Evidence review

Clinical and technical evidence

Regulatory bodies

A search of the Medicines and Healthcare Products Regulatory Agency (MHRA) website revealed no manufacturer Field Safety Notices or Medical Device Alerts for this device.

Clinical evidence

A literature search identified 2 published studies that used the Xpert Carba‑R test: 1 comparative diagnostic accuracy study (Findlay et al. 2015) and 1 non‑comparative diagnostic accuracy study (Anandan et al. 2015). Another non‑comparative diagnostic accuracy study using Xpert MDRO, an earlier version of the test (Tenover et al. 2013), is also included in this briefing. The Xpert MDRO was designed to detect only 3 of the 5 major carbapenemase gene families (KPC, NDM and VIM) and so this study is only considered relevant to the technical performance of the test in identifying these gene families. Detection capabilities for the IMP and OXA gene families were added to the later generation Carba‑R test.

The comparative diagnostic accuracy study by Findlay et al. (2015) aimed to assess the performance of 3 commercial molecular assays for detecting carbapenemase genes (including the 5 major carbapenemase families: KPC, NDM, VIM, IMP, and OXA) in pure bacterial isolates. A panel of 450 carbapenem resistant isolates were tested using the Check‑Direct CPE kit (on 2 different platforms), the eazyplex SuperBug complete A kit and the Xpert Carba‑R kit. The carbapenemase genes had been previously detected and identified using in‑house PCR assays, a commercial microarray (Check‑MDR CT102) or a combination of both tests. These data were regarded as the gold standard against which the commercial assays were compared. The panel of 450 isolates comprised 100 samples each of KPC, NDM, VIM and OXA‑48 variants, 2 isolates co‑producing NDM and OXA‑48 variants, 24 IMP producers and 24 isolates that were carbapenem‑resistant, but did not contain a known carbapenemase gene. The overall test sensitivity was reported as 94.3% (402/426). The Xpert Carba‑R test successfully detected the correct carbapenemase gene in all 302 isolates with a KPC, NDM or VIM enzyme and so achieved 100% sensitivity for these targets. In those isolates with an OXA‑48 variant carbapenemase gene (n=102), 18 of them (17 with the OXA‑48 variant alone and 1 with an OXA‑48‑variant co‑produced with NDM) were not detected by Xpert Carba‑R. PCR and sequencing identified an OXA‑181 gene in each of the 18 false‑negative isolates. A modified Xpert Carba‑R version 2 kit was subsequently provided and correctly identified the 18 OXA‑181 producers. The Xpert Carba‑R test detected 71% (17/24) IMP producers; however, PCR confirmed the presence of a IMP gene in the 7 false negative isolates that were closely matching IMP‑4, IMP‑7, IMP‑8, IMP‑13 and IMP‑14, which Xpert Carba‑R does not cover. The authors concluded that the commercial tests offer a reliable means of detecting bacteria with clinically significant carbapenemases. A summary of the study and results can be found in tables 1 and 2 of the appendix.

The diagnostic accuracy study by Tenover et al. (2013) aimed to determine the sensitivity and specificity of Xpert MDRO, an earlier version of the Xpert Carba‑R test, by comparing it with a culture method with and without a broth enrichment step. The authors collected 328 clinical specimens from human rectal, perirectal and stool samples. The presence of carbapenemase genes was confirmed using culture on a MacConkey agar plate followed by Check‑Points microarray detection. For a control test, 41 bacterial isolates containing a variety of carbapenemases were tested with the Xpert MDRO cartridge, which gave positive results for the KPC, NDM and VIM genes and no false positives. For the 328 clinical samples, the sensitivity, specificity, positive predictive value and negative predictive value for the VIM gene were reported as 100%, 99.4%, 81.8% and 100% respectively. The sensitivity, specificity, positive predictive value and negative predictive value for the KPC gene were reported as 100%, 99.0%, 93.0% and 100% respectively. False‑positives were observed in 2 and 3 cases for VIM and KPC genes respectively. None of the 328 specimens contained the NDM gene. Therefore, 66 contrived stool samples were prepared using various dilutions of 3 Klebsiella pneumoniae isolates containing NDM. The experiments were done in 5 replicates for the 2 lowest dilutions of 150 and 300 colony forming unit/ml (CFU/ml) and 4 replicates for the 3 highest dilutions of 600, 1200 and 1800 CFU/ml. Xpert MDRO showed 100% positivity at dilutions from 300 CFU/ml to 1800 CFU/ml and 93.3% at 150 CFU/ml. The authors concluded that the Xpert MDRO test can detect the KPC, NDM and VIM carbapenemase genes directly from rectal swab samples. They considered that Xpert MDRO could provide valuable information for infection control programs designed to limit the spread of multi‑drug resistant organisms in healthcare settings (tables 3 and 4).

The diagnostic accuracy study by Anandan et al. (2015) evaluated the performance of Xpert Carba‑R using clinical isolates and faecal specimens when compared with conventional multiplex PCR. One hundred and twenty clinical isolates of carbapenem resistant E. coli (n=32) and K. pneumoniae (n=88) were collected from people with bloodstream infections. These isolates were concurrently investigated for the 5 clinically relevant carbapenemase coding genes KPC, NDM, IMP, VIM and OXA‑48, using conventional multiplex PCR. Additionally, 26 of the faecal specimens were tested for the presence of carbapenemase genes using the Xpert Carba‑R test. Conventional PCR identified NDM in 40% (48/120) of isolates, OXA‑48 variants in 39.2% (47/120) of isolates and co‑producers of NDM and OXA‑48 variants in 12.5% (15/120) of isolates. Notably, 8.3% (10/120) of isolates were negative for all 5 tested genes, and all tested isolates were negative for IMP, VIM and KPC genes. The Xpert Carba‑R test identified NDM in 55% (66/120) isolates, but like the Findlay et al. study (2015) it did not identify either OXA‑48 variants or co‑producers of OXA‑48 variants and NDM in the isolates. The identified OXA‑48 variant was found to be OXA‑181 by sequencing. The authors reported the sensitivity, specificity, positive predictive value and negative predictive value of the Xpert Carba‑R test to be 100%, 77%, 96% and 100% respectively (the OXA‑48 variant results were excluded from these figures, owing to the missed OXA‑181 target). Of the tested faecal samples, 46% (12/26) contained carbapenem‑resistance genes, 9 had NDM, 2 had both NDM and VIM and 1 had both NDM and KPC. The authors concluded that the Xpert Carba‑R test would be useful for the prompt detection of people infected or colonised with strains of bacteria that may harbour carbapenemase‑encoded genes. However, they highlighted that incorporating additional OXA‑48 variant specific sequences in the panel may help to improve its sensitivity and maximise the coverage of the assay (tables 5 and 6).

Recent and ongoing studies

No ongoing or in‑development trials on Xpert Carba‑R for detecting CPOs were identified.

Costs and resource consequences

Individual Xpert Carba‑R tests and the GeneXpert system are considerably more expensive than current microbiological culture techniques and would therefore represent an additional acquisition cost to the NHS. However, the faster turnaround time could lead to quicker detection, faster infection control implementation and faster optimal treatment. This may in turn lower the risk of spread and reduce overall costs such as unnecessary bed days in isolation. Xpert Carba‑R would be used in place of standard culture‑based tests and as an adjunct to supplementary and confirmatory tests such as antimicrobial resistance testing.

One conference abstract with limited evidence on resource consequences for Xpert Carba‑R was identified. Delbarre et al. (2014) reported in a French study that the cost of reagent plus technical time was $54.60 (about £35) per test, compared with $116.10 (about £75) for the culture plus PCR. The results were available in 48 minutes compared with at least 30 hours for the culture plus PCR. No specialised staff were needed. No further detail is available and these figures should therefore be interpreted with caution.

Strengths and limitations of the evidence

The available evidence on the use of Xpert Carba‑R to detect CPOs is currently limited in quantity and quality. Three diagnostic accuracy studies, 1 of which compared Xpert Carba‑R to alternative rapid identification methods, were identified. No studies were identified that reported clinical or healthcare‑related outcomes.

The comparative diagnostic accuracy study by Findlay et al. (2015) compared the performance of 3 commercial assays for detecting carbapenemases in bacterial isolates. One of the main strengths of this study was the large panel of bacterial isolates used (n=450), with carbapenemase resistance mechanisms defined through appropriate methods. The 24 samples which were carbapenem‑resistant through other mechanisms and were therefore negative for CPO genes could have allowed for a calculation of specificity. This was not reported and the published results were limited to sensitivities and false negatives. The Xpert Carba‑R test was able to identify carbapenemase genes in 402 of 426 isolates (overall sensitivity of 94.3%). The isolates were selected for geographical, temporal and carbapenemase diversity; however, because of the nature of diagnostic accuracy studies, there is no risk of selection bias. The study was done in the UK at the Antimicrobial Resistance and Healthcare Associated Infections (AMRHAI) Reference Unit and so results would be generalisable to the NHS. Due to the selection of isolates for carbapenemase diversity, the distribution of the carbapenemases observed in this study would not mimic the natural distribution of carbapenemases. The original Xpert Carba‑R assay did not detect OXA‑48 variant genes but the authors later acquired a modified version to detect OXA‑181 genes.

The diagnostic accuracy study by Tenover et al. (2013) evaluated the Xpert MDRO test, an earlier version of Xpert Carba‑R, and therefore could only report on the performance characteristics for 3 of the 5 major carbapenemase gene families (KPC, NDM and VIM). The authors stated that they chose a broth enrichment culture and sequencing of the target genes as the reference standard. However, 2 of the 3 samples that tested positive for the KPC gene using Xpert MDRO had been classified as negative by broth culture but positive by sequencing, and the other sample had not been sequenced at all. As the authors stated there were only 40 true positives, it is evident that the status of these 3 samples had only been based on broth culture alone, which is not a suitable reference standard. In addition, the reported specificities and negative predictive values for both VIM and KPC genes could not be independently replicated from the results presented in the published paper, raising concerns of reporting bias in this study. The authors highlighted that the number of organisms they used to establish the sensitivity and specificity of the Xpert MDRO assay did not cover the entire range of genes that could be detected in each of the 3 carbapenemase families. However, they observed that by using computer modelling, there were no mismatches between the primers chosen for the assay and any mutations associated within the individual target genes. Additionally, the culture media they used in the evaluation were limited to MacConkey agar and did not include a broad range of agar plates that may have supported the growth of other organisms for the 3 Xpert MDRO‑positive, culture‑negative samples. It is also possible that some organisms lost viability during transport to the central testing laboratory.

The diagnostic accuracy study by Anandan et al. (2015) aimed to evaluate the performance of Xpert Carba‑R using clinical isolates and faecal specimens directly. The authors used conventional multiplex PCR as the reference standard for the diagnostic accuracy study, but did not specify which technology was used. It was also apparent that the Xpert Carba‑R test identified more isolates containing NDM than the reference standard, although the authors made no attempt to explain this discrepancy. It is also not clear whether the 26 faecal samples tested by Xpert Carba‑R were confirmed by a reference standard. The performance characteristics that were reported by the authors could not be independently replicated and it is unclear which raw data were used to calculate these. The authors noted that among the 10 OXA‑48 variants, the Xpert Carba‑R test is designed to only identify 4 variants. They highlighted that this provides an explanation for why the Xpert Carba‑R failed to identify the OXA‑48 variants within the clinical isolates. Overall, the study was very poorly reported and results should be interpreted with caution.