Rapid screening for carbapenemase-producing carbapenem-resistant Enterobacterales: clinical implementation of an immunochromatographic test using broth-enriched rectal swabs

ABSTRACT

Carbapenemase-producing carbapenem-resistant Enterobacterales (CP-CRE) poses a public health issue. Rapid detection of CP-CRE colonization is challenging; existing methods are either expensive or time-consuming. We evaluated an immunochromatographic test (ICT) for detecting carbapenemases directly from broth-enriched rectal swabs. One hundred intensive care patients provided 178 pairs of rectal swabs. One swab was tested using the GeneXpert Carba-R PCR assay; the other was inoculated into brain-heart infusion broth. After 4 and 6 h of incubation at 37°C, the broth was tested with the RESIST-5 O.K.N.V.I. ICT for Klebsiella pneumoniae carbapenemase (KPC), New Delhi metallo-β-lactamase (NDM), oxacillinase-48 (OXA-48), Verona integron-encoded metallo-β-lactamase (VIM), and imipenemase (IMP) carbapenemases. Broths were subcultured after overnight incubation, and recovered carbapenem-resistant Enterobacterales isolates were tested using GeneXpert Carba-R PCR. Sensitivity, specificity, and accuracy were calculated in comparison to culture positive for CP-CRE, confirmed by PCR for KPC, NDM, OXA-48, VIM, and IMP. Of the 178 swabs, 60 were culture positive for CP-CRE. After 6 h, the ICT demonstrated a sensitivity of 65%, specificity of 97.5%, and accuracy of 86.8%. Among heavily soiled swabs, sensitivity reached 81.8% for ICT after 6 h, and the specificity was 100%. The mean execution time for carbapenemase detection using ICT was reduced by 60 h compared to culture. The ICT after 6 h incubation offers reduced execution time for detecting CP-CREs. This method may serve as a valuable rapid screening tool, especially in resource-limited settings.

IMPORTANCE

The rapid spread of multidrug-resistant bacteria requires innovative solutions for early detection and prevention measures. In this study, we present a simple protocol for the direct detection of carbapenemases in rectal swabs using an immunochromatographic assay. By optimizing the assay conditions, we achieved rapid and high-accuracy identification of five clinically important carbapenemases. This method can broaden access to rapid CP-CRE detection of fecal colonization—even in laboratories with limited resources—enabling the implementation of faster and more effective infection control measures, potentially reducing the spread of resistance.

INTRODUCTION

Carbapenems are broad-spectrum β-lactam antibiotics often used as last-resort treatments against multidrug-resistant gram-negative bacteria. However, global carbapenem resistance has emerged (1), largely due to carbapenemases, enzymes borne on mobile genetic elements (2). The most commonly encountered carbapenemases identified include Klebsiella pneumoniae carbapenemase (KPC), New Delhi metallo-β-lactamase (NDM), oxacillinase-48 (OXA-48), Verona integron-encoded metallo-β-lactamase (VIM), and imipenemase (IMP) (3–5).

Carbapenemase-producing carbapenem-resistant Enterobacterales (CP-CRE) are a major public health concern due to their rapid spread, severe infections, and limited treatment options, despite new antibiotics (6, 7). Screening for CP-CRE colonization is crucial to control outbreaks, but rapid and accurate detection—especially for low-hydrolysis carbapenemases—remains challenging. An ideal screening test must be sensitive, specific, fast, capable of detecting multiple carbapenemases, and cost-effective (8, 9).

Culture-based methods of carbapenem-resistant Enterobacterales (CRE) colonization screening, such as direct MacConkey or chromogenic agars, are affordable but time-consuming and labor intensive, detecting only resistance without identifying carbapenemase production, which requires additional testing. Phenotypic methods like the modified carbapenem inactivation method are slow, while faster options like CarbaNP and Blue Carba lack sensitivity and specificity for low-hydrolysis carbapenemases (9–11). Molecular tests offer high sensitivity and specificity, identifying carbapenemase genes even at low expression levels. These include isolate-based tests requiring cultured organisms and direct tests that detect genes directly from clinical samples for faster results, but they are costly and require specialized equipment (12).

Immunochromatographic tests (ICTs) are a rapid diagnostic method that uses antibodies on a test strip to detect specific antigens, providing visual results within minutes. They are commonly used for quick, point-of-care detection of infectious agents and specific proteins directly from patient specimens such as nasal swabs (e.g., severe acute respiratory syndrome coronavirus 2) and urine (e.g., human chorionic gonadotropin). ICTs for identifying KPC, NDM, OXA-48, VIM, and IMP carbapenemases, such as the O.K.N.V.I. RESIST-5 ICT (Coris Bioconcept, Gembloux, Belgium), offer a fast, more affordable option but are only cleared for use with bacterial isolates. Our group developed a protocol using ICT to test broth from enriched mock rectal swabs, demonstrating good in vitro sensitivity and specificity (13). The objective of this study was to evaluate the performance of the ICT on rectal swabs from hospitalized patients, following enrichment with broth at different incubation times.

MATERIALS AND METHODS

Study setting and design

The Instituto Central of the Hospital das Clínicas (ICHC) of the Faculdade de Medicina da Universidade de São Paulo, Brazil, is a 906-bed tertiary care hospital, including 100 intensive care unit (ICU) beds. Routine screening for CRE has been in place since 2014. This protocol involves collecting rectal swabs for culture at ICU admission and weekly thereafter, with patients placed under contact precautions until screening results are available. We conducted a clinical validation study to evaluate an approach for using an ICT to detect carbapenemases (i.e., KPC, NDM, OXA-48, VIM, and IMP) from broth-enriched rectal swabs in patients admitted to all ICUs at ICHC from June to October 2023.

Specimen collection and processing

Dual rectal swabs (Copan with Stuart medium) from each patient in the study ICUs were collected at admission (or within 48 h of admission) and weekly throughout ICU stay. Collected swabs were immediately sent to the laboratory for processing. Upon receipt in the laboratory, each swab was visually inspected for the presence of fecal matter, which was used as an indicator of adequate sample quality. Swabs without visible fecal material were classified as “clean” and rejected from further analysis. Only paired swabs with visible fecal content, defined as the presence of brownish discoloration or particulate matter, were deemed “soiled” and accepted for testing. Soiling was categorized into lightly or heavily soiled using a visual scale Fig. S1 at https://doi.org/10.6084/m9.figshare.29594543.v1.

To ensure consistency, rectal swabs were evaluated by trained laboratory personnel, with second opinions obtained if needed before final determination. Swabs determined to be clean were rejected from the study and not included in the final analyses.

Laboratory testing

Of the included paired swabs, one was tested for carbapenemase genes (bla_KPC, bla_OXA-48, bla_NDM, bla_IMP, and bla_VIM) using the GeneXpert Carba-R assay (Cepheid, Sunnyvale, CA, USA) according to the manufacturer’s instructions. The other swab was inoculated into 10  mL of brain-heart infusion (BHI) broth and incubated at 37°C for 4 and 6 h in ambient air, as validated in a previous protocol (13).

After each incubation period, we centrifuged 2 mL aliquots of the enrichment broth for 5 min at 11,000 × g. The supernatant was discarded; the pellet was mixed with 30 µL of lysis buffer C and eight drops of lysis buffer A; 100 µL of this suspension was added to each O.K.N. and V.I. cassette (“K-Set”) from the O.K.N.V.I. RESIST-5 kit. We called this methodology ICT from broth-enriched rectal swabs. Results were read after 15 min by two independent observers. The result was considered positive when the control band and at least one band for one of the five target carbapenemases were visible and negative if only the control band was visible. This approach was previously developed in vitro using mock rectal swabs spiked with samples from the U.S. Centers for Disease Control and Prevention and Food and Drug Administration Antimicrobial Resistance (CDC and FDA AR) Bank and the microbiology lab of the Faculdade de Medicina de São Paulo, containing different variants of the five most common carbapenemases (KPC, NDM, OXA-48, IMP, and VIM).

After 24 h of incubation, according to local protocol, all enrichment broths were subcultured onto Mueller-Hinton agar plates with imipenem 10 µg, ertapenem 10 µg, and meropenem 10 µg disks. After overnight incubation, isolates growing within the inhibition zones (ertapenem ≤21 mm, meropenem and imipenem ≤22 mm) were considered carbapenem resistant (14), identified using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, and tested for carbapenemase genes using the GeneXpert Carba-R assay. This method was used to define CP-CRE for this study and was considered the “gold standard” for comparisons.

Quality control of the ICT was performed each day that patient samples were tested, including one negative control and one positive control for each carbapenemase using control strains from the CDC and FDA AR Bank. Quality control of the Cepheid GeneXpert Carba-R was performed for each carbapenemase and with a negative control weekly according to established hospital laboratory protocols. The execution time for each method was measured from the moment of specimen arrival at the laboratory until the completion of the test.

Statistical analysis

Each method (direct Carba-R, 4 h ICT, and 6 h ICT) was compared to the gold standard, CP-CRE recovery from culture. Sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and accuracy were calculated for each CP-CRE detection method and for each enrichment period. Ninety-five percent confidence intervals were calculated using the exact Clopper-Pearson methodology or standard logit 95% confidence intervals (15), when appropriate.

A sensitivity analysis was conducted to evaluate the impact of swab content levels on ICT performance. Rectal swabs were classified as heavily or lightly soiled based on visual assessment of fecal content, including 26 heavily soiled and 152 lightly soiled swabs. The positivity rates for each group were compared using a chi-squared test. Sensitivity, specificity, NPV, PPV, and accuracy of the ICT, with 95% confidence intervals, were calculated for both groups after 4 and 6 h of incubation to assess differences in performance. The execution time of each test was compared to each other. Statistical analysis was performed using STATA version 17 (StataCorp, College Station, TX, USA).

RESULTS

We included 100 patients and 222 rectal swab pairs in the study. Six (6%) patients signed the informed consent but did not provide any samples. Of the 222 swabs collected, 44 (19.8%) swabs had insufficient collection (“clean swabs”) and were excluded from further analyses per study protocol. As a result, 178 pairs of soiled rectal swabs were tested. Among the 94 patients included in the analysis, 30 (32%) had CP-CRE colonization.

We recovered 61 CP-CRE from culture, including 57 Klebsiella pneumoniae, 1 Enterobacter cloacae, 2 Escherichia coli, and 1 Serratia marcescens. We also recovered three carbapenem-resistant VIM-producing Pseudomonas aeruginosa, which were not included in the comparison analyses. PCR testing yielded 56 KPC and 4 KPC/NDM co-detections. Isolates were recovered from culture for all ICT and/or PCR positive swabs with the exception of one swab that was PCR positive for NDM but was ICT negative and culture negative.

After 4 h of incubation, the ICT detected KPC enzymes in 34 broth aliquots and both KPC and NDM enzymes in 1 aliquot. Following 6 h of incubation, the ICT detected KPC enzymes in 41 aliquots, and both KPC and NDM enzymes in 1 aliquot. The direct Carba-R test detected KPC genes in 60 swabs, NDM genes in 1 swab, both KPC and NDM genes in 4 swabs, and VIM genes in 3 swabs. No OXA-48 or IMP genes or enzymes were detected by any method.

In comparison with the gold standard, the ICT demonstrated variable sensitivity, depending on the incubation time. The ICT with 4 h of incubation had a sensitivity of 56.7%, a specificity of 99.2%, and an accuracy of 85.1%. Increasing the incubation time to 6 h increased the sensitivity to 65% and the accuracy to 86.8%, while the specificity was slightly lower (97.5%). The NPV and PPV for the 6 h ICT were 85.0% and 92.7%, respectively. These ICT false positives were observed in two isolates that were not phenotypically resistant to carbapenems but tested positive for the KPC gene by PCR. In contrast, the direct Carba-R test had a sensitivity of 88.3%, a specificity of 87.4%, and an accuracy of 87.7% (Table 1).

TABLE 1.

Performance of colonization screening methods for carbapenemase-producing and carbapenem-resistant organisms compared to the “gold standard”^a

Parameters	Result for
	ICT 4 h incubation	ICT 6 h incubation	GeneXpert Carba-R PCR
Sensitivity (%) (95% CI)	56.7 (43.2–69.4)	65.0 (51.6–76.9)	88.3 (77.4–95.2)
Specificity (%) (95% CI)	99.2 (95.41–99.9)	97.5 (92.8–99.5)	87.4 (80.1–92.8)
Negative predictive value (%) (95% CI)	82.3 (77.7–86.1)	85.0 (80.0–88.9)	93.8 (88.3–96.8)
Positive predictive value (%) (95% CI)	97.1 (82.3–99.6)	92.7 (80.4–97.5)	77.5 (68.1–84.8)
Accuracy (%) (95% CI)	85.1 (79.1–90.0)	86.8 (80.9–91.4)	87.7 (82.0–92.1)

^a The gold standard method was broth-enriched CRE screening culture followed by carbapenemase PCR on recovered isolates. CI, confidence interval; ICT, immunochromatographic test; PCR, polymerase chain reaction.

When analyzing the CP-CRE detection according to the degree of soiling of the swab, heavily soiled swabs were positive in 42.3% (11 out of 26) of cases, while lightly soiled swabs were positive in 32.2% (49 out of 152) of cases. The ICT demonstrated better performance in heavily soiled swabs. In this group, sensitivity and NPV reached 72.7% and 88.2% after 4 h of incubation, increasing to 81.8% and 91.8% after 6 h. In contrast, among lightly soiled swabs, the sensitivity was lower: 53.1% with 4 h of incubation and 61.2% with 6 h. Notably, specificity and PPV were 100% in heavily soiled swabs at both incubation times. The overall accuracy of the ICT was 91% at 4 h and 94% at 6 h. The mean execution times for CP-CRE detection by direct Carba-R, 4 and 6 h ICT, and culture for CP-CRE were 2.0, 4.5, 6.5, and 66.0 h, respectively.

DISCUSSION

In this study, we demonstrated that the combination of the broth enrichment method and the O.K.N.V.I. RESIST-5 ICT offers good sensitivity and high specificity for CP-CRE detection in rectal swabs after 4 and 6 h incubation periods. Sensitivity was highest with the longer incubation time, reaching 65%, and improved with heavier fecal content on the swab, reaching 81.8%. The diagnostic accuracy of the ICT was comparable to that of the Carba-R assay across both incubation periods; however, in the subset of heavily soiled swabs, the ICT demonstrated superior accuracy. These findings suggest that both extending the incubation time and increasing fecal content on the swab enhance the test’s ability to detect CP-CRE, particularly in samples with lower bacterial loads. Factors such as stool inoculum size, sample quality, and bacterial expression levels can significantly impact the accuracy of ICTs. Vasilakopoulou et al. noted that “poor” rectal swabs with low fecal content required longer incubation to detect carbapenemases, while turbid samples were positive without incubation (16). In our study, although the overall positivity rates were similar between heavily soiled swabs and lightly soiled swabs, the performance of the ICT improved markedly with heavier soiling. After 6 h of incubation, the sensitivity for heavily soiled swabs increased to 81.8%, whereas lightly soiled swabs showed a sensitivity of 61.2%. This suggests that adequate fecal content on the rectal swab mitigates the sensitivity limitations of ICTs, allowing them to perform effectively. This underscores the importance of proper swab collection techniques and ensuring sufficient sample volume to maximize the sensitivity and accuracy of ICTs.

Our findings differ from those reported by other researchers. Fauconnier et al. incubated 149 rectal swabs in trypticase soy broth containing meropenem (0.25 mg/L) for 2.5 h, then concentrated and tested the broth using the O.K.N. K-Set ICT (Coris Bioconcept). They reported an overall sensitivity of 96% for detecting OXA-48, KPC, and NDM enzymes, using multiplex PCR as the reference standard. The majority of strains evaluated in their study were OXA-48 producers (17). Similarly, Wang et al. incubated 100 CP-CRE-positive rectal swabs in lysogeny broth for up to 4 h, then concentrated and tested the broth using the NG-Test CARBA 5. After 4 h of incubation, they achieved sensitivity and specificity values of 96% and 100%, respectively, for the five enzymes tested by PCR (18). Interestingly, their sample set included a higher proportion of NDM-producing strains, whereas in our study, KPC was the most prevalent carbapenemase.

The discrepancies between our findings and those of other studies may stem from different methodologies, including the type of enrichment broth, absence of antibiotics in our protocol, and variations in incubation times. Additionally, gene prevalence among populations studied is an important factor: our sample had a particularly high prevalence of KPC, which may also contribute to differences in findings. A study done at the same hospital in 2016 among patients admitted to the emergency departments or ICUs found a prevalence of 95.6% KPC and 4.4% NDM, respectively, on carbapenem-resistant organism (CPO)-positive rectal swabs using Xpert Carba-R assay (19).

When comparing the specificity of ICT and direct Carba-R to the gold standard results, the ICT demonstrated excellent specificity, while the PCR showed lower specificity. The reduced specificity of PCR may reflect over-detection or high sensitivity inherent to molecular methods, possibly due to the detection of carbapenemase genes in non-viable or non-cultivable microorganisms or in minority bacterial populations of unclear clinical significance. In only one case, an NDM-positive sample was detected with direct Carba-R but missed by ICT. The “lightly” soiled rectal swab showed no growth in culture, suggesting that the positive result in Carba-R may have been due to the detection of DNA fragments rather than live bacteria (17, 18).

The ICT’s quick execution time and ease of use after 6 h of incubation have important practical applications. The rapid and accurate identification of carbapenemase-producing organism carriers is crucial for infection prevention and control. The ICT’s sensitivity and negative predictive value make it a promising screening method because it allows the majority of individuals colonized with CP-CRE to be identified, with a low likelihood of false-negative results. The ICT’s high specificity ensures that positive results are reliable, minimizing false positives that could lead to unnecessary isolation or treatment. The reduced execution time allows healthcare facilities to implement timely interventions, such as patient isolation and contact precautions, to prevent the spread of resistant organisms. During outbreak situations, the ICT can serve as a frontline screening tool to rapidly identify colonized patients, facilitating swift infection prevention responses.

The ICT’s ability to deliver quick results without the need for specialized equipment makes it suitable for widespread use, including in resource-limited settings, possibly being the only method available for rapid carbapenemase detection when PCR is not available.

While direct Carba-R testing showed higher sensitivity, its higher cost and need for specialized equipment may limit its accessibility. In resource-limited settings, the availability of cartridges and, more critically, access to the shared Cepheid Systems console can create testing backlogs, even when cartridges are on hand. Our protocol using the ICT after a 6 h incubation presents a viable alternative that balances accuracy with practicality. However, practical implementation in routine laboratory settings may face challenges. In this study, samples were processed by dedicated staff within a single shift, contributing to the reduced execution time. In real-world scenarios, particularly in labs without a second microbiology shift, turnaround time for results could be impacted. For example, swabs from patients admitted after the morning cutoff would need to wait until the next day for ICT processing to avoid exceeding the 6 h incubation window. Meanwhile, those same samples could be set up for culture or GeneXpert testing on the same day, potentially reducing the difference in execution time between culture and ICT results. While the ICT method has potential, the 6 h incubation period may limit its practicality in some settings, especially where immediate processing is not always feasible.

This study has several limitations. A significant limitation is the small sample size of only 100 patients and 178 paired swabs. As the sample size increases, the results may vary, potentially affecting the accuracy and generalizability of the findings. In particular, the improved sensitivity observed with heavily soiled swabs was based on a limited number of samples. Therefore, larger studies are needed to confirm this effect and to determine whether the increased sensitivity is consistent across different populations and settings. Nonetheless, demonstrating this potential improvement is important, as it emphasizes the significant impact that sample quality and collection techniques can have on test performance.

We did not collect any rectal swabs positive for OXA-48-like or IMP carbapenemases, and very few swabs were positive for NDM and VIM enzymes. As a result, our performance evaluation of the O.K.N.V.I RESIST-5 ICT is primarily limited to KPC enzymes and one combination of carbapenemases (KPC/NDM). This limitation restricts the generalizability of our findings to other settings where different carbapenemases, such as OXA-48 or IMP, may predominate.

We only evaluated two incubation periods (4 and 6 h) with a single enrichment broth that did not contain carbapenem antibiotics. Different incubation times, types of enrichment media, or selective antibiotics might affect bacterial growth and enzyme expression, potentially influencing the test’s performance. We explored these variables in our previous in vitro study and, to some extent, in this current study. However, the relationship between these factors and test performance is complex and not entirely straightforward. Therefore, while we have investigated these aspects, further research is needed to fully understand their impact. Furthermore, ICT only detects the enzymes included in the test design and does not identify CPO species, thus limiting its utility in epidemiological investigation.

The ICT after a 6 h BHI incubation period emerges as a valuable tool for CP-CRE detection due to its good sensitivity and specificity and rapid execution time, making it particularly useful for quick responses. The improved sensitivity observed with heavily soiled swabs suggests that optimizing sample collection could enhance ICT performance. Evaluation of this protocol with a larger number of patients and specimens across different hospitals will be important. Nevertheless, in resource-limited settings, our direct ICT protocol offers a faster and more accessible alternative compared to traditional culture and PCR methods, respectively.

ETHICS APPROVAL

The study protocol was reviewed and approved by the institutional review boards of Hospital das Clínicas (CAAE 47011921.9.0000.0068), the Centers for Disease Control and Prevention (0900f3eb8208d8e8), and Johns Hopkins (20294). All patients and/or their legal representatives were fully informed about the procedures, including potential benefits and risks, and provided written informed consent.