Bacteriophage Assay

8/6/2019 Bacteriophage Assay

1/13

Bacteriophage-based assays for the rapid detectionof rifampicin resistance in Mycobacteriumtuberculosis : a meta-analysisMadhukar Pai a,b, *, Shriprakash Kalantri a,c , Lisa Pascopella d ,Lee W. Riley a , Arthur L. Reingold a

a Division of Epidemiology, School of Public Health, University of California, 140, Warren Hall, Berkeley,CA 94720, USAb Division of Pulmonary and Critical Care Medicine, San Francisco General Hospital, University of California,San Francisco, CA, USAc Department of Medicine, Mahatma Gandhi Institute of Medical Sciences, Sevagram, Indiad Surveillance and Epidemiology Section, Tuberculosis Control Branch, California Department of HealthServices, Berkeley, CA, USA

Accepted 16 May 2005

KEYWORDS

Tuberculosis;Multi-drug resistanttuberculosis;Rifampicin resistance;Bacteriophage;Phage;Diagnosis;Accuracy;Sensitivity andspecicity

Abstract Objective : To summarize, using meta-analysis, the accuracy of bacterio-

phage-based assays for the detection of rifampicin resistance in Mycobacteriumtuberculosis .

Methods : By searching multiple databases and sources we identied a total of 21studies eligible for meta-analysis. Of these, 14 studies used phage amplicationassays (including eight studies on the commercial FASTPlaque-TB w kits), and sevenused luciferase reporter phage (LRP) assays. Sensitivity, specicity, and agreementbetween phage assay and reference standard (e.g. agar proportion method orBACTEC 460) results were the main outcomes of interest.

Results : When performed on culture isolates ( N Z 19 studies), phage assays appearto have relatively high sensitivity and specicity. Eleven of 19 (58%) studies reportedsensitivity and specicity estimates R 95%, and 13 of 19 (68%) studies reported R 95%agreement with reference standard results. Specicity estimates were slightly lowerand more variable than sensitivity; 5 of 19 (26%) studies reported specicity ! 90%.

Only two studies performed phage assays directly on sputum specimens; althoughone study reported sensitivity and specicity of 100 and 99%, respectively, anotherreported sensitivity of 86% and specicity of 73%.

Conclusions : Current evidence is largely restricted to the use of phage assays forthe detection of rifampicin resistance in culture isolates. When used on culture

Journal of Infection (2005) 51 , 175187

www.elsevierhealth.com/journals/jinf

0163-4453/$30.00 Q 2005 The British Infection Society. Published by Elsevier Ltd. All rights reserved.doi:10.1016/j.jinf.2005.05.017

* Corresponding author. Address: Division of Epidemiology, School of Public Health, University of California, 140, Warren Hall,Berkeley, CA 94720, USA. Tel.: C 1 510 388 7137; fax: C 1 510 643 4927.

E-mail address: [email protected] (M. Pai).
http://www.elsevierhealth.com/journals/jinfhttp://www.elsevierhealth.com/journals/jinf


2/13


3/13

available, whereas LRP tests are still underdevelopment and restricted to research settings.In both systems, drug resistance is diagnosed whenM. tuberculosis is detected in samples that containthe drug (e.g. RIF). When phage-based assays fail todetect M. tuberculosis in drug containing samples,the strains are classied as drug-sensitive. 8,9

Phage-based assays are available as commercialkits (e.g. FASTPlaque-TB

w

and PhageTek MBw

, avariant of the FASTPlaque-TB, Biotec LaboratoriesLtd, U. K.) and as in-house (laboratory-developed)assays. 8,9 In-house assays use either amplicationtechnology (e.g. PhaB) or LRPs. Some of the phage-based assays are specically designed to rapidlydetect rifampicin resistance (e.g. FASTPlaque-TB-MDRiw , earlier called FASTPlaque-TB-RIF w ) inculture isolates. Newer kits are being developedfor the rapid detection of drug resistance directlyfrom clinical specimens such as sputum (e.g.FASTPlaque-TB-Response w ).

To evaluate the accuracy of phage-based assaysfor rifampicin resistance, we conducted a systema-tic review and meta-analysis. A secondary objectivewas to assess if studies produced heterogeneousestimates of accuracy, and to evaluate if suchheterogeneity is due to variability in test charac-teristics and study methodology.

Methods

Our meta-analysis was conducted according to apre-specied protocol. The outline of the methodsused in our systematic review has been describedelsewhere. 14,15 Briey, the review involved thefollowing major steps: (1) formulation of the reviewquestion, (2) a comprehensive, systematic searchand selection of primary studies, (3) criticalappraisal of included studies for quality, (4) data

extraction and contact with authors for additionalinformation, (5) synthesis and summary of studyresults, and (6) interpretation of the results.

Study eligibility

We aimed to include original clinical studies thatevaluated the sensitivity and specicity of phage-based assays (either phage amplication or LRP)for the detection of rifampicin-resistantM. tuberculosis in clinical specimens or isolates.Both commercial tests and in-house assays were

included, but separately analysed. The followingstudies were excluded: animal experiments, proofof principle studies on development of new assays,review articles, letters, and commentaries, andstudies that did not evaluate both sensitivity andspecicity in the same population.

With respect to study design, we includeddiagnostic studies (either case-control or cross-sectional) that evaluated the accuracy (sensitivityand specicity) of phage-based assays against areference standard. Any of the following conven-tional drug susceptibility tests were accepted asreference standards: absolute concentrationmethod, proportion method, resistance ratiomethod, and radiometric BACTEC 460 method.

Search strategy

We searched the following electronic databases forprimary studies and conference abstracts: PubMed(19852004), Web of Science (19852004), Embase(19882004), BIOSIS (1993 2004), Cochrane Library(2004), and Google Scholar (December 2004). Allsearches were up to date as of December 2004.

Figure 1 An overview of the phage amplication assay.Source: Hazbon, 2004 6 ; reproduced with permission fromBiomedica.

Figure 2 An overview of the luciferase reporter phageassay.

Phage assays for drug-resistant tuberculosis 177


4/13

We did not impose language restrictions in oursearches. The keywords and search terms usedincluded tuberculosis, mycobacterium tubercu-losis , mycobacteria, bacteriophages, myco-bacteriophage, phage*, Fastplaque, phageamplication, phage-based, bacteriophage-based, luciferase, sensitivity and specicity,

accuracy and predictive value. We also con-tacted experts in the eld to identify ongoing andunpublished studies, and searched the referencelists from the primary studies and review articles.To identify all relevant studies of commercialassays, we contacted and obtained lists of studiesfrom commercial kit manufacturers.

Study selection

Two reviewers (SP and MP) independently screenedthe citations. Citations that were relevant on the

rst screen of titles and abstracts were evaluatedfurther by review of full-text reports. Disagree-ments between the reviewers were resolved byconsensus. A list of excluded studies, along with thereasons for exclusion is available from the authorson request.

Data extraction and assessment of studyquality

The nal set of included articles was assessed byone reviewer (MP), who extracted data from all the

studies using a piloted data extraction form. Asecond reviewer (SP) independently extracted datafrom a subset (one-fourth) of the included studies.The inter-rater agreement between the tworeviewers for sensitivity and specicity data was100%. Data retrieved from the reports includedmethodological quality, phage test characteristics,reference standards employed, outcome data(sensitivity, specicity, and agreement betweenphage results and conventional DST), and pro-portion of results that were indeterminate, con-taminated or had intermediate sensitivity. Sincediscrepant analysis (where discordant resultsbetween index test and reference standard areresolved, post-hoc, using clinical or other labora-tory data) may be a pot en tial source of bias indiagnostic evaluations, 16 we preferentiallyincluded unresolved data where available. Thisapproach is likely to result in more conservativeestimates of accuracy, and more likely to reectthe performance of tests in routine practice ratherthan research settings.

We assessed the quality of the studies by usingthe following criteria: selection bias (consecutive

or random sampling of patients/specimens versusstudies that used neither method), blinding (single/double blind versus unblinded interpretation ofphage test and reference standard results), andpotential for verication bias (complete versuspartial/differential verication of index test resultsby reference standards). We contacted authors,

when necessary, to obtain additional informationon study quality and results. Additional informationwas obtained from the authors for 16 of 21 (76%)studies.

Data synthesis and meta-analysis

To compute sensitivity and specicity of phage-based assays, we cross-tabulated the phage drugsusceptibility results against those obtained fromthe conventional DST (e.g. BACTEC). Sensitivity(true positive rate) refers to the proportion of

rifampicin-resistant strains that are correctlyidentied as resistant by phage-based assays.Specicity (true negative rate) refers to theproportion of rifampicin-susceptible strains thatare correctly identied as susceptible by phage-based assays. For the computation of thesemeasures, most studies excluded test results thatwere uninterpretable, contaminated, indetermi-nate, or had intermediate drug sensitivity. Inaddition to sensitivity and specicity, we computedmeasures of agreement to evaluate the correlationbetween conventional DST and phage assay results.

Simple proportion agreement and kappa estimateswere computed for each study, along with 95%condence intervals (95% CI).

We used standard met hods recommended fordiagnostic meta-analyses. 15,17,18 Data were ana-lysed using Meta-DiSc (version 1.1.1) software. Ouranalyses focused on sensitivity (true positive rate[TPR]), and specicity (1-false positive rate [FPR])for the detection of rifampicin resistance. BecauseTPR and FPR are correlated and vary with thethresholds (cut-points for determining test posi-tives) employed in the original studies, we did notpool the sensitivity and specicity estimatesseparately; instead we analysed TPR and FPR aspairs, and explored the effect of variability in cut-points on study results. 15,17,18

We summarized the joint distribution of TPR andFPR using the summar y receiver operating charac-teristic (SROC) curve. 19 Unlike a traditional ROCplot that explores the effect of varying thresholdson sensitivity and specicity in a single study, eachdata point in the SROC space represents anindividual study. As described by Littenberg andMoses, the SROC curve is obtained by tting a

M. Pai et al.178


5/13

regression curve to pairs of TPR and FPR. 19 Tofacilitate the SROC analysis, a correction was madeto account for zero values in any cell of the 2 ! 2table generated from each study0.5 was added toall cells in the 2 ! 2 table. The SROC plots,therefore, display the zero cell-corrected data.

The SROC curve and the area under it present an

overall summary of test performance, and displaythe trade off between sensitivity and specicity. Asymmetric, shoulder-like SROC curve suggests thatvariability in thresholds employed c ould, in part,explain variability in study results. 15,1719 It alsosuggests a common, homogeneous underlying diag-nostic odds ratio that does not change with thediagnostic threshold. The area under the ROC curve(AUC) and the Q* index are useful global sum mariesof the SROC curve and test performance. 19,20 Anarea under the ROC c urve of 1.0 indicates perfectdiscriminatory ability. 19,20 The Q* index, dened bythe point where sensitivity equals specicity on theSROC curve, is that point on the SROC curve that isintersected by the anti-diagonal. A Q* value of 1.0indicat es 100% accuracy (sensitivity and specicityof 1.0). 19,20 Thus, the closer the Q* value is to 1.0,the more accurate the test.

In meta-analyses, heterogeneity refers to asubstantial degree of variability in study results. 21

Such heterogeneity could be due to variability inthresholds, disease spectrum , assay methods, andstudy quality across studies. 21 In the presence ofsignicant heterogeneity, pooled, summary esti-mates from meta-analyses are hard to interpret.We

investigated heterogeneity using subgroup (strati-ed) analyses. To account for the major differencesin assay methods, we analysed phage amplicationassays separately from LRP tests. Within phageamplication assays, commercial assays were eval-uated separately from in-house assays. In addition,tests performed on culture isolates were evaluatedseparately from assays performed on clinical speci-mens. These subgroups were identied a priori andpre-specied in the study protocol. Subgroupanalyses also allowed us to determine if anyparticular phage-based method was more likely tobe associated with higher accuracy.

Results

We identied a total of 540 unique citations from allliteratu re sea rches. Of these, a total of 21articles 12,2241 (17 English, 1 Chinese, an d 1Turkish), including two conference abstracts 22,41

(English) were eligible for inclusion in our meta-analysis. However, one study was excluded from

the analysis because it had only a single case ofconrmed rifampicin resistance. 28 One articlereported outcome data separately for two variantsof the luciferase reporter phage assay (B ronx boxversus luminometric detection methods). 32 Thesewere counted as separate studies. Our nalanalyses, therefore, had 21 studies.

Description of included studies

The Table 1 presents the characteristics of the 21included studies, along with data on methodologi-cal quality and outcomes. O f these studies, 14studies 12,2225,2931,33,34,3638,40 evaluated a mpli-cation- based assays, whereas seven studies 26,27,32,35,39,41 evaluated LRP assays. Among the 14 phageamplication assays, eight studies 2225,29,33,34,38

used commercial assays (FASTP laque-TB), an d sixstudies used in-house assays. 12,30,31,36,37,40 The

average sample size was 85 specimens or isolates(range 22201). On average, studies included fewerresistant than sensitive strains (mean 27 versus 58,respectively).

BACTEC 460 and proportion methods were themost commonly employed reference tests for con-ventional DST. Eight of 21 (38%) studies recruitedpatients or specimens using a random or consecutivesampling strategy. Twelve of 21 (57%) studies used astudy design that included blinded interpretation ofeither the phage test or the reference standardresults (singleor double blinded). None of the studieshad potential for verication bias.

With respect to assay characteristics, moststudies used culture isolates rather than clinicalspecimens for the phage assays. Also studiesemployed varying denitions of drug resistance.For example, a majority of the studies on theFASTPlaque-TB-RIF assay used R 50 plaques on theRIFK containing plate for classifying a strain asresistant. One study used a cut-point of R 20plaques. Cut-points used in LRP studies also varied,depending on the method used for detection ofluciferase activity (luminometry versus photo-graphic methods). The proportion of phage results

that were indeterminate, and/or contaminated orhad intermediate sensitivity, at least on initialtesting, varied from 0 to 17%.

Accuracy of phage amplication assays

Of the 14 studies that evaluated phage ampli-cation assays, commercial kits were used in eightstudies. These kits included the FASTPlaque-TB-RIF(now called FASTPlaque-TB-MDRi), and FASTPlaque-TB-Response assays. The remaining six studies



6/13


7/13

evaluated D-29 phage-based in-house assays. Withthe exception of two studies, all amplicationassays used culture isolates. Two studies directlyapplie d th e FASTPlaque-TB assays on sputum speci-mens. 24,29

Fig. 3 presents the forest plot of sensitivity andspecicity estimates for the eight commercial phageamplication assays. The results are also displayed asa SROC curve in Fig. 4(A). As seen in Fig. 3, thesensitivity estimates ranged from 0.81 to 1.0. Five of8 (63%) studies had a perfect sensitivity of 100%. Thespecicity estimates, on the other hand, were morevariable than sensitivity (range 0.731.0). Although 4of 8 (50%) studies had specicity estimates O 95%,there were three studies with relatively pooraccuracy (specicity ranging from 0.730.82). TheSROC curve (Fig. 4(A)) shows a high level of overallaccuracy with an area of 0.99, and Q* of 0.95. Of thetwo studies that performed ph age assays directly onsputum specimens, one study 24 reported sensitivityand specicity of 1.0and0.99, respectively, whereasanot her reported sensitivity of 0.86 and specicity of0.73. 29

Fig. 5 presents the forest plot for the six in-housephage amplication assays. Both sensitivity (range0.971.0) and specicity (range 0.841.0) estimateswere consistently high. The SROC curve ( Fig. 4(B))shows a high level of overall accuracy with an areaof 0.99, and Q* of 0.98.

The Table 1 presents the data on agreementbetween conventional DST and phage amplication(commercial and in-house) assays. Agreementestimates were high11 of 14 (79%) studies hadagreement O 90%. The kappa estimates were alsohigh, and ranged from 0.56 to 1.0.

Accuracy of luciferase reporter phage assays

A total of seven studies evaluated phage a ssays thatwer e based on luciferase reporter phages. 26,27,32,35,39,41 All of these were in-house assays, and all usedculture isolates. None of the studies directlyapplied the LRP tests on sputum specimens. Withthe exception of two studies, all the LRP studiesused luminometric detection of the light signal.Two studies used the Bronx box (a bo x containingPolaroid lm) to detect light emission. 32,39

Fig. 6 presents the forest plot of sensitivity andspecicity estimates for the seven LRP assays. Theresults are also displayed as a SROC curve inFig. 4(C). As seen in Fig. 6, all studies except onehad a sensitivity of 100%. Specicity estimates, onthe other hand, were slightly lower and morevariable (range 0.891.0). The SROC plot(Fig. 4(C)) suggests high accuracy with an area of0.98 and Q* of 0.95.

The Table 1 presents the data on agreementbetween conventional DST and LRP assays.

Figure 3 Forest plot of sensitivity and specicity for commercial phage amplication assays (FASTPlaque-TB tests)performed on culture isolates ( N Z 6 studies), and clinical specimens ( N Z 2 studies). (A) Sensitivity. (B) Specicity. Thepoint estimates of sensitivity and specicity from each study are shown as solid circles (when performed on cultureisolates) and squares (when directly performed on sputum specimens). Error bars are 95% condence intervals (CI).



8/13

Agreement estimates were exceptionally highallseven studies had agreement O 90%. The kappaestimates were ranged from 0.79 to 1.0.

Impact of assay methodology on accuracy

Overall, the SROC plots stratied by type of assay

(Fig. 4(A)(C)) were fairly similar for all three assays(commercial and in-house phage amplication, andLRP assays), although commercial assays appear toproduce lower and more variable accuracy esti-mates. The area under the SROC curve and Q*estimates were fairly similar, indicating compar-able diagnostic accuracy.

Isoniazid resistance among rifampicin-resistant isolates

Seven studies provided data on INH resistance by

conventional DST methods among strains th at wereresistant to rifampicin. 12,24,25,29,31,37,40 Thesestudies showed that, on average, 96% of therifampicin-resistant isolates were also resistant toINH. Rifampicin resistance, therefore, was a fairlygood surrogate marker of MDR-TB in the populationsstudied.

Discussion

Principal ndings and clinical implications

Our meta-analysis of 21 studies (with all but twostudies using culture isolates) on phage-basedassays for the detection of rifampicin resistancesuggests that phage assays are associated withfairly high sensitivity and specicity, when appliedto culture isolates. These results indicate that, as agroup, phage assays show promise for the detectionof rifampicin resistance in culture isolates whencompared to conventional agar-based drug suscep-tibility tests and rapid methods such as BACTEC 460.Once an isolate was drawn from a growing culture,the average turnaround time for phage-basedassays was 4872 h, whereas with conventionalDST, the turnaround time might vary from 12weeks for BACTEC and MGIT, and 34 weeks for solidmedia-based methods. In addition to rapidity,phage-based assays have the advantage of beingless expensive and technologically simpler thanmolecular (e.g. PCR) and liquid-media based tests(e.g. MGIT, BACTEC). This offers some advantagesin resource-limited settings.

Although these results suggest that phage-basedassays are fairly sensitive and specic for rifampicin

Figure 4 Summary ROC plots for phage-based assays,stratied by type of assay. (A) Commercial amplicationassays (FASTPlaque-TB) (N Z 8 studies). (B) In-houseamplication assays (D29 phage-based assays) ( N Z 6studies). (C) Luciferase reporter phage assays ( N Z 7

studies). Each solid circle represents each study in themeta-analysis. The curve is the regression line thatsummarizes the overall diagnostic accuracy. SROC,summary receiver operating characteristic; AUC, areaunder the curve; SE (AUC), standard error of AUC; Q*, anindex dened by the point on the SROC curve where thesensitivity and specicity are equal, which is the pointclosest to the top-left corner of the ROC space; SE(Q*),standard error of Q* index.

M. Pai et al.182


9/13

resistance, it is worth noting that phage-basedassays do not show such high accuracy for the directdetec ti on of M. tuberculosis in clinical speci-mens. 13 In a previous meta-analysis, we summar-ized data from 13 studies that evaluatedcommercial and in-house phage-b ased assays forthe diagnosis of pulmonary TB. 13 The resultssuggested that phage-based assays have highspecicity (range 0.831.00), but lower and variable

sensitivity (range 0.210.88). This meta-analysis

also showed that phage-based assays have perform-ance characteristics that are fairly similar to that ofsputum microscopyhigh specicity, but modestand variable sensitivity. 13 A major differencebetween the two meta-analyses is type of speci-mens used in the studies included; almost allstudies (19 of 21) in the current meta-analysisused culture isolates, whereas all the studies (13 of13) analysed in the previous meta-analysis directly

performed the tests on sputum specimens. This

Figure 5 Forest plot of sensitivity and specicity for in-house phage amplication assays, all performed on cultureisolates ( N Z 6 studies). (A) Sensitivity. (B) Specicity. The point estimates of sensitivity and specicity from each studyare shown as solid circles. Error bars are 95% condence intervals (CI).

Figure 6 Forest plot of sensitivity and specicity for luciferase reporter phage assays, all performed on cultureisolates ( N Z 7 studies). (A) Sensitivity. (B) Specicity. The point estimates of sensitivity and specicity from each studyare shown as solid circles. Error bars are 95% condence intervals (CI).



10/13

difference in the specimens used has majorimplications for accuracy of phage-based assaysand their clinical utility.

Although phage assays for rifampicin resistanceare usually performed after primary isolation of M.tuberculosis , their reasonably high accuracy hasgreater clinical implications if they that can be

directly applied to sputum specimensa necessarycondition in resource-limited settings where TBburden is greatest, and where cultures are not easyto obtain. Unfortunately, only two of 21 studies inour meta-analysis eval uate d phage assays directlyon sputum specimens. 24,29 Albert and colleaguesevaluated the new FASTPlaque-TB-Response testdirectly on sputum specimens and demonstrated asensit ivity and specicity of 100 and 99%, respect-ively. 24 However, Butt and co-workers using theFASTPlaque-TB-RIF test showed a more modestsensitivity of 86% and specicity of 73%. 29 Thisdifference in accuracy could be partly due to thevariation in denitions used for rifampicin resist-ance and cut-points used for valid results. In thestudy by Albert and colleagues, assay results wereconsidered valid if R 100 plaques were observed onthe RIFK plate, and a strain was determined to beresistan t if R 50 plaques were seen on the RIF Csample. 24 In the study by Butt and colleagues,results were considered valid if R 20 plaques wereseen on the RIF K plate, and a strain wasdetermined to be resistant when R 20 plaqueswere seen on RIF C plates. 29 This variability incut-points emphasizes the need for better stan-

dardization of phage assays, and for furtherresearch to identify optimal thresholds for validityand test-positivity.

Our results suggest that more research is neededto establish the accuracy of phage-based assayswhen clinical specimens are used for testing. If theyare found to be highly accurate for sputum speci-mens, they have the potential to make a signicantclinical impact, particularly by reducing the time todiagnosis. They can be used to identify patients thatare most likely to have MDR-TB, fail treatment ornot responding to conventional anti-tuberculosistreatment. They can also be used to individualizedrug regimens to suit the resistance pattern of theorganism. This can be very helpful in MDR-TB hotspots wh ere the prevalence of drug resistance isvery high. 1

However, before phage assays can be success-fully used in routine practice, several concerns haveto be addressed. These include unexplained lowsensitivity and specicity in some studies, andpotential for contamination and indeterminate/uninterpretable results. These concerns have beenraised previously in the context of phage assays for

TB diagnosis. 13,42 They will also be important in thecontext of DST. Some studies reported low speci-city estimates (in the ra nge of 80%), and this iscause for concern. 22,36,38 In settings with a rela-tively lower prevalence of MDR-TB, false positiverates in the range of 1020% can result in signicantfalse over-reporting of MDR-TB. Because bacterio-

phages can replicate in non-tuberculous mycobac-teria (NTM) as well as M. tuberculosis , there isalways a potential for false positive results whenphage assays are directly applied to sputum speci-mens. Some non-tuberculous mycobacteria (e.g. M.smegmatis ) are naturally resistant to rifampicin,and in populations with high background prevalenceof NTM, there is a risk of over-diagnosing MDR-TB.Because of the toxicity associated with MDR-TBtreatment, it is important to study this issue furtherand determine the causes of false positive results invarious settings. As noted earlier, rifampicinresistance may not be a pe rfect surrogate markerof MDR-TB in all settings.2 In such populations,phage assay results for rifampicin, if used inisolation, might result in over-diagnosis of MDR-TB. For all these reasons, to minimize false positiveresults, a second conrmatory test (e.g. conven-tional DST for rst line drugs) may be necessary toconrm and validate all positive phage results.However, this facility may not be easily available inresource-limited settings.

Strengths and limitations of the review

Our systematic review had several strengths. Animportant strength of our study was its compre-hensive search strategy. We searched severaldatabases and sources, and also identied studiesby contacting authors, experts and test manufac-turers. We attempted to limit publication bias byincluding journal articles as well as conferenceabstracts, and by including non-English languagearticles. Screening, study selection, and qualityassessment were done independently and reprodu-cibly by two reviewers. We minimized the problemof missing data by contacting authors. A majority ofthe authors responded by sharing additional infor-mation on their primary studies. We performedmeta-analyses and exploration of hetero geneit y inaccordance with published guidelines. 14,17,18 Toaccountforvariation inassaymethods, weperformedsubgroup analyses and analysed the phage ampli-cation assaysseparatelyfrom theLRPassays, andalsoexamined tests performed on cultures separatelyfrom those used on clinical specimens.

Our review had some limitations. Firstly, we didnot address issues such as cost-effectiveness,

M. Pai et al.184


11/13


12/13

Rodriguez, Manzour Hazbon, Niaz Banaiee, and PaulRiska. We thank Heidi Albert, Richard Mole, PaulRiska, Ruth McNerney and Niaz Banaiee for theirassistance with identifying additional studies, andJavier Zamora for his support with the Meta-DiScsoftware. Lastly, we thank Ruth McNerney, PuneetDewan, and Ed Desmond for reviewing the draft

version of this manuscript.

References

1. Espinal MA. The global situation of MDR-TB. Tuberculosis(Edinb) 2003; 83 (13):4451.

2. World Health Organization. Anti-tuberculosis drug resist-ance in the world. Third global report . Geneva: WorldHealth Organization; 2004.

3. Fisher M. Diagnosis of MDR-TB: a developing world problemon a developed world budget. Expert Rev Mol Diagn 2002;2(2):1519.

4. Heifets LB, Cangelosi GA. Drug susceptibility testing ofMycobacterium tuberculosis : a neglected problem at theturn of the century. Int J Tuberc Lung Dis 1999; 3(7):56481.

5. Parsons LM, Somoskovi A, Urbanczik R, Salnger M. Labora-tory diagnostic aspects of drug resistant tuberculosis. FrontBiosci 2004;9 :2086105.

6. Hazbon MH. Recent advances in molecular methods for earlydiagnosis of tuberculosis and drug-resistant tuberculosis.Biomedical 2004; 24 (Suppl 1):14962.

7. Somoskovi A, Parsons LM, Salnger M. The molecular basis ofresistance to isoniazid, rifampin, and pyrazinamide inMycobacterium tuberculosis . Respir Res 2001;2(3):1648.

8. Mole RJ,Maskell T. Phage as a diagnosticthe useof phage inTB diagnosis. J Chem Technol Biotechnol 2001;76 (7):6838.

9. Trollip A, Albert H, Maskell T. Bacteriophage-based tech-nologies for the rapid diagnosis and drug susceptibilitytesting of tuberculosis. Am Clin Lab 2001; 20 (9):3942.

10. Jacobs Jr WR, Barletta RG, Udani R, Chan J, Kalkut G,Sosne G, et al. Rapid assessment of drug susceptibilities ofMycobacterium tuberculosis by means of luciferase reporterphages. Science 1993;260 (5109):81922.

11. McNerney R. TB: the return of the phage. A review of ftyyears of mycobacteriophage research. Int J Tuberc Lung Dis1999;3(3):17984.

12. Wilson SM, Al-Suwaidi Z, McNerney R, Porter J,Drobniewski F. Evaluation of a new rapid bacteriophage-based method for the drug susceptibility testing ofMycobacterium tuberculosis . Nat Med 1997;3 (4):4658.

13. Kalantri SP. Bacteriophage-based tests for the detection ofMycobacterium tuberculosis in clinical specimens: a sys-tematic review and meta-analysis. Masters in Public Health(MPH) Thesis, University of California, Berkeley; 2005.

14. Pai M, McCulloch M, Enanoria W, Colford Jr JM. Systematicreviews of diagnostic test evaluations: whats behind thescenes? ACP J Club 2004; 141 (1):A11A13.

15. Pai M, McCulloch M, Gorman JD, Pai N, Enanoria W,Kennedy G, et al. Systematic reviews and meta-analyses:an illustrated, step-by-step guide. Natl Med J India 2004;17 (2):8695.

16. Hadgu A. Discrepant analysis: a biased and an unscienticmethod for estimating test sensitivity and specicity. J ClinEpidemiol 1999;52 (12):12317.

17. Deeks JJ. Systematic reviews in health care: systematicreviews of evaluations of diagnostic and screening tests. BMJ2001;323 (7305):15762.

18. Irwig L, Tosteson AN, Gatsonis C, Lau J, Colditz G,Chalmers TC, et al. Guidelines for meta-analyses evaluatingdiagnostic tests. Ann Intern Med 1994;120 (8):66776.

19. Littenberg B, Moses LE. Estimating diagnostic accuracy frommultiple conicting reports: a new meta-analytic method.Med Decis Making 1993;13 (4):31321.

20. Walter SD. Properties of the summary receiver operatingcharacteristic (SROC) curve for diagnostic test data. StatMed 2002;21 (9):123756.

21. Lijmer JG, Bossuyt PM, Heisterkamp SH. Exploring sources ofheterogeneity in systematic reviews of diagnostic tests. StatMed 2002;21 :152537.

22. Aktepe OC, Altindis M, Esen N. A rapid methodfor rifampicinsusceptibility testing of Mycobacterium tuberculosis 11thEuropean Congress of Clinical Microbiology and InfectiousDiseases (ECCMID), 2001, Istanbul, Turkey .

23. Albert H, Heydenrych A, Mole R, Trollip A, Blumberg L.Evaluation of FASTPlaque TB-RIF, a rapid, manual test forthe determination of rifampicin resistance from Mycobac-terium tuberculosis cultures. Int J Tuberc Lung Dis 2001;

5(10):90611.24. Albert H, Trollip A, Seaman T, Mole RJ. Simple, phage-based(FASTPplaque) technology to determine rifampicin resist-ance of Mycobacterium tuberculosis directly from sputum.Int J Tuberc Lung Dis 2004; 8(9):11149.

25. Albert H, Trollip AP, Mole RJ, Hatch SJB, Blumberg L. Rapidindication of multidrug-resistant tuberculosis from liquidcultures using FASTPlaqueTB-RIF, a manual phage-basedtest. Int J Tuberc Lung Dis 2002;6(6):5238.

26. Banaiee N, Bobadilla-del-Valle M, Riska PF, Bardarov S,Small PM, Ponce-de-Leon A, et al. Rapid identication andsusceptibility testing of Mycobacterium tuberculosis fromMGIT cultures with luciferase reporter mycobacteriophages. J Med Microbiol 2003;52 (Pt 7):55761.

27. Banaiee N, Bodadilla-del-Valle M, Bardarov S, Riska PF,

Small PM, Ponce-de-Leon A, et al. Luciferase reportermycobacteriophages for detection, identication, andantibiotic susceptibility testing of Mycobacterium tubercu-losis in Mexico. J Clin Microbiol 2001;39 (11):38838.

28. BardarovS, Dou H, Eisenach K, BanaieeN, Ya S, Chan J, et al.Detection and drug-susceptibility testing of M-tuberculosisfrom sputum samples using luciferase reporter phage:comparison with the mycobacteria growth indicator tube(MGIT) system. Diagn Microbiol Infect Dis 2003; 45 (1):5361.

29. Butt T, Ahmad RN, Afzal RK, Mahmood A, Anwar M. Rapiddetection of rifampicin susceptibility of Mycobacteriumtuberculosis in sputum specimens by mycobacteriophageassay. J Pak Med Assoc 2004; 54 (7):37982.

30. Eltringham IJ, Drobniewski FA, Mangan JA, Butcher PD,Wilson SM. Evaluation of reverse transcription-PCR and abacteriophage-based assay far rapid phenotypic detection ofrifampin resistance in clinical isolates of Mycobacteriumtuberculosis . J Clin Microbiol 1999;37 (11):35247.

31. Gali N, Dominguez J, Blanco S, Prat C, Quesada AD, Matas L,et al. Utility of an in-house mycobacteriophage-based assayfor rapid detection of rifampin resistance in Mycobacteriumtuberculosis clinical isolates. J Clin Microbiol 2003;41 (6):26479.

32. Hazbon MH, Guarin N, Ferro BE, Rodriguez AL, Labrada LA,Tovar R, et al. Photographic and luminometric detection ofluciferase reporter phages for drug susceptibility testing ofclinical Mycobacterium tuberculosis isolates. J Clin Micro-biol 2003; 41 (10):48659.

M. Pai et al.186


13/13

33. Kisa O, Albay A, Bedir O, Baylan O, Doganci L. Evaluation ofFASTPlaquetb-RIF for determination of rifampicin resistancein Mycobacterium tuberculosis complex isolates. Int J Tuberc Lung Dis 2003;7 (3):2848.

34. Krishnamurthy A, Rodrigues C, Mehta AP. Rapid detection ofrifampicin resistance in M. tuberculosis by phage assay. Ind J Med Microbiol 2002;20 (4):2114.

35. Lu B, Fu Z, Xu S. Rapid rifampicin susceptibility test by usingrecombinantmycobacteriophages. ZhonghuaJie HeHe Hu XiZa Zhi 2000; 23 (8):4804.

36. Mani C, Selvakumar N, Kumar V, Narayanan S, Narayanan PR.Comparison of DNA sequencing, PCR-SSCP and PhaB assayswith indirect sensitivity testing for detection of rifampicinresistance in Mycobacterium tuberculosis . Int J Tuberc LungDis 2003;7 (7):6529.

37. McNerney R, Kiepiela P, Bishop KS, Nye PM, Stoker NG. Rapidscreening of Mycobacterium tuberculosis for susceptibilityto rifampicin and streptomycin. Int J Tuberc Lung Dis 2000;4(1):6975.

38. Oguz VA, Guneri S, Erdenizmenli M, Yapar N, Kuruuzum Z,Cakmak R, et al. Investigation of rifampin resistance in

Mycobacterium tuberculosis by two different methods.Toraks Dergisi 2002; 3(2):1737.

39. Riska PF, Su Y, Bardarov S, Freundlich L, Sarkis G, Hatfull G,et al. Rapid lm-based determination of antibiotic suscep-tibilities of Mycobacterium tuberculosis strains by using aluciferase reporter phage and the Bronx box. J Clin Microbiol1999; 37 (4):11449.

40. Simboli N, Takiff H, McNerney R, Lopez B, Martin A,Palomino JC, et al. In-house phage amplication assay is asound alternative for detecting rifampin-resistant Mycobac-terium tuberculosis in low-resource settings. Antimicrob Agents Chemother 2005; 49 (1):4257.

41. Banaiee N, January V, Barthus C, Lambride M, RoDiti D,Behr MA, et al. Semi-automation and evaluation of luciferase reporter mycobacteriophages for susceptibility testing of Mycobacterium tuberculosis cultures in SouthAfrica. American Society for Microbiology 104th generalmeeting 2004, New Orleans, LA 2004 [Abstract U-018].

42. Takiff H, Heifets L. In search of rapid diagnosis and drug-resistance detection tools: is the FASTPlaqueTB test theanswer? Int J Tuberc Lung Dis 2002; 6(7):5601.


Bacteriophage Assay

Documents

Transcript of Bacteriophage Assay