which included 4 infected cases (3 PJIs and 1 fracture-related infection (FRI) (20)) and 11 non-infected cases (see Table 2). All included cases required implant removal, underwent sonicated fluid culture (SFC) testing, and had residual sonication fluid (18) properly preserved in the clinical laboratory. Research method; next-generation sequencing. TABLE 2Clinical characteristics of the 15 patients enrolled in the NGS studyaCase no.AgeSexCRP (mg/L)Observation (months)Reason for surgeryJointReceived antibioticsAspiration joint fluid (preoperative) culture164F27695PJIHip+Streptococcus sp.273F353LooseningHip−–355M9240FRIHip−CNS483M1172PJIHip−–558F199LooseningHip−ND670M3099PJIKnee+CNS*750M576OAHip−–848M2102DislocationHip−ND946M199OAHip−–1082F315OAHip−–1161F0101LooseningHip−–1273F1101LooseningKnee−ND1369F1106OAHip−ND1482F136OAHip−ND1574F293PainHip−ND a NGS, next-generation sequencing; PJI, prosthetic joint infection; FRI, fracture-related infection; OA, osteoarthritis; CRP, C-reactive protein; ND, not done; +, presence; −, absence. *Coagulase-negative Staphylococcus sp. Clinical characteristics of the 15 patients enrolled in the NGS studya NGS, next-generation sequencing; PJI, prosthetic joint infection; FRI, fracture-related infection; OA, osteoarthritis; CRP, C-reactive protein; ND, not done; +, presence; −, absence. *Coagulase-negative Staphylococcus sp. Samples for qPCR and NGS were aliquoted from the remnant sonication fluid and stored in the microbiology laboratory. In total, 26 clinical samples and nine controls were analyzed in a single sequencing run. The controls consisted of four quantitative positive controls (E. coli, 10⁴–10⁷ CFU/mL), three negative controls (one nuclease-free [NF] water sample for library preparation and two sterile saline samples for sonication), and two mock community DNA (mock) samples (12–14, 21–25) (Fig. 2). The relationship between the qPCR and NGS results was subsequently compared with the culture test results. The primary objective was to analyze clinical specimens with varying bacterial DNA levels to identify which bacteria predominate under low-biomass conditions and which are likely contaminants. The secondary objective was to determine the lowest concentration at which bacteria could be consistently detected as the dominant organism, representing the highest proportion of NGS reads. Based on our previously proposed quantitative framework (14), we utilized the corresponding Cq value as a threshold to define reliable detection, whereby the targeted bacterium is identified as the dominant taxon in the NGS results. Using this definition, we validated this cutoff point and assessed the ability of NGS to distinguish true biological signals from background contamination. Third, clinical samples were divided into low-biomass and high-biomass groups according to this cutoff value. For each group, we analyzed the top ten bacterial taxa identified by NGS and evaluated their correlation with quantitative measurements. Furthermore, we reassessed the top ten bacterial genera after applying bioinformatic decontamination tools (15, 16). Finally, patients underwent long-term clinical follow-up to monitor their clinical courses, particularly for cases where definitive conclusions could not be drawn solely from the acute clinical presentation and NGS results. Pre/postoperative assessment Preoperatively, synovial aspirates were obtained from patients exhibiting clinical evidence of infection, such as elevated serum C-reactive protein levels, persistent or increasing pain, implant loosening, and fever. OIAI, which includes PJI and FRI, was diagnosed according to the MSIS criteria (19) and the FRI descriptive flowchart (20). Intraoperative sample collection During surgery, multiple peri-implant tissue sites were sampled for histological analysis, and implants were removed under sterile conditions. Individual components of the extracted implants and adjacent bone or tissue were separately subjected to sonication (18). Other peri-implant specimens were transferred to the institutional microbiology laboratory for routine culture (18). Additionally, peri-implant tissues were evaluated intraoperatively using frozen section analysis (18, 19). Antibiotic treatment Routine antibiotic prophylaxis comprised 1 g of cefazolin administered intravenously at the induction of anesthesia and at 3-h intervals throughout surgery (18). If a pathogen had been detected prior to surgery, targeted antibiotic treatment was initiated after intraoperative tissue sampling, guided by antimicrobial susceptibility testing results. Otherwise, empirical antibiotic treatment was administered post-sampling. Previous antimicrobial therapy was defined as the administration of antimicrobial agents for at least 1 day within the 14 days preceding implant removal (18). In cases of suspected infection, preoperative antibiotics were withheld until culture samples had been collected. 16S rRNA gene qPCR To facilitate post-sequencing analyses, qPCR was performed using serially diluted, known quantities of spiked input DNA (10, 12–14). The qPCR threshold used for clinical screening differed from the more stringent threshold adopted for downstream NGS interpretability analyses. The V1–V2 region of the 16S rRNA gene was selected to ensure accurate identification of a broad range of clinically relevant bacteria (12, 21, 22, 24, 25). Broad-range universal primers targeting this region were 8F (5′-AGAGTTTGATCCTGGCTCAG-3′) and 338R (5′-TGCTGCCTCCCGTAGGAGT-3′) (12–14, 24). Bacterial DNA was extracted from sonicate fluid samples using the QIAamp DNA Mini Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s protocol. Broad-spectrum qPCR assays were conducted with SYBR Green on a Light Cycler Nano instrument (Roche Diagnostics, Basel, Switzerland) using the FastStart Essential DNA Green Master (Roche Diagnostics, Burgess Hill, UK). The thermocycling conditions comprised an initial denaturation at 95°C for 10 min; 40 cycles of denaturation at 94°C for 1 min, annealing at 58°C for 1 min, and extension at 68°C for 1 min; and a final extension at 95°C for 10 s. Cq values were calculated using the LightCycler Nano Software (Roche Molecular Systems, Inc., Pleasanton, CA, USA). Next-generation sequencing Initially, DNA was amplified via PCR using custom primers targeting the V1–V2 region (12–14, 24), identical to those used for 16S rRNA gene qPCR. To achieve optimal amplicon yields and minimize background artifacts based on negative-control qPCR findings, the thermocycling profile was reduced from 40 to 30 cycles. Amplicon libraries were prepared using the Ion Plus Fragment Library Kit (Thermo Fisher Scientific, Waltham, MA, USA), barcoded with IonXpress Barcode Adapters (Thermo Fisher Scientific), and processed according to the Ion Amplicon Library Preparation (Fusion Method) protocol (26). Purification was performed using Agencourt AMPure XP beads (Beckman Coulter, High Wycombe, UK), and DNA concentration was quantified via capillary electrophoresis using an Agilent 2100 Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA, USA). Sample DNA was subsequently loaded onto Ion Sphere Particles for emulsion PCR using the Ion Chef instrument, the Ion PGM Hi-Q View Chef 400 Kit, and the Ion 318 Chip v2 (Life Technologies, Carlsbad, CA, USA). Sequencing was performed on the Ion Torrent PGM platform (Thermo Fisher Scientific) (4–7, 12–14, 21, 22, 25, 26). Bioinformatic processing of 16S rRNA gene sequencing data Sequence data were quality-filtered and analyzed using QIIME (27) version 1.8.0. For comparative analysis, a total of 10,000 reads were randomly subsampled per specimen. Chimeric sequences were identified and removed using USEARCH version 6.1.544, and operational taxonomic units (OTUs) were clustered at a 97% sequence similarity threshold using UCLUST. Taxonomic assignment was performed against the Greengenes reference database (version 13.8). To minimize non-bacterial off-target signals, a post-classification organellar filtering step was applied. OTUs assigned to chloroplast or mitochondrial lineages—defined a priori by taxonomy strings containing “c__Chloroplast” or “Mitochondria”—were identified from the taxonomy OTU table and removed from the BIOM table prior to secondary analyses. This filtering was applied uniformly to all organellar OTUs and was not restricted to Streptophyta-labeled reads. Following organellar filtering, a prevalence-based contamination analysis was conducted using the R package decontam (15–17) as an exploratory yet reproducible computational assessment of candidate contaminant OTUs. The three negative controls (two sterile saline controls and one NF water control) were defined as negative references for the algorithm. Prevalence thresholds of 0.1, 0.2, 0.3, 0.4, and 0.5 were systematically evaluated. Because the top mock-community OTUs were retained at a threshold of 0.3 but began to be classified as candidate contaminants at thresholds of 0.4 and above, 0.3 was selected as the strictest threshold to preserve the expected mock-community signal (see Fig. S1). OTUs identified as candidate contaminants at this selected threshold were subsequently removed before secondary comparisons. Mock bacterial community To evaluate sequencing accuracy, mock community genomic DNA was amplified and sequenced in parallel with the clinical samples. The following reagents were obtained from BEI Resources: Genomic DNA from Microbial Mock Community B (Even, Low Concentration 1), v5.1L (HM-782D) and Genomic DNA from Microbial Mock Community B (Staggered, Low Concentration 1), v5.1L (HM-783D). These mock samples contain 17 genera (20 species) provided at even and staggered concentrations, respectively (12–14, 21–25). Quantitative positive control Controls generated by serial dilution of DNA of known composition were included to determine the biomass level at which contamination becomes a dominant feature of sequencing results (10, 12–14). NGS microflora data from mock samples previously demonstrated a low relative abundance of Gram-negative bacteria when primers targeting the V1–V2 variable regions of the 16S rRNA gene were utilized (14, 25). Therefore, we selected E. coli (ATCC 25922) as a model organism to evaluate its sequencing-based detectability (12–14). Simulated infection models with varying bacterial concentrations were prepared using E. coli suspended in sterile saline. Briefly, E. coli was cultured in broth medium overnight, diluted with sterile saline, and adjusted based on McFarland standards to yield four distinct concentrations: 10⁴, 105, 106, and 107 CFU/mL (12–14). Negative control Negative controls comprised three samples: two aliquots of sterile saline (serving as background controls for the sonication process and the dilution of the quantitative positive control) and one aliquot of NF water processed identically through the NGS workflow (12–14). These three samples were utilized not only for the descriptive characterization of background taxa but also as designated negative references for the prevalence-based decontam analysis (15, 16). Strategies for the identification of pathogenic and contaminant bacteria For culture-positive specimens obtained from the same patient, a bacterium was classified as the causative pathogen if the bacterial count was ≥10⁵ CFU/mL and the organism was detected in the post-decontamination NGS results. Similarly, if the same bacterium was quantitatively detected across two or more NGS samples, it was defined as the causative pathogen. In cases where only a single sample was qPCR-positive, the bacterium was deemed the causative pathogen if it accounted for 100% of the total sequencing reads. Conversely, it was classified as a contaminant if its relative abundance was equal to or less than the baseline percentage determined prior to running the decontam algorithm (15, 16). All remaining bacterial taxa were categorized as contaminants. Statistical analysis Correlations were evaluated using Spearman’s rank correlation coefficient. Test reliability was assessed using generalizability theory based on intraclass correlation coefficients (ICCs), which were interpreted as follows: <0.40, poor; 0.41 to 0.75, fair to good; and >0.75, excellent (28). For the regression equation, the coefficient of determination was obtained via analysis of variance, and prediction accuracy was evaluated. The clinical sample size for qPCR was determined based on previous reports (6, 18). Assuming a 30% difference in sensitivity between molecular methods and culture, a power of 80%, and a two-sided alpha error of 0.05, a minimum sample size of 55 patients was required. The Shapiro–Wilk test was used to assess data normality. Continuous variables were compared between groups using Student’s t-test or the Mann–Whitney U test, as appropriate. Categorical variables were evaluated using 2 × 2 contingency tables and Pearson’s chi-squared test, or Fisher’s exact test if any expected cell frequency was less than 5. All statistical analyses were performed using IBM SPSS Statistics version 31.0 (IBM Corp., Armonk, NY, USA). A P value < 0.05 was considered statistically significant for all two-sided tests. A microbial test result was classified as a true positive if it was concordant with any of the infection-defined criteria described above (19, 20). Sensitivity and specificity were calculated for culture testing, qPCR, and their combination, using the infection-defined criteria as the reference standard. The optimal clinical Cq threshold was determined by maximizing Youden’s J statistic, which was calculated as sensitivity + specificity − 1. Receiver operating characteristic (ROC) analysis was performed using SPSS. The 95% confidence intervals (CIs) for sensitivity and specificity were generated using the Clinical Calculator 1 (18, 29). The sensitivity and specificity of the different diagnostic methods were compared using McNemar’s test. RESULTS Reproducibility and reliability of qPCR assays The qPCR assays demonstrated excellent reproducibility and reliability, with an ICC [1, 1] for Cq values of 0.961 (95% CI, 0.881–0.997) (see Fig. 3). The mean (range) Cq values at E. coli concentrations of 107, 106, 105, and 104 CFU/mL were 17.2 (15.84–18.3), 20.6 (19.12–21.7), 24.0 (22.42–25.0), and 26.8 (25.16–28.2), respectively (see Fig. 3). qPCR assay validation. Summary of quantification cycle (Cq) values, standard deviations (SD), and intrarater reliability across serial dilutions of Escherichia coli (104–107 CFU/mL; n = 20 replicates per concentration). The lower table presents the ICC [1,1] alongside its 95% CIs (lower and upper bounds) to verify measurement consistency. Clinical-based Cq cutoff value obtained from statistical analysis The baseline characteristics and clinical comparisons between the OIAI and non-OIAI groups (see Table 1)—including serum CRP values, histopathological findings, and microbiological culture results—are summarized in Table 3. Data were analyzed from 175 qPCR derivation samples, consisting of 80 laboratory-derived samples and 95 clinical sonicated fluid samples (which included 30 clinical validation samples). The clinical study evaluated a total of 276 culture results obtained from 56 patients (see Fig. 2). These consisted of 30 preoperative joint aspirate cultures and 246 peri-implant tissue cultures; the latter comprised 95 implant SFCs and 151 conventional cultures. Culture results for patients diagnosed with OIAI are summarized in Tables 3 and 4. Sensitivity and specificity were calculated for culture testing, qPCR, and their combination, using the OIAI-defined criteria as the reference standard. Based on the Youden J statistic, the clinical Cq cutoff was determined to be 26.4 with a sensitivity of 0.773 and specificity of 0.765 (see Fig. 4). However, dilution experiments demonstrated that reliable target-organism dominance in NGS profiles was maintained only at bacterial concentrations corresponding to Cq values ≤ 25.5. Therefore, a conservative threshold of Cq ≤ 25.4 was adopted for downstream NGS interpretability analyses (see Fig. 5). TABLE 3Clinical and microbiological characteristics of the 56 patients enrolled in the qPCR studya,b,d,gCharacteristicResultOIAI (n = 20)Non-infection (n = 36)P valuecSites with tissue evidence of inflammation ≥5 neutrophils/high-power field (%)12/142/28<0.001fLaboratory findings: serum CRP (mg/L) Median (IQR): before operation21.6 (3.0–85.3)1.1 (0.45–3.2)<0.001e Sites with CRP > 10.0 mg/ L (%)14/20 (70)2/36 (6)<0.001hPreoperative joint aspirate cultures (n = 30) Sites with > 1 culture positive (%)10/15 (67)0/12 (0)<0.001h Detection rate (%)10/16 (63)0/13 (0)<0.001hIntraoperative peri-implant tissue cultures (n = 151) Mean (range) of samples taken3.5 (1–9)3 (0–7)0.601eSites with > 1 culture positive (%) All cases12/20 (60)1/36 (3)<0.001f Antibiotic recipients before sampling case5/10 (50)0Detection rate (%) All cases17/79 (22)0/72 (0)<0.001h Antibiotic recipients before sampling case10/31 (32)0Sonicate fluid cultures (n = 95) Median (IQR) of samples taken2.0 (1.5–2.0)2.0 (1.0–2.0)0.107eSites with > 1 culture positive (%) All cases12/20 (60)0/36 (0)<0.001f Antibiotic recipients before sampling case5/10 (50)0Detection rate (%) All cases14/37 (36)0/58 (0)<0.001h Antibiotic recipients before sampling case7/20 (35)0Sonicate fluid PCR (n = 95) Sites with > 1 PCR positive (%) All cases (Cq ≤ 25.4)16/20 (80)10/36 (28)<0.001hAntibiotic recipients before sampling case Positive rate (%)8/10 (80)0 All cases (Cq ≤ 25.4)15/20 (75)0 a OIAI, orthopedic implant-associated infection; CRP, C-reactive protein; IQR, interquartile range. b Where the denominator is shown, data were not available for all study patients. c P values were calculated by comparing definite OIAI groups with non-infection groups. d Patient who received a minimum of 1 day of antibiotic therapy within 14 days prior to removal of the implants. e Mann–Whitney U test. f Fisher's exact test. g Student's t-test. h Pearson’s chi-squared test. Clinical and microbiological characteristics of the 56 patients enrolled in the qPCR studya,b,d,g OIAI, orthopedic implant-associated infection; CRP, C-reactive protein; IQR, interquartile range. Where the denominator is shown, data were not available for all study patients. P values were calculated by comparing definite OIAI groups with non-infection groups. Patient who received a minimum of 1 day of antibiotic therapy within 14 days prior to removal of the implants. Mann–Whitney U test. Fisher's exact test. Student's t-test. Pearson’s chi-squared test. TABLE 4Diagnostic and microbiological results of individual study subjects with orthopedic implant-associated infectionsaNo.AgeSexJointDisease/treatmentCRP (mg/L)Sinus tractPurulenceReceived antibioticsHistopathology results (n = 14)Culture resultsbPCR resultsTerm from last operation (yrs)Follow-up (month)Preoperative (n = 16)Peri-implant tissue (n = 79)Sonicate fluid (n = 37)≤25.4c170MKneePJI/2st.30.0−++NDCNS (1)/1N (0)/2CNS (1)/2P (1)/26337264FHipPJI/2st.66.0−+++Streptococcus sp. (1)/1Streptococcus sp. (1)/1Streptococcus sp. (2)/2P (2)/21.237355MHipOIAI/debridement92.0−+−+CNS (1)/1CNS (1)/5CNS (1)/2P (1)/20.540483MHipOIAI/1st.11.0−−−+N (0)/1N (0)/3Corynebacterium sp. (2)/2P (2)/23.130553MAnkleOIAI/debridement247.0+++NDNDMRSA (1)/2MRSA (1)/3P (3)/30.145685MHipOIAI/debridement115.0−++NDGBS (1)/1GBS (1)/2N (0)/1P (1)/10.237772FHipPJI/1st.2.3−+−+S. maltophilia (1)/2N (0)/3N (0)/2N (0)/22051874FHipPJI/2st.274.0−−++NDCNS (2)/2CNS (2)/2P (2)/2139920MAnkleOIAI/debridement0.1+++NDNDN (0)/2N (0)/1N (0)/12331074FHipPJI/2st.1.6−+−+N (0)/1N (0)/8N (0)/3P (1)/39451176FHipPJI/1st.50.5−++NDStreptococcus sp. (1)/1N (0)/5N (0)/4P (2)/418381278FHipPJI/2st.12.1+−++N (0)/1Corynebacterium sp. (1)/7N (0)/1P (1)/17361363MHIpPJI/1st.154.0−+++Streptococcus sp. (1)/1Streptococcus sp. (2)/4N (0)/3P (3)/39361443MHipPJI/2st.22.3−+−+Campylobacter fetus subsp. (1)/1N (0)/6N (0)/1N (0)/18341580FHipPJI/1st.10.2−−−+N (0)/1Streptococcus sp. (3)/9N (0)/1P (1)/125241681FKneeOIAI/1st.20.9−−−NDCNS (1)/1CNS (1)/3N (0)/1P (1)/11.2371780MElbowPJI/1st.47.5−−++Streptococcus sp. (1)/1N (0)/4Streptococcus sp. (1)/1N (0)/110.741887MHipPJI/1st.2.5−−−−N (0)/1CNS (2)/5CNS (2)/2P (2)/21.4461974FHipPJI/1st.4.6−−−+NDCNS (1)/4CNS (2)/2P (1)/212352022MHipOIAI/debridement1.7−−−−NDMSSA (1)/2MSSA (1)/1P (1)/11.329 a PJI, prosthetic joint infection; OIAI, orthopedic implant-associated infection; 1st., one-stage exchange; 2st., two-stage exchange; ORIF, open reduction internal fixation; CRP, C-reactive protein; P, positive; N, negative; ND, not done; +, presence; −, no presence; CNS, coagulase-negative Staphylococcus sp.; VGS, viridans group Streptococcus sp.; MSSA, methicillin-susceptible Staphylococcus aureus; MRSA, methicillin-resistant Staphylococcus aureus; coryne. sp., Gram-positive bacillus resembling Corynebacterium sp.; GBS, Group B Streptococcus sp.; S. maltophilia, Stenotrophomonas maltophilia. b (Number of positive specimens)/Number of samples. c PCR positive criteria; ≤25.4 (Cq value); established analytical cutoff via next-generation sequencing validation. Diagnostic and microbiological results of individual study subjects with orthopedic implant-associated infectionsa PJI, prosthetic joint infection; OIAI, orthopedic implant-associated infection; 1st., one-stage exchange; 2st., two-stage exchange; ORIF, open reduction internal fixation; CRP, C-reactive protein; P, positive; N, negative; ND, not done; +, presence; −, no presence; CNS, coagulase-negative Staphylococcus sp.; VGS, viridans group Streptococcus sp.; MSSA, methicillin-susceptible Staphylococcus aureus; MRSA, methicillin-resistant Staphylococcus aureus; coryne. sp., Gram-positive bacillus resembling Corynebacterium sp.; GBS, Group B Streptococcus sp.; S. maltophilia, Stenotrophomonas maltophilia. (Number of positive specimens)/Number of samples. PCR positive criteria; ≤25.4 (Cq value); established analytical cutoff via next-generation sequencing validation. ROC analysis of qPCR Cq values for distinguishing contamination-dominated low-biomass samples from higher-biomass samples. The optimal cutoff (Cq = 26.4) was determined using the Youden index. Area under the curve (AUC), 0.82; 95% CI, 0.74–0.91. Conceptual diagram illustrating the relationship between bacterial quantity and molecular diagnostic thresholds. The schematic workflow highlights the inverse correlation between qPCR Cq values (top scale) and absolute bacterial counts (bottom scale) across low-biomass and high-biomass conditions. The clinical-based cutoff (Cq = 26.4), corresponding to 104 CFU/mL, defines the threshold for infection diagnosis via qPCR, whereas the NGS-based cutoff (Cq = 25.5), corresponding to 105 CFU/mL, represents the analytical limit for reliable bacterial identification. The diagram delineates two distinct analytical zones: a low-biomass region where specimens diagnosed as infected via qPCR may remain unidentifiable or prone to artifact contamination under NGS single-testing a high-biomass region where NGS can provide confirmatory evidence. Characteristics of sequencing results near the quantitative limit and determination of the NGS cutoff Mock analysis detected microbes from all 17 target genera. However, the sequencing results demonstrated a low relative abundance of Gram-negative bacteria; specifically, NGS analysis of sample HM-782D identified several Gram-negative taxa with varying relative abundances, including Acinetobacter (0.09%), Enterobacteriaceae (0.08%), and Pseudomonadaceae (0.01%) (see Table S1). Regarding the bacteria identified in both the positive control (serially diluted E. coli) and the negative controls, seven of the top ten bacteria in the quantitative positive control (highlighted in bold in Table S2) were also present in the negative controls. In the NGS results for the serially diluted E. coli, members of the family Enterobacteriaceae were most frequently detected at concentrations ranging from 107 to 105 CFU/mL. However, at 104 CFU/mL, Enterobacteriaceae (presumed to represent E. coli) was no longer the dominant taxon, accounting for only 4% of the total reads (see Table S2). Consequently, a Cq value of 25.5 (10⁵ CFU/mL) was determined as the operational threshold for reliable bacterial detection. Because higher Cq values reflect lower bacterial biomasses that are prone to contamination, the analytical cutoff value for reliable detection was conservatively established as Cq ≤ 25.4. Diagnostic performance of qPCR using the NGS-derived cutoff value The sensitivities of synovial fluid culture, peri-implant tissue culture, and SFC were 67%, 60%, and 50%, respectively, whereas their specificities were 100%, 97%, and 100%, respectively (see Table 5). In comparison, when applying the NGS-derived cutoff value (Cq ≤ 25.4), the qPCR assay achieved a sensitivity of 80% and a specificity of 72% (see Table 5). To enhance diagnostic accuracy, we evaluated the performance of combining these microbial culture methods with the qPCR assay (see Table 6). Crucially, when defining positivity as either a single positive culture or a positive qPCR result (Cq ≤ 25.4), the combination method achieved the highest sensitivity at 95.0% (95% CI: 73.1%–99.7%), albeit with a reduced specificity of 69.4% (95% CI: 51.7%–83.1%). Conversely, when applying a stricter consensus criterion requiring either two positive cultures or one positive culture plus a positive qPCR result, the combination strategy demonstrated a highly balanced diagnostic profile, achieving a sensitivity of 80.0% (95% CI: 55.7%–93.3%) and a perfect specificity of 100% (95% CI: 88.0%–100%), alongside a positive predictive value of 100% and a negative predictive value of 90.0% (see Table 6). TABLE 5Comparison of microbiologic testa,b,c,d,eTestCases (N = 56)SensitivitySpecificityPositive predictive valueNegative predictive valueOIAINon-infectedNo. of patients with positive specimensb(95% CI)Preoperative culture Synovial fluid10/150/1266.7 (38.7–87.0)100 (69.8–100)100 (65.5–100)70.1 (44.0–88.6) Peri-implant tissue culture ≥1 positive culture12/201/3660.0 (36.4–80.0)97.2 (83.8–99.9)92.3 (62.1–99.6)81.4 (66.1–91.1) ≥2 positive culturesc4/190/3621.1 (7.0–46.1)100 (90.0–100)100 (39.6–100)70.6 (56.0–82.1) Sonicate fluid cultured ≥1 positive culture10/200/3650.0 (27.9–72.1)100 (88.0–100)100 (65.5–100)78.3 (63.2–88.5) ≥2 positive culturesc5/120/3641.7 (16.5–71.4)100 (88.0–100)100 (46.3–100)83.7 (68.7–92.7)Sonicate fluid PCRe Cq ≤ 25.4 ≥1 positive PCR16/2010/3680.0 (55.7–93.4)72.2 (54.6–85.2)61.5 (40.7–79.1)86.7 (68.4–95.6) ≥2 positive PCR7/121/1458.3 (28.6–83.5)92.9 (64.2–99.6)87.5 (46.7–99.3)72.2 (46.4–89.3) a OIAI, orthopedic implant-associated infection; PCR, polymerase chain reaction; Cq, quantification cycle. Clinical Calculator 1 was used for the calculation of sensitivity, specificity, positive predictive value, negative predictive value, and their corresponding 95% confidence intervals (18, 29). b Where the denominator is shown, data were not available for all study patients. c The number of cultures positive for the same microorganism is given. d Sonicate fluid culture with (centrifugation ≥ 1 colony-forming unit/plate); the number of colony-forming units per agar plate growing on either aerobic or anaerobic plates (whichever yielded higher counts) is given. e Positive PCR was defined as qPCR with a Cq value of ≤25.4, representing the NGS-derived cutoff. Comparison of microbiologic testa,b,c,d,e OIAI, orthopedic implant-associated infection; PCR, polymerase chain reaction; Cq, quantification cycle. Clinical Calculator 1 was used for the calculation of sensitivity, specificity, positive predictive value, negative predictive value, and their corresponding 95% confidence intervals (18, 29). Where the denominator is shown, data were not available for all study patients. The number of cultures positive for the same microorganism is given. Sonicate fluid culture with (centrifugation ≥ 1 colony-forming unit/plate); the number of colony-forming units per agar plate growing on either aerobic or anaerobic plates (whichever yielded higher counts) is given. Positive PCR was defined as qPCR with a Cq value of ≤25.4, representing the NGS-derived cutoff. TABLE 6Comparison of diagnostic performance among combined culturea,dTestCases (N = 56)SensitivitySpecificityPositive predictive valueNegative predictive valueOIAINon–infectedNo. of patients with positive specimensb(95% CI)Conventional culture method(Pre-operative culture and Tissue culture) ≥1 Positive culture17/201/3685.0 (61.1-96.0)97.2 (83.8-99.9)94.4 (70.6-99.7)92.1 (77.5-97.9) ≥2 Positive culturesc8/200/3640.0 (20.0-63.6)100 (90.0-100)100 (59.8-100)75.0 (60.1-85.9)Combination culture method(Pre-operative culture and tissue culture, sonicate fluid culture) ≥1 Positive culture18/201/3690.0 (66.9-98.2)97.2 (83.8-99.9)94.7 (71.9-99.7)94.6 (80.5-99.1) ≥2 Positive culturesc14/200/3670.0 (45.7-87.2)100 (88.0-100)100 (73.2-100)85.7 (70.8-94.1)Combination culture and PCR method (culture method and PCR) ≥1 Positive culture or PCR positivee cut-off ≤ 25.4 (Cq)19/2011/3695.0 (73.1-99.7)69.4 (51.7-83.1)63.3 (43.9-79.5)96.2 (78.4-99.8) ≥ 2 positive culturesc or one positive culture and 1 PCR positivee Cut-off ≤ 25.4 (Cq)16/200/3680.0 (55.7-93.3)100 (88.0-100)100 (75.9-100)90.0 (75.4-96.7) a OIAI, orthopedic implant-associated infection; Cq, quantification cycle. Clinical Calculator 1 was used for the calculation of sensitivity, specificity, positive predictive value, negative predictive value, and their corresponding 95% confidence intervals (18, 29). b Where the denominator is shown, data were not available for all study patients. c The number of cultures positive for the same microorganism is given. d Sonicate fluid culture with centrifugation (≥1 colony-forming unit (CFU)/plate); the number of CFUs per agar plate growing on either aerobic or anaerobic plates (whichever yielded higher counts) is given. e Positive PCR was defined as qPCR with a Cq value of ≤25.4, representing the NGS-derived cutoff. Comparison of diagnostic performance among combined culturea,d OIAI, orthopedic implant-associated infection; Cq, quantification cycle. Clinical Calculator 1 was used for the calculation of sensitivity, specificity, positive predictive value, negative predictive value, and their corresponding 95% confidence intervals (18, 29). Where the denominator is shown, data were not available for all study patients. The number of cultures positive for the same microorganism is given. Sonicate fluid culture with centrifugation (≥1 colony-forming unit (CFU)/plate); the number of CFUs per agar plate growing on either aerobic or anaerobic plates (whichever yielded higher counts) is given. Positive PCR was defined as qPCR with a Cq value of ≤25.4, representing the NGS-derived cutoff. Clinical characteristics and diagnostic outcomes of the NGS cohort A subset of patients from the qPCR-tested group who underwent implant removal within a defined timeframe was selected for subsequent NGS analysis. This cohort included 83 culture results from 15 patients: 6 with suspected infection and 9 with suspected non-infected loosening based on preoperative clinical records. Patient-level diagnostic findings, organism-level concordance between culture and NGS, and the diagnostic performance of sonicate fluid qPCR and qPCR combined with the NGS criterion in the 15-patient NGS cohort are summarized in Tables 7–9. These results comprised 9 preoperative joint aspirate cultures, 26 SFCs, and 48 conventional peri-implant tissue cultures (see Fig. 2). Three of the nine preoperative joint aspirate cultures were positive, all of which were confirmed as true infections according to postoperative diagnostic criteria (see Table 2). Among the 15 patients, 5 had positive cultures from either preoperative aspirates or surgical specimens; 4 of these 5 patients were ultimately diagnosed with infection based on clinical criteria (see Table 7). TABLE 7Detailed intraoperative microbiological, histopathological, and molecular results for the 15 patients in the NGS cohortaCaseSuspected infection: Yes/No (preoperative clinical records)Diagnosis with 2018 MSIS/ICMImplications of molecular testingEquivocal results with culture onlyHistopathologyPeri-implant culturebSonicated culture resultsCq (qPCR)Culture-based diagnosisCombination with qPCR and NGSSonicated sample typeSonicate fluid cultureb1YesInfection: PJIInfection: PJIConfirmedNoPositive(0)/1Artificial joint (2)Streptococcus (2)/217.7, 18.72YesNon.Non.SuggestiveYesNegative(0)/7Artificial joint (2)(0)/221.0, 27.43YesInfection: FRIInfection: FRIConfirmedNoPositiveCNS (1)/5Nail, peri-implant tissueCNS (1)/228.5, 21.44YesInfection: PJIInfection: PJIConfirmedNoPositive(0)/3Artificial joint (2)Corynebacterium (2)/221.6, 25.65YesNon.Non.SuggestiveYesNegative(0)/4Artificial joint (2)(0)/223.4, 30.86YesNon.Infection: PJIConfirmedYesND(0)/2Artificial joint (2)CNS (1)/224.0, 26.37NoNon.Non.SuggestiveYesND(0)/2Screw (1), peri-implant tissue (2)(0)/330.0, 30.7, 24.38NoNon.Non.ConfirmedNoNegative(0)/4Artificial joint(0)/125.69NoNon.Non.ConfirmedNoNDNDScrew(0)/125.910NoNon.Non.ConfirmedNoND(0)/3Artificial joint(0)/128.211NoNon.Non.ConfirmedNoND(0)/5Artificial joint (4)(0)/430.8, 28.2, 30.9, 30.712NoNon.Non.ConfirmedNoNegativeNDArtificial joint(0)/128.413NoNon.Non.ConfirmedNoNegative(0)/3Artificial joint(0)/129.414NoEquivocalNon.ConfirmedYesNegativeCNS (1)/4Artificial joint(0)/129.415NoNon.Non.ConfirmedNoNegative(0)/5Plate(0)/131.7 a MSIS/ICM, Musculoskeletal Infection Society/International Consensus Meeting; PJI, periprosthetic joint infection; FRI, fracture-related infection; qPCR Cq, quantification cycle; NGS, next-generation sequencing; ND, not done; CNS, coagulase-negative Staphylococcus sp. Note: genus-level pathogens identified in the table (e.g., Streptococcus, Corynebacterium) are presented in italics in accordance with international microbiological standards. b The ratio represents the number of positive specimens relative to the total number of samples collected. Detailed intraoperative microbiological, histopathological, and molecular results for the 15 patients in the NGS cohorta MSIS/ICM, Musculoskeletal Infection Society/International Consensus Meeting; PJI, periprosthetic joint infection; FRI, fracture-related infection; qPCR Cq, quantification cycle; NGS, next-generation sequencing; ND, not done; CNS, coagulase-negative Staphylococcus sp. Note: genus-level pathogens identified in the table (e.g., Streptococcus, Corynebacterium) are presented in italics in accordance with international microbiological standards. The ratio represents the number of positive specimens relative to the total number of samples collected. TABLE 8Comparison of conventional culture results and NGS taxonomic assignments in culture-positive infected cases using the R package decontamaCase no. (Table 2)AgeSexCRP (mg/L)Reason for surgeryJointReceived antibioticsAspiration joint fluid (preoperative) culturePeri-implant cultureSonicated implantSonicate fluid (positive number)Concordance with NGS results164F276PJIHip+Streptococcus sp.NegativeArtificial joint; stem, UHMWPE fiber cableStreptococcus sp. (2)Exact355M92FRIHip−CNSCNSNailCNSPartial483M11PJIHip−–NegativeArtificial joint (2); cup, stemCoryne. sp. (2)None670M30PJIKnee+CNSNegativeArtificial joint; tibial insertCNSPartial a CRP, C-reactive protein; +, presence; −, absence; UHMWPE fiber cable, ultrahigh-molecular-weight polyethylene fiber cable; CNS, coagulase-negative Staphylococcus sp.; Coryne. sp., Gram-positive bacillus resembling Corynebacterium sp. Comparison of conventional culture results and NGS taxonomic assignments in culture-positive infected cases using the R package decontama CRP, C-reactive protein; +, presence; −, absence; UHMWPE fiber cable, ultrahigh-molecular-weight polyethylene fiber cable; CNS, coagulase-negative Staphylococcus sp.; Coryne. sp., Gram-positive bacillus resembling Corynebacterium sp. TABLE 9Clinical performance report and comparison of diagnostic tests in the next-generation sequencinga,fTestCases (N = 15)SensitivitySpecificityPositive predictive valueNegative predictive valueOIAINon–infectedNo. of patients with positive specimensb(95% CI)Conventional cultureJoint-fluid, Peri-implant tissue culture ≥1 Positive culture3/41/1175 (22-99)91 (57-100)75 (22-99)91 (57-100) ≥2 Positive culturesb1/40/1125 (1-78)100 (68-100)77 (49-95)79 (49-94)Sonicate fluid culturec ≥1 Positive culture4/40/11100 (40-100)100 (68-100)100 (40-100)100 (68-100) ≥2 Positive culturesb2/40/1150 (9-91)100 (68-100)100 (20-100)85 (54-97)Sonicate fluied PCR d ≥1 Positive PCR4/43/11100 (40-100)73 (39-93)57 (20-88)100 (60-100) ≥2 Positive PCR1/40/1125 (1-78)100 (68-100)79 (49-95)79 (49-94)Sonicate fluied PCR with NGSe ≥1 Positive PCR with NGS3/40/1175 (22-99)100 (68-100)100 (31-100)92 (60-100) ≥2 Positive PCR with NGS1/40/1125 (1-78)100 (68-100)77 (49-95)79 (49-94) a Clinical Calculator 1 was used for calculation of sensitivity, specificity, positive and negative predictive value, and confidence intervals (18, 29). b The number of cultures positive for the same microorganism is given. c Sonicate-fluid culture with centrifugation (≥1 CFU/plate); the number of colony-forming units (CFUs) per agar plate growing on either aerobic or anaerobic plates (whichever yielded higher counts) is given. d Positive PCR was defined as qPCR with a Cq value of ≤25.4. e Positive PCR with NGS; a single positive PCR result with NGS occurs when the same bacteria are detected by both PCR and NGS in preoperative and intraoperative culture tests. Alternatively, a positive result occurs when the same bacteria are detected in two or more NGS samples, even if they are not present in the culture tests. f NGS, next-generation sequencing. Clinical performance report and comparison of diagnostic tests in the next-generation sequencinga,f Clinical Calculator 1 was used for calculation of sensitivity, specificity, positive and negative predictive value, and confidence intervals (18, 29). The number of cultures positive for the same microorganism is given. Sonicate-fluid culture with centrifugation (≥1 CFU/plate); the number of colony-forming units (CFUs) per agar plate growing on either aerobic or anaerobic plates (whichever yielded higher counts) is given. Positive PCR was defined as qPCR with a Cq value of ≤25.4. Positive PCR with NGS; a single positive PCR result with NGS occurs when the same bacteria are detected by both PCR and NGS in preoperative and intraoperative culture tests. Alternatively, a positive result occurs when the same bacteria are detected in two or more NGS samples, even if they are not present in the culture tests. NGS, next-generation sequencing. The surgical specimen culture results for all patients are summarized in Table 7. In all four infected cases, SFCs were positive. Notably, in three of these cases, bacteria were detected exclusively in the SFCs, whereas in the remaining case, pathogens were identified in both peri-implant tissue and SFCs. Among the cases with suspected infection, qPCR-positive results were observed in two patients presenting with hip prosthesis loosening. Furthermore, in a non-infected patient who underwent primary total hip arthroplasty for post-traumatic osteoarthritis, a qPCR-positive result was observed in the residual implant samples from the initial trauma surgery. Preoperative NGS analysis revealed multiple bacterial species in one case of suspected septic loosening, whereas Cutibacterium was detected in another case. Crucially, in a patient with post-traumatic osteoarthritis in whom infection was not suspected at the time of the primary THA, PJI developed 2 years after surgery during the 3-year clinical follow-up. Retrospective analysis confirmed that the Staphylococcus species detected at the onset of PJI had already been identified in the baseline NGS results from the initial surgery. Bacterial dominance and potential contaminants in low-biomass samples The complete NGS results for all clinical specimens are detailed in Tables S2 and S3. Using the predefined cutoff value of Cq ≤ 25.4 (corresponding to 10⁵ CFU/mL of E. coli), the clinical samples were stratified into two groups: low-biomass and high-biomass. To compare the microbial profiles between these groups, the top ten taxa were selected based on their relative abundances at the order, family, and genus levels (see Table S3). In the low-biomass samples, Streptococcus (rank 1) and Cutibacterium (rank 2) accounted for disproportionately large fractions of the sequencing reads, suggesting that these taxa represent typical background contamination under low-biomass conditions. Concordance between SFC and NGS for pathogen detection Among the six culture-positive SFCs, the same bacterial genus was identified via NGS as the dominant taxon in four samples across three cases (cases 1, 3, and 6; Table 8). Conversely, in the remaining two culture-positive samples, discrepancies were observed between conventional culture and post-decontamination NGS results; specifically, conventional cultures identified Corynebacterium, whereas NGS identified Cutibacterium as the dominant genus (case 4; Table 8). NGS analysis of qPCR-positive cases with occult infections We conducted a detailed taxonomic analysis at the genus level and below using a three-step framework: (i) genus-level classification, (ii) pre-decontamination profiling (excluding chloroplast and plant-related lineages), and (iii) post-decontamination analysis using the decontam algorithm. In one culture-negative case (case 2, sample 3 [see Fig. S2]), qPCR and initial NGS analysis identified Streptophyta. However, after filtering out these chloroplast- and plant-derived contaminant sequences, decontam analysis excluded Streptophyta and revealed a polymicrobial profile; the most abundant genus was Streptococcus (20%), followed by Cutibacterium (17%) and Anaerococcus (14%) (see Table S5). Other identified genera included Paracoccus (11%), Staphylococcus (7%), Burkholderia (7%), and Candidatus Rhodoluna (5%), with minor taxa accounting for the remaining 19% of the total reads (see Table S5). In the two cases described below, removing these contaminant signals and performing decontam analysis did not alter the detected bacterial taxa. Instead, the relative abundance of each remaining taxon increased (see Table S5). For instance, in another culture-negative, qPCR-positive sample from case 5 (see Table S6), NGS identified Cutibacterium (32% pre- and 36% post-decontamination) and Streptococcus (28% vs 32%, respectively) (see Table S6). Similarly, a primary total hip arthroplasty (THA) case (case 7) for post-traumatic osteoarthritis (see Fig. 6; Table S7) yielded a qPCR-positive sample (Cq = 24.3, sample 13, Table S7) that contained Streptococcus (60% pre- and 63% post-decontamination), Cutibacterium (20% vs 22%), and Staphylococcus (5% vs 5%). Heatmap visualization of microbial profiles and bacterial DNA quantities. The heatmap illustrates the relative taxonomic abundance across clinical specimens and control samples. Relative abundance values are color-coded based on a three-color scale: minimum (0%, blue), midpoint (50%, white), and maximum (100%, red). Samples are ordered on the horizontal axis by bacterial DNA quantity corresponding to the Cq values. On the vertical axis, bacterial taxa are listed in descending order of their cumulative relative abundance across all sequencing reads, reflecting operational taxonomic classifications. Black arrows indicate the quantitative positive controls (Escherichia coli dilutions from 107 to 104 CFU/mL), while the black triangle and double triangles denote the NF water and sterile saline negative controls, respectively. DISCUSSION A key finding of this study is the strong technical reliability and reproducibility of our 16S rRNA gene qPCR protocol, which serves as a crucial foundation for accurate downstream NGS analysis. As demonstrated by our validation using serial E. coli dilutions, Cq measurements exhibited minimal variability, with an ICC [1,1] of 0.961 (see Fig. 3). This high level of reproducibility strongly aligns with our previous methodological validation, which reported near-perfect reliability with a single-measure ICC of 0.991 and an average-measure ICC of 0.998 (14). The statistical significance of these values, further supported by the robust F-test results (F = 441.520, P < 0.001), underscores that our molecular assay yields highly consistent results with negligible technical noise across different bacterial biomass levels. In the context of microbiome-based diagnostics, establishing such rigorous intra-assay consistency is paramount. High-throughput sequencing platforms, particularly when applied to low-biomass clinical specimens like orthopedic sonication fluids, are highly susceptible to false positives arising from background noise or reagent contamination (12, 30). By confirming that our qPCR pre-screening method delivers tightly reproducible Cq values, we demonstrate its utility not only as a standalone diagnostic aid but also as an essential quality-control filter prior to NGS testing (10). Utilizing this protocol as a screening tool allows clinicians to objectively assess bacterial load thresholds, ensuring that only high-confidence samples proceed to expensive sequencing workflows, thereby enhancing both diagnostic accuracy and cost-effectiveness in clinical practice. Clinical-based Cq cutoff value obtained from statistical analysis The ROC curve (see Fig. 4) exhibited an area under the curve (AUC) of 0.820 (95% CI: 0.735–0.905), suggesting that this genetic test possesses good discriminatory ability for the target disease. The result was statistically significant (P < 0.001), rejecting the null hypothesis that the true AUC is 0.5. The ROC curve demonstrates a favorable balance between sensitivity and specificity across different thresholds, suggesting that qPCR is a useful tool for distinguishing between positive and negative cases in this setting. Although variations in amplification efficiency in samples near the limit of detection can affect the sensitivity–specificity trade-off due to the nature of qPCR, the fact that the lower limit of the CI is 0.735 indicates a high level of reliability for clinical implementation. Characteristics of sequencing results near the quantitative limit and the impact of biomass on microbial profiles Regarding bacterial taxa other than E. coli, the analytical outcomes after applying decontam and filtering out chloroplast-associated reads revealed critical insights into low-biomass sequencing behavior. At a dilution concentration of 104 CFU/mL, the sequencing results were nearly identical to the NGS profiles of NF water samples. As the bacterial concentration gradually decreased from 107 CFU/mL, the taxonomic composition showed a progressive trend toward mimicking the NF water sequencing data (see Fig. 7). This finding demonstrates that under low-biomass conditions, even when sterile physiological saline is utilized for sample dilution, the resulting profile inherently converges with the background noise obtained from NF water within the NGS processing workflow (see Table S5). The relative abundance of bacterial taxa was highly dependent on the initial DNA concentration of the samples. As shown in Fig. 7, samples with low Cq values (reflecting high DNA concentrations, or high-biomass conditions) were heavily dominated by the target biological signal. However, a progressive increase in the proportions of Cutibacterium and Streptococcus was observed as the DNA concentration decreased (corresponding to increasing Cq values). This inverse correlation strongly suggests that low-biomass samples are particularly susceptible to background contamination, even after computational decontamination procedures. Our results highlight the inherent challenges of interpreting microbial profiles in low-biomass environments, indicating that when the target DNA concentration falls below a critical threshold, the true biological signal becomes indistinguishable from background noise. This necessitates a cautious interpretation of taxonomic data in high-Cq samples. While high-biomass samples maintained a clear and reliable profile, the prominent emergence of Streptococcus and Cutibacterium in the highly diluted series (see Fig. 7) suggests that current decontamination methods may not completely eliminate reagent-derived or laboratory-acquired DNA. Taken together, these findings underscore the paramount importance of including negative controls (see Fig. 6 and 7) and implementing concentration-based filtering thresholds in metagenomic pipelines (see Table 9). Tipping point of microbial profile convergence after processing with the decontam package. The stacked bar chart illustrates the relative taxonomic abundance of the top ten bacterial genera across serially diluted Escherichia coli positive controls and negative controls after processing with the decontam package. Samples are arranged on the horizontal axis in order of decreasing bacterial biomass. The composition of the highly diluted positive controls progressively transitions and converges with the background contamination profiles observed in the NF water and sterile saline negative controls. Bacterial dominance, contamination risk, and the necessity of a Cq-based cutoff in low-biomass environments Moderate correlations were observed between Cq values and the relative abundances of Sediminibacterium, MLE1-12, and Sinobacteraceae among the top ten most abundant bacterial taxa identified by NGS in the implant sonicated fluid samples. A weak correlation was also observed for Sphingomonadaceae (see Table S3). Consistent with our previous analyses demonstrating a positive correlation between Cq values and the relative abundance of specific background bacteria, these taxa were classified as likely contaminants and systematically removed using decontam (see Table S3). Furthermore, seven of the top ten taxa were consistently detected in both the high- and low-biomass groups (see Table S3). Notably, Streptococcus and Cutibacterium were observed at high frequencies, even in samples with exceptionally low bacterial DNA concentrations. This finding mirrors the contamination profiles observed in our negative controls (see Fig. 6). Because these two genera are well-documented opportunistic pathogens capable of causing orthopedic infections, they must be rigorously differentiated from background contamination. However, their high prevalence in low-biomass specimens makes it challenging to determine whether they represent true causative pathogens or experimental artifacts based on NGS data alone. These findings highlight a critical limitation of NGS in low-biomass samples: the potential distortion of microbial community compositions by contaminating background bacteria (9, 10). Such contamination can obscure true biological signals, leading to the misinterpretation of high-throughput sequencing results. Our data indicate that even after applying computational contamination removal tools, the resolution remains incomplete. While these bioinformatic tools are highly effective for high-biomass samples, they do not necessarily elucidate the true microbial profile under low-biomass conditions. To address this technical limitation, establishing a Cq-based quality control cutoff is essential. In this study, a concentration-dependent tipping point was identified between Cq values of 25.5 (104–105 CFU/mL) and 26.4 (104 CFU/mL), at which the microbial composition shifted from target-organism dominance to contamination dominance. Consequently, NGS data obtained from samples yielding Cq values ≥ 25.5 (104–105 CFU/mL) likely reflect experimental artifacts rather than the actual in vivo microbial composition of the specimen. Our results also revealed a critical discrepancy between the clinical qPCR diagnostic threshold and the analytical detection limit of NGS. At the cutoff value utilized to infer infection via qPCR, the bacterial concentration falls within the low-biomass range, resulting in compromised accuracy when evaluated by NGS (see Fig. 5). Therefore, integrating complementary quantitative methods, such as the present qPCR assay, is essential to safeguarding the accuracy of NGS-based bacterial analyses in clinical settings (see Table 9). Concordance between culture and NGS methods for pathogen identification In infected cases, four of the six culture test results were fully consistent with the post-decontamination NGS outcomes (see Table 8). One patient (case 4) presented with two unmatched results: conventional cultures identified Corynebacterium spp., whereas NGS identified Cutibacterium spp. as the dominant genus (sample 7; Fig. 6; Fig. S2). This discrepancy is likely attributable to a conventional, culture-based identification error rather than an analytical artifact. Several lines of evidence support this interpretation. First, previous studies utilizing identical 16S rRNA universal primers (8F–338R) confirmed the effective amplification and detection of both Corynebacterium and Cutibacterium lineages (24). Second, Cutibacterium was successfully detected in our baseline validation at the expected frequency (see Table S1). Third, the present NGS workflow has been previously validated for the successful identification of Cutibacterium in cases of chronic osteomyelitis (13). Fourth, Corynebacterium can be robustly sequenced with primers targeting the same V1–V2 variable regions of the bacterial 16S rRNA gene using the Ion Torrent PGM platform (25). Overall, these findings strongly suggest that the observed discrepancy did not arise from limitations in primer design, sequencing hardware, or bioinformatic pipelines. Instead, the misidentification likely stems from the inherent limitations of conventional phenotyping. The biochemical characteristics of Corynebacterium species are notoriously difficult to confirm using routine bacteriological testing, making accurate species-level identification highly challenging. Previous research has underscored the systemic difficulty of identifying coryneform bacteria within clinical microbiology laboratories using traditional phenotypic assays (31). High-throughput sequencing can effectively overcome these diagnostic hurdles by providing reliable, sequence-based classification of fastidious bacteria with overlapping biochemical or phenotypic characteristics, which are frequently misclassified or rendered unidentifiable by conventional culture-based systems. NGS analysis of qPCR-positive cases with occult infections In case 2 (see Fig. S3), microbiome analysis of the loosened prosthetic hip joint after applying the decontam package revealed a highly diverse bacterial composition supported by a high bacterial load. Although the prominent background signals made it challenging to definitively evaluate the pathogenic roles of Streptococcus and Cutibacterium, the results suggest that multiple bacterial species, including Staphylococcus, may have colonized the implant surface. Since these samples met validated quantification thresholds, these bacteria should not be immediately dismissed as technical artifacts. The identified Streptococcus lineages are dominant species in the human oral microbiome, and these molecular findings suggest that asymptomatic colonization by the oral commensal microbiota may contribute to the pathogenesis of delayed implant loosening through low-level, multi-species biofilm structures, rather than sterile mechanical damage (32–34). Similarly, in case 5 (see Fig. 6; Fig. S4), a one-stage revision THA was performed for suspected septic loosening. Although all conventional cultures were negative, one of the two intraoperative samples was qPCR-positive. NGS analysis of this high-biomass sample (sample 9; Table S6) identified Cutibacterium spp. (32%) and Streptococcus spp. (28%). Notably, the microbial composition shifted markedly between samples with differing bacterial DNA content (samples 9 and 10; Table S6). Specifically, the relative abundance of Streptococcus increased from 28% to 70% as the bacterial DNA content decreased, suggesting that Streptococcus represents a background contaminant in low-biomass environments. Conversely, the abundance of Cutibacterium decreased from 32% to 12% alongside declining bacterial DNA content, which aligns with the hypothesis that true biological signals diminish concurrently with reductions in total bacterial DNA (13, 14). For this reason, Cutibacterium was considered the true causative pathogen. Furthermore, post-decontamination profiling revealed Burkholderia at 3% in sample 9, and although the proportions of Cutibacterium and Streptococcus increased slightly, the overall biomass-dependent trends remained unaltered. In contrast, the low-biomass sample (sample 10; Table S6) exhibited a pattern highly prone to contamination. This contrast illustrates the complex, biomass-dependent difficulty of interpreting low-biomass specimens, rather than a simple binary distinction between pathogen and contaminant. A third patient (case 7; Fig. S5, Table S7) underwent primary THA for post-traumatic osteoarthritis, which was initially managed as a non-infected case. Crucially, 2 years after the primary THA, this patient developed a delayed PJI requiring a two-stage revision procedure, during which Staphylococcus was identified as the causative pathogen via culture. At the time of the primary surgery, both histopathology and conventional cultures were negative; however, a single screw-derived sample yielded a positive qPCR result (sample 13; Table S7, Cq = 24.3), with a Cq value below the NGS-derived cutoff of 25.4, indicating a bacterial DNA load above the corresponding threshold. Baseline NGS analysis of this sample identified Streptococcus spp. (60%), Cutibacterium spp. (21%), and Staphylococcus spp. (5%) (see Table S7). Although sample 13 was classified as high-biomass based on the Cq threshold, its taxonomic profile partially overlapped with the background noise of the NF water negative control, unlike the true negative samples (samples 14 and 15). This finding suggests that even samples satisfying the qPCR-based Cq threshold cannot be interpreted in isolation; instead, they must be comprehensively evaluated alongside taxonomic plausibility, contamination-aware filtering, and clinical context. Given the patient’s history of severe post-traumatic osteoarthritis following an acetabular fracture, the subclinical presence of bacteria on the implants was not entirely unexpected. A systematic review assessing THA outcomes following acetabular fractures reported that the risk of infection is substantially higher than in conventional hip arthroplasties, particularly in patients with multiple prior surgeries and retained internal fixation hardware (35). Furthermore, Knabl et al. demonstrated asymptomatic bacterial colonization of osteosynthesis implants in healthy patients without any clinical or laboratory signs of infection (36). Similar colonization phenomena have been documented in non-orthopedic fields, such as the subclinical colonization of cardiac electrophysiological devices (37). These observations underscore the extreme complexity of distinguishing true low-abundance biological signals from environmental contaminants, especially in post-traumatic reconstructions. Among the three patients in our cohort, two were managed intraoperatively under the clinical suspicion of infection and achieved uneventful long-term outcomes, whereas the remaining patient (case 7), in whom infection was not initially suspected or treated, subsequently required major revision surgery for delayed PJI. Although we cannot definitively conclude whether the baseline NGS signals directly reflected an active initial infection, the presence of qPCR-positive samples in all three cases cannot be ignored. The fact that the patient who did not receive proactive antimicrobial management later developed an indisputable PJI supports the potential clinical utility of our molecular screening framework to detect occult pathogens before they manifest clinically. Addressing false positives and contamination: past challenges and future perspectives A fundamental challenge in interpreting NGS data is that the methodology provides relative taxonomic proportions rather than absolute bacterial quantification. When a specimen contains a high bacterial DNA concentration, the resulting NGS profile is typically characterized by a high proportion of reads corresponding to a dominant pathogen. Indeed, in cases of overt bacterial infection, one or two bacterial genera usually dominate the sequencing output. In contrast, specimens with low bacterial abundance or those from non-infected cases frequently yield polymicrobial NGS profiles composed of multiple genera, significantly complicating clinical interpretation. Consequently, these methodological limitations have posed a long-standing barrier to the reliable interpretation of NGS findings. To date, detailed sequencing analyses focusing specifically on low-biomass specimens have not been rigorously conducted within the field of orthopedics. Although the sample size of the present study is relatively small, to our knowledge, this is the first orthopedic NGS investigation to specifically address and systematically stratify low-biomass samples using rigorous negative controls. Therefore, our findings are expected to serve as a foundational benchmark for future NGS research evaluating the implant-associated microbiota. Limitations of the study This study has several methodological limitations that should be considered when interpreting our findings. First, the clinical Cq values and established thresholds reported herein are inherently subject to inter-facility and inter-instrument variations, meaning they cannot be directly generalized without local calibration. The precise reliability and reproducibility of these absolute Cq metrics must be independently validated against characterized bacterial loads at individual testing facilities. Second, our total sample size was relatively limited, as the high-throughput NGS data were derived from a single sequencing run encompassing a specific patient cohort. To refine the operational limits of bacterial detection under low-biomass conditions, even stricter quantitative and qualitative quality control criteria may need to be implemented in larger prospective studies. Furthermore, extensive validation of identical clinical specimens under diverse experimental conditions remains necessary. This includes evaluations across different laboratories and on alternative sequencing platforms—such as short-read technologies like Illumina or long-read platforms like Oxford Nanopore vs the Ion Torrent platform utilized here. Comprehensive monitoring and multi-center cross-validation are required to further validate the universal robustness and reproducibility of our proposed diagnostic framework. Conclusion In conclusion, this study supports our hypothesis that NGS may support bacterial identification with negligible background noise when sample selection is guided by a designated Cq threshold. In overtly infected cases, NGS-based identification demonstrated high concordance with conventional cultures, while simultaneously exposing potential classification errors inherent to traditional phenotypic methods. Furthermore, our molecular workflow highlighted the capacity of NGS to resolve occult infections in culture-negative cases. However, because high-throughput sequencing provides relative taxonomic proportions rather than absolute quantification, distinguishing likely causative pathogens from environmental or reagent contaminants remains challenging in low-biomass environments. Our findings underscore that interpreting microbiome data from low-biomass specimens requires exceptional caution, as computational decontamination tools cannot completely eliminate low-level background artifacts. To address this technical hurdle, we demonstrated that integrating qPCR with NGS and applying a stringent analytical Cq cutoff effectively mitigates false-positive signals. Therefore, when utilizing NGS for the diagnosis of PJI, incorporating quantitative assessments alongside rigorous negative and positive controls is paramount. For cases with borderline or inconclusive sequencing profiles, long-term clinical monitoring remains indispensable for definitive determination. Ultimately, larger multi-institutional prospective studies are warranted to refine these diagnostic thresholds, enhance NGS resolution, and firmly establish its clinical applications in orthopedic practice. Supplementary Material Reviewer comments