The clinical diagnostic value of Xpert MTB/RIF for the detection of Mycobacterium tuberculosis in gastric aspirates

Abstract Background: At present, the infection and prevalence rates of tuberculosis (TB) are still high in worldwide. The Xpert MTB/RIF technology has improved the diagnosis speed of Mycobacterium tuberculosis (MTB) and facilitated the rapid treatment of TB patients. Methods: We searched experimental data derived from Xpert MTB/RIF for detecting MTB in gastric aspirates in PubMed, Embase, Web Of Science, and the Cochrane Library databases between January 2012 to April 2019. A summary receiver operating characteristic curve (SROC curve) was used to analyze the pooled sensitivity, pooled specificity, PLR, NLR, and DOR for determining the accuracy of the test. Results: Our database search resulted in 10 relevant articles. The pooled sensitivity of Xpert MTB/RIF for detecting TB in GA was 86% (95% CI, 83–89%), and I2 = 93.4%. The pooled specificity was 92% (95% CI, 90–93%) and I2 = 97.8%. In addition, the positive LR was 12.12 (95% CI, 5.60–26.21), negative LR was 0.20 (95% CI, 0.11–0.36), and the diagnostic odds ratio (DOR) was 147.04 (95% CI, 37.20–581.19). Using the SROC curve, the AUC was 0.9730 and Q* was 0.9248 (SE = 0.0261). The publication bias was P=0.517 (P>0.05). Conclusions: The Xpert MTB/RIF for detecting MTB in gastric aspirates was highly accurate. In addition, we observed that the publication bias in the present study was low. Hence, the Xpert MTB/RIF technology is highly accurate and has the advantage of rapid testing for MTB in clinical samples.


Introduction
Tuberculosis (TB) remains a globally essential and leading cause of infectious disease. It was estimated that around 3 million TB patients were undiagnosed in 2013, with 10 million people diagnosed with TB in 2017 [1]. Approximately, 95% of TB infections and 99% of deaths due to TB infections occur in developing countries, with South Africa being the most affected country [2]. Hence, the early MTB diagnosis is very important. Sputum samples are the most commonly used material for diagnosis. Culturing MTB is the gold standard for TB diagnosis. However, it usually takes several weeks to produce the results needed for a proper diagnosis. Because the results cannot be quickly generated, timely diagnosis and patient treatment are delayed [3][4][5]. The WHO recommended method for the initial diagnosis of MTB is smear microscopy. However, due to the low sensitivity of microscopic examinations, the majority of patients will be misdiagnosed [6]. Hence, new methods are urgently needed to develop early diagnosis methods, shorten treatment times, and improve treatment efficacy and prevention.
In 2011, the MTB rapid molecular diagnostic Xpert MTB/RIF test was introduced. It has improved diagnostic sensitivity for detecting MTB [7]. At present, the Xpert MTB/RIF is used to diagnose bacterial infections in patients more accurately. Globally, the Xpert MTB/RIF test has reduced clinical misdiagnosis and has diminished the empirical treatment of patients with negative results. As previously demonstrated by Penz et al. [8], the Xpert MTB/RIF test has different sensitivities and specificities for detecting MTB obtained from lymphatic, pleural effusion, gastrointestinal tract, genitourinary system, cerebrospinal fluid, and other samples. Hence, the purpose of the present study was to perform a research from recently published studies that used the Xpert MTB/RIF test. The overall diagnostic accuracy for detecting MTB in gastric aspirates was then determined for its suitability for clinical diagnosis.

Method Data sources and search strategy
The following search terms were used: Gastro Enteric, Gastrointestinal, Gastro-Intestinal, GI, Gastric Aspirate, GA, Xpert, Xpert MTB/RIF, Xpert RIF/MTB. The search strategy was performed independently by the authors using Pubmed, Embase, Web Of Science, and the Cochrane Library. Afterwards, the authors presented and compared their individual results. If they were inconsistencies, the search strategy was re-evaluated to resolve any differences. Through discussions, relevant studies were selected and used for the analysis.

Study selection and data extraction
The inclusion and exclusion criteria are formulated first before reading the retrieved studies. The inclusion criteria as follow: (1) Each included study used Xpert MTB/RIF for detection of MTB and a standard test (culture). Based on pre-established inclusion and exclusion criteria, the authors independently read the article title, abstract, keywords, etc. to determine whether the article was relevant. Afterwards, the authors compared their results. Any inconsistencies in article selection required all authors to read the full manuscript. The article in question was included or excluded after a consensus was reached through discussions.
The authors independently read the full article after the initial screening to determine whether the data were relevant and extractable. The extractable data included the name of the first author, publication year, country, sample size, gold standard, and true positive (TP), false positive (FP), false negative (FN), and true negative (TN) compared with Xpert and gold standard data. After data extraction, the authors collectively discussed the results and re-read articles with conflicting results. The authors sought additional advice from experts to resolve any differences. Afterwards, the extracted data were used to formulate the data feature table.

Statistical analysis
Meta-disc (version 1.4) was used to analyze the data and determine the sensitivity, specificity, positive LR, negative LR, diagnostic OR, and SROC curve tables.
Publication bias was determined using STATA (version 12.0) at a test level of α = 0.05.

Study identification and characteristics
A total of 161 studies are retrieved: 48 studies from PubMed, 54 studies from WOS, 50 studies from Embase, 9 studies from the Cochrance Library, and 67 of the 161 studies are duplicates. The remaining 94 articles are excluded after reading the title and abstract of the article based on the inclusion and exclusion criteria, including 3 case reports, 8 reviews, 1 meta-analysis, and 47 unrelated studies. The remaining 35 articles exclude 25 articles after reading the full text, including 1 article lacking a gold standard comparison, and 24 articles cannot be used to form a 2 × 2 table for analysis based on their data. In the end, we got 10 articles for analysis ( Figure 1). The 10 selected articles [9][10][11][12][13][14][15][16][17][18] spanned from 2012 to 2018, of which 6 were published in Asia [9][10][11][12][13][14], 2 were published in Europe [15,16], and 2 were published in South Africa (Table 1) [17,18]. The number of study samples in each article ranged from 48 to 788, with a median of 245.5.

Study quality
The overall methodological quality evaluation of the Xpert MTB/RIF and bacterial cultures used in the study are summarized in Figure 2. The risk of bias due to patient selection, index testing, reference standards, procedures, and time were considered low.

Sensitivity and specificity of Xpert MTB/RIF for diagnosis of MTB in GA
Ten studies included a comparison of the Xpert MTB/RIF versus cultures in 2774 GA samples for detecting tuberculosis (Figures 3 and 4).    (Figures 5 and 6). The diagnostic odds ratio (DOR) was 147.04 (95% CI, 37.20-581.19) (Figure 7).

SROC curve
The AUC was 0.9730 determined from the SROC curve, indicating a good diagnostic value (Figure 8). In addition, Q* was 0.9248 (SE = 0.0261). The Q * value was close to 1, which indicated the high accuracy of Xpert MTB/RIF for MTB detection.

Publication bias
The Deeks funnel plot was generated using the STATAS software. The P-value was 0.517 (P>0.05), indicating low publication bias (Figure 9).

Discussion
In the present study, we extracted Xpert MTB/RIF data from 2774 GA samples obtained from 10 studies. The TB test was performed using GA samples, with bacterial cultures used as the standard. We found that the pooled sensitivity of the Xpert test was 86% (95% CI, 83-89%), and the pooled specificity was 92% (95% CI, 90-93%). In general, the Xpert MTB/RIF TB test performed using GA samples generated results that were relatively accurate compared with the culture method. The research performed by Penz et al. [8] in 2015, demonstrated that the Xpert MTB/RIF test had a sensitivity of 86% (95%CI, 67-98%), and specificity of 98% (95% CI, 98-100%) for analyzing GA samples. This was comparable to our results. However, their positive LR was 65 and negative LR = 0.18, which was different from our results (positive LR = 12.12). The reason may be that we included 10 studies for systematic analysis, while Penz el al. had only 5 studies in their research. In addition, the total sample size was much higher in our study. Hence, the use of the Xpert MTB/RIF test for detecting MTB in GA samples was highly accurate and could be used in clinical settings or public health systems to detect MTB infections in patients. The standard diagnostic method for detecting MTB is through bacterial cultures [3][4][5]. We compared the Xpert MTB/RIF technology with the culture method. The Xpert MTB/RIF detects GA differently from the culture method. It is not affected by the acidity of the GA fluid, treatment delay of the bacteria, and excessive bacterial growth [9]. The diagnostic yield of Xpert MTB/RIF (31.4%) was significantly higher compared with the culture method (24.0%) [9]. Singh et al. [14] demonstrated that Xpert MTB / RIF could be used for diagnosing childhood TB even when the GA samples were stored for 4 years. Singh et al. also said in smear and culture-positive GA samples, the sensitivity of the Xpert MTB/RIF assay was 95.6% and was similar to the sensitivity reported for fresh clinical samples. Previous studies have demonstrated that Xpert MTB/RIF had a high detection rate and accuracy. We performed a rigorous search, using stringent screening criteria, and analyzed the data using strict statistical methods to improve the accuracy of our results.
Based on previous studies, the detection rate of MTB by microscopy was 48.94% and was prone to be affected by several factors [19,20]. The detection rate of Xpert MTB/RIF greatly exceeded the detection rate of microscopic examinations. Besides, the Xpert MTB/RIF test is convenient and rapid to perform.
Our work had a few limitations. The selected publications generated inconsistent data due to different instruments and technical protocols used. The low sensitivity and specificity caused from some of the studies had a significant influence on our research. Walters et al. [18] did not describe the technical details of processing their gastrointestinal fluid samples. Hence, we could not make a more objective evaluation of the low sensitivities they observed in their study. Aslams et al. [9] found that the specificity was only 53%. This could probably have been due to patients receiving anti-tuberculosis drugs. After treatment, some patients may have been cured. Hence there is a need to carefully evaluate the positive results from the Xpert MTB/RIF test [21]. Pang et al. [10] demonstrated relatively low levels of sensitivity and specificity. This may have been due to the long storage time of the samples (>3 days), and the MTB in the GA fluids being partially inactivated by the high acidic environment of the samples. It has been proven that MTB may be inactivated in an acid environment [22,23]. Their positive results from bacterial cultures were relatively low.
Additional studies using larger patient cohort samples should be performed to determine whether the Xpert MTB/RIF test could accurately detect MTB in GA samples.
The advantages of the Xpert MTB/RIF test are the reduced cost and faster turn-around times to generate results. This will prevent patients from being misdiagnosed and receive the appropriate treatment in a shorter period. Quicker and a more accurate diagnosis will lead to better patient prognosis, a better quality of life and reduced mortality.