key: cord-0065686-5ju0p0nu authors: Getachew, Emnet; Adebeta, Tsegaye; Gebrie, Desye; Charlie, Loveness; Said, Bibie; Assefa, Dawit Getachew; Wanjiru, Cathrine Lydiah; Zeleke, Eden Dagnachew; Tesfahunei, Hanna Amanuel; Abebe, Mekdelawit; Joseph, Michele; Manyazewal, Tsegahun title: Pyrosequencing for diagnosis of multidrug and extensively drug-resistant tuberculosis: A systemic review and meta-analysis date: 2021-06-29 journal: J Clin Tuberc Other Mycobact Dis DOI: 10.1016/j.jctube.2021.100254 sha: 0855b65705d71b6aa90aeb60b3da0ccb5431cfb6 doc_id: 65686 cord_uid: 5ju0p0nu BACKGROUND: Multidrug and extensively drug-resistant tuberculosis (M/XDR-TB) pose major threats to global health. Diagnosis accuracy and delay have been the major drivers for the upsurge of M/XDR-TB. Pyrosequencing (PSQ) is a novel, real-time DNA sequencing for rapid detection of mutations associated with M/XDR-TB. We aimed to systematically synthesize the evidence on the diagnostic accuracy of PSQ for M/XDR-TB. METHODS: We conducted an electronic search of PubMed, Embase, Biosis, Web of Science, and Google Scholar up to March 2020. We used the QUADAS‐2 (Quality Assessment of Diagnostic Accuracy Studies) tool to assess the quality of studies, the BRMA (bivariate random‐effects meta-analysis) model to synthesize diagnostic accuracies, and the Rev-Man 5.4 software to perform the meta-analyses. We analyzed dichotomous data using the risk ratio (RR) with a 95% confidence interval. PROSPERO Registration ID: CRD42020200817. RESULTS: The analysis included seven studies, with a total sample of 3,165. At 95% confidence interval, the pooled sensitivity and specificity of PSQ were 89.7 (CI: 83.5–93.8) and 97.8 (CI: 94.9–99.1) for Isoniazid, 94.6 (CI: 90.9–96.8) and 98.5 (CI: 96.5–99.3) for Rifampicin, 87.9 (CI: 81.2–92.4) and 98.8 (CI: 97.2–99.5) for Fluoroquinolone, 83.5 (CI: 72.8–90.5) and 99.4 (CI: 98.3–99.8) for Amikacin, 79 (CI: 67–8-87) and 97.9 (CI: 95.5–99) for Capreomycin, and 69.6 (CI: 57–79.8) and 98.2 (CI: 95.9–99.2) for Kanamycin. The overall pooled sensitivity and specificity were 85.8 (CI: 76.7–91.7) and 98.5 (CI: 96.5–99.3), respectively. CONCLUSION: According to the pooled data, PSQ is highly sensitive and specific for detecting M/XDR-TB, both from clinical specimens and culture isolates, and within a shorter turnaround time. We suggest a continued synthesis of the evidence on the cost-effectiveness and technical feasibilities of PSQ in low-income countries context, including sub-Saharan Africa. Tuberculosis (TB) remained a major threat to global health [1] [2] [3] . According to the World Health Organization's global TB report 2020 [4] , an estimated 10.0 million people fell ill with TB in 2019 and close to half a million people developed rifampicin-resistant TB (RR-TB), of which 78% had multidrug-resistant TB (MDR-TB). The upsurge of multidrug and extensively drug-resistant tuberculosis (M/XDR-TB), as well as the totally drug-resistant TB (TDR-TB)" is hindering the global TB control efforts [5, 6] . The major gap has been the insufficient capability to diagnose M/XDR-TB in a timely and cost-effective manner [7, 8] . Reliable molecular diagnostics, which can detect specific mutations that confer drug resistance, are the most promising technologies for rapid identification of M/XDR-TB. These technologies, most remarkably the Hain MTBDR Plus line probe assay [9, 10] and the Gene Xpert MTB/ RIF assay [11, 12] , are being used globally to screen for the most prevalent mutations associated with resistance to isoniazid (INH) and/or rifampin (RIF). However, they cannot detect all mutations in specific region and are unable to distinguish silent mutations [13] . Therefore, there has been little progress towards the broad application of these tools for M/XDR-TB. Pyrosequencing (PSQ) emerges as a real-time assay for rapid sequencing of small segments of genomic DNA to detect M/XDR-TB accurately and reliably [14] [15] [16] . It was reported that it overcomes the limitation of previous M/XDR-TB assays and detects resistance to more drugs in a shorter time [17, 18] . It not only determines the presence or absence of these mutations but also displays the exact sequence data to guide treatment decisions. However, studies are lacking that synthesized the evidence on the diagnostic accuracy of PSQ for M/XDR-TB. Multiple trials are evaluating novel therapeutic agents, repurposed agents, and novel care cascades to transform the landscape for M/XDR-TB [19] [20] [21] [22] . Looking for new diagnostic tools and following-up of diagnostic accuracy of existing ones are equally important to harness the End-TB pathway. Thus, we aimed to synthesize the evidence on the diagnostic accuracy of PSQ for M/XDR-TB. The study followed the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 guidelines [23] for reporting results, QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies) tool [24] to assess the quality of studies, and the BRMA (bivariate random-effects meta-analysis) model [25] to synthesize diagnostic accuracies. The protocol has been registered on PROSPERO, ID: CRD42020200817 [26] . We included studies that met the study's predefined criteria: • Types of studies: Multi or single-center, prospective, head-to-head clinical evaluation • Participants: Patients of any age who had: isoniazid, rifampicin, fluoroquinolones, amikacin, kanamycin and capreomycin resistant TB. We searched the following databases to find abstracts and their fulltext articles: PubMed, Embase, Biosis, Web of Science, and Google Scholar. We searched up to March 2020 using the keywords and search terms "Tuberculosis* AND Pyrosequencing*", Multidrug-Resistant", "Drug Resistance", "Drug Resistance, Bacterial", "extensively drug resistance", "pyrosequencing*", "Tuberculosis second-line drug-resistant", "PSQ", "phenotype DST*", "BACTEC MGIT960*", "Tuberculosis total drug resistance*", and "pyrosequencing, XDR TB". Two review authors independently extracted data and assessed the quality of included studies using QUADAS-2 tool. We had combined data for PSQ and assessed sensitivity and specificity for anti-M/XDR-TB drugs. We explored the influence on accuracy estimates of individual drugs within a drug class of the same reference standards. We carried out most analyses using a bivariate random-effects model with culturebased DST as a reference standard. Where sufficient data were available, we have compiled data applying meta-analysis. Plotted estimates of sensitivities and specificities in forest plots with 95% Confidence Intervals (Cis). We used Rev-Man 5.4 software to perform the metaanalyses. Two review authors independently examined titles and abstracts of identified citations. Titles and/or abstracts of all citations were screened further for a full-text review using the predefined criteria. For full-text articles, we resolved any discrepancies by consensus. We extracted results from the selected studies that evaluated the accuracy (sensitivity and specificity) of PSQ for M/XDR-TB in clinical specimens or isolates. We considered the following pre-determined criteria to include studies in our review: • PSQ was utilized for rapid M/XDR-TB detection • Detection was performed in clinical specimens or isolates • The study evaluated accuracy (sensitivity and specificity) of PSQ • Phenotypic DST on either liquid or solid media used as reference standard • Sample size should include at least 20, with at least 10 specimens showing M. TB with drug resistance, and at least 10 with drugsusceptible strains. We customized a data extraction form designed to extract and capture the following information: • Details of the study: first author; publication year; country where testing was performed; specimen country origin; setting; participant selection procedure; and the number of participants enrolled • Target conditions: resistance to the following drugs: isoniazid, rifampicin, fluoroquinolones, amikacin, kanamycin and capreomycin. • Resistances to individual drugs: Isoniazid, rifampin fluoroquinolone, and capreomycin, Kanamycin, and Amikacin. • Reference standard: type • Specimen: type; condition (fresh or frozen); definition of a positive test; type of testing • Outcomes: sensitivity and specificity Included articles were assessed by the first author, who extracted data using the extraction form. A second reviewer independently extracted data from a subset of the included studies and differences between the two reviewers were reconciled by consensus. Quality of included studies was assessed using the QUADAS criteria. Due to the limited number of included studies, it is difficult to assess the source of heterogeneity using meta-regression analysis. We have scrutinized heterogeneity through visual examination of forest plots of sensitivity and specificity. Then, if adequate studies were available, we have explored the possible influence of the following pre-specified categorical covariate: resistance to the following drugs: Isoniazid, Rifampicin, fluoroquinolone, Amikacin, Kanamycin, and capreomycin resistance and based on drug group. We had planned to explore the effect of HIV status, the condition of the specimen (fresh or frozen), sample volume, and population (patients thought to have MDR-TB or XDR-TB) on summary estimates of sensitivity and specificity in a meta-regression analysis by adding covariate terms to the bivariate model. However, there was inadequate data that specify all the above information for these additional analyses. For our analyses using the culture-based DST reference standard, we have conducted sensitivity analyses for QUADAS-2 items to explore methodological quality of the studies, checking whether (Table 3 ): • A consecutive or random sample of patients/specimens were enrolled • A case-control design was avoided • Index test results were interpreted without knowledge of the results of the reference standard • The test was applied in the way recommended by the manufacturer We did not assess for publication bias of included data as such techniques were not useful for biases within diagnostic test accuracy studies. We summarized data on inter-reader variability; but not for intra-reader variability as information were not described in the included studies. We had also intended to summarize two patient outcomes, time-to-diagnosis, and time-to-treatment initiation; however, time-to-diagnosis was the only outcome described in the included studies. We found 50 studies from the initial search, of which seven [27] [28] [29] [30] [31] [32] [33] met the inclusion criteria of this review (Fig. 1) . Figs. 2 and 3 show the risk of bias and applicability concerns for each of the 7 included studies. In the patient selection domain, we judged that 1 study (14.3%) had an unclear risk of bias. Since the report on one study did not provide detailed information on the procedure of patient selection, we judged that 6 studies (85.7%) [28] [29] [30] [31] [32] [33] had a low risk of bias in the index test. We have judged that only one study (14.3%) as a low risk of bias in flow and timing of sample collection and study procedure, while the authors of the other studies did not provide information on the timing of sample collection (Fig. 2) . Regarding applicability, six studies [27] [28] [29] [31] [32] [33] had low concern in patient selection and only one study [29] had unclear concerns in the index test domain. In the reference-standard domain, one study (14.3%) [27] had unclear selection and application of reference standard since it was impossible to find out which reference standard had been used (Fig. 3 ). The seven studies that were included had a total sample of 3165. Of the seven studies, one study reported results using both clinical specimens and clinical isolates, 4 studies (57%) evaluated patients from multi-center of different countries, and the rest (43%) were from a single country. The median sample size (interquartile range) was 593 (57 to 1128). Five (72%) studies reported results of INH, RIF, FQ, AMK, KAN, and CAP. One study [30] did not include the result of AMK, CAP, and KAN, and one study [33] did not include the results of AMK and INH. Table 1 summarized all relevant data extracted from included studies. The overall pooled sensitivity 85.8% (95% CI: 76.7-91.7) and pooled specificity 98.5 (95% CI: 96.5-99.3) ( Table 3 4 Here, we conducted a systematic review and meta-analysis on the value of PSQ for drug susceptibility of M. tuberculosis in direct specimens and isolates of 3265 enrolled patient samples. We found a high overall pooled sensitivity of 85.8% and a pooled specificity of 98.5%. There is a critical need for rapid and accurate diagnosis of DR-TB. Earlier detection of DR-TB if followed by prompt and appropriate treatment changes could potentially reduce disease progression and transmission. As our pooled analysis indicated, PSQ is highly specific and comparable to diagnostic tools currently in use such as phenotypic DST, LPA, or Xpert. PSQ has an added value as it is more rapid, can be applied directly from the clinical specimen, make available exact data about all known mutations at once, is open and adaptable to new mutations, and allows extensive post-processing analyses [34, 35] . Previous studies reported that PSQ is more affordable for less affluent settings, typically relevant for areas around the world that are highly burdened with drug-susceptible as well as drug-resistant TB, easily distinguish between missense and silent mutation and detect outside the target region compared to phenotypic Compared with DNA sequencing, the concordance rate of PSQ for first-line drugs was 100% [35, 36] . The overall concordance between PSQ and GenoType MTBDR plus in clinical strains was 99.1% [27] . Thus, PSQ for conferring resistance to RIF and other first-line anti-TB drugs is reliable. Compared with other methods, such as PCR-single strand conformational polymorphism assay 20 and LiPA (line probe assay), PSQ is shown to save time and labor [37] . For sequencing, it is possible to run 96 samples in 1-2 h after PCR amplification and can analyze 96 single nucleotide polymorphisms in 10 min [17, 35, 36] . PSQ can therefore be used as a high-throughput and rapid assay for M/XDR-TB detection. All the included studies reported PSQ as a rapid and accurate method for detecting MTB resistance to INH, RIF, and FQ and SLIs. One study reported low sensitivity of SLIs (KAN and CAP), unlike the other studies. This could be for a reason that PSQ did not detect "eis" mutations. Nextgeneration PSQ assays have now included "eis" mutations that improved their capacity to confer phenotypic KAN resistance [29] . This review had some limitations. Despite searching several sources, some eligible studies may not have been identified and relatively small numbers of eligible studies were included, hampering the subgroup analysis and meta-regression methods. Furthermore, the study does not address some major first-choice drugs included in the updated WHO recommendations for the treatment of MDRTB, including drug susceptibility testing of bedaquiline, clofazimine, linezolid, delamanid and pretomanid. Another limitation is it does not address some of the variations in techniques used under the umbrella of pyrosequencing; over the last decades, the technique has made several improvements that potentially impact the results. On contrary, this review has several strengths. It included studies assessing the diagnostic accuracy of PSQ for both first-line and secondline anti-TB drugs, including second-line injectable drugs. The results were based on strict and careful literature searches, study inclusion, and data extraction. The work strictly followed internationally recognized guidelines and procedures. According to the pooled data, PSQ is highly sensitive and specific for detecting M/XDR-TB, both from clinical specimens and culture isolates, and within a shorter turnaround time. We suggest a continued synthesis of the evidence on the cost-effectiveness and technical feasibilities of PSQ in low-income countries context, including sub-Saharan Africa. Current and future treatments for tuberculosis Burden of tuberculosis and challenges related to screening and diagnosis in Ethiopia The fight to end tuberculosis must not be forgotten in the COVID-19 outbreak World Health Organization (WHO) Differential yield of universal versus selective drug susceptibility testing of patients with tuberculosis in highburden countries: a systematic review and meta-analysis Exploring local realities: Perceptions and experiences of healthcare workers on the management and control of drug-resistant tuberculosis in Addis Ababa Molecular epidemiology of drug resistant Mycobacterium tuberculosis in Africa: a systematic review Challenge to treat pre-extensively drug-resistant tuberculosis in a low-income country: a report of 12 cases Evaluation of genotype MTBDRplus VER 2.0 line probe assay for the detection of MDR-TB in smear positive and negative sputum samples The Performance of the Abbott Real Time MTB RIF/INH Compared to the MTBDRplus V2 for the Identification of MDR-TB Among Isolates New Xpert MTB/XDR: added value and future in the field Extract from used Xpert MTB/RIF Ultra cartridges is useful for accurate second-line drugresistant tuberculosis diagnosis with minimal rpoB-amplicon cross-contamination risk Laboratory diagnosis of tuberculosis: advances in technology and drug susceptibility testing Evaluation of pyrosequencing for extensive drug resistance-defining anti-tuberculosis drugs for use in public healthcare PyroTyping, a novel pyrosequencing-based assay for Mycobacterium tuberculosis genotyping Pyrosequencing for rapid detection of Mycobacterium tuberculosis second-line drugs and ethambutol resistance Singlenucleotide polymorphism-based differentiation and drug resistance detection in Mycobacterium tuberculosis from isolates or directly from sputum Prospective evaluation of pyrosequencing for the rapid detection of isoniazid and rifampin resistance in clinical Mycobacterium tuberculosis isolates Current and future treatments for tuberculosis Closing gaps in the tuberculosis care cascade: an action-oriented research agenda Electronic pillbox-enabled self-administered therapy versus standard directly observed therapy for tuberculosis medication adherence and treatment outcomes in Ethiopia (SELFTB): protocol for a multicenter randomized controlled trial Containment of COVID-19 in Ethiopia and implications for tuberculosis care and research Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: elaboration and explanation QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies Bivariate randomeffects meta-analysis and the estimation of between-study correlation Diagnostic accuracy of pyrosequencing versus phenotypic DST for detection of multi and extensively drugresistant tuberculosis in MDR/XDR Tb Patients CRD42020200817 Available from Pyrosequencing for rapid detection of extensively drug-resistant Mycobacterium tuberculosis in clinical isolates and clinical specimens Evaluation of pyrosequencing for detecting extensively drug-resistant Mycobacterium tuberculosis among clinical isolates from four high-burden countries Performance comparison of three rapid tests for the diagnosis of drug-resistant tuberculosis Pyrosequencing for rapid detection of Mycobacterium tuberculosis resistance to rifampin, isoniazid, and fluoroquinolones Detection of first-and second-line drug resistance in Mycobacterium tuberculosis clinical isolates by pyrosequencing Shedding light on the performance of a pyrosequencing assay for drug-resistant tuberculosis diagnosis Pyrosequencing: a rapid and effective sequencing method to diagnose drug-resistant tuberculosis Rapid detection of rifampin resistance in Mycobacterium tuberculosis by Pyrosequencing technology Pyrosequencing-based approach for rapid detection of rifampin-resistant Mycobacterium tuberculosis Pyrosequencing for the rapid detection of rifampicin resistance in Mycobacterium tuberculosis: a meta-analysis A commercial line probe assay for the rapid detection of rifampicin resistance in Mycobacterium tuberculosis: a systematic review and meta-analysis The authors are grateful to the Center for Innovative Drug Development and Therapeutic Trials for Africa (CDT-Africa), College of Health Sciences, Addis Ababa University for supporting this review.