Effect of sputum quality and role of Xpert® MTB/ RIF assay for detection of smear-negative pulmonary tuberculosis in same-day diagnosis strategy in Addis Ababa, Ethiopia

Background There is limited information on the performance of the Xpert® MTB/RIF test for diagnosis of smear-negative pulmonary tuberculosis (SNPT) and rifampicin resistance (RR) in the same-day diagnosis approach. The effects of sputum quality and other factors affecting the Xpert performance are also under-investigated. Objective This study aimed to determine the performance of the Xpert® MTB/RIF test for detection of SNPT and RR in the same-day diagnosis strategy and the effect of sputum quality and other factors on its performance. Methods A cross-sectional study was conducted from August 2017 to January 2018 across 16 health facilities in Addis Ababa, Ethiopia. Two spot sputum samples were collected from 418 presumptive SNPT patients, tested with Xpert® MTB/RIF, then compared to tuberculosis culture. Additionally, culture isolates were tested for RR by BACTEC MGIT™ 960 drug susceptibility testing (DST) and MTBDRplus version 2. Results The Xpert® MTB/RIF test detected 24 (5.7%) SNPT cases, with a sensitivity of 92.3% (75.9% – 97.9%) and specificity of 99.2% (97.8% – 99.7%) compared with tuberculosis culture. Xpert® MTB/RIF also detected three (11.58%) RR strains with 100.0% concordance with BACTEC MGIT™ 960 DST and MTBDRplus results. Three blood-stained SNPT samples were positive by Xpert (30.0%), which was 6.9 times higher compared to salivary sputum (odds ratio: 6.9, 95% confidence interval: 1.36–34.96, p = 0.020). Conclusion The performance of the Xpert® MTB/RIF to detect SNPT and RR in same-day diagnosis is high. However, SNPT positivity varies among sputum qualities, and good sample collection is necessary for better test performance.

http://www.ajlmonline.org Open Access 2017 to January 2018. 7 Mortality due to SNPT is also substantial, especially among the immunocompromised. In Mozambique, the mortality of HIV co-infected patients reached 55.8%. 8 Thus, early detection with sensitive diagnostic tools that simultaneously detect drug-resistant tuberculosis is critical.
However, diagnosis of SNPT and smear-negative drugresistant tuberculosis is challenging in low-and middleincome countries (LMICs) like Ethiopia, 9 mainly due to a lack of sensitive diagnostic tools. In most LMICs, smear microscopy, which has low sensitivity, is still the first line of diagnosis. The Xpert ® MTB/RIF test is a cartridge-based, fully automated DNA testing platform that, in less than 2 h, simultaneously detects tuberculosis and mutations conferring rifampicin resistance (RR). 10 The technology detects RR using five probes targeting mutations in the rpoB region of the Mycobacterium tuberculosis genome 11,12 : rpoB mutations are responsible for RR in over 99.5% of RR strains. 10 The Xpert ® MTB/RIF test currently resolves the challenges of initial smear microscopy in tuberculosis case detection, which improves the clinical management of tuberculosis cases. 13,14 In addition, the Xpert ® MTB/RIF is advantageous over smear microscopy and M. tuberculosis culture by its higher sensitivity, simultaneous detection of RR, shorter turnaround time (TAT) (2 h), and a minimal safety requirement. 15 By comparison, culture-based RR confirmation takes more than two weeks to get results. 16 Nevertheless, various factors impact the Xpert ® MTB/RIF's performance, such as the type of specimen, 17 sputum quality, 18,19,20 sample collection and diagnosis strategy, 21,22,23 and clinical characteristics of patients. 17,24,25 Since 2012, more than 314 GeneXpert ® instruments have been installed in Ethiopia. Also, the diagnostic strategy changed from spotmorning-spot, which required three samples collected over two days, to same-day diagnosis (spot-spot diagnosis), which requires two samples collected on the same day. 26 However, information is lacking on the performance of the Xpert ® MTB/RIF test to diagnose SNPT, smear-negative RR tuberculosis, the effects of sputum quality and other factors in the same-day tuberculosis diagnosis approach in Addis Ababa, Ethiopia. Therefore, this study aimed to determine the performance of the Xpert ® MTB/RIF test to diagnose SNPT and RR against conventional tuberculosis culture and drug susceptibility testing (DST) in the spot-spot diagnostic strategy. It also intended to determine the SNPT positivity in sputum samples with varying quality and the effect of sputum quality and other factors on Xpert ® MTB/RIF test performance.

Ethical considerations
The study was reviewed and approved by the Institutional

Study setting and design
This cross-sectional study was conducted from August 2017 to January 2018 in Addis Ababa, Ethiopia. Addis Ababa is the federal capital of Ethiopia and has a population of more than three million. 27 The study was conducted in 16 systematically selected governmental and private directly observed treatment short-course tuberculosis sites. These sites participate in blind rechecking, and most participate in the on-site supervision programme by the Addis Ababa City Administration health research and laboratory service.

Data collection
Socio-demographic information, clinical presentations, comorbidities, and other factors were collected using a structured questionnaire by trained clinicians through interviews. The questionnaire was piloted following the on-site training of the data collectors.

Patient enrollment
A total of 418 presumptive SNPT patients who were negative for two consecutive spot sputum smears were enrolled. Adults and children with presumptive PTB who visited the sites during the study period were eligible. However, those taking anti-tuberculosis drugs for more than one week or unable to submit two spot sputum samples were excluded.
Trained laboratorians gave patients detailed instructions on sputum collection. Each presumptive SNPT patient collected two sputum spot ≥ 3 mL samples into sterile 50 mL falcon tubes. Sputum samples were transported with a triple packaging system to the National Tuberculosis Reference Laboratory (NTRL) of the Ethiopian Public Health Institute (EPHI) for laboratory investigation. Sputum samples were stored at 2 °C -8 °C until transported by triple packaging system for safety, with a thermometer inside for temperature monitoring. The transportation from study sites to NTRL takes less than one hour since it is within the city. Samples were submitted to NTRL within a maximum of two days. Macroscopic sputum quality was evaluated and categorised as blood-stained, purulent, mucoid, and salivary based on the Global Laboratory Initiative (GLI) guidance. 28

Laboratory investigations
The two consecutive spot sputum samples collected from each patient were pooled into one and homogenised. Quality assurance and quality control The sterility and performance of LJ media-manufactured inhouse were verified before use for testing using controls by randomly selecting some prepared LJ medium, putting it in the LJ incubator and monitoring for any contaminant growth. If no growth was observed within 56 days, it was considered as sterile or safe for use. Likewise, a new lot or batch of BACTEC™ MGIT™ 960 media was also verified. Each of the culture and identification procedures was performed in a certified Class II biosafety cabine. Preventive maintenance of the equipment, temperature monitoring, and instrument operation checks were performed. In each batch of sample cultured, reagent sterility and process contamination check control were incorporated based on the NTRL standard procedures. A proficiency testing scheme continuously monitored all study test methods. Also, the NTRL is International Organization for Standardization 15189 accredited by the Ethiopian National Accreditation Office for Xpert ® MTB/RIF.

Data entry and analysis
Double data entry was perfomed on EpiData statistical software version 3.0 (EpiData Association, Odense, Denmark). The clean data were transferred to and analysed using IBM Statistical Package for Social Sciences software version 20.0 (Chicago, Illinois, United States). Data entry and cleaning were performed by the NTRL data manager. The characteristics of the study participants were analysed using descriptive statistics. The sensitivity and specificity of the Xpert ® MTB/ RIF test were calculated at 95% confidence intervals (CI). Pearson chi-square, Fisher's exact test, or binary logistic regression was used to determine the associations between dependent and independent variables; variables with a p ≤ 0.2 were selected for multivariable analysis. The strength of associations was measured by odds ratios, and a p < 0.05 was taken as statistically significant.

Socio-demographic and clinical characteristics of study participants
Most of the participants (231; 55.3%) were female.  (Table 1 and Table 2).
The majority of the positive Xpert ® MTB/RIF tests, 14 (51.9%), were low and very low in bacillary load. There were 20 (4.8%) unsuccessful Xpert MTB/RIF results (sum of errors, invalids and no results). The most unsuccessful results were errors (15; 3.6%) ( Table 4). The error code 5007 was the most common (      Table 1).

Discussion
Smear microscopy is still a front-line tuberculosis diagnostic tool in developing countries like Ethiopia. 1,21,29 However, because of its low sensitivity, smear microscopy misses many tuberculosis cases, resulting in delayed diagnosis and treatment initiation. 15,30,31 Therefore, this study evaluated the effect of sputum quality and Xpert ® MTB/RIF performance for SNPT detection in same-day diagnosis.
Sputum quality has effects on the detection of PTB. 19,20,32,33 However, many studies do not consider attributes of sputum quality in MTB testing by the Xpert ® MTB/RIF test. 34,35 In the present study, the M. tuberculosis positivity by Xpert ® MTB/ RIF significantly varied across sputum qualities. The majority of sputum, 147 (35.2%), submitted was salivary with a positivity rate of 6.1%. Although the blood-stained sample was the least submitted, the positivity was very high at 30.0%. In 95% CI, blood-stained samples showed 6.9 (p = 0.020) times more positive than salivary.
Many laboratories reject blood-stained sputum, assuming that it brings unreliable Xpert results due to polymerase chain reaction inhibition; however, this study revealed the highest SNPT positivity in blood-stained sputum. Contrary to the present study, a study showed that the Xpert result is only valid at less than 2.0% blood contamination of the sputa. If sputum contamination with blood is beyond 5.0%, the result will be unreliable and absolute inhibition occurs at 20.0% blood contamination. 18 However, studies found that patients who are MTB positive in a blood sample have a higher risk of death. 36 Therefore, we are missing the most important sputum quality. The current finding is different from a study in Uganda, where SNPT was more in salivary sputum and lowest in blood-stained sputum. 20 This might be as a result of the fact that the majority of the Ugandan study participants were HIV positive and most of them had a very low CD4 count (≤ 200 cells/µL). The second highest SNPT positivity, 7.7%, was diagnosed from purulent sputum. Although the difference is not statistically significant, blood-stained sputum showed greater positivity than purulent sputum, whereas purulent sputum is considered the best sputum for MTB detection. The difference in macroscopic sputum appearance significantly varied for MTB positivity by Xpert ® MTB/RIF test, implying a need for proper sputum collection, similar to other reports. 20,33,34 In Ethiopia, the sputum sample collection strategy for MTB diagnosis was changed from spot-morning-spot to spot-spot (same-day diagnosis) in 2017. 26 Same-day diagnosis stops multiple visits to the health facilities by the patient to submit sputum and receive a result. However, it is 2.8% less sensitive with a lower dropout rate than the conventional spot-morning-spot strategy. 21 In the spot-morning-spot strategy, three smear slides are made. The first slide made from the first spot sputum, the second slide made from the morning sputum and the third slide from the second spot sputum. 22   The diagnostic sensitivity of the Xpert ® MTB/RIF test in this study was 92.3%, while specificity was 99.2% reference to tuberculosis culture. The present study revealed higher sensitivity and specificity than the Uganda report 20 because most of Uganda's study participants were HIV-positive and used a spot-morning-spot diagnostic approach. Another justification might be tuberculosis prevalence is higher in Ethiopia. Similarly, the present study revealed a very high sensitivity compared to a study in Jigjiga, Ethiopia: 48.5% for smear-negative PTB. 37 Likewise, the current study showed higher sensitivity and specificity relative to a review conducted in Liverpool, United Kingdom, in 2013, which was 67.0% -74.0% pooled sensitivity and 99.0% pooled specificity. 38 However, the sensitivity of the Xpert ® to diagnose SNPT showed considerable variability in different studies. 39,40,41 The possible reasons for the variabilities in sensitivity and specificity might be differences in study design, study population, sample collection strategy, study period, study area or location, tuberculosis prevalence, comorbidity and laboratory performance. The findings in this study are commendable since they facilitate early tuberculosis diagnosis and universal access to DST, 42 which are pivotal to the endTB strategy. Furthermore, the World Health Organization approved that tuberculosis diagnostic technologies such as Xpert ® MTB/RIF need to be scaled up to become the first line of diagnosis. With the scaleup, patients will benefit from early diagnosis and initiation of treatment.
More than half of the positive Xpert ® MTB/RIF results (14; 51.9%) had a low or very low bacillary load, implying that SNPT patients have a paucibacillary load. Twenty (4.8%) Xpert ® MTB/RIF results were unsuccessful (the sum of errors, invalids and no results), which is lower than a report from Nigeria. 43 The Nigeria study did not include only presumptive smear-negative SNPT cases. Most of our unsuccessful results were due to error results (15; 3.6%), but in the Nigerian study, these were considered invalid results. 43 The error rate in the present study was higher than the GLI recommendation (< 3.0%), 44 but lower than a report from Addis Ababa, which reported 8.9%. 45 The discrepancy in the study reports might result from the difference in the testers' expertise and experience, study population, sample type, and study period. The error code 5007 was the most common (12; 2.9%), often caused by viscous sputum or wrong sample volume, improper filling of cartridge reaction tube, bubbles, or probe integrity issues. Errors are a loss of valuable time and money for the patients and the laboratory. Thus, patient training on quality sputum collection and laboratory staff refresher training may minimise these errors. Of the three RR cases detected by the Xpert ® MTB/RIF test, two were due to probe E (codons 529-533) missing, which is the most frequent type of mutation in the rpoB region of the mycobacteria. 46,47 The other was probe B missing (codons 511-518; mutation).

Limitations
The current study did not explain the performance of the Xpert ® MTB/RIF test in the same-day diagnosis to diagnose smear-negative RR as it did not include enough RR cases, which is a limitation of the study.

Conclusion
The performance of the Xpert ® MTB/RIF test to detect SNPT in spot-spot samples and rapid detection of RR were high. However, the diagnostic performance of the test significantly varied across different sputum qualities. Thus, good patient instruction or training and close follow-up on sputum sample collection are essential for getting quality sputum for better Xpert ® MTB/RIF yield. We recommend testing sputum samples by Xpert ® MTB/RIF irrespective of sputum quality.