ORIGINAL CONTRIBUTION

Comparison of a Whole-Blood Interferon ␥

Assay With Tuberculin Skin Testing for
Detecting Latent Mycobacterium
tuberculosis Infection
Gerald H. Mazurek, MD Context Identifying persons with latent tuberculosis infection (LTBI) is crucial to the
Philip A. LoBue, MD goal of TB elimination. A whole-blood interferon ␥ (IFN-␥) assay, the Quanti-
FERON-TB test, is a promising in vitro diagnostic test for LTBI that has potential ad-
Charles L. Daley, MD
vantages over the tuberculin skin test (TST).
John Bernardo, MD Objectives To compare the IFN-␥ assay with the TST and to identify factors asso-
Alfred A. Lardizabal, MD ciated with discordance between the tests.
William R. Bishai, MD, PhD Design and Setting Prospective comparison study conducted at 5 university-
affiliated sites in the United States between March 1, 1998 and June 30, 1999.
Michael F. Iademarco, MD, MPH
Participants A total of 1226 adults (mean age, 39 years) with varying risks of My-
James S. Rothel, PhD cobacterium tuberculosis infection or documented or suspected active TB, all of whom
underwent both the IFN-␥ assay and the TST.

T
UBERCULOSIS (TB) IS THE SINGLE
Main Outcome Measure Level of agreement between the IFN-␥ assay and the
leading microbial killer of TST.
adults in the world with a death
Results Three hundred ninety participants (31.8%) had a positive TST result and 349
toll of more than 2 million per-
(28.5%) had a positive IFN-␥ assay result. Overall agreement between the IFN-␥ as-
sons per year. The World Health Or- say and the TST was 83.1% (␬=0.60). Multivariate analysis revealed that the odds of
ganization estimates that one third of having a positive TST result but negative IFN-␥ assay result were 7 times higher for
the world’s population is infected with BCG-vaccinated persons compared with unvaccinated persons. The IFN-␥ assay pro-
the causative organism, Mycobacte- vided evidence that among unvaccinated persons with a positive TST result but nega-
rium tuberculosis complex.1 While the tive IFN-␥ assay result, 21.2% were responding to mycobacteria other than M tuber-
majority of M tuberculosis infections are culosis.
kept in check by the host’s immune de- Conclusions For all study participants, as well as for those being screened for LTBI,
fenses and remain latent, some latent the IFN-␥ assay was comparable with the TST in its ability to detect LTBI, was less
infections progress to active and con- affected by BCG vaccination, discriminated responses due to nontuberculous myco-
tagious disease.2 bacteria, and avoided variability and subjectivity associated with placing and reading
The number of persons with latent the TST.
JAMA. 2001;286:1740-1747 www.jama.com
TB infection (LTBI) in the United States
is estimated to range from 10 million
to 15 million and a large number of tools with which to identify persons creased risk of subsequently develop-
cases of active TB arise from this pool with LTBI and those at greatest risk of ing, or currently having, active TB.7-9
of infected persons.3 Identifying per- developing active TB.5 However, despite its widespread use and
sons with LTBI is crucial to the goal of Until recently, skin testing with pu- a large body of data on its standardiza-
TB elimination, because the develop- rified protein derivative (PPD) of tuber- tion, the TST is subject to considerable
ment of active TB in these persons can culin was the only practical way of de-
effectively be prevented with treat- tecting latent M tuberculosis infections. Author Affiliations and Financial Disclosures are listed
at the end of this article.
ment, thereby stopping further spread In the United States, the tuberculin skin Corresponding Author and Reprints: Gerald H.
of disease.4 In recognition of this, a re- test (TST) is used as an initial screening Mazurek, MD, Division of Tuberculosis Elimination,
Centers for Disease Control and Prevention, 1600
cent Institute of Medicine report gave test for both LTBI and active TB.6 A posi- Clifton Rd, Mailstop E10, Atlanta, GA 30333 (e-mail:
high priority to the development of tive TST result is indicative of an in- gym6@cdc.gov).

1740 JAMA, October 10, 2001—Vol 286, No. 14 (Reprinted) ©2001 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a University Of Connecticut Health Center User on 05/28/2015

 INTERFERON ␥ ASSAY VS TUBERCULIN SKIN TESTING

variations and other limitations. False- giene and Public Health, Baltimore, Md; and residence or work (paid or un-
positive TST responses may result from University of California at San Fran- paid) in a health care setting, prison,
contact with environmental mycobacte- cisco; New Jersey Medical School, New- homeless shelter, drug rehabilitation
ria that share common antigens with M ark; and University of California at San unit, or other group housing. Based on
tuberculosis or may result from prior BCG Diego, using a common protocol. These responses to the questionnaire and a re-
vaccinations.7,10,11 Errors in placement sites were randomly coded as A-E in the view of available medical records, per-
and reading of the TST can also yield analysis. Ethical approval for the study sons were categorized into 4 study
false-positive results. A multitude of con- was obtained from the institutional re- groups: (1) low-risk for LTBI, sub-
ditions may blunt the response to tuber- view boards at the Centers for Disease jects receiving preemployment or pre-
culin, most notably human immunode- Control and Prevention (CDC), which school enrollment TST with no iden-
ficiency virus (HIV)–associated supported the study, and the 5 study tified risks for LTBI; (2) high-risk for
immunosuppression, but perhaps the sites prior to enrolling any subjects. All LTBI, asymptomatic subjects with risk
largest cause of erroneous TST results lies participants provided written in- of LTBI including contacts of patients
with the subjective nature of placement formed consent. with TB; persons from countries where
and reading of the test.6,12 Digit prefer- Persons recruited for the study were tuberculosis is prevalent (⬎10 cases per
ence (eg, rounding measures of TST in- 18 years or older and included per- 100000 population)26; intravenous drug
duration to the nearest multiple of 5 mm) sons requesting a preemployment or users; persons who lived, worked, or
and interpretation bias can signifi- preschool enrollment TST; persons volunteered on a regular basis in a
cantly affect TST results.13 being screened with a TST because they homeless shelter, prison, drug reha-
Discovery of the role of T lympho- were considered to be at high risk for bilitation unit, hospital, or nursing
cytes and interferon ␥ in the immune LTBI; persons in whom TB was clini- home; and persons determined to be at
process has led to the development of cally suspected and who had received increased risk by prior local investiga-
an in vitro assay for cell-mediated im- fewer than 6 weeks of anti-TB therapy; tions; (3) TB suspects, subjects being
mune reactivity to M tuberculosis.14 This and persons who previously had ac- evaluated for active TB who had re-
whole-blood interferon ␥ (IFN-␥) as- tive TB, confirmed by a positive cul- ceived fewer than 6 weeks of anti-TB
say is marketed in Australia as the ture, and who had completed a course therapy; and (4) culture-confirmed TB,
QuantiFERON-TB test (Cellestis Lim- of multidrug anti-TB therapy within the subjects who completed treatment for
ited, St Kilda, Australia) for the detec- prior 2 years. Subjects were excluded culture-confirmed TB within the prior
tion of LTBI. Like the TST, the IFN-␥ from the study if they self-reported as 2 years.
assay detects cell-mediated immunity pregnant or HIV-positive; had a his- Persons enrolled during preemploy-
to tuberculin. This IFN-␥ assay is based tory of severe reaction to tuberculin; ment or preschool enrollment exami-
on the quantification of interferon ␥ that were immunocompromised due to leu- nations were assigned to group 2 if risk
is released from sensitized lympho- kemia, lymphoma, or Hodgkin dis- factors for LTBI were identified dur-
cytes in whole blood when it is incu- ease; or had taken immunosuppres- ing questioning. However, to main-
bated overnight with PPD from M tu- sive drugs (eg, corticosteroids, tain the integrity of group 1 as truly low-
berculosis and control antigens. methotrexate, azathioprine) during the risk for LTBI, persons considered to be
Results of animal and human stud- preceding 3 months. at high-risk for LTBI at enrollment were
ies of the IFN-␥ assay conducted world- After providing written informed assigned to group 2 even when risk fac-
wide have been encouraging, but the consent, enrolled persons completed a tors were denied.
test has not been widely evaluated in detailed questionnaire about possible
the United States.15-25 Therefore, we risk factors for exposure to M tubercu- Tuberculin Skin Testing
compared the IFN-␥ assay with TST re- losis. Subjects were also asked to indi- The TST was administered by the Man-
sults from persons at 5 sites in the cate results of any prior TST, whether toux method using 0.1 mL (5 TU) of Tu-
United States with varying degrees of they had received BCG vaccination, de- bersol (Connaught Laboratories Inc,
risk for M tuberculosis infection and in tails of any contact with a person hav- Toronto, Ontario) and interpreted by
persons with documented and sus- ing TB, any risk factors associated with trained health care workers according to
pected active TB. Multivariate analy- HIV infection, and whether they had American Thoracic Society (ATS)/
sis was used to identify subject- any other medical conditions. When ap- CDC guidelines.6 Transverse indura-
related and test-related factors plicable, data were also collected from tion at the TST site was measured 48 to
associated with test discordance. medical records about findings on chest 72 hours after injection of PPD. TST re-
radiography, results and dates of cul- sults were interpreted using the risk-
METHODS tures for mycobacteria, and details of stratified interpretation of induration, as
The study was conducted at 5 sites: Bos- treatment for TB. Data were collected recommended by the ATS/CDC guide-
ton University School of Medicine, on subjects’ age, race, place of birth, lines, unless otherwise stated that the
Mass; Johns Hopkins School of Hy- residence outside of the United States, cutoff for a positive reaction was 10 mm.6
©2001 American Medical Association. All rights reserved. (Reprinted) JAMA, October 10, 2001—Vol 286, No. 14 1741

Downloaded From: http://jama.jamanetwork.com/ by a University Of Connecticut Health Center User on 05/28/2015

 INTERFERON ␥ ASSAY VS TUBERCULIN SKIN TESTING

IFN-␥ Assay terpretations were performed using an measurements recorded as 5 mm with

Blood for the IFN-␥ assay was drawn Excel spreadsheet (Microsoft Corp, the average number of measurements
into 10-mL heparinized tubes before a Redmond, Wash). recorded as 3, 4, 5, 6, and 7 mm. Digit
TST was placed. The assay was per- preference was considered significant
formed and interpreted according to the Data Collection and if the number of measurements at a par-
manufacturer’s instructions using pre- Statistical Analysis ticular value exceeded the average num-
viously described cut-points to iden- Information from the standardized ber of surrounding measurements by
tify infected persons.16,27 Specifically, questionnaire, the TST record, and the more than 50%.
within 12 hours of collection, 1-mL ali- IFN-␥ assay record were entered into
quots of heparinized whole blood were a dBASE IV spreadsheet (dBASE Inc, RESULTS
stimulated with 3 drops of the stan- Vestal, NY) using a double data entry Between March 1, 1998, and June 30,
dard antigens provided in the test kit method for verification. Statistical 1999, a total of 1471 subjects were en-
and incubated for 12 to 24 hours at analysis was conducted using SPSS sta- rolled and information conforming to
37°C. The antigens included a saline tistical software (version 7.5.1, SPSS Inc, the study protocol was available for
control (nil), PPD from M tuberculosis Chicago, Ill). Concordance between 1226 adults. For 133 subjects the TST
(human PPD), PPD from M avium TST and IFN-␥ assay results was as- was not placed, read, or recorded as
(avian PPD), and phytohemaggluti- sessed using ␬ coefficients, where ␬ val- specified. For 97 subjects the IFN-␥ as-
nin (mitogen). The human PPD and ues greater than 0.75 represented ex- say was not performed or recorded as
avian PPD included in the test kits are cellent agreement beyond chance; ␬ specified. For 2 subjects, complete data
prepared by CSL, Limited (Parkville, values less than 0.4 were considered to for both the TST and the IFN-␥ assay
Australia) specifically for IFN-␥ assay represent poor agreement beyond were unavailable. Other critical infor-
testing. After incubation, plasma (200- chance; and ␬ values between 0.4 and mation, including results of mycobac-
300 µL) was collected from above the 0.75 were considered to represent fair terial culture, were missing for 11 sub-
settled blood cells. Plasma samples were to good agreement beyond chance.28 jects. Data from 2 subjects with an
stored at 2°C to 8°C for up to 14 days For these comparisons, IFN-␥ assay re- indeterminate IFN-␥ assay result were
before the concentration of IFN-␥ in 50 sults indicating reactivity to M avium not included in the analysis. Subjects
µL of each sample was quantified by en- were categorized as assay–negative for included in the analysis ranged in age
zyme-linked immunosorbent assay. The M tuberculosis reactivity. We did not from 18 to 87 years (mean, 39 years).
amount of IFN-␥ produced in re- analyze results using receiver operat- Half of the subjects were female; 72%
sponse to the human PPD in excess of ing characteristic curves because there of subjects were born in the United
the saline control (human – nil) was cal- is no “gold standard” for LTBI. Bivar- States; and 38% were white, 35% were
culated, as was the amount of IFN-␥ iate and logistic regression analyses black, 13% were Hispanic, 12% were
produced in excess of the saline con- were used to identify subject-related Asian, and 2% were other races.
trol by the avian PPD and mitogen- and test-related factors associated with As shown in TABLE 1, 87 subjects had
stimulated blood cultures, (avian – nil) test discordance in the group of sub- culture-confirmed TB and completed
and (mitogen – nil), respectively. A jects consisting of those at low- and treatment within the prior 2 years
positive test result for M tuberculosis in- high-risk for LTBI combined (groups (group 4); 94 were suspects being
fection was defined by the following 2 1 and 2). Variables included age, sex, evaluated for active TB who had re-
criteria: race, history of BCG vaccination, HIV ceived anti-TB therapy for fewer than
(1) [(human – nil) / (mitogen – nil) risk, TB exposure, TST in the prior year, 6 weeks (group 3). Additionally, 947
ⱖ 0.15] and time from phlebotomy until incuba- subjects were considered to be at high-
(2) [{(human – nil) – (avian – nil)} / tion of blood with antigens, duration risk for infection with M tuberculosis
(human – nil) ⱖ −0.10]. of incubation, delay to ELISA testing, (group 2), and 98 subjects were con-
An IFN-␥ assay result indicating re- discordance in TST and IFN-␥ assay re- sidered to be at low-risk for M tuber-
activity to M avium complex was de- sults, immune reactivity to M avium culosis infection (group 1). The TST was
fined by the following criteria: complex by IFN-␥ assay, timing of TST interpreted as positive for 390 sub-
(1) (avian – nil) / (mitogen – nil) ⱖ reading, and site where enrolled. jects (31.8%) based on induration and
0.20 and Digit preference for recording TST re- risk strata. Responses less than 15 mm
(2) [(human − nil) – (avian – nil)] / sults of 4, 5, 9, 10, 14, and 15 mm was but more than 4 mm were considered
(human – nil) ⬍ −0.10. assessed by comparing the number of positive for 108 subjects because of an
The IFN-␥ assay result was consid- measurements at each of these values increased risk of infection, suspected
ered to be “indeterminate” if (mitogen with the average number of measure- TB, or culture-confirmed TB (Table 1).
– nil) was less than 0.5 IU. All other ments around the value. For example, Significant digit preference was ob-
IFN-␥ assay result profiles were con- preference for a TST result of 5 mm was served in reporting a TST response of
sidered negative. Calculations and in- assessed by comparing the number of 10 mm at 3 study sites (sites A, B, and
1742 JAMA, October 10, 2001—Vol 286, No. 14 (Reprinted) ©2001 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a University Of Connecticut Health Center User on 05/28/2015

 INTERFERON ␥ ASSAY VS TUBERCULIN SKIN TESTING

C) and a TST response of 15 mm at 3 persons (Table 3) reveals that BCG vac- tive IFN-␥ assay was the site of enroll-
sites (sites B, C, and D). cination was associated with a dispro- ment (TABLE 5). Other factors that were
The IFN-␥ assay indicated that 349 portionate number (n=35) of positive examined but were not statistically as-
subjects (28.5%) had immune reactiv- TST/negative IFN-␥ assay results. Ad- sociated with discordance of either type
ity to M tuberculosis and 101 subjects ditionally, 7 of the 33 nonvaccinated in- were age, sex, HIV risk, TB exposure,
(8.2%) had immune reactivity to M dividuals (21.2%) with positive TST/ and TST in the prior year. Within the
avium complex. The IFN-␥ assay re- negative IFN-␥ assay discordance time periods stipulated by the IFN-␥ as-
vealed no mycobacterial immune reac- demonstrated M avium complex reac- say manufacturer, there was no signifi-
tivity for 776 subjects. tivity by IFN-␥ assay. cant association between discordance
Overall agreement between the IFN-␥ Multivariable analysis was per- and the time from phlebotomy until in-
assay and the TST, using the risk- formed to identify other factors asso- cubation of the blood with the anti-
stratified interpretation of induration, ciated with TST and IFN-␥ assay dis- gens, time of incubation, or delay to
was 83.1% (␬ = 0.60). As shown in cordance. This analysis was confined to ELISA testing. Similarly, there was no
TABLE 2, interpretation of the TST re- the intended population for the IFN-␥ association between discordance and
sult using a single cutoff of 10 mm al- assay, those persons being screened for timing of TST reading within the stipu-
tered the degree of agreement mini- LTBI. Factors statistically associated lated 48 to 72 hours (data not shown).
mally. Agreement between the TST and with a positive TST but negative IFN-␥
the IFN-␥ assay was 91.8% (␬=0.17) for assay included history of BCG vacci- COMMENT
subjects with no identified risk (group nation, Asian race, site of study enroll- The goal of this study was to evaluate
1); 84.9% (␬=0.55) for subjects at high- ment, and evidence of M avium com- the IFN-␥ assay in detecting LTBI.
risk of infection (group 2); 78.7% plex immune reactivity by IFN-␥ assay Evaluation of diagnostic tests for LTBI
(␬=0.41) for the TB suspects (group 3); (TABLE 4). The only factor statistically in humans is hampered by the lack of
and 69.0% (␬=0.16) for subjects with associated with a negative TST but posi- a “gold standard.” As a result, new tests
prior culture-confirmed TB (group 4).
When the analysis was confined to
Table 1. Results of Tuberculin Skin Test (TST)*
persons for whom the IFN-␥ assay is
intended, those being screened for LTBI No. of Subjects With Indicated Induration (No. TST Positive)†
(eg, subjects in groups 1 and 2) and not Induration, mm Group 1 Group 2 Group 3 Group 4
those with suspected or confirmed TB, ⬍5 94 (0) 703 (0) 14 (0) 4 (0)
agreement of the IFN-␥ assay with the 5-9 0 (0) 16 (1)‡ 1 (1) 3 (3)
TST was 84.7% (␬ = 0.55) (TABLE 3). 10-14 2 (0) 80 (76)§ 17 (17) 10 (10)
Within this group, agreement was ⱖ15 2 (2) 148 (148) 62 (62) 70 (70)
88.1% (␬ = 0.50) for subjects with no Total 98 (2) 947 (225) 94 (80) 87 (83)
history of BCG vaccination and 70.1% *Group 1 indicates no known latent tuberculosis infection (LTBI) risk; group 2, high LTBI risk; group 3, suspected TB;
and group 4, prior culture-confirmed TB. See “Methods” section for more details.
(␬=0.41) for those who had received †Risk-stratified interpretation of TST induration.
the BCG vaccine. An examination of ‡One subject was infected with human immunodeficiency virus, hence a 5-mm cutoff was used.
§Four subjects perceived at enrollment to be high-risk for LTBI subsequently denied risk factors and their TST was
TST and IFN-␥ assay results for these interpreted as negative because induration was less than 15-mm cutoff.

Table 2. Agreement Between the Whole-Blood Interferon ␥ (IFN-␥) Assay and Tuberculin Skin Test (TST)*
Overall Stratified Overall 10-mm
Group 1 Group 2 Group 3 Group 4 TST Cutoff TST Cutoff
(n = 98) (n = 947) (n = 94) (n = 87) (n = 1226) (n = 1226)
Positive TST and positive 1 146 63 56 266 265
IFN-␥ assay
Negative TST and negative 89 649 11 4 753 751
IFN-␥ assay
Negative TST and positive 7 73 3 0 83 84
IFN-␥ assay
Positive TST and negative 1 79 17 27 124 126
IFN-␥ assay
Agreement, %
Overall 91.8 84.9 78.7 69.0 83.1 82.9
␬ Coefficients (95% CI) 0.17 (0.04-0.30) 0.55 (0.50-0.61) 0.41 (0.25-0.56) 0.16 (0.06-0.26) 0.60 (0.55-0.65) 0.59 (0.55-0.64)
Positive TST 50.0 64.9 78.8 67.5 68.2 67.8
Negative TST 92.7 89.9 78.6 100.0 90.1 89.9
*Group 1 indicates no known latent tuberculosis infection (LTBI) risk; group 2, high LTBI risk; group 3, suspected TB; and group 4, prior culture-confirmed TB. CI indicates confi-
dence interval.

©2001 American Medical Association. All rights reserved. (Reprinted) JAMA, October 10, 2001—Vol 286, No. 14 1743

Downloaded From: http://jama.jamanetwork.com/ by a University Of Connecticut Health Center User on 05/28/2015

 INTERFERON ␥ ASSAY VS TUBERCULIN SKIN TESTING

are commonly compared with the TST,

Table 3. Impact of BCG Vaccination on Tuberculin Skin Test (TST) and Whole-Blood
Interferon ␥ (IFN-␥) Assay Agreement in Intended Population* despite its well-documented limita-
Unknown tions. Owing to the lack of a definitive
Intended BCG Vaccination standard, the IFN-␥ assay was evalu-
Population Vaccinated Unvaccinated Status ated on the basis of its agreement with
(n = 1045) (n = 157) (n = 770) (n = 118)
Positive TST and positive 147 56 60 31
the TST in persons with varying de-
IFN-␥ assay grees of risk for M tuberculosis infec-
Negative TST and negative 738 54 618 66 tion and in persons with documented
IFN-␥ assay and suspected active TB. Overall agree-
Negative TST and positive 80 12 59 9 ment between the TST and IFN-␥ as-
IFN-␥ assay
Postive TST and negative 80 35 33 12
say was good (83.1%, ␬=0.60) as was
IFN-␥ assay agreement when the analysis was lim-
Agreement, % ited to persons for whom the test is in-
Overall 84.7 70.1 88.1 82.2 tended, those subjects being screened
␬ Coefficient (95% CI) 0.55 (0.50-0.60) 0.41 (0.29-0.54) 0.50 (0.44-0.56) 0.61 (0.46-0.76) for LTBI, groups 1 and 2 combined
Positive TST 64.8 61.5 64.5 72.1 (84.7%, ␬=0.55). Test concordance was
Negative TST 90.2 81.8 91.3 88.0 65% for persons with a positive TST and
*CI indicates confidence interval. The intended population is persons being screened for latent infection.
90% for those with a negative TST. Pot-
tumarthy et al27 found a similar level of
Table 4. Factors Associated With Positive Tuberculin Skin Test (TST)/Negative Interferon ␥ agreement between the 2 tests in a study
(IFN-␥) Assay Discordance Using a Multivariable Model With Group 1 and Group 2 Subjects
Combined* involving New Zealand health care
Odds Ratio
workers and immigrants, whereas
Variable No. of Subjects (95% Confidence Interval) P Value Streeton et al16 reported a somewhat
Race better concordance of 98% for per-
White 396 1.0 sons with no known exposure and a
Hispanic 103 1.24 (0.52-2.99) .63 negative TST, and 90% for untreated
Black 346 1.69 (0.82-3.48) .15 TST reactors.
Asian 91 2.33 (1.05-5.21) .04
The agreement between the IFN-␥
Other 23 0.61 (0.07-5.36) .66
assay and TST in this study is similar
History of BCG vaccine
None 707 1.0 to the agreement found when mul-
Unknown 108 2.49 (1.07-5.76) .03 tiple TSTs are administered simulta-
Vaccinated 144 6.92 (3.56-13.43) ⬍.001 neously using different PPD prepara-
Site where enrolled tions. Villarino et al 2 9 found ␬
A 201 1.0 coefficients of 0.46 to 0.53 when com-
B 137 1.48 (0.44-5.00) .52 paring Tubersol PPD (Pasteur Merieux
C 221 3.52 (1.15-10.76) .03 Connaught USA, Swiftwater, Pa) with
D 182 3.30 (1.10-9.95) .03 Aplisol PPD (Parkdale Pharmaceuti-
E 218 4.29 (1.50-12.31) .01 cals, Rochester, Mich), Aplisol PPD
Mycobacterium avium complex with PPD-S1 (produced by Seibert and
by whole-blood IFN-␥ assay
No 866 1.0 Glenn in 1941 and available from the
Yes 93 2.64 (1.28-5.42) .01 Food and Drug Administration), or Tu-
*Odds ratios were adjusted for the variables listed and for age, sex, human immunodeficiency virus risk, tuberculosis bersol with PPD-S1 (M. Elsa Villa-
exposure, and TST in the prior year. rino, MD, written communication, Feb-
ruary 2001). This is similar to the ␬
Table 5. Factors Associated With Negative Tuberculin Skin Test (TST)/Positive Whole-Blood coefficient of 0.60 we observed in the
Interferon ␥ (IFN-␥) Assay Discordance Using a Multivariable Model With Group 1 and Group
2 Subjects Combined* present study. Additionally, the IFN-␥
Variable No. of Subjects Odds Ratio (95% Confidence Interval) P Value assay and the TST measure different pa-
Site where enrolled rameters of the immune response,
A 221 1.0 which are not exclusively linked. This
B 152 1.54 (0.78-3.03) .22 is demonstrated in IFN-␥ knockout
C 206 0.28 (0.10-0.76) .01 mice, which are able to mount a de-
D 174 0.55 (0.26-1.20) .13 layed-type hypersensitivity reaction to
E 207 0.58 (0.28-1.21) .15 PPD despite the absence of IFN-␥ pro-
*Odds ratios were adjusted for the variables listed and for age, sex, human immunodeficiency virus risk, tuberculosis duction.30 The discordant results seen
exposure, and TST in the prior year.
for the IFN-␥ assay and the TST in our
1744 JAMA, October 10, 2001—Vol 286, No. 14 (Reprinted) ©2001 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a University Of Connecticut Health Center User on 05/28/2015

 INTERFERON ␥ ASSAY VS TUBERCULIN SKIN TESTING

study could be due to measurement of assay. Thus, the IFN-␥ assay offers a po- most of these people with no identi-
different immune parameters by the 2 tentially significant improvement in fied risk for LTBI had a false-positive
tests, but it may also be influenced by specificity with the consequent ben- IFN-␥ assay. However, evidence that
the use of different PPD preparations. efit of avoiding LTBI treatment for per- some negative TST/positive IFN-␥ as-
The inherent problem in compar- sons not infected with M tuberculosis. say results reflect greater sensitivity for
ing the IFN-␥ assay with the TST is that, In the future, the use of antigens in the the IFN-␥ assay comes from prior com-
by virtue of the trial design, the assay IFN-␥ assay that are found in M tuber- parisons of the tests. Converse et al19
cannot be perceived to perform better culosis but not in nontuberculous my- provided evidence that the IFN-␥ as-
than the TST. Because there is no gold cobacteria or BCG (such as ESAT-6) is say is more sensitive than the TST for
standard for LTBI, comparisons can- likely to improve the assay’s ability to injection drug users with and without
not demonstrate which test is supe- discriminate among LTBI, BCG vacci- HIV infection. Nine of 24 subjects who
rior for LTBI. Discrepancies encoun- nation, and nontuberculous mycobac- were positive by IFN-␥ assay but nega-
tered may be the result of limitations terial reactivity.20 tive by TST had a history of a positive
in the TST and not limitations in the Multivariable analysis showed that TST in the past. Kimura et al17 also re-
IFN-␥ assay. In recognition of this, lo- there was an association between 3 of ported finding more intravenous drug
gistic regression analysis was used to the study sites and test discordance. Al- users having immune reactivity to PPD
identify subject-related and test- though this could possibly be ex- with the IFN-␥ assay than with the TST.
related factors associated with test dis- plained by population differences be- Additional evidence that the IFN-␥ as-
cordance. Factors examined included tween the sites, 2 observations suggest say may be more sensitive for the de-
those known to adversely affect TST ac- that differences in testing methods were tection of LTBI comes from studies of
curacy, such as BCG vaccination and involved. The first is documentation of the bovine version of the assay; the
reactivity to nontuberculous mycobac- digit preference in TST results at 4 sites, cattle can be killed and cultures can be
teria. 2 of which were associated with posi- used as the gold standard. Wood et al15
A history of BCG vaccination was tive TST/negative IFN-␥ assay discor- found that 37 of 67 cattle (55.2%) that
strongly associated with positive TST/ dance. The rounding up of TST indu- were positive in the bovine IFN-␥ as-
negative IFN-␥ assay discordance. BCG ration measures to 10 or 15 mm would say but TST negative were culture-
vaccination is known to induce reac- lead to some persons being classified positive for M tuberculosis complex, spe-
tivity to PPD and can cause false- falsely as TST positive, according to the cifically M bovis. In contrast, only 2 of
positive TST reactions.31 Our results ATS/CDC interpretation criteria.6 The the 53 (3.8%) animals with a positive
suggest that the IFN-␥ assay may be less evidence of digit preference is a good TST and negative IFN-␥ assay result
affected than TST by prior BCG vacci- example of the subjectivity often en- were culture positive. These findings are
nation. However, the IFN-␥ assay is also countered with interpreting the TST re- supported by data from New Zealand,
affected by BCG, as was demonstrated sult. Persons reading TST reactions in where the bovine IFN-␥ assay is rou-
by Johnson et al20 in a study of medi- the present study were not blinded to tinely used to detect infected cattle that
cal students tested prior to and 5 patient histories and may have been un- are undetected by the TST. 37 Pub-
months after vaccination. The effect consciously biased toward a positive or lished data demonstrate that approxi-
may be relatively short-lived and may negative result. In contrast, the IFN-␥ mately 50% of negative TST/positive
diminish with time, as shown in ani- assay is not subject to operator bias. The IFN-␥ assay animals are truly infected
mals.32-34 second observation was that the sites with M bovis, as confirmed by culture
Reactivity to nontuberculous myco- statistically associated with positive of the organism.38,39
bacteria is also known to cause false- TST/negative IFN-␥ assay discor- Agreement between the TST and the
positive TST results.35,36 The IFN-␥ as- dance had less negative TST/positive IFN-␥ assay was less for subjects with
say measures the cellular immune IFN-␥ assay discordance. Again, this culture-confirmed TB (both current and
response to both human PPD and avian may be caused by the subjectivity of previously treated) as compared with
PPD. Avian PPD is included in the as- reading the TST result. It is unlikely that those subjects without positive cul-
say as an indicator for nontubercu- this variation is due to differences in tures. The observation that active TB is
lous mycobacterial reactivity, akin to IFN-␥ assay methods because all sites associated with a decrease in PPD-
the use of Battey Bacillus sensitin used identical standards and quality specific IFN-␥ production may ex-
(PPD-B) in a comparative skin test. Re- control documentation. plain these differences.40-42 Although
sults from the IFN-␥ assay suggested In the current study, more persons both TST and IFN-␥ assay responses
that reactivity to nontuberculous my- with no identified risk for LTBI were can be reduced in cases of active dis-
cobacteria may be the cause of a posi- negative for TST and positive for IFN-␥ ease, TST responses are usually re-
tive TST result in one fifth of the non- assay than expected. Although some of stored within 2 weeks after initiation
BCG vaccinated subjects that were these individuals may have false- of antimycobacterial chemotherapy and
positive by TST but negative by IFN-␥ negative TST results, it is likely that nutritional supplementation.43 Follow-
©2001 American Medical Association. All rights reserved. (Reprinted) JAMA, October 10, 2001—Vol 286, No. 14 1745

Downloaded From: http://jama.jamanetwork.com/ by a University Of Connecticut Health Center User on 05/28/2015

 INTERFERON ␥ ASSAY VS TUBERCULIN SKIN TESTING

ing treatment, PPD-specific IFN-␥ re- Health, Baltimore, Md (Dr Bishai); and Biosciences Di- culosis in HIV-infected drug users with cutaneous an-
vision, CSL Limited, Parkville, Australia (Dr Rothel). Dr ergy. JAMA. 1992;268:504-509.
sponses usually increase but can re- LoBue is now with the Division of Tuberculosis Elimi- 10. Judson FN, Feldman RA. Mycobacterial skin tests
main low or absent for at least a year nation, Centers for Disease Control and Prevention, in humans 12 years after infection with Mycobacte-
Atlanta, Ga, and assigned to the San Diego County rium marinum. Am Rev Respir Dis. 1974;109:544-
following successful completion of Tuberculosis Control Program; Dr Rothel is now with 547.
therapy.44,45 These findings suggest that Cellestis Limited, St Kilda, Australia. 11. Snider DE Jr. Bacille Calmette-Guerin vaccina-
Financial Disclosure: Dr Rothel owns stock and is em- tions and tuberculin skin tests. JAMA. 1985;253:3438-
the results found in the present study ployed by Cellestis Limited, which markets the Quan- 3439.
could be skewed in favor of the TST tiFERON-TB test used in this study. 12. Huebner RE, Schein MF, Bass JBJ. The tuberculin
since all but a few individuals with cul- Author Contributions: Study concept and design: Ma- skin test. Clin Infect Dis. 1993;17:968-975.
zurek, LoBue, Daley, Bernardo, Lardizabal, Bishai, Iade- 13. Bearman JE. A study of variability in tuberculin test
ture-confirmed TB had received therapy marco, Rothel. reading. Am Rev Respir Dis. 1964;90:913-919.
for at least 4 weeks. In persons who Acquisition of data: Mazurek, LoBue, Daley, Ber- 14. Rothel JS, Jones SL, Corner LA, Cox JC, Wood PR.
nardo, Lardizabal, Bishai, Iademarco, Rothel. A sandwich enzyme immunoassay for bovine inter-
have already been diagnosed with TB Analysis and interpretation of data: Mazurek, feron-gamma and its use for the detection of tuber-
and begun chemotherapy, the use of ei- LoBue, Daley, Bernardo, Lardizabal, Bishai, Iade- culosis in cattle. Aust Vet J. 1990;67:134-137.
marco, Rothel. 15. Wood PR, Corner LA, Rothel JS, et al. Field com-
ther test is unlikely. To accurately com- Drafting of the manuscript: Mazurek, LoBue, Daley, parison of the interferon-␥ assay and the intradermal
pare the sensitivity of the TST with the Bernardo, Lardizabal, Bishai, Iademarco, Rothel. tuberculin test for the diagnosis of bovine tuberculo-
IFN-␥ assay for the diagnosis of active Critical revision of the manuscript for important in- sis. Aust Vet J. 1991;68:286-290.
tellectual content: Mazurek, LoBue, Daley, Ber- 16. Streeton JA, Desem N, Jones SL. Sensitivity and
TB, subjects should be tested prior to nardo, Lardizabal, Bishai, Iademarco, Rothel. specificity of a ␥ interferon blood test for tuberculosis
the initiation of anti-TB therapy. Statistical expertise: Mazurek, Iademarco. infection. Int J Tuberc Lung Dis. 1998;2:443-450.
Obtained funding: Mazurek, LoBue, Daley, Ber- 17. Kimura M, Converse PJ, Astemborski J, et al. Com-
In this study, the IFN-␥ assay was nardo, Lardizabal, Bishai. parison between a whole blood interferon-␥ release assay
comparable with the TST in its ability Administrative, technical, or material support: Ma- and tuberculin skin testing for the detection of tuber-
zurek, LoBue, Daley, Bernardo, Lardizabal, Bishai,
to detect latent M tuberculosis infec- Rothel.
culosis infection among patients at risk for tuberculosis
exposure. J Infect Dis. 1999;179:1297-1300.
tion. The assay has several logistic ad- Study supervision: Mazurek, LoBue, Daley, Ber- 18. Desem N, Jones SL. Development of a human ␥
vantages over the TST. Unlike the TST, nardo, Lardizabal, Bishai, Rothel. interferon enzyme immunoassay and comparison with
Funding/Support: This work was funded by the Cen- tuberculin skin testing for detection of Mycobacte-
the IFN-␥ assay requires a single pa- ters for Disease Control and Prevention. rium tuberculosis infection. Clin Diagn Lab Immu-
tient visit, an important benefit be- Acknowledgment: We thank Saundra Barnes, RN, nol. 1998;5:531-536.
Cindy Merrifield, RN, Grace Link-Barnes, RN, MPH, 19. Converse PJ, Jones SL, Astemborski J, Vlahov D,
cause the proportion of persons return- Muppy Haigler, RN, Crystal Carter, Richard E. Chais- Graham NM. Comparison of a tuberculin inter-
ing to have their TST read is very low son, MD, Ken Dansbury, Tarek Elbeik, PhD, Richard
feron-␥ assay with the tuberculin skin test in high-
Stephens, PhD, John Stephens, Chiew Ko, ScM, Wei
in some settings.46 The IFN-␥ assay as- Lu, Marilyn Owens, RN, Peach Francisco, RN, Pat Jun-
risk adults: effect of human immunodeficiency virus
infection. J Infect Dis. 1997;176:144-150.
sesses responses to multiple antigens si- quera, Sue Ann Diprofio, RN, PHN, Tony Catanzaro,
20. Johnson PD, Stuart RL, Grayson ML, et al. Tu-
MD, and Savita Rao, PhD, for technical assistance; Ann
multaneously and includes avian PPD Lanner, Elsa Villarino, MD, MPH, and Rick O’Brien,
berculin-purified protein derivative-, MPT-64-, and
to discriminate responses due to reac- ESAT-6-stimulated ␥ interferon responses in medical
MD, for editorial assistance; and Damian Jolley, MSc,
students before and after Mycobacterium bovis BCG
tivity to nontuberculous mycobacte- Yong-Cheng Wang, PhD, Lorna Thorpe, PhD, Kayla
vaccination and in patients with tuberculosis. Clin Di-
Laserson, ScD, Kenneth Castro, MD, and Jose Becerra,
ria from those due to LTBI. The assay agn Lab Immunol. 1999;6:934-937.
MD, MPH, for analytical and statistical assistance.
21. Wood PR, Corner LA, Rothel JS, et al. A field evalu-
does not boost amnestic immune re- ation of serological and cellular diagnostic tests for bo-
sponses, eliminates the subjectivity of vine tuberculosis. Vet Microbiol. 1992;31:71-79.
REFERENCES 22. Domingo M, Liebana E, Vilafranca M, et al. A field
the TST, and can be completed in less evaluation of the interferon-␥ assay and the intrader-
1. Raviglione MC, Snider DE Jr, Kochi A. Global epi-
than 24 hours. The finding that BCG demiology of tuberculosis: morbidity and mortality of mal tuberculin test in dairy cattle in Spain. In: Griffin
vaccination has less effect on the IFN-␥ a worldwide epidemic. JAMA. 1995;273:220-226. F, De Lisle G, eds. Tuberculosis in Wildlife and Do-
2. Nardell EA. Pathogenesis of tuberculosis. In: Rei- mestic Animals. Dunedin, New Zealand: University of
assay than on the TST is promising, chman LB, Hershfield ES, eds. Tuberculosis: A Com- Otago Press; 1995:304-306.
given that a large proportion of TB cases prehensive International Approach. New York, NY: 23. Dondo A, Goria M, Moda G, et al. Gamma in-
Marcel Dekker Inc; 1993:103-122. terferon assay for the diagnosis of bovine tuberculo-
in the United States are in persons origi- 3. Recommendations of the Advisory Council for the sis: field evaluation of sensitivity and specificity.
nating from countries where BCG vac- Elimination of Tuberculosis. Screening for tuberculosis Medicina Veterinaria Preventiva. 1996;13:14-19.
and tuberculosis infection in high-risk populations. 24. Monaghan M, Collins D, McMurray C, Kelly A.
cination is commonplace. Better per- Field trials of the ␥ interferon assay for the diagnosis
MMWR Morb Mortal Wkly Rep. 1995;44:19-34.
formance may be seen when TB- 4. Bass JB Jr, Farer LS, Hopewell PC, et al. Treatment of bovine tuberculosis in the Republic of Ireland. In:
specific antigens are included in the of tuberculosis and tuberculosis infection in adults and Griffin F, De Lisle G, eds. Tuberculosis in Wildlife and
children. Am J Respir Crit Care Med. 1994;149:1359- Domestic Animals. Dunedin, New Zealand: Univer-
IFN-␥ assay. 1374. sity of Otago Press; 1995:319-320.
5. Institute of Medicine. Ending Neglect: The Elimi- 25. Monaghan M, Quinn PJ, Kelly AP, et al. A pilot
Author Affiliations: Division of Tuberculosis Elimina- nation of Tuberculosis in the United States. Wash- trial to evaluate the ␥-interferon assay for the detec-
tion, Centers for Disease Control and Prevention, and ington, DC: National Academy Press; 2000. tion of Mycobacterium bovis infected cattle under Irish
Pulmonary and Critical Care Division, Department of 6. Diagnostic standards and classification of tuber- conditions. Ir Vet J. 1997;50:229-232.
Medicine, Emory University School of Medicine (Drs culosis in adults and children. Am J Respir Crit Care 26. World Health Organization. Global Tuberculo-
Mazurek and Iademarco), Atlanta, Ga; Pulmonary and Med. 2000;161:1376-1395. sis Control. Geneva, Switzerland: WHO; 1999. WHO/
Critical Care Division, University of California at San 7. Edwards PQ, Edwards LB. Story of the tuberculin CDS/CPC/TB/99.259.
Diego (Dr LoBue); Pulmonary and Critical Care Divi- skin test from an epidemiologic viewpoint. Am Rev 27. Pottumarthy S, Morris AJ, Harrison AC, Wells VC.
sion, University of California at San Francisco (Dr Da- Respir Dis. 1960;81:1-47. Evaluation of the tuberculin ␥ interferon assay: po-
ley); Pulmonary Center, Boston University School of 8. Antonucci G, Girardi E, Raviglione MC, Ippolito G. tential to replace the Mantoux skin test. J Clin Mi-
Medicine, Mass (Dr Bernardo); New Jersey Medical Risk factors for tuberculosis in HIV-infected persons. crobiol. 1999;37:3229-3232.
School National Tuberculosis Center, Newark (Dr Lar- JAMA. 1995;274:143-148. 28. Fleiss JL. The measurement of interrater agree-
dizabal); Department of International Health, Johns 9. Selwyn PA, Sckell BM, Alcabes P, Friedland GH, ment. In: Bradley RA, Hunter JS, Kendal DG, Watson
Hopkins University School of Hygiene and Public Klein RS, Schoenbaum EE. High risk of active tuber- GS, eds. Statistical Methods for Rates and Propor-

1746 JAMA, October 10, 2001—Vol 286, No. 14 (Reprinted) ©2001 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a University Of Connecticut Health Center User on 05/28/2015

 INTERFERON ␥ ASSAY VS TUBERCULIN SKIN TESTING

tions. New York, NY: John Wiley & Sons Inc; 1981: ity with ability of strains to grow in macrophages. In- 41. Dlugovitzky D, Bay ML, Rateni L, et al. In vitro
212-236. fect Immun. 1999;67:2172-2177. synthesis of interferon-␥, interleukin-4, transforming
29. Villarino ME, Burman W, Wang YC, et al. Com- 35. Chaparas SD. Immunologically based diagnostic growth factor-␤ and interleukin-1 ␤ by peripheral blood
parable specificity of 2 commercial tuberculin re- tests with tuberculin and other mycobacterial anti- mononuclear cells from tuberculosis patients: rela-
agents in persons at low risk for tuberculous infec- gens. In: Kubica GP, Wayne LG, eds. The Mycobac- tionship with the severity of pulmonary involvement.
tion. JAMA. 1999;281:169-171. teria: A Source Book. New York, NY: Marcel Dekker Scand J Immunol. 1999;49:210-217.
30. Flynn JL, Chan J, Triebold KJ, Dalton DK, Stew- Inc; 1984:195-220. 42. Swaminathan S, Gong J, Zhang M, et al. Cyto-
art TA, Bloom BR. An essential role for interferon ␥ in 36. Von Reyn CF, Williams DE, Horsburgh CR Jr, et kine production in children with tuberculous infec-
resistance to Mycobacterium tuberculosis infection. al. Dual skin testing with Mycobacterium avium sen- tion and disease. Clin Infect Dis. 1999;28:1290-
J Exp Med. 1993;178:2249-2254. sitin and purified protein derivative to discriminate pul- 1293.
31. Menzies R, Vissandjee B. Effect of bacille Calmette- monary disease due to M avium complex from pul- 43. Rooney JJ Jr, Crocco JA, Kramer S, Lyons HA. Fur-
Guerin vaccination on tuberculin reactivity. Am Rev monary disease due to Mycobacterium tuberculosis. ther observations on tuberculin reactions in active tu-
Respir Dis. 1992;145:621-625. J Infect Dis. 1998;177:730-736. berculosis. Am J Med. 1976;60:517-522.
32. Buddle BM, Keen D, Thomson A, et al. Protec- 37. Wood PR, Jones SL. BOVIGAM: an in vitro cel- 44. Ellner JJ. Review: the immune response in hu-
tion of cattle from bovine tuberculosis by vaccination lular diagnostic test for bovine tuberculosis. Tubercu- man tuberculosis–implications for tuberculosis con-
with BCG by the respiratory or subcutaneous route, losis. 2001;81:147-155. trol. J Infect Dis. 1997;176:1351-1359.
but not by vaccination with killed Mycobacterium vac- 38. Neill SD, Cassidy J, Hanna J, et al. Detection of 45. Hirsch CS, Toossi Z, Othieno C, et al. Depressed
cae. Res Vet Sci. 1995;59:10-16. Mycobacterium bovis infection in skin test-negative T-cell interferon-␥ responses in pulmonary tubercu-
33. Buddle BM, De Lisle GW, Pfeffer A, Aldwell FE. cattle with an assay for bovine interferon-␥. Vet Rec. losis: analysis of underlying mechanisms and modu-
Immunological responses and protection against My- 1994;135:134-135. lation with therapy. J Infect Dis. 1999;180:2069-
cobacterium bovis in calves vaccinated with a low dose 39. Ryan T, Livingstone P. Risk analysis: movement 2073.
of BCG. Vaccine. 1995;13:1123-1130. of cattle from tuberculosis-infected herds. Surveil- 46. Chaisson RE, Keruly JC, McAvinue S, Gallant JE,
34. Wedlock DN, Aldwell FE, Collins DM, De Lisle GW, lance. 2000;27:8-10. Moore RD. Effects of an incentive and education pro-
Wilson T, Buddle BM. Immune responses induced in 40. Sodhi A, Gong J, Silva C, Qian D, Barnes PF. Clini- gram on return rates for PPD test reading in patients
cattle by virulent and attenuated Mycobacterium bo- cal correlates of interferon ␥ production in patients with with HIV infection. J Acquir Immune Defic Syndr Hum
vis strains: correlation of delayed-type hypersensitiv- tuberculosis. Clin Infect Dis. 1997;25:617-620. Retrovirol. 1996;11:455-459.

In all my work what I try to say is that as human beings

we are more alike than we are unalike.
—Maya Angelou (1928- )

©2001 American Medical Association. All rights reserved. (Reprinted) JAMA, October 10, 2001—Vol 286, No. 14 1747

Downloaded From: http://jama.jamanetwork.com/ by a University Of Connecticut Health Center User on 05/28/2015