Development and testing of a rapid diagnostic test for bubonic and
Transcription
Development and testing of a rapid diagnostic test for bubonic and
ARTlCLES Development pneumonic and testing of a rapid diagnostic plague Suzanne Chanteau, E/isabeth Carnie/, Farida Nato U/a Raha/ison, La/ao Ra/afiarisoa, Jeanine Summary Methods We developed a test that used monoclonal antibodies to the F1 antigen of Yersinia pestis. Sensitivity and specificity were assessed with a range of bacterial cultures and clinical samples, and compared with findings from available ELlSA and bacteriological tests for plague. Samples from patients thought to have plague were tested with the RDTin the laboratory and by health workers in 26 pilot sites in Madagascar. Findings The RDTdetected concentrations of F1 antigen as low as 0.5 ngjmL in up to 15 min, and had a shelf life of 21 days at 60°C. Its sensitivity and specificity were both 100%. RDT detected 41.6% and 31% more positive clinical specimens than did bactel;i°logical methods and EUSA, respectively. The agreement rate between tests dane at remate centres and in the laboratory was 89.8%. With the combination of bacteriological methods and F1 ELlSA as reference standard, the positive and negative predictive values of the RDT were 90.6% and 86.7%, respectively. Interpretation Our RDTis a specific, sensitive, and reliable test that can easily be dane by health workers at the patient's bedside, for the rapid diagnosis of pneumonic and bubonic plague. This test will be of key importance for the contrai of plague in endemic countries. Lancet 2003; 361: 211-16 See Commentary page 191 ., ., I i ! \ i I Fou/on, Mahery Ratsitorahina, La/a Ratsifasoamanana, Introduction Background Plague is often fatal without prompt and appropriate treatment. It affects mainly poor and remate populations. Late diagnosis is one of the major causes of human death and spread of the disease, since it limits the effectiveness of contrai measures. We aimed to develop and assess a rapid diagnostic test (RDT)for plague. i I test for bubonic and .FI :.. Rapid diagnostic tests (RDTs) for infectious diseases are of much value in facilitation of major improvements in disease management, especially in developing countries. Some RDTs are based on immunochromatography, with conjugated gold or latex particles used to detect specific antigens from bacterial, viral, or parasitic agents.'-3 Plague, a flea-borne rodent disease that is occasionally transmitted to man,' is still prevalent in more than 20 countries, mainly in Africa.' Madagascar is one of the most active plague foci in the world.6-8 Prompt diagnosis of this fulminant disease is of key importance for control of the mortality rate and for effective controlo The diagnosis of bubonic or pneumonic plague is still based on clinical symptoms in most endemic areas. However, such diagnosis is not reliable. Biological samples are usually assessed by microscopy in regional hospitaIs, but in remote areas such tests are often not done at alI or are unreliable. Resource-poor regions do not have the technology and staff needed for bacteriological identification methods, ELISA, and PCR for diagnosis of plague during the acute phase of the disease.9,lO In such endemic foci, the development of simple, quick, and specific methods by which local health workers could test biological samples from patients with symptoms of plague, or dead rodents, is a key priority. Indeed, such a test would eliminare doubt in suspected cases of plague, facilitating the immediate treatment of patients and contacts, and implementation of compulsory flea control measures. Such action is crucial ifplague morbidity and lethality are to be reduced. Very early diagnosis of plague would algo be essential in cases of bioterrorism. Yersinia pestis is one of the bacteriological agents that could potentially be used for biological warfare.11 If transmitted in the forro of dro?lets that are t?en inhaled ~r spread by ae~osols, the bacIllus causes hlghly contaglous pneumornc plague, which is always fatal if treatment is not started within 24 h óf onset of symptoms.'2 antigen is specific to Y pestis and is present in large amounts in blood and bubo samples from Institut Pasteur de Madagascar,WHOCollaboratingCentre for Plague,Antananarivo,Madagascar(S ChanteauPhO, L RahalisonPhO,M RatsitorahinaMO);Plague National Contrai Programme,Ministry of Health,Antananarivo,Madagascar (L Ralafiarisoa,L RatsifasoamananaMO);and Institut Pasteur, WHOCollaboratingCentre for Yers/nia,Paris, France (JFoulon,E CarnielMO,F Nato PhD) infected people.'~" It is stable in tropical climates, and can be detected even after the patient has been treated for several days.15,16An RDT for plague based on FI antigen has been tested in the laboratory, and gave promising results.16 However, since this test is not readily available, another rapid test is needed for use in the plague foci of African, Asian, and North and South American countries, and in non-endemic countries, in cases of suspected imported plague or of biological att~ck. T~ere!ore, we aime~ to d~velop a sensitive FI Correspondenceto: Dr SuzanneChanteau,CERMES, POBox10887, Niamey,Niger antlgen dIPStlck, and to validare It both at the plague reference laboratory and at 26 remote gires in (e-mail: [email protected]) Madagascar. , THE LANCET. Vol361 .January18,2003. www.thelancet.com 211 ARllCLES Methods Oevelapmentaf plague ROT To develop monoclonal antibodies (Mab) against Fi antigen, we immunised rnice (Biozzi BP/2 strain) with Fi antigen that was purified as Chen and Meyer describedl7 from Y pestisstrain 185S/97 (Madagascar).The fusion and screening methods used have been described.18 67 hybridomas with strong IgG responsesagainst F1were obtained, 16 of which were cloned and characterised.For development of the plague RDT, two monoclonal antibodies (IgG1 and K chain isotype) that bound to two difIerent epitopes of the FI antigen19 were selected on the basisof their high affinity, and were purified by ammonium sulphate precipitation and fast protein liquid chromatography (Pharmacia, Bois d'Arcy, France) according to the column manufacturer's instructions. Mab B18-1 (kD <5X1Q-1ImoI) was conjugatedto gold particles as the mobile phase, and Mab G6-18 (kD=lX10~10 moI) was deposited on nitrocellulose as the capture antibody. The hybridomas Bl8-1 and G6-18 were stored at the Collection National desMicroorganismes (Paris,France). Our RDT was derived from a prototype developedby the Naval Medical Research Institute (Bethesda, Maryland, USA) and assessed in Madagascar.16 For this new test, we used a combination of B 18-1 and G6-18 produced at the Institut Pasteur, whereas the original RDT combined F1--O4-A-G1 Mab with a polyclonal rabbit antiserum, and was based on one-step, vertical-tlow immunochromatography.'oThe colloidal gold particles (40 nrn diameter) were conjugated to anti-F1 Mab B18-1 (British Biocell Intemational CardifI, UK) and lyophilised (A54Qnrn=3) onto polyester release pads (Filtreri, Ambérieu en Bugey, France). An automatic thin layer chromatography sampler (CAMAG, Muttenz, Switzerland) was used to spray the anti-F1 capture Mab G6-18 in a tine, at a concentrationof 2 ~cm, onto nitrocellulose membrane (Irnmunopore FP, Wh}tman Intemational, Chateau Giron, France); together with the contraI capture tine, which was affinity-purified goat anti-mouse IgG (ICN Biomedicals, Aurora, Orno, USA), sprayed as a liDe higher up the strip at a concentration of 0,8 Ilg/cm. Cellulose filter paper was used for the wicking pad and samplepads (Schull and Schleicher, Ecquevilly, France). The immunostrips were trimmed to a width of 5 mm and stored in a waterproof bag at 4°C, or in 5 rnL disposable plastic tubes at roam temperature at non-central sites.The test was dane in the plastic tube with a sample volume of about 200 j.l.L. After 10-15 min, a positive result appeared as two pink tines (upper controltine and lower F1-positive liDe), and a negative result as a single upper pink contraI tine. Semiquantitative analysis was possible, with the intensity afilie test tine scored from 0'5 to 4 (indicating the concentration ofF1 antigen in the test sample). The cutof! and the range of FI concentration that could be detected was measured with tenfold dilutions (50 j.l.g/rnL to 0,05 ng/rnL) of purified FI antigen. The reliablility of the RDT was assessedwith two concentrations of purified FI antigen (0'5 ng/rnL and 1 j.l.g/rnL). The shelf life of the strips in the laboratory was assessed by testing twice per week for 3 weeksafter storageat 60°C, 4°C, -20°C, and :-80°C.'0The re~iabilityof the RDT was assessedby repeanngthe test ten nInes on the samesample and in ten difIerent seriesoftests. Referencetestsand samples We used two reference standard tests for the diagnosisof plague. The first was gram staining and isolation of Y pestis, either direcrly from the patients' samples or after mouse inoculation. A specimen was judged to be confinned as 212 being plague when the culture was positive, presumptive when culture was negativebut rnicroscopywas positive, and negative when both tests were negative. Confinned and presumptive cases were notified to WHO. The second method was the immunocapture EUSA for detection of FI antigen with F1--O4-A-G1 anti-F1 Mab.16The lowest concentration of antigen detectable with this test was 2 ng/rnL. To assessthe sensitivity of the plague RDT we used 85 strains of Y pestisthat had been cultured in brain-heart broth at 37°C for 72 h. 55 of these were isolated in Madagascar from 1996 to 2000 (33 from patients with bubonic plague, four from those with pneumonic plague, seven from rats [Rattus norvegicus], three from tleas [Xenopsyllacheopis],seven from shrews [Suncus murinus], and one from a hedgehog [Setifersetosus].30 strains carne from 14 other countries (collection of the WHO Collaborating Centre for Yersinia, Institut Pasteur, Paris, France): tive of biotype Antiqua (four from Kenya, one former USSR), tive ofbiotype Medievalis (Kurdistan [four], Turkey [anel), and 20 of biotype Orientalis (Germany [two], Morocco [anel, Turkey [anel, India [anel, Narnibia [two], Vietnam [seven], Brazil [two], Argentina [anel, Senegal [anel, South Africa [anel, Congo [anel). Additionally, we used 198 clinical specimens from plague patients that had tested positive by both bacteriology and FI EUSA: 17 sputum samples,35 serum samples,43 urine samples,and 103 bubo aspirates. To assessthe specificity of the test we used 189 cultures of yersinia and other enterobacteria grown in brain-heart broth, at 37°C for 72 h. These sampleswere: 71 cultures of enteropathogenicyersinia, consisting of Y pseudotuberculosis (44 strains, serotypesI-V from 17 difIerent animal species and environrnental samples isolated in 22 difIerent countries), and Yenterocolitica(27 strains isolated in seven countries and belonging to six biotypes [lA-5] and to various serotypes); 52 non-pathogenic yersinia of various serotypesbelonging to eight speciesand isolated in various countries (Y kristenseni [nine), Y intermedia [ten], Y mollaretii [ten], Y frederiksenii[ten], Y bercovieri[ten], Y rhodei [anel, Y ruckeri [anel, and Yaldovae [anel; and 66 enterobacteriabelonging to 19 genera and 55 species (Escherichiaspp [eight], Enterobacterspp [ten], Erwinia sp [anel, Hafnia sp [anel, Serraria spp [ten], Klebsiellaspp [ten], Salmonellaspp [four], Shigellaspp [four], Levinea sp [anel, Citrobacterspp [two], Edwa~iella spp [two], Proteus spp [two], Providencia spp [two], Morganella sp [anel, Kluyvera spp [two], Budvicia sp [anel, Cedeceaspp [two], Leminorella spp [two], and Obesumbacteriumsp [one]). Additionally we used 137 clinical specimens that were negative for plague: 47 sputum samples from suspected tuberculosis patients (14 positive for Mycobacterium tuberculosis), 66 normal serum samples diluted 1/10, and 24 undiluted normal urine samples. For ethical reasons, bubo aspirateswere not taken from patients who were not thoUghtto have plague. The negative and positive predictive values (NPV and PPV, respectively)of the FI dipstick were calculated by Bayes' theorem, and 95% CI by Fleiss' quadratic approximation method (Sp=sensitivity,Se=sensitivity): Sp (l-prevalence) NPV= Sp (l-prevalence) + (l-Se)prevalence Se (prevalence) PPV= Se(prevalence)+ (l-Sp) (1- prevalence) The agreementbetween RDT and reference tests results (bacteriology or FI EUSA, and bacteriology combined THELANCET.Vol361.January 18,2003'www.thelancet.com --- ARTICLES O Plague endemic areas Pilotsites I I 200 km I Locationof 26 pilot sites inplagueendemicdistricts of Madagasc ar i wirll FI ELISA) was assessedwirll rlle percentage of concordant results and rlle Kappa coefficient wirll confidence intervals. The correlation between FI ELISA and RDT results was assessedby Kendall's correlation coefficient. Finally, we used rlle RDT to test 283 bubo aspirates,which were stored at rlle Institut Pasteur,rllat had been isolated from patients who were suspectedon clínical grounds to have plague but tested negative wirll rlle two referencestandard assays. We did PCR with any cultures of Y pestisrllat tested negative on RDT, to find out wherller cafl gene, which encodes for FI antigen, was present. The cafl gene was amplified by PCR, wirll heat-killed bacteria as templates and two oligonucleotide primers: forward 5'-CAGTTC CGTTATCGCCATTGC-3', and reverse 5/-TATTGG TTAGATACGGTTACGGT-3'. PCR amplification was done in a final volume of 50 ILL, in a PTC 100 rllermal cycler (MJ Research,Basel,Switzerland). We mixed 7 ILL of boiled bacterial suspension(target DNA), 200 ltInol of each deoxyribonucleotide triphosphate, 2 mmol MgCI. 0.1 ILmol of each primer, and 1 U of rllermostable DNÀ polymerase (Roche, Meylan, France) in the corresponding IX polymerase buffer. The gene was amplified over 25 cycles, each consisting of denaturation for 1 min at 95°C, annealing for 1 min at 55°C, and polymerisation for 2 min at 72°C; followed by one cycle at 55°C for 1 min and 72°C for 3 minoPlasmid DNA was extracted21 and digested wirll EcoRV. PCR products and digested plasmid DNA were electrophoresed in agarose gels and stained wirll ethidium bromide. THELANCET.Vol361.January 18,2003'www.thelancet.com Assessmentof ROr We undertook a prospectivestudy to assessrlle effectiveness of rlle FI dipstick during rlle 200G-Ql plague season in Madagascar, as part of rlle national plague control programme. We obtained a biological sample (bubo aspirate, sputum, or post-mortem organ puncture, as appropriate) from patients wirll suspectedplague, and rlle patients were treated wirll streptomycin. Sampleswere sent on a swabin Cary Blair agar, at room temperature,to rlle Central Laboratory at rlle Institut Pasteurde Madagascar. On arrival, rlle specimenwas washed out of rlle swab by incubation in 1 mL phosphate-buffered saline, and was tested wirll rlle two referencemerllods and wirll RDT. The mean transporttimes of rlle positive and negativesamplesto rlle Central Laboratory were compared wirll rlle Kruska11Wallis non-parametric testo To test rlle RDT in remote areas,a network of 26 pilot sites was set up from Dec 1, 2000, to Jan 25, 2001, consistingof six district hospitaIsand 20 healrll care centres in remote areas (figure). After rlle efficiency of rlle plague RDT had been assessedat rlle Central Laboratory, kitseachcomposed of one syringe,one needIe,PBS, one plague RDT in a disposabletest tube wirll silica gel, one Cary Blair tube, one swab,a user's manual, and a questionnaire-were distributed to rllese field sites. The robustness of rlle test under field conditions was assessed,based on several criteria: reliability of rlle results obtained by non-biologist staff, presentation of rlle kit, ease of manipulation and interpretation, shelf life under real field conditions, lengili and content ofrlle training programme, and intelligibility rlle illustrated manual. of The RDTs were stored at room temperature during rlle study (2o-30°C). 29 medical doctors, 19 nurses,and nine healrll workers participated in rlle assessment.They were trained onsite for 3 h as to how to obtain clinical samples and to use, read, and archive rlle dipsticks. An illustrated instruction guidebook, in French and Malagasy,was ~ven to eachcentre. The plague RDTwere wasrllen doneasked at rlletopatlent's bedside. The trained personnel send a sample of rlle specimen to rlle Central Laboratory for a dipstick test wirll rlle same batch of strips, bacteriological identification, and FI ELISA assay.The technicians who did rlle testsat rlle Central Laboratory were not awareof rlle results obtained in rlle pilot centres. From May to June, 2001 (end of rlle plague season),rlle 26 sites were visited again, RDT results were compared, and rlle registerswere checked.The healrll personnelwere askedto fil1 in an individual user's questionnaire,to provide criticisms, comments, and observations about rlle plague RDT and rlle instruction book. Role of the fundingsource The sponsorsofrlle study had no role in study design,data collection, data analysis,data interpretation, or writing of rlle reporto Results The lower detection threshold of rlle newly developed plague RDT was 0'5 ng/mL, and rlle range of FI concentrations extended from 0'5 ng/mL to 50 lLg/mL (5 log), wirllout any hook effect (in which no signal is detected for high concentrations). Results obtained were much rlle same after storageof RDT for 21 days at 60°C, 4°C, -20°C, and -80°C. The RDT had a sensitivityof 100% on fresh1yisolated Y pestis strains and positive control clínical samples: alI 55 bacterial cultures and all198 positive specimenstestedwere positive wirll rlle RDT. Positive RDT results were obtained for 28 of 30 Y pestisstrains from various continents. PCR 213 ARTlCLES Plague RDT Bacteriology Confirmed Presumptive Negative Fi ELlSA Positive Negative Comblnatlon of reference tests Positive Negative Positive Negative 279(40%) 412(60%) 151(83%) 12(80%) 31 (17%) 3 (20%) 116(24%) 378(77%) 212(100%)67 (14%) 0(0%) 412(86%) 232(87%) 34 (13%) 47 (11%) 378(89%) Total 691 182 494 212 266 425 Number (%) samples 15 479 Table1: ComparisonbetweenplagueRDT,bacteriology,and Fj. ELISAfor specimenstested at CentralLaboratory assaysdone on the two RDT -negativestrains with primers internal to the cafl gene yielded no amplification product, indicating that this gene (or the entire pFra plasmid) was deleted in these two strains. Comparison of the plasmid restriction profiles of these strains with that of wild-type strain 6/69 from Madagascarconfirmed the absenceof the restriction fragments corresponding to pFra (data not shown). The specificity of the plague RDT was algO100% for alI bacterial cultures and alI negative control clinical specimens.Of the 283 bubo aspiratesfrom suspectedplague patients thar testednegativein bacteriologyand EUSA, 256 (90.5%) algo tested negative by RDT, but 27 (9'5%) gave weak positive resultswith this test (scoreof 0'5). From Dec 1, 2000, to May 30, 2001, 671 cases of suspectedplague were declared to the national surveillance system:598 (89%) survived,61 (9%)died, 12 (2%) had an unknown outcome. 642 (96%) patientshad bubonic plague, 20 (3%) had pneumonic plague, and nine (1 %) had an unknown clipical status. In total, 691 clinical specimens were sentto the Central Laboratory in Cary Blair agar: 643 (93%) were bubo aspirates,13 (12%) sputa, and 35 (5%) post mortem lung or tiver puncture samples.The median transport time to the Central Laboratory of the 691 samples was 8 days, ranging from O to 66 days (25th percentile 5 days, 75th percentile 14 days). The mean transport time was 10.6 days (SD 9.6) for bacteriologically confirmed specimens and 12.1 days (11'9) for negative specimens (non-significant difference, Kruskall-Wallis non-parametric test). ~ After recovery of lhe samples from Cary Blair media, Y pestiswas cultured from 182 (26%) samples (classified as confinned specimens),whereas 15 (2%) were negative by culture but positive by microscopy (classified as presumptive specimens; table 1). Thus, the total number of plague patients identified by bacteriology was 197 of 691 suspected. With RDT, 279 (40%) specimens were positive for plague, 116 (42%) of which tested negative both by microscopy and culture. The concordance between lhe results of bacteriological and plague RDT results was 78,3% (K=0'52, 95% CI 0.50-0'54; moderate agreement). Therefore, use of plague RDT alone led to identification of 41.6% more cases than when only bacteriology was used (279 vs 197 positive tests). However, 31 (17%) samples tested negative with the plague RDT were positive by bacteriology. When FI EUSA was used as the reference method, results obtained with this technique and with RDT were highly concordant (concordance 90'3%, results ofthe two tests agreed for 624 of691 specimens tested)and strongly correlated (Kendall's correlation coefficient r=0.834, p<10-3).K was 0.79 (95% CI 0'77-0.81; good agreement). AlI EUSA-positive samples algo tested positive on RDT. Moreover, RDT detected 67 weak1y positive samples (score of 0'5) among lhe 479 samples that tested negative by EUSA (table 1). Therefore, use of RDT resulted in 31 % (67 of212) more positive results than were obtained with FI EUSA alone. When we used a combination of bacteriological tests and FI EUSA as lhe reference standard (table 1), 266 of the 691 specimens tested positive, whereas with RDT alone, 279 tested positive (K=0.75, 95% CI 0.73-0'77). At the 26 remote sites, 128 patients (122 survivors and six patients who died) were suspected, on clinical grounds, to have plague (123 bubonic and tive pneumonic fonns) and were treated. Results of RDTs done by health workers at lhe patient's bedside, and by skilled technicians at the Central Laboratory, on the garoespecimens (table 2) were concordant in 115 of 128 patients (90%). Four test~ done at remote sites (3%) were not valid since the controlline was absent (table 3). With bacteriology and FI EUSA as reference standards, 58 samples were identified as positive among the 128 tested, whereas RDTs done at remote sites and at lhe Central Laboratory identified 53 and 55 positive samples, respectively. Agreement between results from RDTs done at the field sites and bacteriology was moderate (K=0.62, 95% CI 0'57-0.66), but agreement was very good when RDT in the field was compared with FI EUSA (0.83, 0'78-0.87). The false negatives with RDT corresponded with samplesthat tested positive in culture after amplification in two mice (low bacterial load) but negative with FI EUSA. Five patients who were identified at remote sites as plague-infected were negative for plague with the combination of reference tests. On the assumption that the combination of lhe two reference tests is lhe gold standard (which is not lhe case), lhe calculated PPV and NPV would be 90.6% (95% CI 78.6-96.5) and 86.7% (76.4-93.1), respectively, at lhe pilot sites, and 92.7% (81.6-97.6) and 90.4% (80.7-95.7), respectively, at lhe Central Laboratory. There was 85% concordance between lhe results of the combined reference methods and those for the plague RDT in the field (K=0'76, 95% CI 0'71-0.80), and 91 % concordance between lhe combined and RDT methods at lhe Central Laboratory (0.82, 0'77-0.86). The 57 individual user's questionnaires gave a mean overall satisfaction score of 17'6 of 20 [range 14--20}. Most (51 of 57,90%) ofthe health staffthought that lhe RDT would help to reduce plague morbidity and mortality in Madagascar. Sputum obtained from 16 patients with suspected pneumonic plague were tested with lhe plague RDT as well as with lhe two reference tests for comparison (table 4): 13 sputa were tested only at lhe Central Laboratory; lhe remaining three sputum samples were algo tested at lhe field sites. The eight patients for whom the diagnosis of plague was confinned by bacteriology algo tested positive by RDT (one tested negative by ELISA). However, two patients who tested negative by bacteriological methods gave a weak positive result with lhe plague RDT, one of whom algo tested positive by EUSA. ROT,rematesites Tot~1 Positive Negative Invalid patlents RDT,CentralLaboratory Positive (n=55) 49 (92.5%%)5 (7%) 1 (25%) 55 (43%) Negative (n=73) 4 (7'5%) 66 (93%) 3 (75%) 73 (57%) Totalpatients (%) 53 71 4 128 Table2: ComparisonbetweenplagueRDTsdone at remote sites and at CentralLaboratoryon the samespecimens 214 THELANCET'Vo1361'January 18,2003'www.thelancet.com ~ AR11CLES Total patients(%) Combination Positive Negative Total RDT,remotesites RDT,CentralLaboratory Positive Negative Invalid Positive Negative 48 (83%) 5 (7%) 10 (17%) 61 (87%) O (0%) 4 (6%) 51 (88%) 4 (6%) 7 (12%) 66 (94%) 53 (41%) 71 (56%) 4 (11%) 55 (43%) 73 (57%) bacteriology/ELlSA 58 (100%) 70 (100%) 128 (100%) Table 3: Comparison between plague RDTdone at remote sites and at Central Laboratory and combination of bacteriology and Fi ELiSA Discussion We have shown that our RDT is as specific as, and at least as sensitive as, the two available standard methods. The excellent specificity of the RDT, its low detection threshold, and the higher number of positive specimens detected among samples from patients with suspected plague, suggest a greater sensitivity than bacteriology and ELISA. Thus, in the operational conditions ofthe plague control programme at the Central Laboratory, we recommend a combination of bacteriological tests and RDT for diagnosis and surveillance of plague in Madagascar. Our new test detected FI antigen at a wider range of concentrations than other tests, in 10-15 min, without prozone effect. Absence of prozone phenomenon is essential to avoid false-negative results, especially in postmortem samples or sputum samples that contain very high concentrations of FI antigen. The highest FI concentration ever noted in clinical samples was 30 J.lg/mL (S Chanteau, unpublished). The existing reference tests, which can only be done in the laboratory-the isolation of Y pestis and the detection of FI antigen by ELISA-are specific, but lack sensitivity in ali countries in which plague is endemic. The failure of these standard assays is due to the deterioration, contaminati~, or both, of specimens during their long transport time to the laboratory, antibiotic treatment of patients before sampling, and the diffusion of FI antigen in the Cary Blair transport medium. The difference between the numbers of suspected plague samples found positive by RDT and culture could algo be accounted for by diffusion ofthe FI antigen in the Cary Blair transport medium, a difficulty that is not encountered if these tests are directly done on clinical samples by health staff. Furthermore, other investigators have shown that some patients who were originally diagnosed as negative with (degree PlagueRDT of positivity) F:1 (ngjmL) ELlSA I Bacteriology Patlent'snumber 230/01 Neg Neg Neg 307/01 Neg Neg Neg 369/01 Neg Neg Neg 382/01 Neg Neg Neg 967/00 Neg Neg Neg 1058/00* Neg Neg Neg 23:1/01 Pos(0.5+) Pos(5) Neg :1052/00 Pos(0.5+) Neg Neg 969/00 Pos(0.5+) Pos(:13) Confirmed :1027/00 Pos(1+) Pos(44) Confirmed 999/00 Pos(2+) Neg Confirmed 1157/00 Pos(3+) Pos(125) Confirmed :1:157/00* Pos(3+) Pos(125) Confirmed 974/00 Pos(4+) Pos(>:125) Confirmed 1:158/00 Pos(4+) Pos(>125) Confirmed 1:158/00* Fios(4+) Pos(>125) Confirmed Totalpositive/tested 10/16 8/16 8/16 Confirmed;Y pestisisolated.Neg;negative. Pos;positive. 0.5+to 10+;score (seetext). 'Patientalgotestedat remotesites. Table 4: Diagnosis of pneumonic plague in sputum with plague RDT, Fi ELiSA, and bacteriology TOELANCET. Vol361 .January18,2003. www.thelancet.com the reference tests (and positive with the initial plague dipstick) subsequently underwent FI antibody seroconversion.16 Conversely, 116 samples that tested positive by plague RDT gave negative results in bacteriological tests, probably for the reasons given above. These reasons explain the fair concordance between RDT and bacteriology. Similarly, 67 samples that tested negative with FI ELISA gave weak positive results with RDT, probably because the detection threshold of RDT was lower than that of EilSA. Finally, 27 of the 283 clinical specimens classified as negative with the combined bacteriological and ELISA tests tested positive by RDT. Since the RDT was highly specific and had a low detection threshold, we suggest that these 27 specimens were true plague-positive specimens, and that the plague RDT is thus more sensitive than the two reference assays. Of the 30 Y pestis strains of the three biotypes and from different continents, most tested positive by RDT, indicating that this test is potentially applicable to plague foci worldwide. The two strains that tested negative proved to have total or part deletions of the cal] gene; this deficiency seems to be due to the loss of the 110 kb plasmid, probably during the long storage time of the strains in the laboratory. The invalid tests in the conditions of the remote sites (no control liDe in 3% of the specimens) were due to moistening of the dipsticks, caused by the humidity of the air. This drawback can be overcome by improvement of the RDT packaging, with waterproof bags instead of a disposable test tube. We expect the shelflife ofthe RDT in such conditions to be about 2 years at room temperature, on the basis of stability after 21 days at 60°C.21 The test kit should therefore tolerate storage at room temperature for long periods in plague endemic zones. The other discordant results were due either to weak positive bands that were not seen in the field, or to a mlsmterpretatlon ofthe control and pOSltlve lIDes. ThlS difficulty could be solved by a second training session focusing on interpretation of the test results. ... Dunng non-central evaluatlon of the RDT, thlS test was applied to the diagnosis of pneumonic plague as well as bubonic plague. Although sputum contains high concentrations of FI antigen, we thought that its density and viscosity might inhibit antigen-antibody interactions . and obstruct the flow of lmmune complexes through the dipstick. This difficulty was overcome by diluting sputum samples with saline or PBS (1 in 2 to 1 in 10 dilution). Since the sensitivity and specificity of the test with samples from patients with confirmed and suspected . 1 h. pneumornc pague was ~gh, the RDT seems ~o b~ a useful method for rapld, easy, and non-mvaSlve confirmation of this disease. The availability of a reliable onsite diagnostic test should result in rapid treatment of patients and efficient administration of chemoprophylaxis to the contact population; these measures should, in turn, help to prevent the spread of pneumonic plague in endemic countries or in cases of bioterrorist use. 215 ARTlCLES Assessment of the RDT results and of the users' questionnaires showed that the technical transfer of the test to remote pilot sites was successful and allowed the reliable diagnosis of bubonic and pneumonic plague. Th find' I d I fr om hea Ith sta ff ese mgs e to a genera request k I fi fi li . b d . 1 bl wor mg m pague OCl or lS test to e ma e aval a e at their centres. Bacteriological and EUSA tests are between .. . References 1 Cole RA, Lu HM, Shi YZ, WangJ, De-Hua T, Zhou AT. Clinical evaluationof a rapid immunochromatographicassaybasedon me 38kDa antigen of Mycobacteriumtuberculosis on patients wim pulmonary tuberculosisin China. TuberLungDis 1996,77: 363--{í8. 2 W iIGJ Lammi PT W . N Th ICTFiI ..' d" e, e J' elSS .e antigen test for diagnosis ofbancroftian 13: 401-04. anaS1Stest: a rapl .onnat filariasis. Parasito! Today 1997 ' tive and 50 times more expensive than the RDT, and they 3 Makler MT, Palmer C], Ager AL. A reviewofpractical techniquesfor cannot be used in remQte locations. me diagnosisof malaria. Annals TropMed Parasito11998; 92: 419-33. Overall, we have sh~wn that with our test, the rapid and 4 Per;ry~' r:ertherstonlD. Yersiniapestis, etiologic agentofplague. ffi d fb b d I ChnMlCrobiolRev1997; 10: 35-66. .. cost-e ecttve lagnosls o U ornc an pneumornc P ague 5 Th WorId Heal-LUl OrganlZanon. Out break news. 'UTL' ..e w"'y E","Á-' "~ff'W I ReC could easlly be achleVed by health workers in remote sites. 2000, 75: 337-44. Use of the test could help to reduce mortality (through 6 ChanteauS, Ratsifasoamanana L, RasoamananaB, et ai. Plague,a rapid and efficient treatment of patients), morbidity (by reemergingdiseasein Madagascar.EmerglnfectDis 1998,4: 101-04. rapid implementation of preventive measures), and 7 ChanteauS, Rats~torahinaM, RahalisonL, et ai. Current epidemiology insecticide resistance of fleas (through rational use of ofh~an plagu: m Madagascar.MlCrobeslnfect2000; 2: 25-~1. " .. d ) Th ki 11b d b d I 8 Bolsler P, RahallsonL, RasolomaharoM, et ai. Four succeSSlve armual expe~slv~ msecttcl es .lS t Wl e lstn u~e to a I outbreaksofbubonic plaguein me coastal city of Mahajanga the distnct and health care centres of the endemlc plague (Madagascar):epidemiologicalfeatures.EmerglnfectDis 2002, 8: areas ofMadagascar in 2002. The next step will be, with 311-16. the assistance of international organisations to make the 9 Williams JE, Arntzen L, Tyndal GL, lsaacsonM. Application of enzyme I RDT .1 bl th .' . th d .immunoassays for me confirmation of clinically suspectplaguein p ague ~val a e to o er countnes Wl en em1C Namibia 1982.Bull WorldHealth Organ 1986,64: 745-52. plague worldWlde. 10 RahalisonL, Vololonirina E, RatsitorahinaM, ChanteauS. Diagnosisof . .. . . . . .. Contrioutors S Chanteau designedme study, developedanti-F1 Mabs and RDT, planned me user's guide book, trained healm staff in me field, analysed data, and wrote me manuscript. L Rahalison supervisedassaysat Central Laboratory, trained healm staffin me field, and made me user's guide book (French and Ma)agasylanguages). L Ralafiarisoaproduced RDT kits, trained healm staffin me field, and did RDT and EUSA tests at me Central Laboratory.] Foulon assessedRDT wim bacterial cultures and did PCR. L Ratsifasoamananacoordinated me plague national contrai programme and me network ofhealm centres. E Carniel designedme study, supervised o.fme assessment~f RDT wim b~cterial cultures, and wro,teme man~scnpt. F Nato coordmated me In~ntut Pasteur.RDT pro)ect, supervlsedme ~evelopment and producnon afilie ann-F1 Mabs, and wrote me manuscnpt. Confli~ of interest statement None declared. Acknowledgments We mank me US Naval Medical ResearchInstitute (Bemesda-MDUSA) which sharedwim us meir experiencein me development ofRDTs and provided me immunocapture EUSA assayfor evaluation at me InstituI Pasteurde Madagascar. We mank]ean Luc Guesdon for me purification by FPLC afilie monoclonal antibodies, me technicians afilie Central Plague Laboratory, and me healm staffin me pilot sitesfor meir contribution to me validation afilie test under field conditions. This investigation was funded by me InstituI Pasteur, Paris (Grant PTR 2000-11), InstituI Pasteur de Madagascar, and me MinisIry ofHealm of Madagascar (World Bank IDA 3302 MAG). bubonic plagueby PCR in Madagascar.J Clin Microbio12000;38: 260-63. 11 Bellamy R], Friedlander AR Bioterrorism. QJM 2001, 94: 227-34. 12 RatsitorahinaM, ChanteauS, RabalisonL, Ratsifasoamanana L, Boisier P. Epidemiological and diagnosticaspectsafilie outbreak ofpneumonic plaguein Madagascar.Lancet2000; 355: 111-13. 13 Baker EE, Sommer H, Foster LE, Meyer E, Meyer KF. The isolation and characterisationafilie soluble antigen of PasteureUa pestis.lnfect lmmun 1952,68: 131-45. 14 BrubakerRR. Interconversionof purine mononucleotidesin Pasteurella pestis.lnfect lmmun 1970, 1: 446-54. 15 ChanteauS, RabarijaonaL, Hager ], Boisier P, Burans], Rasolomaharo M. F1 antigenemiain bubonic plaguepatients, a marker of gravity and efficacyoftreatrnent. TransR Soc TropMed Hyg 1998,92: 572-73. 16 Ch S Rah li L Ra . ahin M ai E I di . anteau, a son, tsltor a , et .ar y agnoslsof bubonic plagueusingF1 antigen capture EUSA assayand rapid immunogold dipstick. IntJ Med Microbio12000;290: 279-83. 17 Chen TH, Meyer KF. An evaluationof Pasteurella pestisfraction l-specific antibody for me confinnation of plagueinfection. Bull WorldHealth Organ1966; 34: 911-18. 18 Ko?ler G, Milstein C. Con~~ous cultures offused cells secreting annbody ofpredefined specificlty. Nature 1975,265: 495-97. 19 Phalipon A, Arondel], Nato F, Rouyre S, Mazie ]C, SansonettiP]. ldentification and characterizationofB-cell epitopes afilie IPA C invasin of Shigeilaflexneri.lnfectlmmun 1992,60: 1919-26. 20 PaekSH, Lee SH, Cho]H, Kim YS. Development ofrapid one-step immuno-chromatographicassay.Methods2000; 22: 53--{í0. 21 Bimboim HC, Doly]. A rapid alka1ineextraction procedure for screeningrecombinant plasmidDNA. NucleicAcids Res1979,7: 1513-23. .j 216 THE LANCET. Vol361 .]anuary 18,2003. www.melancet.com