False rifampicin resistant results using Xpert MTB/RIF on urine samples in hospitalised HIV-infected patients

Background A small proportion of false rifampicin resistant results have previously been reported using GeneXpert MTB/RIF version G4 on sputum samples; however, this has not been investigated for urine samples in HIV-associated tuberculosis (TB). Objectives We sought to determine the proportion of false rifampicin resistant results using Xpert MTB/RIF version G4 on urine samples among HIV-infected inpatients investigated for TB. Methods Hospitalised HIV-infected patients undergoing systematic TB testing from two cohorts in Cape Town, South Africa, were enrolled. All patients with ≥1 urine Xpert result available were included. Rifampicin resistant urine Xpert results were classified into three mutually exclusive groups: (1) true rifampicin resistance, (2) false rifampicin resistance or (3) unknown after review of available microbiologic and clinical data. Results Overall, 1171 patients were included, from whom a total of 1704 urine Xpert results were available on unconcentrated and/or concentrated urine samples. There were 416 samples positive for TB (24.4% [95% CI 22.4–26.5]), of which 43/413 (10.4% [95% CI 7.6–13.8]) were rifampicin resistant (after excluding three results that were falsely positive due to contamination). Of 43 rifampicin resistant Xpert results (among 40 patients), 30 were classified as true resistance, 11 as false resistance and 2 could not be classified. Excluding unclassifiable results, 30/41 results were confirmed as true-positive urine Xpert rifampicin resistance (positive predictive value: 73.2% [95% CI 57.1–85.8]). Conclusion Urine Xpert testing showed a high proportion of false rifampicin resistance results. Urine Xpert rifampicin resistant results should be interpreted cautiously and confirmed when possible.


Introduction
Tuberculosis (TB) remains the leading cause of death in people living with HIV, contributing to one-in-three AIDS-related deaths. 1 Timely diagnosis of TB in such patients remains challenging because of non-specific presentations and disseminated disease. 2,3,4 Gene Xpert MTB/RIF, an automated nucleic acid amplification test, is capable of providing results in a few hours and represents an important breakthrough for diagnosing HIV-associated TB. Importantly, Xpert also rapidly detects rifampicin resistance, without need for an additional sample or cartridges. It has been endorsed by the World Health Organization (WHO) since 2010. Sputum Xpert (or Xpert Ultra where available) is currently recommended by the WHO as the initial diagnostic test in patients with suspected HIV-associated TB or multi-drug resistant (MDR) TB. 5 In those with microbiologically confirmed TB, Xpert MTB/RIF is also recommended by the WHO as a first-line assay for the rapid detection of rifampicin-resistance. It is therefore an important tool in tackling the growing global health challenge of drug resistant (DR)-TB. However, the WHO does not currently have a recommendation regarding the use of Xpert MTB/RIF in urine owing to an insufficient amount of data on the performance and utility of this assay in urine specimens. 6 In concordance with WHO guidelines, sputum Xpert was implemented as the initial diagnostic evaluation in those with suspected TB and DR-TB in South Africa as well as other countries 7 and in South Africa, it has now been replaced with the updated Xpert Ultra cartridge. Although http://www.sajhivmed.org.za Open Access Xpert has not been associated with a mortality reduction in most trials to date, 8,9,10 its implementation has been associated with overall shorter times to starting anti-TB therapy, including DR TB. 8,11,12,13 It has also increased the diagnostic yield by 1.4% -15% (compared to sputum microscopy) in clinical trials in Sub-Saharan Africa, Brazil and Indonesia. 8,9,10,11,13,14,15,16 Against the backdrop of improved case detection, previous studies have reported on false rifampicin resistance results associated with the Xpert MTB/RIF assay, and meta-analyses found the overall specificity of the Xpert for rifampicinresistance in sputum samples to be 98% (i.e. 2% showed false rifampicin resistant results) and 99% in extra-pulmonary samples. 17,18 This however appeared to be associated in part with earlier Xpert cartridge generations. 19 An implementation study from South Africa found the Xpert G4 cartridge to have excellent positive predictive value for rifampicin resistance of 99.5% (95% CI 98.5-100) in sputum samples. 20 We have previously found that among HIV-patients requiring acute medical hospitalisation, testing of a single concentrated urine sample detected 2.2 times more TB cases than sputum Xpert testing, largely because of the inability of sick inpatients to produce a sputum sample. 21 Additionally, a recent randomised, multi-country trial found that the addition of rapid urine-based assays (including urine Xpert) to sputum Xpert testing was associated with reduced mortality among hospitalised HIVinfected patients in sub-group analyses. 22 This suggests that urine-based testing using Xpert may have an important role in the TB diagnostic algorithm among hospitalised patients with advanced HIV, especially those too ill to produce a sputum sample. However, the proportion of false-positive rifampicin resistance results using Xpert on urine samples has not been reported. We sought to determine the proportion of urine Xpert false rifampicin resistance results among hospitalised HIV-infected patients being investigated for HIV-associated TB in Cape Town, South Africa.  , that were clinically indicated  and collected by the medical teams were also recorded -for  example, lymph node aspirates, cerebrospinal fluid TB  cultures, pleural TB cultures and urine TB cultures. In the first study, two cases of false urine rifampicin resistance occurred 3 months after study initiation (Appendix Table 1

Laboratory methods
Urine Xpert testing for both studies was performed at the Groote Schuur Hospital National Health Laboratory Service laboratory using Xpert MTB/RIF Assay G4 version 5. All specimens were processed using standardised protocols and quality assurance procedures as previously described. 24 In brief, for the GF Jooste Hospital study, Xpert testing of urine samples was conducted in two ways on each sample. The first method (unconcentrated) utilised 2.0 mL of fresh urine that was centrifuged, resuspended in 0.75 mL phosphate buffer and then tested using Xpert. 25 The second method (concentrated) used a 30 mL -40 mL urine sample that was centrifuged at 3000 g for 15 min. The resultant supernatant was removed and the pellet was resuspended in the residual urine volume (without the addition a phosphate buffer); 0.75 mL was then tested using Xpert. 21 For both methods, Xpert sample reagent (1.5 mL) was added to the samples as per manufacturer's instructions. The Khayelitsha Hospital study only used Xpert testing on concentrated urine samples and was undertaken using the same methods as described above. The reference standard for drug resistance, including rifampicin resistance for both studies, was a molecular line probe assay (MTBDR plus; Hain Lifescience Nehren, Germany) undertaken on culture isolates from any clinical specimen (not necessarily urine).

Analysis
Patient populations were from overlapping referral areas in the Cape Town townships and both cohorts included

Ethical consideration
Approval for both studies was obtained from the University of Cape Town Faculty of Health Sciences Human Research Ethics Committee and patients provided written informed consent according to the approved study protocols.

Results
There were 585 patients from the GF Jooste Hospital cohort and 586 patients from the Khayelitsha Hospital cohort with urine Xpert results available for a total of 1171 hospitalised HIV-infected patients. Overall 1704 urine Xpert results were available from 1171 patients, of which 554 were performed on unprocessed urine samples and 1150 on concentrated urine samples ( Figure 1). Baseline characteristics of the two cohorts were similar. (Table 1).  We were unable to obtain information about the probe features for samples of the JTBS study.
Three patients (n = 3/40, 7.5%) with a confirmed rifampicin resistant urine Xpert result had evidence of likely heteroresistant infection. The first patient (Appendix Table 1-A1 -KDHTB203) cultured a drug susceptible isolate from blood (MycoF/lytic bottle), sputum and urine samples but also a rifampicin resistant isolate from sputum during the same admission. The second patient (KDHTB531) cultured a drug-sensitive isolate from blood as well as a drug-resistant isolate from sputum during the same admission. The third patient (JTBS463) was originally started on drug-sensitive TB treatment after a prior sputum Xpert and abscess aspirate culture both showed rifampicin susceptible isolates. One month after starting TB treatment, the patient was admitted for TB immune reconstitution inflammatory syndrome (IRIS). Shortly after discharge, the patient was readmitted for gastroenteritis and was clinically deteriorating despite drug-sensitive TB treatment. At this time, two urine Xpert results showed rifampicin resistance; however, the patient died shortly after receipt of urine Xpert results.

Discussion
In this study, which included hospitalised HIV-infected patients systematically investigated for TB, the overall proportion of urine Xpert rifampicin resistance results was 10.4% (n = 43/413); however, the positive predictive value of urine Xpert MTB/RIF for rifampicin resistance was only 73.2% (n = 30/41).
The correct identification of drug-resistant TB has important implications for both the individuals' health as well as for public health. For the patient, a false rifampicin resistance result may result in not only over-treatment with more toxic drugs that are less efficacious for drugsensitive TB, but also significantly and unnecessarily prolong treatment times. In high burden, under-resourced settings, a false rifampicin resistance may have important resource implications by resulting in additional drug susceptibility testing, significantly more expensive treatment costs and unnecessary community contact tracing. 26 Thus, any test that detects DR TB should ideally have very high specificity. Under the best-case scenario when results were restricted to those not receiving TB treatment, we found that Xpert testing of rifampicin resistance on urine samples did not achieve sufficiently high positive predictive value (86%) to be the sole/ Xpert Ultra is an updated, next-generation sample cartridge for the Xpert platform that is now recommended by the WHO as a replacement for the current Xpert MTB/RIF cartridge 27,28 and has been implemented in South Africa. It provides increased sensitivity for the detection of MTB in sputum (especially smear-negative and pauci-baciliary disease). Xpert Ultra utilises a new melt curve analysis to detect RIF-resistance; however, its diagnostic accuracy (including specificity) for the detection of rifampicin resistance is similar to that of Xpert. 29 The results of this study suggest that urine Xpert Ultra rifampicin resistance results should be interpreted cautiously and confirmed by alternative drug susceptibility testing (either phenotypic or alternative genotypic assays) until the specificity of Xpert Ultra for rifampicin resistance detection has been confirmed to be adequately high to warrant stand-alone testing on urine samples.
Of interest, in this cohort we describe three patients with a confirmed urine Xpert rifampicin resistance result who also had drug-sensitive strains from independent samples during the same admission suggesting likely heteroresistance (either polyclonal infection or acquired heteroresistance). The prevalence of heteroresistance in MTB infections has previously been described. 30,31,32 Although not well-studied, these are likely associated with increased rates of treatment failure for the individual 31 and could complicate TB control efforts at a population level. Xpert may miss heteroresistance if used as a standalone test for the detection of rifampicin resistance, however, early studies show that Xpert Ultra may detect heteroresistance when the resistant DNA comprises 5% or more of the sample. 28 Strengths of this study include a large number of urine Xpert rifampicin results from two geographically and clinically comparable cohorts where patients were prospectively recruited and underwent systematic testing for TB. Additionally, all TB assays including urine Xpert testing were performed at the same laboratory according to standard protocols. After an error yielded two likely false Xpert rifampicin resistant urine cases due to contamination soon after recruitment initiation, disposable bedpans (single-use) were implemented for the duration of both studies. We therefore recommend that clinicians use single-use specimen collection bedpans and containers when utilising Xpert or Xpert Ultra testing on urine to prevent DNA-crosscontamination between samples.
The reason(s) for the high proportion of false positive urine rifampicin resistance is not entirely clear, but the proportion was higher among those already receiving TB therapy. A limitation of this study is that we did not have data available to systematically evaluate the Xpert probe features associated with our classification of false rifampicin resistance. Different methods of drug susceptibility testing could explain discrepant results in some cases. 33,34 The majority of drug susceptibility testing on cultured isolates in both studies was PCR-based; however, we also captured results of all TB tests performed in-service and cannot reliably differentiate between drug susceptibility testing performed with other methods such as liquid or solid media for all samples for the duration of the study. Because of the early implementation of disposable bedpans, we do not suspect undetected contamination beyond that described above. Furthermore, because most patients with positive urine rifampicin results did not have paired urine culture isolates available for further genotypic or phenotypic drug-susceptibility testing, patients classified as having false positive rifampicin results may have had heteroresistance with compartmentalised true rifampicin-resistant urinary TB and rifampicin-susceptible TB at other anatomic sites. However, the favourable clinical course of most of these patients on first-line drug-sensitive TB treatment counts against this possibility. Notably, a large proportion of false positive rifampicin results were among those already receiving anti-TB therapy, where 50% of urine Xpert rifampicin resistance results were classified as false resistance; this suggests that further caution should be applied when interpreting urine Xpert rifampicin resistance results in treatment-experienced patients.
An additional limitation of the study is that sequencing of isolates was not performed as part of either study. Sequencing of the rpoB gene would have been particularly useful in the cases that we could not classify as true or false resistance and the heteroresistant cases. Furthermore, urine TB cultures were not routinely performed in either study and it may have been useful to compare drug susceptibility results on isolates cultured from urine samples collected at the same time as the urine Xpert samples.
In conclusion, urine testing using Xpert provides important diagnostic yield for hospitalised HIV-infected patients being investigated for HIV-associated TB, especially in those unable to produce sputum samples. Although the overall proportion of patients with urine Xpert rifampicin resistance in this cohort was relatively low, the proportion of those classified as false rifampicin resistance was substantially higher than has previously been reported on sputum. Urine Xpert rifampicin resistant results should therefore be interpreted with caution, repeated on a second sample in patients at lowrisk for drug resistant TB (as currently recommended by the WHO for sputum samples) and confirmed using additional culture-based or molecular assays when possible. Whether these findings apply to Xpert Ultra is an issue that requires further study.

Data availability statement
The data sets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Disclaimer
The views expressed in the article are those of the authors and not an official position of the institution or funder.