Stool Xpert® MTB/RIF test for the diagnosis of childhood pulmonary tuberculosis at primary clinics in Zimbabwe

SUMMARY OBJECTIVE: To evaluate the diagnostic performance of Xpert® MTB/RIF on stool samples from children with clinical suspicion of pulmonary tuberculosis (PTB) at primary care clinics. DESIGN: A cross-sectional diagnostic evaluation enrolling 5–16 year olds from whom one induced sputum (IS) sample was tested for microbiological TB confirmation. Results of a single stool sample tested using Xpert were compared against microbiologically confirmed TB, defined as a positive result on sputum microscopy and/or culture and/or IS Xpert. RESULTS: Of 222 children enrolled, 218 had complete microbiological results. The median age was 10.6 years (interquartile range 8–13). TB was microbiologically confirmed in 19/218 (8.7%) children. Of these, respectively 5 (26%), 9 (47%) and 15 (79%) were smear-, culture- and IS Xpert-positive. Stool Xpert was positive in 13/19 (68%) microbiologically confirmed cases and 4/199 (2%) microbiologically negative cases. Stool Xpert detected 76.9% (10/13) of human immunodeficiency virus (HIV) infected and 50% (3/6) of non-HIV-infected children with microbiologically confirmed TB (P = 0.241). CONCLUSION: Stool Xpert is a potential alternative screening test for children with suspected TB if sputum is unavailable. Strategies to optimise the diagnostic yield of stool Xpert assay need further study.

C O N C L U S I O N : Stool Xpert is a potential alternative screening test for children with suspected TB if sputum is unavailable. Strategies to optimise the diagnostic yield of stool Xpert assay need further study. K E Y W O R D S : GeneXpert; faecal specimen; children CHILDHOOD TUBERCULOSIS (TB) contributes 10% of the total 9 million incident cases worldwide, with an estimated 136 000 deaths among children. 1 The childhood TB burden is largely due to undiagnosed and late diagnosis of adult TB, which creates a reservoir for transmission to children. 2 The principal challenge to timely diagnosis of childhood TB is the difficulty in obtaining spontaneously expectorated sputum, necessitating sputum induction, which requires skill and infrastructure. Furthermore, sputum from children is often paucibacillary, as children are less likely to form cavitary lesions in lungs to contain the bacilli. 2,3 The sensitivity of the standard diagnostic tests, sputum microscopy and culture, is therefore lower. A more easily accessible sample that can be obtained with minimal skill at primary health care level and an alternative diagnostic tool is therefore required for the diagnosis of TB in children.
Mycobacterium tuberculosis DNA can be detected by an automated molecular assay, Xpert w MTB/RIF (Cepheid, Sunnyvale, CA, USA). Automation includes sample processing, nucleic acid extraction, polymerase chain reaction (PCR) and the detection of specific M. tuberculosis codon sequences within a closed system. 4,5 The assay is rapid, producing a result in 3 h, with little hands-on time in sample processing.
Previous studies have shown high sensitivity of Xpert when sputum, 6-8 gastric lavage, 9,10 bronchoalveolar lavage 11 and nasopharyngeal aspirates 12 were used. Xpert was therefore endorsed by the World Health Organization as an initial test for diagnosing TB in children. However, invasive procedures are required to obtain such samples.
Stool is a potential sample that can be collected non-invasively. Children tend to swallow sputum when they cough and M. tuberculosis DNA has been shown to survive the harsh acidic and digestive environment of the gastro-intestinal tract. 13 We therefore evaluated the diagnostic performance of Xpert using stool in children with clinically suspected pulmonary TB (PTB) presenting to primary care clinics in Harare, Zimbabwe. The diagnostic yield of stool Xpert was assessed against confirmed microbiological PTB determined by testing induced sputum (IS) with microscopy, solid culture and Xpert.

Participants
Participants were recruited from eight primary care clinics in Harare between September 2013 and October 2014, with guardian consent and participant assent. Participants were children aged 5-16 years presenting with a chronic cough of .2 weeks and any one of the classic signs and symptoms of TB, including weight loss, loss of appetite, persistent fever without an apparent cause, night sweats or history of close contact with a TB index patient, defined as living in close proximity (sharing a room within a household) with an adult diagnosed with TB within the preceding 12 months.
Those aged 0-5 years were not included. At the time of seeking ethical approval, evidence on the safety and tolerability of sputum induction in children was requested, as sputum induction is rarely practised. The available literature did not meet the satisfaction of the ethical committee to allow the procedure to be performed in children aged ,5 years.
Participants were excluded if they had an already confirmed TB diagnosis, had been on anti-tuberculosis treatment for .72 h before enrolment, required emergency medical attention or were resident outside Harare.

Study procedures
A questionnaire was used to collect data on demographics and physical examination. Human immunodeficiency virus (HIV) status was obtained from documented proof of test result. In case of unknown HIV status, participants were offered the test through the clinic's HIV testing protocols. All participants were asked to provide a stool sample and to undergo sputum induction to obtain one sputum sample.
Sputum induction was performed by a trained nurse following the method described by Zar et al., 14 except where it was contraindicated. 15 The procedure was performed in an open area after a 2-3 h fast. Two doses of 100 lg salbutamol were administered via an aerochamber to prevent bronchoconstriction, followed by nebulisation with 5% hypertonic saline for 5 min, or until the participant started to cough. Oxygen saturation was monitored during and 30 min after the procedure. IS was collected in a sterile container by nasopharyngeal suction, if required. If the participant failed to produce sputum, the procedure was repeated once, 5 min after the first attempt.
The stool sample (about 5 g) was collected into a wide mouthed specimen collection jar on the spot or submitted the following day. Samples were transported at 4-88C to the laboratory within 8 h of collection.
Anti-tuberculosis treatment was commenced by the residing medical officer at the clinic following Zimbabwe's national TB guidelines, in which diagnosis is based on clinical history and examination, chest radiography (CXR) and/or microbiological results. Study test results on microbiological confirmation were made available to the medical officer to assist in participant management.

Laboratory methods
Sputum was processed for microscopy, culture and Xpert testing. It was decontaminated by adding an equal volume of 4% sodium hydroxide for 15 min, neutralised with phosphate buffered saline (PBS) and centrifuged at 3200x g for 20 min to obtain a sediment which was resuspended in 1.5 ml PBS. One drop of the sediment was used to make a smear, which was stained using the Ziehl-Neelsen (ZN) method. All smears were re-read by a second reader blinded to the first reading. Discordant results were resolved through a third reading by a different reader. IS sediment (0.1 ml) was cultured on Löwenstein-Jensen (LJ) slopes at 378C for 8 weeks, with weekly monitoring for growth. At least one colony on LJ was considered to be positive for M. tuberculosis growth, confirmed by ZN stain and speciation using the SD Bioline TB Ag MPT64 antigen rapid test (Standard Diagnostics Inc., Kyonggi, South Korea). Discordant ZN-positive/MPT64 antigen-negative samples were speciated using colony morphology, growth at different temperatures and on para-nitrobenzoic acid-containing LJ slopes. Confirmed cultures underwent phenotypic direct drug susceptibility testing (DST) on LJ slopes for 4 weeks at 378C. For Xpert testing, 0.5 ml of IS sediment was mixed with 1.5 ml Xpert reagent, and the mixture was tested according to the manufacturer's instructions.
Stool samples were processed within 2 days of collection, as described by Nicol et al. 16 An aliquot of 0.15 g was measured into a centrifuge tube using a sterile disposable plastic loop that was left inside the tube; 2.4 ml PBS was then added, and the mixture was vortexed to remove stool particles from the loop. The loop was removed and the sample mixture was left undisturbed at room temperature for 20 min. Two aliquots of 1 ml supernatant were removed. One aliquot was used for immediate testing and another was stored for repeat testing if required. Before testing, the processed stool sample was centrifuged at 3200x g for 15 min; the sediment was then resuspended in 1 ml PBS, mixed with 2 ml Xpert reagent and tested according to the manufacturer's instructions. Interpretation of results was autogenerated by the Xpert system. Personnel carrying out the index test were blinded to sputum test results.

Data analysis
Sample size was based on an anticipated sensitivity of the stool Xpert test of 80%, 10% level of precision and 4% anticipated loss to follow-up. During sample size computation, no data were available on TB prevalence among children in our setting; prevalence was therefore set at 29%, adopted from a community study of paediatric TB contacts in South Africa. 17 Data were analysed using STATA, version 13.0 (StataCorp, College Station, TX, USA). Descriptive statistics were used to characterise the study population. The v 2 test was used to test for associations. Performance of stool Xpert was assessed as the proportion of stool Xpert-positive cases among microbiologically confirmed and clinically diagnosed TB cases. Microbiological TB was defined as a positive result on ZN smear microscopy and/or culture and/or Xpert test on IS. Indeterminate TB status results and samples without complete data on index and reference tests were excluded from the analysis. The study also aimed to evaluate the performance of stool Xpert in TB clinical case definitions as defined by Graham et al. 18 However, the number of children who underwent CXR was too small (n ¼ 24) to provide meaningful classification. According to the Zimbabwe national TB guidelines, a CXR is required for children with clinical suspicion of PTB but without microbiological confirmation; CXR data were therefore dependent on the clinic's operational systems. Ethics

RESULTS
Of the total 222 participants recruited, 4 were excluded from the analysis: 1 participant was withdrawn by the guardian, 2 were unable to provide IS samples and 1 had an IS Xpert test that produced an error result despite repeated testing. The median age of the 218 remaining participants was 10.6 years (interquartile range [IQR] 8-13); 44% were male. The median Z-scores for weight-for-age and heightfor-age were respectively -1.09 (IQR -1.99 to -0.46) and -1.88 (IQR -3.70 to -0.73). HIV status was available for 201 (91%) participants, of whom 111 (51%) were HIV-infected, with 55 (49.5%) undergoing antiretroviral therapy (ART). The distribution of demographic characteristics between microbiological TB and no TB groups is shown in Table 1.

Diagnostic performance of stool Xpert
Stool Xpert was positive in 17/218 (8.3%) children. Among HIV-infected children, stool Xpert was  Table 2). The four children with a stool Xpert but negative IS Xpert and culture were all HIV-infected, had no history of TB and no history of contact with a TB index patient. Two (50%) of the four children received a clinical diagnosis and were the only ones with a positive stool Xpert among 13 clinically diagnosed children. Of these two children, one (50%) had clinical improvement at 2 months of anti-tuberculosis treatment (determined after the study). Of 186 children not commenced on treatment, two (1.1%) were stool Xpert-positive; clinical improvement could not be reliably determined after the study.

DISCUSSION
The main finding of the study is that Xpert testing on stool identified 68% of microbiologically confirmed TB and was negative in 98% of children with negative microbiological results. Similar to findings by Nicol et al., 16 this study showed that Xpert on stool was more sensitive in HIV-infected than in non-HIVinfected children. Although the test failed to identify 32% of TB cases, it may be of more value in HIVinfected children, who run a high risk of acquiring TB, as well as high risk of disease progression. Marcy et al. reported a sensitivity of 62% for stool Xpert testing among HIV-infected children. 19 Smear microscopy, the routinely available test at primary care centres, has a sensitivity of ,10-15% in children with proven TB, 20-22 and its performance is even poorer in those with HIV infection. 23 Xpert is currently being rolled out in many parts of southern Africa for use as an initial test for TB diagnosis. The test offers rapid diagnosis and detects more cases than smear microscopy in children. [7][8][9]21 However, its limitation is that it requires a sputum sample, which is difficult to obtain freely expectorated from children. At primary care level, sputum induction in children is rarely attempted, and when it is performed the quality of the specimen varies depending on the personnel, resulting in variable test sensitivity. Welday et al. showed that stool Xpert testing in children detected all of the sputum smearpositive cases. 24 In our study, stool Xpert testing helped to identify most of the IS Xpert-positive cases, suggesting that when a sputum sample is not readily available a stool sample, which is easier to collect, may be used with the Xpert test. However, the diagnostic yield of the stool Xpert test still requires optimisation.
The inability of stool Xpert to detect six of the microbiologically confirmed cases may be due to n (%)

B)
Clinically diagnosed TB interference by PCR inhibitors present in stool samples. Emerging evidence suggests that using larger sample volumes, of 0.6 g 25 and 2 cm 3 , 26 and pretreatment with a stool-processing buffer to inactivate PCR inhibitors, 25 achieves greater diagnostic yield of the stool Xpert test. The stool Xpert test performed poorly in the clinically diagnosed children. Xpert testing has consistently shown poor performance in clinically defined probable TB cases in children. 27,28 This is more likely to indicate a limitation of the assay's detection limit than a limitation of the sample used, as clinical microbiologically negative TB cases are more likely to have paucibacillary disease. The two clinically diagnosed children identified using stool Xpert were both HIV-infected, again indicating the potential use of the test in HIV infection.
The main limitation of our study was the lower than expected TB prevalence, which limited study power. Second, we used one stool sample collected at random; a second sample has been shown to have incremental value in detecting additional cases. 6,7,22 Our findings cannot be generalised to all children, as the performance of the stool Xpert test was assessed only in children aged 5-16 years. Different results may occur with children aged ,5 years, in whom TB disease presentation is markedly different.
In conclusion, the stool Xpert test holds promise as an alternative to sputum Xpert testing, particularly in HIV-infected children. Stool collection is easier and relatively safe compared to sputum collection, and may potentially have diagnostic utility if its sensitivity is optimised. Further studies are required to determine optimal stool collection times, optimal sample volumes, the need for a second sample, concentration techniques that allow maximum yield and test combination algorithms.
Xpert testing on stool samples from children i