improving evaluations of t-cell assays for diagnosing active mycobacterium tuberculosis infection
TRANSCRIPT
252 Letters to the Editor
patients underwent MRI brain imaging, 4/5 had normalstudies and 1 patient had periventricular white matterchanges (JC and BK virus negative).
All of the patients improved clinically from startingantiretroviral therapy or switching from their original com-bination to a regime with a higher CNS penetration score.3
CSF analysis for HIV-1 RNA is not frequently performed inmost centres. We advise that any patient who presents witheven vague neurological symptoms should have a lumbarpuncture to exclude discordant CSF and plasma viraemia.
References
1. McArthur JC, McClernon DR, Cronin MF, Nance-Sproson TE,Saah AJ, St Clair M, et al. Relationship between human immuno-deficiency virus-associated dementia and viral load in cerebro-spinal fluid and brain. Ann Neurol 1997;42:689e98.
2. Garvey LJ, Everitt A, Winston A, Mackie NE, Benzie A. Detect-able cerebrospinal fluid HIV RNA with associated neurologicaldeficits, despite suppression of HIV replication in the plasmacompartment. AIDS 2009;23(11):1443e4.
3. Letendre S, Marquie-Beck J, Capparelli E, Best B, Clifford D,Collier AC, et al. Validation of the CNS penetration-effective-ness rank for quantifying antiretroviral penetration into thecentral nervous system. Arch Neurol 2008;65:65e70.
A. Scourfield*L. WatersM. Nelson
Department of HIV medicine, Chelsea and WestminsterHospital, 1st Floor St Stephens Centre, 369 Fulham Road,
London SW10 9NH, UK*Corresponding author. Tel.: þ44 020 846 6322.
E-mail address: [email protected](A. Scourfield)
15 December 2009
Available online 23 December 2009
ª 2009 The British Infection Society. Published by Elsevier Ltd. Allrights reserved.doi:10.1016/j.jinf.2009.12.009
Improving evaluations of T-cell assays for diagnosingactive Mycobacterium tuberculosis infection
KEYWORDSMycobacterium;Tuberculosis;Evaluation;T-cell
than TB, acute presentation; compared to some whohave had a high pretest probability (90%), e.g. HIV infe
Evaluations of T-cell assays for diagnosing activeMycobacterium tuberculosis (TB) infection were recentlypublished in the Journal of Infection by Liao et al., Kim
et al. and Clark et al. (December 2009).1,2 These evalua-tions presented evidence of T-cell assays clinical accuracy,i.e. can the tests detect the presence and absence of activeTB infection. Unfortunately, knowledge of T-cell assaysclinical accuracy offers only moderate help in deciding ifthe tests are clinically useful in an individual patient. Tobetter understand clinical usefulness, evaluations shouldinvestigate how a patient’s management is altered by theiruse, that is: does the test help you treat the patientquicker, cheaper, or better; does it make no difference ordoes it lead to harm? At present the debate about T-cellassays for diagnosing active TB infection seems focusedon their clinical accuracy. A simple Medline search, basedon the term ELISPOT and Tuberculosis, demonstrates this.This search identified seven studies into the clinical accu-racy of T-cell assays to diagnose active TB infection.3e8
Only one reported clinical usefulness. It included just sevenpatients with active TB,9 with the T-cell assays promptingearly initiation of anti-tuberculous therapy in threepatients. This provides preliminary evidence that despitethe limitations of the T-cell assays: their clinical sensitivitybeing reported as low as 70% by Liao et al., their clinicalusefulness is worthy of investigation.
Evaluating the clinical usefulness of T-cell assays, aswith other diagnostic tests, will be a difficult area toresearch.10 Ideally, a trial would be completed composingtwo cohorts: with and without T-cell assays, with clinicaloutcomes compared. Such a study would be limited bya lack of blinding and it may be ethically difficult to with-hold a diagnostic test if the patient may benefit. The alter-native is to prospectively evaluate the impact the test hason clinical decision making and subsequent clinicaloutcomes. However evaluations are completed, they shouldinform clinicians how clinically useful a test will be for apatient under their care. Clinical usefulness of a T-cell as-say is likely to be derived from the short time the test takescompared to Mycobacterial culture and identification. Ben-efits will only be realised if a test influences clinical deci-sion making. Decision making in an individual patientshould be based on the predictive values derived froma test. Predictive values are a function of clinical accuracyand pretest probability (the chance of having a conditionbased on all factors, e.g. clinical examination and radiolog-ical findings, available before a test is completed). By usingthe evaluation of a T-cell assay by Kim et al., who evalu-ated the clinical accuracy of a T-cell assay in diagnosingabdominal TB, the importance of pretest probability onpredictive values can be demonstrated. If we considerwithin their patients the pretest probability of somepatients was low (10%), e.g. probable alternative diagnosis
mayction
with no other obvious diagnosis, we can calculate thepredictive values in these two groups (Tables 1 and 2).11
This analysis shows that in patients with a high pretestprobability, a positive result, giving a positive predictivevalue of 97%, should result in TB therapy being initiated.This is compared to those with a low pretest probabilitywhere a positive result would give a positive predictivevalue of 32%; here it would probably be inappropriate tostart TB treatment based on the T-cell assay. In both thesesituations it could be argued the tests have not materially
Table 1 Positive and negative predictive values (PPV/NPV) of the ELISPOT assay for abdominal Mycobacteriumtuberculosis (TB) where the pretest probability of infectionis low (10%).
Pretestprobability 10%
Active TB
þve (10%) �ve (n Z 90%)
T-cell assay þve 8.9 18.9 PPV Z 32%T-cell assay �ve 1.1 71.1 NPV Z 99%
Sensitivity 89% Specificity 79%
Sensitivity and specificity based of results published by Kimet al.2
Table 2 Positive and negative predictive values (PPV/NPV) of the ELISPOT assay for abdominal Mycobacteriumtuberculosis (TB) where the pretest probability of infectionis high (90%).
Pretestprobability 90%
Active TB
þve (90%) �ve (10%)
T-cell assay þve 80.1 2.1 PPV Z 97%T-cell assay �ve 9.9 7.9 NPV Z 44%
Sensitivity 89% Specificity 79%
Sensitivity and specificity based of results published by Kimet al.2
Table 3 Positive and negative predictive values (PPV/NPV) of the ELISPOT assay for abdominal Mycobacteriumtuberculosis (TB) where the pretest probability of infectionis moderate (60%).
Pretestprobability 60%
Active TB
þve (60%) �ve (40%)
T-cell assay þve 53.4 8.4 PPV Z 86%T-cell assay �ve 6.6 31.6 NPV Z 82%
Sensitivity 89% Specificity 79%
Sensitivity and specificity based of results published by Kimet al.2
Letters to the Editor 253
affected the pretest probability and would not haveaffected clinical decision making, i.e. the tests have noclinical usefulness. Where tests are often of value is wherethere is diagnostic uncertainty. In a case with a moderatepretest probability (60%), we can see the PPV would be86% (Table 3). This is likely to be sufficient to alter clinicaldecision making. The question then remains, will thisaltered decision making be of benefit or detriment to thepatients tested.
Guidelines for completing evaluations of tests intoinfectious diseases are available for those consideringundertaking such evaluations.12 In addition, the importancegiven to the evaluations of diagnostic tests is increasing andthis is recognised by the National Institute of HealthResearch in the UK who now offers specific funding into
evaluations of diagnostic tests (Diagnostic tests and testtechnologies programme).
In conclusion, researchers into TB diagnostics need todevelop transferable criteria for assigning a pretest prob-ability of active infection to patients. Estimates of clinicalaccuracy can then be used to prospectively assign patientspost test probabilities and report how this value impacts onthe clinical care of patients with suspected TB.
Funding
None.
Acknowledgements
None.
References
1. Liao CH, Chou CH, Lai CC, Huang YT, Tan CK, Hsu HL, et al.Diagnostic performance of an enzyme-linked immunospotassay for interferon-gamma in extrapulmonary tuberculosisvaries between different sites of disease. J Infect 2009;59(6):402e8.
2. Kim SH, Cho OH, Park SJ, Ye BD, Sung H, Kim MN, et al.Diagnosis of abdominal tuberculosis by T-cell-based assays onperipheral blood and peritoneal fluid mononuclear cells.J Infect 2009;59(6):409e15.
3. Liebeschuetz S, Bamber S, Ewer K, Deeks J, Pathan AA,Lalvani A. Diagnosis of tuberculosis in South African childrenwith a T-cell-based assay: a prospective cohort study. Lancet2004;364(9452):2196e203.
4. Kim SH, Song KH, Choi SJ, Kim HB, Kim NJ, Oh MD, et al. Diag-nostic usefulness of a T-cell-based assay for extrapulmonarytuberculosis in immunocompromised patients. Am J Med2009;122(2):189e95.
5. Lee LN, Chou CH, Wang JY, Hsu HL, Tsai TH, Jan IS, et al.Enzyme-linked immunospot assay for interferon-gamma inthe diagnosis of tuberculous pleurisy. Clin Microbiol Infect2009;15(2):173e9.
6. Dosanjh DP, Hinks TS, Innes JA, Deeks JJ, Pasvol G,Hackforth S, et al. Improved diagnostic evaluation of suspectedtuberculosis. Ann Intern Med 2008;148(5):325e36.
7. Kim SH, Choi SJ, Kim HB, Kim NJ, Oh MD, Choe KW. Diagnosticusefulness of a T-cell based assay for extrapulmonary tubercu-losis. Arch Intern Med 2007;167(20):2255e9.
8. Adetifa IM, Lugos MD, Hammond A, Jeffries D, Donkor D,Adegbola RA, et al. Comparison of two interferon gammarelease assays in the diagnosis of Mycobacterium tuberculo-sis infection and disease in The Gambia. BMC Infect Dis2007;7:122.
9. Gooding S, Chowdhury O, Hinks T, Richeldi L, Losi M, Ewer K,et al. Impact of a T cell-based blood test for tuberculosis infec-tion on clinical decision-making in routine practice. J Infect2007;54(3):169e74.
10. Scott IA, Greenberg PB, Poole PJ. Cautionary tales in theclinical interpretation of studies of diagnostic tests. InternMed J 2008;38(2):120e9.
11. Douglass CW. Evaluating diagnostic tests. Adv Dent Res 1993;7(2):66e9.
12. Banoo S, Bell D, Bossuyt P, Herring A, Mabey D, Poole F, et al.Evaluation of diagnostic tests for infectious diseases: generalprinciples. Nat Rev Microbiol 2006;4(Suppl. 9):S21eS31.
254 Letters to the Editor
Andrew Kirby*School of Infection & Host Defence,
University of Liverpool,8th floor Duncan Building,
Daulby Street,Liverpool L69 3GA, UK
*Tel.: þ44 151 7064396; fax: þ44 151 7065805.E-mail address: [email protected]
10 December 2009
Available online 22 December 2009
ª 2009 The British Infection Society. Published by Elsevier Ltd. Allrights reserved.doi:10.1016/j.jinf.2009.12.006
Gram stain/aolc screening for detection of catheterrelated bloodstream infection
Dear Editor,
Conventional methods for diagnosing CRBSI such as quanti-tative or semiquantitative catheter-tip cultures requireremoval of the CVC. However, CVCs removed on suspicionof CRBSI prove to be the source of blood stream infection inonly 15% of cases.1 It has been shown that clinically sus-pected CRBSI can be detected by the differential time topositivity (DTP) method or Gram/acridine-orange leukocytecytospin (AOLC) test without catheter removal.2,3 Whethermicroorganisms can be detected in blood drawn from CVCsbefore symptoms of CRBSI become manifest using theGram/AOLC method has not been investigated. The aimof this prospective pilot study was to investigate whetherthe Gram stain/AOLC method might be a potentially usefulway to anticipate the diagnosis of CRBSI.
Patients included in this study were recruited at ourbone marrow transplantation unit: patients had to (1)undergo HSCT (2 patients were prepared for but sub-sequently did not receive HSCT; both were neutropenic),(2) obtain a central venous access and (3) provide writteninformed consent (study was approved by our institutionalreview board). A catheter episode was defined as theperiod from the insertion of the catheter or access toimplanted ports to the complete removal of the CVC or portsystems. Patients were daily observed for clinical signs ofbloodstream infection (i.e. core temperature of >38.0 �Cor <36.0 �C, tachycardia> 90/min, tachypnea> 20/min,or transient hypotension).
Gram stain/AOLC screening started from the insertion ofthe catheter or access to implanted ports to the completeremoval of the CVC or port systems. The treating physicianswere unaware of screening results. For screening by Gram/AOLC test, EDTA blood samples were drawn from randomlyselected catheter lumina three times a week previous toroutine scheduled blood sampling for e.g. clinical chemistryor blood count and were processed as described previously.4
When CRBSI was suspected by the treating physicians
routine diagnostic measures were taken such as the DTPmethods and routine Gram/AOLC tests as described pre-viously.2e6 When CVCs were removed the Brun-Buissonmethod was routinely performed.6 CRBSI was defined bythe presence of signs of bloodstream infection (see above)plus a positive DTP result 2 and/or a positive routineGram/AOLC test result3 and/or a positive Brun-Buisson cul-ture result.6,7
Fifty-two patients and 60 catheter episodes were pro-spectively analyzed. The mean observation time per cath-eter episode was 31� 18 days (range 6e103 days) resultingin 1859 CVC days. Five of fifty-two patients died during theobservation time, all deaths were unrelated to CRBSI. Eightepisodes of CRBSI were identified. Six of eight episodeswere caused by coagulase negative staphylococci, one byEscherichia coli and one by Pseudomonas sp. Three casesof bloodstream infections unrelated to CVCs were detectedcaused by Candida krusei, E. coli and Staphylococcusaureus.
Local signs of infection at the catheter insertion siteoccurred in 19/60 catheter episodes. One of these 19 caseswas identified as CRBSI. In 7/8 CRBSI episodes no local signsof infection were detected. There was no statisticallysignificant difference between CRBSI cases and non-CRBSIcases in terms of local signs of infection at the catheterinsertion site (p Z 0.21).
Eight cases of CRBSI were detected in our study. Fiveof eight CRBSI cases were detected by Gram/AOLCscreening test on average 48 hours before the diagnosiswas established by routine measures (range 0e96 hoursbefore, median 48 hours). In all of these 5 cases controlcultures from screening tests yielded the same specieswith identical antibiotic susceptibility pattern as obtainedby routine microbiological tests. Three cases of CRBSIwere not detected by our screening method. All of thosewere subsequently detected by the routinely performedDTP (3 cases) and routine Gram/AOLC test (additionallypositive to DTP in 1 case). One patient with positiveGram/AOLC screening subsequently had no signs of bac-teremia, therefore had no routinely performed DTP androutine Gram/AOLC test and was considered to have noCRBSI. The sensitivity and specificity of Gram/AOLCscreening were 62.5% and 98% respectively. The PPVand NPV were 83% and 94.4% respectively.
These data suggest that patients with negative screeningresults were very unlikely to experience a CRBSI. Negativescreening results indicate that the CVC is very rarely thesource of blood stream infection in a symptomatic patientwhich could enable clinicians to focus diagnostic measuresmore efficiently on other origins of infections. Anotheradvantage of screening by Gram/AOLC test was that screen-ing results turned positive on average 48 hours before thediagnosis of CRBSI was established by routine methods.
Sensitivity however was moderate. The Gram/AOLCtest has an absolute threshold of 103e104 CFU/mLblood,3e5 whereas the DTP method has a relative thresh-old since it measures the difference in the microbial loadof peripheral and hub-blood cultures. Compared to theGram/AOLC test, DTP may therefore already indicateCRBSI when the microorganisms in the catheter bloodhave not yet reached the threshold necessary to obtainpositive Gram/AOLC test results. In addition, randomized