Duplex Recombinase Polymerase Amplification‐Lateral Flow Assay for Rapid Visual Detection of Critical Antibiotic Resistance Genes in Escherichia coli and Klebsiella pneumoniae From Positive Blood Cultures Authors: Variya Nemidkanam, Pattamaporn Deechart, Nuntaree Chaichanawongsaroj Published: 2026 ABSTRACT ABSTRACTBackgroundBloodstream infections (BSIs) caused by multidrug‐resistant bacteria (MDR) are a leading cause of mortality and morbidity among hospitalized patients. Extended‐Spectrum Beta‐Lactamase (ESBL), carbapenem‐resistant Enterobacterales (CRE), and colistin resistance are critically important antibiotic‐resistant bacteria. The rapid identification of these critical antibiotic resistance genes from positive blood cultures is essential for accelerating the diagnostic turnaround time, enabling early therapy, and supporting effective antibiotic stewardship programs.MethodsRecombinase Polymerase Amplification (RPA) integrated with an in‐house Lateral Flow Assay (LFA) for duplex detection of bla CTX‐M/mcr‐1 and bla oxa‐48‐like/bla NDM was established in this study. 5′ modified primers compatible with those immobilized antibodies on LFA were designed. The primer concentrations, temperature, and time were optimized at 0.2–0.6 μM, 36°C–39°C, and 20–30 min, respectively. The duplex RPA amplicons were diluted in buffer and visually detected on in‐house lateral flow strips within 10–15 min. The limit of detection (LOD), specificity, and method validation were evaluated.ResultsThe optimal primer concentrations of bla CTX‐M/mcr‐1 (0.35 μM/0.35 μM) and bla oxa‐48‐like/bla NDM (0.2 μM/0.6 μM) provided balanced signal intensities for all target genes at 37°C–38°C within 25 min. The LOD across incubation time of 0–8 h was determined as 107–102 CFU/mL. No cross‐reactivity was observed with other common Gram‐negative pathogens. Sensitivity and specificity of duplex RPA‐LFA for bla CTX‐M, mcr‐1, and bla oxa‐48‐like/bla NDM detection were 85.71%–98.28% and 93.62%–100%. The accuracy was 93.33%–96.13%.ConclusionThis developed duplex RPA‐LFA assay is well‐suited for rapid screening in positive blood cultures, providing a cost‐effective tool for the treatment, surveillance, and control of MDR in critical infections.Keywords: bla CTX‐M , bla NDM , bla oxa‐48‐like , Escherichia coli , Klebsiella pneumoniae , lateral flow assay, mcr‐1, recombinase polymerase amplification Introduction Escherichia coli is the most common cause of bacteremia, particularly in high‐income countries, exceeding other major pathogens associated with bacteremia. The community‐acquired bacteremia accounted for the highest bacteremia episode, followed by community‐onset healthcare‐associated and hospital‐acquired bacteremia. The extraintestinal pathogenic E. coli infection in the bloodstream causes severe complications in hospitalized patients, leading to high mortality rates. The incidence rate and mortality risk are increased with age. A systematic review showed that the mortality rate of bloodstream infections (BSIs) ranged from 8% to 48% per year [1, 2]. Beta‐lactam antibiotics are the most common first‐line drugs for Gram‐negative bacteria treatment due to their minimal side effects and broad spectrum. However, various mechanisms are responsible for beta‐lactam resistance, particularly beta‐lactamase production. Extended‐spectrum beta‐lactamase (ESBL) is a typical resistance mechanism of the Enterobacteriaceae family, particularly E. coli and Klebsiella pneumoniae [3]. The percentages of ESBL‐producing E. coli and K. pneumoniae increased significantly over time in BSIs [4]. With the emergence of ESBL‐producing bacteria, carbapenem is the drug of choice for the treatment of ESBL‐producing bacteria and a last resort for curing multidrug‐resistant (MDR) Enterobacterales. Colistin is also a last resort regimen for the management of systemic MDR Gram‐negative bacterial infections, including BSIs. The global escalation of carbapenem‐ and colistin‐resistant strains over the past decade represents a critical challenge [5]. In Thailand, colistin resistance in K. pneumoniae isolates has been reported as one of the highest rates in the region at levels as high as 19.2%. Furthermore, the pooled prevalence of MDR of K. pneumoniae from BSIs has reached 80.1% [6]. The clinical consequences of these MDR are severe. The 30‐day mortality rate for patients with BSIs caused by antibiotic‐resistant bacteria has been reported as high as 32% [7]. Systemic studies have identified bla CTX‐M as the most prevalent ESBL gene in clinical E. coli isolates [3, 8, 9]. Moreover, the prevalence of bla CTX‐M‐producing E. coli strains co‐harboring carbapenem resistance genes has increased, with bla NDM as the most frequently co‐detected resistance gene [8]. Regarding carbapenem‐resistant K. pneumoniae (CRKP) in BSIs, these pathogens rank as the second most prevalent in healthcare‐onset cases, following only Acinetobacter baumanii. Among CRKP blood isolates, bla oxa‐48‐like has emerged as the most widely distributed carbapenemase determinant [5]. Additionally, the global prevalence of colistin resistance in E. coli and K. pneumoniae isolates from BSI patients has also steadily increased [5, 6]. The mcr‐1 gene remains the primary mobile mechanism for colistin resistance, showing a notably higher prevalence in E. coli compared to K. pneumoniae [10]. Notably, the rising incidence of MDR E. coli and K. pneumoniae in BSIs is a major public health crisis, requiring the development of rapid molecular diagnostic tools. Conventional culture, identification, and antibiotic susceptibility testing (AST) remain the gold standard diagnostic practices. However, these processes are time‐consuming and labor‐intensive. Rapid and accurate identification of antimicrobial resistance genes is critical for the early diagnosis of effective therapy, which reduces unnecessary prescription of broad‐spectrum antibiotics and improves patient outcomes. While various molecular technologies have the potential to shorten turnaround times, their implementation remains a challenge. Current diagnosis methods of positive blood culture in routine microbiological laboratories are mostly PCR‐based, such as T2 Biosystems, SeptiTest, BioFire FilmArray, and Verigene. Furthermore, emerging technologies designed for direct specimen analysis include digital PCR/IC3D, qPCR‐HRM, flow cytometry, PNA‐FISH technology, and next‐generation sequencing [11]. However, those technologies require complex, expensive instruments and highly trained laboratory personnel. Isothermal amplification methods have been developed to overcome the limitations of thermocycler‐dependent assays. Loop‐mediated isothermal amplification (LAMP) has been used for the rapid detection of Group B Streptococcus [12] and methicillin‐resistant Staphylococcus aureus in BSIs [13]. However, LAMP requires a complex set of 4–6 primers, which challenges the multiplex amplification and increases the risk of non‐specific binding and false‐positive results [14]. Recombinase polymerase amplification (RPA) is a simple isothermal amplification that has been widely developed for pathogen and antimicrobial resistance gene detection [15]. Recently, the biphasic blood drying method combined with LAMP or RPA has been shown to further enhance the sensitivity of bacterial detection in clinical samples [16]. Furthermore, the integration of RPA with lateral flow assays (LFA) provides a rapid, visual readout suitable for point‐of‐care diagnostics. RPA‐LFA was recently established to detect spiked blood samples containing S. aureus , K. pneumoniae , Pseudomonas aeruginosa , and Hemophilus influenza within less than 15 min [17]. Thus, this study aimed to develop duplex RPA‐LFA for the rapid detection of bla CTX‐M/mcr‐1 genes in E. coli and bla oxa‐48‐like/bla NDM in K. pneumoniae . The limit of detection and specificity of the developed assay were evaluated. Both sets of duplex RPA‐LFA targeting all four antimicrobial genes were examined with leftover and spiked blood samples. The results were compared with antimicrobial susceptibility testing obtained from the Sensititre automation system. This rapid and simple method is valuable for the management of BSIs in clinical practices. Materials and Methods Bacterial Strains Positive control strains harboring the target genes were used as follows: E. coli PG003 (bla CTX‐M and mcr‐1) and K. pneumoniae N009 (bla oxa‐48‐like and bla NDM). To validate mcr‐1, and bla NDM and bla oxa‐48‐like genes, 28 E. coli isolates harboring the mcr‐1 gene and 30 K. pneumoniae isolates harboring bla NDM and bla oxa‐48‐like were used to spike into negative hemoculture samples to mimic the real specimens. All these E. coli and K. pneumoniae pure isolates were characterized in our previous studies [18, 19]. All bacterial strains were cultured on MacConkey agar (Oxoid, Basingstoke, UK) at 37°C for 18–24 h. Bacteria stocks were preserved in 10% glycerol skim milk and stored at −70°C. Leftover Hemoculture Samples All leftover clinical hemoculture samples were kindly provided by Associate Professor Dr. Pitak Santanirand, Microbiology Laboratory, Department of Pathology, Faculty of Medicine Ramathibodi Hospital, Mahidol University, Bangkok, Thailand. All samples were collected between August 2024 and November 2024. A total of 105 leftover positive hemoculture samples were utilized for duplex RPA‐LFA validation, comprising of ESBL‐producing E. coli (58 samples) and non‐ESBL‐producing bacteria (47 samples). Additionally, 180 negative (no‐growth) hemoculture samples were collected for spiking experiments to determine the limit of detection (LOD) and specificity, as well as to perform the validation of duplex RPA‐LFA of mcr‐1, bla oxa‐48‐like, and bla NDM genes. The study protocol was approved by the Research Ethics Review Committee for Research Involving Human Research Participants, Group I, Chulalongkorn University, Thailand (COA No. 192/68). DNA Extraction Bacterial DNA extraction from hemoculture samples was performed using the QIAamp BiOstic Bacteremia DNA Kit (Qiagen, USA) according to the manufacturer's recommendations. Briefly, 1.8 mL of the sample was subjected to centrifugation at 13,000× g for 2 min. Subsequently, 450 μL of MBL was added, and the lysate was transferred into a 2 mL PowerBead. The mixture was then heated at 70°C for 15 min and centrifuged at 10,000× g for 1 min. IRS solutions were added and centrifuged at 10,000× g for 1 min. 1 mL of BB was added to the supernatant, then eluted with 50 μL of elution buffer through an MB spin column. Colonies of positive E. coli and K. pneumoniae strains were subjected to DNA extraction using the boiling method to obtain positive DNA control. Extracted DNA was stored at −20°C until further amplification. Optimization of Duplex RPA for bla CTX ‐M/mcr‐1 and bla oxa‐48‐like/ bla NDM Detection Two sets of duplex RPA reactions were performed using TwistAmp Basic Kit (TwistDX, Maidenhead, UK), following the manufacturer's recommendations (www.globalpointofcare.abbott/us/en/lp/twistdx/support.html). The bla CTX‐M and mcr‐1 gene primers were combined in the first reaction tube, while the bla oxa‐48‐like and bla NDM primers were combined in the second reaction tube. The primer sequences were modified at the 5′ end with biotin, fluorescein (FAM), or digoxigenin (DIG) as shown in Table 1. Parameters for each duplex RPA reaction were optimized. Initially, the concentrations of bla CTX‐M/mcr‐1 primers (0.35/0.35 μM, 0.40/0.35 μM, 0.45/0.35 μM, 0.35/0.28 μM, 0.35/0.32 μM) and bla oxa‐48‐like/bla NDM primers (0.2/0.6 μM, 0.3/0.6 μM, 0.2/0.4 μM, 0.2/0.5 μM, 0.2/0.6 μM) were optimized while temperature and time were fixed at 37°C for 25 min. Using the selected optimal primer concentration, the incubation temperature varied from 36°C to 39°C with the incubation time fixed at 25 min. Finally, with the optimal primer concentration and temperature, the incubation time was evaluated at 20, 25, and 30 min. TABLE 1Sequences of 5′ modified primers and their product size for duplex RPA‐LFA.Target geneSequence (5′–3′)Size (bp)References bla CTX‐M F: Biotin‐ATGTGCAGYACCAGTAARGTKATGGC R: FAM‐TGGGTRAARTARGTSACCAGAAYCAGCGG 593This study mcr‐1 F: Biotin‐AGTCCGTTTGTTCTTGTGGC R: DIG‐AGATCCTTGGTCTCGGCTTG 320[19] bla oxa‐48‐like F: Biotin‐GCGTGGTTAAGGATGAACAC R: DIG‐CATCAAGTTCAACCCAACCG 438This study bla NDM F: FAM‐GGTTTGGCGATCTGGTTTTC R: Biotin‐ACCGCCCATCTTGTCCTGAT 251This study Note: Degenerate base symbols: Y = C/T, R = A/G, K = G/T, S = C/G. Sequences of 5′ modified primers and their product size for duplex RPA‐LFA. F: Biotin‐ATGTGCAGYACCAGTAARGTKATGGC R: FAM‐TGGGTRAARTARGTSACCAGAAYCAGCGG F: Biotin‐AGTCCGTTTGTTCTTGTGGC R: DIG‐AGATCCTTGGTCTCGGCTTG F: Biotin‐GCGTGGTTAAGGATGAACAC R: DIG‐CATCAAGTTCAACCCAACCG F: FAM‐GGTTTGGCGATCTGGTTTTC R: Biotin‐ACCGCCCATCTTGTCCTGAT Note: Degenerate base symbols: Y = C/T, R = A/G, K = G/T, S = C/G. Each primer set was prepared as a master mix with 29.5 μL of rehydration buffer, which was then added to a freeze‐dried RPA pellet. Lastly, 1 μL of 50 ng DNA template and 1.25 μL of 280 mM magnesium acetate were added, resulting in a final reaction volume of 25 μL. The optimized reaction was incubated at 37°C for 25 min in a heat‐dried bath. Positive DNA control for all four target genes and sterile distilled water were included as positive and no template control (NTC) tubes, respectively. The RPA amplicons were detected by 2% agarose gel electrophoresis and with in‐house LFA. In‐House LFA Detection of Duplex RPA The LFA is composed of two test lines (coated with anti‐FAM and anti‐DIG, respectively) and one control line (coated with goat anti‐mouse IgG), as illustrated in Figure 1A. The RPA amplicon was diluted in running buffer at a ratio of 1:10 in a total volume of 150 μL. The diluted RPA amplicon was applied to the sample pad as in a previous study [19]. Results were observed by the appearance of a red line within 10–15 min. The presence of three red lines on bla CTX‐M/mcr‐1 duplex RPA‐LFA or bla oxa‐48‐like/bla NDM duplex RPA‐LFA indicates the presence of both target genes in that specific LFA (Figure 1B). The appearance of red lines at the specific position for bla CTX‐M, mcr‐1, bla oxa‐48‐like, or bla NDM, along with the control line, indicates the presence of the individual target gene. The absence of a red line at the control line position is interpreted as an invalid test result. To determine the optimal RPA‐LFA conditions, the intensities of the LFA lines were quantified using 8‐bit grayscale densitometry (ImageJ, USA). The signal intensity of the test line was normalized against that of the control line and reported as relative grayscale intensity (Test/Control ratio). The LFA was stored in an airtight bag containing a desiccant at 4°C. The shelf life of our in‐house LFA is approximately 1 year. Standard handling precautions were followed, including allowing the LFA and the running buffer to reach room temperature before use. All strips were used within 30 min of removal from the bag to ensure optimal capillary flow and prevent atmospheric moisture interference. Schematic representation and interpretation of the duplex lateral flow assay (LFA). Schematic illustration of the LFA components and the principle of detection (A). Visual interpretation of duplex detection results on the LFA strip, describing the patterns for negative, single‐positive, double‐positive, and invalid outcomes (B). Limit of Detection (LODs) of Duplex RPA‐LFA The positive bacterial strains of E. coli PG003 and K. pneumoniae N009 were initially adjusted to a turbidity of 0.5 McFarland standard (approximately 108 CFU/mL) [20, 21]. Ten‐fold dilutions were then performed to prepare inoculum concentrations ranging from 100 to 108 CFU/mL. Negative hemoculture bottles were spiked with 0.5 mL of each bacterial dilution. Five spiked hemoculture batches were prepared and incubated at 37°C. Each spiked hemoculture batch was subsequently retrieved for analysis at 0, 2, 4, 6, and 8 h. The LODs of each duplex RPA‐LFA reaction at each time point were determined in triplicate. All spiked hemoculture samples were subjected to DNA extraction, and the resulting genomic DNA was evaluated by bla CTX‐M/mcr‐1 and bla oxa‐48‐like/bla NDM duplex RPA‐LFA. Positive DNA control and NTC tubes were included in every experiment. The LODs were defined as the lowest bacterial concentration (CFU/mL) at which both test lines were clearly visualized by the naked eye. Specificity of Duplex RPA‐LFA The specificity of the duplex RPA‐LFA was determined using a panel of non‐target bacterial isolates. These panels were selected to ensure no cross‐reactivity with other common BSI pathogens that do not carry the target resistance genes. The following 12 clinical bacterial isolates were tested for the bla CTX‐M/mcr‐1 and bla oxa‐48‐like/bla NDM duplex RPA‐LFA: Salmonella paratyphi , Citrobacter freundii , Edwarsiella tarda, Pseudomonas aeruginosa , Enterococcus cloacae, Morganella morganii, Enterobacter aerogenes , Vibrio vulnificus , Klebsiella pneumoniae , E. coli , Salmonella typhi, and Klebsiella oxytoca . Each bacterial isolate was prepared in peptone water and spiked into negative hemoculture bottles (0.5 mL of 108 CFU/mL). To fully validate the assay's specificity, target‐positive strains were similarly spiked to serve as matrix‐matched positive controls. All spiked hemoculture samples were subjected to DNA extraction and evaluated by bla CTX‐M/mcr‐1 and bla oxa‐48‐like/bla NDM duplex RPA‐LFA. Positive DNA control and NTC tubes were included in every experiment. Validation of Duplex RPA‐LFA for bla CTX ‐M, mcr‐1, bla oxa‐48‐like, or bla NDM Detection The duplex RPA‐LFA for bla CTX‐M detection was validated using clinical samples positive for ESBL‐producing E. coli (58 samples). A total of 47 clinical hemoculture samples containing non‐ESBL‐producing bacteria were utilized as negative controls to evaluate the specificity of all target gene detections. Due to the low prevalence of colistin‐ and carbapenem‐resistant bacteria in BSIs, these clinical samples are very limited. To ensure robust validation for mcr‐1, bla oxa‐48‐like, or bla NDM, no‐growth hemoculture samples were spiked with 28 strains of mcr‐1 positive E. coli , and 30 strains of bla oxa‐48‐like and bla NDM‐positive K. pneumoniae . Each bacterial isolate was prepared in peptone water and spiked into negative hemoculture bottles (0.5 mL of 108 CFU/mL). All spiked hemoculture samples were subjected to DNA extraction, and the resulting genomic DNA was evaluated by bla CTX‐M/mcr‐1 and bla oxa‐48‐like/bla NDM duplex RPA‐LFA. Positive DNA control and NTC tubes were included in every experiment. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated. The gold standard for comparison was the phenotypic antimicrobial susceptibility results generated by an automated Sensititre ARIS HiQ system (Thermo Scientific, Waltham, MA, USA). Statistical analysis was performed using MedCalc software (https://www.medcalc.org/calc/diagnostic_test.php), accessed in October 2025. The diagnostic parameters were calculated using the following formulas based on a 2 × 2 contingency table. The formulas used for these calculations are as follows: Results Optimization of RPA‐LFA for Duplex bla CTX ‐M /mcr‐1 Genes and bla oxa‐48‐like/ bla NDM Genes Detection The optimal conditions for each assay were selected based on the grayscale densitometry method. The highest signal intensity was observed for both test lines on the LFA, while ensuring the absence of background signal in the no template control (NTC). Robust duplex RPA amplification was successfully detected by LFA across all tested primer concentrations, temperature ranges (36°C–39°C), and incubation time (20–30 min) (Figures 2 and 3). Optimization of duplex RPA‐LFA assays for bla CTX‐M/mcr‐1 genes detection. bla CTX‐M/mcr‐1 primers concentration (A), incubation temperature (B), and incubation time (C) were optimized. LFA images are displayed alongside their corresponding relative grayscale intensities, as determined by densitometric quantification. Data are presented as the mean of three replicates, with error bars representing the standard deviation. NTC, no template control. Optimization of duplex RPA‐LFA assays for bla oxa‐48‐like/bla NDM genes detection. bla oxa‐48‐like/bla NDM primers concentration (A), incubation temperature (B), and incubation time (C) were optimized. LFA images are displayed alongside their corresponding relative grayscale intensities, as determined by densitometric quantification. Data are presented as the mean of three replicates, with error bars representing the standard deviation. NTC, no template control. For the bla CTX‐M /mcr‐1 duplex RPA‐LFA assay, an equal concentration ratio of 0.35 μM of both primers provided the most balanced and intense signal for both target genes. Densitometric analysis confirmed that this concentration yielded the highest relative grayscale intensity for both bla CTX‐M and mcr‐1 test lines (Figure 2A). Regarding incubation temperature, all temperatures exhibited similar intensity (Figure 2B). However, the highest relative intensities for bla CTX‐M and mcr‐1 were observed at 38°C and 37°C, respectively. Consequently, 37°C was selected as the optimal temperature, as it produced the highest cumulative relative intensity across both targets. While the assay demonstrated a time‐dependent increase in signal intensity (Figure 2C), an incubation time of 25 min was selected as the optimal condition to minimize the total diagnostic turnaround time. Regarding the bla oxa‐48‐like/bla NDM duplex RPA‐LFA assay, a concentration ratio of 0.2 μM for bla oxa‐48‐like and 0.6 μM for bla NDM yielded the most distinct and balanced test lines of both target genes. This specific primer concentration and incubation temperature of 38°C showed the highest cumulative relative intensity across both targets (Figure 3A,B). Similar time‐dependent manners were also observed in this duplex assay (Figure 3C). Therefore, an incubation time of 25 min was also selected as the optimal condition. Limit of Detection for Duplex RPA‐LFA The LOD for bla CTX‐M/mcr‐1 duplex RPA‐LFA assay was determined to be 107, 106, 105, 103, and 102 CFU/mL at 0, 2, 4, 6, and 8 h, respectively (Figure 4A and Table 2). For bla oxa‐48‐like/bla NDM, LOD was determined to be 107, 107, 105, 103, and 102 at 0, 2, 4, 6, and 8 h, respectively (Figure 4B and Table 2). Limit of detection (LOD) of duplex RPA‐LFA assays over a time range from 0 to 8 h. Representative LFA images from three replicates of bla CTX‐M/mcr‐1 detection (A), bla oxa‐48‐like/bla NDM detection (B). TABLE 2Limit of detection (LOD) of duplex RPA‐LFA assays over a time range from 0 to 8 h.HourLimit of detection (LOD) (CFU/mL) bla CTX‐M /mcr‐1 bla oxa‐48‐like /bla NDM 0107 107 2106 107 4105 105 6103 103 8102 102 Limit of detection (LOD) of duplex RPA‐LFA assays over a time range from 0 to 8 h. Specificity of Duplex RPA‐LFA The specificity of the two duplex RPA‐LFA assays was evaluated using genomic DNA extracted from negative hemoculture samples spiked with 12 common Gram‐negative bacterial strains in BSIs. There was no amplification or cross‐reactivity observed when testing the bla CTX‐M/mcr‐1 and bla oxa‐48‐like/bla NDM duplex RPA‐LFA assay with DNA from S. paratyphi , C. freundii , E. tarda , P. aeruginosa , E. cloacae , M. morganii , E. aerogenes , V. vulnificus , K. pneumoniae , E. coli , S. typhi , and K. oxytoca (Figure 5A,B). The spiked target‐positive control (matrix‐matched control) showed a valid result, consistent with the NTC and positive DNA controls. Accordingly, these results demonstrate that both duplex RPA‐LFA assays are highly specific for their target genes, showing no cross‐reactivity with other common BSI Gram‐negative pathogens. Specificity of duplex RPA‐LFA against other common Gram‐negative BSI pathogens. Evaluation of the bla CTX‐M/mcr‐1 (A) and bla oxa‐48‐like/bla NDM (B) genes. NTC, no template control. Validation of Duplex RPA‐LFA With Clinical and Spiked Samples The performance of the duplex RPA‐LFA assays was validated against phenotypic antimicrobial susceptibility testing results. For bla CTX‐M detection in clinical samples, the duplex RPA‐LFA demonstrated a sensitivity and specificity of 98.28% [CI 90.76% to 99.96%] and 93.62% [CI 82.46% to 98.66%] with an accuracy of 96.19% (Table 3). TABLE 3Validation of duplex RPA‐LFA with the phenotypic antimicrobial susceptibility testing for bla CTX‐M detection in clinical sample.Phenotypic antimicrobial susceptibility testingDuplex RPA‐LFADuplex‐RPA‐LFA efficiency for bla CTX‐M detection bla CTX‐M+ bla CTX‐M−ESBL+571 Sensitivity 98.28% (95% CI: 90.76% to 99.96%) Specificity 93.62% (95% CI 82.46% to 98.66%) PPV 95% (95% CI: 86.40% to 98.27%) NPV 97.78% (95% CI: 86.29% to 99.68%) Accuracy 96.19% (95% CI: 90.53% to 98.95%) ESBL−344Total6045 Validation of duplex RPA‐LFA with the phenotypic antimicrobial susceptibility testing for bla CTX‐M detection in clinical sample. Sensitivity 98.28% (95% CI: 90.76% to 99.96%) Specificity 93.62% (95% CI 82.46% to 98.66%) PPV 95% (95% CI: 86.40% to 98.27%) NPV 97.78% (95% CI: 86.29% to 99.68%) Accuracy 96.19% (95% CI: 90.53% to 98.95%) Regarding the validation of mcr‐1 and bla oxa‐48‐like/bla NDM using spiked hemoculture samples, the duplex RPA‐LFA achieved a sensitivity and specificity of 85.71% [CI 67.33% to 95.97%] and 97.87% [CI 88.71% to 99.95%] for mcr‐1 detection (Table 4). For bla oxa‐48‐like /bla NDM, the assay achieved a sensitivity of 90.0% [CI 73.47% to 97.89%] and a specificity of 100% [CI 92.45% to 100%] (Table 4). The accuracy of duplex RPA‐LFA for mcr‐1 and bla oxa‐48‐like /bla NDM was 93.33% and 96.10%, respectively. TABLE 4Validation of duplex RPA‐LFA with the phenotypic antimicrobial susceptibility testing for mcr‐1 and bla oxa‐48‐like /bla NDM detection in spiked hemoculture sample.Phenotypic antimicrobial susceptibility testingDuplex RPA‐LFADuplex‐RPA‐LFA efficiency for mcr‐1 detection mcr‐1+ mcr‐1−CoR+244 Sensitivity 85.71% (95% CI: 67.33% to 95.97%) Specificity 97.87% (95% CI: 88.71% to 99.95%) PPV 96% (95% CI: 77.44% to 99.41%) NPV 92% (95% CI: 82.26% to 96.61%) Accuracy 93.33% (95% CI: 85.12% to 97.80%) CoR−146Total2550 bla oxa‐48‐like or bla NDM+ bla oxa‐48‐like or bla NDM−Duplex‐RPA‐LFA efficiency for bla oxa‐48‐like/bla NDM detectionCRE+273 Sensitivity 90% (95% CI: 73.47% to 97.89%) Specificity 100% (95% CI: 92.45% to 100%) PPV 100% (95% CI: 87.23% to 100%) NPV 94% (95% CI: 84.26% to 97.86%) Accuracy 96.10% (95% CI: 89.03% to 99.19%) CRE−047Total2750 Validation of duplex RPA‐LFA with the phenotypic antimicrobial susceptibility testing for mcr‐1 and bla oxa‐48‐like /bla NDM detection in spiked hemoculture sample. Sensitivity 85.71% (95% CI: 67.33% to 95.97%) Specificity 97.87% (95% CI: 88.71% to 99.95%) PPV 96% (95% CI: 77.44% to 99.41%) NPV 92% (95% CI: 82.26% to 96.61%) Accuracy 93.33% (95% CI: 85.12% to 97.80%) Sensitivity 90% (95% CI: 73.47% to 97.89%) Specificity 100% (95% CI: 92.45% to 100%) PPV 100% (95% CI: 87.23% to 100%) NPV 94% (95% CI: 84.26% to 97.86%) Accuracy 96.10% (95% CI: 89.03% to 99.19%) Discussion Bloodstream infections (BSIs) caused by antibiotic‐resistant bacteria are a significant risk for increased mortality rate, prolonged hospital stays, increased intensive care unit (ICU) admission, and escalating healthcare costs [22]. This clinical challenge requires rapid molecular diagnostic techniques to improve therapeutic efficiency, reduce antimicrobial resistance, and enhance infection control [23]. In addition, diagnostic tests should ideally meet the ASSURED criteria (Affordable, Sensitive, Specific, User‐friendly, Rapid and Robust, Equipment‐free, and Deliverable to end‐users) to be effective for point‐of‐care testing or field applications [24]. In this study, we developed two duplex RPA‐LFA assays for the simultaneous detection of bla CTX‐M /mcr‐1 and bla oxa‐48‐like/bla NDM in BSIs. These visual diagnostic tests target critical antimicrobial drug resistance genes in E. coli and K. pneumoniae , which are WHO priority pathogens. The 2024 WHO Bacterial Priority Pathogens has listed carbapenem‐resistant K. pneumoniae (CRKP) as the priority pathogen, followed by third‐generation cephalosporin‐resistant E. coli [25]. Furthermore, data from the WHO Global Antimicrobial Resistance and Use Surveillance System (GLASS) demonstrated a statistically significant correlation between high consumption of beta‐lactams, cephalosporins, and fluoroquinolones and the prevalence of MDR E. coli and K. pneumoniae in BSIs [26]. Multiplex diagnostics, which allow for the simultaneous testing of a single clinical sample for multiple pathogens or antimicrobial resistance genes, are useful for serious infections such as BSIs. By increasing the diagnostic output per sample input, multiplexing offers superior efficiency compared to single‐plex detection [27]. Our duplex RPA‐LFA assay successfully amplifies and detects two distinct gene pairs, bla CTX‐M/mcr‐1 and bla oxa‐48‐like/bla NDM, in separate reaction tubes. These duplex RPA assays effectively overcome the limitations of a single‐tube multiplex amplification assay, where non‐specific interactions among multiple oligonucleotides frequently occur [27]. The optimum primer concentrations, temperature, and incubation time in our duplex RPA were in the recommended ranges, which are 0.15–0.6 mM each primer, 750–2000 nM of total primer concentration, 36°C–39°C, 20–30 min, respectively [27, 28]. Several rapid and visual detection methods have been integrated with RPA, such as SYBR green I [29], LFA [19], and polymer flocculation sedimentation [30]. Among these, LFA is the most practical for point‐of‐care or onsite field testing due to its rapid detection, simplicity, low cost, and portability [28]. The principle of RPA‐LFA used involves labelling the amplicons with FAM or DIG on one strand, and biotin on the other. The labelled RPA amplicons react with anti‐biotin conjugated colloidal gold nanoparticles (AuNPs) and form a complex at the conjugate pad. As these complexes flow, they are captured by anti‐FAM or anti‐DIG on the test lines. The accumulation of AuNPs results in a visible red line, indicating a positive result [19, 31]. RPA‐LFA provides results within 30–45 min without sophisticated instruments needed (20–30 min for amplification and 10–15 min for LFA detection). Our in‐house LFA, which is capable of detecting two targets with internal control, is estimated to cost approximately 3–5$/test. This represents a significant cost‐effective compared to commercial LFA (~35$/test) [32]. Additionally, the total cost of RPA‐LFA is expected to decrease further as the demand for RPA kits increases and the original patents for the technology expire, potentially allowing for more competitive pricing and wider availability of reagents [33]. The detection efficiency of our duplex RPA‐LFA was evaluated by determining the LOD. In spiked blood samples, the LOD is significantly influenced by the enrichment time. High initial bacterial load (108 CFU/mL) could be directly detected; however, initial bacterial load as low as 102–103 CFU/mL could be detected following a 6–8 h additional incubation. This requirement for enrichment is consistent with the conventional blood culture methods, which require significant incubation time to generate positive signals in automated instrumentation. Similarly, a previous study has shown that 92% of cases were accurately identified by PCR and pyrosequencing after an enrichment time of 5.8 ± 2.9 h [34]. The sensitivity of these assays can be further enhanced by removal of human DNA from blood samples, which has allowed real‐time PCR to detect bacterial load as low as 10 CFU/mL of E. coli [35]. Furthermore, the pathogen enrichment step prior to DNA extraction, including centrifugation, filtration, and microfluidic systems, has been found to enhance sensitivity while depleting potential PCR inhibitors [36]. The LOD for RPA‐LFA varies depending on the target pathogens [28]. Other studies have demonstrated that RPA‐LFA assays can detect S. aureus , K. pneumoniae , P. aeruginosa, and H. influenzae at LODs of 600 CFU/mL, 60 CFU/mL, and 6 CFU/mL, respectively, in spiked blood samples. In comparison, the LOD for real‐time PCR in similar conditions ranged from 6 × 103 to 60 CFU/mL [17]. These results indicate that our duplex RPA‐LFA offers a level of sensitivity that is highly comparable to other platforms. The sensitivity and specificity of our duplex RPA‐LFA ranged from 85% to 98% and 94% to 100%, respectively, depending on the target antimicrobial‐resistant genes. These values are slightly lower than those reported in our previous studies, which may be caused by the complexity of DNA extraction directly from whole blood compared to isolated colonies. In our previous study, the detection of the mcr‐1 gene from pure E. coli colonies isolated from stool samples achieved 100% sensitivity and specificity using RPA followed by in‐house LFA [19]. Similarly, a multiplex RPA combined with a commercial LFA, single‐stranded tag hybridization chromatographic printed‐array strip (STH‐PAS), achieved 99.2%–100% sensitivity and 100% specificity for the detection of four common ESBL genes from E. coli colonies [37]. The high clinical performance of RPA‐LFA is further supported by a study demonstrating the detection of methicillin‐resistant S. aureus in positive blood cultures with 100% sensitivity and 97.7% specificity [38]. A critical technical factor that possibly influenced the performance of this assay is the constant shaking throughout the incubation period. It has been shown to enhance the reaction rate and improve overall sensitivity [27]. False‐positive results in the RPA‐LFA for detecting antimicrobial resistance genes may arise from non‐specific amplification products and primer‐dimers that bind to LFA strips, effectively mimicking target gene signals. The high analytical sensitivity of the RPA‐LFA is also prone to cross‐contamination from previous RPA products [39]. Conversely, false‐negative results might result from other resistance mechanisms. Although bla CTX‐M, mcr‐1, bla oxa‐48‐like, or bla NDM genes are the most common in Enterobacteriaceae, other beta‐lactamases belonging to Ambler classes A‐D (e.g., TEM, SHV, IMP, VIM, KPC) and other mcr variants are also geographically distributed in various clinical settings [40, 41]. For example, a study of ESBL‐producing E. coli in Southern Thailand identified bla TEM (14.2%) and bla SHV (5.8%) as the predominant genes following bla CTX‐M (50.8%) [42]. Furthermore, phenotypic resistance to colistin and carbapenems can be driven by intrinsic resistance mechanisms such as lipopolysaccharide (LPS) modifications or efflux pump mutations [43, 44]. Technical factors related to sample preparation, such as residual DNA extraction buffers or components of the bacterial enrichment medium, can lead to both false‐negative and false‐positive results. Cetyltrimethylammonium ammonium bromide (CTAB) DNA extraction buffer has been identified as an RPA inhibitor, potentially leading to decreased sensitivity or false negative results. In contrast, selenite cystine broth promotes primer‐dimer formation, resulting in false‐positives on LFA [27]. Although there is some discordance between phenotypic drug resistance and the duplex RPA‐LFA results, our assay is still promising for point‐of‐care testing in resource‐limited areas. Beyond clinical settings, the prevalence of bla CTX‐M, mcr‐1, and bla NDM is frequently reported in livestock, highlighting the critical need for rapid surveillance tools in a One Health framework [45, 46, 47]. Our findings regarding the high sensitivity of the duplex RPA‐LFA provide a robust alternative for settings with limited laboratory instruments. Furthermore, the emergence of plasmid‐mediated tigecycline resistance (tet(X4)) in porcine isolates represents a significant threat to last‐resort antibiotic efficacy, particularly when co‐carried within high‐risk E. coli lineages [45, 46]. Consequently, the modular nature of the RPA‐LFA platform allows for the future integration of tet(X4) detection, offering a versatile tool for comprehensive AMR monitoring. Conclusion Our duplex RPA‐LFA represents a promising tool for the detection of multiple critical antimicrobial resistance genes (bla CTX‐M, mcr‐1, bla NDM, and bla oxa‐48‐like) in BSIs. The assays demonstrated high sensitivity and specificity, comparable to benchmark methods. Furthermore, the implementation of an enrichment step successfully enhanced the limit of detection, ensuring reliability even at lower initial bacterial loads. The ability to rapidly screen for these critical resistance genes is essential for effective therapeutic management of life‐threatening infections. Beyond BSIs, this duplex RPA‐LFA is potentially adaptable to other clinical specimens, offering a scalable solution to improve precision medicine and reduce the spread of antimicrobial resistance. Funding This research was funded by the Thailand Science Research and Innovation Fund, Chulalongkorn University (HEAF67370051). Ethics Statement The protocol was exemption approved by the Research Ethics Review Committee for Research Involving Human Research Participants, Group I, Chulalongkorn University, Thailand (COA No. 192/68). No patient's personal information was obtained. Conflicts of Interest The authors declare no conflicts of interest. Acknowledgements We thank the Second Century Fund (C2F), Postdoctoral Fellowship, Chulalongkorn University, Bangkok, Thailand. Data Availability Statement The authors confirm that the data supporting the findings of this study are available within the article. References Associated Data Data Availability Statement The authors confirm that the data supporting the findings of this study are available within the article.