The Cost Effectiveness of Interferon Gamma Release Assays Versus Tuberculin Skin Tests in Health Care

Transcription

1 ORIGINAL INVESTIGATION Cost-effectiveness of Interferon Gamma Release Assays vs Tuberculin Skin Tests in Health Care Workers Marie A. de Perio, MD; Joel Tsevat, MD, MPH; Gary A. Roselle, MD; Stephen M. Kralovic, MD, MPH; Mark H. Eckman, MD, MS Background: Interferon gamma release assays (IGRAs) offer alternatives to tuberculin skin tests (TSTs) for diagnosing latent tuberculosis infection (LTBI). Unlike TSTs, IGRAs require only a single patient visit and are not affected by prior BCG vaccination, providing greater specificity. Of 2 Food and Drug Administration approved IGRAs, the newer QuantiFERON-TB Gold in Tube test (QFT-GIT) requires less manual processing time than the QuantiFERON-TB Gold test (QFT-G). We compared the cost-effectiveness of the QFT-G, QFT-GIT, and TST for detecting LTBI in new health care workers (HCWs). Methods: A Markov state-transition decision analytic model using the societal perspective and lifetime horizon was constructed to compare costs and qualityadjusted life-years (QALYs) associated with the 3 strategies for hypothetical 35-year-old HCWs with or without prior BCG vaccination. Direct and indirect costs and probabilities were based on manufacturer data, national Veterans Health Administration records, and the published literature. Future costs and QALYs were discounted at 3% per year. Results: Both IGRAs were more effective and less costly than the TST, whether or not the HCW had been vaccinated with BCG previously. The incremental costeffectiveness ratio of the QFT-G compared with the QFT- GIT was $14 092/QALY for non BCG-vaccinated HCWs and $ /QALY for BCG-vaccinated HCWs. There was no prevalence of LTBI at which the TST became the most effective or least costly strategy. If the sensitivity of the QFT-GIT exceeds that of the QFT-G, then the QFT- GIT is the most effective and least costly strategy. Conclusion: Use of the QFT-G and QFT-GIT leads to superior clinical outcomes and lower costs than the TST and should be considered in screening non BCGvaccinated and BCG-vaccinated new HCWs for LTBI. Arch Intern Med. 2009;169(2): Author Affiliations: Divisions of Infectious Diseases (Drs de Perio, Roselle, and Kralovic) and General Internal Medicine (Drs Tsevat and Eckman), University of Cincinnati College of Medicine, and Cincinnati Veterans Affairs Medical Center (Drs Tsevat, Roselle, and Kralovic), Cincinnati, Ohio; and Veterans Affairs Central Office Infectious Diseases Program, Washington, DC (Drs Roselle and Kralovic). HEALTH CARE WORKERS (HCWs) are at increased risk for Mycobacterium tuberculosis infection. 1 The 2005 Centers for Disease Control and Prevention (CDC) guidelines for preventing tuberculosis (TB) transmission in health care settings recommend a comprehensive program consisting of administrative, environmental, and personal respiratory protection controls. 2 One cornerstone of TB control in health care settings involves routinely screening HCWs for latent TB infection (LTBI) using the tuberculin skin test (TST) and administering isoniazid treatment to HCWs testing positive. In 2005, the Food and Drug Administration (FDA) approved the Quanti- FERON-TB Gold test (QFT-G) (Cellestis Limited, Victoria, Australia), a wholeblood interferon gamma release assay (IGRA) used to diagnose active TB and LTBI. 3 The QFT-G is an enzyme-linked immunosorbent assay (ELISA) test that measures the release of interferon gamma in blood from sensitized persons. The antigens consist of synthetic peptides representing 2 M tuberculosis proteins, early secretory antigenic target 6 (ESAT-6) and culture filtrate protein 10 (CFP-10). Blood is incubated with the antigens, and interferon gamma released by sensitized leukocytes is measured. 4 The CDC guidelines published in 2005 indicate that the QFT-G can be used in any instance in which the TST is used. 3 In 2007, the FDA approved the nextgeneration IGRA, the QuantiFERON-TB Gold in Tube test (QFT-GIT) (Cellestis Limited). This test contains an extra antigen, TB7.7, which theoretically improves sensitivity and circumvents the time-consuming step of manually stimulating lymphocytes, as the tubes already contain the antigens. 179

2 QFT + LTBI QFT determinate QFT-G No LTBI QFT adequate QFT LTBI QFT indeterminate Repeat QFT Newly employed health care worker QFT-GIT No LTBI QFT inadequate Repeat QFT TST + Second TST read Second TST + LTBI TST read Second TST placed Second TST TST TST Second TST not read No LTBI Second TST not placed TST not read Decision node Chance node Markov node Common subtree Die of hepatitis Dead Hepatitis Accept isoniazid treatment Survive hepatitis Finish isoniazid treatment QFT-G + QFT-GIT + TST- + CXR Isoniazid treatment indicated Refuse isoniazid treatment No hepatitis Not finish isoniazid treatment Isoniazid treatment not indicated Figure 1. Simplified diagram of the Markov state-transition decision analytic model. CXR indicates chest radiograph; LTBI, latent tuberculosis infection; QFT-G, QuantiFERON-TB Gold test (Cellestis Limited, Victoria, Australia); QFT-GIT, QuantiFERON-TB Gold in Tube test (Cellestis Limited); TST, tuberculin skin test. Advantages of both IGRAs over the TST include that they necessitate only a single patient visit, results are available in 24 hours, and the findings are not subject to reader bias. The ESAT-6, CFP-10, and TB7.7 proteins are absent from all BCG vaccine strains and from many nontuberculous mycobacteria. 5 Therefore, among previously BCGvaccinated and non BCG-vaccinated subjects, the IGRAs have high specificity. 6 Major disadvantages of the IGRAs include their high relative cost and the need for an equipped laboratory. 7 Blood samples must be processed within 12 hours with the QFT-G, and errors in collecting, transporting, or running the assay can lead to inadequate test results. Consequently, IGRA tests can be nondiagnostic, necessitating second testing. Finally, data are limited in children, recently exposed subjects, and immunocompromised individuals. 3 Because HCWs are at risk of contracting and spreading TB, guidelines recommend that they be tested for LTBI when starting work and annually while working in a health care setting. 2 The objective of this study was to compare the cost-effectiveness of 3 strategies for detecting LTBI in new HCWs, namely, the QFT-G, QFT-GIT, and TST. METHODS STUDIED POPULATION In 2007, the national Veterans Health Administration (VHA) system included 153 hospitals, 135 nursing homes, and 47 domiciles. 8 The number of full-time employees in 2007 totaled , including physicians, 847 dentists, registered nurses, licensed practical nurses, 7868 nonphy- 180

3 sician providers, health technicians and allied health personnel, wage board and purchase and hire employees, and other employees. 8 In the VHA system, TSTs are typically performed during preplacement orientation. In recent years, the number of new workers, including volunteers, receiving a preemployment TST has ranged from to each year (G.A.R., unpublished data, February 20, 2007). The VHA policy calls for performing 2-step TSTs in individuals who were without a TST in the previous year. 9 Given the heterogeneity of the VHA HCW population, it was necessary to choose a prototypical HCW, a standard technique in population-based decision analyses. The base-case population in our analysis consisted of hypothetical 35-yearold HCWs beginning work at a VHA health care facility. Because female registered nurses make up the bulk of the HCW population, their pay schedule was chosen for the base-case analysis. MODEL DESIGN We constructed a Markov state-transition decision analytic model using the societal perspective and a lifetime horizon. Direct costs, costs of missed work time, and probabilities were based on data from the IGRA manufacturer, from the VHA, and from the published literature. Effectiveness was measured in quality-adjusted life-years (QALYs). We calculated the incremental costs per QALY gained for the 3 strategies, discounting future costs and QALYs at 3% per year. 10 Analyses were conducted using commercially available software (Decision Maker 4.0; Decision Maker Software- Beta version, Stephen G. Pauker, MD, Frank Sonnenberg, MD, John B. Wong, MD, C. Greg Hagerty, PhD, creators, New England Medical Center, Boston, Massachusetts). 11 LTBI SCREENING MODEL Our Markov state-transition decision analytic model included all steps of LTBI screening (Figure 1). We conducted separate analyses for HCWs previously vaccinated with BCG and for HCWs not previously vaccinated. Strategies included the QFT-G, QFT-GIT, or TST. The HCWs entered the model with underlying LTBI, multidrug-resistant LTBI (data not shown), or neither. We accounted for inadequate and indeterminate outcomes of both QFTs, for failure to return for TST reading, and for 2-step testing. We incorporated the medical evaluation process, including whether isoniazid treatment is indicated, accepted, and taken for a full 9 months and whether the HCW develops hepatitis. Evaluation of transmission from HCWs to others and annual testing were beyond the scope of this model. The Markov state-transition decision analytic model analysis simulated the movement of the HCW population among several health states (Figure 2). The HCWs were subject to a risk of death because of demographic factors and to an annual risk of developing active TB. Each Markov cycle length was 1 year. As HCWs cycled through this model, costs and QALYs were accumulated. SIMPLIFYING ASSUMPTIONS The following simplifying assumptions were included: (1) No HCW undergoing screening has underlying human immunodeficiency virus or active TB. (2) A second QFT test result is always adequate, and a second indeterminate QFT test result is treated as negative. (3) The HCWs who develop hepatitis or who do not finish a full 9-month course of isoniazid treatment take isoniazid for 3 months. (4) Survivors of active TB experience no long-term morbidity and do not develop a second case. LTBI, isoniazid treatment, complete LTBI, isoniazid treatment, partial LTBI, no isoniazid treatment No LTBI, isoniazid treatment, complete No LTBI, isoniazid treatment, partial No LTBI, no isoniazid treatment (5) IGRA tests are performed onsite at each local VHA medical center. BASE-CASE INPUT VARIABLES Prevalence and Test Characteristics Base-case parameter values were derived from the VHA and from the published literature (Table 1). Absent a current gold standard for the diagnosis of LTBI, the true prevalence of LTBI is unknown. Our base-case estimate of 5% was explored through sensitivity analyses, ranging from 0% to 100%. We calculated test sensitivity and specificity from individual studies 6,17-24,26-30 published as of February 2007 and pooled the results to generate weighted mean test characteristics. We separated out indeterminate IGRA results from positive and negative results and modeled them independently. We calculated specificity separately for non BCG-vaccinated and BCG-vaccinated subjects. The TST characteristics were based on studies that compared the TST and QFT testing directly. 6,17-22,26,28,29 Because of the scarcity of data regarding the proportions of QFT-Gs that are adequately obtained, the specificity of the QFT- GIT, and the probability of an indeterminate second QFT-GIT result, we used equivalent values for these input variables for both IGRAs. We used statistics from the national VHA database to determine the probabilities that a HCW returns to have a first and second TST read and that a second TST is placed. We used the following published exponential regression model 13 to calculate the probability of developing TB in any given year following detection of a positive test result 35 : Annual Probability= e t, where e is the mathematical constant for exponentiation and t is the number of years after exposure. In a sensitivity analysis, we varied this probability by factors from 1 to 10 to determine the effect of different probabilities on the outcome. To calculate death rates from other causes, we converted age-, sex-, and race/ethnicityassociated mortality rates reported in 2003 life tables into probabilities. 36 Costs Active TB Dead Figure 2. Simplified diagram of the major Markov state-transition decision analytic model states. Arrows indicate the direction in which the health care worker can move from one health state to another each year. LTBI indicates latent tuberculosis infection; TB, tuberculosis. Our analysis included all direct costs and costs associated with missing work related to each strategy (Table 2). Costs were converted to 2007 US dollars 41 with 3% discounting. 10 Cost estimate sources included the published literature, Medi- 181

4 Table 1. Base-Case Parameter Values Input Variable Base-Case Value for Non BCG-Vaccinated Cohort (BCG-Vaccinated Cohort, if Different) Source Prevalence Prevalence of LTBI 0.05 Estimate Prevalence of multidrug-resistant LTBI 0.01 Centers for Disease Control and Prevention, Lifetime probability of developing active TB 0.01 Tsevat et al, Active TB case fatality rate 0.05 Centers for Disease Control and Prevention, Multidrug-resistant active TB case fatality rate 0.12 Chan et al, Test Characteristics Probability of adequate QFT-G or QFT-GIT sample 0.97 Brock et al, Probability of determinate first QFT-G result 0.97 Mori et al, ; Kang et al, ; Ferrara et al, ; Ferrara et al, ; Lee et al, ; Brock et al, ; Kobashi et al, ; Dewan et al, Probability of determinate first QFT-GIT result 1.00 Pai et al, Probability of determinate second QFT result 0.63 Grinsdale and Kawamura, Sensitivity of QFT-G 0.83 Mori et al, ; Kang et al, ; Ferrara et al, ; Ferrara et al, ; Lee et al, ; Brock et al, ; Kobashi et al, ; Dewan et al, ; Ravn et al, ; Goletti et al, Specificity of QFT-G 1.00 (0.96) Mori et al, ; Lee et al, ; Brock et al, ; Kobashi et al, ; Johnson et al, ; Taggart et al, Sensitivity of QFT-GIT 0.76 Pai et al, ; Tsiouris et al, Specificity of QFT-GIT 1.00 (0.96) Mori et al, ; Lee et al, ; Brock et al, ; Kobashi et al, ; Johnson et al, ; Taggart et al, Sensitivity of first TST 0.67 Mori et al, ; Kang et al, ; Ferrara et al, ; Ferrara et al, ; Lee et al, ; Kobashi et al, Specificity of first TST 0.98 (0.70) Mori et al, ; Kang et al, ; Lee et al, ; Brock et al, ; Kobashi et al, ; Ravn et al, ; Johnson et al, ; Taggart et al, Probability that first TST is read 0.88 National VHA data a Probability that second TST is placed 0.45 National VHA data a Probability that second TST is read 0.90 National VHA data a Sensitivity of second TST 0.90 Estimate Specificity of second TST 0.95 (0.50) Estimate Treatment of LTBI Probability that isoniazid is indicated 0.96 LoBue and Catanzaro, Probability that isoniazid is accepted by HCW 0.68 LoBue and Catanzaro, Probability of developing hepatitis Kopanoff et al, Hepatitis case fatality rate 0.02 Kopanoff et al, Probability that HCW finishes isoniazid 0.48 LoBue and Catanzaro, Efficacy of 3 mo of isoniazid 0.16 International Union Against Tuberculosis Committee on Prophylaxis, Efficacy of 9 mo of isoniazid 0.90 International Union Against Tuberculosis Committee on Prophylaxis, Efficacy of BCG vaccine 0.00 (0.51) Colditz et al, Utilities Well 1.00 Tsevat et al, Taking 9 mo of isoniazid therapy 0.99 Tsevat et al, Active hepatitis (1-mo duration) 0.85 Tsevat et al, Fatal hepatitis (1-mo duration) 0.40 Tsevat et al, Active TB (12-mo duration) 0.85 Tsevat et al, Fatal TB (6-mo duration) 0.50 Tsevat et al, Dead 0.00 Tsevat et al, Abbreviations: HCW, health care worker; LTBI, latent tuberculosis infection; QFT-G, QuantiFERON-TB Gold test (Cellestis Limited, Victoria, Australia); QFT-GIT, QuantiFERON-TB Gold in Tube test (Cellestis Limited); TB, tuberculosis; TST, tuberculin skin test; VHA, Veterans Health Administration. a Gary A. Roselle, MD; unpublished data; February 20, care reimbursement rates for professional and technical costs, and the VHA cost-accounting system for supplies. We used national VHA pay schedules to obtain HCW hourly wages and to calculate lost income because of missed work. 37 In our base-case analysis, we calculated the cost of running the QFT-G or QFT-GIT based on batches of 20 tests at a time and varied the number of samples run in sensitivity analyses across a range of 1 to 20. Utilities To account for short-term morbidities of disease, we used qualityof-life adjustments based on expert opinion from a study by 182

5 Table 2. Costs Variable Base-Case Value, $ Source TST First TST placement (includes costs of PPD, PPD supplies, EHN time to Smith and Velez, ; hospital accounting screen and place) Second TST placement (includes first TST placement costs, HCW time Smith and Velez, absent from work) TST reading (includes EHN time to read, clerical time, HCW time absent from work) Smith and Velez, QFT First QFT-G (includes phlebotomy supplies, EHN time to screen, Smith and Velez, ; hospital accounting; manufacturer phlebotomist time, QFT-G kit and standards, laboratory supplies, laboratory scientist time, clerical time) First QFT-GIT (includes phlebotomy supplies, EHN time to screen, Smith and Velez, ; hospital accounting; manufacturer phlebotomist time, QFT-GIT kit, laboratory supplies, laboratory scientist time, clerical time) QFT-G or QFT-GIT interpretation (includes EHN time to interpret and 5.11 Smith and Velez, contact, clerical time) Second QFT-G (includes first QFT-G cost, HCW time absent from work) Smith and Velez, ; hospital accounting; manufacturer Second QFT-GIT (includes first QFT-GIT cost, HCW time absent from Smith and Velez, ; hospital accounting; manufacturer work) Medical evaluation after positive test result (includes MD visit, CXR, HCW time absent from work) Smith and Velez, ; Salpeter et al, ; Medicare 39 Isoniazid Treatment Isoniazid (3 mo) (includes cost of treatment, 3 RN visits, HCW time Smith and Velez, ; Salpeter et al, absent from work) Isoniazid (9 mo) (includes cost of treatment, RN visits, HCW time absent from work) Smith and Velez, ; Salpeter et al, Mild hepatitis (includes 2 MD visits, 2 liver enzyme tests, HCW time absent from work) Severe hepatitis (includes 7 hospital days, 1 MD visit, HCW time absent from work) Treatment of active TB (includes 20 hospital days, outpatient medication, MD visits, daily observed treatment, contact tracing, HCW time absent from work) Treatment of MDR-TB (includes hospital days, outpatient medication, MD visits, daily observed treatment, contact tracing, HCW time absent from work) Hepatitis Smith and Velez, ; Salpeter et al, Smith and Velez, ; Salpeter et al, TB Smith and Velez, ; Salpeter et al, Smith and Velez, ; Rajbhandary et al, Abbreviations: CXR, chest radiograph; EHN, employee health nurse; HCW, health care worker; MD, doctor of medicine; MDR-TB, multidrug-resistant tuberculosis; PPD, purified protein derivative; QFT-G, QuantiFERON-TB Gold test (Cellestis Limited, Victoria, Australia); QFT-GIT, QuantiFERON-TB Gold in Tube test (Cellestis Limited); RN, registered nurse; TB, tuberculosis; TST, tuberculin skin test. Tsevat and colleagues 13 (Table 1). This factor was multiplied by the duration of illness or therapy to calculate the number of QALYs, which was then discounted at 3% per year. 10 SENSITIVITY ANALYSIS We performed 1-way deterministic sensitivity analyses, varying all of the input probabilities, costs, and utilities across clinically plausible ranges. We examined changes in the age of VHA HCWs and their wage rate to generalize our findings to the entire HCW population. We also performed probabilistic sensitivity analyses by means of second-order Monte Carlo simulation, using beta distributions to describe the uncertainty surrounding test characteristics for the 3 strategies. RESULTS BASE CASE Our base-case results show that the IGRAs are more effective and less costly than the TST, whether or not the HCW has been previously BCG vaccinated (Table 3). For non BCG-vaccinated HCWs, the incremental costeffectiveness ratio (ICER) of the QFT-G compared with the QFT-GIT is $14 092/QALY. For BCG-vaccinated HCWs, the ICER of the QFT-G compared with the QFT- GIT is $ /QALY. DETERMINISTIC SENSITIVITY ANALYSIS For both cohorts modeled, the TST strategy remains more costly and less effective than either QFT strategy across a wide range of values for each model variable, including the prevalence of LTBI and the probability of developing active TB. Varying the test characteristics of the TST, QFT-G, and QFT-GIT had little effect on our results, confirming that the TST is more costly and less effective than either QFT strategy. Results were not sensitive to changes in the probability that a second TST is necessary and that a HCW returns to have a first or sec- 183

6 Table 3. Base-Case Results Cohort Test Strategy Cost, $ Effectiveness, QALY Incremental Cost, $ Incremental Effectiveness Incremental Cost-effectiveness Ratio, $ Non BCG vaccinated QFT-GIT NA NA NA QFT-G TST DS BCG vaccinated QFT-GIT NA NA NA QFT-G TST DS Abbreviations: DS, dominated strategy (least effective, most costly); NA, not applicable; QALY, quality-adjusted life-year; QFT-G, QuantiFERON-TB Gold test (Cellestis Limited, Victoria, Australia); QFT-GIT, QuantiFERON-TB Gold in Tube test (Cellestis Limited); TST, tuberculin skin test. Incremental Cost-effectiveness Ratio, $/QALY QFT-GIT least costly QFT-G most effective Base case Non BCG-vaccinated cohort BCG-vaccinated cohort QFT-GIT dominates Threshold Probability That a QFT Is Cost-effective Compared With QFT-GIT, % Willingness to Pay, $/QALY BCG-vaccinated cohort Non BCG-vaccinated cohort Sensitivity of QFT-GIT Figure 3. Sensitivity analyses comparing the QuantiFERON-TB Gold test (QFT-G) (Cellestis Limited, Victoria, Australia) with the QuantiFERON-TB Gold in Tube test (QFT-GIT) (Cellestis Limited), while varying the sensitivity of the QFT-GIT. QALY indicates quality-adjusted life-year. ond TST read. Varying the ages of HCWs entering the VHA workforce from 25 to 55 years revealed that both IGRA strategies remain more effective and less costly than the TST in all age groups. When comparing the QFT-GIT and QFT-G strategies, if the sensitivity of the QFT-GIT exceeds that of the QFT-G, then the QFT-GIT becomes the most effective and least costly strategy. This is summarized in Figure 3. In sensitivity analyses varying the number of QFT samples run per batch, the QFT-G and QFT-GIT strategies were less costly and more effective than the TST as long as the number of samples run per batch is at least 12 for non BCG-vaccinated HCWs and at least 4 for BCG-vaccinated HCWs. If fewer samples are run per batch, the TST is the least costly but still the least effective. Sensitivity analyses varying the cost of the QFT-G and QFT-GIT kits revealed that each IGRA is less costly and more effective than the TST as long as the cost of a QFT-G kit is $32 or less and the cost of a QFT-GIT kit is $36 or less. Figure 4. Cost-effectiveness acceptability curves demonstrating the probability that the QuantiFERON-TB Gold test (QFT-G) (Cellestis Limited, Victoria, Australia) is cost-effective compared with the QuantiFERON-TB Gold in Tube test (QFT-GIT) (Cellestis Limited) by willingness to pay based on 1000 Monte Carlo simulations. QALY indicates quality-adjusted life-year. PROBABILISTIC SENSITIVITY ANALYSIS We performed probabilistic sensitivity analyses to examine uncertainty associated with test characteristics for the 3 strategies (Figure 4). The QFT-G and QFT-GIT strategies were cost saving compared with the TST strategy in 100% of our 1000 Monte Carlo simulations for both cohorts. In the non BCGvaccinated population, when comparing the QFT-G and QFT-GIT strategies, the QFT-G was cost saving 30% of the time, while the QFT-GIT was cost saving 3% of the time. The QFT-G had an ICER of less than $50 000/QALY gained in 47% of simulations. In the BCG-vaccinated population, the QFT-G was cost saving 21% of the time, while the QFT-GIT was cost saving 18% of the time. The QFT-GIT had an ICER of less than $50 000/QALY in 13% of simulations. COMMENT With the advent of new diagnostic technologies for screening LTBI and subsequent recommendations for their use, it is important to determine how they compare with the traditional TST clinically and economically. The

7 CDC guidelines 3 recommend that the QFT-G can be used in any instance in which the TST is used, including HCW screening. At the time of publication of these guidelines, the QFT-GIT had not yet been approved by the FDA. One limitation of these guidelines is that they were based on few published studies. The CDC guidelines postulate that the QFT-G might represent a cost-effective alternative to the TST and emphasize the need for an economic evaluation. 3 As such, we performed a cost-effectiveness analysis of 2 IGRA tests vis-à-vis the TST among newly hired HCWs and incorporated a significant number of subsequently published studies. We found that the QFT-G and QFT-GIT strategies are more effective and less costly than the TST, whether or not the HCW has been previously BCG vaccinated. Our findings are robust and insensitive to changes across a wide range of probabilities, costs, and utility estimates. Our sensitivity analyses indicate that the IGRA strategies are clinically and economically worthwhile among low- and high-prevalence populations. Our analysis supports the CDC recommendation of use of the QFT-G in HCW screening. To our knowledge, this study is the first to assess the cost-effectiveness of IGRAs vs the TST for detecting LTBI in HCWs. We modeled each step of the LTBI screening process, including the probabilities of having the TST read, of needing to do a 2-step TST, and of obtaining an indeterminate QFT test result. Time costs associated with missing work were included to truly reflect a societal perspective. We used QALYs as the effectiveness measure, the standard in cost-effectiveness analyses. 10 Our sensitivity analyses reveal that batch size is important in comparing the costs of IGRAs vs the TST. The QFT-G and QFT-GIT are less costly and more effective than the TST as long as the number of samples run per batch is at least 12 for non BCG-vaccinated HCWs and at least 4 for BCG-vaccinated HCWs. Therefore, efforts should be made to run the maximum number of QFT samples in a batch. Some institutions have designated certain mornings for blood to be collected for QFTs among HCWs to maximize batch size and to overcome the 12- hour limit posed by the QFT-G. The QFT-GIT kit eliminates this time limit, as it contains the antigens in the collection tubes. IGRA samples can also be saved and stored after the lymphocytes have been stimulated for up to 14 days before running the ELISA. Our calculations for the sensitivity of the QFT-GIT were based on only 2 existing published studies 24,30 in adults. Its sensitivity is expected to be greater than that of the QFT-G because the QFT-GIT contains an extra antigen. Yet, this was not the case because the sample sizes in the QFT-GIT studies were small and a substantial proportion of patients with active TB in 1 study 30 had coexisting human immunodeficiency virus, leading to a likely underestimation of the true sensitivity of the test among HCWs. Varying the sensitivity of the QFT-GIT shows that if its sensitivity is greater than that of the QFT-G, as expected, the QFT-GIT is the least costly and most effective strategy. With FDA approval of the QFT-GIT in October 2007, the QFT-G is being phased out and replaced by the QFT-GIT. We analyzed the QFT-G also because the data for this test are more extensive. Two previous cost-effectiveness analyses examining IGRAs focused on immigrants and on close contacts of patients with active TB. In the first analysis, Oxlade et al 42 studied immigrants to Canada and close contacts of patients with active TB. Strategies considered for the immigrant cohort included no screening, chest radiograph, TST, QFT-G, and TST followed by QFT-G if positive. Strategies considered for the close-contact cohort included no screening, TST, and QFT-G. The authors assumed that the TST and QFT-G had equivalent test characteristics in most groups but varied the specificities in the population receiving BCG vaccine as children. They also assumed the sensitivity of the QFT-G for LTBI to be 95% and calculated incremental cost per case prevented rather than per QALY. The authors found that sequential screening with the TST followed by QFT testing was more cost-effective in terms of cases prevented among immigrants and found that use of the TST led to greater cost savings among close contacts, except among previously BCG-vaccinated individuals. The second cost-effectiveness analysis performed by Diel et al 43 examined close contacts of patients with TB in Germany. Strategies compared included TST using a 5- or 10-mm cutoff, QFT-G, and TST followed by QFT-G if positive. Unlike our study, which pooled sensitivity and specificity data from multiple published studies, they derived QFT-G characteristics from a single study 6 in which the sensitivity estimate was high. Also, they estimated values for TST characteristics instead of deriving them from the literature 43 and measured effectiveness using life expectancy unadjusted for quality. Like the analysis by Oxlade et al, 42 their analysis found that the strategy combining the TST (with a 5-mm cutoff) followed by QFT-G testing if positive was the least costly and was equally effective, making it more cost-effective. Sequential TST and QFT testing was found to be the most cost-effective for the specific populations studied in the prior analyses. 42,43 We did not consider this appropriate for screening our new HCW population for many reasons. The prevalence of LTBI is expected to be lower among HCWs than among immigrants and close contacts. Unlike our analysis, the previous analyses did not include 2-step testing and assumed a 100% return rate for TST reading. Enforcement of TST reading might be more stringent among the immigrant and close-contact populations, rendering better adherence to TST reading more plausible. In our analysis, adding the direct costs for additional visits to the costs associated with missed work made the TST strategy more costly for both of our cohorts. Therefore, it would be even more costly to add QFT testing as a potential third or fifth step. Our findings reinforce the recommendation in the CDC guidelines that the QFT-G be used in place of (and not in addition to) the TST. 3(p52) An inherent limitation of decision analyses is their dependence on the quality and accuracy of their base-case modeling parameter values. In contrast to the previous cost-effectiveness analyses by Oxlade et al 42 and by Diel et al, 43 we used pooled data from multiple studies for our modeling parameter values. The publication of more studies, especially regarding the test characteristics of the QFT- GIT, would increase the quality and accuracy of future cost-effectiveness analyses. 185

8 To our knowledge, no longitudinal studies to date have examined the predictive value of IGRAs, 7 making it difficult to predict the incidence of active TB in persons with positive and negative IGRA results compared with the TST. As discussed in the CDC guidelines, 3 this information requires years of observational study. Because there is no gold standard for the diagnosis of LTBI, assessing the sensitivity and specificity of each test through extrapolation from the published literature was difficult. Studies 6,17-24,26-30 calculating sensitivity used patients with culture-confirmed active TB as the gold standard. Such patients represent less-thanoptimal surrogates because of the known reduction in cell-mediated response in patients with active TB Hence, the sensitivity of the IGRAs for LTBI is likely underestimated. Specificity calculations used populations consisting of healthy lifelong residents of low-incidence countries without any occupational, travel, or other exposures to TB. 6,17,20-22,26,28,29 The lack of a gold standard prevents verification of LTBI, and using these subjects may have overestimated specificity. Our study used data for some input variables from the VHA, which can be interpreted as an advantage and as a disadvantage. We used actual national VHA pay schedules 37 and data on HCW TST screening. Although some may question the generalizability of our results to their HCW populations, our probabilities for HCW return for TST reading and the need for 2-step testing are comparable to the published findings at other institutions Furthermore, our findings can be generalized across different institutions that may have a higher or lower prevalence of LTBI, as varying this input variable did not change the results of our analysis. Although we chose a 35-year-old female nurse as the prototypical HCW, our sensitivity analyses varying her age and wages across wide ranges did not affect our findings that both IGRAs were more effective and less costly than the TST. Therefore, our findings are generalizable to the whole HCW population. Our model did not examine the transmission of TB, which would have added societal costs and benefits. We did not include strategies examining subsequent annual TST or QFT testing, focusing instead on the first screening test for new HCWs. Another potential benefit of QFT testing is that more HCWs will believe a positive test result and accept isoniazid treatment than with a positive TST result. If so, then QFT screening would be more effective regarding QALYs gained. However, to our knowledge, no current data comparing acceptance rates of a positive QFT test vs TST result are available, rendering it necessary to exclude this from our analysis. We conclude that the QFT-G and QFT-GIT are clinically and economically worthwhile alternatives to the TST in testing HCWs for LTBI, as both IGRA strategies are more effective and less costly than the TST strategy. IGRA tests should be considered for screening non BCGvaccinated and BCG-vaccinated new HCWs for LTBI. Efforts should be made to maximize the number of samples per IGRA batch to minimize costs. Accepted for Publication: June 15, Correspondence: Marie A. de Perio, MD, Division of Infectious Diseases, University of Cincinnati College of Medicine, 231 Albert Sabin Way, Box , Cincinnati, OH (deperiom@hotmail.com). Author Contributions: Dr de Perio had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: de Perio, Tsevat, Roselle, and Kralovic. Acquisition of data: de Perio, Roselle, and Kralovic. Analysis and interpretation of data: de Perio, Tsevat, and Eckman. Drafting of the manuscript: de Perio, Tsevat, and Eckman. Critical revision of the manuscript for important intellectual content: de Perio, Tsevat, Roselle, Kralovic, and Eckman. Statistical analysis: de Perio and Eckman. Administrative, technical, and material support: de Perio, Roselle, and Kralovic. Study supervision: Tsevat, Roselle, Kralovic, and Eckman. Financial Disclosure: None reported. Previous Presentations: This study was presented in part at the 45th Annual Meeting of the Infectious Diseases Society of America; October 5, 2007; San Diego, California (abstract 370); and at the 29th Annual Meeting of the Society for Medical Decision Making; October 22, 2007; Pittsburgh, Pennsylvania. REFERENCES 1. Maloney SA, Pearson ML, Gordon MT, Del Castillo R, Boyle JF, Jarvis WR. Efficacy of control measures in preventing nosocomial transmission of multidrugresistant tuberculosis to patients and health care workers. Ann Intern Med. 1995; 122(2): Jensen PA, Lambert LA, Iademarco MF, Ridzon R; CDC. Guidelines for preventing the transmission of Mycobacterium tuberculosis in health-care settings, MMWR Recomm Rep. 2005;54(RR-17): Mazurek GH, Jereb J, Lobue P, Iademarco MF, Metchock B, Vernon A; Division of Tuberculosis Elimination, National Center for HIV, STD, and TB Prevention, Centers for Disease Control and Prevention (CDC). Guidelines for using the QuantiFERON-TB Gold test for detecting Mycobacterium tuberculosis infection, United States, 2005 [published correction appears in MMWR Morb Mortal Wkly Rep. 2005;54(50):1288]. MMWR Morb Mortal Wkly Rep. 2005;54(RR-15): Pai M, Riley LW, Colford JM Jr. Interferon- assays in the immunodiagnosis of tuberculosis: a systematic review. Lancet Infect Dis. 2004;4(12): Andersen P, Munk ME, Pollock JM, Doherty TM. Specific immune-based diagnosis of tuberculosis. Lancet. 2000;356(9235): Mori T, Sakatani M, Yamagishi F, et al. Specific detection of tuberculosis infection: an interferon- based assay using new antigens. Am J Respir Crit Care Med. 2004;170(1): Menzies D, Pai M, Comstock G. Meta-analysis: new tests for the diagnosis of latent tuberculosis infection: areas of uncertainty and recommendations for research [published correction appears in Ann Intern Med. 2007;146(9):688]. Ann Intern Med. 2007;146(5): Department of Veterans Affairs. Volume II: medical programs & information technology programs: congressional submission, FY /summary/2009/volume_2-medical_programs_and_information_technology.pdf. Accessed April Department of Veterans Affairs. VA Handbook 5019: Occupational Health Services. Accessed February 20, Gold MR, Siegel JE, Russell LB, Weinstein MC, eds. Cost-effectiveness in Health and Medicine. New York, NY: Oxford University Press Inc; Lau J, Kassirer JP, Pauker SG. Decision Maker 3.0: improved decision analysis by personal computer. Med Decis Making. 1983;3(1): Centers for Disease Control and Prevention (CDC). Trends in tuberculosis incidence United States, MMWR Morb Mortal Wkly Rep. 2007;56(11): Tsevat J, Taylor WC, Wong JB, Pauker SG. Isoniazid for the tuberculin reactor: take it or leave it. Am Rev Respir Dis. 1988;137(1): Centers for Disease Control and Prevention. Reported Tuberculosis in the United 186

9 States, Atlanta, GA: US Dept of Health and Human Services; September Chan ED, Laurel V, Strand MJ, et al. Treatment and outcome analysis of 205 patients with multidrug-resistant tuberculosis. Am J Respir Crit Care Med. 2004; 169(10): Brock I, Rugwald M, Lundgren B, Westh H, Mathiesen LR, Ravn P. Latent tuberculosis in HIV positive, diagnosed by the M. tuberculosis specific interferon- test. Resp Res. 2006;7(1):e56. Accessed October 8, Kang YA, Lee HW, Yoon HI, et al. Discrepancy between the tuberculin skin test and the whole-blood interferon assay for the diagnosis of latent tuberculosis infection in an intermediate tuberculosis burden country. JAMA. 2005;293 (22): Ferrara G, Losi M, D Amico R, et al. Use in routine clinical practice of two commercial blood tests for diagnosis of infection with Mycobacterium tuberculosis: a prospective study. Lancet. 2006;367(9519): Ferrara G, Losi M, Meacci M, et al. Routine hospital use of a commercial whole blood interferon- assay for tuberculosis infection. Am J Respir Crit Care Med. 2005;172(5): Lee JY, Choi HJ, Park IN, et al. Comparison of two commercial interferon- assays for diagnosing Mycobacterium tuberculosis infection. Eur Respir J. 2006; 28(1): Brock I, Munk ME, Kok-Jensen A, Andersen P. Performance of whole blood IFNgamma test for tuberculosis diagnosis based on PPD or the specific antigens ESAT-6 and CPT-10. Int J Tuberc Lung Dis. 2001;5(5): Kobashi Y, Obase Y, Fukuda M, Yoshida K, Miyashita N, Oka M. Clinical reevaluation of the QuantiFERON TB-2G test as a diagnostic method for differentiating active tuberculosis from nontuberculous mycobacteriosis. Clin Infect Dis. 2006; 43(12): Dewan PK, Grinsdale J, Kawamura LM. Low sensitivity of a whole-blood interferon- release assay for detection of active tuberculosis. Clin Infect Dis. 2007; 44(1): Pai M, Joshi R, Bandyopadhyay P, et al. Sensitivity of a whole-blood interferongamma assay among patients with pulmonary tuberculosis and variations in T-cell responses during anti-tuberculosis treatment. Infection. 2007;35(2): Grinsdale J, Kawamura LM. Community-based tuberculosis screening using QuantiFERON-TB Gold: the San Francisco experience. Paper presented at: Rethinking the Epidemiology of Tuberculosis Infection: The First Global Symposium on Interferon-gamma Assays; February 22, 2007; Vancouver, British Columbia, Canada. 26. Ravn P, Munk ME, Andersen AB, et al. Prospective evaluation of a whole-blood test using Mycobacterium tuberculosis specific antigens ESAT-6 and CFP-10 for diagnosis of active tuberculosis. Clin Diagn Lab Immunol. 2005;12(4): Goletti D, Vincenti D, Carrara S, et al. Selected RD1 peptides for active tuberculosis diagnosis: comparison of a gamma interferon whole-blood enzyme-linked immunosorbent assay and an enzyme-linked immunospot assay. Clin Diagn Lab Immunol. 2005;12(11): Johnson PD, Stuart RL, Grayson ML, et al. Tuberculin-purified protein derivative, MPT-64, and ESAT-6 stimulated gamma interferon responses in medical students before and after Mycobacterium bovis BCG vaccination and in patients with tuberculosis. Clin Diagn Lab Immunol. 1999;6(6): Taggart EW, Hill HR, Ruegner RG, Litwin CM. Evaluation of an in vitro assay for interferon gamma production in response to the Mycobacterium tuberculosis synthesized peptide antigens ESAT-6 and CFP-10 and the PPD skin test. Am J Clin Pathol. 2006;125(3): Tsiouris SJ, Coetzee D, Toro PL, Austin J, Stein Z, El-Sadr W. Sensitivity analysis and potential uses of a novel gamma interferon release assay for the diagnosis of tuberculosis. J Clin Microbiol. 2006;44(8): LoBue PA, Catanzaro A. Effectiveness of a nosocomial tuberculosis control program at an urban teaching hospital. Chest. 1998;113(5): Kopanoff DE, Snider DE, Caras GJ. Isoniazid-related hepatitis. Am Rev Respir Dis. 1978;117(6): International Union Against Tuberculosis Committee on Prophylaxis. Efficacy of various durations of isoniazid preventive therapy for tuberculosis: five years of follow-up in the IUAT trial. Bull World Health Organ. 1982;60(4): Colditz GA, Brewer TF, Berkey CS, et al. Efficacy of BCG vaccine in the prevention of tuberculosis: meta-analysis of the published literature. JAMA. 1994; 271(9): Comstock GW, Woolpert SF, Livesay VT. Tuberculosis studies in Muscogee County, Georgia: twenty-year evaluation of a community trial of BCG vaccination. Public Health Rep. 1976;91(3): Arias E. United States life tables, Natl Vital Stat Rep. 2006;54(14): Smith MW, Velez JP. A guide to estimating wages of VHA employees. July Accessed March 1, Salpeter SR, Sanders GD, Salpeter EE, Owens DK. Monitored isoniazid prophylaxis for low-risk reactors older than 35 years of age: a risk-benefit and costeffectiveness analysis. Ann Intern Med. 1997;127(12): Centers for Medicare and Medicaid Services. National physician fee schedule relative value file Accessed February 20, Rajbhandary SS, Marks SM, Bock NN. Costs of patients hospitalized for multidrugresistant tuberculosis. Int J Tuberc Lung Dis. 2004;8(8): Bureau of Labor Statistics, US Department of Labor. Consumer Price Index. http: //data.bls.gov/cgi-bin/cpicalc.pl. Accessed March Oxlade O, Schwartzman K, Menzies D. Interferon-gamma release assays and TB screening in high-income countries: a cost-effectiveness analysis. Int J Tuberc Lung Dis. 2007;11(1): Diel R, Nienhaus A, Loddenkemper R. Cost-effectiveness of interferon- release assay screening for latent tuberculosis infection treatment in Germany. Chest. 2007;131(5): Holden M, Dubin MR, Diamond PH. Frequency of negative intermediatestrength tuberculin sensitivity in patients with active tuberculosis. N Engl J Med. 1971;285(27): Battershill JH. Cutaneous testing in the elderly patient with tuberculosis. Chest. 1980;77(2): Howard WL, Klopfenstein MD, Steininger WJ, Woodruff CE. The loss of tuberculin sensitivity in certain patients with active pulmonary tuberculosis. Chest. 1970;57(6): Frenzel EC, Thomas GA, Hanna HA. The importance of two-step tuberculin skin testing for newly employed healthcare workers. Infect Control Hosp Epidemiol. 2006;27(5): Hallak KM, Schenk M, Neale AV. Evaluation of the two-step tuberculin skin test in health care workers at an inner-city medical center. J Occup Environ Med. 1999; 41(5): Sepkowitz KA, Feldman J, Louther J, Rivera P, Villa N, DeHovitz J. Benefit of twostep PPD testing of new employees at a New York City hospital. Am J Infect Control. 1997;25(3): Swinker M. Prevalence of positive tuberculin skin reactions in a two-step testing program at a tertiary-care center in eastern North Carolina. Infect Control Hosp Epidemiol. 2000;21(1): Rattner SL, Fleischer JA, Davidson BL. Tuberculin positivity and patient contact in healthcare workers in the urban United States. Infect Control Hosp Epidemiol. 1996;17(6): Ball R, Van Wey M. Tuberculosis skin test conversion among health care workers at a military medical center. Mil Med. 1997;162(5):