Automated real-time nucleic acid amplification technology ...apps.who.int/iris/bitstream/10665/112472/1/9789241506335_eng.pdf · 4.2 Using Xpert MtB/riF to diagnose eXtrapUlMonary

Automated real-time nucleic acid amplification technology for rapid

and simultaneous detection of tuberculosis and rifampicin resistance:

Xpert MTB/RIF assay for the diagnosis of pulmonary and extrapulmonary TB

in adults and children

Global TB ProgrammeWorld Health OrganizationAvenue Appia 20, CH-1211 Geneva-27, Switzerland

Information Resource Centre HTM/GTB:Email: [email protected]: www.who.int/tb

ISBN 978 92 4 150633 5

POLICY UPDATE

PE

RF

OR

MA

NC

EA

CC

UR

AC

Y ACCURACYRECOMMENDATIONS

MOLECULAR DIAGNOSTICS

TUBERCULOSISDRUG-RESISTANCE

NEW DIAGNOSTIC TESTS RIF

AM

PIC

INP

UL

MO

NA

RY T

B

RA

PID

TB

TE

ST

RESISTANCEDIAGNOSIS

MYCOBACTERIUM

TB

TB/HIV

COVER policy_update_OK.indd 1 11/04/2014 14:50

Automated real-time nucleic acid amplification technology for rapid and simultaneous detection

of tuberculosis and rifampicin resistance:Xpert MTB/RIF assay for the diagnosis of pulmonary and extrapulmonary TB


Policy update

WHO Library Cataloguing-in-Publication Data

Automated real-time nucleic acid amplification technology for rapid and simultaneous detection of tuberculosis and rifampicin resistance: Xpert MTB/RIF assay for the diagnosis of pulmonary and extra-pulmonary TB in adults and children. Policy update.

Revision of Automated real-time nucleic acid amplification technology for rapid and simultaneous detec-tion of tuberculosis and rifampicin resistance: Xpert MTB/RIF system. Policy statement. Geneva, World Health Organization, 2011.

1.Tuberculosis, Multidrug-resistant - diagnosis. 2.Tuberculosis - diagnosis. 3.Rifampin - pharmacology. 4.Mycobacterium tuberculosis - isolation and purification. 5.HIV infections - diagnosis. 6.Sensitivity and specificity. 7.Guideline. I. World Health Organization.

ISBN: 978 92 4 150633 5 (NLM classification: WF 310)

© World Health Organization 2013

All rights reserved. Publications of the World Health Organization are available on the WHO web site (www.who.int) or can be purchased from WHO Press, World Health Organization, 20 Avenue Appia, 1211 Geneva 27, Switzerland (tel.: +41 22 791 3264; fax: +41 22 791 4857; e-mail: [email protected]).

Requests for permission to reproduce or translate WHO publications – whether for sale or for noncommercial distribution – should be addressed to WHO Press through the WHO web site (http://www.who.int/about/licensing/copyright_form/en/index.html).

The designations employed and the presentation of the material in this publication do not imply the expression of any opinion whatsoever on the part of the World Health Organization concerning the legal status of any country, territory, city or area, or of its authorities, or concerning the delimitation of its frontiers or boundaries. Dotted lines on maps represent approximate border lines for which there may not yet be full agreement.

The mention of specific companies or of certain manufacturers’ products does not imply that they are endorsed or recommended by the World Health Organization in preference to others of a similar nature that are not mentioned. Errors and omissions excepted, the names of proprietary products are distinguished by initial capital letters.

All reasonable precautions have been taken by the World Health Organization to verify the information contained in this publication. However, the published material is being distributed without warranty of any kind, either expressed or implied. The responsibility for the interpretation and use of the material lies with the reader. In no event shall the World Health Organization be liable for damages arising from its use.

Designed by GPS Publishing

Printed in France

WHO/HTM/TB/2013.16

Contents

EXECUTIVE SUMMARY XI

WHO’s pOlicy recOmmendatiOns XV

1. BACKGROUND 1

2. METHODS 3

2.1 eVidence syntHesis 3

2.2 eXpert GrOup meetinG 7

2.3 eXternal reVieW 7

2.4 preparinG tHe pOlicy update 7

3. SCOPE 8

3.1 date Of reVieW: 2017 8

4. EVIDENCE BASE FOR POLICY FORMULATION 9

4.1 Using Xpert MtB/riF to diagnose pUlMonary tB and riFaMpicin resistance in adUlts 9

4.1.1 usinG Xpert mtB/rif as a replacement test fOr smear micrOscOpy 10

4.1.2 usinG Xpert mtB/rif as an add-On test fOllOWinG micrOscOpy 12

4.1.3 usinG Xpert mtB/rif tO detect smear-pOsitiVe culture-pOsitiVe tB 12

4.1.4 usinG Xpert mtB/rif tO detect smear-neGatiVe culture-pOsitiVe tB 12

4.1.5 usinG Xpert mtB/rif tO detect pulmOnary tB in HiV-neGatiVe and HiV- pOsitiVe indiViduals 12

4.1.6 usinG Xpert mtB/rif tO detect rifampicin resistance 14

4.1.7 effect Of tHe VersiOn Of tHe Xpert mtB/rif assay 15

4.1.8 accuracy Of tHe Xpert mtB/rif G4 cartridGe 15

4.1.9 accuracy Of tHe reference standards 16

4.2 Using Xpert MtB/riF to diagnose eXtrapUlMonary tB and riFaMpicin resistance in adUlts and children 17

4.2.1 detectinG lympH nOde tB in samples frOm BiOpsy Or fine-needle aspiratiOn 18

4.2.2 detectinG pleural tB in pleural fluid 19

4.2.3 detectinG tB in samples Of csf 20

4.2.4 detectinG tB in Gastric fluid 22

4.2.5 detectinG tB in tissue samples 22

4.2.6 detectinG rifampicin resistance 23

4.3 Using Xpert MtB/riF to diagnose tB and riFaMpicin resistance in children 23

4.3.1 usinG Xpert mtB/rif tO diaGnOse pulmOnary tB in cHildren 24

4.3.2 Xpert mtB/rif cOmpared WitH smear micrOscOpy 26

4.3.3 perfOrmance Of Xpert mtB/rif in smear-pOsitiVe and smear-neGatiVe cHildren 27

4.3.4 Xpert mtB/rif in cHildren aGed 0–4 years and 5–15 years 29

4.3.5 Xpert mtB/rif in HiV-pOsitiVe and HiV-neGatiVe cHildren 30

4.3.6 usinG Xpert mtB/rif tO detect peripHeral lympH nOde tB in cHildren 32

4.3.7 usinG Xpert mtB/rif tO detect tB meninGitis in cHildren 32

4.3.8 usinG Xpert mtB/rif tO detect rifampicin resistance in cHildren 32

4.4 aFFordaBility and cost eFFectiveness oF Using Xpert MtB/riF to diagnose tB 33

5. WHO’S POLICY RECOMMENDATIONS 38

5.1 usinG Xpert mtB/rif tO diaGnOse pulmOnary tB and rifampicin resistance in adults and cHildren 38

5.2 usinG Xpert mtB/rif tO diaGnOse eXtrapulmOnary tB and rifampicin resistance in adults and cHildren 39

6. IMPLEMENTATION CONSIDERATIONS 39

6.1 researcH needs 42

6.2 plans fOr suppOrtinG scale-up Of tHe implementatiOn Of Xpert mtB/rif 42

7. GRADE TABLES 44

ANNEXES 74

anneX 1. memBers Of tHe eXpert GrOup 74

anneX 2. WHO staff memBers 76

anneX 3. memBers Of tHe strateGic and tecHnical adVisOry GrOup fOr tuBerculOsis (staG-tB) 76

anneX 4. declaratiOns Of interests 78

Tables

taBle 1. meta-analysis Of tHe sensitiVity and specificity Of Xpert mtB/rif in diaGnOsinG eXtrapulmOnary tB and rifampicin resistance in adults and cHildren cOmpared aGainst culture as a reference standard as Well as aGainst a cOmpOsite reference standard, By type Of eXtrapulmOnary specimen xiii

taBle 2. pOOled sensitiVity and specificity Of tHe Xpert mtB/rif assay fOr detectinG pulmOnary tB and rifampicin resistance 12

taBle 3. meta-analysis Of tHe estimated sensitiVity and specificity Of smear micrOscOpy in diaGnOsinG pulmOnary tB, peripHeral lympH nOde tB and tB meninGitis in cHildren cOmpared aGainst culture as a reference standard in puBlisHed and unpuBlisHed studies 27

taBle 4. meta-analysis Of tHe sensitiVity and specificity Of Xpert mtB/rif in diaGnOsinG pulmOnary tB, peripHeral lympH nOde tB and tB meninGitis cOmpared aGainst culture as a reference standard in smear-neGatiVe and smear-pOsitiVe cHildren in puBlisHed and unpuBlisHed studies 29

taBle 5. meta-analysis cOmparinG Xpert mtB/rif fOr diaGnOsinG pulmOnary tB, peripHeral lympH nOde tB and tB meninGitis usinG culture as a reference standard in HiV-pOsitiVe and HiV-neGatiVe cHildren, stratified By smear status 31

taBle 6. metareGressiOn mOdel fOr Xpert mtB/rif usinG samples Of eXpectOrated Or induced sputum frOm cHildren, cOntrOllinG fOr smear status and HiV status 31

taBle 7. OVerVieW Of studies cOmparinG tHe cOsts Of usinG Xpert mtB/rif and fOllOW-On tests WitH current diaGnOstic alGOritHms fOr diaGnOsinG tB and mdr-tB 34

taBle 8. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in diaGnOsinG pulmOnary tB in adults 44

taBle 9. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in diaGnOsinG pulmOnary tB in sputum smear-pOsitiVe adults 45

taBle 10. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in diaGnOsinG pulmOnary tB in sputum smear-neGatiVe adults 46

taBle 11. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in diaGnOsinG pulmOnary tB in adults liVinG WitH HiV 47

taBle 12. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in diaGnOsinG pulmOnary tB in adults WitHOut HiV infectiOn 48

taBle 13. Grade eVidence prOfile: tHe incremental yield Of Xpert mtB/rif cOmpared WitH micrOscOpy in patients WitH culture-cOnfirmed tB 49

taBle 14. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in diaGnOsinG pulmOnary tB in adults as an add-On test fOllOWinG neGatiVe sputum-smear micrOscOpy 50

taBle 15. sensitiVity Of Xpert mtB/rif in smear-neGatiVe culture-cOnfirmed pulmOnary tB in indiViduals, By HiV status 51

taBle 16. Grade eVidence prOfile: additiOnal yield Of Xpert mtB/rif OVer micrOscOpy in smear-neGatiVe tB 51

taBle 17. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in detectinG rifampicin resistance, WHere Xpert mtB/rif replaces pHenOtypic culture-Based druG-susceptiBility testinG as tHe initial test 52

taBle 18. accuracy Of Xpert mtB/rif in detectinG tB in lympH nOde fluid and tissue (a. eVidence prOfile, B. summary Of findinGs) 53

taBle 19. accuracy Of Xpert mtB/rif in detectinG tB in pleural fluid (a. eVidence prOfile, B. summary Of findinGs) 55

taBle 20. accuracy Of Xpert mtB/rif in detectinG tB in cereBrOspinal fluid (a. eVidence prOfile, B. summary Of findinGs) 57

taBle 21. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in detectinG tB in Gastric fluid 59

taBle 22. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in detectinG tB in tissue samples 60

taBle 23. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in detectinG rifampicin resistance in nOnrespiratOry specimens 61

taBle 24. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in detectinG tB in cHildren cOmpared WitH culture as a reference standard (a. eXpectOrated sputum and induced sputum, B. Gastric laVaGe Or aspirate, c. summary Of findinGs) 62

taBle 25. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in detectinG tB in cHildren cOmpared WitH a clinical reference standard (a. eXpectOrated sputum and induced sputum, and Gastric laVaGe and aspirate, B. summary Of findinGs) 63

taBle 26. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in detectinG tB in cHildren fOllOWinG neGatiVe smear micrOscOpy (a. eVidence prOfile, B. summary Of findinGs, c. additiOnal yield Of Xpert mtB/rif OVer micrOscOpy) 64

taBle 27. incremental yield Of Xpert mtB/rif cOmpared WitH smear micrOscOpy in cHildren WitH culture-cOnfirmed tB (a. eXpectOrated sputum and induced sputum, B. Gastric laVaGe Or aspirate) 70

taBle 28. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in detectinG rifampicin resistance in respiratOry specimens frOm cHildren 71

taBle 29. Grade eVidence prOfile: accuracy Of Xpert mtB/rif in detectinG peripHeral lympH nOde tB in cHildren 72

taBle 30. Grade eVidence prOfile and summary Of findinGs: accuracy Of Xpert mtB/rif in detectinG tB meninGitis in cHildren 73

Figures

FigUre 1. steps in usinG tHe Xpert mtB/rif assay 2

FigUre 2. selectiOn Of studies eValuatinG tHe accuracy Of Xpert mtB/rif in diaGnOsinG pulmOnary tB and rifampicin resistance in adults: flOW diaGram Of studies identified By tHe initial literature searcHes 9

FigUre 3. selectiOn Of studies eValuatinG tHe accuracy Of Xpert mtB/rif in diaGnOsinG pulmOnary tB and rifampicin resistance in adults: flOW diaGram Of studies identified By tHe updated literature searcH 10

FigUre 4. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif fOr detectinG pulmOnary tB in 27 studies (36 study centres) 11

FigUre 5. fOrest plOts Of tHe sensitiVity and specificity Of Xpert mtB/rif fOr detectinG pulmOnary tB in HiV-neGatiVe indiViduals suspected Of HaVinG tB (9 studies, 18 study centres) and HiV-pOsitiVe indiViduals suspected Of HaVinG tB (10 studies, 16 centres) 13

FigUre 6. fOrest plOts Of tHe sensitiVity and specificity Of Xpert mtB/rif fOr detectinG rifampicin resistance WHen Xpert mtB/rif Was used as an initial test replacinG pHenOtypic culture-Based druG-susceptiBility testinG in 24 studies (33 study centres). (studies are presented in Order Of decreasinG sensitiVity and decreasinG numBer Of true-pOsitiVe results) 14

FigUre 7. selectiOn Of studies eValuatinG tHe accuracy Of Xpert mtB/rif in diaGnOsinG eXtrapulmOnary tB and detectinG rifampicin resistance in adults and cHildren: flOW diaGram Of studies included in tHe reVieW 17

FigUre 8. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG eXtrapulmOnary tB in lympH nOde samples (tissue Or aspirate) cOmpared WitH culture as tHe reference standard 18

FigUre 9. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG eXtrapulmOnary tB in lympH nOde samples (tissue Or aspirate) cOmpared WitH a cOmpOsite reference standard 19

FigUre 10. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG tB usinG pleural fluid cOmpared WitH culture as a reference standard 20

FigUre 11. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG tB in pleural fluid cOmpared WitH a cOmpOsite reference standard 20

FigUre 12. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG tB in cereBrOspinal fluid cOmpared WitH culture as a reference standard 21

FigUre 13. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG tB in cereBrOspinal fluid cOmpared WitH a cOmpOsite reference standard 21

FigUre 14. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG tB in Gastric fluid cOmpared WitH culture as a reference standard 22

FigUre 15. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG tB in tissue samples (OtHer tHan frOm a lympH nOde) cOmpared WitH culture as a reference standard 23

FigUre 16. selectiOn Of studies eValuatinG tHe accuracy Of Xpert mtB/rif in diaGnOsinG tB and rifampicin resistance in cHildren: flOW diaGram Of studies included in tHe reVieW 24

FigUre 17. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG pulmOnary tB, peripHeral lympH nOde tB and tB meninGitis in cHildren cOmpared aGainst culture as a reference standard, By study and specimen type 25

FigUre 18. fOrest plOt Of tHe sensitiVity and specificity Of smear micrOscOpy in detectinG pulmOnary tB, peripHeral lympH nOde tB and tB meninGitis in cHildren cOmpared aGainst culture as a reference standard , By study and specimen type 26

FigUre 19. fOrest plOt Of tHe sensitiVity Of Xpert mtB/rif in diaGnOsinG pulmOnary tB, peripHeral lympH nOde tB and tB meninGitis in smear-pOsitiVe and smear-neGatiVe cHildren, By study and specimen type 28

FigUre 20. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif in detectinG pulmOnary tB, peripHeral lympH nOde tB and tB meninGitis in HiV-pOsitiVe and HiV-neGatiVe cHildren, By study and specimen type 30

FigUre 21. fOrest plOt Of tHe sensitiVity and specificity Of Xpert mtB/rif fOr detectinG resistance tO rifampicin, peripHeral lympH nOde tB and tB meninGitis in cHildren, By study and specimen type 32

XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE ix

Abbreviations

AFB acid-fast bacilli

CI confidence interval

CrI credible interval

CRS composite reference standard

CSF cerebrospinal fluid

DOI Declaration of Interests

DST drug-susceptibility testing

FIND Foundation for Innovative New Diagnostics

FNA fine needle aspiration

GRADE Grading of Recommendations Assessment, Development and Evaluation

HIV human immunodeficiency virus

MDR-TB multidrug-resistant tuberculosis

MGIT mycobacterial growth indicator tube

NAAT nucleic acid amplification test

PCR polymerase chain reaction

PRISMA Preferred Reporting Items for Systematic Reviews and Meta-analyses

rpoB gene encoding for the ß-subunit of the DNA-dependent RNA polymerase of Mycobacterium tuberculosis

QUADAS Quality Assessment of Diagnostic Accuracy Studies

STAG-TB Strategic and Technical Advisory Group for Tuberculosis

TB tuberculosis

WHO World Health Organization

XDR-TB extensively drug-resistant tuberculosis

x XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE

Acknowledgements

This document was prepared by a WHO Steering Group comprising Christopher Gilpin, Karin Weyer, Wayne van Gemert and Fuad Mirzayev (all from WHO’s Global TB Programme) on the basis of con-sensus agreed at an Expert Group Meeting convened by WHO in Geneva during 20–21 May 2013.

The findings and recommendations from the meeting were presented to WHO’s Strategic and Techni-cal Advisory Group for Tuberculosis (STAG-TB) in June 2013 (Annex 3). STAG-TB agreed with the recommendations made by the Expert Group on using Xpert MTB/RIF to diagnose TB and rifampicin resistance in pulmonary and extrapulmonary TB in adults and children, and advised WHO to produce a policy update.

This document was finalized following consideration of all comments and suggestions from the partici-pants of the Expert Group and the members of STAG-TB.

WHO gratefully acknowledges the contributions of the Chairperson of the Expert Group (Holger Schünemann), the members of the Expert Group (Annex 1) and STAG-TB, as well as the WHO staff members who developed this policy update (Annex 2). Karen Steingart (systematic reviewer for pul-monary TB), Claudia Denkinger (systematic reviewer for extrapulmonary TB), Anne Detjen and Anna Mandalakas (systematic reviewers for paediatric TB) and Andrea Pantoja (reviewer for affordability and cost effectiveness of the test ) are thanked for preparing the systematic reviews and presenting their findings to the members of the Expert Group.

Funding from the United States Agency for International Development is gratefully acknowledged through USAID-WHO Consolidated Grant No. GHA-G-00-09-00003/US 2012 0392.

Declarations of Interests

The members of the Expert Group, technical resource consultants and members of STAG-TB completed Declarations of Interests (DOIs). These were reviewed by the WHO Steering Group prior to the meeting of the Expert Group and prior to preparing the current policy update. The review of each DOI assessed whether an interest had been declared and, if so, whether it was insignificant or potentially significant. If the interest was assessed as being significant or potentially significant, the declaration was referred to WHO’s Legal Department, and their advice on the meeting’s procedures was followed. A summary of DOI statements is provided in Annex 4. Selected individuals with intellectual or research involvement in Xpert MTB/RIF, or both, were invited as observers to provide technical input and answer technical questions. These individuals did not participate in the GRADE evaluation process at the meeting nor in the final discussions when recommendations were developed. Also, they were not involved in develop-ing the report of the Expert Group’s meeting, nor in preparing documentation for the STAG-TB or in drafting WHO’s policy update.

XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE xi

Executive summary

The global priorities for tuberculosis (TB) care and control are to improve case-detection and to detect cases earlier, including cases of smear-negative disease which are often associated with coinfection with the human immunodeficiency virus (HIV) and young age, and to enhance the capacity to diagnose multidrug-resistant tuberculosis (MDR-TB). In September 2010, the World Health Organization (WHO) convened an Expert Group to review the evidence on the accuracy of the Xpert MTB/RIF assay (Cepheid, Sunnyvale, CA, United States) for the purpose of formulating recommendations to guide the use of the test. Policy recommendations on using Xpert MTB/RIF were issued by WHO early in 2011,1 supported by an operational how-to document2 and a checklist for implementation at the country level.3

WHO’s current policies and guidance recommend that Xpert MTB/RIF be used as an initial diagnostic test in individuals suspected of having MDR-TB or HIV-associated TB (strong recommendation, moderate quality of evidence). The guidance also provides a conditional recommendation that Xpert MTB/RIF be used as a follow-on test to smear microscopy in settings where MDR-TB or HIV are of lesser concern, especially for further testing of smear-negative specimens. In acknowledgement of the difficulties of obtaining microbiological confirmation of the diagnosis in children, this recommendation generalizes from data on adults to include the use of Xpert MTB/RIF in children.

Since 2010, more than 85 peer-reviewed research papers have been published on using Xpert MTB/RIF to diagnose pulmonary, extrapulmonary and paediatric TB, and studies continue to be performed. Given the amount of additional data on Xpert MTB/RIF that have emerged since 2010, an update of WHO’s policies and guidance was warranted. WHO’s Global TB Programme therefore commissioned three systematic reviews to update and revise the guidance; these reviews examined the utility of Xpert MTB/RIF in diagnosing TB and rifampicin resistance in pulmonary, extrapulmonary and paediatric TB. Published studies on the affordability and cost effectiveness of Xpert MTB/RIF were also reviewed. WHO convened an Expert Group to review the evidence at Les Pensierès, Veyrier-du-Lac, France during 20–21 May 2013. The major findings and recommendations of this Expert Group are summarized below, and a detailed meeting report is available at:http://www.who.int/tb/laboratory/policy_statements/en/

1 Automated real-time nucleic acid amplification technology for rapid and simultaneous detection of tuberculosis and rifampicin resistance: Xpert MTB/RIF system. Policy statement. Geneva, World Health Organization, 2011 (http://whqlibdoc.who.int/publications/2011/9789241501545_eng.pdf).

2 Rapid implementation of the Xpert MTB/RIF diagnostic test. Technical and operational ‘how-to’: practical considerations. Geneva, World Health Organization, 2011 (http://whqlibdoc.who.int/publications/2011/9789241501569_eng.pdf).

3 Prerequisites to country implementation of Xpert MTB/RIF and key action points at country level: checklist. Geneva, World Health Organization, 2011 (http://whqlibdoc.who.int/hq/2011/WHO_HTM_TB_2011.12_eng.pdf).

xii XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE

Using Xpert MTB/RIF to diagnose pulmonary TB and rifampicin resistance in adults

Twenty seven unique studies involving 9 558 participants were included in the review. The reference standards for detecting pulmonary TB were solid culture or liquid culture. The reference standard for detecting rifampicin resistance was phenotypic culture-based drug-susceptibility testing (DST).

When used as an initial diagnostic test replacing smear microscopy, Xpert MTB/RIF achieved an overall pooled sensitivity of 88% (95% credible interval [CrI], 84–92%)4 and a pooled specificity of 99% (95% CrI, 98– 99%) (22 studies, 9008 participants).

When used as an add-on test following a negative smear-microscopy result, Xpert MTB/RIF yielded a pooled sensitivity of 68% (95% CrI, 61–74%) and a pooled specificity of 99% (95% CrI, 98–99%) (23 studies, 7151 participants).

For smear-positive culture-positive TB, the pooled sensitivity of Xpert MTB/RIF was 98% (95% CrI, 97–99%) (23 studies, 1 952 participants); for smear-negative culture-positive TB, the pooled sensitivity was 68% (95% CrI, 61–74%) (23 studies, 7 151 participants).

For people living with HIV, the pooled sensitivity of Xpert MTB/RIF was 79% (95% CrI, 70–86%) (7 studies, 1 789 participants); for people without HIV infection, the pooled sensitivity was 86% (95% CrI, 76–92%) (7 studies, 1470 participants).

When used to detect rifampicin resistance, Xpert MTB/RIF achieved a pooled sensitivity of 95% (95% CrI, 90–97%) (17 studies, 555/2624 total specimens) and a pooled specificity of 98% (95% CrI, 97–99%) (24 studies, 2414 specimens, including true negatives and false positives).

Using Xpert MTB/RIF to diagnose extrapulmonary TB and rifampicin resistance in adults and children

Fifteen published studies and 7 unpublished studies, involving 5 922 samples, were included in the review. The majority of studies (59%) were performed in settings with a high burden of TB. Due to the heterogeneity of sample types included in the studies, prespecified subgroups of samples (pleural fluid, lymph node samples [biopsy and aspirate combined], other tissues and cerebrospinal fluid [CSF]) were included in the meta-analysis with a comparison against culture and against a composite reference standard (CRS). In the different studies, the CRS included some combination of a nucleic acid amplification test (NAAT) other than Xpert MTB/RIF, histology, smear, culture, biochemical testing, presenting signs, or a response to treatment with anti-TB therapy (Table 1).

4 The credible interval (CrI) is the Bayesian equivalent of the confidence interval, or CI.

XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE xiii

Table 1. Meta-analysis of the sensitivity and specificity of Xpert MTB/RIF in diagnosing extrapulmonary TB and rifampicin resistance in adults and children compared against culture as a reference standard as well as against a composite reference standard, by type of extrapulmonary specimen

Specimen type Comparison(No. of studies, No. of samples)

Median (%) pooled sensitivity (pooled 95% CrI)

Median (%) pooled specificity (pooled 95% CrI)

Lymph node tissue and aspirate

Xpert MTB/RIF compared against culture(14 studies, 849 samples)

84.9 (72–92)

92.5 (80–97)

Xpert MTB/RIF compared against a composite reference standard(5 studies, 1 unpublished)

83.7 (74–90)

99.2 (88–100)

Cerebrospinal fluid

Xpert MTB/RIF compared against culture (16 studies, 709 samples)

79.5 (62–90)

98.6 (96–100)

Xpert MTB/RIF compared against a composite reference standard(6 studies, 512 samples)

55.5 (51–81)

98.8 (95–100)

Pleural fluid


43.7 (25–65)

98.1 (95–99)

Xpert MTB/RIF compared against a composite reference standard(7 studies, 698 samples)

17 (8–34)

99.9 (94–100)

Gastric lavage and aspirate


83.8 (66–93)

98.1 (92–100)

Other tissue samples


81.2 (68–90)

98.1 (87–100)

CrI, credible interval; the CrI is the Bayesian equivalent of the confidence interval.

The data for additional sample types (such as, ascitic fluid, pericardial fluid, urine, blood and stool) were limited and therefore not considered in the analysis.

xiv XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE

Using Xpert MTB/RIF to diagnose pulmonary TB and rifampicin resistance in children

Sixteen studies (12 published and 4 unpublished) were included in the review. All studies were performed at higher levels of care, and the children included in the studies were mainly inpatients.

Pulmonary TB was evaluated in 13 studies that included 2603 participants. The overall pooled sensitivity of Xpert MTB/RIF compared against culture as a reference standard in children presumed to have TB was 66% in 10 studies where expectorated sputum or induced sputum was used (95% CrI, 52–77%); the pooled sensitivity was 66% in 7 studies where samples from gastric lavage or aspiration were used (95% CrI, 51–81%). The pooled specificity of Xpert MTB/RIF compared against culture as the reference standard was at least 98%, with narrow confidence intervals.

The pooled sensitivity of Xpert MTB/RIF in culture-negative specimens from children compared against clinical TB used as the reference standard was very low at 4% for samples of expectorated or induced sputum (8 studies), and 15% for samples from gastric lavage or aspiration (3 studies), both sensitivities had wide confidence intervals. It is likely that the apparently poor performance of Xpert MTB/RIF was the result of a reference standard for clinical TB that lacked specificity. The sensitivity of Xpert MTB/RIF to detect rifampicin resistance in specimens from children was 86% (95% CrI, 53–98%).

Affordability and cost effectiveness of using Xpert MTB/RIF to diagnose TB

Twelve published papers were identified that compared the costs of current diagnostic algorithms for diagnosing TB and MDR-TB with the costs of using Xpert MTB/RIF as the initial diagnostic test or as a follow-on test to microscopy . The setting for the majority of analyses was South Africa; two studies in-cluded other countries in sub-Saharan Africa (Botswana, Lesotho, Namibia, Swaziland and Uganda); one study included countries in the former Soviet Union; and one global analysis included all countries. Seven of the 12 studies analysed costs, and 5 were cost–effectiveness analyses. Wide variations in the methods used, the underlying assumptions, and the intended use of Xpert MTB/RIF made a systematic review impossible.

XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE xv

WHO’s policy recommendations

Box 1. Using Xpert MTB/RIF to diagnose pulmonary TB and rifampicin resistance in adults and children

These recommendations should be read in conjunction with the remarks in section 5.1.

• Xpert MTB/RIF should be used rather than conventional microscopy, culture and DST as the initial diagnostic test in adults suspected of having MDR-TB or HIV-associated TB (strong recommendation, high-quality evidence).

• Xpert MTB/RIF should be used rather than conventional microscopy, culture and DST as the initial diagnostic test in children suspected of having MDR-TB or HIV-associated TB (strong recommendation, very low-quality evidence).

• Xpert MTB/RIF may be used rather than conventional microscopy and culture as the initial diagnostic test in all adults suspected of having TB (conditional recommendation acknowledging resource implications, high-quality evidence).

• Xpert MTB/RIF may be used rather than conventional microscopy and culture as the initial diagnostic test in all children suspected of having TB (conditional recommendation acknowledging resource implications, very low-quality evidence).

• Xpert MTB/RIF may be used as a follow-on test to microscopy in adults suspected of having TB but not at risk of MDR-TB or HIV-associated TB, especially when further testing of smear-negative specimens is necessary (conditional recommendation acknowledging resource implications, high-quality evidence).

Box 2. Using Xpert MTB/RIF to diagnose extrapulmonary TB and rifampicin resistance in adults and children

These recommendations should be read in conjunction with the remarks in section 5.2.

• Xpert MTB/RIF should be used in preference to conventional microscopy and culture as the initial diagnostic test for CSF specimens from patients suspected of having TB meningitis (strong recommendation given the urgency for rapid diagnosis, very low-quality evidence).

• Xpert MTB/RIF may be used as a replacement test for usual practice (including conventional microscopy, culture or histopathology) for testing specific nonrespiratory specimens (lymph nodes and other tissues) from patients suspected of having extrapulmonary TB (conditional recommendation, very low-quality evidence).

XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE 1

POLICY UPDATE

1. Background

The global priorities for tuberculosis (TB) care and control are to improve case-detection and to detect cases earlier, including cases of smear-negative disease which are often associated with coinfection with the human immunodeficiency virus (HIV) and young age, and to enhance capacity to diagnose multidrug-resistant tuberculosis (MDR-TB). In September 2010, the World Health Organization (WHO) convened an Expert Group to review the evidence on the Xpert MTB/RIF assay (Cepheid, Sunnyvale, CA, United States) in order to formulate recommendations on using the test. Policy recommendations on using the Xpert MTB/RIF assay were issued by WHO early in 20115, supported by an operational how-to document6 and a checklist for implementation at the country level7.

WHO’s current policies and guidance recommend that Xpert MTB/RIF be used as an initial diagnostic test in individuals suspected of having MDR-TB or HIV-associated TB (strong recommendation, moderate quality of evidence). The guidance also provides a conditional recommendation that Xpert MTB/RIF be used as a follow-on test to smear microscopy in settings where MDR-TB or HIV are of lesser concern, especially for further testing of smear-negative specimens. In acknowledgement of the difficulties of obtaining microbiological confirmation of the diagnosis in children, the recommendation generalizes from data on adults to include the use of Xpert MTB/RIF in children.

Extrapulmonary TB accounts for about 25% of all cases of TB and an even higher

percentage of cases in children and in people who are immunocompromised. Diagnosing extrapulmonary TB is often challenging, requiring the clinician to obtain specimens for microscopy, culture and histopathology from the suspected sites of involvement. However, the availability of these tests is limited and the need for alternative tests to use to diagnose TB in nonrespiratory samples is great. In 2011, the global burden of TB in children was estimated to be 500 000 cases, representing approximately 6% of all cases of TB. However, in all likelihood this burden has been underestimated due to the difficulties associated with obtaining microbiological confirmation of the diagnosis of TB in children.

The Xpert MTB/RIF assay remains the only fully automated cartridge-based real-time DNA-based test that can detect both TB and resistance to rifampicin in less than 2 hours, and it is the the only mature technology representing a new generation of automated platforms for molecular diagnosis.

Since 2010, at least 85 peer-reviewed research papers have been published about using Xpert MTB/RIF to diagnose pulmonary, extrapulmonary and paediatric TB, and studies continue to be performed. Given the amount of additional data on Xpert MTB/RIF that have emerged since 2010, an update of WHO’s policies and guidance was warranted. WHO’s Global TB Programme therefore commissioned three systematic reviews to update and revise the guidance; these reviews assessed the utility of Xpert MTB/RIF

5 Automated real-time nucleic acid amplification technology for rapid and simultaneous detection of tuberculosis and rifampicin resistance: Xpert MTB/RIF system. Policy statement. Geneva, World Health Organization, 2011 (http://whqlibdoc.who.int/publications/2011/9789241501545_eng.pdf).

6 Rapid implementation of the Xpert MTB/RIF diagnostic test. Technical and operational ‘how-to’: practical considerations. Geneva, World Health Organization, 2011 (http://whqlibdoc.who.int/publications/2011/9789241501569_eng.pdf).

7 Prerequisites to country implementation of Xpert MTB/RIF and key action points at country level: checklist. Geneva, World Health Organization, 2011 (http://whqlibdoc.who.int/hq/2011/WHO_HTM_TB_2011.12_eng.pdf).

2 XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE

for diagnosing TB and rifampicin resistance in pulmonary, extrapulmonary and paediatric TB. Published studies on the affordability and cost effectiveness of Xpert MTB/RIF were also reviewed. WHO convened an Expert Group to review the evidence at Les Pensierès, Veyrier-du-Lac, France, during 20–21 May 2013.

Xpert MTB/RIF is an automated polymerase chain reaction (PCR) test (that is, a molecular test) utilizing the GeneXpert platform (Cepheid, Sunnyvale, CA, United States). Xpert MTB/RIF is a single test that can detect both Mycobacterium tuberculosis complex and rifampicin resistance within 2 hours after starting the assay, with minimal hands-on technical time (Figure 1). Unlike conventional nucleic acid amplification tests (NAATs), in Xpert MTB/RIF sample processing, PCR amplification and detection are integrated into a single self-enclosed test unit, which is the Xpert MTB/RIF cartridge. Following sample loading, all steps in the assay are automated and

contained within the cartridge. In addition, the assay’s sample reagent, used to liquefy sputum, is tuberculocidal (that is, it has the ability to kill TB bacteria), which largely eliminates concerns about biosafety during the test procedure. These features allow the technology to be taken out of a central laboratory or reference laboratory and to be used nearer to patients. However, Xpert MTB/RIF requires an uninterrupted and stable electrical power supply, temperature control and yearly calibration of the instrument’s modules8.

The test procedure may be used directly on clinical specimens, either fresh sputum samples or sputum pellets (also called sputum sediment), which are obtained after decontaminating and concentrating the sputum. In both cases, the test material is combined with the reagent, mixed by hand or vortex, and incubated at room temperature for 15 minutes. After incubation, 2 ml of the treated sample are transferred to the cartridge, and the run is initiated.

Figure 1. Steps in using the Xpert MTB/RIF assaya

a Figure used with permission from the Foundation for Innovative New Diagnostics (FIND).

8 Rapid implementation of the Xpert MTB/RIF diagnostic test. Technical and operational ‘how-to’:practical considerations. Geneva; World Health Organization, 2011(http://whqlibdoc.who.int/publications/2011/9789241501569_eng.pdf).


Xpert MTB/RIF uses molecular beacon technology to detect rifampicin resistance. Molecular beacons are nucleic acid probes that recognize and report the presence or absence of the normal, rifampicin-susceptible wild-type sequence of the rpoB gene of TB. Five different coloured beacons are used, each covering a separate nucleic acid sequence within the amplified rpoB gene. When a beacon binds to the matching sequence, it fluoresces (or lights up), which indicates the presence of one of the gene sequences that is characteristic of rifampicin-susceptible TB. If a beacon fails to bind

to the matching sequence or if binding is delayed, the sample is potentially resistant to rifampicin. The number of positive beacons and the timing of their detection (when the fluorescent signal rises above a predetermined baseline cycle threshold), as well as the results of sample processing controls, allow the test to distinguish among the following results: no TB; TB detected, rifampicin resistance detected; TB detected, no rifampicin resistance detected; TB detected, rifampicin resistance indeterminate; and an invalid result.

2. Methods

2.1 Evidence synthesis

In May 2013, a guideline development group (referred to as the Expert Group in this document) was convened by WHO’s Global TB Programme to assess the data on Xpert MTB/RIF with a view to updating WHO’s 2011 policy recommendations on its use. WHO commissioned three systematic reviews on the use of Xpert MTB/RIF to diagnose pulmonary, extrapulmonary and paediatric TB and to detect rifampicin resistance, as well as a review of the affordability and cost effectiveness of Xpert MTB/RIF.

In accordance with WHO’s standards for assess-ing evidence when formulating policy recommen-dations, the Grading of Recommendations As-sessment, Development and Evaluation (GRADE)

system9 was used for the evidence synthesis pro-cess to provide a systematic, structured framework for evaluating both the accuracy of the test and the test’s impact on patients and public health. The evaluations used the GRADE system to deter-mine the quality of the evidence and provide in-formation on the strength of the recommendations using a priori questions (that is, PICO questions) agreed by the Expert Group. PICO refers to the following four elements that should be included in questions that govern a systematic search of the evidence: the Population targeted by the action or intervention; the Intervention; the Comparator; and the Outcome.The PICO questions for each review are given in Box 3.

9 Schünemann HJ et al. Grading quality of evidence and strength of recommendations for diagnostic tests and strategies.BMJ, 2008, 336:1106-1110.


Box 3. PICO questions for the four systematic reviews evaluating the accuracy of the Xpert MTB/RIF assay in diagnosing TB

Review 1. Updated systematic review: Xpert MTB/RIF for diagnosis of pulmonary tuberculosis and rifampicin resistance in adults1. What is the diagnostic accuracy of Xpert MTB/RIF for detection of pulmonary TB in adults, where Xpert MTB/RIF is used as a replacement test for smear microscopy? 2. What is the diagnostic accuracy of Xpert MTB/RIF for detection of pulmonary TB in adults, where Xpert MTB/RIF is used as an add-on test following a negative smear-microscopy result?3. What is the diagnostic accuracy of Xpert MTB/RIF for detection of smear-positive pulmonary TB in adults?4. What is the diagnostic accuracy of Xpert MTB/RIF for detection of smear-negative (culture-positive) pulmonary TB in adults?5. What is the diagnostic accuracy of Xpert MTB/RIF for detection of pulmonary TB in people living with HIV (adults)?6. What is the diagnostic accuracy of Xpert MTB/RIF for detection of pulmonary TB in adults without HIV infection?7. What is the diagnostic accuracy of Xpert MTB/RIF for detection of rifampicin resistance, where Xpert MTB/RIF is used as an initial test replacing phenotypic culture-based drug-susceptibility testing?

Review 2. Systematic review: Xpert MTB/RIF for diagnosis of tuberculosis and detection of rifampicin resistance on nonrespiratory samples (extrapulmonary TB)1. What is the diagnostic accuracy of Xpert MTB/RIF overall compared with culture for nonrespiratory specimens, where Xpert MTB/RIF is used as a replacement test for usual practice?a

2. What is the diagnostic accuracy of Xpert MTB/RIF overall compared with a combined clinical and laboratory reference standard for nonrespiratory specimens, where Xpert MTB/RIF is used as a replacement test for usual practice? 2a. What is the diagnostic accuracy of Xpert MTB/RIF for lymph node fluid and tissue, where Xpert MTB/RIF is used as a replacement test for usual practice?2b. What is the diagnostic accuracy of Xpert MTB/RIF for pleural fluid, where Xpert MTB/RIF is used as a replacement test for usual practice?2c. What is the diagnostic accuracy of Xpert MTB/RIF for cerebrospinal fluid, where Xpert MTB/RIF is used as a replacement test for usual practice?2d. What is the diagnostic accuracy of Xpert MTB/RIF for gastric fluid, where Xpert MTB/RIF is used as a replacement test for usual practice? 2e. What is the diagnostic accuracy of Xpert MTB/RIF for tissue samples, where Xpert MTB/RIF is used as a replacement test for usual practice? 3. What is the diagnostic accuracy of Xpert MTB/RIF for detection of rifampicin resistance in nonrespiratory specimens, where Xpert MTB/RIF is used as an initial test replacing phenotypic culture-based drug-susceptibility testing?

Review 3. Systematic review: Xpert MTB/RIF for diagnosis of tuberculosis and rifampicin resistance in children1. What is the diagnostic accuracy of Xpert MTB/RIF for detection of TB in children compared with culture, where Xpert MTB/RIF is used as a replacement test for usual practice?b


2. What is the diagnostic accuracy of Xpert MTB/RIF for detection of TB in children compared with a combined clinical and laboratory reference standard, where Xpert MTB/RIF is used as a replacement test for usual practice? 3. What is the diagnostic accuracy of Xpert MTB/RIF for detection of TB in children, where Xpert MTB/RIF is used as an add-on test following a negative smear-microscopy result? 4. What is the diagnostic accuracy of Xpert MTB/RIF compared with smear microscopy for detection of TB in children?5. What is the diagnostic accuracy of Xpert MTB/RIF for detection of rifampicin resistance in children, where Xpert MTB/RIF is used as an initial test replacing phenotypic culture-based drug-susceptibility testing?6. What is the diagnostic accuracy of Xpert MTB/RIF for detection of peripheral lymph node TB in children, where Xpert MTB/RIF is used as a replacement test for usual practice? 7. What is the diagnostic accuracy of Xpert MTB/RIF for detection of TB meningitis in children, where Xpert MTB/RIF is used as a replacement test for usual practice?

Review 4. Systematic review: Affordability, cost effectiveness and resource implications for Xpert MTB/RIF scale up1. For which diagnostic and screening algorithms is the Xpert MTB/RIF assay an affordable and a cost-effective intervention?

a Most analyses were performed using two reference standards: culture (the current reference standard) and a combined clinical and laboratory reference standard chosen by the study’s authors (given the technical limitations of using culture for diagnosis).b Given the difficulties of diagnosing TB in children, usual practice refers to customary practice in the field, which may vary from setting to setting. The usual practice for children (aged 0–15 years) suspected of having intrathoracic TB (that is, pulmonary, pleural, and mediastinal or hilar lymph node TB) normally requires bacteriological confirmation through examination of sputum (obtained by expectoration, gastric washings, or induction) for smear microscopy and culture. In the event of negative bacteriological results, a diagnosis of TB may be based on the presence of abnormalities consistent with TB on chest radiography, a history of exposure to an infectious case, evidence of TB infection (that is, a positive tuberculin skin test or interferon-g release assay) and clinical findings suggestive of TB. For children suspected of having extrapulmonary TB, appropriate specimens from the suspected sites of involvement may be obtained for microscopy, and for culture and histopathological examination.

Three systematic reviews were conducted according to the standards outlined by the Cochrane Collaboration in the Cochrane handbook.10 A comprehensive search of the following databases was performed: Cochrane Infectious Diseases Group Specialized Register, MEDLINE, Embase, Thomson Reuters (formerly ISI) Web of Knowledge, Medion, LILACS, BIOSIS and Scopus. Searches of the metaRegister of Controlled Trials and the search portal of WHO’s International Clinical Trials Registry were also performed to identify any continuing trials. Searches included all studies without restrictions on language.

A comprehensive search of the literature for studies on the cost effectiveness and affordability of Xpert MTB/RIF included PubMed, the Health Economics Evaluation Database, the United Kingdom’s National Health Service Economic Evaluation Database, the Cost-effectiveness Analysis Registry and the Research Papers in Economics database.

Where feasible, meta-analysis was used to summarize the results of independent studies, and these results have been displayed as forest plots. Where meta-analysis was not feasible due to heterogeneity, the evidence has been presented in a narrative synthesis.

10 Higgins JPT, Green S, eds. Cochrane handbook of systematic reviews for interventions, version 5.1.0. Cochrane Collaboration, 2011 (available at http://www.cochrane-handbook.org).


Each reviewer prepared GRADE evidence profiles for each PICO question. GRADE evidence profiles were prepared to assess the diagnostic accuracy of Xpert MTB/RIF for each systematic review. In systematic reviews that assess the accuracy of diagnostic tests, the choice of an optimal reference standard is critical, since the reference standard is used to determine the presence or absence of the target condition. The reference standards used for the different systematic reviews are described below.

• Using Xpert MTB/RIF to detect pulmonary TB and rifampicin resistance: the reference standards used were conventional culture and drug-susceptibility testing (DST). Culture using either solid media or commercial liquid media, as recommended by WHO, was considered to be an acceptable reference standard. The reference methods for DST for rifampicin resistance were those recommended by WHO, and these included molecular line probe assays.

• Using Xpert MTB/RIF to detect extrapulmonary TB in adults and children: the reference standard used was either conventional culture (as described above) or a composite reference standard (CRS) defined by the authors of the individual studies that were included in the systematic review; in the different studies the CRS included some combination of NAAT other than Xpert MTB/RIF, histology, smear, culture, biochemical testing, presenting signs, or a response to treatment with anti-TB therapy.

• Using Xpert MTB/RIF to detect TB and rifampicin resistance in children: the reference standard used was either conventional culture (as described above) or a clinical TB reference standard, recognizing the limitations of mycobacterial culture in children. Children were categorized as positive using the clinical TB standard if they had started anti-TB therapy as a result of a clinical diagnosis of

TB. This broad clinical reference standard was chosen in order to accommodate the heterogeneous methods and clinical definitions used in the studies. Children assigned to the group clinical not TB (that is, negative according to the reference standard clinical TB) either (1) did not have another diagnosis assigned, or (2) did not start anti-TB treatment but nonetheless improved, or their condition did not worsen after at least 1 month of follow-up after enrolment.

Using the GRADE framework, results for sensitivity and specificity were used as proxy measures for outcomes seen as important to patients; these outcomes were based on the relative importance or impact of false-positive and false-negative results. Poor sensitivity would result in false-negative results so that patients with TB or MDR-TB would be missed, and this would have negative consequences in terms of morbidity, mortality and transmission of disease. Poor specificity would result in false-positive results so that patients without TB or MDR-TB would be prescribed unnecessary treatment, and this would have negative consequences that might include serious adverse events related to treatment with second-line anti-TB agents.

Rates for true positives, true negatives, false positives and false negatives were calculated based on pretest probabilities – that is, an assumed prevalence of TB of 2.5%, 5% and 10% among patients suspected to have TB who were being screened, and an assumed prevalence of rifampicin resistance of 5% and 15% (as a proxy for MDR-TB) among patients with confirmed TB.

The evaluation of the impact on patients was based on a balance among the following values:

• true positives – the benefit to patients from rapid diagnosis and treatment;

• true negatives – the benefit to patients who would be spared unnecessary treatment; the benefit of reassurance and alternative diagnosis;


• false positives – the likelihood of anxiety and morbidity caused by additional testing, unnecessary treatment, or both; the chance that a false positive may halt further diagnostic evaluation;

• false negatives – the increased risk of morbidity and mortality, and the continued risk of community transmission of TB.

For each outcome, the quality of evidence according to GRADE was initially regarded as high since all studies were cross-sectional or cohort studies, prospectively enrolling patients suspected of having TB or MDR-TB. The quality of the evidence and the limitations of the studies were assessed using six GRADE criteria: (1) study design, (2) risk of bias, (3) directness, (4) inconsistency, (5) imprecision, and (6) publication or reporting bias.

Each review used the Quality Assessment of Diagnostic Accuracy Studies (QUADAS -2) tool to appraise the studies.11 This tool consists of four domains: patient selection, index test, reference standard, and flow and timing.

2.2 Expert Group meeting

PICO questions were drafted by the WHO Steering Group and were presented to the Expert Group for discussion and modification. The Steering Group also prepared an initial list of relevant outcomes, including desirable effects and undesirable effects, and requested the Expert Group to identify any other important outcomes.

A webinar was conducted with members of the Expert Group prior to the meeting to refine and finalize the proposed outcomes seen as important to patients, and to rate their relative importance. The following outcomes for each PICO question were determined, and the ratings of their importance were unanimously agreed by the Expert Group:

• critical outcomes – diagnostic accuracy as reflected by true-positive, true-negative, false-positive and false-negative results; time to diagnosis;

• important outcome – cost.

The meeting was chaired by an expert in evidence synthesis. Decisions were based on consensus. Concerns raised by members were noted and included in the final report of the meeting. The detailed report was prepared by the Steering Group; the report went through several iterations before being finally signed off by members of the Expert Group.

2.3 External review

The findings and recommendations from the Expert Group’s meeting were presented in June 2013 to WHO’s Strategic and Technical Advisory Group for TB (STAG-TB), members of which served as the external review group. STAG-TB agreed with the Expert Group’s recommendations, and advised WHO to develop and disseminate an updated policy on using Xpert MTB/RIF to diagnose TB and rifampicin resistance in pulmonary and extrapulmonary TB in adults and children. STAG-TB also recommended that WHO should continue its global coordination and implementation plan aimed at scaling up the use of the Xpert MTB/RIF assay.12

2.4 Preparing the policy update

The draft policy update, which was based on the consensus recommendations made by the Expert Group, was subsequently prepared by the Steering Group and circulated to the Expert Group and members of STAG-TB following an iterative process similar to that described above.

11 Whiting PF et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Annals of Internal Medicine, 2011, 155:529-536.12 Strategic and Technical and Advisory Group for Tuberculosis (STAG-TB): report of the 13th meeting. Geneva, World Health Organization, 2013 (WHO/HTM/TB/2013.09) available at http://www.who.int/tb/advisory_bodies/STAG_report2013.pdf)


3. Scope

This document replaces WHO’s 2011 policy statement on Xpert MTB/RIF13, and provides a pragmatic summary of the updated evidence and recommendations on using Xpert MTB/RIF to diagnose pulmonary and extrapulmonary TB, and to detect rifampicin resistance in adults and children. It should be read in conjunction with the detailed findings from the 2013 report of the Expert Group’s meeting14 and the framework for implementing TB diagnostics. These documents are available at http://www.who.int/tb/dots/laboratory/policy/en; they provide guidance on implementing the diagnostic tools and methods approved by WHO within the context of a country’s infrastructure, resources, epidemiology of TB and MDR-TB, and on TB policy reform. This policy update will be supported by the second edition of WHO’s Xpert MTB/RIF implementation manual (a publication derived from the guidance on policy).15

None of the existing tools for diagnosing TB are mutually exclusive, and they can be implemented in various combinations in screening and diagnostic algorithms, which can be tailored to be highly specific to each country’s settings and resources. Therefore, input from experts working in TB laboratories is needed to define the most cost-effective and efficient algorithms for individual countries; these algorithms should be guided by WHO’s standards and procedures, and

implemented within a framework of integrated activities aimed at strengthening laboratories.

This policy guidance should be used to support the implementation of Xpert MTB/RIF technology to diagnose TB and detect rifampicin resistance within TB and TB–HIV programmes. The policies are intended to be used by managers and laboratory directors working in these programmes in coordination with external laboratory consultants, donor agencies, technical advisers, laboratory technicians, procurement officers for laboratory equipment, service providers in the private sector, relevant government sectors, and implementation partners that are involved in country-level strengthening of TB laboratories. Individuals responsible for programme planning, budgeting, mobilizing resources and implementing training activities for TB and TB–HIV diagnostic services may also benefit from this document.

3.1 Date of review: 2017

Additional data from implementation sites will be reviewed annually, and the guidance will continue to be refined based on more extensive field evaluations of the new technology that will be conducted after implementation; these evaluations will include country-specific cost–effectiveness and cost–benefit analyses.

13 Automated real-time nucleic acid amplification technology for rapid and simultaneous detection of tuberculosis and rifampicin resistance: Xpert MTB/RIF system. Policy statement. Geneva, World Health Organization, 2011 (http://whqlibdoc.who.int/publications/2011/9789241501545_eng.pdf).14 Using the Xpert MTB/RIF assay to detect pulmonary and extrapulmonary tuberculosis and rifampicin resistance in adults and children: Expert Group meeting report. Geneva, World Health Organization, 2013 (available at http://www.who.int/tb/laboratory/policy_statements/en/)15 Xpert MTB/RIF implementation manual. Technical and operational ‘how-to’: practical considerations. 2nd ed. Geneva, World Health Organization, 2014 (available at http://www.who.int/tb/laboratory/policy_statements/en/)


4. Evidence base for policy formulation

4.1 Using Xpert MTB/RIF to diagnose pulmonary TB and rifampicin resistance in adults

A Cochrane review published 31 January 201316 evaluated 18 studies (Figure 2). An additional literature search was performed on

7 February 2013, which identified 9 additional studies (Figure 3). From the combined literature searches, 27 relevant studies in 36 study centres (26 published studies and 1 unpublished study) were identified and evaluated.

Figure 2. Selection of studies evaluating the accuracy of Xpert MTB/RIF in diagnosing pulmonary TB and rifampicin resistance in adults: flow diagram of studies identified by the initial literature searches

139 records identified through database search

25 September 2011

5 additional records identified through other sources

137 records screened after duplicates removed

60 full-text articles assessed for eligibility

18 studies included in qualitative synthesis

15 studies included in meta-analysis of TB detection; 11 studies

included in meta-analysis of detection of rifampicin resistance

81 records identified through database search 15 December 2011; no new

records identified

42 full-text articles excludedReasons• Abstract only: 10• Case–control study: 1• Correspondence with authors: 1• Cost-effectiveness analysis: 1• Could not obtain: 1• Duplicate data: 1• Editorial or comment article: 13 • Extrapulmonary TB: 5• Paediatric TB: 1• Review article: 6• Technical article: 2

77 records excluded based on title and abstract

16 Steingart KR et al. Xpert® MTB/RIF assay for pulmonary tuberculosis and rifampicin resistance in adults. Cochrane Database of Systematic Reviews, 2013, (1):CD009593 (doi:10.1002/14651858.CD009593.pub2).


Figure 3. Selection of studies evaluating the accuracy of Xpert MTB/RIF in diagnosing pulmonary TB and rifampicin resistance in adults: flow diagram of studies identified by the updated literature search

343 records identified through database search

7 February 2013

2 additional records identified through other sources

130 records screened after duplicates removed

54 full-text articles assessed for eligibility

9 studies included in qualitative synthesis

7 studies included in quantitative synthesis

(meta-analysis)

45 full-text articles excludedReasons• Extrapulmonary TB: 7• Paediatric TB: 6• Correspondence with authors: 5• Duplicate data: 4• Impact study: 4• Study did not enrol patients

suspected of having TB: 3• Technical article: 3• Case–control study: 2• Case report: 2• Editorial or comment article: 2• Data insufficient: 2• Treatment monitoring: 2• Abstract: 1• Reference standard not satisfied: 1• Relevance: 1

76 records excluded based on title and abstract

To assess the accuracy of Xpert MTB/RIF in detecting TB, data were reviewed from 27 studies that included 9558 participants; 22 of the studies, involving 9008 participants, were included in the meta-analysis. Five studies that enrolled primarily smear-positive or smear-

negative patients were excluded. The pooled sensitivity of Xpert MTB/RIF for detecting TB was 88% (95% credible interval [CrI], 84–92%); the pooled specificity for detecting TB was 99% (95% CrI, 98–99%) (Figure 4).17

17 The credible interval (CrI) is the Bayesian equivalent of the confidence interval (CI).


Figure 4. Forest plot of the sensitivity and specificity of Xpert MTB/RIF for detecting pulmonary TB in 27 studies (36 study centres)a

TP, true positive; FP, false positive; FN, false negative; TN, true negative; CI, confidence interval.

a The figure shows the estimated sensitivity and specificity of each study (blue square) and its 95% CI (black horizontal line). Values for test results are the number of each type of result (true positive, false positive, false negative, true negative).

4.1.1 Using Xpert MTB/RIF as a replacement test for smear microscopy

Twenty-one studies (8 880 participants) provided data that compared the sensitivity of Xpert MTB/RIF with smear microscopy. For smear microscopy, the pooled sensitivity was 65% (95% CrI, 57–

72%). For Xpert MTB/RIF, the pooled sensitivity was 88% (95% CrI, 84–92%). Therefore, in comparison with smear microscopy, Xpert MTB/RIF increased TB detection among culture-confirmed cases by 23% (95% CrI, 15–32%) (Table 2).


Table 2. Pooled sensitivity and specificity of the Xpert MTB/RIF assay for detecting pulmonary TB and rifampicin resistance

Type of analysis (No. of studies, No. of participants)

Median (%) pooled sensitivity (95% CrI)

Median (%) pooled specificity (95% CrI)

Xpert MTB/RIF used as an initial test for TB detection replacing microscopy (22, 9008)

88 (84–92) 99 (98–99)

Xpert MTB/RIF used as an add-on test for TB detection following a negative smear-microscopy result (23, 7151)

68 (61–74) 99 (98–99)

Xpert MTB/RIF used as an initial test for detecting rifampicin resistance replacing conventional drug-susceptibility testing as the initial testa

95 (90–97) 98 (97–99)


a The pooled sensitivity estimates and specificity estimates for detecting rifampicin resistance were determined separately by univariate analyses. The pooled sensitivity analysis included 17 studies (555 participants); the pooled specificity analysis included 24 studies (2414 participants).

4.1.2 Using Xpert MTB/RIF as an add-on test following microscopy

When Xpert MTB/RIF was used as an add-on test following a negative smear-microscopy result (23 studies, 7 151 participants), the pooled sensitivity was 68% (95% CrI, 61–74%); the pooled specificity was 99% (95% CrI, 98– 99%). In other words, 68% of smear-negative culture-confirmed cases of TB were detected using Xpert MTB/RIF following smear microscopy, which increased case-detection by 68% (95% CrI, 61–74%) in this group (Table 2).

4.1.3 Using Xpert MTB/RIF to detect smear-positive culture-positive TB

There was little variation in the sensitivity estimates (95–100%) for studies reporting data on smear-positive cases (24 studies, 33 study centres, 2071 participants). In the meta-analysis, the pooled sensitivity for smear-positive culture-positive TB was very high at 98% (95% CrI, 97–99% (23 studies, 1952 participants). Estimates of the pooled specificity of Xpert MTB/RIF were not performed because participants in the smear-positive subgroup were considered to be true cases of TB.

4.1.4 Using Xpert MTB/RIF to detect smear-negative culture-positive TB

Twenty four studies (33 study centres, 7 247 participants) reported data on smear-negative cases. There was considerable variability in the sensitivity estimates (range, 43–100% ). Specificity estimates showed less variation (range, 86–100%). The meta-analysis included 23 studies that allowed for direct comparison between subgroups that were smear-positive and those that were smear-negative. The pooled estimate of sensitivity for smear-negative culture-positive TB was 68% (95% CrI, 61–74%).

4.1.5 Using Xpert MTB/RIF to detect pulmonary TB in HIV-negative and HIV- positive individuals

Nine studies (18 study centres, 2555 participants) reported data on HIV-negative individuals, and 10 studies (16 study centres, 2378 participants) reported on HIV-positive individuals (Figure 5). There was variability in sensitivity in both the HIV-negative subgroup (range, 56–100%) and HIV-positive subgroup (range, 0–100%). The small number of participants in several studies may have contributed to some of the variability. Specificity varied less than sensitivity in both subgroups: from 96% to 100% in the HIV-negative subgroup, and from 92% to 100% in the HIV-positive subgroup.


Figure 5. Forest plots of the sensitivity and specificity of Xpert MTB/RIF for detecting pulmonary TB in HIV-negative individuals suspected of having TB (9 studies, 18 study centres) and HIV-positive individuals suspected of having TB (10 studies, 16 centres)a

TP, true positive; FP, false positive; FN, false negative; TN, true negative; CI, confidence interval. a The figure shows the estimated sensitivity and specificity of each study (blue square) and its 95% CI (black horizontal line). Values for test results are the number of each type of result (true positive, false positive, false negative, true negative).

The meta-analysis included 7 studies that provided data on both HIV-negative individuals (1470 participants) and HIV-positive individuals (1789 participants). The pooled sensitivity for the HIV-negative subgroup was 86% (95% CrI, 76–92%); for the HIV-positive subgroup it was 79% (95% CrI, 70–86%) .

The corresponding pooled specificities were similar: for the HIV-negative subgroup the specificity was 99% (95% CrI, 98–100%); for the HIV-positive subgroup it was 98% (95% CrI, 96–99%). When adjusting for the percentage of smear-positive patients in each study, the impact

of the HIV covariate decreased, suggesting that some of the differences between the HIV-positive subgroup and the HIV-negative subgroup could be attributed to differences in smear status.

Five studies reported data from which it was possible to assess the accuracy of Xpert MTB/RIF in HIV-positive individuals with culture-positive smear-negative TB. The sensitivity of Xpert MTB/RIF in HIV-positive individuals with smear-negative culture-positive TB ranged from 43% to 93%; in HIV-positive individuals with smear-positive culture-positive TB it ranged from 91% to 100%. Data were sufficient to perform a univariate meta-


analysis to assess the sensitivity of Xpert MTB/RIF. Among people living with HIV, the pooled sensitivity of Xpert MTB/RIF for smear-negative culture-positive TB was 61% (95% CrI, 42–79%) compared with 97% (95% CrI, 91–99%) for smear-positive culture-positive TB; this was a statistically significant result (data not shown). Hence, among people who are coinfected with HIV and TB, those with smear-positive disease were more likely to be diagnosed with TB using Xpert MTB/RIF than those with smear-negative disease.

4.1.6 Using Xpert MTB/RIF to detect rifampicin resistance

Of the 27 studies, 24 studies (33 study centres, 2 969 participants) provided data on detecting

rifampicin resistance, and included 555 rifampicin-resistant specimens. Figure 6 shows the forest plots of sensitivity and specificity for this analysis. Although there was variability in the estimates of sensitivity (range, 33–100%), in general the poorer estimates of sensitivity were related to study centres that had only a low number of rifampicin-resistant specimens. There was less variability in specificity (range, 83–100%). The pooled sensitivity by univariate analysis was 95% (95% CrI, 90–97%); the pooled specificity was 98% (95% CrI, 97–99%). The pooled sensitivity and specificity were the same when bivariate analysis was used for the subset of studies that provided data on both sensitivity and specificity (17 studies, 2624 participants).

Figure 6. Forest plots of the sensitivity and specificity of Xpert MTB/RIF for detecting rifampicin resistance when Xpert MTB/RIF was used as an initial test replacing phenotypic culture-based drug-susceptibility testing in 24 studies (33 study centres)a


a The figure shows the estimated sensitivity and specificity of each study (blue square) and its 95% CI (black horizontal line). Values for test results are the number of each type of result (true positive, false positive, false negative, true negative).


4.1.7 Effect of the version of the Xpert MTB/RIF assay

The basis in the Xpert MTB/RIF system for detecting rifampicin resistance is the difference between the first M. tuberculosis-specific beacon or probe (the early-cycle threshold) and the last beacon (the late-cycle threshold). This difference is referred to as the delta-cycle threshold. The original configuration of the Xpert MTB/RIF system reported rifampicin resistance when the delta-cycle threshold was higher than 3.5 cycles; rifampicin sensitivity was reported when the delta-cycle threshold was 3.5 cycles or lower (using the Xpert MTB/RIF G1 cartridge). After May 2010, the manufacturer modified the delta-cycle threshold cut-off value to improve the specificity of Xpert MTB/RIF in detecting rifampicin resistance (using the Xpert MTB/RIF G2 and G3 cartridges). Another modification was implemented in late 2011 (using the Xpert MTB/RIF G4 cartridge), which changed the molecular beacon sequence of Probe B to improve detection of rifampicin resistance when there were fluctuations in the annealing temperature. Changes to fluidics and software virtually eliminated the signal-loss detection error (known as the 5011 error), and allowed high sensitivity and specificity to be maintained when detecting TB and rifampicin resistance. These enhancements to the assays were considered to be part of a routine process of product improvement. Cepheid, the Foundation for Innovative New Diagnostics (FIND) and the University of Medicine and Dentistry of New Jersey will continue to monitor the clinical performance of the Xpert MTB/RIF test. By 2013, the G4 cartridges were the only type of cartridge available.

The effect of the version of Xpert MTB/RIF on the sensitivity and specificity for detecting rifampicin resistance was investigated. The pooled sensitivity for studies using Xpert MTB/RIF G2, G3 or G4 cartridges (13 studies) was 93% (95% CrI, 87–97%); for studies using the Xpert MTB/RIF G1 cartridge (4 studies) it was 97% (95%

CrI, 91–99%) . The pooled specificity for studies using Xpert MTB/RIF G2, G3 or G4 cartridges (15 studies) was 98% (95% CrI, 96–99%); for studies using the Xpert MTB/RIF G1 cartridge (4 studies) it was 99% (95% CrI, 98–100%). The overlapping credible intervals indicate that there was no statistically significant difference in the estimates of either sensitivity or specificity for the Xpert MTB/RIF G1 cartridge when compared with later versions of the cartridge.

4.1.8 Accuracy of the Xpert MTB/RIF G4 cartridge

Two studies used the Xpert MTB/RIF G4 cartridge and provided data suitable to determine specificity. One study observed a specificity of 100% (10/10 tests) (95% confidence interval [CI], 69–100%); the second study reported a specificity of 95% (42/44 tests) (95% CI, 85–99%) (Figure 6).

FIND evaluated the diagnostic accuracy of the G4 cartridge18 in a study that tested 233 archived sputum specimens that had been stored in Borstel, Germany, and were from individuals suspected of having TB; additionally, there were 184 frozen sediments from Lima, Peru, that were positive for acid-fast bacilli (AFB), as well as frozen sputum specimens from a further 231 patients consecutively enrolled from Baku, Azerbaijan. All of the samples were shipped to and tested in Germany using the G4 cartridge. Fresh sputum samples from 30 patients were tested using both the G3 cartridge and the G4 cartridge in Kampala, Uganda; a further 218 specimens were evaluated using both the G3 cartridge and the G4 cartridge in Cape Town, South Africa.

The reference standard used across all sites included at least one Löwenstein–Jensen culture and at least one culture using the BACTEC MGIT (mycobacterial growth indicator tube) 960 Mycobacterial Detection System (Becton Dickinson, Franklin Lakes, NJ, United States), with M. tuberculosis species confirmed using Capilia (Tauns Laboratories, Shizuoka, Japan), GenoType

18 Report: performance of Xpert MTB/RIF version G4 assay. Geneva, Foundation for Innovative New Diagnostics, 2011 (available at: http://www.stoptb.org/wg/gli/assets/documents/map/findg4cartridge.pdf).


MTBDRplus (Hain Lifescience, Nehren, Germany) or GenoType Mycobacterium CM/AS (Hain Lifescience, Nehren, Germany). Conventional testing for rifampicin resistance was performed using either the Löwenstein–Jensen proportion method or the BACTEC MGIT 960 and, in a few cases, using only the Genotype MTBDRplus assay. Genetic sequencing was performed on results discordant between Xpert MTB/RIF and conventional DST. Six patients (smear-negative and culture-negative) were started on anti-TB treatment and excluded from the analysis. Genetic sequencing was used to resolve discordant results to determine sensitivity and specificity.

The overall sensitivity for rifampicin-resistant TB was 98.9% (87/88 tests) (95% CI, 93.8–99.8%); the overall specificity for rifampicin-sensitive TB was 99.8% (433/434 tests) (95% CI, 98.7– 100.0%). For four cases in which results were discordant (Xpert MTB/RIF identified samples as rifampicin-sensitive but DST identified them as resistant), the rpoB region was sequenced; the discordant results resolved in three of these cases in favour of Xpert MTB/RIF. For nine cases in which results were discordant and Xpert MTB/RIF identified the specimens as rifampicin resistant but DST identified them as rifampicin sensitive, sequencing of the rpoB region was performed; discordant results resolved in eight of these cases in favour of Xpert MTB/RIF.

4.1.9 Accuracy of the reference standards

Culture is regarded as the best reference standard for active TB, and was the reference standard used for TB in the systematic review on pulmonary TB. Phenotypic culture-based DST methods, using WHO’s recommended critical concentrations, were the reference standards for rifampicin resistance.19

Three recent studies have raised concerns about using phenotypic DST to detect rifampicin

resistance, in particular the automated BACTEC MGIT 960 system. Van Deun and colleagues reported that the BACTEC 460 and the BACTEC MGIT 960 missed certain strains associated with low-level rifampicin resistance.20 Using Xpert MTB/RIF and gene sequencing, Williamson and colleagues identified four patients (three with clinical information available) whose TB isolates contained mutations to the rpoB gene but whose results from the BACTEC MGIT 960 indicated that the isolates were rifampicin susceptible. In this study, 2/49 (4.1%) patients whose isolates did not have apparent rpoB gene mutations experienced treatment failure compared with 3/3 (100%) patients whose isolates did have rpoB gene mutations and had been deemed rifampicin-susceptible using phenotypic methods.21

In a study involving retreatment patients, Van Deun and colleagues found that disputed rpoB mutations conferring low-grade resistance were often missed by rapid phenotypic DST, particularly with the BACTEC MGIT 960 system, but to a lesser extent also by conventional slow DST.22

The authors suggested this may be the reason for the perceived insufficient specificity of molecular DST for rifampicin. Although the study involved retreatment patients, the results also appear to be similar for individuals newly diagnosed with TB (A Van Deun, personal communication, 2013).

Therefore, using only phenotypic DST as a reference to determine the specificity of a molecular method of DST may underestimate the specificity of the molecular method of DST. In light of these findings, it is unclear whether and to what extent Xpert MTB/RIF might outperform phenotypic methods of DST in assessing rifampicin resistance.

WHO continues to collect and evaluate emerging data on this issue, and will formally review the

19 Policy guidance on drug-susceptibility testing (DST) of second-line antituberculosis drugs. Geneva, World Health Organization, 2008 (WHO/HTM/TB/2008.392).20 Van Deun A et al. Mycobacterium tuberculosis strains with highly discordant rifampin susceptibility test results. Journal of Clinical Microbiology, 2009,47:3501–3506.21 Williamson DA et al. An evaluation of the Xpert MTB/RIF assay and detection of false-positive rifampicin resistance in Mycobacterium tuberculosis. Diagnostic Microbiology and Infectious Disease, 2012 , 74:207–209.22 Van Deun A et al. Rifampicin drug resistance tests for tuberculosis: challenging the gold standard. Journal of Clinical Microbiology, 2013, 51:2633–2640.


accuracy of phenotypic resistance standards for DST once sufficient data become available.

4.2 Using Xpert MTB/RIF to diagnose extrapulmonary TB and rifampicin resistance in adults and children

A literature search was performed for published and unpublished reports of studies made available

from 1 January 2007 until 14 December 2012. From the literature searches, 22 relevant studies that included 5922 specimens were identified. Figure 7 shows the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) diagram with the flow of the studies.

Figure 7. Selection of studies evaluating the accuracy of Xpert MTB/RIF in diagnosing extrapulmonary TB and detecting rifampicin resistance in adults and children: flow diagram of studies included in the review

194 potentially relevant citations identified in electronic databases

143 excluded at screening step 1Reason: Not relevant based on assessment of title and abstract

51 complete papers retrieved for more detailed evaluation

22 papers (or studies) included in the

systematic review

8 unpublished papers or collections of data added• 6 about to be published• 2 studies continuing

1 unpublished paper or collection of data excludedReasonDid not include any sample type that contributed to the subgroups being analysed

36 excluded at screening step 2 Reasons• < 10 samples per type of extrapulmonary TB: 10• Specificity results lacking: 0 • Abstract only: 12 • Cost–effectiveness analysis: 0 • Did not contain samples of extrapulmonary TB: 5• Duplicate data: 1• Editorial or comment: 2• Inappropriate reference standard: 3• Outcome lacking: 0 • Review article: 3• Technical: 0


Thirteen studies (59%) were conducted in low-income and middle-income countries. All studies were performed in tertiary care centres or reference laboratories. Two studies (one by Bates and one by Walters) included only children;23 nine studies included no children. In the remaining 11 studies the percentages of children in the study population ranged from 2% to 34%. Three published studies and four unpublished studies included only one type of sample (for example, only pleural fluid). The remainder of the studies included varying percentages of different types of samples. Twelve studies reported on only one sample per patient; the other studies either reported on multiple samples per patient or did not report the number of samples per patient. Six studies used frozen, archived samples ; 15 used fresh samples; and 1 study used both fresh and frozen samples.

The studies reviewed were diverse with respect to both the different types of samples tested and their relative percentages in each study. The heterogeneity in the performance characteristics of Xpert MTB/RIF, primarily in sensitivity, across the different types of samples was substantial. Therefore, combining these studies to obtain an overall estimate of the accuracy of Xpert MTB/

RIF in diagnosing extrapulmonary TB would not be meaningful.

An analysis of predefined subgroups of sample types (that is, pleural fluid, lymph node aspirate or tissue, CSF, gastric fluid, and tissue other than lymph node) was undertaken to account for the heterogeneity among the studies. Data on the smear status of samples were not available for the individual types of samples. Therefore, samples included in the subgroups were either smear positive, smear negative or of unknown smear status.

4.2.1 Detecting lymph node TB in samples from biopsy or fine-needle aspiration

Fourteen studies were identified that tested the accuracy of Xpert MTB/RIF on samples from lymph node biopsies or fine-needle aspiration (FNA) compared against culture as a reference standard (Figure 8). A meta-analysis was performed for each sample type if at least 4 studies had at least 10 samples in each study. For the 11 studies with more than 10 samples (total, 849 samples) the estimates for sensitivity ranged from 50% to 100%. The pooled sensitivity across studies was 84.9% (95% CI, 72.1–92.4%); the pooled specificity was 92.5% (95% CI, 80.3–97.4%).

Figure 8. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting extrapulmonary TB in lymph node samples (tissue or aspirate) compared with culture as the reference standarda


a The figure shows the estimated sensitivity and specificity of each study (blue square) and its 95% CI (black horizontal line). The names of unpublished studies have been obscured. Values for test results are the number of each type of result (true positive, false positive, false negative, true negative).

23 Additional information about the studies can be found in the annexes to Using the Xpert MTB/RIF assay to detect pulmonary and extrapulmonary tuberculosis and rifampicin resistance in adults and children: Expert Group meeting report. Geneva, World Health Organization, 2013 (available at: http://www.who.int/tb/laboratory/policy_statements/en/)


Five studies (one unpublished) assessed the accuracy of Xpert MTB/RIF on samples from lymph nodes compared against an author-defined CRS (Figure 9). In the different studies the CRS included some combination of NAAT other than Xpert MTB/RIF, histology, smear,

culture, biochemical testing, presenting signs,

or a response to treatment with anti-TB therapy.

The pooled sensitivity was estimated to be

83.7% (95% CI, 73.8–90.3%), and the pooled

specificity was 99.2% (95% CI, 88.4–100%).

Figure 9. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting extrapulmonary TB in lymph node samples (tissue or aspirate) compared with a composite reference standarda



Studies that used fresh samples showed a slightly higher sensitivity and a lower specificity than those that used frozen samples; however, the precision of these estimates was low because data were limited. Only nine studies included information on the prevalence of HIV, and only two studies included more than 10% HIV-positive patients. Accuracy estimates for these studies did not differ substantially from those that included fewer HIV-positive patients. Given the limited amount of data for the group that had a prevalence of HIV greater than 10%, a summary estimate was not determined.

4.2.2 Detecting pleural TB in pleural fluid

Seventeen studies (1385 samples, 217 culture positive) provided data that could be used to estimate the sensitivity and specificity of Xpert MTB/RIF in testing pleural fluid. Results from the assessment of the accuracy of Xpert MTB/RIF using samples from pleural biopsy were integrated into the assessment of Xpert MTB/RIF for testing tissue biopsies of all kinds other than

lymph node.

The sensitivity of Xpert MTB/RIF in testing pleural fluid varied from 0% to 100% among the studies. The outliers at the lower end of the range and the upper end were studies with few culture-confirmed cases of TB. One study was excluded from the meta-analysis because the sensitivity could not be estimated; a second study was also excluded because it included fewer than 10 specimens of pleural fluid. The pooled sensitivity was low at 43.7%, with wide confidence intervals (95% CI, 24.8–64.7%); the pooled specificity was high at 98.1% (95% CI, 95.3–99.2%) (Figure 10).

Seven studies (4 published and 3 unpublished) with 698 samples (188 culture positive) evaluated Xpert MTB/RIF in testing pleural fluid compared with a CRS. In the different studies the CRS included some combination of NAAT other than Xpert MTB/RIF, histology, smear, culture, biochemical testing, presenting signs, or a response to treatment with anti-TB therapy. Compared with studies that used culture as the


Figure 10. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting TB using pleural fluid compared with culture as a reference standarda

TP, true positive; FP, false positive; FN, false negative; TN, true negative; CI, confidence interval.a The figure shows the estimated sensitivity and specificity of each study (blue square) and its 95% CI (black horizontal line). The names of unpublished studies have been obscured. Values for test results are the number of each type of result (true positive, false positive, false negative, true negative).

reference standard, the CRS subgroup yielded an even lower pooled sensitivity (17.0%; 95% CI, 7.5–34.2%), albeit with a high specificity (99.9%; 95% CI, 93.7–100.0%) (Figure 11).

Pleural fluid is not regarded as a suitable specimen for the microbiological diagnosis of pleural TB. Pleural biopsy is the preferred sample type for diagnosing pleural TB.

Figure 11. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting TB in pleural fluid compared with a composite reference standarda


4.2.3 Detecting TB in samples of CSF

In total, 709 CSF samples in 16 studies were tested with Xpert MTB/RIF, and the results were compared against culture as a reference standard (13 studies had more than 10 samples, and 10 of these provided information on both sensitivity and specificity). Only 117 culture-confirmed cases of TB were found. Estimates of sensitivity varied

widely and ranged from 51% to 100%; one study with 19 samples (3 false negatives) was an outlier at 0%. The pooled sensitivity across studies was 79.5% (95% CI, 62.0–90.2%) and the pooled specificity was 98.6% (95% CI, 95.8–99.6%), suggesting good performance of Xpert MTB/RIF in detecting TB in CSF when tested against culture as a reference standard (Figure 12).


Ten of the sixteen studies comparing Xpert MTB/RIF with culture as a reference standard used a concentration step in processing the sample. Six studies did not use a concentration step. A concentration step appeared to increase the

sensitivity of Xpert MTB/RIF (82%; 95% CI, 71–93% for concentrated samples versus 56%; 95% CI 36-77% for unconcentrated samples), although the confidence intervals overlapped. The use of a concentration step did not affect the specificity.

Figure 12. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting TB in cerebrospinal fluid compared with culture as a reference standarda


Only 6 studies (3 unpublished) assessed the accuracy of using Xpert MTB/RIF to test CSF samples compared against an author-defined CRS; sensitivity estimates ranged from 20% to 86% (Figure 13). The pooled sensitivity was estimated to be 55.5% (95% CI, 44.2–66.3%), and the

pooled specificity was estimated to be 98.8% (95% CI, 94.5–99.8%). The reduced sensitivity of Xpert MTB/RIF compared with the CRS versus culture as a reference standard suggests that either the CRS was too broad or that culture as the single reference standard is inadequate.

Figure 13. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting TB in cerebrospinal fluid compared with a composite reference standarda


4.2.4 Detecting TB in gastric fluid

Twelve studies (1258 samples) examined the performance of Xpert MTB/RIF in samples of

gastric fluid and compared the results against culture as a reference standard (8 studies had more than 10 samples). Two studies included only children. The remaining 10 studies included


adults and children (the proportion of children included across sample types ranged from 0% to 33.5%). One study with 788 samples that had valid results using Xpert MTB/RIF accounted for 62.6% of all gastric-fluid samples. The estimates of sensitivity varied from 69% to 100%; specificity

varied from 98% to 100%, with one unpublished

study an outlier reporting 52% specificity. The

pooled sensitivity across studies was 83.8% (95%

CI, 65.9–93.2%), and the pooled specificity

was 98.1% (95% CI, 92.3–99.5%) (Figure 14).

Figure 14. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting TB in gastric fluid compared with culture as a reference standarda



All studies used a concentration step during specimen processing and prior to using the Xpert MTB/RIF assay; therefore, this was not considered to be a source of heterogeneity within the data. The condition of the gastric-fluid specimen (fresh versus frozen) did not appear to affect the performance of Xpert MTB/RIF. The five studies testing fresh specimens achieved a pooled sensitivity of 83% (95% CI, 71–93%). A further three studies used frozen specimens and had a pooled sensitivity of 79% (95% CI, 57–100%). The condition of the specimen (fresh or frozen) did not affect the specificity.

4.2.5 Detecting TB in tissue samples

There were 12 studies (699 samples) that tested Xpert MTB/RIF using tissue samples from any

site other than a lymph node, and compared the results against culture as a reference standard (10 studies had more than 10 samples). The estimates of sensitivity varied widely and ranged from 42% to 100%. The pooled estimate of sensitivity was calculated as 81.2% (95% CI, 67.7–89.9%). The pooled specificity was 98.1% (95% CI, 87.0–99.8%) (Figure 15).

The condition of the specimen (fresh versus frozen) did not appear to affect the performance of Xpert MTB/RIF. The five studies testing fresh specimens achieved a pooled sensitivity of 79% (95% CI, 64–94%). A further three studies used frozen specimens and had a pooled sensitivity of 76% (95% CI, 58–94%). The condition of the specimen (fresh or frozen) did not affect the specificity.


Figure 15. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting TB in tissue samples (other than from a lymph node) compared with culture as a reference standarda



4.2.6 Detecting rifampicin resistance

Data on detecting rifampicin resistance were used only from published studies because data collection was incomplete in some of the unpublished studies. Furthermore, data from a study were included only if DST had been done for all samples that were positive by culture and Xpert MTB/RIF because the selective confirmation of results could have introduced bias.

In total, data on resistance testing were available for 566 samples from 13 studies. Forty one samples were confirmed to be rifampicin resistant by phenotypic DST. Given the limited amount of data, no summary estimate was calculated. The average prevalence of rifampicin resistance across the studies was 5.4%, with the highest prevalence reported from India (25.6%). Xpert MTB/RIF did not identify 2 of the 41 phenotypically rifampicin-resistant samples. Six of the 41 samples that were identified as rifampicin resistant by Xpert MTB/RIF were identified as susceptible by phenotypic DST. Five of these six samples underwent sequencing of the rpoB gene, and four were found to have a

mutation in the same region of the rpoB gene at codon 533. Hence, Xpert MTB/RIF detected four additional rifampicin-resistant strains that would have been missed by phenotypic DST alone.

See section 4.1.9 for more information about the accuracy of the reference standards used to detect rifampicin resistance.

4.3 Using Xpert MTB/RIF to diagnose TB and rifampicin resistance in children

An electronic literature search was performed initially on 24 January 2013. In total, 39 articles were identified. Of these, 26 were excluded based on a review of their title and abstract. Of the remaining 12 articles that underwent a full-text review, 10 were included in this review. An additional 2 published studies were included that had been identified during a final electronic search conducted on 3 April 2013. An additional four unpublished studies were included that had been identified by querying networks of people working in childhood TB and contacting authors. In total, 16 studies were included (Figure 16).


Figure 16. Selection of studies evaluating the accuracy of Xpert MTB/RIF in diagnosing TB and rifampicin resistance in children: flow diagram of studies included in the review

39 articles identified through initial database search

January 29, 2013

39 articles screened by title and abstract

12 full text articles assessed for eligibility

16 studies included (12 published, 4 unpublished)

for qualitative analysis/synthesis• 13 pulmonary TB• 6 Rifampicin resistance• 4 peripheral lymph node TB• 5 TB meningitis

26 articles excluded based on title and

abstract

2 full-text articles excluded

(> 15 years only)

2 published articles included after final

database search and review on 3 April, 2013

4 unpublished studies identified and included

through additional searches

4.3.1 Using Xpert MTB/RIF to diagnose pulmonary TB in children

The utility of Xpert MTB/RIF in diagnosing paediatric pulmonary TB was evaluated in 13 studies that included 2603 participants. Studies either collected the same specimen type from all children or different types of specimens from different subgroups of children (for example, samples of expectorated sputum were collected from older children; samples of induced sputum or gastric lavage or aspirate were collected from younger children). In three studies different types of specimens were collected from each child. As a result, a total of 3347 specimens

were assessed: expectorated sputum (4 studies, 270 children), induced sputum (7 studies, 1279 children), nasopharyngeal aspirate (1 study, 474 children), gastric lavage or aspirate (6 studies, 1324 children).

For samples of expectorated sputum, sensitivity varied from 55% to 90%; for induced sputum sensitivity varied from 40% to 100%; and for gastric lavage or aspirate it varied from 40% to 100%. Confidence intervals overlapped for each specimen type, suggesting that no specimen type was superior. Specificities for all studies and specimen types ranged from 93% to 100% (Figure 17).


Figure 17. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting pulmonary TB, peripheral lymph node TB and TB meningitis in children compared against culture as a reference standard, by study and specimen typea

TP, true positive; FP, false positive; FN, false negative; TN, true negative; CI, confidence interval; ES, expectorated sputum; GLA, gastric lavage or aspirate; IS, induced sputum; NPA, nasopharyngeal aspirate.


One study examined the yield of specimens of nasopharyngeal aspirate. The sensitivity of Xpert MTB/RIF for these specimens was 44% (95% CI, 33–55%). In the same group of children, the sensitivity for induced sputum was 60% (95% CI, 49–70%).

The sensitivity and specificity of Xpert MTB/RIF were compared against mycobacterial culture as the reference standard as well as against the clinical TB reference standard.

4.3.1.1 Differences in the reference standards used

In all of the studies included in the review, 13.2% of children had culture-confirmed TB. The proportion of children with culture-confirmed TB varied by study and specimen type (range, 0–54.2%). In the majority of studies (9/13; 69%), multiple cultures were performed on samples from single participants. Hence, the definition of culture positive was based on the presence of at least one positive culture result

out of as many as six cultures performed. The average bacteriological yield in studies using multiple cultures was increased compared with the group of four studies that defined a participant as culture-positive based only on one culture result. Studies also used different culture techniques, but the impact of this potential source of bias was not evaluated.

Children were categorized as positive using the clinical TB reference standard if they were culture negative and had started anti-TB therapy based on a clinical diagnosis of TB. This broad clinical reference standard was chosen in order to accommodate the heterogeneous study methods and clinical definitions used in the studies. Children assigned to the group clinical not TB (that is, negative according to the clinical TB reference standard) either (1) did not have another diagnosis assigned, or (2) did not start anti-TB treatment but nonetheless improved or did not worsen after at least 1 month of follow-up after enrolment.


4.3.1.2 Accuracy of Xpert MTB/RIF compared with culture

The pooled sensitivity of Xpert MTB/RIF compared against culture was 66% for samples of expectorated or induced sputum (95% CrI, 52–77%), and 66% for samples of gastric lavage or aspirate (95% CrI, 51–81%). The width of the credible intervals indicated there was a high level of heterogeneity among the studies. The specificity values for Xpert MTB/RIF compared against culture as the reference standard were all at least 98%, with narrow credible intervals.

4.3.1.3 Accuracy of Xpert MTB/RIF compared with clinical TB as the reference standard

The sensitivity of Xpert MTB/RIF in culture-negative samples from paediatric patients compared against clinical TB as the reference standard was 4% for samples of expectorated or induced sputum and 15% for gastric lavage or aspirate, with all confidence intervals being wide and

therefore indicating a high level of heterogeneity. It is likely that the apparently poor performance of Xpert MTB/RIF was the result of a clinical TB reference standard that lacked specificity. The specificity values of Xpert MTB/RIF compared against the clinical TB reference standard were at least 99%, with narrow confidence intervals.

4.3.2 Xpert MTB/RIF compared with smear microscopy

The diagnostic accuracy of smear microscopy was calculated and compared against culture as the reference standard. Sensitivities varied from 30% to 60% for samples of expectorated sputum, 0% to 50% for induced sputum, 0% to 50% for gastric lavage or aspirate, and the sensitivity was 18% in the one cohort providing samples of nasopharyngeal aspirate. The 95% confidence intervals were wide and overlapping. The specificity was 95% or higher for all studies and specimen types (Figure 18).

Figure 18. Forest plot of the sensitivity and specificity of smear microscopy in detecting pulmonary TB, peripheral lymph node TB and TB meningitis in children compared against culture as a reference standard , by study and specimen typea

TP, true positive; FP, false positive; FN, false negative; TN, true negative; CI, confidence interval; ES, expectorated sputum; GLA, gastric lavage or aspirate; IS, induced sputum; NPA, nasopharyngeal aspirate.



For expectorated and induced sputum the pooled sensitivity was 29% (95% CrI, 16–42%); for gastric lavage or aspirate it was 22% (95% CrI, 12–35%); the pooled credible intervals were wide for both expectorated or induced sputum and gastric lavage or aspirate, indicating a high

level of heterogeneity. The pooled specificity of

Xpert MTB/RIF for samples of gastric lavage or

aspirate was 99% (95% CrI, 97–100%); for

expectorated and induced sputum it was 100%

(95% CrI, 99–100%) (Table 3).

Table 3. Meta-analysis of the estimated sensitivity and specificity of smear microscopy in diagnosing pulmonary TB, peripheral lymph node TB and TB meningitis in children compared against culture as a reference standard in published and unpublished studies

ComparisonSpecimen type (No. of studies, No. of children)



Smear microscopy compared against culture as a reference standard (published and unpublished studies)

Expectorated sputum and induced sputum (10, 1546)

29 (16–42)

100 (99–100)

Gastric lavage or aspirate (7, 1319)

22 (12–35)

99 (97–100)


These data suggest that in comparison with smear microscopy, the sensitivity of Xpert MTB/RIF was 37% higher if performed on samples of expectorated or induced sputum and 44% higher if performed on samples of gastric lavage or aspirate.

4.3.3 Performance of Xpert MTB/RIF in smear-positive and smear-negative children

Seven studies reporting data from 1083 children were included in the analysis of using Xpert MTB/RIF to test samples of expectorated or induced sputum.

Six studies with data from 1259 children were included in the analysis of using Xpert MTB/RIF to test samples of gastric lavage or aspirate. All studies that were included reported results for both smear-positive and smear-negative children (Figure 19).

The sensitivity of Xpert MTB/RIF on samples of expectorated or induced sputum from children with smear-negative results ranged from 25% to 86%. In contrast, the sensitivity of Xpert MTB/RIF in testing samples of either expectorated or induced sputum from smear-positive children ranged from 92% to 100%.

The pooled estimate of sensitivity for smear-positive children was 96% (95% CrI, 90–99%); for smear-negative children it was 55% (95% CrI, 41–69%).

The findings were similar when Xpert MTB/RIF was used to test samples of gastric lavage or aspirate: for smear-positive children the overall sensitivity was 95% (95% CrI, 83–99%), and for smear-negative children it was 62% (95% CrI, 44–80%). The credible intervals were wide (indicating variability), but did not overlap (Table 4).


Figure 19. Forest plot of the sensitivity of Xpert MTB/RIF in diagnosing pulmonary TB, peripheral lymph node TB and TB meningitis in smear-positive and smear-negative children, by study and specimen typea

TP, true positive; FP, false positive; FN, false negative; TN, true negative; CI, confidence interval; ES, expectorated sputum; GLA, gastric lavage or aspirate; IS, induced sputum.



Table 4. Meta-analysis of the sensitivity and specificity of Xpert MTB/RIF in diagnosing pulmonary TB, peripheral lymph node TB and TB meningitis compared against culture as a reference standard in smear-negative and smear-positive children in published and unpublished studies

ComparisonSpecimen type (No. of studies, No. of children)



Xpert MTB/RIF in smear-positive children

Expectorated sputum or induced sputum (7, 68)a

96 (90–99) c

Gastric lavage or aspirate (6, 32)b

95 (83–99) c

Xpert MTB/RIF in smear-negative children

Expectorated sputum or induced sputum (7, 1008)a

55 (41–69)

98 (96–99)

Gastric lavage or aspirate (6, 1204)b

62 (44–80)

99 (97–99)

CrI, credible interval; the CrI is the Bayesian equivalent of the confidence interval; additional information about the studies can be found in the annexes to Using the Xpert MTB/RIF assay to detect pulmonary and extrapulmonary tuberculosis and rifampicin resistance in adults and children: Expert Group meeting report. Geneva, World Health Organization, 2013 (available at: http://www.who.int/tb/laboratory/policy_statements/en/)

a This analysis included studies by Bates 2013, Nhu 2013, Nicol 2011, Rachow 2012 (expectorated sputum only), Sekadde 2013, an unpublished study by Walters and Zar 2012.b This analysis included studies by Bates 2013, Causse 2011, an unpublished study by Chisti, Nhu 2013, Tortoli 2012 and Walters 2012.c There were not enough data to calculate specificity.

As expected, smear status was associated with the performance of Xpert MTB/RIF, indicating that the sensitivity of Xpert MTB/RIF was greater in children who had a higher mycobacterial burden than in those with paucibacillary disease. Xpert MTB/RIF detected 55% of smear-negative culture-positive children from samples of expectorated or induced sputum, and 62% of smear-negative culture-positive children from samples of gastric lavage or aspirate.

4.3.4 Xpert MTB/RIF in children aged 0–4 years and 5–15 years

Five of the seven studies reported results for children aged 0–4 years and 5–15 years. Data from 976 children were included in the analysis of using Xpert MTB/RIF to test samples of expectorated or

induced sputum. The estimated accuracy of Xpert MTB/RIF in testing samples of gastric lavage or aspirate was reported only for children aged 0–4 years (5 studies, 957 children).

The sensitivity of Xpert MTB/RIF in both age groups ranged from 0% to 100%. The pooled sensitivity among children aged 0–4 years was 57% for Xpert MTB/RIF used to test samples of expectorated or induced sputum (95% CrI, 36–74%) and gastric lavage or aspirate (95% CrI, 38–75%). The pooled sensitivity for samples of expectorated or induced sputum was higher in children aged 5–15 years (83%; 95% CrI, 68–92%). The pooled specificity was at least 98% for all groups assessed, with relatively narrow credible intervals.


4.3.5 Xpert MTB/RIF in HIV-positive and HIV-negative children

Seven studies reporting data from 1074 children

were included in the analysis of the accuracy of

Xpert MTB/RIF in testing samples of expectorated

or induced sputum. All studies reported results for both HIV-positive and HIV-negative children.

The sensitivity of the test among HIV-positive children ranged from 20% to 100%; among HIV-negative children it ranged from 33% to 100% (Figure 20).

Figure 20. Forest plot of the sensitivity and specificity of Xpert MTB/RIF in detecting pulmonary TB, peripheral lymph node TB and TB meningitis in HIV-positive and HIV-negative children, by study and specimen typea

TP, true positive; FP, false positive; FN, false negative; TN, true negative; CI, confidence interval; ES, expectorated sputum; GLA, gastric lavage or aspirate; IS, induced sputum.


The pooled sensitivity among HIV-positive children (75%; 95% CrI, 57–88%) was higher than the sensitivity for HIV-negative children (57%; 95% CrI, 41–71%); however, the credible intervals were wide and overlapping. The pooled specificity was 98% for both groups.An analysis of the performance of Xpert MTB/

RIF stratified by smear status and HIV status demonstrated that the test had high sensitivity among smear-positive children regardless of their HIV status. The sensitivity of Xpert MTB/RIF was lowest among HIV-negative children, although the credible intervals were wide and overlapping (Table 5).


Table 5. Meta-analysis comparing Xpert MTB/RIF for diagnosing pulmonary TB, peripheral lymph node TB and TB meningitis using culture as a reference standard in HIV-positive and HIV-negative children, stratified by smear status

CategorySpecimen type (No. of studies, No. of children)

Pooled sensitivity (%) (95% CrI)

HIV positive and smear positive Expectorated sputum or induced sputum (5, 21)a

97 (85–100)

HIV positive and smear negativeExpectorated sputum or induced sputum (5, 25)a

69 (46–87)

HIV negative and smear positiveExpectorated sputum or induced sputum (5, 29)a

94 (81–99)

HIV negative and smear negativeExpectorated sputum or induced sputum (5, 85)a

48 (29–67)

CrI, credible interval; the CrI is the Bayesian equivalent of the confidence interval; additional information about the studies can be found in the annexes to Using the Xpert MTB/RIF assay to detect pulmonary and extrapulmonary tuberculosis and rifampicin resistance in adults and children: Expert Group meeting report. Geneva, World Health Organization, 2013 (available at: http://www.who.int/tb/laboratory/policy_statements/en/)

a The studies included in this analysis were Bates 2013, Nicol 2011, Rachow 2012 (expectorated sputum), Sekadde 2013 and Zar 2012.

A metaregression model that simultaneously controlled for smear status and HIV status using Xpert MTB/RIF to test samples of expectorated or induced sputum showed that the odds of test positivity were fourfold greater in smear-positive

children than in smear-negative children. The odds of Xpert MTB/RIF positivity were not statistically significant for HIV-positive children compared with HIV-negative children (Table 6).

Table 6. Metaregression model for Xpert MTB/RIF using samples of expectorated or induced sputum from children, controlling for smear status and HIV status

Node Mean SD MC error 2.5% Median 97.5%

Beta 0 0.06201 0.385 0.0054 –0.8159 –0.064 0.7059

Beta 1 (HIV) 0.5863 0.5551 0.008862 –0.4919 0.5789 1.705

Beta 2 (Smear) 3.98 1.076 0.02855 2.159 3.878 6.399

Pool

ed s

ensit

ivity

Smear negative and HIV negative

0.485 0.09264 0.0013 0.3066 0.484 0.6695

Smear positive and HIV negative

0.9694 0.03031 6.402 0.8873 0.9785 0.9983

Smear negative and HIV positive

0.6213 0.1101 0.001197 0.3944 0.6257 0.8216

Smear positive and HIV positive

0.988 0.01977 3.951 0.9284 0.9879 0.9991

SD, standard deviation.


There were insufficient data to perform a meta-analysis comparing the performance of Xpert MTB/RIF when using samples of gastric lavage or aspirate from HIV-positive and HIV-negative children.

4.3.6 Using Xpert MTB/RIF to detect peripheral lymph node TB in children

The use of samples from FNA or lymph node biopsy to diagnose peripheral lymph node TB was evaluated in five studies. Two studies were excluded from the meta-analysis because their samples included fewer than five participants. Therefore, the analysis included 3 studies with data from 172 children (Figure 21). The pooled sensitivity of Xpert MTB/RIF compared against culture as the reference standard was 86% (95% CrI, 65–96%); the pooled specificity was 81% (95% CrI, 54–93%). Credible intervals were wide for both sensitivity and specificity, indicating heterogeneity.

4.3.7 Using Xpert MTB/RIF to detect TB meningitis in children

The use of CSF to diagnose TB meningitis was evaluated in five studies that included 61 children.

In total 7/61 (11.5%) children had TB meningitis confirmed by CSF culture. Of these, 3 children were positive by Xpert MTB/RIF (3/61; 4.9%). Two studies were excluded from the meta-analysis because they did not include any culture-positive children. One of the remaining studies had a subgroup sample size that included fewer than five children. Hence, there were insufficient data to calculate sensitivity from the two remaining studies in which 2/6 (33%) culture-positive children had positive results from Xpert MTB/RIF (Figure 21). The pooled specificity, which included 3 studies and 51 children, was 95% (95% CrI, 81–99%), with relatively wide credible intervals.

4.3.8 Using Xpert MTB/RIF to detect rifampicin resistance in children

In total, seven studies provided data on using Xpert MTB/RIF to detect rifampicin resistance (Figure 21). Four studies used conventional phenotypic DST, and three used line probe assays. A meta-analysis of 3 studies that included 176 participants showed a pooled sensitivity of 86% (95% CrI, 53%-98%) and a pooled specificity of 98% (95% CrI, 94–100%).

Figure 21. Forest plot of the sensitivity and specificity of Xpert MTB/RIF for detecting resistance to rifampicin, peripheral lymph node TB and TB meningitis in children, by study and specimen typea

RIF, rifampicin; TP, true positive; FP, false positive; FN, false negative; TN, true negative; CI, confidence interval; FNA, fine needle aspiration; CSF, cerebrospinal fluid. a The figure shows the estimated sensitivity and specificity of each study (blue square) and its 95% CI (black horizontal line). The names of unpublished studies have been obscured. Values for test results are the number of each type of result (true positive, false positive, false negative, true negative).


4.4 Affordability and cost effectiveness of using Xpert MTB/RIF to diagnose TB

Twelve published papers were identified that compared the costs of using Xpert MTB/RIF and follow-on tests to diagnose TB and MDR-TB with current diagnostic algorithms. The setting for most of these analyses was South Africa (10 studies); 2 of these studies also included other countries in sub-Saharan Africa (Botswana, Lesotho, Namibia, Swaziland and Uganda); 1 study included countries in the former Soviet Union; and 1 global analysis was done (Table 7).

On a global level, the review of the evidence showed that using Xpert MTB/RIF to diagnose TB and MDR-TB was cost effective for all individuals suspected of having TB including those suspected to be coinfected with HIV when compared with current practices. Using Xpert MTB/RIF to diagnose MDR-TB would cost US$ 0.09 billion per year globally, which is less than the cost of conventional diagnostics used globally and in all countries that have a high burden of TB.

Diagnosing TB in HIV-positive people using Xpert MTB/RIF would cost about US$ 0.10 billion per

year, and would cost less than the conventional diagnostics used globally in almost all of the countries with a high burden of TB. Testing everyone who has signs and symptoms of TB would cost almost US$ 0.47 billion per year globally, which is much more than the cost of conventional diagnostics. Nevertheless, in European countries, Brazil and South Africa, the cost would represent less than 10% of current funding for TB programmes.

Introducing Xpert MTB/RIF to diagnose MDR-TB and to diagnose TB in HIV-positive people is warranted in many countries. Using it to test everyone who has signs and symptoms of TB is affordable in several middle-income countries, but its financial viability in low-income countries would require large increases in TB funding or further reductions in the price of the test, or both.

At the country level, the majority of costing and cost–effectiveness studies were from South Africa, so there remains a need for further evidence to be collected from other countries and epidemiological settings.

34 XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATETa

ble

7. O

verv

iew

of s

tudi

es c

ompa

ring

the

costs

of u

sing

Xpe

rt M

TB/R

IF a

nd fo

llow

-on

tests

with

cur

rent

dia

gnos

tic a

lgor

ithm

s fo

r di

agno

sing

TB

and

MD

R-TB

Refe

renc

eaPr

imar

y re

sear

ch

ques

tion

Setti

ngPo

pula

tion

Base

line

diag

nost

ic

algo

rithm

Mai

n co

mpa

rison

Tim

e ho

rizon

TB

trans

miss

ion

cons

ider

ed

Conc

lusi

on

Abi

mbo

la

et a

l 201

2C

ost-e

ffect

iven

ess

anal

ysis

of u

sing

Xper

t MTB

/RI

F to

re

duce

ear

ly m

orta

lity

in in

divi

dual

s in

itiat

ing

ART

Sub-

Saha

ran

Afri

ca

HIV

-pos

itive

in

divi

dual

s in

itiat

ing

ART

Smea

r mic

rosc

opy,

C

XR (f

or s

putu

m s

mea

r-ne

gativ

e in

divi

dual

s),

cultu

re

Xper

t MTB

/RI

F as

firs

t-line

test

6 m

onth

sYe

sXp

ert M

TB/

RIF

is co

st ef

fect

ive

as a

firs

t-line

te

st

And

rew

s

et a

l 201

2C

ost–

effe

ctiv

enes

s an

alys

is of

usin

g Xp

ert M

TB/

RIF

to

scre

en fo

r TB

in H

IV-

posit

ive

indi

vidu

als

initi

atin

g A

RT

Sout

h A

frica

H

IV-p

ositi

ve

indi

vidu

als

initi

atin

g A

RT

Smea

r mic

rosc

opy,

cu

lture

Xper

t MTB

/RI

F as

firs

t-line

test

Lifet

ime

Yes

Xper

t MTB

/RI

F is

cost

effe

ctiv

e as

a fi

rst-li

ne

test

Men

zies

et

al 2

012

Cos

t–ef

fect

iven

ess

anal

ysis

of u

sing

Xper

t MTB

/RI

F to

di

agno

se T

B

Botsw

ana,

Le

soth

o,

Nam

ibia

, So

uth

Afri

ca a

nd

Swaz

iland

Indi

vidu

als

susp

ecte

d

to h

ave

TB

Smea

r mic

rosc

opy,

cu

lture

for s

putu

m

smea

r-neg

ativ

e in

divi

dual

s

Xper

t MTB

/RI

F as

firs

t-line

test

10 y

ears

an

d 20

ye

ars

Yes

Xper

t MTB

/RI

F is

cost

effe

ctiv

e as

a fi

rst-li

ne

test

Vass

all

et a

l 201

1C

ost–

effe

ctiv

enes

s an

alys

is of

usin

g Xp

ert M

TB/

RIF

Indi

a,

Sout

h A

frica

and

U

gand

a

Indi

vidu

als

susp

ecte

d

to h

ave

TB

Sput

um s

mea

r-neg

ativ

e in

divi

dual

s w

ith

2 ne

gativ

e sm

ears

pl

us p

ositi

ve c

linic

al

diag

nosis

and

/or

CXR

an

d cu

lture

-bas

ed D

ST

for r

etre

atm

ent c

ases

(1) X

pert

MTB

/RI

F in

add

ition

to

sm

ear

mic

rosc

opy;

(2)

Xper

t MTB

/RI

F re

plac

ing

smea

r m

icro

scop

y

Lifet

ime

No

Xper

t MTB

/RI

F is

cost

effe

ctiv

e in

bot

h co

mpa

rison

s

XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE 35W

inet

sky

et a

l 20

12

Cos

t–ef

fect

iven

ess

anal

ysis

of u

sing

Xper

t MTB

/RI

F

Priso

ns in

th

e fo

rmer

So

viet

U

nion

Indi

vidu

als

susp

ecte

d to

hav

e TB

(p

rison

ers

in a

se

tting

with

a

high

bur

den

of

MD

R-TB

)

Sym

ptom

scr

eeni

ng,

self-r

efer

ral,

mas

s m

inia

ture

radi

ogra

phy

Xper

t MTB

/RI

F as

firs

t-line

test

10 y

ears

Yes

Xper

t MTB

/RI

F is

cost

effe

ctiv

e as

a fi

rst-li

ne

test

Mey

er-

Rath

et a

l 20

12

Cos

t of s

calin

g up

us

e of

Xpe

rt M

TB/

RIF

natio

nally

Sout

h A

frica

A

ll in

divi

dual

s su

spec

ted

to

have

TBT

Smea

r mic

rosc

opy,

C

XR fo

r spu

tum

sm

ear-

nega

tive

indi

vidu

als,

cu

lture

Xper

t MTB

/RI

F as

firs

t-line

test

Tim

e ta

ken

to o

btai

n a

diag

nosis

(d

urat

ion

of

diag

nosti

c pa

thw

ay)

No

Scal

ing

up X

pert

MTB

/RI

F ad

ds 3

5% m

ore

to

budg

et in

201

1 (U

S$

293

milli

on v

ersu

s U

S$

218

milli

on);

scal

e-up

in

crea

ses

subs

tant

ially

th

e nu

mbe

r of T

B ca

ses

and

MD

R-TB

ca

ses

diag

nose

d, a

nd

patie

nts

starte

d on

tre

atm

ent

Mey

er-

Rath

et

al 2

011

(pol

icy

brie

f)

Cos

t of s

calin

g up

us

e of

Xpe

rt M

TB/

RIF

natio

nally

Sout

h A

frica

A

ll in

divi

dual

s su

spec

ted

to

have

TB

Smea

r mic

rosc

opy,

C

XR fo

r spu

tum

sm

ear-

nega

tive

indi

vidu

als,

cu

lture

Xper

t MTB

/RI

F as

firs

t-line

test

Tim

e ta

ken

to o

btai

n a

diag

nosis

(d

urat

ion

of

diag

nosti

c pa

thw

ay)

No

Ann

ual a

dditi

onal

bu

dget

requ

ired

is ar

ound

US$

34–

79

milli

on fo

r acc

eler

ated

sc

ale-

up

36 XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATEPa

ntoj

a

et a

l 20

12

Cos

t ana

lysis

of

usin

g Xp

ert M

TB/

RIF

for t

arge

t pop

ulat

ions

Glo

bal

anal

ysis,

an

d 36

co

untri

es

with

a

high

bu

rden

of

TB a

nd

a hi

gh

burd

en o

f M

DR-

TB

All

indi

vidu

als

susp

ecte

d to

hav

e TB

T,

HIV

-pos

itive

in

divi

dual

s su

spec

ted

to h

ave

TB,

indi

vidu

als

susp

ecte

d to

ha

ve M

DR-

TB

Smea

r mic

rosc

opy,

C

XR fo

r spu

tum

sm

ear-

nega

tive

indi

vidu

als,

cu

lture

, DST

Xper

t MTB

/RI

F as

firs

t-line

test

for e

ach

targ

et

popu

latio

n

1 ye

arN

oU

sing

Xper

t MTB

/RI

F to

dia

gnos

e M

DR-

TB a

nd T

B in

HIV

-po

sitiv

e in

divi

dual

s co

sts le

ss th

an c

urre

nt

conv

entio

nal d

iagn

osis;

us

ing

Xper

t MTB

/RIF

for

all i

ndiv

idua

ls su

spec

ted

to h

ave

TB is

muc

h m

ore

costl

y bu

t is

affo

rdab

le

in m

iddl

e-in

com

e co

untri

esSc

hnip

pel

et a

l 20

12

Plac

emen

t of X

pert

MTB

/RI

F: c

ost a

t la

bora

tory

leve

l and

at

clin

ic le

vel

Sout

h A

frica

A

ll in

divi

dual

s su

spec

ted

to

have

TB

Not

app

licab

leXp

ert M

TB/

RIF

at la

bora

tory

le

vel a

nd a

t cl

inic

leve

l

1 ye

arN

oC

ost o

f Xpe

rt M

TB/

RIF

at p

oint

-of-tr

eatm

ent

(clin

ic) l

evel

is 5

1%

mor

e ex

pens

ive

than

pl

acem

ent a

t sub

distr

ict

labo

rato

ries

(US$

10

7 m

illion

ver

sus

US$

71

milli

on)

Schn

ippe

l et

al

2013

Cos

t of a

sec

ond

Xper

t MTB

/RI

F te

st to

in

divi

dual

s w

ho w

ere

nega

tive

by fi

rst X

pert

MTB

/RI

F

Sout

h A

frica

H

IV-p

ositi

ve

indi

vidu

als

susp

ecte

d to

ha

ve T

B

Cul

ture

for i

ndiv

idua

ls ne

gativ

e by

Xpe

rt M

TB/

RIF

Seco

nd X

pert

MTB

/RI

F te

st fo

r ind

ivid

uals

nega

tive

by fi

rst

Xper

t MTB

/RI

F

Tim

e ta

ken

to o

btai

n a

diag

nosis

(d

urat

ion

of

diag

nosti

c pa

thw

ay)

No

Cos

t per

TB

patie

nt o

n tre

atm

ent o

f offe

ring

seco

nd X

pert

MTB

/RI

F te

st to

thos

e w

ho

initi

ally

teste

d ne

gativ

e is

12%

less

than

of

ferin

g al

tern

ativ

e cu

lture

XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE 37Th

eron

et

al 2

012

Acc

urac

y an

d co

st of

tests

use

d in

co

mbi

natio

n w

ith

Xper

t MTB

/RI

F

Sout

h A

frica

A

ll in

divi

dual

s su

spec

ted

to

have

TB

Smea

r mic

rosc

opy,

C

XR, i

nter

fero

n γ

rele

ase

assa

ys (a

djun

ct

tests

)

Xper

t MTB

/RI

F co

mbi

ned

with

ad

junc

t tes

ts

Not

av

aila

ble

No

Usin

g Xp

ert M

TB/

RIF

for i

ndiv

idua

ls w

ho a

re

sput

um s

mea

r-neg

ativ

e ha

d th

e lo

wes

t cos

t of

any

strat

egy,

and

the

high

est a

ccur

acy

Van

Rie

et

al 2

013

Cos

t, yi

eld

and

turn

arou

nd ti

me

of

usin

g Xp

ert M

TB/

RIF

to d

iagn

ose

sput

um

smea

r-neg

ativ

e in

divi

dual

s

Sout

h A

frica

A

ll in

divi

dual

s su

spec

ted

to h

ave

TB:

sput

um s

mea

r-ne

gativ

es

Smea

r mic

rosc

opy

and

cultu

re

Xper

t MTB

/RI

F us

ed to

te

st sp

utum

sm

ear-n

egat

ive

indi

vidu

als

susp

ecte

d to

ha

ve T

B

Not

av

aila

ble

No

Cos

t per

cas

e di

agno

sed

is sim

ilar i

n bo

th s

trate

gies

; Xpe

rt M

TB/

RIF

is co

st-sa

ving

fo

r pat

ient

s

ART

, ant

iretro

vira

l the

rapy

; MD

R-TB

, mul

tidru

g-re

sista

nt tu

berc

ulos

is; C

XR, c

hest

X-ra

y; D

ST, d

rug-

susc

eptib

ility

testi

ng.

a A

dditi

onal

info

rmat

ion

abou

t the

stu

dies

can

be

foun

d in

the

anne

xes

to U

sing

the

Xper

t MTB

/RI

F as

say

to d

etec

t pul

mon

ary

and

extra

pulm

onar

y tu

berc

ulos

is an

d rif

ampi

cin

resis

tanc

e in

adu

lts a

nd

child

ren:

Exp

ert G

roup

mee

ting

repo

rt. G

enev

a, W

orld

Hea

lth O

rgan

izat

ion,

201

3 (a

vaila

ble

at: h

ttp:/

/w

ww

.who

.int/

tb/

labo

rato

ry/

polic

y_sta

tem

ents/

en/

).


5. WHO’s policy recommendations

The GRADE process confirmed that there is a solid evidence base to support the widespread use of Xpert MTB/RIF to detect TB and rifampicin resistance. WHO therefore recommends the use of Xpert MTB/RIF as described below.

5.1 Using Xpert MTB/RIF to diagnose pulmonary TB and rifampicin resistance in adults and children

• Xpert MTB/RIF should be used rather than conventional microscopy, culture and DST as the initial diagnostic test in adults suspected of having MDR-TB or HIV-associated TB (strong recommendation, high-quality evidence).

• Xpert MTB/RIF should be used rather than conventional microscopy, culture and DST as the initial diagnostic test in children suspected of having MDR-TB or HIV-associated TB (strong recommendation, very low-quality evidence).

• Xpert MTB/RIF may be used rather than conventional microscopy and culture as the initial diagnostic test in all adults suspected of having TB (conditional recommendation acknowledging resource implications, high-quality evidence).

• Xpert MTB/RIF may be used rather than conventional microscopy and culture as the initial diagnostic test in all children suspected of having TB (conditional recommendation acknowledging resource implications, very low-quality evidence).

• Xpert MTB/RIF may be used as a follow-on test to microscopy in adults suspected of having TB but who are not at risk of MDR-TB or HIV-associated TB, especially when further testing of smear-negative specimens is necessary (conditional recommendation acknowledging resource implications, high-quality evidence).

RemarksThese recommendations apply to the use of Xpert MTB/RIF for specimens of processed and unprocessed sputum.

These recommendations also apply to specimens of gastric lavage and aspirate from adults and children; the recommendation for adults is based on the generalization of data from children.

These recommendations support the use of a single sputum specimen for diagnostic testing, acknowledging that processing multiple specimens increases the sensitivity of Xpert MTB/RIF but has resource implications.

Children suspected of having pulmonary TB but who have had a single negative result by Xpert MTB/RIF should undergo further diagnostic testing, and a child for whom there is a high clinical suspicion of TB should be treated even if an Xpert MTB/RIF result is negative or if the test is not available.

Conventional microscopy and culture remain essential for monitoring therapy and for performing DST for anti-TB agents other than rifampicin (including for isoniazid and second-line anti-TB agents).

Expanding the scope of the use of Xpert MTB/RIF and its placement in diagnostic algorithms will have significant implications for operational implementation, and its use should be phased in within the context of national strategic plans for TB.

Emerging data have shown that Xpert MTB/RIF detects some rifampicin-resistant strains that are identified as susceptible by phenotypic DST. Sequencing these discordant results usually resolves in favour of Xpert MTB/RIF, and patients missed by phenotypic DST have poor treatment outcomes on first-line treatment.


5.2 Using Xpert MTB/RIF to diagnose extrapulmonary TB and rifampicin resistance in adults and children

• Xpert MTB/RIF should be used in preference to conventional microscopy and culture as the initial diagnostic test for CSF specimens from patients suspected of having TB meningitis (strong recommendation given the urgency of rapid diagnosis, very low-quality evidence).

• Xpert MTB/RIF may be used as a replacement test for usual practice (including conventional microscopy, culture or histopathology) for testing specific nonrespiratory specimens (lymph nodes and other tissues) from patients suspected of having extrapulmonary TB (conditional recommendation, very low-quality evidence).

RemarksIndividuals suspected of having extrapulmonary TB but who have had a single negative result from Xpert MTB/RIF should undergo further diagnostic testing, and those for whom there is a high clinical suspicion for TB (especially children) should be treated even if an Xpert MTB/RIF result is negative or if the test is not available.

For CSF specimens, Xpert MTB/RIF should be preferentially used instead of culture if the sample

volume is low or if additional specimens cannot be obtained in order to make a quick diagnosis. If sufficient volume of material is available, concentration methods should be used to increase the yield.

Pleural fluid is a suboptimal sample for the bacterial confirmation of pleural TB regardless of the method used. A pleural biopsy is the preferred sample. The sensitivity of Xpert MTB/RIF in testing samples of pleural fluid is very low. Nevertheless, any individual with a positive result from pleural fluid tested by Xpert MTB/RIF should be treated for pleural TB; those with a negative result from Xpert MTB/RIF should have other tests.

Conventional microscopy and culture are essential for monitoring therapy and for performing DST for anti-TB agents other than rifampicin (including for isoniazid and second-line anti-TB agents).

Emerging data have shown that Xpert MTB/RIF detects some rifampicin-resistant strains that are found to be susceptible by phenotypic DST. Sequencing these discordant results usually resolves in favour of Xpert MTB/RIF, and patients missed by phenotypic DST have poor treatment outcomes on first-line treatment.

These recommendations do not apply to samples of stool, urine or blood, given the lack of data on the utility of Xpert MTB/RIF for these specimens.

6. Implementation considerations

A range of implementation issues were identified that need to be addressed to ensure that the use of Xpert MTB/RIF will be optimal.

• Although Xpert MTB/RIF is suitable for use at all levels of the health system, and testing with Xpert MTB/RIF does not require additional laboratory equipment, the sophisticated nature of the device requires

care in handling – that is, a stable and uninterrupted electrical supply is necessary to avoid interrupting the procedure and losing results; security against theft is necessary as is adequate storage space for the cartridges; staff must be assigned to perform the testing; and biosafety procedures must be put in place that are similar to those used for microscopy.


• Scaling up the implementation of Xpert MTB/RIF does not eliminate the need for capacity for conventional TB microscopy, culture and DST. Microscopy or culture, or both, remain essential for monitoring treatment since molecular tests based on DNA detection have not been shown to be suitable. Therefore, Xpert MTB/RIF should not be used to monitor treatment. In addition, conventional culture and DST are required to detect resistance to anti-TB agents other than rifampicin.

• The decision to scale-up the implementation of Xpert MTB/RIF should be made by ministries of health within the context of national plans for managing TB, MDR-TB and HIV-associated TB; plans for scale-up should consider the country’s specific epidemiology, the screening strategies used, how to ensure timely access to quality-assured first-line and second-line anti-TB agents, and whether care-delivery mechanisms are appropriate.

• Countries already using the molecular line probe assay to rapidly diagnose rifampicin resistance may introduce Xpert MTB/RIF at lower levels of the laboratory service, given that the line probe assay is suitable for high-throughput testing at the central or regional laboratory level (for additional information, see http://www.who.int/tb/laboratory/policy_statements/en/).

• The negative predictive value of Xpert MTB/RIF24 used to test for rifampicin resistance is more than 99% both in settings with a low prevalence of rifampicin resistance and in those with a high prevalence of rifampicin resistance – that is, a negative result accurately excludes the possibility of rifampicin resistance and no further testing is needed to confirm the negative results.

• The positive predictive value25 of Xpert MTB/RIF used to test for susceptibility to rifampicin exceeds 90% in settings or patient groups where the underlying prevalence of rifampicin resistance is greater than 15%. In settings or patient groups where rifampicin resistance is rare, the positive predictive value is adversely affected. The positive predictive values ranges from 71% to 84% in settings where the prevalence of rifampicin resistance is between 5% and 10%, and it diminishes to less than 70% when the prevalence of underlying rifampicin resistance falls below 5%. The positive predictive value can be greatly improved by engaging in a careful risk assessment for individual patients and by using targeted testing.

• It is important to differentiate between new cases of TB and previously treated cases; previously treated cases have a much higher likelihood of MDR-TB. Even in groups in which the prevalence of MDR-TB is low, testing previously treated TB cases with Xpert MTB/RIF will result in a high positive predictive value for detecting rifampicin resistance. Testing new cases not at risk for MDR-TB in groups in which the prevalence of MDR-TB is low will result in a lower positive predictive value for rifampicin susceptibility.

• It is uncommon to detect rifampicin resistance in groups in which there is a low prevalence of MDR-TB. In patients in a low-prevalence group who initially test positive by Xpert MTB/RIF, a second Xpert MTB/RIF test may be used to control for preanalytical and postanalytical errors, and to improve the clinicians’ confidence in deciding on an appropriate regimen. – In patients whose Xpert MTB/RIF result repeatedly detects rifampicin resistance,

24 The negative predictive value for rifampicin resistance is the proportion of cases diagnosed as rifampicin-susceptible that are truly susceptible.25 The positive predictive value for rifampicin resistance is the proportion of cases diagnosed as rifampicin-resistant that are truly resistant.


a WHO-recommended treatment regimen for MDR-TB (including isoniazid) should be initiated and optimized following confirmatory testing for rifampicin resistance and additional DST to test for resistance to isoniazid and second-line anti-TB agents. – Patients with discordant results for rifampicin resistance reported by Xpert MTB/RIF should be commenced on a WHO-recommended regimen of first-line agents.

• Based on the considerations above, confirmatory testing for rifampicin resistance in groups with a low prevalence of MDR -TB shown by Xpert MTB/RIF using other genotypic testing or conventional DST techniques may be unreliable and can produce discordant results which require resolution by DNA sequencing (section 4.1.9).

• Because Xpert MTB/RIF detects resistance only to rifampicin, countries with documented or suspected cases of extensively drug-resistant TB (XDR-TB) should establish or expand their capacity for conventional culture and DST for second-line agents to ensure quality assured testing of second-line agents, following WHO’s policy guidance.

• Scale-up of Xpert MTB/RIF should be implemented within the context of national plans for laboratory strengthening, considering that the GeneXpert system may also provide a technology platform for other diagnostic services, and reduce the costs involved in providing integrated laboratory services.

• Xpert MTB/RIF cartridges and the specimen reagent should be stored at 2–28 ˚C, following the manufacturer’s recommendations. The cartridges are bulky and require substantial storage space.

• The maximum capacity of a single, 4-module GeneXpert instrument is 16–20 specimens per day. Therefore, busier sites will either need several 4-module instruments or larger instruments (16 modules or more),

and these additional requirements will have associated implications for cost and storage.

• The manufacturer’s recommended ambient operating temperature for the GeneXpert instrument is limited to a maximum of 30 °C. In settings where the ambient temperature regularly exceeds 30 ˚C, the room where the assay is being done may need to be air conditioned.

• The Xpert MTB/RIF cartridges have a shelf-life of 12 months, posing a challenge in relatively inaccessible areas that have complex customs-clearance procedures. Therefore, inventory will need to be managed carefully by considering the usage, shelf-life and lead time for the delivery of orders.

• The GeneXpert modules require annual calibration, which must be performed by an authorized service provider or carried out by exchanging the modules. A remote calibration option is available from the manufacturer. This option does not require that modules be exchanged, and it can be done using a specially designed calibration kit. The calibration kit contains special cartridges that can be run on each module (without a sample) when calibration is due. During this run (which lasts approximately 20 minutes), the instrument will be automatically calibrated. However, if this calibration fails, it will be necessary to exchange the module. The calibration kit has the same shelf-life as the usual cartridges; in 2013 the cartridges and kit could be ordered together at the preferential pricing of US$ 450.00 for a kit sufficient to calibrate up to 4 modules.

• The initial capital cost for the GeneXpert unit (US$ 17 000 per 4-module desktop unit or US$ 17 500 per 4-module laptop unit) is significantly higher than the cost of microscopy (around US$ 1 500 per microscope), but it is much lower than


for conventional culture and DST (up to US$ 1.4 million for a new laboratory or up to US$ 300 000 for an established laboratory), given the need for extensive biosafety equipment and the infrastructure required for conventional testing.

• A detailed commercial sales contract and customer support plan should be negotiated with the supplier, guaranteeing an ample and continual supply of cartridges, customs clearance, maintenance and calibration, and appropriate repair and replacement.

• Mechanisms for rapidly reporting Xpert MTB/RIF results to clinicians, and offering timely access to appropriate treatment, must be established to provide patients with the benefits of an early diagnosis.

• Health-care staff must be trained how to properly select persons to test with the Xpert MTB/RIF assay in different epidemiological settings; they must also be trained to interpret the results of rifampicin-resistance testing for persons with and without risk factors for drug-resistant TB, understand when to initiate a WHO-recommended MDR-TB regimen based on the results from Xpert MTB/RIF, and how to use additional testing (either a second Xpert MTB/RIF assay or another test) to ensure early initiation of the appropriate WHO-recommended first-line regimen or MDR-TB treatment regimen.

6.1 Research needs

A series of operational research questions have been identified related to the introduction and scale-up of Xpert MTB/RIF, and its impact on the diagnosis of TB, MDR-TB and the management of patients.

No additional studies, at country level, on the diagnostic accuracy of Xpert MTB/RIF are needed.

Further operational research should focus on the following priorities:

• evaluation of diagnostic algorithms in different epidemiological and geographical settings, and patient populations;

• country-specific cost–effectiveness and cost–benefit analyses of Xpert MTB/RIF in different programmatic settings;

• prospective evaluations of using Xpert MTB/RIF for nonrespiratory specimens, including blood, urine and stool;

• evaluation of the impact of Xpert MTB/RIF in reducing the diagnostic delay and the time until appropriate treatment is initiated;

• evaluation of the impact of Xpert MTB/RIF on case-detection, treatment access and treatment outcomes in hard-to-reach populations.

6.2 Plans for supporting scale-up of the implementation of Xpert MTB/RIF

WHO will continue to evaluate the usefulness and impact of Xpert MTB/RIF, with active collaboration from the Global Laboratory Initiative of the Stop TB Partnership26, the WHO and Global Laboratory Initiative’s TB Supranational Reference Laboratory Network, and several organizations that have implemented the Xpert MTB/RIF system (including technical agencies, research organizations, nongovernmental organizations and donors). Implementing organizations are brought together by WHO during an annual meeting to share their experiences, identify operational constraints and solutions, and contribute to the global collection of data by WHO on the scale-up of Xpert MTB/RIF in different settings.

In addition, WHO’s Global TB Programme maintains a website27 dedicated to monitoring the roll-out of Xpert MTB/RIF in order to inform and facilitate coordination among implementing organizations including countries, technical

26 For additional information, see http://www.stoptb.org/wg/gli/.27 For additional information, see http://www.who.int/tb/laboratory/mtbrifrollout.


agencies, nongovernmental agencies and other partners.

A second edition of WHO’s implementation document for Xpert MTB/RIF is being developed by WHO’s Global TB Programme based on this policy update; it will be disseminated widely through e-mail and on the web sites of stakeholders. The Xpert MTB/RIF implementation manual outlines the requirements for systematic scale-up of Xpert MTB/RIF in varying epidemiological and resource settings. The document suggests ways to developing diagnostic algorithms and manage patients; it also discusses the operational and logistical aspects that need to be addressed during implementation of Xpert MTB/RIF. This policy update will also be incorporated into a comprehensive training package on Xpert MTB/RIF (being developed by WHO and the Global Laboratory Initiative) that will be made available to country and technical partners.

WHO and the Global Drug Facility have initiated

a unified forecasting initiative for Xpert MTB/RIF that collects forecasts of orders from major programmes and projects; these forecasts are submitted quarterly to the manufacturer. Reliable forecasting is needed to ensure effective scale-up of Xpert MTB/RIF by aiding the manufacturer in planning to meet demand.

Given the public-health importance and the ease of use of the Xpert MTB/RIF technology, expanded uptake at country level is anticipated and, therefore, the impact on TB case-detection will be monitored by WHO’s Global TB Programme. In addition, improvements in diagnosing drug-resistant TB using this technology will need to be accompanied by scale-up of access to appropriate treatment, and it will be essential that scale-up is closely coordinated with care-delivery mechanisms at the country level. Therefore, implementation of the new policy guidance will be harmonized with the Global TB Programme’s efforts to scale-up treatment of MDR-TB.

Box 4. Online resources

Strategic and Technical Advisory Group for Tuberculosis (STAG-TB): report of the 13th meeting. Geneva, World Health Organization, 2013 (WHO/HTM/TB/2013.09)(http://www.who.int/entity/tb/advisory_bodies/STAG_report2013.pdf).

Using the Xpert MTB/RIF assay to detect pulmonary and extrapulmonary tuberculosis and rifampicin resistance in adults and children: Expert Group meeting report. Geneva, World Health Organization, 2013 (available at http://www.who.int/tb/laboratory/policy_statements/en/).

Policy framework for implementing new tuberculosis diagnostics. Geneva, World Health Organization, 2010(available at: http://www.who.int/tb/laboratory/whopolicy_framework_mar2011.pdf).

Xpert MTB/RIF implementation manual. Technical and operational ‘how-to’: practical considerations, 2nd ed. Geneva, World Health Organization, 2014 (available at http://www.who.int/tb/laboratory/policy_statements/en/).

Xpert MTB/RIF training package. Geneva, Global Laboratory Initiative, 2014 (available at http://www.stoptb.org/wg/gli/).

44 XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE7.

GRA

DE ta

bles

Tabl

e 8.

GRA

DE

evid

ence

pro

file:

acc

urac

y of

Xpe

rt M

TB/R

IF in

dia

gnos

ing

pulm

onar

y TB

in a

dults

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r det

ectio

n of

pul

mon

ary

TB in

adu

lts, w

here

Xpe

rt M

TB/

RIF

is us

ed a

s a

repl

acem

ent

test

for s

mea

r mic

rosc

opy?

Parti

cipa

nts:

Adu

lts s

uspe

cted

of h

avin

g pu

lmon

ary

TB. S

ettin

g: M

ainl

y in

term

edia

te-le

vel l

abor

ator

ies

and

prim

ary

heal

th-c

are

faci

litie

s.Ta

rget

con

ditio

n: P

ulm

onar

y TB

. Ref

eren

ce s

tand

ard:

Sol

id c

ultu

re o

r liq

uid

cultu

re.

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 22

(900

8). P

oole

d se

nsiti

vity

: 88%

(95%

CrI,

84–

92%

); po

oled

spe

cific

ity: 9

9% (9

5% C

rI, 9

8– 9

9%)

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

a

Lim

itatio

nsIn

dire

ctne

ss I

ncon

sist

ency

Impr

ecis

ion

Publ

icat

ion

bias

Prev

alen

ce

25/1

000b

Prev

alen

ce

50/1

000b

Prev

alen

ce

100/

1000

bPr

eval

ence

30

0/10

00b

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-se

ctio

nal

Non

ecN

oned

Non

e N

one

Und

etec

tede

Hig

h

22(2

1–23

)44

(42–

46)

88(8

4–92

)26

4(2

52–2

76)

False

neg

ativ

es

(indi

vidu

als

inco

rrect

ly

clas

sified

as

not h

avin

g TB

)

Cro

ss-

sect

iona

l N

onec

Non

edN

one

Non

eU

ndet

ecte

deH

igh

3

(2–4

)6

(4–8

)12

(8–1

6)36

(24–

48)

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cl

assifi

ed a

s ha

ving

TB

Cro

ss-

sect

iona

l N

onec

Non

edN

one

Non

eU

ndet

ecte

deH

igh

10

(10–

20)

10(1

0–19

)9

(9–1

8)7

(7–1

4)

True

neg

ativ

es

(indi

vidu

als

with

out T

B)C

ross

-se

ctio

nal

Non

ecN

one4

Non

eN

one

Und

etec

tede

Hig

h

965

(956

–965

)94

1 (9

31–9

41)

891

(882

–891

)69

3(6

86–6

93)

CrI,

cre

dibl

e in

terv

al.

a Th

e ex

pect

ed n

umbe

r of X

pert

MTB

/RI

F re

sults

was

bas

ed o

n th

e se

nsiti

vity

and

spe

cific

ity e

stim

ates

cal

cula

ted

from

a c

ompa

rison

with

cul

ture

for d

iffer

ent p

reva

lenc

es o

f TB.

b Th

e es

timat

es o

f TB

prev

alen

ce w

ere

prov

ided

by

the

WH

O S

teer

ing

Gro

up.

c Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

The

maj

ority

of s

tudi

es e

nrol

led

indi

vidu

als

cons

ecut

ivel

y, a

nd a

sses

sed

the

resu

lt of

the

refe

renc

e sta

ndar

d bl

inde

d to

the

resu

lt fro

m X

pert

MTB

/RI

F. Th

e Xp

ert M

TB/

RIF

resu

lt is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

. d

The

ratin

g of

the

qual

ity o

f the

evi

denc

e m

ay b

e lo

wer

ed if

ther

e ar

e im

porta

nt d

iffer

ence

s in

the

tests

stu

died

and

in th

e ex

perti

se o

f tho

se a

pply

ing

the

tests

in th

e stu

dies

com

pare

d w

ith th

e se

tting

s fo

r whi

ch th

e re

com

men

datio

ns a

re in

tend

ed. T

he m

ajor

ity o

f stu

dies

(15/

22; 6

8%) e

valu

ated

Xpe

rt M

TB/

RIF

in s

ettin

gs w

here

it w

as in

tend

ed to

be

used

. Alth

ough

stu

dies

of d

iagn

ostic

acc

urac

y m

ay n

ot p

rovi

de d

irect

evi

denc

e ab

out p

atie

nt-im

porta

nt o

utco

mes

, tw

o stu

dies

pro

vide

d in

form

atio

n ab

out t

he ti

me

until

trea

tmen

t ini

tiatio

n. In

Boe

hme

2011

, for

sm

ear-n

egat

ive

cultu

re-p

ositi

ve T

B,

the

med

ian

dela

y un

til b

egin

ning

trea

tmen

t bef

ore

Xper

t MTB

/RI

F w

as in

trodu

ced

was

56

days

(int

erqu

artil

e ra

nge

[IQR]

, 39

–81

days

) com

pare

d w

ith 5

day

s (IQ

R, 2

–8 d

ays)

afte

r Xpe

rt M

TB/

RIF

was

intro

duce

d. In

Van

Rie

201

3, fo

r sm

ear-n

egat

ive

cultu

re-p

ositi

ve T

B pa

tient

s w

ith p

ositi

ve re

sults

from

Xpe

rt M

TB/

RIF,

treat

men

t was

beg

un o

n th

e sa

me

day

com

pare

d w

ith a

fter 1

3 da

ys fo

r pa

tient

s di

agno

sed

by o

ther

met

hods

.e

One

unp

ublis

hed

study

was

incl

uded

in th

e an

alys

is. N

o fo

rmal

ass

essm

ent o

f pub

licat

ion

bias

was

con

duct

ed u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

bec

ause

suc

h te

chni

ques

hav

e no

t be

en fo

und

to b

e he

lpfu

l in

studi

es o

f dia

gnos

tic a

ccur

acy.

How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to b

e m

inim

al s

ince

Xpe

rt M

TB/R

IF is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.

XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE 45Ta

ble

9. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

iagn

osin

g pu

lmon

ary

TB in

spu

tum

sm

ear-

posi

tive

adul

ts

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r det

ectio

n of

pul

mon

ary

TB in

sm

ear-p

ositi

ve in

divi

dual

s?Pa

rtici

pant

s: A

dults

who

are

sm

ear p

ositi

ve a

nd c

ultu

re p

ositi

ve

Setti

ng: M

ainl

y in

term

edia

te-le

vel l

abor

ator

ies

and

prim

ary

heal

th-c

are

faci

litie

s Ta

rget

con

ditio

n: P

ulm

onar

y TB

Re

fere

nce

stand

ard:

Sol

id c

ultu

re o

r liq

uid

cultu

reN

umbe

r of s

tudi

es (n

umbe

r of p

artic

ipan

ts): 2

3 (1

952)

Pool

ed s

ensit

ivity

: 98%

(95%

CrI,

97–

99%

); po

oled

spe

cific

ity: N

ot e

stim

ated

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

25

/100

0Pr

eval

ence

50

/100

0Pr

eval

ence

10

0/10

00

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-se

ctio

nal

Non

eaN

oneb

Non

ecN

one

Und

etec

tedd

Hig

h

25(2

4-25

)49

(49-

50)

98(9

7-99

)Fa

lse n

egat

ives

(in

divi

dual

s in

corre

ctly

cl

assifi

ed a

s no

t hav

ing

TB)

Cro

ss-

sect

iona

l N

onea

Non

ebN

onec

Non

eU

ndet

ecte

ddH

igh

1

(0-1

)1

(1-2

)2

(1-3

)

False

pos

itive

se (in

divi

dual

s in

corre

ctly

cl

assifi

ed a

s ha

ving

TB

––

––

––

––

––

True

neg

ativ

ese

(indi

vidu

als

with

out T

B)–

––

––

––

––

–

CrI,

cre

dibl

e in

terv

al.

a Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

The

maj

ority

of s

tudi

es e

nrol

led

indi

vidu

als

cons

ecut

ivel

y, a

nd a

sses

sed

the

resu

lt of

the

refe

renc

e sta

ndar

d bl

inde

d to

the

resu

lt fro

m X

pert

MTB

/RI

F. Th

e Xp

ert M

TB/

RIF

resu

lt is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

. b

The

maj

ority

of s

tudi

es (1

6/23

; 70%

) eva

luat

ed X

pert

MTB

/RI

F in

set

tings

whe

re it

was

inte

nded

to b

e us

ed. A

lthou

gh s

tudi

es o

f dia

gnos

tic a

ccur

acy

may

not

pro

vide

dire

ct e

vide

nce

abou

t pat

ient

-im

porta

nt o

utco

mes

, the

qua

lity

of th

e ev

iden

ce w

as n

ot d

owng

rade

d.c

The

estim

ates

of s

ensit

ivity

wer

e hi

ghly

con

siste

nt.

d O

ne u

npub

lishe

d stu

dy w

as in

clud

ed in

the

anal

ysis.

No

form

al a

sses

smen

t of p

ublic

atio

n bi

as w

as c

ondu

cted

usin

g m

etho

ds s

uch

as fu

nnel

plo

ts or

regr

essio

n te

sts b

ecau

se s

uch

tech

niqu

es h

ave

not b

een

foun

d to

be

help

ful f

or s

tudi

es o

f dia

gnos

tic a

ccur

acy.

How

ever

, re

porti

ng b

ias

was

con

sider

ed to

be

min

imal

sin

ce X

pert

MTB

/RI

F is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.

e Th

e po

oled

spe

cific

ity fo

r Xpe

rt M

TB/

RIF

was

not

esti

mat

ed in

thes

e stu

dies

bec

ause

alm

ost a

ll pa

rtici

pant

s w

ere

cons

ider

ed to

be

true

posit

ives

for T

B.


ble

10. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

iagn

osin

g pu

lmon

ary

TB in

spu

tum

sm

ear-

nega

tive

adul

ts

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r det

ectio

n of

cul

ture

-con

firm

ed p

ulm

onar

y TB

in s

mea

r-neg

ativ

e in

divi

dual

s?Pa

rtici

pant

s: A

dults

who

wer

e sm

ear-n

egat

ive

and

susp

ecte

d of

hav

ing

pulm

onar

y TB

Se

tting

: Mai

nly

inte

rmed

iate

-leve

l lab

orat

orie

s an

d pr

imar

y he

alth

-car

e fa

cilit

ies

Targ

et c

ondi

tion:

Pul

mon

ary

TB

Refe

renc

e sta

ndar

d: S

olid

cul

ture

or l

iqui

d cu

lture

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 23

(715

1)

Pool

ed s

ensit

ivity

: 68%

(95%

CrI,

61–

74%

); po

oled

spe

cific

ity: 9

9% (9

5% C

rI, 9

8–99

%)

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

smea

r-ne

gativ

e in

divi

dual

s te

sted

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

25

/100

0Pr

eval

ence

50

/100

0Pr

eval

ence

10

0/10

00

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-se

ctio

nal

Non

eaN

oneb

Serio

us (–

1)c

Non

edU

ndet

ecte

deM

oder

ate

17

(15-

19)

34(3

1-37

)68

(61-

74)

False

neg

ativ

es

(indi

vidu

als

inco

rrect

ly

clas

sified

as

not h

avin

g TB

)

Cro

ss-

sect

iona

l N

onea

Non

ebSe

rious

(–1)

cN

oned

Und

etec

tede

Mod

erat

e

8(7

-10)

16(1

3-20

)32

(26-

39)

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cl

assifi

ed a

s ha

ving

TB

Cro

ss-

sect

iona

l N

onea

Non

ebN

one

Non

eU

ndet

ecte

deH

igh

10

(10-

20)

10(1

0-19

)9

(9-1

8)

True

neg

ativ

es

(indi

vidu

als

with

out T

B)C

ross

-se

ctio

nal

Non

eaN

oneb

Non

eN

one

Und

etec

tede

Hig

h

965

(956

-965

)94

1(9

31-9

41)

891

(882

-891

)

CrI,

cre

dibl

e in

terv

al.

a Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

The

maj

ority

of s

tudi

es e

nrol

led

indi

vidu

als

cons

ecut

ivel

y, a

nd a

sses

sed

the

resu

lts o

f the

refe

renc

e sta

ndar

d bl

inde

d to

the

resu

lt fro

m X

pert

MTB

/RI

F. Th

e Xp

ert M

TB/

RIF

resu

lt is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

. b

The

maj

ority

of s

tudi

es (1

6/23

; 70%

) eva

luat

ed X

pert

MTB

/RI

F in

set

tings

whe

re it

was

inte

nded

to b

e us

ed. A

lthou

gh s

tudi

es o

f dia

gnos

tic a

ccur

acy

may

not

pro

vide

dire

ct e

vide

nce

abou

t pat

ient

-im

porta

nt o

utco

mes

, the

qua

lity

of th

e ev

iden

ce w

as n

ot d

owng

rade

d.c

Ther

e w

as s

ome

varia

bilit

y in

sen

sitiv

ity e

stim

ates

acr

oss

studi

es.

This

hete

roge

neity

cou

ld n

ot b

e ex

plai

ned

by s

tudy

qua

lity

or b

y re

mov

ing

the

study

by

Law

n 20

11 fr

om th

e an

alys

is. T

he s

tudy

by

Law

n, w

hich

foun

d th

e lo

wes

t sen

sitiv

ity, e

valu

ated

the

use

of X

pert

MTB

/RI

F to

scr

een

HIV

-pos

itive

pat

ient

s w

ith a

dvan

ced

imm

unod

efici

ency

, reg

ardl

ess

of th

eir s

ympt

oms,

who

wer

e en

rollin

g in

an

tiret

rovi

ral t

hera

py. T

he o

bser

ved

sens

itivi

ty m

ay h

ave

varie

d am

ong

studi

es a

s a

resu

lt of

cha

ract

erist

ics

asso

ciat

ed w

ith th

e pa

rtici

pant

s, b

ut th

ere

wer

e in

suffi

cien

t dat

a to

inve

stiga

te th

is po

ssib

ility.

Th

e qu

ality

of t

he e

vide

nce

was

dow

ngra

ded

by o

ne p

oint

.d

At a

pre

test

prob

abili

ty o

f 10%

, the

con

fiden

ce in

terv

als

for t

rue

posit

ives

and

false

neg

ativ

es w

ere

rela

tivel

y na

rrow

.e

One

unp

ublis

hed

study

was

incl

uded

in th

e an

alys

is. N

o fo

rmal

ass

essm

ent o

f pub

licat

ion

bias

was

con

duct

ed u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

bec

ause

suc

h te

chni

ques

hav

e no

t bee

n fo

und

to b

e he

lpfu

l in

studi

es o

f dia

gnos

tic a

ccur

acy.

How

ever

, re

porti

ng b

ias

was

con

sider

ed to

be

min

imal

sin

ce X

pert

MTB

/RI

F is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.


ble

11. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

iagn

osin

g pu

lmon

ary

TB in

adu

lts li

ving

with

HIV

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r det

ectio

n of

pul

mon

ary

TB in

peo

ple

livin

g w

ith H

IV?

Parti

cipa

nts:

Adu

lts li

ving

with

HIV

and

sus

pect

ed o

f hav

ing

pulm

onar

y TB

Se

tting

: Mai

nly

inte

rmed

iate

-leve

l lab

orat

orie

s an

d pr

imar

y he

alth

-car

e fa

cilit

ies

Targ

et c

ondi

tion:

Pul

mon

ary

TB

Refe

renc

e sta

ndar

d: S

olid

cul

ture

or l

iqui

d cu

lture

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 7 (1

789)

Pool

ed s

ensit

ivity

: 79%

(95%

CrI,

70–

86%

); po

oled

spe

cific

ity: 9

8% (9

5% C

rI, 9

6–99

%)

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

25

/100

0Pr

eval

ence

50

/100

0Pr

eval

ence

10

0/10

00

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-se

ctio

nal

Non

eaN

oneb

Serio

us (–

1)c

Non

edU

ndet

ecte

deM

oder

ate

20

(18-

22)

40(3

5-43

)79

(70-

86)

False

neg

ativ

es

(indi

vidu

als

inco

rrect

ly

clas

sified

as

not h

avin

g TB

)

Cro

ss-

sect

iona

l N

onea

Non

ebSe

rious

(–1)

cN

oned

Und

etec

tede

Mod

erat

e

5(4

-8)

11(7

-15)

21(1

4-30

)

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cl

assifi

ed a

s ha

ving

TB

Cro

ss-

sect

iona

l N

onea

Non

ebN

one

Non

eU

ndet

ecte

deH

igh

20

(10-

39)

19(1

0-38

)18

(9-3

6)

True

neg

ativ

es

(indi

vidu

als

with

out T

B)C

ross

-se

ctio

nal

Non

eaN

oneb

Non

eN

one

Und

etec

tede

Hig

h

956

(936

-965

)93

1(9

12-9

41)

882

(864

-891

)

CrI,

cre

dibl

e in

terv

al.

a Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

All

studi

es e

nrol

led

indi

vidu

als

cons

ecut

ivel

y, a

nd a

sses

sed

the

resu

lt of

the

refe

renc

e sta

ndar

d bl

inde

d to

the

resu

lt fro

m X

pert

MTB

/RI

F. Th

e Xp

ert M

TB/

RIF

resu

lt is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

.b

The

ratin

g of

the

qual

ity o

f the

evi

denc

e m

ay b

e lo

wer

ed if

ther

e ar

e im

porta

nt d

iffer

ence

s in

the

tests

stu

died

and

the

expe

rtise

of t

hose

app

lyin

g th

e te

sts in

the

studi

es c

ompa

red

with

the

setti

ngs

for w

hich

the

reco

mm

enda

tions

are

inte

nded

. The

maj

ority

of s

tudi

es (6

/7;

86%

) eva

luat

ed X

pert

MTB

/RI

F in

set

tings

whe

re it

was

inte

nded

to b

e us

ed. A

lthou

gh s

tudi

es o

f dia

gnos

tic a

ccur

acy

may

no

t pro

vide

dire

ct e

vide

nce

abou

t pat

ient

-impo

rtant

out

com

es, t

he q

ualit

y of

the

evid

ence

was

not

dow

ngra

ded.

c Th

ere

was

som

e va

riabi

lity

in s

ensit

ivity

esti

mat

es a

cros

s stu

dies

but

this

hete

roge

neity

cou

ld n

ot b

e ex

plai

ned

by s

tudy

qua

lity.

The

obs

erve

d se

nsiti

vity

may

hav

e va

ried

amon

g stu

dies

as

a re

sult

of c

hara

cter

istic

s as

soci

ated

with

the

parti

cipa

nts,

but

ther

e w

ere

insu

ffici

ent d

ata

to in

vesti

gate

this

poss

ibili

ty. T

he q

ualit

y of

evi

denc

e w

as d

owng

rade

d by

one

poi

nt.

d A

t a p

rete

st pr

obab

ility

of 1

0%, t

he c

onfid

ence

inte

rval

s fo

r tru

e po

sitiv

es a

nd fa

lse n

egat

ives

wer

e re

lativ

ely

narro

w.

e N

o fo

rmal

ass

essm

ent o

f pub

licat

ion

bias

was

con

duct

ed u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

bec

ause

suc

h te

chni

ques

hav

e no

t bee

n fo

und

to b

e he

lpfu

l in

studi

es o

f dia

gnos

tic

accu

racy

. How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to b

e m

inim

al s

ince

Xpe

rt M

TB/

RIF

is a

new

test

that

has

had

con

sider

able

atte

ntio

n an

d sc

rutin

y.


ble

12. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

iagn

osin

g pu

lmon

ary

TB in

adu

lts w

ithou

t HIV

infe

ctio

n

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r det

ectio

n of

pul

mon

ary

TB in

adu

lts w

ithou

t HIV

infe

ctio

n?Pa

rtici

pant

s: A

dults

who

wer

e H

IV n

egat

ive

and

susp

ecte

d of

hav

ing

pulm

onar

y TB

Se

tting

: Mai

nly

inte

rmed

iate

-leve

l lab

orat

orie

s an

d pr

imar

y he

alth

-car

e fa

cilit

ies

Targ

et c

ondi

tion:

Pul

mon

ary

TB

Refe

renc

e sta

ndar

d: S

olid

cul

ture

or l

iqui

d cu

lture

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 7 (1

470)

Pool

ed s

ensit

ivity

: 86%

(95%

CrI,

76–

92%

); po

oled

spe

cific

ity: 9

9% (9

5% C

rI, 9

8–10

0%)

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

25

/100

0Pr

eval

ence

50

/100

0Pr

eval

ence

10

0/10

00

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-se

ctio

nal

Non

eaN

oneb

Serio

us (–

1)c

Non

edU

ndet

ecte

deM

oder

ate

22

(19-

23)

43(3

8-46

)86

(76-

92)

False

neg

ativ

es

(indi

vidu

als

inco

rrect

ly

clas

sified

as

not h

avin

g TB

)

Cro

ss-

sect

iona

l N

onea

Non

ebSe

rious

(–1)

cN

oned

Und

etec

tede

Mod

erat

e

4(2

-6)

7(4

-12)

14(8

-24)

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cl

assifi

ed a

s ha

ving

TB

Cro

ss-

sect

iona

l N

onea

Non

ebN

one

Non

eU

ndet

ecte

deH

igh

10

(0-2

0)10

(0-1

9)9

(0-1

8)

True

neg

ativ

es

(indi

vidu

als

with

out T

B)C

ross

-se

ctio

nal

Non

eaN

oneb

Non

eN

one

Und

etec

tede

Hig

h

956

(956

-975

)94

1(9

31-9

50)

891

(882

-900

)

CrI,

cre

dibl

e in

terv

al.

a Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

All

studi

es e

nrol

led

indi

vidu

als

cons

ecut

ivel

y, a

nd a

sses

sed

the

resu

lt fro

m th

e re

fere

nce

stand

ard

blin

ded

to th

e re

sult

from

Xpe

rt M

TB/

RIF.

The

Xper

t MTB

/RI

F re

sult

is au

tom

ated

and

was

con

sider

ed b

linde

d in

all

studi

es.

b Th

e ra

ting

of th

e qu

ality

of t

he e

vide

nce

may

be

low

ered

if th

ere

are

impo

rtant

diff

eren

ces

in th

e te

sts s

tudi

ed a

nd th

e ex

perti

se o

f tho

se a

pply

ing

the

tests

in th

e stu

dies

com

pare

d w

ith th

e se

tting

s fo

r whi

ch th

e re

com

men

datio

ns a

re in

tend

ed. T

he m

ajor

ity o

f stu

dies

(6/

7; 8

6%) e

valu

ated

Xpe

rt M

TB/

RIF

in s

ettin

gs w

here

it w

as in

tend

ed to

be

used

. Alth

ough

stu

dies

of d

iagn

ostic

acc

urac

y m

ay

not p

rovi

de d

irect

evi

denc

e ab

out p

atie

nt-im

porta

nt o

utco

mes

, the

qua

lity

of th

e ev

iden

ce w

as n

ot d

owng

rade

d.c

Ther

e w

as s

ome

varia

bilit

y in

sen

sitiv

ity e

stim

ates

acr

oss

studi

es b

ut th

is he

tero

gene

ity c

ould

not

be

expl

aine

d by

stu

dy q

ualit

y. T

he o

bser

ved

sens

itivi

ty m

ay h

ave

varie

d am

ong

studi

es a

s a

resu

lt of

cha

ract

erist

ics

asso

ciat

ed w

ith th

e pa

rtici

pant

s, b

ut th

ere

wer

e in

suffi

cien

t dat

a to

inve

stiga

te th

is po

ssib

ility.

The

qua

lity

of th

e ev

iden

ce w

as d

owng

rade

d by

one

poi

nt.

d A

t a p

rete

st pr

obab

ility

of 1

0%, t

he c

onfid

ence

inte

rval

s fo

r tru

e po

sitiv

es a

nd fa

lse n

egat

ives

wer

e re

lativ

ely

narro

w.

e N

o fo

rmal

ass

essm

ent o

f pub

licat

ion

bias

was

con

duct

ed u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

bec

ause

suc

h te

chni

ques

hav

e no

t bee

n fo

und

to b

e he

lpfu

l in

studi

es o

f dia

gnos

tic

accu

racy

. How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to b

e m

inim

al s

ince

Xpe

rt M

TB/

RIF

is a

new

test

that

has

had

con

sider

able

atte

ntio

n an

d sc

rutin

y.


ble

13. G

RAD

E ev

iden

ce p

rofil

e: th

e in

crem

enta

l yie

ld o

f Xpe

rt M

TB/R

IF c

ompa

red

with

mic

rosc

opy

in p

atie

nts

with

cul

ture

-con

firm

ed T

B

PIC

O q

uesti

on: W

hat i

s th

e in

crem

enta

l yie

ld o

f Xpe

rt M

TB/

RIF

com

pare

d w

ith m

icro

scop

y in

pat

ient

s w

ith c

ultu

re-c

onfir

med

TB?

Parti

cipa

nts:

Adu

lts w

ith c

ultu

re-c

onfir

med

TB

Setti

ng: M

ainl

y in

term

edia

te-le

vel l

abor

ator

ies

and

prim

ary

heal

th-c

are

faci

litie

s Ta

rget

con

ditio

n: P

ulm

onar

y TB

Re

fere

nce

stand

ard:

Sol

id c

ultu

re o

r liq

uid

cultu

reN

umbe

r of s

tudi

es (n

umbe

r of p

artic

ipan

ts): 2

1 (8

880)

Pool

ed s

ensit

ivity

for s

mea

r mic

rosc

opy

: 65%

(95%

CrI,

57–

72%

); po

oled

sen

sitiv

ity fo

r Xpe

rt M

TB/

RIF:

88%

(95%

CrI,

84–

92%

)

Out

com

e

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed (9

5% C

rI)a

Qua

lity

of

ev

iden

ceb

Prev

alen

ce 2

5/10

00Pr

eval

ence

50/

1000

Prev

alen

ce 1

00/1

000

Prev

alen

ce 3

00/1

000

Smea

r

mic

rosc

opy

Xper

t M

TB/R

IFSm

ear

m

icro

scop

yXp

ert

MTB

/RIF

Smea

r

mic

rosc

opy

Xper

t M

TB/R

IFSm

ear

m

icro

scop

yXp

ert

MTB

/RIF

True

pos

itive

s (in

divi

dual

s w

ith T

B)16

(14–

18)

22(2

1–23

)33

(29–

36)

44(4

2–46

)65

(57–

72)

88(8

4–92

)19

5 (1

71–2

16)

264

(252

–276

)H

igh

True

pos

itive

s

(abs

olut

e di

ffere

nce)

6 m

ore

11 m

ore

23 m

ore

69 m

ore

False

neg

ativ

es

(indi

vidu

als

inco

rrect

ly

clas

sified

as

not h

avin

g TB

)

9(7

–11)

3(2

–4)

18(1

4–22

)6

(4–8

)35

(28–

43)

12(8

–16)

105

(84–

129)

36(2

4–48

)H

igh

False

neg

ativ

es

(abs

olut

e di

ffere

nce)

6 fe

wer

12

few

er

23 fe

wer

69 fe

wer

CrI,

cre

dibl

e in

terv

al.

a Th

e se

nsiti

vity

resu

lts w

ere

take

n fro

m b

ivar

iate

ana

lyse

s (in

clud

ing

both

sen

sitiv

ity a

nd s

peci

ficity

) to

obta

in th

e va

lues

for t

rue

posit

ives

and

false

neg

ativ

es.

b Th

e G

RAD

E fra

mew

ork

was

use

d to

ass

ess

the

qual

ity o

f the

evi

denc

e. F

or m

icro

scop

y, th

e se

nsiti

vity

esti

mat

es fo

r ind

ivid

ual s

tudi

es w

ere

varia

ble

(rang

e, 2

9–83

%),

and

the

pool

ed s

ensit

ivity

es

timat

e w

as im

prec

ise.

The

mai

n re

ason

for h

eter

ogen

eity

and

impr

ecisi

on in

the

sens

itivi

ty e

stim

ate

was

con

sider

ed to

be

the

varia

bilit

y in

sm

ear-p

ositi

ve s

tatu

s ac

ross

stu

dies

. Se

vera

l add

ition

al

fact

ors

may

hav

e co

ntrib

uted

to th

is he

tero

gene

ity, i

nclu

ding

type

of m

icro

scop

y, m

etho

d of

spe

cim

en p

roce

ssin

g an

d H

IV s

tatu

s. T

he q

ualit

y of

evi

denc

e w

as n

ot d

owng

rade

d.


ble

14. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

iagn

osin

g pu

lmon

ary

TB in

adu

lts a

s an

add-

on te

st fo

llow

ing

nega

tive

sput

um-s

mea

r m

icro

scop

y

PIC

O q

uesti

on:

Wha

t is

the

diag

nosti

c ac

cura

cy o

f Xpe

rt M

TB/

RIF

for d

etec

tion

of p

ulm

onar

y TB

in a

dults

, w

here

Xpe

rt M

TB/

RIF

is us

ed a

s an

add

-on

test

follo

win

g a

nega

tive

smea

r-mic

rosc

opy

resu

lt?

Parti

cipa

nts:

Adu

lts w

ho a

re s

mea

r neg

ativ

e an

d su

spec

ted

of h

avin

g pu

lmon

ary

TB

Setti

ng: M

ainl

y in

term

edia

te-le

vel l

abor

ator

ies

and

prim

ary

heal

th-c

are

faci

litie

s Ta

rget

con

ditio

n: P

ulm

onar

y TB

Re

fere

nce

stand

ard:

Sol

id c

ultu

re o

r liq

uid

cultu

reN

umbe

r of s

tudi

es (n

umbe

r of p

artic

ipan

ts): 2

3 (7

151)

Pool

ed s

ensit

ivity

: 68%

(95%

CrI,

61–

74%

); po

oled

spe

cific

ity: 9

9% (9

5% C

rI, 9

8–99

%)

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

smea

r-ne

gativ

e in

divi

dual

s te

sted

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

25

/100

0Pr

eval

ence

50

/100

0Pr

eval

ence

10

0/10

00

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-se

ctio

nal

Non

eaN

oneb

Serio

us (–

1)c

Non

edU

ndet

ecte

deM

oder

ate

17

(15-

19)

34(3

1-37

)68

(61-

74)

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed a

s no

t ha

ving

TB)

Cro

ss-

sect

iona

l N

onea

Non

ebSe

rious

(–1)

cN

oned

Und

etec

tede

Mod

erat

e

8(7

-10)

16(1

3-20

)32

(26-

39)

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB

Cro

ss-

sect

iona

l N

onea

Non

ebN

one

Non

eU

ndet

ecte

deH

igh

10

(10-

20)

10(1

0-19

)9

(9-1

8)

True

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Cro

ss-

sect

iona

l N

onea

Non

ebN

one

Non

eU

ndet

ecte

deH

igh

96

5(9

56-9

65)

941

(931

-941

)89

1(8

82-8

91)

CrI,

cre

dibl

e in

terv

al.

a Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

The

maj

ority

of s

tudi

es e

nrol

led

indi

vidu

als

cons

ecut

ivel

y, a

nd a

sses

sed

the

resu

lt fro

m th

e re

fere

nce

stand

ard

blin

ded

to th

e re

sult

from

Xpe

rt M

TB/

RIF.

The

Xper

t MTB

/RI

F re

sult

is au

tom

ated

and

was

con

sider

ed b

linde

d in

all

studi

es.

b Th

e m

ajor

ity o

f stu

dies

(16/

23; 7

0%) e

valu

ated

Xpe

rt M

TB/

RIF

in s

ettin

gs w

here

it w

as in

tend

ed to

be

used

. Alth

ough

stu

dies

of d

iagn

ostic

acc

urac

y m

ay n

ot p

rovi

de d

irect

evi

denc

e ab

out p

atie

nt-

impo

rtant

out

com

es, t

he q

ualit

y of

the

evid

ence

was

not

dow

ngra

ded.

c Th

ere

was

som

e va

riabi

lity

in s

ensit

ivity

esti

mat

es a

cros

s stu

dies

but

this

hete

roge

neity

cou

ld n

ot b

e ex

plai

ned

by s

tudy

qua

lity

or b

y re

mov

ing

the

study

by

Law

n 20

11 fr

om th

e an

alys

is. T

he s

tudy

by

Law

n, w

hich

foun

d th

e lo

wes

t sen

sitiv

ity, e

valu

ated

the

use

of X

pert

MTB

/RI

F to

scr

een

HIV

-pos

itive

pat

ient

s w

ith a

dvan

ced

imm

unod

efici

ency

, reg

ardl

ess

of th

eir s

ympt

oms,

who

wer

e en

rollin

g in

an

tiret

rovi

ral t

hera

py. T

he o

bser

ved

sens

itivi

ty m

ay h

ave

varie

d am

ong

studi

es a

s a

resu

lt of

cha

ract

erist

ics

asso

ciat

ed w

ith th

e pa

rtici

pant

s, b

ut th

ere

wer

e in

suffi

cien

t dat

a to

inve

stiga

te th

is po

ssib

ility.

Th

e qu

ality

of t

he e

vide

nce

was

dow

ngra

ded

by o

ne p

oint

.d

At a

pre

test

prob

abili

ty o

f 10%

, the

con

fiden

ce in

terv

als

for t

rue

posit

ives

and

false

neg

ativ

es w

ere

rela

tivel

y na

rrow

.e

One

unp

ublis

hed

study

was

incl

uded

in th

e an

alys

is. N

o fo

rmal

ass

essm

ent o

f pub

licat

ion

bias

was

con

duct

ed u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

bec

ause

suc

h te

chni

ques

hav

e no

t be

en fo

und

to b

e he

lpfu

l in

studi

es o

f dia

gnos

tic a

ccur

acy.

How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to b

e m

inim

al s

ince

Xpe

rt M

TB/R

IF is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.


Table 15. Sensitivity of Xpert MTB/RIF in smear-negative culture-confirmed pulmonary TB in individuals, by HIV status

PICO question: What is the diagnostic accuracy of Xpert MTB/RIF for detection of pulmonary TB in adults, where Xpert MTB/RIF is used as an add-on test following a negative smear microscopy result, stratified by HIV status?Participants: Adults with smear-negative culture-confirmed pulmonary TBSetting: One intermediate-level laboratory and one primary health-care clinicReference standard: Phenotypic culture using solid media or liquid mediaNumber of studies (number of participants): 2 (91)

Study HIV-positive participants (n = 33)

HIV-negative participants (n = 58)

Sensitivity (%)(95% CI)

Sensitivity (%)(95% CI)

Theron 2011 48(27–69)

45(25–67)

Van Rie 2013 60(27–86)

67(13–98)

CI, confidence interval.

Table 16. GRADE evidence profile: additional yield of Xpert MTB/RIF over microscopy in smear-negative TB

PICO question: What is the additional yield of Xpert MTB/RIF over microscopy in smear-negative TB?Participants: Adults who are smear negative and culture positiveSetting: Mainly intermediate-level laboratories and primary health-care facilities Target condition: Pulmonary TB Reference standard: Solid culture or liquid cultureNumber of studies (number of participants): 23 (7151)Polled sensitivity for smear microscopy: 0%; pooled sensitivity for Xpert MTB/RIF : 68% (95% CrI, 61–74%)

Outcome

Number of results/1000 individuals tested (95% CrI)aQuality

of evidenceb

Prevalence 25/1000 Prevalence 50/1000 Prevalence 100/1000

Smear microscopy

Xpert MTB/RIF

Smear microscopy

Xpert MTB/RIF

Smear microscopy

Xpert MTB/RIF

True positives (individuals with TB)

0 17(15–19)

0 34(31–37)

0 68(61–74)

Moderate

True positives (absolute difference)

17 more 34 more 68 more

False negatives (individuals incorrectly classified as not having TB)

25 8(7–10)

50 16(13–20)

100 32(26–29)

Moderate

False negatives (absolute difference)

17 fewer 34 fewer 68 fewer

CrI, credible interval.a The sensitivity results were taken from the bivariate analyses (including both sensitivity and specificity) to obtain the values for true positives and false negatives. b The GRADE framework was used to assess the quality of the evidence. The quality of the evidence was downgraded one point for inconsistency/imprecision.


ble

17.

GRA

DE

evid

ence

pro

file:

acc

urac

y of

Xpe

rt M

TB/R

IF in

det

ectin

g rif

ampi

cin

resi

stanc

e, w

here

Xpe

rt M

TB/R

IF r

epla

ces

phen

otyp

ic c

ultu

re-

base

d dr

ug-s

usce

ptib

ility

testi

ng a

s th

e in

itial

test

PIC

O q

uesti

on:

Wha

t is

the

diag

nosti

c ac

cura

cy o

f Xpe

rt M

TB/

RIF

for d

etec

tion

of ri

fam

pici

n re

sista

nce,

whe

re X

pert

MTB

/RI

F is

used

as

an in

itial

test

repl

acin

g ph

enot

ypic

cul

ture

-bas

ed d

rug-

susc

eptib

ility

testi

ng?

Parti

cipa

nts:

Adu

lts w

ith c

onfir

med

TB.

Set

ting:

Mai

nly

inte

rmed

iate

-leve

l lab

orat

orie

s an

d pr

imar

y he

alth

-car

e fa

cilit

ies.

Targ

et c

ondi

tion:

Rifa

mpi

cin

resis

tanc

e. R

efer

ence

sta

ndar

d: P

heno

typi

c cu

lture

-bas

ed d

rug-

susc

eptib

ility

testi

nga .

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

poo

led

sens

itivi

ty: 1

7 (5

55).

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

poo

led

spec

ifici

ty: 2

4 (2

414)

.Po

oled

sen

sitiv

ity: 9

5% (9

5% C

rI, 9

0–97

%);

Pool

ed s

peci

ficity

: 98%

(95%

CrI,

97–

99%

)b

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

50

/100

0cPr

eval

ence

10

0/10

00c

True

pos

itive

s (in

divi

dual

s w

ith T

B an

d rif

ampi

cin

resis

tanc

e)C

ross

-se

ctio

nal

Non

edN

onee

Non

efN

one

Und

etec

tedg

Hig

h

84(4

5-49

)14

3(1

35-1

46)

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed

as h

avin

g TB

that

is ri

fam

pici

n su

scep

tible

)C

ross

-se

ctio

nal

Non

edN

onee

Non

efN

one

Und

etec

tedg

Hig

h

3(2

-5)

5(8

-15)

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed

as h

avin

g TB

with

rifa

mpi

cin

resis

tanc

e)C

ross

-se

ctio

nal

Non

edN

onee

Non

efN

one

Und

etec

tedg

Hig

h

19

(10-

29)

17

(9-2

6)Tr

ue n

egat

ives

(ind

ivid

uals

with

TB

that

is

rifam

pici

n su

scep

tible

)C

ross

-se

ctio

nal

Non

edN

onee

Non

efN

one

Und

etec

tedg

Hig

h

931

(922

-941

)83

3 (8

25-8

42)

CrI,

cre

dibl

e in

terv

al.

a Ph

enot

ypic

dru

g-su

scep

tibili

ty te

sting

is c

onsid

ered

an

impe

rfect

refe

renc

e sta

ndar

d bu

t its

use

shou

ld n

ot d

owng

rade

the

ratin

g of

the

qual

ity o

f evi

denc

e.

b Es

timat

es o

f sen

sitiv

ity a

nd s

peci

ficity

wer

e de

term

ined

sep

arat

ely

usin

g un

ivar

iate

ana

lyse

s.c

Prev

alen

ce e

stim

ates

wer

e pr

ovid

ed b

y th

e W

HO

Ste

erin

g G

roup

. Th

e up

per l

imit

for t

he p

reva

lenc

e of

rifa

mpi

cin

resis

tanc

e in

new

cas

es w

as e

stim

ated

to b

e 5%

(50/

1000

cas

es);

the

low

er

limit

for t

he p

reva

lenc

e of

rifa

mpi

cin

resis

tanc

e in

pre

viou

sly tr

eate

d ca

ses

was

esti

mat

ed to

be

15%

(150

/10

00 c

ases

).d

The

QU

AD

AS-

2 to

ol w

as u

sed

to a

sses

s th

e ris

k of

bia

s. T

he m

ajor

ity o

f stu

dies

enr

olle

d in

divi

dual

s co

nsec

utiv

ely,

and

ass

esse

d th

e re

sult

from

the

refe

renc

e sta

ndar

d bl

inde

d to

the

resu

lt fro

m X

pert

MTB

/RI

F. Th

e Xp

ert M

TB/

RIF

resu

lt is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

. e

The

ratin

g of

the

qual

ity o

f the

evi

denc

e m

ay b

e lo

wer

ed if

ther

e ar

e im

porta

nt d

iffer

ence

s in

the

tests

stu

died

and

the

expe

rtise

of t

hose

app

lyin

g th

e te

sts in

the

studi

es c

ompa

red

with

the

setti

ngs

for w

hich

the

reco

mm

enda

tions

are

inte

nded

. The

maj

ority

of s

tudi

es (1

2/17

, 71%

for s

ensit

ivity

; and

16/

24, 6

7% fo

r spe

cific

ity) e

valu

ated

Xpe

rt M

TB/

RIF

in s

ettin

gs w

here

it w

as in

tend

ed to

be

used

. Onl

y tw

o stu

dies

eva

luat

ed X

pert

MTB

/RI

F w

ith th

e G

4 ca

rtrid

ge. I

n 20

13, t

he G

4 ca

rtrid

ges

wer

e th

e on

ly ty

pe o

f car

tridg

e av

aila

ble.

It is

pos

sible

that

the

perfo

rman

ce o

f Xpe

rt M

TB/

RIF

usin

g th

e G

4 ca

rtrid

ge w

ill be

diff

eren

t. A

lthou

gh s

tudi

es o

f dia

gnos

tic a

ccur

acy

may

not

pro

vide

dire

ct e

vide

nce

abou

t pat

ient

-impo

rtant

out

com

es, t

he q

ualit

y of

the

evid

ence

was

not

dow

ngra

ded.

f Es

timat

es o

f se

nsiti

vity

and

spe

cific

ity w

ere

cons

isten

t. Se

vera

l stu

dies

hav

e su

gges

ted

that

det

erm

inin

g th

e sp

ecifi

city

of

a m

olec

ular

met

hod

of d

rug-

susc

eptib

ility

testi

ng,

such

as

Xper

t MTB

/RI

F, us

ing

phen

otyp

ic d

rug-

susc

eptib

ility

testi

ng a

lone

may

und

eres

timat

e th

e sp

ecifi

city

of t

he m

olec

ular

test.

Gen

e se

quen

cing

of d

iscor

dant

resu

lts o

btai

ned

usin

g Xp

ert M

TB/

RIF

and

phen

otyp

ic

drug

-susc

eptib

ility

testi

ng (d

iscre

pant

ana

lysis

) is

ofte

n re

solve

d in

favo

ur o

f Xpe

rt M

TB/

RIF,

sugg

estin

g th

at X

pert

MTB

/RI

F ha

s hi

gher

spe

cific

ity. H

owev

er, d

iscre

pant

ana

lyse

s m

ay in

trodu

ce b

ias.

g O

ne u

npub

lishe

d stu

dy w

as in

clud

ed in

the

anal

ysis.

No

form

al a

sses

smen

t of p

ublic

atio

n bi

as w

as c

ondu

cted

usin

g m

etho

ds s

uch

as fu

nnel

plo

ts or

regr

essio

n te

sts b

ecau

se s

uch

tech

niqu

es h

ave

not b

een

foun

d to

be

help

ful i

n stu

dies

of d

iagn

ostic

acc

urac

y. H

owev

er,

repo

rting

bia

s w

as c

onsid

ered

to b

e m

inim

al s

ince

Xpe

rt M

TB/

RIF

is a

new

test

that

has

had

con

sider

able

atte

ntio

n an

d sc

rutin

y.


ble

18. A

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

TB in

lym

ph n

ode

fluid

and

tiss

ue (A

. Evi

denc

e pr

ofile

, B. S

umm

ary

of fi

ndin

gs)

PIC

O q

uesti

on:

Wha

t is

the

diag

nosti

c ac

cura

cy o

f Xpe

rt M

TB/

RIF

for

TB d

etec

tion

in ly

mph

nod

e flu

id a

nd ti

ssue

, w

here

Xpe

rt M

TB/

RIF

is us

ed a

s a

repl

acem

ent t

est f

or u

sual

pra

ctic

e?

A. E

vide

nce

profi

le Out

com

eaSt

udy

desi

gnFa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

eQ

ualit

y of

ev

iden

ceLi

mita

tions

Indi

rect

ness

In

cons

iste

ncy

Impr

ecis

ion

Publ

icat

ion

bias

True

pos

itive

s (in

divi

dual

s w

ith T

B)M

ajor

ity

cros

s-sec

tiona

lSe

rious

(–1)

bN

onec

Serio

us (–

1)d

Serio

us (–

1)e

Und

etec

tedf

Very

low

False

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Maj

ority

cr

oss-s

ectio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Maj

ority

cr

oss-s

ectio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

False

neg

ativ

es (p

atie

nts

inco

rrect

ly c

lass

ified

as

not

hav

ing

TB)

Maj

ority

cr

oss-s

ectio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

a Fo

r eac

h ou

tcom

e, th

e ra

ting

of th

e qu

ality

of e

vide

nce

starte

d as

hig

h w

hen

the

study

des

igns

wer

e ra

ndom

ized

con

trolle

d tri

als

or h

igh-

qual

ity o

bser

vatio

nal s

tudi

es (s

uch

as c

ross

-sect

iona

l stu

dies

w

ith d

iagn

ostic

unc

erta

inty

and

dire

ct c

ompa

rison

of t

he re

sults

of t

he in

dex

test

with

a re

fere

nce

stand

ard)

, or l

ow w

hen

ther

e w

ere

case

–con

trol s

tudi

es. T

he e

vide

nce

was

dow

ngra

ded

one

poin

t w

hen

a se

rious

issu

e w

as id

entifi

ed;

it w

as d

owng

rade

d tw

o po

ints

whe

n a

very

ser

ious

issu

e w

as id

entifi

ed in

any

of t

he o

ther

five

fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce:

lim

itatio

ns,

indi

rect

ness

, inc

onsis

tenc

y, im

prec

ision

or p

ublic

atio

n bi

as.

b Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

The

maj

ority

of s

tudi

es e

nrol

led

patie

nts

cons

ecut

ivel

y, a

nd a

sses

sed

the

resu

lt fro

m th

e re

fere

nce

stand

ard

blin

ded

to th

e re

sult

from

Xpe

rt M

TB/

RIF.

The

Xper

t MTB

/RI

F re

sult

is au

tom

ated

and

was

con

sider

ed b

linde

d in

all

studi

es.

How

ever

, th

e va

riabl

e us

e of

the

com

posit

e re

fere

nce

stand

ard

acro

ss s

tudi

es w

as a

con

cern

for

its

poss

ibili

ty o

f int

rodu

cing

bia

s. T

he e

vide

nce

was

dow

ngra

ded

one

poin

t.c

The

maj

ority

of s

tudi

es u

sed

Xper

t MTB

/RI

F in

terti

ary

care

cen

tres

or re

fere

nce

labo

rato

ries.

Bec

ause

obt

aini

ng a

sam

ple

requ

ires

an in

vasiv

e pr

oced

ure,

the

test

will

likel

y be

per

form

ed a

t hig

her

leve

ls of

car

e th

an a

re fe

asib

le fo

r dia

gnos

ing

pulm

onar

y TB

. Thu

s, th

e stu

dy p

opul

atio

ns a

re p

roba

bly

repr

esen

tativ

e of

the

popu

latio

n th

at w

ill re

ceiv

e th

e te

st. T

he e

vide

nce

was

not

dow

ngra

ded.

d

Une

xpla

ined

het

erog

enei

ty in

the

findi

ngs

mig

ht o

rigin

ate

from

diff

eren

ces

in s

ampl

e pr

oces

sing,

the

cond

ition

of s

ampl

es, a

nd d

iffer

ence

s in

stu

dy p

opul

atio

ns (f

or e

xam

ple,

var

ying

pre

vale

nces

of

TB

or H

IV).

The

evid

ence

was

dow

ngra

ded

one

poin

t.e C

onfid

ence

inte

rval

s ar

e w

ide,

like

ly d

ue in

par

t to

unex

plai

ned

hete

roge

neity

as

desc

ribed

in fo

otno

te d

and

in p

art d

ue to

ver

ifica

tion

bias

bec

ause

of t

he u

se o

f an

impe

rfect

refe

renc

e sta

ndar

d.

The

evid

ence

was

dow

ngra

ded

one

poin

t.f U

npub

lishe

d stu

dies

wer

e in

clud

ed. D

ata

did

not a

llow

for f

orm

al a

sses

smen

t of p

ublic

atio

n bi

as u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

bec

ause

suc

h te

chni

ques

hav

e no

t bee

n fo

und

to b

e he

lpfu

l in

studi

es o

f dia

gnos

tic a

ccur

acy.

How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to b

e m

inim

al s

ince

Xpe

rt M

TB/

RIF

is a

new

test

that

has

had

con

sider

able

atte

ntio

n an

d sc

rutin

y.


B. Summary of findings

Reference standard: CultureNumber of studies (number of samples): 14 (849); 11 studies had more than 10 samplesPooled sensitivity: 85% (95% CI, 72–92); pooled specificity: 93% (95% CI, 80–97)Reference standard: Composite reference standardNumber of studies (number of samples): 5 (409)Pooled sensitivity: 84% (95% CI, 74–90%); pooled specificity: 99% (95% CI, 88–100%)

Outcome

Number of results/1000 individuals tested (95% CrI)Quality

of evidence

Prevalence 2.5% Prevalence 5% Prevalence 10%Culture CRS Culture CRS Culture CRS


21 (18–23)

21 (19–23)

43 (36–46)

42 (37–45)

85 (72–92)

84 (74–90)

Very low

True negatives (individuals without TB)

907 (780–946)

965(858–975)

884(760–922)

941 (836–950)

837(720–873)

891 (792–900)

Very low

False positives (individuals incorrectly classified as having TB)

68 (29–195)

10 (0–117)

67 (29–190)

10 (0–114)

63 (27–180)

9 (0–108)

Very low


4 (2–7)

3 (4–7)

8 (4–14)

8 (5–13)

15 (8–28)

16 (10–26)

Very low

CI, confidence interval; CRS, composite reference standard.


ble

19. A

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

TB in

ple

ural

flui

d (A

. Evi

denc

e pr

ofile

, B. S

umm

ary

of fi

ndin

gs)

PIC

O q

uesti

on:

Wha

t is

the

diag

nosti

c ac

cura

cy o

f Xpe

rt M

TB/

RIF

for T

B de

tect

ion

in p

leur

al fl

uid,

whe

re X

pert

MTB

/RI

F is

used

as

a re

plac

emen

t tes

t fo

r usu

al p

ract

ice?

A. E

vide

nce

profi

le Out

com

eaSt

udy

desi

gnFa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

eQ

ualit

y of

ev

iden

ceLi

mita

tions

Indi

rect

ness

In

cons

iste

ncy

Impr

ecis

ion

Publ

icat

ion

bias

True

pos

itive

s (in

divi

dual

s w

ith T

B)M

ajor

ity

cros

s-sec

tiona

lSe

rious

(–1)

bN

onec

Serio

us (–

1)d

Serio

us (–

1)e

Und

etec

tedf

Very

low

False

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Maj

ority

cr

oss-s

ectio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Maj

ority

cr

oss-s

ectio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

False

neg

ativ

es (p

atie

nts

inco

rrect

ly c

lass

ified

as

not

hav

ing

TB)

Maj

ority

cr

oss-s

ectio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

a Fo

r eac

h ou

tcom

e, th

e ra

ting

of th

e qu

ality

of e

vide

nce

starte

d as

hig

h w

hen

the

study

des

igns

wer

e ra

ndom

ized

con

trolle

d tri

als

or h

igh-

qual

ity o

bser

vatio

nal s

tudi

es (s

uch

as c

ross

-sect

iona

l stu

dies

w

ith d

iagn

ostic

unc

erta

inty

and

dire

ct c

ompa

rison

of t

he re

sults

of t

he in

dex

test

with

a re

fere

nce

stand

ard)

, or

low

whe

n th

ere

wer

e ca

se–c

ontro

l stu

dies

. The

evi

denc

e w

as d

owng

rade

d on

e po

int

whe

n a

serio

us is

sue

was

iden

tified

; it

was

dow

ngra

ded

two

poin

ts w

hen

a ve

ry s

erio

us is

sue

was

iden

tified

in a

ny o

f the

oth

er fi

ve fa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

e: li

mita

tions

, in

dire

ctne

ss, i

ncon

siste

ncy,

impr

ecisi

on o

r pub

licat

ion

bias

. b

The

QU

AD

AS-

2 to

ol w

as u

sed

to a

sses

s th

e ris

k of

bia

s. T

he m

ajor

ity o

f stu

dies

enr

olle

d pa

tient

s co

nsec

utiv

ely,

and

ass

esse

d th

e re

sult

from

the

refe

renc

e sta

ndar

d bl

inde

d to

the

resu

lt fro

m X

pert

MTB

/RI

F. Th

e Xp

ert M

TB/

RIF

resu

lt is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

. H

owev

er,

the

varia

ble

use

of th

e co

mpo

site

refe

renc

e sta

ndar

d ac

ross

stu

dies

was

a c

once

rn fo

r its

po

ssib

ility

of i

ntro

duci

ng b

ias.

The

evi

denc

e w

as d

owng

rade

d on

e po

int.

c Th

e m

ajor

ity o

f stu

dies

use

d Xp

ert M

TB/

RIF

in te

rtiar

y ca

re c

entre

s or

refe

renc

e la

bora

torie

s. B

ecau

se o

btai

ning

a s

ampl

e re

quire

s an

inva

sive

proc

edur

e, th

e te

st w

ill lik

ely

be p

erfo

rmed

at h

ighe

r le

vels

of c

are

than

are

feas

ible

for d

iagn

osin

g pu

lmon

ary

TB. T

hus,

the

study

pop

ulat

ions

are

pro

babl

y re

pres

enta

tive

of th

e po

pula

tion

that

will

rece

ive

the

test.

The

evi

denc

e w

as n

ot d

owng

rade

d.

d U

nexp

lain

ed h

eter

ogen

eity

in th

e fin

ding

s m

ight

orig

inat

e fro

m d

iffer

ence

s in

sam

ple

proc

essin

g, th

e co

nditi

on o

f the

sam

ple,

and

diff

eren

ces

in s

tudy

pop

ulat

ions

(for

exa

mpl

e, v

aryi

ng p

reva

lenc

es

of T

B or

HIV

). Th

e ev

iden

ce w

as d

owng

rade

d on

e po

int.

e C

onfid

ence

inte

rval

s ar

e w

ide,

like

ly d

ue in

par

t to

unex

plai

ned

hete

roge

neity

as

desc

ribed

in fo

otno

te d

and

in p

art d

ue to

ver

ifica

tion

bias

bec

ause

of t

he u

se o

f an

impe

rfect

refe

renc

e sta

ndar

d.

The

evid

ence

was

dow

ngra

ded

one

poin

t.f U

npub

lishe

d stu

dies

wer

e in

clud

ed. D

ata

did

not a

llow

for f

orm

al a

sses

smen

t of p

ublic

atio

n bi

as u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

bec

ause

suc

h te

chni

ques

hav

e no

t bee

n fo

und

to b

e he

lpfu

l in

studi

es o

f dia

gnos

tic a

ccur

acy.

How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to b

e m

inim

al s

ince

Xpe

rt M

TB/

RIF

is a

new

test

that

has

had

con

sider

able

atte

ntio

n an

d sc

rutin

y.



Reference standard: CultureNumber of studies (number of samples): 17 (1384), 16 studies had more than 10 samples Pooled sensitivity: 44% (95% CI, 25–65%); pooled specificity: 98% (95% CI, 95–99%)Reference standard: Composite reference standardNumber of studies (number of samples): 7 (698)Pooled sensitivity: 17% (95% CI, 8–34%); pooled specificity: 100% (95% CI, 94–100%)

Outcome


of evidence



11 (6–16)

4 (2–9)

22 (13–33)

9 (4–17)

44 (25–65)

17 (8–34)

Very low


956(926–965)

975(917–975)

931(903–941)

950 (893–950)

882(855–891)

900 (846–900)

Low


20 (10–49)

0 (0–59)

19 (10–48)

0 (0–57)

18 (9–45)

0 (0–54)

Low


14 (9–19)

21 (17–23)

28 (18–38)

42 (33–46)

56 (35–75)

83 (66–92)

Very low



ble

20. A

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

TB in

cer

ebro

spin

al fl

uid

(A. E

vide

nce

profi

le, B

. Sum

mar

y of

find

ings

)

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r TB

dete

ctio

n in

cer

ebro

spin

al fl

uid,

whe

re X

pert

MTB

/RI

F is

used

as

a re

plac

emen

t te

st fo

r usu

al p

ract

ice?

A. E

vide

nce

profi

le Out

com

eaSt

udy

desi

gnFa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

eQ

ualit

y of

ev

iden

ceLi

mita

tions

Indi

rect

ness

In

cons

iste

ncy

Impr

ecis

ion

Publ

icat

ion

bias

True

pos

itive

s (in

divi

dual

s w

ith T

B)M

ajor

ity

cros

s-sec

tiona

lSe

rious

(–1)

bN

onec

Serio

us (–

1)d

Serio

us (–

1)e

Und

etec

tedf

Very

low

False

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Maj

ority

cr

oss-s

ectio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfLo

w

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Maj

ority

cr

oss-s

ectio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfLo

w

False

neg

ativ

es (p

atie

nts

inco

rrect

ly c

lass

ified

as

not

hav

ing

TB)

Maj

ority

cr

oss-s

ectio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

a Fo

r eac

h ou

tcom

e, th

e ra

ting

of th

e qu

ality

of e

vide

nce

starte

d as

hig

h w

hen

the

study

des

igns

wer

e ra

ndom

ized

con

trolle

d tri

als

or h

igh-

qual

ity o

bser

vatio

nal s

tudi

es (s

uch

as c

ross

-sect

iona

l stu

dies

w

ith d

iagn

ostic

unc

erta

inty

and

dire

ct c

ompa

rison

of t

he re

sults

of t

he in

dex

test

with

a re

fere

nce

stand

ard)

, or l

ow w

hen

ther

e w

ere

case

–con

trol s

tudi

es. T

he e

vide

nce

was

dow

ngra

ded

one

poin

t w

hen

a se

rious

issu

e w

as id

entifi

ed;

it w

as d

owng

rade

d tw

o po

ints

whe

n a

very

ser

ious

issu

e w

as id

entifi

ed in

any

of t

he o

ther

five

fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce:

lim

itatio

ns,

indi

rect

ness

, inc

onsis

tenc

y, im

prec

ision

or p

ublic

atio

n bi

as.

b Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

The

maj

ority

of s

tudi

es e

nrol

led

patie

nts

cons

ecut

ivel

y, a

nd a

sses

sed

resu

lts fr

om th

e re

fere

nce

stand

ard

blin

ded

to th

e re

sult

from

Xpe

rt M

TB/

RIF.

The

Xper

t MTB

/RI

F re

sult

is au

tom

ated

and

was

con

sider

ed b

linde

d in

all

studi

es. H

owev

er, t

he v

aria

ble

use

of th

e co

mpo

site

refe

renc

e sta

ndar

d ac

ross

stu

dies

was

a c

once

rn fo

r its

poss

ibili

ty

of in

trodu

cing

bia

s. T

he e

vide

nce

was

dow

ngra

ded

one

poin

t.c

The

maj

ority

of s

tudi

es u

sed

Xper

t MTB

/RI

F in

terti

ary

care

cen

tres

or re

fere

nce

labo

rato

ries.

Bec

ause

obt

aini

ng a

sam

ple

requ

ires

an in

vasiv

e pr

oced

ure,

the

test

will

likel

y be

per

form

ed a

t hig

her

leve

ls of

car

e th

an a

re fe

asib

le fo

r dia

gnos

ing

pulm

onar

y TB

. Thu

s, th

e stu

dy p

opul

atio

ns a

re p

roba

bly

repr

esen

tativ

e of

the

popu

latio

n th

at w

ill re

ceiv

e th

e te

st. T

he e

vide

nce

was

not

dow

ngra

ded.

d

Une

xpla

ined

het

erog

enei

ty in

the

findi

ngs

mig

ht o

rigin

ate

from

diff

eren

ces

in s

ampl

e pr

oces

sing,

the

cond

ition

of t

he s

ampl

es, a

nd d

iffer

ence

s in

stu

dy p

opul

atio

ns (f

or e

xam

ple,

var

ying

pre

vale

nces

of

TB

or H

IV).

The

evid

ence

was

dow

ngra

ded

one

poin

t.e

Con

fiden

ce in

terv

als

are

wid

e, li

kely

due

in p

art t

o un

expl

aine

d he

tero

gene

ity a

s de

scrib

ed in

foot

note

d a

nd in

par

t due

to v

erifi

catio

n bi

as b

ecau

se o

f the

use

of a

n im

perfe

ct re

fere

nce

stand

ard.

Th

e ev

iden

ce w

as d

owng

rade

d on

e po

int.

f Unp

ublis

hed

studi

es w

ere

incl

uded

. Dat

a di

d no

t allo

w fo

r for

mal

ass

essm

ent o

f pub

licat

ion

bias

usin

g m

etho

ds s

uch

as fu

nnel

plo

ts or

regr

essio

n te

sts b

ecau

se s

uch

tech

niqu

es h

ave

not b

een

foun

d to

be

help

ful i

n stu

dies

of d

iagn

ostic

acc

urac

y. H

owev

er, r

epor

ting

bias

was

con

sider

ed to

be

min

imal

sin

ce X

pert

MTB

/RI

F is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.



Reference standard: CultureNumber of studies (number of samples): 16 (709); 13 studies had more than 10 samplesPooled sensitivity: 80% (95% CI, 62–90%); pooled specificity: 99% (95% CI, 96–100%)Reference standard: Composite reference standardNumber of studies (number of samples): 7 (698)Pooled sensitivity: 56% (95% CI, 44–66%); pooled specificity: 99% (95% CI, 95–100%)

Outcome


of evidence



20 (16–23)

14 (11–17)

40 (31–45)

28 (22–33)

80 (62–90)

56 (44–66)

Very low


965(936–975)

965(926–975)

941(912–950)

941 (903–950)

891(864–900)

891 (855–900)

Low


10 (0–39)

10 (0–49)

10 (0–38)

10 (0–48)

9 (0-36)

9 (0–45)

Low


5 (3-10)

9 (11-14)

10 (5-19)

22 (17-28)

20 (10-38)

56 (44-66)

Very low



ble

21. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

TB in

gas

tric

fluid

PIC

O q

uesti

on:

Wha

t is

the

diag

nosti

c ac

cura

cy o

f Xpe

rt M

TB/

RIF

for T

B de

tect

ion

in g

astri

c flu

id,

whe

re X

pert

MTB

/RI

F is

used

as

a re

plac

emen

t tes

t fo

r usu

al p

ract

ice?

Refe

renc

e sta

ndar

d: C

ultu

reN

umbe

r of s

tudi

es (n

umbe

r of s

ampl

es):

12 (1

258

sam

ples

); 8

studi

es h

ad m

ore

than

10

sam

ples

Po

oled

sen

sitiv

ity: 8

4% (9

5% C

I, 66

– 93

%);

pool

ed s

peci

ficity

: 98%

(95%

CI,

92–1

00%

)

Out

com

eaSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

2.

5%Pr

eval

ence

5%

Prev

alen

ce

10%

True

pos

itive

s (in

divi

dual

s w

ith T

B)M

ajor

ity

cros

s-se

ctio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

21(1

7-23

)42

(33-

47)

84(6

6-93

)

False

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Maj

ority

cr

oss-

sect

iona

l

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

956

(897

-975

)93

1(8

74-9

50)

882

(828

-900

)

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Maj

ority

cr

oss-

sect

iona

l

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

20(0

-78)

19(0

-76)

18(0

-72)

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed a

s no

t ha

ving

TB)

Maj

ority

cr

oss-

sect

iona

l

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

4(2

-9)

8(4

-17)

16(7

-34)

a Fo

r eac

h ou

tcom

e, th

e ra

ting

of th

e qu

ality

of e

vide

nce

starte

d as

hig

h w

hen

the

study

des

igns

wer

e ra

ndom

ized

con

trolle

d tri

als

or h

igh-

qual

ity o

bser

vatio

nal s

tudi

es (s

uch

as c

ross

-sect

iona

l stu

dies

with

dia

gnos

tic u

ncer

tain

ty a

nd d

irect

com

paris

on o

f ind

ex te

st re

sults

with

a re

fere

nce

stand

ard)

, or l

ow w

hen

ther

e w

ere

case

–con

trol s

tudi

es. T

he

evid

ence

was

dow

ngra

ded

one

poin

t whe

n a

serio

us is

sue

was

iden

tified

; it

was

dow

ngra

ded

two

poin

ts w

hen

a ve

ry s

erio

us is

sue

was

iden

tified

in a

ny o

f the

oth

er fi

ve

fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce:

lim

itatio

ns, i

ndire

ctne

ss, i

ncon

siste

ncy,

impr

ecisi

on o

r pub

licat

ion

bias

.b

The

QU

AD

AS-

2 to

ol w

as u

sed

to a

sses

s th

e ris

k of

bia

s. T

he m

ajor

ity o

f stu

dies

enr

olle

d pa

tient

s co

nsec

utiv

ely,

and

ass

esse

d re

sults

from

the

refe

renc

e sta

ndar

d bl

inde

d to

the

resu

lts fr

om X

pert

MTB

/RI

F. Th

e Xp

ert M

TB/

RIF

resu

lt is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

. H

owev

er,

the

varia

ble

use

of th

e co

mpo

site

refe

renc

e sta

ndar

d ac

ross

stu

dies

was

a c

once

rn fo

r its

poss

ibili

ty o

f int

rodu

cing

bia

s. T

he e

vide

nce

was

dow

ngra

ded

one

poin

t.c

The

maj

ority

of s

tudi

es u

sed

Xper

t MTB

/RI

F in

terti

ary

care

cen

tres

or re

fere

nce

labo

rato

ries.

Bec

ause

obt

aini

ng a

sam

ple

requ

ires

an in

vasiv

e pr

oced

ure,

the

test

will

likel

y be

per

form

ed a

t hig

her l

evel

s of c

are

than

are

feas

ible

for d

iagn

osin

g pu

lmon

ary

TB. T

hus,

the

study

pop

ulat

ions

are

pro

babl

y re

pres

enta

tive

of th

e po

pula

tion

that

will

rece

ive

the

test.

The

evi

denc

e w

as n

ot d

owng

rade

d.

d U

nexp

lain

ed h

eter

ogen

eity

in th

e fin

ding

s m

ight

orig

inat

e fro

m d

iffer

ence

s in

sam

ple

proc

essin

g, th

e co

nditi

on o

f the

sam

ples

, an

d di

ffere

nces

in s

tudy

pop

ulat

ions

(for

ex

ampl

e, v

aryi

ng p

reva

lenc

es o

f TB

or H

IV).

The

evid

ence

was

dow

ngra

ded

one

poin

t.e C

onfid

ence

inte

rval

s ar

e w

ide,

like

ly d

ue in

par

t to

unex

plai

ned

hete

roge

neity

as

desc

ribed

in fo

otno

te d

and

in p

art d

ue to

ver

ifica

tion

bias

bec

ause

of t

he u

se o

f an

impe

rfect

refe

renc

e sta

ndar

d. T

he e

vide

nce

was

dow

ngra

ded

one

poin

t.f

Unp

ublis

hed

studi

es w

ere

incl

uded

. D

ata

did

not a

llow

for

for

mal

ass

essm

ent o

f pu

blic

atio

n bi

as u

sing

met

hods

suc

h as

fun

nel p

lots

or r

egre

ssio

n te

sts b

ecau

se s

uch

tech

niqu

es h

ave

not b

een

foun

d to

be

help

ful i

n stu

dies

of d

iagn

ostic

acc

urac

y. H

owev

er, r

epor

ting

bias

was

con

sider

ed to

be

min

imal

sin

ce X

pert

MTB

/RI

F is

a ne

w te

st th

at

has

had

cons

ider

able

atte

ntio

n an

d sc

rutin

y.


ble

22. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

TB in

tiss

ue s

ampl

es

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r TB

dete

ctio

n in

tiss

ue s

ampl

es, w

here

Xpe

rt M

TB/

RIF

is us

ed a

s a

repl

acem

ent t

est

for u

sual

pra

ctic

e?Re

fere

nce

stand

ard:

Cul

ture

Num

ber o

f stu

dies

(num

ber o

f sam

ples

): 12

(699

); 10

stu

dies

had

mor

e th

an 1

0 sa

mpl

es

Pool

ed s

ensit

ivity

: 81%

(95%

CI,

68–

90%

); po

oled

spe

cific

ity: 9

8% (9

5% C

I, 87

–100

%)

Out

com

eaSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

2.

5%Pr

eval

ence

5%

Prev

alen

ce

10%

True

pos

itive

s (in

divi

dual

s w

ith T

B)M

ajor

ity

cros

s-se

ctio

nal

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

20(1

7-23

)41

(34-

45)

81(6

8-90

)

False

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Maj

ority

cr

oss-

sect

iona

l

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

956

(848

-975

)93

1(8

27-9

50)

882

(783

-900

)

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Maj

ority

cr

oss-

sect

iona

l

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

20(0

-127

)19

(0-1

24)

18(0

-117

)

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed a

s no

t ha

ving

TB)

Maj

ority

cr

oss-

sect

iona

l

Serio

us (–

1)b

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eU

ndet

ecte

dfVe

ry lo

w

5(3

-8)

10(5

-16)

19(1

0-32

)

CI,

confi

denc

e in

terv

al.

a Fo

r eac

h ou

tcom

e, th

e ra

ting

of th

e qu

ality

of e

vide

nce

starte

d as

hig

h w

hen

the

study

des

igns

wer

e ra

ndom

ized

con

trolle

d tri

als

or h

igh-

qual

ity o

bser

vatio

nal s

tudi

es (s

uch

as c

ross

-sect

iona

l stu

dies

w

ith d

iagn

ostic

unc

erta

inty

and

dire

ct c

ompa

rison

of i

ndex

test

resu

lts w

ith a

refe

renc

e sta

ndar

d),

or lo

w w

hen

ther

e w

ere

case

–con

trol s

tudi

es.

The

evid

ence

was

dow

ngra

ded

one

poin

t whe

n a

serio

us is

sue

was

iden

tified

; it w

as d

owng

rade

d tw

o po

ints

whe

n a

very

ser

ious

issu

e w

as id

entifi

ed in

any

of t

he o

ther

five

fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce:

lim

itatio

ns, i

ndire

ctne

ss,

inco

nsist

ency

, im

prec

ision

or p

ublic

atio

n bi

as.

b Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

The

maj

ority

of s

tudi

es e

nrol

led

patie

nts

cons

ecut

ivel

y, a

nd a

sses

sed

the

resu

lt fro

m th

e re

fere

nce

stand

ard

blin

ded

to th

e re

sult

from

Xpe

rt M

TB/

RIF.

The

Xper

t MTB

/RI

F re

sult

is au

tom

ated

and

was

con

sider

ed b

linde

d in

all

studi

es.

How

ever

, th

e va

riabl

e us

e of

the

com

posit

e re

fere

nce

stand

ard

acro

ss s

tudi

es w

as a

con

cern

for

its

poss

ibili

ty o

f int

rodu

cing

bia

s. T

he e

vide

nce

was

dow

ngra

ded

one

poin

t.c

The

maj

ority

of s

tudi

es u

sed

Xper

t MTB

/RI

F in

terti

ary

care

cen

tres

or re

fere

nce

labo

rato

ries.

Bec

ause

obt

aini

ng a

sam

ple

requ

ires

an in

vasiv

e pr

oced

ure,

the

test

will

likel

y be

per

form

ed a

t hig

her

leve

ls of

car

e th

an a

re fe

asib

le fo

r dia

gnos

ing

pulm

onar

y TB

. Thu

s, th

e stu

dy p

opul

atio

ns a

re p

roba

bly

repr

esen

tativ

e of

the

popu

latio

n th

at w

ill re

ceiv

e th

e te

st. T

he e

vide

nce

was

not

dow

ngra

ded.

d

Une

xpla

ined

het

erog

enei

ty in

the

findi

ngs

mig

ht o

rigin

ate

from

diff

eren

ces

in s

ampl

e pr

oces

sing,

the

cond

ition

of t

he s

ampl

es, a

nd d

iffer

ence

s in

stu

dy p

opul

atio

ns (f

or e

xam

ple,

var

ying

pre

vale

nces

of

TB

or H

IV).

The

evid

ence

was

dow

ngra

ded

one

poin

t.e

Con

fiden

ce in

terv

als

are

wid

e, li

kely

due

in p

art t

o un

expl

aine

d he

tero

gene

ity a

s de

scrib

ed in

foot

note

d a

nd in

par

t due

to v

erifi

catio

n bi

as b

ecau

se o

f the

use

of a

n im

perfe

ct re

fere

nce

stand

ard.

Th

e ev

iden

ce w

as d

owng

rade

d on

e po

int.

f Unp

ublis

hed

studi

es w

ere

incl

uded

. Dat

a di

d no

t allo

w fo

r for

mal

ass

essm

ent o

f pub

licat

ion

bias

usin

g m

etho

ds s

uch

as fu

nnel

plo

ts or

regr

essio

n te

sts b

ecau

se s

uch

tech

niqu

es h

ave

not b

een

foun

d to

be

help

ful i

n stu

dies

of d

iagn

ostic

acc

urac

y. H

owev

er, r

epor

ting

bias

was

con

sider

ed to

be

min

imal

sin

ce X

pert

MTB

/RI

F is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.


ble

23. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

rifam

pici

n re

sista

nce

in n

onre

spira

tory

spe

cim

ens

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r rifa

mpi

cin-

resis

tanc

e de

tect

ion

in n

onre

spira

tory

spe

cim

ens,

whe

re X

pert

MTB

/RI

F is

used

as

an in

itial

test

repl

acin

g ph

enot

ypic

cul

ture

-bas

ed d

rug-

susc

eptib

ility

testi

ng?

Num

ber o

f stu

dies

(num

ber o

f sam

ples

): 13

(566

)

Out

com

eabSt

udy

desi

gnFa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

eQ

ualit

y of

ev

iden

ceLi

mita

tions

Indi

rect

ness

In

cons

iste

ncy

Impr

ecis

ion

Publ

icat

ion

bias

True

pos

itive

s (in

divi

dual

s w

ith T

B)M

ajor

ity

cros

s-sec

tiona

lN

onec

Serio

us (–

1)d

Serio

us (–

1)e

Serio

us (–

1)f

Und

etec

tedg

Very

low

False

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Maj

ority

cr

oss-s

ectio

nal

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eSe

rious

(–1)

fU

ndet

ecte

dgLo

w

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Maj

ority

cr

oss-s

ectio

nal

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eSe

rious

(–1)

fU

ndet

ecte

dgLo

w

False

neg

ativ

es (p

atie

nts

inco

rrect

ly c

lass

ified

as

not

hav

ing

TB)

Maj

ority

cr

oss-s

ectio

nal

Non

ecSe

rious

(–1)

dSe

rious

(–1)

eSe

rious

(–1)

fU

ndet

ecte

dgVe

ry lo

w

a G

iven

the

limite

d am

ount

of d

ata,

we

did

not c

alcu

late

sum

mar

y es

timat

es. S

ix s

tudi

es re

porte

d a

sens

itivi

ty o

f 100

% (3

4 rif

ampi

cin-

resis

tant

cas

es);

1 stu

dy h

ad 5

0% s

ensit

ivity

(2 ri

fam

pici

n-re

sista

nt

case

s); 1

stu

dy h

ad 0

% s

ensit

ivity

(1 ri

fam

pici

n-re

sista

nt c

ase)

.b

For e

ach

outc

ome,

the

ratin

g of

the

qual

ity o

f evi

denc

e sta

rted

as h

igh

whe

n th

e stu

dy d

esig

ns w

ere

rand

omiz

ed c

ontro

lled

trial

s or

hig

h-qu

ality

obs

erva

tiona

l stu

dies

(suc

h as

cro

ss-se

ctio

nal s

tudi

es

with

dia

gnos

tic u

ncer

tain

ty a

nd d

irect

com

paris

on o

f ind

ex te

st re

sults

with

a re

fere

nce

stand

ard)

, or

low

whe

n th

ere

wer

e ca

se–c

ontro

l stu

dies

. Th

e ev

iden

ce w

as d

owng

rade

d on

e po

int w

hen

a se

rious

issu

e w

as id

entifi

ed; i

t was

dow

ngra

ded

two

poin

ts w

hen

a ve

ry s

erio

us is

sue

was

iden

tified

in a

ny o

f the

oth

er fi

ve fa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

e: li

mita

tions

, ind

irect

ness

, in

cons

isten

cy, i

mpr

ecisi

on o

r pub

licat

ion

bias

. c

The

QU

AD

AS-

2 to

ol w

as u

sed

to a

sses

s th

e ris

k of

bia

s. T

he m

ajor

ity o

f stu

dies

enr

olle

d pa

tient

s co

nsec

utiv

ely,

and

ass

esse

d th

e re

sult

from

the

refe

renc

e sta

ndar

d bl

inde

d to

the

resu

lt fro

m X

pert

MTB

/RI

F. Th

e Xp

ert M

TB/

RIF

resu

lt is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

. The

refe

renc

e sta

ndar

d of

phe

noty

pic

drug

-susc

eptib

ility

testi

ng w

as u

sed

in a

ll stu

dies

. The

evi

denc

e w

as

not d

owng

rade

d. H

owev

er,

ther

e ar

e co

ncer

ns th

at p

heno

typi

c dr

ug-su

scep

tibili

ty te

sting

is n

ot a

per

fect

refe

renc

e sta

ndar

d fo

r det

ectin

g rif

ampi

cin

resis

tanc

e sin

ce s

eque

ncin

g da

ta h

ave

show

n th

at X

pert

MTB

/RI

F m

ight

cor

rect

ly id

entif

y re

sista

nce

that

is n

ot id

entifi

ed b

y ph

enot

ypic

dru

g-su

scep

tibili

ty te

sting

. Acr

oss

all s

tudi

es in

clud

ed h

ere,

six

false

-pos

itive

resu

lts fr

om X

pert

MTB

/RI

F te

sting

w

ere

iden

tified

. Of t

hose

, five

wer

e te

sted

with

seq

uenc

ing,

and

four

out

of fi

ve w

ere

foun

d to

hav

e a

mut

atio

n in

cod

on 5

33.

d Th

e m

ajor

ity o

f stu

dies

use

d Xp

ert M

TB/

RIF

in te

rtiar

y ca

re c

entre

s or

refe

renc

e la

bora

torie

s. B

ecau

se o

btai

ning

a s

ampl

e re

quire

s an

inva

sive

proc

edur

e, th

e te

st w

ill lik

ely

be p

erfo

rmed

at h

ighe

r le

vels

of c

are

than

are

feas

ible

for d

iagn

osin

g pu

lmon

ary

TB. T

hus,

the

study

pop

ulat

ions

are

pro

babl

y re

pres

enta

tive

of th

e po

pula

tion

that

will

rece

ive

the

test.

The

evi

denc

e w

as n

ot d

owng

rade

d.

e U

nexp

lain

ed h

eter

ogen

eity

in th

e fin

ding

s m

ight

orig

inat

e fro

m d

iffer

ence

s in

the

type

s of

spe

cim

en te

sted

and

in s

ampl

e pr

oces

sing.

The

evi

denc

e w

as d

owng

rade

d on

e po

int.

6 Th

ere

wer

e in

suffi

cien

t dat

a to

obt

ain

a su

mm

ary

estim

ate

for r

ifam

pici

n re

sista

nce.

The

resu

lts fr

om in

divi

dual

stu

dies

var

ied

wid

ely;

ther

efor

e, th

e ev

iden

ce w

as d

owng

rade

d on

e po

int.

f Unp

ublis

hed

studi

es w

ere

not i

nclu

ded

in th

e as

sess

men

t of u

sing

Xper

t MTB

/RI

F to

det

ect r

ifam

pici

n re

sista

nce

beca

use

data

col

lect

ion

on d

rug-

susc

eptib

ility

testi

ng w

as n

ot c

ompl

ete

for s

ome

of

the

studi

es. H

owev

er, r

epor

ting

bias

was

con

sider

ed to

be

min

imal

sin

ce X

pert

MTB

/RI

F is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.


ble

24. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

TB in

chi

ldre

n co

mpa

red

with

cul

ture

as a

refe

renc

e sta

ndar

d (A

. Exp

ecto

rate

d sp

utum

and

indu

ced

sput

um, B

. Gas

tric

lava

ge o

r as

pira

te, C

. Sum

mar

y of

find

ings

)

PIC

O q

uesti

on:

Wha

t is

the

diag

nosti

c ac

cura

cy o

f Xpe

rt M

TB/

RIF

for t

he d

etec

tion

of T

B in

chi

ldre

n ag

ains

t cul

ture

, w

here

Xpe

rt M

TB/

RIF

is us

ed a

s a

repl

acem

ent t

est f

or u

sual

pra

ctic

e?

A. E

xpec

tora

ted

sput

um a

nd in

duce

d sp

utum

Parti

cipa

nts:

Chi

ldre

n ag

ed 0

–15

year

s su

spec

ted

of h

avin

g TB

. Set

ting:

Mai

nly

terti

ary

care

refe

rral h

ospi

tals

and

univ

ersit

y ho

spita

lsTa

rget

con

ditio

n: P

ulm

onar

y TB

. Ref

eren

ce te

st: S

olid

cul

ture

or l

iqui

d cu

lture

. Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 10

(154

6)a .

Pool

ed s

ensit

ivity

: 66%

(95%

CrI,

52–

77%

); po

oled

spe

cific

ity: 9

8% (9

5% C

rI, 9

6–99

%)

Out

com

eSt

udy

desi

gnFa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

eQ

ualit

y of

ev

iden

ceLi

mita

tions

Indi

rect

ness

In

cons

iste

ncy

Impr

ecis

ion

Publ

icat

ion

bias

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-sect

iona

l N

oneb

Serio

us (–

1)c

Serio

us (–

1)d

Non

eeU

ndet

ecte

dfLo

w

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed

as n

ot h

avin

g TB

)C

ross

-sect

iona

l N

oneb

Serio

us (–

1)c

Serio

us (–

1)d

Non

eeU

ndet

ecte

dfLo

w

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Cro

ss-se

ctio

nal

Non

ebSe

rious

(–1)

cSe

rious

(–1)

dN

onee

Und

etec

tedf

Mod

erat

e

True

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Cro

ss-se

ctio

nal

Non

ebSe

rious

(–1)

cSe

rious

(–1)

dN

onee

Und

etec

tedf

Mod

erat

e

Tim

e to

dia

gnos

isgO

nly

desc

riptiv

e da

ta a

vaila

ble

Not

rate

d

CrI,

cre

dibl

e in

terv

al.

a Ra

chow

201

2 re

porte

d te

sting

one

coh

ort u

sing

expe

ctor

ated

spu

tum

and

one

usin

g in

duce

d sp

utum

; thi

s w

as c

ount

ed a

s tw

o stu

dies

.b

The

QU

AD

AS-

2 to

ol w

as u

sed

to a

sses

s the

risk

of b

ias.

The

maj

ority

of s

tudi

es e

nrol

led

patie

nts c

onse

cutiv

ely.

Res

ults

from

Xpe

rt M

TB/

RIF

are

auto

mat

ed a

nd w

ere

cons

ider

ed b

linde

d in

all

studi

es. T

he e

xclu

sion

of e

nrol

led

patie

nts

from

ana

lyse

s w

ere

wel

l exp

lain

ed in

mos

t stu

dies

; how

ever

, in

one

study

(LaC

ours

e, u

npub

lishe

d) 9

9/30

0 ch

ildre

n w

ere

excl

uded

bec

ause

they

wer

e lo

st to

follo

w u

p. A

noth

er s

tudy

(Bat

es 2

013)

enr

olle

d a

num

ber o

f chi

ldre

n w

ho

had

alre

ady

initi

ated

ant

i-TB

treat

men

t (77

/10

37 w

ith sa

mpl

es o

f exp

ecto

rate

d or

indu

ced

sput

um, o

r gas

tric

lava

ge o

r asp

irate

): 6

6 of

thes

e ch

ildre

n w

ere

cultu

re n

egat

ive

and

wer

e ex

clud

ed fr

om th

e an

alys

is; 1

1 w

ere

cultu

re p

ositi

ve

and

wer

e in

clud

ed. A

dditi

onal

ly, th

e fo

llow

ing

cons

ider

atio

ns w

ere

not t

aken

into

acc

ount

in ju

dgem

ents

abou

t the

lim

itatio

ns o

f the

stud

ies b

ut th

ey a

re im

porta

nt: c

ultu

re is

an

impe

rfect

refe

renc

e sta

ndar

d. In

chi

ldre

n, c

ultu

re se

nsiti

vity

va

ries f

rom

20%

to 7

0%, d

epen

ding

on

fact

ors s

uch

as th

e ch

ild’s

smea

r sta

tus,

seve

rity

of d

iseas

e, th

e ch

ild’s

age,

the

cultu

re m

etho

d us

ed, a

nd sp

ecim

en ty

pe. I

n or

der t

o im

prov

e th

e di

agno

stic

yiel

d, 9

/10

stud

ies p

erfo

rmed

at l

east

two

cultu

res f

or e

ach

child

, but

som

e pe

rform

ed a

s man

y as

six

cultu

res p

er c

hild

. The

sens

itivi

ty o

f Xpe

rt M

TB/

RIF

may

be

unde

resti

mat

ed w

hen

it is

com

pare

d ag

ains

t cul

ture

as a

refe

renc

e sta

ndar

d. T

he e

vide

nce

was

not

dow

ngra

ded.

c T

he ra

ting

of th

e qua

lity o

f evi

denc

e may

be l

ower

ed if t

here

are

impo

rtant

diff

eren

ces i

n the

pop

ulat

ions

stud

ied

and

the t

ests

studi

ed, a

nd in

the e

xper

tise o

f thos

e usin

g th

e tes

ts in

stud

ies c

ompa

red

with

the s

ettin

gs fo

r whi

ch th

e rec

omm

enda

tions

are

in

tend

ed. A

ll stu

dies

wer

e pe

rform

ed a

t hig

her le

vels

of ca

re –

that

is, h

ighe

r-lev

el re

ferra

l hos

pita

ls an

d un

iver

sity h

ospi

tals.

Six

of th

e 10

stud

ies (

60%

) incl

uded

onl

y inp

atie

nts;

4/

10 st

udie

s inc

lude

d in

patie

nts a

nd o

utpa

tient

s. In

the

paed

iatri

c po

pula

tion,

inpa

tient

s may

pre

sent

with

mor

e se

vere

dise

ase

and

have

a h

ighe

r lik

elih

ood

of b

eing

smea

r pos

itive

; the

y ar

e on

e gr

oup

that

mig

ht b

enefi

t fro

m X

pert

MTB

/RI

F. O

utpa

tient

s may

pre

sent

diff

eren

tly, w

ith le

ss se

vere

dise

ase

and

so b

e le

ss lik

ely t

o be

smea

r pos

itive

or c

ultu

re p

ositi

ve. O

utpa

tient

s are

not

repr

esen

ted

by th

e stu

dies

incl

uded

in th

is re

view

. Onl

y one

stud

y (Ra

chow

201

2) in

clud

ed a

bro

ader

gro

up o

f chi

ldre

n: a

ppro

xim

atel

y 20%

of c

hild

ren

susp

ecte

d of

ha

ving

TB

had

been

iden

tified

thro

ugh

cont

act tr

acin

g. Tw

o un

publ

ished

stud

ies (

Chi

sti a

nd La

Cou

rse)

incl

uded

chi

ldre

n w

ho w

ere

seve

rely

mal

nour

ished

, and

TB

was

one

diff

eren

tial d

iagn

osis.

(Chi

ldre

n in

the

Chi

sti st

udy

addi

tiona

lly h

ad

to h

ave

coug

h la

sting

long

er th

an 2

wee

ks o

r pne

umon

ia o

n X-

ray,

or b

oth.

) The

re w

ere

serio

us c

once

rns a

bout

indi

rect

ness

in th

e m

ajor

ity o

f stu

dies

per

form

ed in

hig

her-l

evel

refe

rral h

ospi

tals.

The

evi

denc

e w

as d

owng

rade

d by

one

poi

nt.

d The

poin

t esti

mat

es fo

r sen

sitiv

ity ra

nged

from

25%

to 1

00%

, ind

icat

ing

a hi

gh le

vel o

f het

erog

enei

ty. H

owev

er, t

he 9

5% c

onfid

ence

inte

rval

s ov

erla

pped

. Sub

grou

p an

alys

es s

ugge

sted

that

ther

e w

as a

n ef

fect

of s

mea

r sta

tus

on

over

all s

ensit

ivity

, w

ith a

poo

led

estim

ate

for s

ensit

ivity

am

ong

smea

r-pos

itive

chi

ldre

n of

96%

(95%

CrI,

90–

99%

) and

for s

mea

r-neg

ativ

e ch

ildre

n of

55%

(95%

CrI,

41–

69%

). Th

ere

was

no

rele

vant

het

erog

enei

ty a

mon

g sm

ear-

posit

ive

child

ren

(poo

led

estim

ates

of s

ensit

ivity

rang

e, 9

2–10

0%),

but t

here

was

con

sider

able

het

erog

enei

ty a

mon

g sm

ear-n

egat

ive

child

ren

(rang

e, 2

5–10

0%).

Subg

roup

ana

lyse

s by

age

gro

up (f

or th

ose

aged

0–4

yea

rs th

e po

oled

sen

sitiv

ity e

stim

ate

was

57%

with

95%

CrI,

36–

74%

; fo

r chi

ldre

n ag

ed 5

–15

year

s th

e po

oled

sen

sitiv

ity e

stim

ate

was

83%

with

95%

CrI,

68–

92%

) and

HIV

sta

tus

(for H

IV-p

ositi

ve c

hild

ren

the

pool

ed s

ensit

ivity

esti

mat

e w

as 7

5% w

ith 9

5% C

rI, 5

7–88

%;

for H

IV-n

egat

ive

child

ren

the

pool

ed s

ensit

ivity

esti

mat

e w

as 5

7% w

ith 9

5% C

rI, 4

1–71

%) i

ndic

ated

that

ther

e w

ere

diffe

renc

es a

mon

g th

ese

grou

ps,

but t

hese

esti

mat

es a

re h

eter

ogen

eous

and

se

em to

be

affe

cted

by

smea

r sta

tus.

Ove

rall,

het

erog

enei

ty c

anno

t be

fully

exp

lain

ed b

y th

e su

bgro

up a

naly

ses;

ther

efor

e, in

cons

isten

cy is

rate

d as

a s

erio

us c

once

rn fo

r the

sen

sitiv

ity e

stim

ates

. The

evi

denc

e w

as d

owng

rade

d by

on

e po

int.

Estim

ates

of s

peci

ficity

did

not

sho

w s

erio

us h

eter

ogen

eity

: poi

nt e

stim

ates

rang

ed fr

om 9

3% to

100

%. T

here

was

no

conc

ern

abou

t spe

cific

ity.

e Five

of t

he 1

0 stu

dies

incl

uded

mor

e th

an 1

00 c

hild

ren,

and

the

yiel

d of

cul

ture

am

ong

all s

tudi

es v

arie

d fro

m 2

% to

52%

. The

con

fiden

ce in

terv

al fo

r sen

sitiv

ity w

as w

ide

(cro

ssin

g +/

– 10

%),

whi

ch is

of c

once

rn. T

he e

vide

nce

was

not

dow

ngra

ded

furth

er g

iven

the

dow

ngra

ding

for i

ncon

siste

ncy.

The

re w

as n

o co

ncer

n ab

out s

peci

ficity

.f T

hree

unpu

blish

ed st

udie

s wer

e inc

lude

d. A

few

othe

r stu

dies

wer

e in p

rogr

ess,

but d

ata w

ere n

ot av

aila

ble f

or an

alys

is. Th

e dat

a tha

t wer

e inc

lude

d in t

he re

view

did n

ot al

low

for fo

rmal

asse

ssm

ent o

f pub

licat

ion b

ias u

sing m

etho

ds su

ch as

funn

el pl

ots

or re

gres

sion t

ests;

such

tech

niqu

es ha

ve no

t bee

n fou

nd to

be he

lpfu

l in st

udie

s eva

luat

ing d

iagn

ostic

accu

racy

. How

ever

, rep

ortin

g bia

s was

cons

ider

ed to

be m

inim

al si

nce X

pert M

TB/

RIF i

s a ne

w te

st th

at ha

s had

cons

ider

able

atte

ntio

n and

scru

tiny.

g Th

e time

to d

iagn

osis

was

mai

nly

repo

rted

for t

he tu

rnar

ound

tim

e in

the

labo

rato

ry ra

ther

than

from

the

day

of s

peci

men

col

lect

ion:

it ra

nged

from

6 d

ays

to 7

wee

ks fo

r cul

ture

(7 s

tudi

es) a

nd fr

om 2

hou

rs to

2.5

hou

rs fo

r Xpe

rt M

TB/

RIF.

Thre

e stu

dies

des

crib

ed th

e tim

e fro

m e

nrol

men

t or s

peci

men

col

lect

ion

to d

iagn

osis

usin

g Xp

ert M

TB/

RIF

(1 d

ay) a

nd c

ultu

re (2

1 da

ys u

sing

the

BAC

TEC

MG

IT 9

60, a

nd 3

0 da

ys u

sing

Löw

enste

in–J

ense

n m

ediu

m).

XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATE 63B.

Gas

tric

lava

ge o

r as

pira

te

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 7 (1

319)

Pool

ed s

ensit

ivity

: 66%

(95%

CrI,

51–

81%

); po

oled

spe

cific

ity: 9

8% (9

5% C

rI, 9

6–99

%)

Out

com

eSt

udy

desi

gnFa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

eQ

ualit

y of

ev

iden

ceLi

mita

tions

Indi

rect

ness

In

cons

iste

ncy

Impr

ecis

ion

Publ

icat

ion

bias

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-sect

iona

l N

onea

Serio

us (–

1)b

Serio

us (–

1)c

Non

e (se

rious

)dU

ndet

ecte

deLo

w

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed

as n

ot h

avin

g TB

)C

ross

-sect

iona

l N

onea

Serio

us (–

1)b

Serio

us (–

1)c

Non

e (se

rious

)dU

ndet

ecte

deLo

w

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Cro

ss-se

ctio

nal

Non

eaSe

rious

(–1)

bSe

rious

(–1)

cN

oned

Und

etec

tede

Mod

erat

e

True

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Cro

ss-se

ctio

nal

Non

eaSe

rious

(–1)

bSe

rious

(–1)

cN

oned

Und

etec

tede

Mod

erat

e

CrI,

cre

dibl

e in

terv

al.

a Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

Thr

ee o

f se

ven

studi

es e

nrol

led

indi

vidu

als

cons

ecut

ivel

y. R

esul

ts fro

m X

pert

MTB

/RI

F ar

e au

tom

ated

and

wer

e co

nsid

ered

blin

ded

in a

ll stu

dies

. Th

e ex

clus

ion

of e

nrol

led

patie

nts

from

ana

lyse

s w

ere

wel

l exp

lain

ed i

n m

ost

studi

es;

how

ever

in

one

study

(Ba

tes

2013

) a

num

ber

of c

hild

ren

who

had

alre

ady

initi

ated

ant

i-TB

tre

atm

ent

wer

e en

rolle

d (7

7/10

37,

for

spec

imen

s fro

m e

xpec

tora

ted

or i

nduc

ed s

putu

m a

nd o

f ga

stric

lav

age

and

aspi

rate

).

A t

otal

of

66 o

f th

ese

child

ren

wer

e cu

lture

neg

ativ

e an

d w

ere

excl

uded

; 11

wer

e cu

lture

pos

itive

and

wer

e in

clud

ed i

n th

e an

alys

is. A

dditi

onal

ly, t

he f

ollo

win

g co

nsid

erat

ions

wer

e no

t ta

ken

into

acc

ount

in

judg

emen

ts ab

out

the

limita

tions

of

the

studi

es b

ut t

hey

are

impo

rtant

: cu

lture

is

an i

mpe

rfect

ref

eren

ce s

tand

ard.

In

child

ren,

cul

ture

sen

sitiv

ity v

arie

s fro

m 2

0% t

o 70

%,

depe

ndin

g on

fac

tors

suc

h as

the

sev

erity

of

dise

ase,

the

ag

e of

the

chi

ld,

the

cultu

re m

etho

ds u

sed,

and

spe

cim

en t

ypes

. In

ord

er t

o im

prov

e th

e di

agno

stic

yiel

d, 9

/10

stu

dies

per

form

ed a

t le

ast

two

cultu

res

for

each

chi

ld,

but

som

e pe

rform

ed

as m

any

as s

ix c

ultu

res

per

child

. Th

e se

nsiti

vity

of

Xper

t MTB

/RI

F m

ay b

e un

dere

stim

ated

whe

n it

is co

mpa

red

agai

nst c

ultu

re a

s a

refe

renc

e sta

ndar

d. T

he e

vide

nce

was

not

dow

ngra

ded.

b

The

ratin

g of

the

qual

ity o

f evi

denc

e m

ay b

e lo

wer

ed if

ther

e ar

e im

porta

nt d

iffer

ence

s in

the

popu

latio

ns st

udie

d an

d th

e te

sts st

udie

d, a

nd in

the

expe

rtise

of t

hose

usin

g th

e te

sts in

the

studi

es c

ompa

red

with

the

setti

ngs f

or w

hich

the

reco

mm

enda

tions

are

inte

nded

. All

studi

es w

ere

perfo

rmed

at h

ighe

r lev

els o

f car

e –

that

is, h

ighe

r-lev

el re

ferra

l hos

pita

ls an

d un

iver

sity

hosp

itals.

Fou

r of t

he se

ven

studi

es

incl

uded

onl

y in

patie

nts;

one

incl

uded

inpa

tient

s and

out

patie

nts;

two

wer

e la

bora

tory

stud

ies t

hat d

id n

ot p

rovi

de in

form

atio

n on

the

patie

nt c

ohor

t. In

the

paed

iatri

c po

pula

tion,

inpa

tient

s may

pre

sent

w

ith m

ore

seve

re d

iseas

e an

d ha

ve a

hig

her l

ikel

ihoo

d of

bei

ng sm

ear p

ositi

ve; t

hey

are

one

grou

p th

at m

ight

ben

efit f

rom

Xpe

rt M

TB/

RIF.

Out

patie

nts m

ay p

rese

nt d

iffer

ently

, with

less

seve

re d

iseas

e an

d so

be

less

like

ly to

be

smea

r pos

itive

or c

ultu

re p

ositi

ve. O

utpa

tient

s are

not

repr

esen

ted

by th

e stu

dies

incl

uded

in th

is re

view

. One

unp

ublis

hed

study

(Chi

sti) i

nclu

ded

seve

rely

mal

nour

ished

chi

ldre

n w

ho

had

coug

h la

sting

long

er th

an 2

wee

ks o

r pne

umon

ia o

n X-

ray,

or b

oth;

TB

was

one

diff

eren

tial d

iagn

osis.

The

re w

ere

serio

us c

once

rns a

bout

indi

rect

ness

. The

evi

denc

e w

as d

owng

rade

d by

one

poi

nt.

c The

poin

t esti

mat

es fo

r sen

sitiv

ity ra

nged

from

40%

to 1

00%

, ind

icat

ing

a hi

gh le

vel o

f het

erog

enei

ty. H

owev

er, t

he 9

5% c

onfid

ence

inte

rval

s ov

erla

pped

. Sub

grou

p an

alys

es s

ugge

sted

that

ther

e w

as a

n ef

fect

of s

mea

r sta

tus

on o

vera

ll se

nsiti

vity,

with

a p

oole

d es

timat

e fo

r sen

sitiv

ity a

mon

g sm

ear-p

ositi

ve c

hild

ren

of 9

5% (9

5% C

rI, 8

3–99

%) a

nd fo

r sm

ear-n

egat

ive

child

ren

of 6

2% (9

%%

CrI,

44

– 80

%).

Ther

e w

as n

o re

leva

nt h

eter

ogen

eity

am

ong

smea

r–po

sitiv

e ch

ildre

n (p

oole

d es

timat

es o

f sen

sitiv

ity ra

nge,

92–

100%

), bu

t the

re w

as c

onsid

erab

le h

eter

ogen

eity

am

ong

smea

r-neg

ativ

e ch

ildre

n (ra

nge,

36–

100%

). A

naly

ses

of s

ubgr

oups

by

age

was

pos

sible

onl

y fo

r chi

ldre

n ag

ed 0

–4 y

ears

(poo

led

estim

ate

of s

ensit

ivity

was

57%

with

95%

CrI,

38–

75%

). Th

ree

studi

es in

clud

ed

only

chi

ldre

n ag

ed 0

–5 y

ears

; the

num

ber o

f gas

tric

lava

ge a

nd a

spira

tions

per

form

ed in

chi

ldre

n ag

ed 5

–15

in th

e re

mai

ning

stu

dies

was

sm

all.

Subg

roup

ana

lysis

for H

IV s

tatu

s w

as n

ot p

ossib

le

for s

peci

men

s fro

m g

astri

c la

vage

or a

spira

tion.

Ove

rall,

het

erog

enei

ty c

ould

not

be

fully

exp

lain

ed b

y th

e su

bgro

up a

naly

ses;

ther

efor

e, in

cons

isten

cy w

as ra

ted

as a

ser

ious

con

cern

for t

he s

ensit

ivity

es

timat

es. T

he e

vide

nce

was

dow

ngra

ded

by o

ne p

oint

. Esti

mat

es o

f spe

cific

ity d

id n

ot sh

ow se

rious

het

erog

enei

ty: p

oint

esti

mat

es ra

nged

from

93%

to 1

00%

. The

re w

as n

o co

ncer

n ab

out s

peci

ficity

. d

Thre

e of

7

studi

es

incl

uded

m

ore

than

10

0 ch

ildre

n,

and

the

yiel

d of

cu

lture

am

ong

all

studi

es

varie

d fro

m

2.8%

to

32

.7%

. Th

e co

nfide

nce

inte

rval

fo

r se

nsiti

vity

w

as

wid

e (c

ross

ing

+/–

10%

), w

hich

is

of

conc

ern.

Th

e ev

iden

ce

was

no

t do

wng

rade

d fu

rther

gi

ven

the

dow

ngra

ding

fo

r in

cons

isten

cy.

Ther

e w

as

no

conc

ern

abou

t sp

ecifi

city.

e Tw

o un

publ

ished

stu

dies

wer

e in

clud

ed. A

few

oth

er s

tudi

es w

ere

in p

rogr

ess,

but

dat

a w

ere

not a

vaila

ble

for a

naly

sis. T

he d

ata

that

wer

e in

clud

ed d

id n

ot a

llow

for f

orm

al a

sses

smen

t of p

ublic

atio

n bi

as u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

; suc

h te

chni

ques

hav

e no

t bee

n fo

und

to b

e he

lpfu

l in

studi

es e

valu

atin

g di

agno

stic

accu

racy

. How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to

be m

inim

al s

ince

Xpe

rt M

TB/

RIF

is a

new

test

that

has

had

con

sider

able

atte

ntio

n an

d sc

rutin

y.

64 XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATEC.

Sum

mar

y of

find

ings

Out

com

e

Xper

t MTB

/RIF

usi

ng s

peci

men

s of

exp

ecto

rate

or

indu

ced

sput

umXp

ert M

TB/R

IF u

sing

spe

cim

ens

of g

astr

ic la

vage

or

asp

irate

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

aQ

ualit

y

of

evid

ence

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

aQ

ualit

y

of

evid

ence

Prev

alen

ce

10/1

000b

Prev

alen

ce

50/1

000b

Prev

alen

ce

100/

1000

bPr

eval

ence

10

/100

0bPr

eval

ence

50

/100

0bPr

eval

ence

10

0/10

00b

True

pos

itive

s (in

divi

dual

s w

ith T

B)7

(5-8

)33

(26-

39)

66(5

2-77

)Lo

w

7(5

-8)

33(2

6-41

)66

(51-

81)

Low

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cl

assifi

ed a

s no

t hav

ing

TB)

3(2

-5)

17(1

2-24

)34

(23-

48)

Low

3

(2-5

)17

(10-

25)

34(1

9-49

)Lo

w

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cl

assifi

ed a

s ha

ving

TB)

20(1

0-40

)19

(10-

38)

18(9

-36)

Very

low

20

(10-

40)

19(1

0-38

)18

(9-3

6)M

oder

ate

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cl

assifi

ed a

s no

t hav

ing

TB)

970

(950

-980

)93

1(9

12-9

41)

882

(864

-891

)Ve

ry lo

w

970

(950

-980

)93

1(9

12-9

41)

882

(864

-891

)M

oder

ate

CrI,

cre

dibl

e in

terv

al.

a The

expe

cted

num

ber o

f Xpe

rt M

TB/

RIF

resu

lts w

as b

ased

on

the

sens

itivi

ty a

nd s

peci

ficity

esti

mat

es c

alcu

late

d fro

m a

com

paris

on w

ith c

ultu

re fo

r diff

eren

t pre

vale

nces

of T

B.

b The

estim

ates

of T

B pr

eval

ence

wer

e pr

ovid

ed b

y th

e W

HO

Ste

erin

g G

roup

.


ble

25. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

TB in

chi

ldre

n co

mpa

red

with

a c

linic

al re

fere

nce

stand

ard

(A. E

xpec

tora

ted

sput

um a

nd in

duce

d sp

utum

, and

gas

tric

lava

ge a

nd a

spira

te, B

. Sum

mar

y of

find

ings

)PI

CO

que

stion

: Wha

t is

the

diag

nosti

c ac

cura

cy o

f Xpe

rt M

TB/

RIF

for d

etec

tion

of T

B in

chi

ldre

n co

mpa

red

with

a c

linic

al re

fere

nce

stand

ard,

whe

re X

pert

MTB

/RI

F is

used

as

a re

plac

emen

t tes

t for

usu

al p

ract

ice?

A. E

xpec

tora

ted

and

indu

ced

sput

um, a

nd g

astr

ic la

vage

and

asp

irate

Parti

cipa

nts:

Cul

ture

neg

ativ

e ch

ildre

n ag

ed 0

-15

year

s su

spec

ted

of h

avin

g TB

. Set

ting:

Mai

nly

terti

ary

care

refe

rral h

ospi

tals,

uni

vers

ity h

ospi

tals.

Targ

et c

ondi

tion:

Pul

mon

ary

TB. R

efer

ence

test:

Clin

ical

TB.

Exp

ecto

rate

d sp

utum

and

indu

ced

sput

um c

ombi

ned.

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 8 (9

95)a .

Pool

ed s

ensit

ivity

: 4%

(95%

CrI,

1–1

2%);

pool

ed s

peci

ficity

100

% (9

5% C

rI, 9

9–10

0%).

Gas

tric

lava

ge a

nd a

spira

te.

Num

ber

of s

tudi

es (n

umbe

r of

par

ticip

ants)

: 3

(269

). Po

oled

sen

sitiv

ity:

15%

(95%

CrI,

5–3

1%);

pool

ed s

peci

ficity

: 99

%

(95%

CrI,

96–

100%

)

Out

com

eSt

udy

desi

gnFa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

eQ

ualit

y of

ev

iden

ceLi

mita

tions

Indi

rect

ness

In

cons

iste

ncy

Impr

ecis

ion

Publ

icat

ion

bias

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-sect

iona

l Se

rious

(–1)

bSe

rious

(–1)

cSe

rious

(–1)

dN

onee

Und

etec

tedf

Very

low

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed

as n

ot h

avin

g TB

)C

ross

-sect

iona

l Se

rious

(–1)

bSe

rious

(–1)

cSe

rious

(–1)

dN

onee

Und

etec

tedf

Very

low

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Cro

ss-se

ctio

nal

Serio

us (–

1)b

Serio

us (–

1)c

Serio

us (–

1)d

Non

eeU

ndet

ecte

dfLo

w

True

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Cro

ss-se

ctio

nal

Serio

us (–

1)b

Serio

us (–

1)c

Serio

us (–

1)d

Non

eeU

ndet

ecte

dfLo

w

CrI,

cre

dibl

e in

terv

al.

a Ra

chow

201

2 re

porte

d te

sting

one

coh

ort u

sing

expe

ctor

ated

spu

tum

and

one

usin

g in

duce

d sp

utum

; thi

s w

as c

ount

ed a

s tw

o stu

dies

.b

The

QU

ADAS

-2 to

ol w

as u

sed

to a

sses

s the

risk

of b

ias.

Six

of s

even

stud

ies i

nclud

ed in

the

met

a-an

alys

is en

rolle

d in

divid

uals

cons

ecut

ively

(seve

n stu

dies

use

d sp

ecim

ens o

f exp

ecto

rate

d or

indu

ced

sput

um; t

hree

use

d sp

ecim

ens f

rom

gas

tric

lava

ge o

r asp

iratio

n). R

esult

s fro

m X

pert

MTB

/RIF

are

auto

mat

ed a

nd w

ere

cons

ider

ed b

linde

d in

all s

tudi

es. C

linic

al T

B is

an im

perfe

ct re

fere

nce

stand

ard.

In th

is re

view

it w

as d

efine

d on

the

basis

of w

heth

er c

ultur

e-ne

gativ

e ch

ildre

n su

spec

ted

of h

avin

g TB

wer

e in

itiate

d on

ant

i-TB

treat

men

t. Da

ta a

re la

ckin

g on

how

wel

l thi

s cli

nica

l ref

eren

ce s

tand

ard

iden

tifies

true

cas

es, b

oth

in te

rms

of o

verd

iagn

osis

and

unde

rdia

gnos

is. T

he ri

sk o

f bia

s fo

r the

refe

renc

e sta

ndar

d w

as ra

ted

as

uncle

ar fo

r all

studi

es. T

here

wer

e ve

ry s

erio

us c

once

rns

abou

t the

lim

itatio

ns o

f the

stu

dies

. The

evid

ence

was

dow

ngra

ded

by o

ne p

oint

.c

The

ratin

g of

the

qual

ity o

f evid

ence

may

be

low

ered

if th

ere

are

impo

rtant

diff

eren

ces

in th

e te

sts s

tudi

ed a

nd th

e ex

perti

se o

f tho

se u

sing

the

tests

in th

e stu

dies

com

pare

d w

ith th

e se

ttings

for w

hich

the

reco

mm

enda

tions

are

inte

nded

. All

studi

es w

ere

perfo

rmed

at h

ighe

r lev

els

of c

are

– th

at is

, hig

her-le

vel r

efer

ral h

ospi

tals

and

unive

rsity

hosp

itals.

Five

of t

he s

even

stu

dies

(71%

) stu

dies

inclu

ded

only

inpa

tient

s; tw

o stu

dies

inclu

ded

inpa

tient

s an

d ou

tpat

ient

s. In

the

paed

iatri

c po

pula

tion,

inpa

tient

s m

ay p

rese

nt w

ith m

ore

seve

re d

iseas

e an

d ha

ve a

hig

her

likel

ihoo

d of

bei

ng s

mea

r po

sitive

. In

one

stu

dy (R

acho

w 2

012)

, ap

prox

imat

ely

20%

of c

hild

ren

susp

ecte

d of

hav

ing

TB w

ere

iden

tified

thro

ugh

cont

act

traci

ng;

this

study

is m

ost r

epre

sent

ative

of t

he p

aedi

atric

pop

ulatio

n th

at m

ight

ben

efit f

rom

Xpe

rt M

TB/R

IF Tw

o un

publ

ished

stu

dies

(Chi

sti a

nd L

aCou

rse) i

nclud

ed c

hild

ren

who

wer

e se

vere

ly m

alno

urish

ed,

and

TB w

as o

ne d

iffer

entia

l di

agno

sis.

Addi

tiona

lly,

child

ren

in th

e C

histi

stu

dy h

ad to

hav

e ha

d co

ugh

lasti

ng lo

nger

than

2 w

eeks

or

pneu

mon

ia o

n X-

ray,

or

both

. Th

ere

wer

e se

rious

con

cern

s ab

out i

ndire

ctne

ss.

The

evid

ence

was

dow

ngra

ded

by o

ne p

oint

. d

The

poin

t es

timat

es

for

sens

itivity

ra

nged

fro

m

0%

to

35%

fo

r te

sting

sp

ecim

ens

of

expe

ctor

ated

or

in

duce

d sp

utum

; fo

r te

sting

sp

ecim

ens

from

ga

stric

la

vage

or

as

pira

tion,

th

e po

int

estim

ates

ra

nged

fro

m

0%

to

20%

; th

ese

resu

lts

indi

cate

m

oder

ate

hete

roge

neity

. H

owev

er,

the

95%

co

nfide

nce

inte

rval

s ov

erla

pped

. N

o su

bgro

up

anal

ysis

was

pe

rform

ed

for

Xper

t M

TB/R

IF co

mpa

red

agai

nst

a cli

nica

l re

fere

nce

stand

ard.

H

eter

ogen

eity

may

pa

rtly

be

the

resu

lt of

di

ffere

nces

in

pa

tient

po

pula

tions

(fo

r ex

ampl

e,

caus

ed

by

the

use

of

diffe

rent

in

clusio

n cr

iteria

). In

cons

isten

cy

was

ra

ted

as

a se

rious

co

ncer

n fo

r th

e es

timat

es

of

sens

itivity

.

The

evid

ence

w

as

dow

ngra

ded

by

one

poin

t. Th

e es

timat

es

of

spec

ifici

ty di

d no

t sh

ow

hete

roge

neity

: po

int

estim

ates

ra

nged

fro

m

99%

to

10

0%.

Ther

e w

as

no

conc

ern

abou

t sp

ecifi

city.

e

The

maj

ority

of

coho

rts i

nclud

ed i

n th

e re

view

had

a s

ampl

e siz

e gr

eate

r th

an 6

0 (5

/8 s

tudi

es u

sing

expe

ctor

ated

or

indu

ced

sput

um,

and

2/3

usin

g ga

stric

lava

ge o

r as

pira

te )

; th

e yi

eld

of c

ultur

e am

ong

all c

ohor

ts va

ried

from

2.4

% t

o 54

.2%

. Th

e co

nfide

nce

inte

rval

for

sen

sitivi

ty fo

r ex

pect

orat

ed o

r in

duce

d sp

utum

was

with

in t

he +

/-10

% m

argi

n; h

owev

er,

for

gastr

ic la

vage

and

asp

irate

it

was

wid

e. T

here

wer

e no

con

cern

s fo

r ex

pect

orat

ed o

r in

duce

d sp

utum

; th

ere

wer

e se

rious

con

cern

s ab

out

gastr

ic l

avag

e an

d as

pira

te.

The

evid

ence

was

not

dow

ngra

ded

furth

er g

iven

the

dow

ngra

ding

for

inc

onsis

tenc

y. T

here

was

no

conc

ern

abou

t th

e ev

iden

ce f

or s

peci

ficity

. f T

wo

unpu

blish

ed s

tudi

es w

ere

inclu

ded.

A fe

w o

ther

stu

dies

wer

e in

pro

gres

s bu

t dat

a w

ere

not a

vaila

ble

for a

nalys

is. T

he d

ata

that

wer

e in

clude

d do

not

allo

w fo

r for

mal

ass

essm

ent o

f pub

licat

ion

bias

usin

g m

etho

ds s

uch

as fu

nnel

plo

ts or

regr

essio

n te

sts; s

uch

tech

niqu

es h

ave

not b

een

foun

d to

be

help

ful i

n stu

dies

eva

luatin

g di

agno

stic

accu

racy

. How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to b

e m

inim

al s

ince

Xpe

rt M

TB/R

IF is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.

66 XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATEB.

Sum

mar

y of

find

ings

Out

com

e

Xper

t MTB

/RIF

usi

ng s

peci

men

s of

exp

ecto

rate

or

indu

ced

sput

umXp

ert M

TB/R

IF u

sing

spe

cim

ens

of g

astr

ic la

vage

or

asp

irate

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

aQ

ualit

y

of

evid

ence

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

aQ

ualit

y

of

evid

ence

Prev

alen

ce

10/1

000b

Prev

alen

ce

50/1

000b

Prev

alen

ce

100/

1000

bPr

eval

ence

10

/100

0bPr

eval

ence

50

/100

0bPr

eval

ence

10

0/10

00b

True

pos

itive

s (in

divi

dual

s w

ith T

B)0

(0-1

)2

(1-6

)4

(1-1

2)Lo

w

4(1

-8)

8(3

-16)

15(5

-31)

Very

low

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cl

assifi

ed a

s no

t hav

ing

TB)

10(9

-10)

48(4

4-50

)96

(88-

99)

Low

21

(17-

24)

43(3

5-48

)85

(69-

95)

Very

low

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cl

assifi

ed a

s ha

ving

TB)

0(1

-10)

0(0

-10)

0(1

-9)

Mod

erat

e

10(0

-39)

10(0

-38)

9(0

-9)

Mod

erat

e

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cl

assifi

ed a

s no

t hav

ing

TB)

0(0

-1)

2(1

-6)

4(1

-12)

Mod

erat

e

965

(936

-975

)94

1(9

12-9

50)

891

(864

-900

)M

oder

ate

CrI,

cre

dibl

e in

terv

al.

a The

expe

cted

num

ber o

f Xpe

rt M

TB/

RIF

resu

lts w

as b

ased

on

the

sens

itivi

ty a

nd s

peci

ficity

esti

mat

es c

alcu

late

d fro

m a

com

paris

on w

ith c

ultu

re fo

r diff

eren

t pre

vale

nces

of T

B.

b The

estim

ates

of T

B pr

eval

ence

wer

e pr

ovid

ed b

y th

e W

HO

Ste

erin

g G

roup

.


ble

26. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

TB in

chi

ldre

n fo

llow

ing

nega

tive

smea

r mic

rosc

opy

(A. E

vide

nce

profi

le, B

. Su

mm

ary

of fi

ndin

gs, C

. Add

ition

al y

ield

of X

pert

MTB

/RIF

ove

r m

icro

scop

y)

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r det

ectio

n of

TB

in c

hild

ren,

whe

re X

pert

MTB

/RI

F is

used

as

an a

dd-o

n te

st fo

llow

ing

a ne

gativ

e sm

ear-m

icro

scop

y re

sult?

A. E

vide

nce

profi

lePa

rtici

pant

s: S

mea

r neg

ativ

e ch

ildre

n ag

ed 0

-15

year

s su

spec

ted

of h

avin

g TB

. Set

ting:

Mai

nly

terti

ary

care

refe

rral h

ospi

tals

and

univ

ersit

y ho

spita

ls.Ta

rget

con

ditio

n: P

ulm

onar

y TB

. Ref

eren

ce te

st: S

olid

cul

ture

or l

iqui

d cu

lture

. Exp

ecto

rate

d sp

utum

and

indu

ced

sput

um c

ombi

ned.

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 7 (1

008)

. Poo

led

sens

itivi

ty 4

4% (9

5% C

rI, 4

1–69

%);

pool

ed s

peci

ficity

: 98%

(95%

CrI,

96–

99%

).G

astri

c la

vage

or a

spira

te.

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 6

(120

4).

Pool

ed s

ensit

ivity

: 62

% (9

5% C

rI, 4

4–80

%);

pool

ed s

peci

ficity

: 99

%

(95%

CrI,

97–

99%

)

Out

com

eSt

udy

desi

gnFa

ctor

s th

at m

ay d

ecre

ase

the

qual

ity o

f evi

denc

eQ

ualit

y of

ev

iden

ceLi

mita

tions

Indi

rect

ness

In

cons

iste

ncy

Impr

ecis

ion

Publ

icat

ion

bias

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-sect

iona

l N

onea

Serio

us (–

1)b

Serio

us (–

1)c

Expe

ctor

ated

or i

nduc

ed

sput

um: n

oned

Gas

tric

lava

ge o

r as

pira

te: s

erio

us (–

1)d

Und

etec

tede

Expe

ctor

ated

or i

nduc

ed

sput

um:

Low

G

astri

c la

vage

or

aspi

rate

: Ve

ry lo

w

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed a

s no

t hav

ing

TB)

Cro

ss-se

ctio

nal

Non

eaSe

rious

(–1)

bSe

rious

(–1)

cEx

pect

orat

ed o

r ind

uced

sp

utum

: non

ed

Gas

tric

lava

ge o

r as

pira

te: s

erio

us (–

1)d

Und

etec

tede

Expe

ctor

ated

or i

nduc

ed

sput

um:

Low

G

astri

c la

vage

or

aspi

rate

: Ve

ry lo

w

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Cro

ss-se

ctio

nal

Non

eaSe

rious

(–1)

bN

onec

Non

edU

ndet

ecte

deM

oder

ate

True

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB

)C

ross

-sect

iona

l N

onea

Serio

us (–

1)b

Non

ecN

oned

Und

etec

tede

Mod

erat

e

CrI,

cre

dibl

e in

terv

al.

a Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess t

he ri

sk o

f bia

s for

all 1

1 stu

dies

that

wer

e in

clud

ed in

this

estim

ate

(7 c

ohor

ts co

ntrib

utin

g sa

mpl

es o

f exp

ecto

rate

d or

indu

ced

sput

um; 4

coh

orts

with

spec

imen

s fro

m g

astri

c la

vage

or a

spira

tion)

. The

maj

ority

of s

tudi

es e

nrol

led

indi

vidu

als

cons

ecut

ivel

y. R

esul

ts fro

m X

pert

MTB

/RI

F re

sult

are

auto

mat

ed a

nd w

ere

cons

ider

ed b

linde

d in

all

studi

es. T

he e

xclu

sion

of e

nrol

led

patie

nts f

rom

ana

lyse

s wer

e w

ell e

xpla

ined

in m

ost s

tudi

es; h

owev

er in

one

stud

y (B

ates

201

3) th

at u

sed

spec

imen

s fro

m g

astri

c la

vage

or a

spira

tion,

a n

umbe

r of c

hild

ren

who

had

alre

ady

initi

ated

ant

i-TB

treat

men

t wer

e en

rolle

d (7

7/10

37 s

ampl

es o

f exp

ecto

rate

d or

indu

ced

sput

um, a

nd s

ampl

es fr

om g

astri

c la

vage

or a

spira

tion)

; 66

of th

ese

child

ren

wer

e cu

lture

neg

ativ

e an

d w

ere

excl

uded

from

the

anal

ysis;

11 w

ere

cultu

re p

ositi

ve a

nd w

ere

incl

uded

. Add

ition

ally,

the

follo

win

g co

nsid

erat

ions

wer

e no

t tak

en in

to a

ccou

nt in

judg

emen

ts ab

out t

he li

mita

tions

of t

he s

tudi

es b

ut th

ey

are

impo

rtant

: cul

ture

is a

n im

perfe

ct re

fere

nce

stand

ard.

In c

hild

ren,

cul

ture

sen

sitiv

ity v

arie

s fro

m 2

0% to

70%

, dep

endi

ng o

n fa

ctor

s su

ch a

s th

e se

verit

y of

dise

ase,

the

age

of th

e ch

ild, t

he c

ultu

re

68 XPERT MTB/RIF FOR THE DIAGNOSIS OF PULMONARY AND EXTRAPULMONARY TB – POLICY UPDATEm

etho

ds u

sed,

and

spec

imen

type

s. In

ord

er to

impr

ove

the

diag

nosti

c yi

eld,

2/

11 st

udie

s (bo

th u

sing

spec

imen

s fro

m g

astri

c la

vage

or a

spira

tion

perfo

rmed

at le

ast t

wo

cultu

res f

or e

ach

child

, but

som

e pe

rform

ed a

s m

any

as s

ix c

ultu

res

per c

hild

. The

sen

sitiv

ity o

f Xpe

rt M

TB/

RIF

may

be

unde

resti

mat

ed w

hen

it is

com

pare

d ag

ains

t cul

ture

as

a re

fere

nce

stand

ard.

The

evi

denc

e w

as n

ot d

owng

rade

d.

b A

ll stu

dies

wer

e pe

rform

ed a

t hig

her l

evel

s of c

are

– th

at is

, at h

ighe

r-lev

el re

ferra

l hos

pita

ls an

d un

iver

sity

hosp

itals.

Eig

ht st

udie

s inc

lude

d on

ly in

patie

nts;

two

studi

es in

clud

ed in

patie

nts a

nd o

utpa

tient

s (tw

o stu

dies

teste

d sp

ecim

ens o

f exp

ecto

rate

d or

indu

ced

sput

um; o

ne st

udy

teste

d sp

ecim

ens f

rom

gas

tric

lava

ge o

r asp

iratio

n); t

wo

studi

es w

ere

labo

rato

ry-b

ased

and

pro

vide

d lit

tle c

linic

al in

form

atio

n (th

ese

teste

d sp

ecim

ens f

rom

gas

tric

lava

ge o

r asp

iratio

n). I

n th

e pa

edia

tric

popu

latio

n, in

patie

nts m

ay p

rese

nt w

ith m

ore

seve

re d

iseas

e an

d ha

ve a

hig

her l

ikel

ihoo

d of

bei

ng sm

ear p

ositi

ve; t

hey

are

one

grou

p th

at m

ight

ben

efit f

rom

Xpe

rt M

TB/

RIF.

Out

patie

nts m

ay p

rese

nt d

iffer

ently

, with

less

seve

re d

iseas

e an

d so

be

less

likel

y to

be

smea

r pos

itive

or c

ultu

re p

ositi

ve. O

utpa

tient

s are

not

repr

esen

ted

by th

e stu

dies

incl

uded

in th

is re

view

. Onl

y on

e stu

dy (R

acho

w 2

012,

usin

g sp

ecim

ens f

rom

exp

ecto

rate

d sp

utum

) rep

rese

nted

a b

road

er g

roup

of c

hild

ren:

app

roxi

mat

ely

20%

of c

hild

ren

susp

ecte

d of

hav

ing

TB h

ad b

een

iden

tified

thro

ugh

cont

act t

raci

ng. O

ne u

npub

lishe

d stu

dy (C

histi

) inc

lude

d se

vere

ly m

alno

urish

ed c

hild

ren

with

cou

gh la

sting

long

er th

an 2

wee

ks o

r pne

umon

ia o

n X-

ray,

or b

oth;

in th

ese

case

s, T

B w

as o

ne d

iffer

entia

l dia

gnos

is. T

here

wer

e se

rious

con

cern

s abo

ut in

dire

ctne

ss in

the

maj

ority

of s

tudi

es p

erfo

rmed

in h

ighe

r-lev

el re

ferra

l hos

pita

ls. T

he e

vide

nce

was

dow

ngra

ded

by o

ne p

oint

.

c No

study

dire

ctly

add

ress

ed th

e que

stion

of in

cons

isten

cy b

y per

form

ing

prio

r testi

ng w

ith m

icro

scop

y and

then

subs

eque

ntly

per

form

ing

Xper

t MTB

/RI

F. Es

timat

es o

f sen

sitiv

ity w

ere v

aria

ble a

cros

s stu

dies

and

sp

ecim

en ty

pes,

and

rang

ed fr

om 3

6% to

86%

for e

xpec

tora

ted

sput

um, f

rom

43%

to 7

0% fo

r ind

uced

sput

um, a

nd fr

om 4

0% to

100

% fo

r spe

cim

ens f

rom

gas

tric l

avag

e or

asp

iratio

n. Th

e he

tero

gene

ity co

uld

not b

e exp

lain

ed; th

e evi

denc

e was

ther

efor

e dow

ngra

ded b

y one

poin

t. Esti

mat

es of

spec

ifici

ty di

d not

show

serio

us he

tero

gene

ity: p

oint

estim

ates

rang

ed fro

m 9

3% to

100

%. T

here

was

no co

ncer

n abo

ut sp

ecifi

city.

d

The

pool

ed e

stim

ate

of s

ensit

ivity

for

spe

cim

ens

from

exp

ecto

rate

d or

ind

uced

spu

tum

had

a w

ide

95%

con

fiden

ce i

nter

val

(gre

ater

tha

n +/

-10%

of

the

poin

t es

timat

e);

the

estim

ate

for

gastr

ic l

avag

e or

asp

iratio

n ha

d a

wid

e 95

% c

onfid

ence

int

erva

l +/

-20%

of

the

poin

t es

timat

e).

The

conc

erns

abo

ut e

xpec

tora

ted

or i

nduc

ed s

putu

m w

ere

serio

us;

the

conc

erns

abo

ut

sam

ples

fro

m g

astri

c la

vage

or

aspi

ratio

n w

ere

very

ser

ious

. Th

e ev

iden

ce w

as n

ot d

owng

rade

d fu

rther

giv

en t

he d

owng

radi

ng f

or i

ncon

siste

ncy.

The

re w

as n

o co

ncer

n ab

out

spec

ifici

ty.

e Tw

o un

publ

ished

stu

dies

wer

e in

clud

ed.

The

data

that

wer

e in

clud

ed in

the

revi

ew d

id n

ot a

llow

for f

orm

al a

sses

smen

t of p

ublic

atio

n bi

as u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

; su

ch te

chni

ques

hav

e no

t bee

n fo

und

to b

e he

lpfu

l in

studi

es e

valu

atin

g di

agno

stic

accu

racy

. How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to b

e m

inim

al s

ince

Xpe

rt M

TB/

RIF

is a

new

test

that

has

had

co

nsid

erab

le a

ttent

ion

and

scru

tiny.

B. S

umm

ary

of fi

ndin

gs

Out

com

e

Xper

t MTB

/RIF

usi

ng s

peci

men

s of

exp

ecto

rate

or

indu

ced

sput

umXp

ert M

TB/R

IF u

sing

spe

cim

ens

of g

astr

ic la

vage

or

asp

irate

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

aQ

ualit

y

of

evid

ence

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

aQ

ualit

y

of

evid

ence

Prev

alen

ce

10/1

000b

Prev

alen

ce

50/1

000b

Prev

alen

ce

100/

1000

bPr

eval

ence

10

/100

0bPr

eval

ence

50

/100

0bPr

eval

ence

10

0/10

00b

True

pos

itive

s (in

divi

dual

s w

ith T

B)6

(4-7

)28

(21-

35)

55(4

1-69

)Lo

w

6(4

-8)

31(2

2-40

)62

(44-

80)

Very

low

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cl

assifi

ed a

s no

t hav

ing

TB)

5(3

-6)

23(1

6-30

)45

(31-

59)

Low

4

(2-6

)19

(10-

28)

38(2

0-56

)Ve

ry lo

w

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cl

assifi

ed a

s ha

ving

TB)

20(1

0-40

)19

(10-

38)

18(9

-36)

Mod

erat

e

10(1

0-30

)10

(10-

29)

9(9

-27)

Mod

erat

e

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cl

assifi

ed a

s no

t hav

ing

TB)

970

(950

-980

)93

1(9

12-9

41)

882

(864

-891

)M

oder

ate

98

0(9

60-9

41)

941

(922

-941

)89

1(8

73-8

91)

Mod

erat

e

CrI,

cre

dibl

e in

terv

al.

a The

expe

cted

num

ber o

f Xpe

rt M

TB/

RIF

resu

lts w

as b

ased

on

the

sens

itivi

ty a

nd s

peci

ficity

esti

mat

es c

alcu

late

d fro

m a

com

paris

on w

ith c

ultu

re fo

r diff

eren

t pre

vale

nces

of T

B.

b The

estim

ates

of T

B pr

eval

ence

wer

e pr

ovid

ed b

y th

e W

HO

Ste

erin

g G

roup

.


C. What is the additional yield of Xpert MTB/RIF compared with microscopy in children with smear-negative culture-positive TB?

Outcome

Number of results/1000 smear-negative culture-positive children tested (95% CrI) Quality

of evidence


Smear microscopy

Xpert MTB/RIF

Smear microscopy

Xpert MTB/RIF

Smear microscopy

Xpert MTB/RIF

Specimens from expectorated or induced sputumTrue positives (patients with TB)

0 6(4–7)

0 28(21–35)

0 55(41–69)

Low



False negatives (patients incorrectly classified as not having TB)

10 5(3–6)

50 23(16–30)

100 45(31–59)

Low



Specimens from gastric lavage or aspirationTrue positives (patients with TB)

0 6(4–8)

0 31(22–40)

0 62(44–80)

Low




10 4(2–6)

50 19(10–28)

100 38(20–56)

Low



CrI, credible interval.


Table 27. Incremental yield of Xpert MTB/RIF compared with smear microscopy in children with culture-confirmed TB (A. Expectorated sputum and induced sputum, B. Gastric lavage or aspirate)PICO question: What is the incremental yield of Xpert MTB/RIF compared with smear microscopy in children with culture-confirmed TB?A. Expectorated sputum and induced sputumParticipants: Children aged 0-15 years with culture-confirmed TBSetting: Mainly tertiary care referral hospitals and university hospitalsTarget condition: Pulmonary TB. Reference standard: Solid culture or liquid culture.Number of studies (number of participants): 10 (1546)aMicroscopyPooled sensitivity: 29% (95% CrI, 16–42%); pooled specificity: 100% (95% CrI, 99–100%)Xpert MTB/RIFPooled sensitivity: 66% (95% CrI, 52–77%); pooled specificity: 98% (95% CrI, 96–99%)

Outcome

Number of results/1000 culture-positive children tested (95% CrI)aQuality

of evidence


Smear microscopy

Xpert MTB/RIF

Smear microscopy

Xpert MTB/RIF

Smear microscopy

Xpert MTB/RIF

True positives (patients with TB)

3 (2–4)

7(5–8)

15 (8–21)

33(26–39)

29 (16–42)

66(52–77)

Low




7 (6–8)

3(2–5)

36 (29–42)

17 (12–24)

71 (58–84)

34 (23–48)

Low



CrI, credible interval.a These are the same studies as those assessed in Table 24 A. Rachow 2012 reported testing one cohort using expectorated sputum and one using induced sputum; this was counted as two studies.

B. Gastric lavage or aspirateNumber of studies (number of participants): 7 (1319)Microscopy Pooled sensitivity: 11% (95% CrI, 12–35%); pooled specificity: 99% (95% CrI, 97–100%)Xpert MTB/RIFPooled sensitivity: 66% (95% CrI, 51–81%); pooled specificity: 98% (95% CrI, 96–99%)

Outcome

Number of results/1000 culture-positive children tested (95% CrI)aQuality

of evidence


Smear microscopy

Xpert MTB/RIF

Smear microscopy

Xpert MTB/RIF

Smear microscopy

Xpert MTB/RIF

True positives (patients with TB)

2 (1–4)

7(5–8)

11 (6–18)

33(26–41)

22 (12–35)

66(51–81)

Low




8 (7–9)

3(2–5)

39 (33–44)

17 (10–25)

78 (65–88)

34 (19–49)

Low



CrI, credible interval.


ble

28. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

rifam

pici

n re

sista

nce

in r

espi

rato

ry s

peci

men

s fro

m c

hild

ren

PIC

O q

uesti

on: W

hat i

s th

e di

agno

stic

accu

racy

of X

pert

MTB

/RI

F fo

r det

ectio

n of

rifa

mpi

cin

resis

tanc

e in

resp

irato

ry s

peci

men

s fro

m c

hild

ren?

Parti

cipa

nts:

Chi

ldre

n ag

ed 0

–15

year

s su

spec

ted

of h

avin

g TB

or m

ultid

rug-

resis

tant

TB

Setti

ng: M

ainl

y te

rtiar

y ca

re re

ferra

l hos

pita

ls an

d un

iver

sity

hosp

itals

Targ

et c

ondi

tion:

Rifa

mpi

cin

resis

tanc

eRe

fere

nce

test:

Cul

ture

and

cul

ture

-bas

ed p

heno

typi

c dr

ug-su

scep

tibili

ty te

sting

(1 s

tudy

) or H

ain

Lifes

cien

ce’s

Gen

otyp

e M

TBD

Rplu

s (2

stu

dies

)N

umbe

r of s

tudi

es (n

umbe

r of p

artic

ipan

ts): 3

(176

) Po

oled

sen

sitiv

ity: 8

6% (9

5% C

rI, 5

3–98

%);

pool

ed s

peci

ficity

: 98%

(95%

CrI,

94–

100%

)

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

50

/100

0Pr

eval

ence

10

0/10

00

True

pos

itive

s (in

divi

dual

s w

ith T

B an

d rif

ampi

cin

resis

tanc

e)C

ross

-se

ctio

nal

Non

eaN

oneb

Non

ecVe

ry s

erio

us

(–2)

dU

ndet

ecte

deLo

w

43(2

7-49

)12

9(8

0-14

7)Fa

lse n

egat

ives

(ind

ivid

uals

inco

rrect

ly c

lass

ified

as

hav

ing

TB th

at is

rifa

mpi

cin

susc

eptib

le)

Cro

ss-

sect

iona

l N

onea

Non

ebN

onec

Very

ser

ious

(–

2)d

Und

etec

tede

Low

7

(1-2

4)21

(3-7

1)Fa

lse p

ositi

ves

(indi

vidu

als

inco

rrect

ly c

lass

ified

as

hav

ing

TB w

ith ri

fam

pici

n re

sista

nce)

Cro

ss-

sect

iona

l N

onea

Non

ebN

onec

Non

eU

ndet

ecte

deH

igh

19

(0

-57)

17

(0-5

1)Tr

ue n

egat

ives

(ind

ivid

uals

with

TB

that

is

rifam

pici

n su

scep

tible

)C

ross

-se

ctio

nal

Non

eaN

oneb

Non

ecN

one

Und

etec

tede

Hig

h

931

(893

-950

)83

3 (7

99-8

50)

CrI,

cre

dibl

e in

terv

al; D

ST, d

rug-

susc

eptib

ility

testi

ng.

a Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

All

thre

e stu

dies

enr

olle

d in

divi

dual

s co

nsec

utiv

ely.

The

resu

lt fro

m X

pert

MTB

/RI

F is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

. Th

ere

wer

e no

con

cern

s ab

out r

isk o

f bia

s.b

The

ratin

g of

the

qual

ity o

f evi

denc

e m

ay b

e lo

wer

ed if

ther

e ar

e im

porta

nt d

iffer

ence

s in

the

popu

latio

ns a

nd te

sts s

tudi

ed, a

nd in

the

expe

rtise

of t

hose

usin

g th

e te

sts in

the

studi

es c

ompa

red

with

th

e se

tting

s fo

r whi

ch th

e re

com

men

datio

ns a

re in

tend

ed. A

ll stu

dies

wer

e pe

rform

ed a

t ter

tiary

car

e re

ferra

l hos

pita

ls or

uni

vers

ity h

ospi

tals,

and

all

child

ren

incl

uded

in th

e stu

dies

wer

e in

patie

nts.

c

For

the

3 stu

dies

, th

e po

int

estim

ates

fo

r se

nsiti

vity

w

ere

67%

, 83

%

and

100%

, in

dica

ting

som

e he

tero

gene

ity.

The

95%

co

nfide

nce

inte

rval

s w

ere

wid

e an

d ov

erla

ppin

g.

The

estim

ates

of

sp

ecifi

city

di

d no

t sh

ow

serio

us

hete

roge

neity

: th

e po

int

estim

ates

ra

nged

fro

m

94%

to

10

0%.

Ther

e w

ere

no

conc

erns

ab

out

spec

ifici

ty.

d O

nly

3 stu

dies

with

a to

tal o

f 176

chi

ldre

n w

ere

incl

uded

in th

e re

view

; 121

chi

ldre

n w

ere

enro

lled

in o

ne s

tudy

. The

con

fiden

ce in

terv

al fo

r sen

sitiv

ity w

as w

ide

(cro

ssin

g +/

– 20

%).

Ther

e w

ere

very

ser

ious

con

cern

s ab

out t

his

evid

ence

. The

evi

denc

e w

as d

owng

rade

d by

two

poin

ts. T

here

wer

e no

con

cern

s ab

out s

peci

ficity

.e

One

unp

ublis

hed

study

was

incl

uded

in th

e re

view

. A

few

add

ition

al s

tudi

es w

ere

in p

rogr

ess

but d

ata

wer

e no

t ava

ilabl

e fo

r ana

lysis

. Th

e da

ta in

clud

ed in

the

revi

ew d

id n

ot a

llow

for f

orm

al

asse

ssm

ent o

f pub

licat

ion

bias

usin

g m

etho

ds s

uch

as fu

nnel

plo

ts or

regr

essio

n te

sts; s

uch

tech

niqu

es h

ave

not b

een

foun

d to

be

help

ful i

n stu

dies

eva

luat

ing

diag

nosti

c ac

cura

cy. H

owev

er, r

epor

ting

bias

was

con

sider

ed to

be

min

imal

sin

ce X

pert

MTB

/RI

F is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.


ble

29. G

RAD

E ev

iden

ce p

rofil

e: a

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

perip

hera

l lym

ph n

ode

TB in

chi

ldre

n

PIC

O q

uesti

on:

Wha

t is

the

diag

nosti

c ac

cura

cy o

f Xpe

rt M

TB/

RIF

for d

etec

tion

of p

erip

hera

l lym

ph n

ode

TB in

chi

ldre

n, w

here

Xpe

rt M

TB/

RIF

is us

ed

as a

repl

acem

ent t

est f

or u

sual

pra

ctic

e?Pa

rtici

pant

s: C

hild

ren

aged

0–1

5 ye

ars

susp

ecte

d of

hav

ing

perip

hera

l lym

ph n

ode

TBSe

tting

: Mai

nly

terti

ary

care

refe

rral h

ospi

tals

and

univ

ersit

y ho

spita

lsTa

rget

con

ditio

n: P

erip

hera

l lym

ph n

ode

TBRe

fere

nce

test:

Sol

id c

ultu

re o

r liq

uid

cultu

re o

f sam

ples

from

fine

nee

dle

aspi

ratio

n or

bio

psy

Num

ber o

f stu

dies

(num

ber o

f par

ticip

ants)

: 3 (1

72)

Pool

ed s

ensit

ivity

: 86%

(95%

CrI,

65–

96%

); po

oled

spe

cific

ity: 8

1% (9

5% C

rI, 5

4–93

%)

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

10

/100

0Pr

eval

ence

50

/100

0Pr

eval

ence

10

0/10

00

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-se

ctio

nal

Non

eaN

oneb

Serio

us (–

1)c

Serio

us (–

1)d

Und

etec

tede

Very

low

22

(16-

24)

43(3

3-48

)86

(65-

96)

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed a

s no

t ha

ving

TB)

Cro

ss-

sect

iona

lN

onea

Non

ebSe

rious

(–1)

cSe

rious

(–1)

dU

ndet

ecte

deVe

ry lo

w

4(1

-9)

7(2

-18)

14(4

-35)

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Cro

ss-

sect

iona

lN

onea

Non

ebSe

rious

(–1)

cSe

rious

(–1)

dU

ndet

ecte

deVe

ry lo

w

185

(68-

449)

181

(67-

437)

171

(63-

414)

True

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Cro

ss-

sect

iona

lN

onea

Non

ebSe

rious

(–1)

cSe

rious

(–1)

dU

ndet

ecte

deVe

ry lo

w

790

(527

-907

)77

0(5

13-8

84)

729

(486

-837

)

CrI,

cre

dibl

e in

terv

al.

a Th

e Q

UA

DA

S-2

tool

was

use

d to

ass

ess

the

risk

of b

ias.

Tw

o of

thre

e stu

dies

enr

olle

d ch

ildre

n co

nsec

utiv

ely;

one

was

a la

bora

tory

-bas

ed s

tudy

and

pro

vide

d no

clin

ical

info

rmat

ion.

The

resu

lt fro

m

Xper

t MTB

/RI

F is

auto

mat

ed a

nd w

as c

onsid

ered

blin

ded

in a

ll stu

dies

. Cul

ture

is a

n im

perfe

ct re

fere

nce

stand

ard.

The

re w

ere

no c

once

rns

abou

t lim

itatio

ns.

b Th

e ra

ting

of th

e qu

ality

of e

vide

nce

may

be

low

ered

if th

ere

are

impo

rtant

diff

eren

ces

in th

e po

pula

tions

and

tests

stu

died

, and

in th

e ex

perti

se o

f tho

se u

sing

the

tests

in th

e stu

dies

com

pare

d w

ith

the

setti

ngs

for w

hich

the

reco

mm

enda

tions

are

inte

nded

. All

studi

es w

ere

perfo

rmed

at h

ighe

r-lev

el c

are

faci

litie

s. T

he p

atie

nt p

opul

atio

n th

at re

ceiv

ed th

e in

dex

test

prob

ably

refle

cts

the

popu

latio

n of

inte

nded

use

. The

re w

ere

no c

once

rns

abou

t ind

irect

ness

.c

For t

he 3

stu

dies

, the

poi

nt e

stim

ates

for s

ensit

ivity

wer

e 77

%, 1

00%

and

100

%, i

ndic

atin

g so

me

hete

roge

neity

. The

95%

con

fiden

ce in

terv

als

over

lapp

ed. T

he e

stim

ates

for s

peci

ficity

wer

e 50

%,

71%

and

96%

. The

re is

no

expl

anat

ion

for t

he h

eter

ogen

eity.

The

re w

ere

serio

us c

once

rns

abou

t thi

s ev

iden

ce. T

he e

vide

nce

was

dow

ngra

ded

by o

ne p

oint

.d

Onl

y 3

studi

es w

ith a

tota

l of 1

72 c

hild

ren

wer

e in

clud

ed; 1

stu

dy h

ad o

nly

5 pa

rtici

pant

s. T

he c

onfid

ence

inte

rval

s fo

r bot

h se

nsiti

vity

and

spe

cific

ity w

ere

wid

e (c

ross

ing

+/–

20%

). Th

ere

wer

e ve

ry s

erio

us c

once

rns

abou

t thi

s ev

iden

ce. H

owev

er, t

he e

vide

nce

was

not

dow

ngra

ded

furth

er g

iven

the

dow

ngra

ding

for i

ncon

siste

ncy.

e O

ne u

npub

lishe

d stu

dy w

as in

clud

ed. A

few

oth

er st

udie

s wer

e in

pro

gres

s but

dat

a w

ere

not a

vaila

ble

for a

naly

sis. T

he d

ata

incl

uded

in th

e re

view

did

not

allo

w fo

r for

mal

ass

essm

ent o

f pub

licat

ion

bias

usin

g m

etho

ds s

uch

as fu

nnel

plo

ts or

regr

essio

n te

sts; s

uch

tech

niqu

es h

ave

not b

een

foun

d to

be

help

ful i

n stu

dies

eva

luat

ing

diag

nosti

c ac

cura

cy. H

owev

er, r

epor

ting

bias

was

con

sider

ed to

be

min

imal

sin

ce X

pert

MTB

/RI

F is

a ne

w te

st th

at h

as h

ad c

onsid

erab

le a

ttent

ion

and

scru

tiny.


ble

30. G

RAD

E ev

iden

ce p

rofil

e an

d su

mm

ary

of fi

ndin

gs: a

ccur

acy

of X

pert

MTB

/RIF

in d

etec

ting

TB m

enin

gitis

in c

hild

ren

PIC

O q

uesti

on: W

hat i

s the

dia

gnos

tic a

ccur

acy

of X

pert

MTB

/RI

F fo

r det

ectio

n of

TB

men

ingi

tis in

chi

ldre

n, w

here

Xpe

rt M

TB/

RIF

is us

ed a

s a re

plac

emen

t te

st fo

r usu

al p

ract

ice?

Parti

cipa

nts:

Chi

ldre

n ag

ed 0

–15

year

s su

spec

ted

of h

avin

g TB

men

ingi

tisSe

tting

: Mai

nly

terti

ary

care

refe

rral h

ospi

tals

and

univ

ersit

y ho

spita

lsTa

rget

con

ditio

n: T

B m

enin

gitis

Refe

renc

e te

st: S

olid

cul

ture

or l

iqui

d cu

lture

of c

ereb

rosp

inal

flui

dN

umbe

r of s

tudi

es (n

umbe

r of p

artic

ipan

ts): 3

(51)

Pool

ed s

ensit

ivity

: not

eno

ugh

data

; poo

led

spec

ifici

ty: 9

5% (9

5% C

rI, 8

1–99

%)

Out

com

eSt

udy

desi

gn

Fact

ors

that

may

dec

reas

e th

e qu

ality

of e

vide

nce

Qua

lity

of

ev

iden

ce

Num

ber

of r

esul

ts/1

000

indi

vidu

als

test

ed

(95%

CrI)

Lim

itatio

nsIn

dire

ctne

ss

Inco

nsis

tenc

yIm

prec

isio

nPu

blic

atio

n bi

asPr

eval

ence

10

/100

0Pr

eval

ence

50

/100

0Pr

eval

ence

10

0/10

00

True

pos

itive

s (in

divi

dual

s w

ith T

B)C

ross

-se

ctio

nal

Insu

ffici

ent d

ataa

No

data

No

data

No

data

False

neg

ativ

es (i

ndiv

idua

ls in

corre

ctly

cla

ssifi

ed a

s no

t ha

ving

TB)

Cro

ss-

sect

iona

lIn

suffi

cien

t dat

aaN

o da

taN

o da

taN

o da

ta

False

pos

itive

s (in

divi

dual

s in

corre

ctly

cla

ssifi

ed a

s ha

ving

TB)

Cro

ss-

sect

iona

lN

onea

Non

ebN

onec

Serio

us (–

1)d

Und

etec

tede

Mod

erat

e

True

neg

ativ

es (i

ndiv

idua

ls w

ithou

t TB)

Cro

ss-

sect

iona

lN

onea

Non

ebN

onec

Serio

us (–

1)d

Und

etec

tede

Mod

erat

e

CrI,

cre

dibl

e in

terv

al.

a Th

ree

studi

es p

rovi

ded

data

abo

ut u

sing

Xper

t M

TB/

RIF

on c

ereb

rosp

inal

flui

d to

dia

gnos

e TB

men

ingi

tis.

Ther

e w

ere

insu

ffici

ent

data

to

calc

ulat

e th

e se

nsiti

vity

of

Xper

t M

TB/

RIF.

In t

wo

studi

es,

two

out

of s

ix c

ultu

re-p

ositi

ve c

hild

ren

wer

e al

so f

ound

to

be p

ositi

ve u

sing

Xper

t M

TB/

RIF.

One

stu

dy d

id n

ot h

ave

any

posit

ive

resu

lts f

or c

ultu

re o

r Xp

ert

MTB

/RI

F. b

The

studi

es w

ere

all p

erfo

rmed

at h

ighe

r lev

els

of c

are

(such

as

refe

rral h

ospi

tals)

, whi

ch re

flect

s th

e po

pula

tion

that

wou

ld b

enefi

t fro

m u

sing

Xper

t MTB

/RI

F on

cer

ebro

spin

al fl

uid.

The

re w

ere

no

conc

ern

for i

ndire

ctne

ss.

c Fo

r the

3 s

tudi

es, t

he p

oint

esti

mat

es fo

r spe

cific

ity w

ere

86%

, 97%

and

100

%, i

ndic

atin

g lit

tle h

eter

ogen

eity.

The

95%

con

fiden

ce in

terv

als

over

lapp

ed. T

here

was

no

conc

ern

abou

t inc

onsis

tenc

y.

d O

nly

3 stu

dies

with

a to

tal o

f 51

child

ren

wer

e in

clud

ed in

the

revi

ew. T

he c

onfid

ence

inte

rval

s fo

r bot

h se

nsiti

vity

and

spe

cific

ity w

ere

wid

e (c

ross

ing

+/–

20%

). Th

ere

wer

e ve

ry s

erio

us c

once

rns

abou

t im

prec

ision

. The

evi

denc

e w

as n

ot d

owng

rade

d fu

rther

giv

en th

e do

wng

radi

ng fo

r inc

onsis

tenc

y.e

One

unp

ublis

hed

study

was

incl

uded

. A fe

w o

ther

stud

ies w

ere

in p

rogr

ess b

ut d

ata

wer

e no

t ava

ilabl

e fo

r ana

lysis

. The

dat

a in

clud

ed in

the

revi

ew d

id n

ot a

llow

for f

orm

al a

sses

smen

t of p

ublic

atio

n bi

as u

sing

met

hods

suc

h as

funn

el p

lots

or re

gres

sion

tests

; suc

h te

chni

ques

hav

e no

t bee

n fo

und

to b

e he

lpfu

l in

studi

es e

valu

atin

g di

agno

stic

accu

racy

. How

ever

, rep

ortin

g bi

as w

as c

onsid

ered

to

be m

inim

al s

ince

Xpe

rt M

TB/

RIF

is a

new

test

that

has

had

con

sider

able

atte

ntio

n an

d sc

rutin

y.


Annexes

Annex 1. Members of the Expert Group

Sevim Ahmedov (Observer)Senior Tuberculosis Technical AdvisorUnited States Agency for International Development 1300 Pennsylvania Ave NWWashington, DC 20523E-mail: [email protected]

Lucia BarreraServicio Micobacterias (Mycobacteria Laboratory)Velez Sarsfield 5631281 Buenos AiresArgentinaE-mail [email protected]

Catharina BoehmeSenior Medical OfficerFoundation for Innovative New DiagnosticsAvenue de Budé 161202 GenevaSwitzerlandE-Mail: [email protected]

Lucy Cheshire Tuberculosis Advocacy Consortium Adalyn CourtSuite C6Ngong RoadNairobi KenyaE-mail: [email protected]

Gavin Churchyard (Observer)The Aurum Institute NPCThe Ridge29 Queens RoadParktownJohannesburgSouth AfricaE-mail: [email protected]

Daniela Maria Cirillo HeadEmerging Bacterial Pathogens Unit San Raffaele Scientific Institute Via Olgettina 58 20132 Milan ItalyE-mail: [email protected]

Frank Cobelens (Observer)Amsterdam Institute of Global Health and Development &Department of Global HealthAcademic Medical CenterAmsterdam The NetherlandsE-mail: [email protected]

Bill Coggin (Observer)Department of State/PEPFAROffice of the US Global AIDS Coordinator2100 Pennsylvania Ave NWWashington, DC 20522United StatesE-mail: [email protected]

Colleen DanielsDirector TB/HIVTreatment Action GroupNew York, NY 10016United StatesE-mail: [email protected]

Claudia Denkinger (systematic reviewer)McGill University4075 Rue CartierMontreal, QuebecCanadaE-mail: [email protected]


Anne Detjen (systematic reviewer)Technical Consultant North America Office The International Union Against Tuberculosis and Lung Disease 61 Broadway, Suite 1720New York, NY 10006 United States E-mail: [email protected]

Mildred FernandoManagement Sciences for Health, Inc. 181 Floro Subdivision Malhacan, MeycauayanBulacan PhilippinesE-mail: [email protected]

Nazir Ismail (Observer)Head of Centre for Tuberculosis National Institute for Communicable Diseases1 Modderfontein RoadSandringhamJohannesburgSouth AfricaE-mail: [email protected]

Moses JolobaDepartment of Medical MicrobiologyMakerere University College of Health Sciences2nd Floor Pathology/Microbiology BLRoom C31, Upper Mulago Hill Road256 KampalaUgandaE-mail: [email protected]

Anna Mandalakas (systematic reviewer)DirectorGlobal Tuberculosis and Mycobacteriology ProgramThe Tuberculosis InitiativeTexas Children’s Hospital1102 Bates St, FC-630 Houston, Texas 77030United StatesE-mail: [email protected]

Andrea Pantoja (systematic reviewer)ConsultantAnkenhofstrasse 23 Oberengstringen 8102 SwitzerlandE-mail: [email protected]

Holger Schünemann ChairDepartment of Clinical Epidemiology & BiostatisticsMcMaster University Health Sciences CentreRoom 2C10B1200 Main Street WestHamilton, Ontario ON L8N 3Z5CanadaE-mail: [email protected]

Moorine Penniah SekaddeNational Tuberculosis and Leprosy ProgrammeKampalaUgandaE-mail: [email protected]

Thomas M ShinnickAssociate Director for Global Laboratory ActivitiesCenters for Disease Control and Prevention1600 Clifton RoadMS-G35, NEAtlanta, GA 30333United StatesE-mail: [email protected]

Karen Steingart (systematic reviewer)EditorCochrane Infectious Diseases Group United StatesE-mail: [email protected]

Sabira TahseenNational coordinatorNational TB Control Programme IslamabadPakistanE-mail: [email protected]


Maarten van CleeffKNCV Tuberculosis FoundationParkstraat 172514 JD The HagueThe NetherlandsE-mail: [email protected]

Francis VaraineMédecins sans Frontières8 rue St Sabin75011 ParisFranceE-mail: [email protected]

Annex 2. WHO staff members

Haileyesus Getahun E-mail: [email protected]

Christopher Gilpin E-mail: [email protected]

Jean Iragena E-mail: [email protected]

Knut Lönnroth Email: [email protected]

Lisa Nelson Email: [email protected]

Fuad Mirzayev E-mail: [email protected]

Wayne van Gemert Email: [email protected]

Karin Weyer E-mail: [email protected]

Matteo Zignol E-mail: [email protected]

Annex 3. Members of the Strategic and Technical Advisory Group for Tuberculosis (STAG-TB)

Dr Draurio BarreiraHeadNational TB Control Program Ministry of HealthBrasilia-DFBrazil

Dr Amy BloomSenior Technical AdviserUS Agency for International Development (USAID)Washington, DCUnited States

Prof. Gavin ChurchyardChief Executive OfficerThe Aurum Institute NPCParktown, Johannesburg South Africa

Dr Daniela M CirilloHeadEmerging Bacterial Pathogens UnitSan Raffaele del Monte Tabor Foundation San Raffaele Scientific InstituteMilanItaly


Professor Elizabeth CorbettTropical EpidemiologyLondon School of Hygiene & Tropical Medicine and Malawi Liverpool Wellcome Trust Clinical Research ProgrammeBlantyre Malawi

Dr Charles L DaleyChiefDivision of Mycobacterial and Respiratory InfectionsNational Jewish Medical and Research CenterDenver, COUnited States

Professor Stephen GrahamInternational Child HealthUniversity of MelbourneDepartment of PaediatricsRoyal Children’s HospitalParkville, MelbourneAustralia

Dr Akramul IslamAssociate DirectorHealth, Nutrition and Population ProgramBangladesh Rural Advancement Committee CentreDhakaBangladesh

Dr Michael KimerlingSenior Program Officer, TBGlobal Health ProgramBill and Melinda Gates FoundationSeattle, WAUnited States

Dr Jaime LagahidNational Center for Disease, Prevention and ControlDepartment of HealthManilaThe Philippines

Dr Chakaya J Muhwa(STAG-TB Chair)Chief Research OfficerCentre for Respiratory Diseases ResearchKenya Medical Research InstituteNairobiKenya

Dr A PrakashJoint SecretaryDisease ControlMinistry of Health and Family WelfareNew DelhiIndia

Dr Ejaz Qadeer ManagerNational TB Control ProgrammeFederal Ministry of HealthIslamabadPakistan

Dr Joseph Sitienei National TB Programme ManagerDivision of Leprosy, Tuberculosis and Lung DiseaseMinistry of HealthNairobiKenya

Dr Alena SkrahinaScientific DirectorRepublican Research and Practical Centre for Pulmonology and TuberculosisMinskBelarus

Dr Soumya SwaminathanDirector National Institute for Research in TuberculosisIndian Council for Medical ResearchChennaiIndia


Dr Francis VaraineCoordinator of MSF Working Group on TuberculosisMédecins Sans FrontièresParisFrance

Dr Maarten van CleeffProgram Director TB Care KNCV Tuberculosis FoundationThe HagueThe Netherlands

Dr Dalene von DelftMedical DoctorTB PROOFSomerset WestCape TownSouth Africa

Annex 4. Declarations of Interests

None declaredLucia Barrera, Lucy Cheshire, Colleen Daniels, Holger Schünemann (Chairperson), Moorine Penniah Sekadde, Sabira Tahseen, Maarten van Cleeff

Declared, insignificantDaniela Maria Cirillo: meeting participation sponsored by Cepheid; research grant support from FIND and ST microelectronics.

Mildred Fernando: Management Sciences for Health contributed approximately US$ 80 to enable her to participate in the Expert Group meeting; the Xpert MTB/RIF assay may have been used to determine if her TB had relapsed.

Thomas M Shinnick: received funding from the United States Government to participate in the meeting.

Francis Varaine: leader of the Médecins sans Frontières working group on TB; required to defend positions related to TB diagnostics.

Declared, significant (observer status)Sevim Ahmedov: United States government employee; the US government has contributed funds to enable reductions in the end-user price of cartridges used for the Xpert MTB/RIF test.

Catharina Boehme: employed by FIND, which had a cofunding agreement with Cepheid supporting the development of Xpert MTB/RIF.

Gavin Churchyard: employed by the Aurum Institute, which received funding from the Bill and Melinda Gates Foundation for trials using the Xpert MTB/RIF assay to diagnose TB in South Africa, and to evaluate the impact of the test and its cost effectiveness during routine roll-out of the test.

Frank Cobelens: consultant to the Bill and Melinda Gates Foundation for research on rolling out and scaling up the use of Xpert MTB/RIF; also conducted a cost-effectiveness analysis of Xpert MTB/RIF for FIND.

Bill Coggin: United States government employee; the US government has contributed funds to enable reductions in the end-user price of cartridges used in the Xpert MTB/RIB test.

Claudia Denkinger: conducted the systematic review of using Xpert MTB/RIF to diagnose extrapulmonary TB

Anne Detjen: conducted the systematic review of using Xpert MTB/RIF to diagnose paediatric TB


Nazir Ismail: employed by the National Health Laboratory Service of South Africa and involved in the roll out of Xpert MTB/RIF testing in South Africa; South Africa purchased more than 1 million cartridges; Nazir Ismail received no financial gain from this process.

Moses Joloba: laboratory manger of the National TB Reference Laboratory in Kampala which was used by FIND as one testing site for the initial Xpert MTB/RIF demonstration studies.

Anna Mandalakas: conducted the systematic review of using Xpert MTB/RIF to diagnose paediatric TB.

Andrea Pantoja: conducted the review of the affordability and cost effectiveness of Xpert MTB/RIF.

Karen Steingart: conducted the systematic review of using Xpert MTB/RIF to diagnose pulmonary TB.

Automated real-time nucleic acid amplification technology for rapid

and simultaneous detection of tuberculosis and rifampicin resistance:

Xpert MTB/RIF assay for the diagnosis of pulmonary and extrapulmonary TB


Global TB ProgrammeWorld Health OrganizationAvenue Appia 20, CH-1211 Geneva-27, Switzerland

Information Resource Centre HTM/GTB:Email: [email protected]: www.who.int/tb

ISBN 978 92 4 150633 5

POLICY UPDATE

PE

RF

OR

MA

NC

EA

CC

UR

AC

Y ACCURACYRECOMMENDATIONS

MOLECULAR DIAGNOSTICS

TUBERCULOSISDRUG-RESISTANCE

NEW DIAGNOSTIC TESTS RIF

AM

PIC

INP

UL

MO

NA

RY T

B

RA

PID

TB

TE

ST

RESISTANCEDIAGNOSIS

MYCOBACTERIUM

TB

TB/HIV

COVER policy_update_OK.indd 1 11/04/2014 14:50

Documents

Automated real-time nucleic acid amplification technology ...apps.who.int/iris/bitstream/10665/112472/1/9789241506335_eng.pdf · 4.2 Using Xpert MtB/riF to diagnose eXtrapUlMonary