RESEARCH ARTICLE Open Access
Assessment and comparative analysis of a rapiddiagnostic test (Tubex
®
) for the diagnosis of typhoid fever among hospitalized children inrural Tanzania
Benedikt Ley
1,8*
, Kamala Thriemer
1
, Shaali M Ame
2
, George M Mtove
3,4
, Lorenz von Seidlein
1,4,5
, Ben Amos
4,6
,Ilse CE Hendriksen
4,7
, Abraham Mwambuli
4
, Aikande Shoo
4,6
, Deok R Kim
1
, Leon R Ochiai
1
, Michael Favorov
1
,John D Clemens
1
, Harald Wilfing
8
, Jacqueline L Deen
1,4
and Said M Ali
2
Abstract
Background:
 Typhoid fever remains a significant health problem in many developing countries. A rapid test with aperformance comparable to that of blood culture would be highly useful. A rapid diagnostic test for typhoid fever, Tubex
®
, is commercially available that uses particle separation to detect immunoglobulin M directed towards
Salmonella
 Typhi O9 lipopolysaccharide in sera.
Methods:
 We assessed the sensitivity and specificity of the Tubex test among Tanzanian children hospitalized withfebrile illness using blood culture as gold standard. Evaluation was done considering blood culture confirmed
 S
. Typhi with non-typhi salmonella (NTS) and non - salmonella isolates as controls as well as with non-salmonellaisolates only.
Results:
 Of 139 samples tested with Tubex, 33 were positive for
 S
. Typhi in blood culture, 49 were culture-confirmed NTS infections, and 57 were other non-salmonella infections. Thirteen hemolyzed samples wereexcluded. Using all non -
 S
. Typhi isolates as controls, we showed a sensitivity of 79% and a specificity of 89%.When the analysis was repeated excluding NTS from the pool of controls we showed a sensitivity of 79% and aspecificity of 97%. There was no significant difference in the test performance using the two different controlgroups (p > 0.05).
Conclusion:
 This first evaluation of the Tubex test in an African setting showed a similar performance to thoseseen in some Asian settings. Comparison with the earlier results of a Widal test using the same samples showedno significant difference (p > 0.05) for any of the performance indicators, irrespective of the applied control group.
Keywords:
 Salmonella, Tubex®, Widal, Africa, Rapid Diagnostic Test
Background
Typhoid fever remains a significant health problem inmany developing countries. Estimates suggest an incidencerate of more than 21.5 million cases globally in the year2000 [1]. Recent data from Tanzania mainland have founda strong variation of prevalence rates among blood culturepositive isolates collected in local hospitals, ranging from9% [2] to 21.4% [3] for
 Salmonella
 enterica serovar Typhi(
. Typhi), no data from Zanzibar are available to date. Asthe clinical picture of typhoid fever is often unspecific,misdiagnosis and insufficient or inadequate treatment arepotential risks associated with the disease. In the absenceof difficult-to-obtain bone marrow specimens, microbiolo-gic culture of a blood sample is considered to be the cur-rent state-of-the art test for the diagnosis of typhoid fevereven though its sensitivity may be as low as 40% [4,5]. Cul- ture may take up to seven days and requires a well-runand equipped laboratory, which is often not available in
* Correspondence: bnley@ivi.int
1
 Translational Research Division, International Vaccine Institute, Seoul, KoreaFull list of author information is available at the end of the article
Ley
 et al 
.
 BMC Infectious Diseases
 2011,
 11
:147http://www.biomedcentral.com/1471-2334/11/147
© 2011 Ley et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative CommonsAttribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction inany medium, provided the srcinal work is properly cited.
 
settings with endemic typhoid fever. The widely in useWidal test provides a cost efficient alternative [6] for sero-logical diagnosis, however its performance remains unsa-tisfying with sensitivity reported from Tanzania of 75%using blood culture as the gold standard and applying acut off titer of 1:80 [7]. The test further requires the estab-lishment of a local cut off titer prior to use which is com-plicated. Therefore, a rapid test with a performancecomparable to that of blood culture would be desirable.A rapid diagnostic test for typhoid fever, Tubex
®
iscommercially available that uses particle separation todetect immunoglobulin M (IgM) directed towards
 Sal-monella
 enterica serovar Typhi (
. Typhi) O9 lipopoly-saccharide in patient sera. Performance of the test haspreviously been evaluated in a number of studies in Asiabut none in Africa. Using blood culture results for com-parison, we assessed the sensitivity and specificity of theTubex test among Tanzanian children hospitalized withfebrile illness and compared our results with those fromprevious studies.
Methods
For evaluation of the Tubex test, we used a selectedsubset of serum samples that was obtained for a feversurveillance study [2] from Teule Hospital in MuhezaDistrict, Tanzania. In order to accommodate therequired sample size for the test validation, we includedrandomly selected and age-matched
 Salmonella enterica
serotype Typhi (
. Typhi) positive serum samples from asecond fever surveillance study conducted at ChakeChake Hospital in Pemba, Zanzibar. All samples werecollected from children between the ages of 2 months to14 years from 2008 to 2009.At Teule Hospital in Muheza, sera and blood was col-lected for culture from children with a history of threedays of fever, or a history of less than three days of fever but with at least one of the following severity cri-teria: respiratory distress; deep breathing; respiratory distress in combination with severe pallor; prostration;capillary refill
 
3 seconds; temperature gradient; systolicblood pressure <70 mm Hg; coma defined by Glasgow Coma Scale (GCS)
 
 10 or Blantyre Coma Scale (BCS)
 2; severe jaundice; history of two or more convulsionsin the last 24 hours; blood glucose <3 mmol associatedwith clinical signs; neck stiffness; bulging fontanel; oroxygen saturation <90% [2].At Chake Chake Hospital in Pemba, sera and bloodwas collected for culture from children with a recordedbody temperature of >37.5°C for outpatients and any history of fever for inpatients. Duration of fever was notconsidered for study recruitment.About 3 to 5 milliliters (ml) of blood (depending onbody weight) was collected and inoculated in a BactA-LERT
 Pediatric-fan bottle (Teule Hospital) or aBacTec Peds PLUS
/F bottle (Chake Chake Hospital)and incubated in the respective machine (BacT/ALERT3D or BacTec 9050). Bacterial growth was evaluated fol-lowing standard procedures.The Tubex
®
test (IDL - Sweden) was conductedaccording to the manufacturer
s instructions, which areas follows. Forty-five microliters (
μ
l) of antigen coveredparticles were added to the Tubex Reaction Well Stripand 45
 μ
l of non-hemolyzed serum was added. Aftertwo minutes of incubation time, 90
 μ
l of magnetic anti-body coated solution was added, and the strip wassealed and shaken for two minutes. The strip was thenplaced on a magnetic tray for five minutes, separatingthe particles if a positive sample had been added. Theresulting color change of the solution was read andcategorized on a scale from 0 to 10. The results wereinterpreted as positive for scores of 4 or greater and asnegative for scores of 2 or below as per the manufac-turer
s instructions. Samples with a color correspondingto the value of 3 were interpreted as indeterminate. Allblood culture isolates from individuals that matched theinclusion criteria and that were not considered a con-taminant were included in the analysis.We performed the Tubex test on non-hemolyzedserum samples from the patients of the two surveillancestudies who had blood culture-confirmed
 
. Typhi(defined as group 1), randomly selected cases of non-Typhi serotypes of 
 S. enterica
 (NTS) (defined as group2), and randomly selected cases with other (non-
Salmo-nellae
) pathogenic bacteria (defined as group 3). Staff members performing the Tubex test were blinded to theblood culture results.For the analysis, sensitivity (true-positive rate) wasdefined as the probability that the Tubex test result willbe positive when there is blood culture-confirmedtyphoid fever (group 1) and specificity (true-negativerate) was defined as the probability that the Tubex
®
testresult will be negative when
 
. Typhi is not isolatedfrom blood culture (groups 2 and 3). We conducted asecondary analysis using only group 3 as the controlgroup. Comparison of test performance using differentcontrol groups was done using the Yates Chi-SquareTest corrected for continuity.We conducted a literature review in order to compareour findings with those from previous studies. Weincluded studies of the Tubex test, which were identifiedby directly searching the MEDLINE database throughPubMed. All articles since the first publication of thetest [8] were included. We also conducted a supplemen-tary search of references in retrieved articles. Abstractswere reviewed, and if relevant, the article was included.A comparison of performance of the Tubex
®
test withearlier published Widal test results obtained from thesame samples was done using McNemar
s Test for
Ley
 et al 
.
 BMC Infectious Diseases
 2011,
 11
:147http://www.biomedcentral.com/1471-2334/11/147Page 2 of 6
 
Correlated Proportions http://faculty.vassar.edu/lowry/VassarStats.html.The fever surveillance studies at Chake-Chake andTeule Hospitals were approved by their respective localethical review boards (Tanzania and Zanzibar), as wellas by the International Vaccine Institute
s InstitutionalReview Board. Written informed consent was obtainedfrom legal guardians of all participants prior to any sam-ple or data collection.
Results
A total of 139 samples were tested with Tubex. Thirty-three were found positive for
 
. Typhi in blood culture(group 1), 49 were culture-confirmed non-
. Typhi(NTS) infection (group 2), and 57 were other non-
Salmo-nella
 infections that were not contaminants (group 3).Thirteen hemolyzed samples were excluded (Figure 1).Of the 33 blood culture-positive
 
. Typhi cases, 26had a positive Tubex result and were considered as truepositives. Of the 106 blood culture confirmed NTS andnon-salmonella cases (groups 2 and 3), 94 yielded anegative Tubex result and were considered as true nega-tives. Considering only the 57 non-
Salmonella
 cases(group 3) as controls, resulted in 54 true negative cases.Using groups 2 and 3 as controls showed a sensitivity of 79% and a specificity of 89% (Table 1). The sameanalysis was repeated excluding NTS from the pool of controls and showed 79% and 97% for sensitivity andspecificity, respectively. There was no significant differ-ence in the test performance using the two differentcontrol groups (all were p > 0.05 using the Chi squaretest).A total of 14 articles were retrieved and evaluated forinclusion into the review. All of the reported studieswere performed in Asia; none in Africa. A total of sixarticles were excluded: two evaluated the test for non-typhoidal
 Salmonella
 [9] or S. Paratyphi [10], three did not evaluate the sensitivity and specificity of the test[11-13], and one was a letter to the editor [14]. Thus, eight publications were included in the review (Table 2).
 
*Age-matched, randomly selected **Randomly selected *** 17 from Zanzibar and 16 from Teule
+
 Other bacterial growth included: 11x
Escherichia coli 
, 10x
Haemophilus influenzae
 type B, 8x
Streptococcus pneumoniae
, 6x
Pseudomonas
spp., 5x
 Acinetobacter baumannii 
, 3x
Streptococcus
 beta
 –
 hemolytic group A, 3x
Haemophilus influenzae
, 2x
Staphylcoccus aureus
, 1x
Burkholderia cepacia
, 1x
Streptococcus
 beta-hemolytic group C , 1x
Campylobacter
spp., 1x
Pateurella multocida
, 1x gram negative rods(not identified further), 1x
Citrobacter braakii 
, 1x
Stenotrophomonas
, 1x
Haemophilus parainfluenzae
, 1x
Streptococcus
beta - hemolytic
 
189 serum samples with positive blood culture results from Teule 13 excluded (hemolyzed) 33
S. Typhi 
*** => group 1 57 other bacterial growth included** => group 3 49 Nontyphi => group 2 54 other bacterial growth excluded** 17 serum samples with positive blood culture results for
Salmonella
 Typhi from Pemba, Zanzibar*
Figure 1
 Specimen assembly
.
Table 1 Performance of Tubex 
®
using group 1 as truepositives and two different control groups as truenegatives
Control GroupGroup 2 + 3* Group 3*Sensitivity (95% CI)
 0.79 (0.52-0.81) 0.79 (0.62-0.90)
(absolute numbers)
 (26/33) (26/33)
Specificity (95% CI)
 0.89 (0.81-0.94) 0.97 (0.85-0.99)
(absolute numbers)
 (94/106) (94/97)
*Group 1 =
 S 
. Typhi (n = 33), Group 2 = all non-yphi Salmonella (n = 49),Group 3 = all blood culture-positive non-
Salmonella
 cases (n = 57).
Ley
 et al 
.
 BMC Infectious Diseases
 2011,
 11
:147http://www.biomedcentral.com/1471-2334/11/147Page 3 of 6
 
Five of the included articles reported findings of testperformance that were similar to our results [15-19]. Two publications showed considerably lower sensitivity and specificity [20,6], and one reported higher values [8] (Table 2).
Discussion
We found Tubex has a sensitivity of 79% using eithercontrol group (95%CI: 52-81% for groups 2 and 3, and62-90% for group 3 only) and a specificity of 89-97%(95%CI: 81-94% for groups 2 and 3 and 85-99% for
Table 2 Comparison of the performance of the Tubex 
®
test from published reports
Author Year Journal SampleSizeLocation Tubex
®
cut off Sens Spec True neg.definitionReader GoldstandardStudypopulation
Ley, B. etal Thispaper ThisJournal139 Tanzania >4 79% 89% All non-typhibacteriamiaInvestigator Blood Culture(BACTEC)>2 months +>37.5°(inpatients) & history of fever(outpatients)88 79% 97% All non-salmonellabacteriamiaNaheed,A. et al2008 DiagnMicrobiolInfect Dis.867 Bangladesh
 
4 60% 58% Other confirmedbacteremiaICDDRB lab Manual BloodCultureActivesurveillance Temp
 
38°C60% 64% Blood culture neg& otherbacteremiaRahman,M. et al2007 DiagnMicrobiolInfect Dis.243 Bangladesh >4 91.2% 82.3% Other febrilepatientsICDDRB lab,min. 2independentlab techsManual BloodCultureOutpatients,all ages withhistory of feverNo.Pos89.5% Healthy subjects HealthysubjectsDong, B.et al2007 Epidemiol.Infect.1732 China
 
2 100% 43% Paratyphoid cases - Blood culture(BACTEC)Age 5-60with reportedhistory of fever for 3days
4 69% 95%
6 62% 95%
8 23% 100%
10 15% 100%Kawano,R. L. et al2006 JCM. 177 Philippines
 
2 94.7% 80.4% Blood cultureneg.n/A Manual BloodCulture&BACTECClinicallysuspectedtyphoid casesDutta, S.et al2006 DiagnMicrobiolInfect Dis.495 India
 
4 56% 88% Paratyphoid andmalaria casesn/A Blood CultureBACTECOutpatients,all ages, Patwith historyof fever for 3daysOhlsen,S. J. et al2004 JCM. 79 Vietnam Accordingtoprotocol78% 94% Other lab-confirmed febrileillnessesn/A Manual BloodCulture/ BACTECPat
 
3 yearand history of 
4 day feverHouse,D. et al2001 JCM. 127 Vietnam >2 87% 76% Febrilehospitalizedpatientslabtech Culture Children andadultsLim et al 1998 JCM. 105 HongKong & Malaysia>2 100% 100% Healthyindividuals andpat with otherbacterial diseasesand autoimmunediseaselabtechs Cultureconfirmed(56% of pos.),clinical picture,various othertestsClinicalpicture,cultureconfirmed,
Ley
 et al 
.
 BMC Infectious Diseases
 2011,
 11
:147http://www.biomedcentral.com/1471-2334/11/147Page 4 of 6
 
group 3 only) irrespective of control group. To ourknowledge, this is the first evaluation of the test in anAfrican population. Our results were similar to thoseobserved in previous studies (five out of eight studies) inAsia assessing the performance of the test [15-19], though Kawano
 et al.
 [17] and House
 et al.
 [19] used alower cut-off titer than is recommended in the manual.In contrast, two studies [20,6] found the performance of  Tubex to be poorer than our findings, despite using asimilar cut-off value, gold standard, and inclusion cri-teria. The extremely good performance of Tubexobserved by Lim
 et al.
 [8] has not been reproducedsince.An important limitation of this study is that the seraare combined from two different patient populationsand the purposeful selection of samples included in thethree groups. During the preparation of the study, wecalculated the sample sizes of true positive sera and truenegative sera that are required for validation of theTubex test and for comparison with the Widal test per-formance. The number of true positive sera from eitherhospital alone was insufficient for the validation. Thus,we included
 
. Typhi blood culture-confirmed sera fromPemba. Analysis of the results by hospital was not possi-ble because of insufficient sample size.In a sub-analysis in assessing cross reactivity withNTS, blood culture-confirmed NTS cases were consid-ered as true positives, and all other positive isolates,excluding
 
. Typhi, were considered as true negatives. Inthis sub-analysis Tubex had a sensitivity of 18% and aspecificity of 95% (analysis not shown).Comparison with a Widal test that was earlier con-ducted using the same samples [7] revealed no signifi-cant difference (p > 0.05) for any of the performanceindicators, irrespective of the applied control group. Butcompared to the Widal test, Tubex is easier and quickerto perform. The Widal test requires 16 - 20 hours untilthe results are obtained while the complete procedurefor the Tubex test is approximately 20 minutes. Tubexis more expensive at approximately 2.15 USD per testcompared to <0.80 USD per test for the Widal tubeagglutination test [6].Interpreting the Tubex test results was found to bedifficult and the results were prone to inter - reader var-iation. Assessing the color change according to the pro- vided color scale requires experience and standardizedgood lighting conditions. The Tubex test can only beapplied to non-hemolyzed and non-icteric serum sam-ples, thus limiting its general application. However Tam
et al.
 [12] have described a method that includes awashing step and thereby addresses the problem of tur-bid serum. This method requires double the amount of antigen-coated particles as well as glycine buffered saline(GBS), thereby increasing the price per sample toapproximately 4.50 USD and reducing its feasibility asan easy-to-perform test. While neither of the tests canbe performed by untrained staff, interpretation of resultsis considered easier for the Widal test compared toTubex.
Conclusion
The advantages of Tubex over the Widal test and thegold standard of blood culture is the short time itrequires to obtain a result, and it does not require estab-lishing a local cut-off value as with the Widal. In set-tings that can afford the relatively high cost of Tubexand that require instant individual diagnoses to supportthe clinical diagnosis of typhoid fever, Tubex is superiorto the Widal tube agglutination test. For screening andsurveillance purposes, as well as in settings with limitedfinancial and technical resources, the Widal tube agglu-tination test is a possible alternative that can provide asimilar performance as Tubex at a lower cost though itrequires more time. Our evaluation of Tubex showedthat any result must be handled with precaution. Resultsshould be considered as indicative, not confirmatory.The test may be used to exclude disease though. In con-clusion, the need for a reliable, fast, cheap, and easy-to-apply rapid diagnostic test for typhoid fever remains inhigh demand.
Acknowledgements
We thank Hugh Reyburn of the London School of Hygiene and TropicalMedicine who supported this project and provided his scientific expertise.We also thank Rajabu Malahiyo and Steven Magesa for their support. This work was supported by a grant from the Korea InternationalCooperation Agency (KOICA) and the Swedish International DevelopmentCooperation Agency (SIDA) to the International Vaccine Institute. This study is published with permission from the Director General of the Tanzanian National Institute for Medical Research, Dar-Es-Salaam. We aregrateful to the patients and their parents who made this work possible. Wethank all of the technical staff and research assistants who were involved inthe study.
Author details
1
 Translational Research Division, International Vaccine Institute, Seoul, Korea.
2
Laboratory Division, Public Health Laboratory (Pemba) - Ivo de Carneri,Chake Chake, Tanzania.
 3
Amani Centre, National Institute for MedicalResearch, Tanga, Tanzania.
 4
Joint Malaria Program, Tanga, Tanzania.
 5
AsiaPacific Malaria Elimination Network (APMEN), Menzies School of HealthResearch, Casuarina, Australia.
 6
 Teule Hospital, Muheza, Tanga, Tanzania.
7
Oxford Research Unit, Mahidol University, Bangkok, Thailand.
 8
Biocenter,University of Vienna, Vienna, Austria.
Authors
 contributions
BL performed the TUBEX test, analyzed results and wrote the manuscript, KT performed TUBEX tests, literature search and contributed to the manuscript,SMA supervised the laboratory work in Pemba, GM was in charge of theimplementation and study management in Teule, LvS provided scientificsupport to study staff and contributed to the manuscript, BA supervised thelaboratory work in Teule, ICEH was involved in the clinical care of patients,AM was in charge of data management, AS performed blood cultureprocedures, DRK performed statistical analyses, RLO provided scientificsupport to the manuscript, MF provided scientific support to the manuscript,JDC provided scientific support to the manuscript, HW provided scientificsupport to the manuscript, JLD provided major scientific support to the
Ley
 et al 
.
 BMC Infectious Diseases
 2011,
 11
:147http://www.biomedcentral.com/1471-2334/11/147Page 5 of 6
of 6