0
Original Research: RESPIRATORY INFECTIONS |

Cost-effectiveness of Interferon-γ Release Assay Screening for Latent Tuberculosis Infection Treatment in Germany* FREE TO VIEW

Roland Diel, MD, MPH; Albert Nienhaus, MD, MPH; Robert Loddenkemper, MD, FCCP
Author and Funding Information

*From the School of Public Health (Dr. Diel), University of Düsseldorf, Düsseldorf, Germany; Institution for Statutory Accident Insurance and Prevention in the Health and Welfare Services (Dr. Nienhaus), Hamburg, Germany; and the German Central Committee Against Tuberculosis (Dr. Loddenkemper), Berlin, Germany.

Correspondence to: Roland Diel, MD, MPH, School of Public Health, c/o Institute for Medical Sociology, Heinrich Heine University, Post Box 101007, D-40001 Düsseldorf, Germany; e-mail: Roland.Diel@uni-duesseldorf.de



Chest. 2007;131(5):1424-1434. doi:10.1378/chest.06-2728
Text Size: A A A
Published online

Objectives: To assess the cost-effectiveness of the new QuantiFERON-TB Gold In-Tube (QFT-G) [Cellestis; Carnegie, VIC, Australia] assay for screening and treating of persons who have had close contact with tuberculosis (TB) patients and are suspected of having latent tuberculosis infection (LTBI) [hereafter called close-contacts] in Germany.

Methods: The health and economic outcomes of isoniazid treatment of 20-year-old close-contacts were compared in a Markov model over a period of 20 years, using two different cutoff values for the tuberculin skin test (TST), the QFT-G assay alone, or the QFT-G assay as a confirmatory test for the TST results.

Results: QFT-G assay-based treatment led to cost savings of $542.9 and 3.8 life-days gained per LTBI case. TST-based treatment at a 10-mm induration size cutoff gained $177.4 and 2.0 life-days gained per test-positive contact. When the cutoff induration size for the TST was reduced to 5 mm, the incremental cost-effectiveness ratio fell below the willingness-to-pay threshold ($30,170 per life-years gained) but resulted in unnecessary treatment of 77% of contacts owing to false-positive TST results. Combination with the 5-mm induration size TST cutoff value compared to the results of the QFT-G assay alone reduced the total costs per 1,000 contacts by 1.8% to $222,869. The number treated to prevent 1 TB case was 22 for the two QFT-G assay-based procedures, 40 for the TST at a cutoff induration size of 10 mm, and 96 for the TST at a cutoff induration size of 5 mm. When the sensitivity rates of the TST and the QFT-G assay were compounded, the QFT-G assay strategy alone was slightly less costly (0.6%) than the two-step approach.

Conclusions: Using the QFT-G assay, but especially combining the QFT-G assay following the TST screening of close-contacts at a cutoff induration size of 5 mm before LTBI treatment is highly cost-effective in reducing the disease burden of TB.

Figures in this Article

Many clinical studies15 of the screening of those persons for latent tuberculosis infection (LTBI) who have come into close contact with tuberculosis (TB) patients [called hereafter close-contacts] using the new, highly specific interferon-γ release assays (IGRAs) have been published, but no study has investigated the long-term economic consequences of subsequent LTBI treatment. Two recent cost-minimization studies have suggested possible cost savings using the T-SPOT.TB assay (Oxford Immunotec; Abingdon, UK)6 and the QuantiFERON-TB Gold In-Tube assay (QFT-G) [Cellestis; Carnegie, VIC, Australia]7 for initial screening, but did not investigate the cost implications of treatment. Therefore, we conducted a cost-effectiveness analysis of alternative LTBI screening strategies followed by isoniazid (INH) treatment under a range of different conditions, comparing existing programs based on the tuberculin skin test (TST) and the QFT-G assay results.

In some European countries, namely, the United Kingdom and Switzerland, the currently published recommendations89 suggest the implementation of IGRAs as confirmatory tests for TST-positive contacts to determine the number of true LTBI patients with respect to treatment intentions. In Germany, new guidelines are in preparation, and to aid that process this study was based on current German epidemiologic and cost data.

This study is a theoretical modeling study; therefore, it was not necessary to seek any patient consent or the approval of any internal review boards.

Screening Tests

The following strategies were compared: (1) performing the TST as a stand-alone tool to diagnose LTBI using the German standard induration cutoff size (> 5 mm); (2) using the TST with a higher cutoff induration size of > 10 mm; (3) completely replacing the TST by using the QFT-G assay alone; and (4) using the TST with a cutoff induration size of > 5 mm, followed by a QFT-G assay in all TST-positive individuals before considering LTBI treatment.

Modeling Screening Variables of QFT-G Assay and Mantoux TST

Method-related data for this analysis were taken from a recent prospective side-by-side study10 comparing the TST with the QFT-G assay among 309 immunocompetent adult close-contacts with a mean (± SD) age of 28.5 ± 10.5 years under routine program conditions at the Department of Public Health in Hamburg between May 1 and October 31, 2005. This population contained a high proportion of persons who had been vaccinated with bacillus Calmette-Guerin (BCG)10and comprised nearly equivalent numbers of male contacts (48.2%) and female contacts (51.8%). The Mantoux TST was applied according to the German national guidelines11; those contacts who were considered to be positive for LTBI (if induration size was > 5 mm) were offered preventive treatment with a dosage of 300 mg of INH for 9 months.12 For the QFT-G enzyme-linked immunosorbent assay, a 2-mL blood sample was taken and analyzed in a local laboratory according to the instructions of the manufacturer for the QFT-G method (www.cellestis.com); the cutoff value for a positive test was an interferon-γ concentration of ≥ 0.35 IU/mL in the TB antigen sample, after subtracting the background level.

In persons who are at low risk for TB infection, the specificity of the QFT-G assay overall has been found to be > 98%.2 As this group may contain TB-infected persons, the specificity of the QFT-G assay for LTBI (defined as the probability of a negative test result if the infection is truly absent) may approach 100%. Taking 100% as the best possible specificity value, the number of QFT-G assay-positive contacts determines reciprocally (as the “gold standard”) the prevalence of true-positive (TP) results, because the proportion of contacts without infection who are incorrectly classified as test-positive is zero.

The TST has a well-documented lack of specificity at cutoff values of 5 and 10 mm, particularly in populations with significant numbers of persons who have been vaccinated with BCG, which allows the presumption that if there are differences between the number of QFT-G assay-positive and TST-positive contacts, all positive TST results at the particular cutoff beyond this number may be considered to be false-positive (FP) results.

On the other hand, given the high specificity of the QFT-G assay, if contacts who tested positive with the QFT-G assay are missing a corresponding positive TST result, it is reasonable to estimate the sensitivity of the TST (defined as the probability of screening positive if the infection is truly present) relative to the QFT-G assay. In this assumption, a QFT-G assay positive result is almost certainly indicative of true infection. For the QFT-G assay itself, however, the absolute sensitivity for LTBI cannot be clearly defined due to the lack of a reference test. Thus, it is deemed acceptable to use the sensitivity of the QFT-G assay in active TB cases, which is equal to a baseline value of 90% in current studies.13

The raw data from the results of both tests used in our reference study (with two TST results being obtained to consider the two possible cutoff values) were normalized to hypothetical cohort sizes of 1,000 contact individuals each. We were then also able to calculate the positive predictive value (PPV) for each strategy and, by dividing the assumed number of contacts with false-negative (FN) test results by the number of contacts who tested negative and then subtracting the quotient from 100%, the negative predictive value (NPV) for each strategy.

As shown in Table 1 , of the 1,000 contacts, 443 (44.3%) and 207 (20.7%), respectively, had positive TST results at cutoff values of 5 and 10 mm; 100 contacts (10.0%) had positive QFT-G assay findings. Only 20.3% of the TST-positive individuals (90 of 443) and 39.1% of the TST-positive individuals (81/207), respectively, at cutoff values of 5 and 10 mm also had positive QFT-G assay results, suggesting a low PPV, which is defined as the number of TP results divided by the number of TP and FP results. There were 10 and 19 individuals, respectively, who were QFT-G assay positive but TST result negative at respective cutoff values of 5 and 10 mm, indicating that the TST test has a sensitivity similar to the QFT-G assay of 90.0% (at TST cutoff induration size of 5 mm) and 81.0% (at TST cutoff induration size of 10 mm).

In this study, there were 25 TST-positive contacts (8.1%) with a cutoff induration size of > 15 mm. Of these contacts, all 6 non–BCG-vaccinated persons had simultaneously positive QFT-G assay results, as did 8 of the 19 persons who were vaccinated with BCG. This suggests that in the remaining 11 QFT-G assay-negative contacts in question the TST reactions could be attributed to the previous BCG vaccination. Although, principally, positive TST reactions of > 15 mm are considered to represent the presence of LTBI, a study published in 200514 has suggested that 18 mm is a more appropriate induration size for such a supposition. Considering that, and that our reference study found all 10 subjects with TST results of > 18 mm induration size to be also QFT-G assay positive, we saw no evidence that TST-positive/QFT-G assay discordance (ie, FN QFT-G assay results) should be considered in our model.

Assuming a sensitivity for the QFT-G assay of 90%, the 31 contacts (10%) who tested positive on the QFT-G assay comprise only 90% of the TP results within the contact investigation. Therefore, a further 1% of total contacts tested have to be considered as having FN test results, and the total number of LTBI cases will expand to 110 persons per 1,000 contacts to be screened. Since in our model of the QFT-G screening strategy 100 of a cohort of 1,000 contacts (ie, 10%) have tested positive and 900 have tested negative, the 10 contacts with FN test results will result in a NPV, which is defined as the number of true-negative (TN) test results divided by the number of TN results and FN test results, of 890 of 900 contacts with true-negative test results (ie, a NPV of 98.88%) and a corresponding proportion of 10 of 900 contacts who tested FP (0.0111).

Starting our calculations with the same number of 1,000 LTBI patients to be screened in each strategy, there are two modeling possibilities. The first option is to assume that the 10 assumed FN QFT-G assay results “hide” within the same number of subjects with positive TST results (ie, they are picked up not by the QFT-G assay, but by the TST and are misclassified as having FP results). Choosing this alternative would slightly increase the PPV of the TST, which would yield 100 of 443 contacts with TP results instead of 90 of 443 contacts with TP results at a cutoff of 5 mm. Since the sensitivity of the TST at a cutoff of 5 mm and that of the QFT-G assay would then be identical, making the strategies more easily comparable, we used this option to provide base-case probabilities.

The alternative option is to assume that the contacts with FN QFT-G assay results are not picked up by either the QFT-G assay or the TST, and thus the second option is simply to add the number of the 10 FN QFT-G assay results to the negative results of the TST. Since the FN TST results are directly related to the side-by-side QFT-G assay values as conditional probabilities, this would correspondingly further reduce TST sensitivity. The latter option comparator was taken for sensitivity analysis (see “Sensitivity Analysis” section).

Decision Analysis Model

We developed a computer-based Markov model (TreeAge Pro 2006 Healthcare Module; TreeAge Software Inc; Williamstown, MA) to simulate the economic and clinical outcomes of screening and LTBI treatment or nontreatment dependence on the test results of each strategy in a cohort of 20-year-old close-contacts. One of the four alternative decision trees (showing the QFT-G assay screening strategy) is presented in Figure 1 .

If test results are positive, subjects are assumed to have LTBI and, after active TB has been excluded by a chest radiograph examination, they are offered treatment with INH (see preceding section). It is assumed that this course of INH will provide protection over 20 years15 and that reinfection with Mycobacterium tuberculosis (MTB) will not occur. Accordingly, the following two subgroups can be followed prospectively according to the PPV of each strategy (which is in this case 100% for the QFT-G assay): (1) those who completed LTBI treatment; and (2) those who refused to receive LTBI treatment.

Contacts who test negative are considered not to be infected with MTB, with the exception of those whose test results will be FN, depending on the assumed sensitivity of the test. Whereas the costs associated with the first subgroup of the correctly identified contacts whose test results were negative will be limited to the cost of screening, the latter FN group will incur active TB cases and disease-associated costs over the next 20 years with the same probability as will the persons whose test results are truly positive who refuse to be treated.

Five clinical states with their associated costs were included in the model representing the various possible states of close-contacts after they have been infected with MTB. These states were as follows: (1) asymptomatic LTBI, incurring only the cost of initial screening (TST based or QFT-G assay based) and, depending on whether they accepted LTBI treatment or not, the cost of INH medication and corresponding clinical controls; (2) active illness due to reactivation of the disease, to which some of these LTBI patients progress with a transition probability (tpReact); (3) death due to the disease itself (including consequent conditions) with the transition probability (tpDcm) or in contrast to this (4) survival after recovery without sequelae (1 − tpDcm); or (5) death due to all-cause mortality, excluding TB disease represented by age-dependent life expectancy, with a probability (tpDn) that affects all members of the cohort equally and is only modified by the members′s age in each circle. This study considers an immediately implemented LTBI treatment (or nontreatment) following testing and is divided into 20 equal yearly increments during which transitions between the various health states may occur.

Probabilities

Transition probabilities (ie, probabilities of passing from one health state to another) and the estimation of INH efficacy were derived from Diel et al15and are shown in Table 2 . The risk parameters of TB reactivation (tpReact), depending largely on the age of the infected person and the size of the induration produced by a TST, were derived by Horsburgh16 for the group of contacts who were 25 to 35 years of age and were separated by the three induration diameters in question (ie, cutoff induration sizes of 5 and 10 mm as the TST baseline and 15 mm for a sensitivity analysis).

Compliance

In contrast to other Western industrialized countries (eg, the United States17or Switzerland18), there has to date been no assessment of valid data on compliance for starting LTBI treatment among close-contacts in Germany. Therefore, we conducted a sensitivity analysis considering this important key parameter.

Secondary LTBI Patients

To avoid possible bias in comparing the benefits of different screening tools, we included the number of subsequent infections and TB cases averted by LTBI treatment within the period of 20 years in a dynamic model approach. Thus, we integrated subsequent LTBI cases into to the decision tree, assuming that each secondary LTBI patient would be screened and treated (or not treated) in the same way as a primary LTBI patient. However, the number of new latent cases depends on the number of the contacts as well as on the degree of infectiousness of the active TB case; therefore, they may differ substantially. In Germany in 2004, 57.9% of all TB cases were culture-confirmed pulmonary TB cases; of these, nearly one half (46.3%) were smear-positive for TB.19Cautiously assuming that a typical smear-positive TB patient will infect five contacts20but that a smear-negative, culture-confirmed TB patient will infect only one contact,21 the weighted average of new LTBI cases for each person with newly developed active TB is three. Consequently, this number was taken for the sensitivity analysis.

Derivation of Costs of TB Disease

All costs are expressed in 2004 US dollars. Costs were assessed from the societal perspective, and included direct medical costs, measuring both inpatient and outpatient costs, and the indirect costs arising from the loss of productivity. This analysis, calculating total costs due to the disease of $23,413 has been published in detail elsewhere.15

Costs of LTBI Screening and Treatment

The costs of LTBI testing and treatment were recently published in a cost-minimization study6 from the German Public Health Service perspective. Asymptomatic infection is assumed to produce no cost (except the cost of testing). The cost of testing comprised the labor cost for the staff performing the Mantoux TST, the material cost of the vial and associated consumables for each TST, and the costs of an initial chest radiograph to rule out active TB prior to treatment and a medical consultation at the end of the course of treatment ($145.99). For the QFT-G assay, besides the costs for a chest radiograph and consultation, the cost for drawing blood, the cost of the screening kit, reagents, and laboratory technician’s fees for each QFT-G assay test ($171.78) have to be considered. If a contact chooses the option of LTBI treatment, the average costs are $250.28, including the cost of 9 months of INH therapy, the cost of visits to the clinician, and the cost of liver-function tests before and during the treatment period.15

A 9-month course of INH chemoprevention was not assigned a cost value because toxic hepatitis as a side effect is highly age related22 and would mostly occur in those persons > 50 years of age and in those persons with abnormal evaluation of liver values (ie, glutamic-oxaloacetic transaminase, glulamic-pyruvic transaminase, and bilirubin levels). The German guidelines do not recommend INH treatment for such patients

Cost-effectiveness and the Number Needed To Treat

Following the recommendations of the Panel of Cost-effectiveness in Health and Medicine,23 we measured the comparative performance of alternative intervention possibilities by using the incremental cost-effectiveness ratios (ICERs) of the different strategies, defined as (CST − CSN) divided by (EST − ESN), where CST − CSN is the difference between the sum of the costs of each LTBI screening test and treatment minus the costs for screening by this test and no treatment over the 20-year period, and EST − ESN is the difference between the effectiveness of these interventions. Effectiveness in this context was measured in terms of saved life expectancy (generally converted to life-years gained [LYG]) to yield the net cost required to increase by 1 life-year (LY) compared with the next less costly intervention. Negative numbers will identify cost savings (ie, if an intervention costs less and is more effective than its comparator), while positive numbers indicate additional expenditure per outcome unit. That means that the higher the ratio, the less cost-effective the intervention.

Future costs and LYG were discounted at an annual rate of 3%. As commonly used, a rough benchmark of $50,000 per LYG (called willingness-to-pay) is considered to be the range in which an intervention is supposed to be cost-effective.24

Furthermore, we determined the average cost-effectiveness, which was defined as the costs per case prevented with a given strategy, and demonstrated the total costs and cost components for each strategy presented (ie, the contribution of cost by treatment, the cost due to negative test results, and cost of overlooked TB cases among contacts with FN results with undetected LTBI due to the different NPVs of a strategy). Finally, we then calculated the number needed-to-treat (NNT) [ie, the number of subjects treated to prevent one TB case that is equal to 1 divided by the absolute risk difference]. Thus, for the NNT we determined the number of contacts whose test results were positive by each screening tool and were subsequently treated with INH and the number of TB cases occurring with and without treatment of these contacts.

Sensitivity Analysis

One-way sensitivity analysis involves varying the value of a key parameter over a wide range to determine whether the optimal strategy changes. The variables chosen in our model included the sensitivity of the QFT-G assay (0.95 instead of 0.90, as described above), the sensitivity of the TST (see “Modeling Screening Variables of QFT-G Assay and Mantoux TST” section), the annual probability of progression to disease (assuming that of QFT-G assay to be the same as that of a 15-mm TST induration), the compliance with LTBI treatment, and the number of secondary LTBI patients arising from active TB cases among our subject contacts during the course of our study. We also performed a threshold analysis to determine what reduction in total cost of TB treatment a screening strategy must bring in order to result in cost neutrality (the point at which the reduction in the total costs of illness equal the incremental screening and treatment costs per LYG vs the nontreatment costs of this strategy). We did not vary the specificity of the QFT-G assay (eg, to 0.98 instead of 1.00), because in this case the number of LTBI cases would decrease and the cost of each strategy accordingly in the same proportion.

The projected base-case outcomes of the different screening strategies are presented in Table 3.

Using QFT-G Assay Alone

In the presence of the highly specific screening made possible by the QFT-G assay, we projected (per thousand 20-year-old close-contacts) a total of 5.8 TB cases that would arise within 20 years, if the 100 QFT-G assay-positive and truly MTB-infected contacts remained untreated. Of these, 4.6 cases could be prevented by a full course of treatment; but in addition, assuming a 0.90 sensitivity as the base-case value, 10% of all predicted cases of TB (ie, 0.58 cases) would also have been unrecognized because of FN test results.

For the high reactivation probability scenario for the QFT-G assay, which alternatively supposed the same reactivation probability as that of a TST result of at least 15 mm in this age group, 10.6 cases would occur, and 8.4 of them would be prevented by INH therapy; 1.06 cases arising among the 10 LTBI patients with FN test results would be missed. Owing to the previous finding that 90% of close-contacts had a negative test result, the NNT in order to prevent 1 TB case was 22 (95% confidence interval [CI], 16 to 33) for a normal reactivation probability and 12 (95% CI, 10 to 16) for a high reactivation probability. The ICER, comparing those patients who were screened and treated following a positive QFT-G assay test result to the nontreated patients, was $542.9 less expensive per contact treated in addition to a saving of 0.0104 LYs. The screening and treatment strategy, therefore, dominated the screening-but-not-treating option. This option further improved to lower expenditures of $1,192.20 and to raise life expectancy to 0.0190 LY per contact treated when the high reactivation probability is used (because of the absolutely higher number of predicted TB cases at only a small incremental increase in treatment costs compared with untreated contacts). Thus, the annual cost avoided due to the treatment of the 100 QFT-G assay-positive contacts, calculated as ($542.9 × 100)/20 years, was $2,715 or $5,961, depending on the expected reactivation rates.

Investigating the compartment costs of screening, under base-case assumptions, $164,550 of a total of $227,014 in costs (72.5%) is expended on negative test results, and $11,663 is expended on the consequences of FN results (5%). Even reducing the cost of TB treatment in sensitivity analysis to a threshold of $7,378 (ie, by more than two thirds to what appears to be an extremely unrealistic value) would result in a cost saving when LTBI treatment is offered. This would still be the case even if compliance with starting treatment is low, since each single LTBI case that does not progress to TB disease matters, and any treatment level dominates no treatment. A compliance with starting treatment of only 10% saves $54.29 per treated contact and increases LYGs by 0.001 years; thus, although the greater the compliance the better, a minimum limit of compliance is not necessary from a cost-effectiveness point of view.

However, although it is even more cost-effective than the baseline value of reactivation assumption, the total strategic costs of the high reactivation probability are higher. The rationale for this seemingly paradoxical result is that in absolute numbers fewer active cases can be prevented given the same efficacy of 0.8 (2.2 cases left instead of 1.2), and more cases will be missed given the same NPV (1.06 instead of 0.58) [Table 3]; this, however, includes the cost of more TB cases, even though more are averted. But, if more TB cases really do occur in the population, this will also be the situation using TST testing.

Assuming that there are, on average, three secondary MTB infections of contacts due to each TB case occurring within the 20 years of our study, the number of cases predicted in our cohort was enlarged by only 0.44 active cases (ie, by about 7.6%), representing an increase of 0.4% in total screening costs (from $227,014 to $227,932). Increasing the sensitivity of the QFT-G assay from 90 to 95% reduces the total costs by 2.2% (from $227,014 to $222,041).

TST Cutoff Induration Size of > 5 mm

Performing the TST with a cutoff induration size of 5 mm resulted in a total of 443 positive results (ie, 4.9 times as many positive results as obtained by the 90 positive QFT-G assay results at the same time). Correcting the number of assumed FP TST (cutoff > 5 mm) results by adding to the 90 TP (QFT-G assay) results the assumed 10 FN (QFT-G assay) results, 343 of the 443 contacts (77.4%) would be given INH unnecessarily, reflecting the lower PPV of the TST (cutoff > 5 mm) compared with that of QFT-G assay. That means that more than four times (4.36) as many contacts need to be treated in order to avoid one active TB case in comparison with the QFT-G assay-based strategy (NNT, 96 vs 22, respectively). Assuming that there are 20 persons with FN test results in the sensitivity analysis (Table 4 ), as described in the “Materials and Methods” section, the number of cases being missed doubled (1.16) compared to those determined with the QFT-G assay approach, but this only resulted in a higher expenditure of $11,163 and 2.8% greater total costs. The crucial point, however, is that the treatment costs are more than the three times (3.08) those of the QFT-G assay. Thus, the ICER was $30,170 per LYG, and the annual cost due to treatment ($1,563) was an additional expenditure. Although the ICER of an LTBI treatment based on this screening approach falls just within the willingness-to-pay range, a cost savings would be achieved only with unrealistically high TB costs of $30,224 (reference option) and $36,150 (second option) or higher, or with a reduction in the cost of LTBI treatment of about one fourth and one third, respectively (from $250.3 to $193.4 and $164.0, respectively).

TST Cutoff Induration Size of > 10 mm

A cutoff induration size of 10 mm for the TST substantially reduces the treatment costs or ICER compared to TST cutoff induration size of 5 mm, as here the ratio of 2.3 between the 207 contacts who initially had positive test results and the 91 (81 plus 10) LTBI cases is markedly reduced in comparison with the TST performed with a cutoff induration size of > 5 mm, indicating a further decrease in the proportion of unnecessarily treated individuals (116 of 207 individuals; 56.0%). The treatment cost was 46.1% lower (second option, 46.9%) and clearly outweighs the costs due to FN results that is 2.27 times higher (second option, 1.7 times higher) following not only the comparably higher specificity than the 5-mm induration size cutoff, but also the higher reactivation probability of 0.0037 a year. This results in the prevention of nearly one TB case more (0.82). This approach is almost as highly cost-effective as the QFT-G assay screening, with an annual cost savings due to treatment of $1,836.

TST > 5 mm Followed by QFT-G Assay

Screening first by TST with a cutoff at > 5 mm followed by the QFT-G assay as a confirmation test has no impact on the ICER compared with using the QFT-G assay alone. If the assumed 10 FN QFT-G assay results (not included in Table 1 as described above) are picked up by the same number of subjects with positive TST results under base-case assumptions, in contrast to the second option the number of treated contacts presumed to be infected by MTB will not be reduced by 10% (90 vs 100) after preselection by the TST, but will result in 100 subjects with positive QFT-G assay results, as if the QFT-G assay were being performed alone. Since the costs of TST prescreening must then be added (plus $2,390), the cost due to the FN results will be decisive.

Following the base-case option, the costs due to negative results are clearly less in comparison with using the QFT-G assay alone (minus $6,528), reducing the total costs per 1,000 contacts compared to using the TST alone by 1.8% to $222,869; however, following the second option, as described in the section “Modeling Screening Variables of QFT-G Assay and Mantoux TST,” the lower sensitivity of the TST leads to half a TB case more that is missed (0.56), and therefore to additional costs of $11,163 due to FN results that more than outweigh the lower cost generated by TN TST results in comparison with the performance of the QFT-G assay test alone ($147,373 vs $152,887, respectively). Thus, in the base-case option the combination with the 5-mm TST cutoff induration size will be the least expensive of the four strategies, while in the second approach the lease expensive strategy will be to use the QFT-G assay strategy alone. All in all, however, the cost of the program based on combining the two tests is only marginally lower than the total cost of the program based on QFT-G assay alone per 1,000 close-contacts by approximately $1,397 (0.61%).

Until now, cost-effectiveness analyses of LTBI treatment have been based on unreliable assumptions regarding the specificity of the TST and consequently regarding the PPV, which is (along with true LTBI prevalence) highly dependent on specificity as a methodologic parameter. Even in the pre-IGRA era it was common wisdom that FP TST results may occur in persons who have been infected with nontuberculous mycobacteria and in persons vaccinated with BCG, and that test specificity will increase with the reaction induration size defined as a “positive” test.17

Most economics publications,25,26 however, either implicitly assume a specificity value of 100% by suggesting that there is LTBI treatment for each contact with a positive TST results or perform the TST with baseline values that are nearly as high (eg, 95%15 or 99%27), which may for this reason lead to a systematic bias by overestimating the number of cases prevented as the numerator of the ICER. Only Dasgupta and Menzies28 have suggested in a newer analysis a reduction in specificity of up to 60%. Our analysis tries to give a realistic answer after considering the data provided by a German real-world IGRA study on this topic.10

Of course, when implementing programs for preventing infectious diseases, the lowest possible cost per case of illness prevented should be the standard. In 20-year-old close-contacts, the baseline strategy of screening, combining the TST performed with an induration cutoff size of 5 mm and subsequently performing the QFT-G assay, was the least costly and the most cost-effective alternative and was very closely followed by the use of the QFT-G assay alone. Using the cost of $50,000 per LYG by LTBI treatment as a benchmark, the TST strategy with a 5- mm induration size cutoff also fulfilled this criterion, albeit at the price of producing the highest total cost due to a low PPV and therefore an unacceptably high rate of unnecessarily treated contacts who were not infected with MTB.

All in all, the probability of TB developing after infection, given a positive QFT-G test result, appeared to be the most sensitive parameter, resulting in an increase in total screening costs from $227,014 to $252,685 and thereby changing the ranking in favor of the TST performed at an cutoff induration size of 10 mm, although of course this also reflects the costs of more TB cases. In the absence of better available data, we used values for this probability from one metaanalysis16 determining the reactivation risk of LTBI cases with respect to age group and induration diameters only of the TST. However, the studies analyzed in the metaanalysis16 may have been more or less confounded by the presence of more persons vaccinated with BCG and more persons with nontuberculous mycobacteria infections, comprising also persons who were uninfected but who had a FP test result. Therefore, these data may have underestimated the true risk of disease and consequently also underestimated the benefits of screening in our analyses of long-term clinical consequences. Since induration size can also be a relatively strong predictor among noninfected persons who are vaccinated with BCG,14 our analysis adopted a cautious approach, starting from a risk of disease progression if the QFT-G assay result is positive that is comparable to that of a TST performed with an induration size cutoff of 5 mm and comparing the outcomes when that risk is increased to that of having an induration size of 15 mm.

Another underestimation of the benefits of the QFT-G assay occurred through our decision not to include the costs of the side effects of INH chemoprevention. These costs need not be considered if the German guidelines (ie, no person > 50 years of age should be treated and any person with signs of previous hepatopathy should also be excluded from treatment) are regularly used. However, under other circumstances, a considerable number of persons would be given chemoprevention as a result of FP TST results, which could prove to be a significant cost factor.

Neither raising the sensitivity of the QFT-G assay to 95% nor the dynamic model approach comprising the consequences of new LTBI cases incurred during the time frame had any substantial influence on the outcomes of this analysis. If the total cost of active TB disease increased by more than one quarter (28.2%), this would also bring the TST-based strategy with a 5-mm induration size cutoff into the cost-saving range; this scenario, however, seems to be beyond reality.

Both QFT-G assay-based strategies needed only 22 contacts who tested positive to avoid one TB case, which may appear to be a striking argument from an ethical point of view. Irrespective of economic considerations, this approach stands in contrast to both TST-based strategies that need nearly twice as many contacts (TST 10-mm cutoff, 40 contacts) or more than the four times as many contacts (TST 5-mm cutoff, 96 contacts) to be treated to achieve the same target.

Although the two TST-based strategies also provide a remarkable clinical benefit compared with no LTBI treatment, they were, when performed alone, in each case more costly and less effective than the QFT-G assay, the higher cost of implementation of which was outweighed by the averted cost of unnecessarily treating contacts who otherwise would have been wrongly classified as LTBI cases. Only if the predictive value of a positive QFT-G assay result with respect to the probability of progression to TB disease should prove to be the same as that of a positive TST result at a cutoff induration size of 15 mm could a TST-based strategy at a cutoff induration size of 10 mm be even more cost-effective. Overall, these results provide evidence that the QFT-G assay (performed alone or, especially, in combination with the TST) can provide both superior clinical and economic outcomes at a comparatively small additional cost when performed as an additional test before starting LTBI treatment. In an era of increasingly limited resources that are available for health care, it is worthwhile reconsidering the economic implications of this strategy.

Abbreviations: BCG = bacillus Calmette-Guerin; CI = confidence interval; FN = false-negative; FP = false-positive; ICER = incremental cost-effectiveness ratio; IGRA = interferon-γ release assay; INH = isoniazid; LTBI = latent tuberculosis infection; LY = life-year; LYG = life-year gained; MTB = Mycobacterium tuberculosis; NNT = number needed to treat; NPV = negative predictive value; PPV = positive predictive value; QFT-G = QuantiFERON-TB Gold In-Tube; TB = tuberculosis; TN = true-negative; TP = true-positive; tpDcm = probability of death due to TB; tpReact = transition probability for a progression to manifest TB; TST = tuberculin skin test

The authors have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.

Table Graphic Jump Location
Table 1. Results of TST and QFT-G Testing in a Population of 309 Close-Contacts, Converted to 1,000 Contacts and Separated by the Cutoff of Induration Diameter
Figure Jump LinkFigure 1. Simplified Markov model for predicting the total costs and the occurrence of TB due to LTBI by screening a cohort of 20-year-old close-contacts and treating those with positive test results with INH using the QFT-G assay-based strategy (branches representing secondary LTBI and TB disease cases are not shown due to space constraints). A decision node (□) is the decision after a positive QFT-G assay skin test to accept INH or not. Branches from a node change (o) represent the possible outcomes of an event; branches from a Markov node (M) represent the possible different health states. A terminal node (◃) represents a state from which an individual will jump to the next cycle Only the state “death” is an absorbing one and cannot be departed from. The bold horizontal lines indicate that the following subtree is cloned (ie, a copy is attached to a node in another branch of the tree). The cloned subtrees, denoted 1 and 2, are attached to the (1-NPV) and the NPV-node, respectively. Probabilities (p) are defined as follows: pQFT = probability of QFT-G assay being positive; tpDn_20 = probability of death due to causes other than TB among contacts with an initial age of 20; # = complementary probability (all probabilities of chance node branches to sum to 1.0); Effect = efficacy of INH (in percentage) to prevent progression to manifest TB.Grahic Jump Location
Table Graphic Jump Location
Table 2. Epidemiologic and Methodologic Estimates Used in Cost-effectiveness Analysis
* 

Values are given as the estimate from the baseline assumption (estimate from the alternative assumption).

Table Graphic Jump Location
Table 3. Base-Case Assumptions in the Health and Economic Outcomes of the Cohort of 20-Year-Olds
* 

If each active TB case arising during the study period in addition infects three further contact persons.

 

If treatment is both more effective and less expensive than nontreatment, the latter option is “dominated.”

 

Values are given as NNT (95% CI).

§ 

This reactivation probability is equivalent to that of a TST assay conducted with an induration diameter of at least 15 mm.

Table Graphic Jump Location
Table 4. Sensitivity Analysis on Decreased PPV of the TST Health and Economic Outcomes of the 20-Year-Old Cohort
* 

If each active TB case arising during the study period infects three further contact persons.

 

If treatment is both more effective and less expensive than nontreatment, the latter option is “dominated.”

 

Values are given as NNT (95% CI).

Ewer, K, Deeks, J, Alvarez, L, et al (2003) Comparison of T-cell-based assay with tuberculin skin test for diagnosis ofMycobacterium tuberculosisinfection in a school tuberculosis outbreak.Lancet361,1168-1173. [PubMed] [CrossRef]
 
Zellweger, JP, Zellweger, A, Ansermet, S, et al Contact tracing using a new T-cell-based test: better correlation with tuberculosis exposure than the tuberculin skin test.Int J Tuberc Lung Dis2005;9,1242-1247. [PubMed]
 
Diel, R, Ernst, M, Döscher, G, et al Avoiding the effect of BCG vaccination in detectingMycobacterium tuberculosisinfection with a blood test.Eur Respir J2006;28,16-23. [PubMed]
 
Lee, JY, Choi, HJ, Park, I-N, et al Comparison of two commercial interferon-γ assays for diagnosingMycobacterium tuberculosisinfection.Eur Respir J2006;28,24-30. [PubMed]
 
Piana, F, Codecasa, LR, Cavallerio, P, et al Use of a T-cell-based test for detection of tuberculosis infection among immunocompromised patients.Eur Respir J2006;28,31-34. [PubMed]
 
Diel, R, Nienhaus, A, Lange, C, et al Cost-optimisation of screening for latent tuberculosis in close contacts.Eur Respir J2006;28,35-44. [PubMed]
 
Wrighton-Smith, P, Zellweger, J-P Direct costs of three models for the screening of latent tuberculosis infection.Eur Respir J2006;28,45-50. [PubMed]
 
Lungenliga, Schweiz. Erkennung der Tuberkuloseinfektion mittels Bluttest (interferon-γ). 2005; Lungenliga Schweiz. Bern, Switzerland:.
 
National Institute for Clinical Excellence. Tuberculosis: national clinical guideline for diagnosis, management, prevention, and control. Available at: www.nice.org.uk/. Accessed October 9, 2006.
 
Diel R, Nienhaus A, Lange C, et al. Tuberculosis contact investigation with a new, specific blood test in a low-incidence population containing a high proportion of BCG-vaccinated persons. Respir Res [serial online] 2006; 7:77.
 
Deutsches Zentralkomitee zur Bekämpfung der Tuberkulose.. Richtlinien für die Umgebungsuntersuchungen bei Tuberkulose [Central Committee for Control of Tuberculosis. Guidelines for environmental contact tracing in tuberculosis].Gesundheitswesen1996;58,657-665. [PubMed]
 
Schaberg, T, Hauer, B, Loddenkemper, R, et al Latent tuberculosis infection: recommendations for preventive therapy in adults in Germany.Pneumologie2004;58,92-102. [PubMed]
 
Mori, T, Sakatani, M, Yamagishi, F, et al Specific detection of tuberculosis infection: an interferon-γ-based assay using new antigens.Am J Respir Crit Care Med2004;170,59-64. [PubMed]
 
Tissot, F, Zanetti, G, Francioli, P, et al Influence of bacille Calmette-Guérin vaccination on size of tuberculin skin test reaction: to what size?Clin Infect Dis2005;40,211-217. [PubMed]
 
Diel, R, Nienhaus, A, Schaberg, T Cost-effectiveness of isoniazid chemoprevention in close contactsEur Respir J2005;26,465-473. [PubMed]
 
Horsburgh, CR, Jr Priorities for the treatment of latent tuberculosis infection in the United States.N Engl J Med2004;350,2060-2079. [PubMed]
 
American Thoracic Society.. Targeted tuberculin testing and treatment of latent tuberculosis infection.Am J Respir Crit Care Med2000;161,S221-247. [PubMed]
 
Lungenliga Schweiz-Bundesamt für Gesundheit.. Handbuch Tuberkulose.Schweiz Med Forum2003;3,487-491
 
Robert-Koch, Institute. Die Epidemiologie der Tuberkulose in Deutschland: Jahrbuch 2004. 2005; Robert-Koch Institute. Berlin, Germany:.
 
Salpeter, EE, Salpeter, SR Mathematical model for the epidemiology of tuberculosis, with estimates of the reproductive number and infection-delay function.Am J Epidemiol1998;142,398-406
 
Behr, MA, Warren, SA, Salamon, H, et al Transmission ofMycobacterium tuberculosisfrom patients smear-negative for acid-fast bacilli.Lancet1999;353,444-449. [PubMed]
 
Nolan, CM, Goldberg, SV, Buskin, SE Hepatotoxicity associated with isoniazid preventive therapy: a 7-year survey from a public health tuberculosis clinic.JAMA1999;281,1014-1018. [PubMed]
 
Weinstein, M, Siegel, J, Gold, M, et al Recommendations of the Panel on Cost-effectiveness in Health and Medicine.JAMA1996;276,1253-1258. [PubMed]
 
Owens, DK Interpretation of cost-effectiveness analyses.J Gen Intern Med1998;13,716-717. [PubMed]
 
Snyder, DC, Paz, EA, Mohle-Boetani, JC, et al Tuberculosis prevention in methadone maintenance clinics: effectiveness and cost-effectiveness.Am J Respir Crit Care Med1999;160,178-186. [PubMed]
 
Dasgupta, K, Schwartzman, K, Marchand, R, et al Comparison of cost-effectiveness of tuberculosis screening of close contacts and foreign-born populations.Am J Respir Crit Care Med2000;162,2079-2086. [PubMed]
 
Porco TC, Lewis B, Marseille E, et al. Cost-effectiveness of tuberculosis evaluation and treatment of newly-arrived immigrants. BMC Public Health [serial online] 2006; 6:157.
 
Dasgupta, K, Menzies, D Cost-effectiveness of tuberculosis control strategies among immigrants and refugees.Eur Respir J2005;25,1107-1116. [PubMed]
 

Figures

Figure Jump LinkFigure 1. Simplified Markov model for predicting the total costs and the occurrence of TB due to LTBI by screening a cohort of 20-year-old close-contacts and treating those with positive test results with INH using the QFT-G assay-based strategy (branches representing secondary LTBI and TB disease cases are not shown due to space constraints). A decision node (□) is the decision after a positive QFT-G assay skin test to accept INH or not. Branches from a node change (o) represent the possible outcomes of an event; branches from a Markov node (M) represent the possible different health states. A terminal node (◃) represents a state from which an individual will jump to the next cycle Only the state “death” is an absorbing one and cannot be departed from. The bold horizontal lines indicate that the following subtree is cloned (ie, a copy is attached to a node in another branch of the tree). The cloned subtrees, denoted 1 and 2, are attached to the (1-NPV) and the NPV-node, respectively. Probabilities (p) are defined as follows: pQFT = probability of QFT-G assay being positive; tpDn_20 = probability of death due to causes other than TB among contacts with an initial age of 20; # = complementary probability (all probabilities of chance node branches to sum to 1.0); Effect = efficacy of INH (in percentage) to prevent progression to manifest TB.Grahic Jump Location

Tables

Table Graphic Jump Location
Table 1. Results of TST and QFT-G Testing in a Population of 309 Close-Contacts, Converted to 1,000 Contacts and Separated by the Cutoff of Induration Diameter
Table Graphic Jump Location
Table 2. Epidemiologic and Methodologic Estimates Used in Cost-effectiveness Analysis
* 

Values are given as the estimate from the baseline assumption (estimate from the alternative assumption).

Table Graphic Jump Location
Table 3. Base-Case Assumptions in the Health and Economic Outcomes of the Cohort of 20-Year-Olds
* 

If each active TB case arising during the study period in addition infects three further contact persons.

 

If treatment is both more effective and less expensive than nontreatment, the latter option is “dominated.”

 

Values are given as NNT (95% CI).

§ 

This reactivation probability is equivalent to that of a TST assay conducted with an induration diameter of at least 15 mm.

Table Graphic Jump Location
Table 4. Sensitivity Analysis on Decreased PPV of the TST Health and Economic Outcomes of the 20-Year-Old Cohort
* 

If each active TB case arising during the study period infects three further contact persons.

 

If treatment is both more effective and less expensive than nontreatment, the latter option is “dominated.”

 

Values are given as NNT (95% CI).

References

Ewer, K, Deeks, J, Alvarez, L, et al (2003) Comparison of T-cell-based assay with tuberculin skin test for diagnosis ofMycobacterium tuberculosisinfection in a school tuberculosis outbreak.Lancet361,1168-1173. [PubMed] [CrossRef]
 
Zellweger, JP, Zellweger, A, Ansermet, S, et al Contact tracing using a new T-cell-based test: better correlation with tuberculosis exposure than the tuberculin skin test.Int J Tuberc Lung Dis2005;9,1242-1247. [PubMed]
 
Diel, R, Ernst, M, Döscher, G, et al Avoiding the effect of BCG vaccination in detectingMycobacterium tuberculosisinfection with a blood test.Eur Respir J2006;28,16-23. [PubMed]
 
Lee, JY, Choi, HJ, Park, I-N, et al Comparison of two commercial interferon-γ assays for diagnosingMycobacterium tuberculosisinfection.Eur Respir J2006;28,24-30. [PubMed]
 
Piana, F, Codecasa, LR, Cavallerio, P, et al Use of a T-cell-based test for detection of tuberculosis infection among immunocompromised patients.Eur Respir J2006;28,31-34. [PubMed]
 
Diel, R, Nienhaus, A, Lange, C, et al Cost-optimisation of screening for latent tuberculosis in close contacts.Eur Respir J2006;28,35-44. [PubMed]
 
Wrighton-Smith, P, Zellweger, J-P Direct costs of three models for the screening of latent tuberculosis infection.Eur Respir J2006;28,45-50. [PubMed]
 
Lungenliga, Schweiz. Erkennung der Tuberkuloseinfektion mittels Bluttest (interferon-γ). 2005; Lungenliga Schweiz. Bern, Switzerland:.
 
National Institute for Clinical Excellence. Tuberculosis: national clinical guideline for diagnosis, management, prevention, and control. Available at: www.nice.org.uk/. Accessed October 9, 2006.
 
Diel R, Nienhaus A, Lange C, et al. Tuberculosis contact investigation with a new, specific blood test in a low-incidence population containing a high proportion of BCG-vaccinated persons. Respir Res [serial online] 2006; 7:77.
 
Deutsches Zentralkomitee zur Bekämpfung der Tuberkulose.. Richtlinien für die Umgebungsuntersuchungen bei Tuberkulose [Central Committee for Control of Tuberculosis. Guidelines for environmental contact tracing in tuberculosis].Gesundheitswesen1996;58,657-665. [PubMed]
 
Schaberg, T, Hauer, B, Loddenkemper, R, et al Latent tuberculosis infection: recommendations for preventive therapy in adults in Germany.Pneumologie2004;58,92-102. [PubMed]
 
Mori, T, Sakatani, M, Yamagishi, F, et al Specific detection of tuberculosis infection: an interferon-γ-based assay using new antigens.Am J Respir Crit Care Med2004;170,59-64. [PubMed]
 
Tissot, F, Zanetti, G, Francioli, P, et al Influence of bacille Calmette-Guérin vaccination on size of tuberculin skin test reaction: to what size?Clin Infect Dis2005;40,211-217. [PubMed]
 
Diel, R, Nienhaus, A, Schaberg, T Cost-effectiveness of isoniazid chemoprevention in close contactsEur Respir J2005;26,465-473. [PubMed]
 
Horsburgh, CR, Jr Priorities for the treatment of latent tuberculosis infection in the United States.N Engl J Med2004;350,2060-2079. [PubMed]
 
American Thoracic Society.. Targeted tuberculin testing and treatment of latent tuberculosis infection.Am J Respir Crit Care Med2000;161,S221-247. [PubMed]
 
Lungenliga Schweiz-Bundesamt für Gesundheit.. Handbuch Tuberkulose.Schweiz Med Forum2003;3,487-491
 
Robert-Koch, Institute. Die Epidemiologie der Tuberkulose in Deutschland: Jahrbuch 2004. 2005; Robert-Koch Institute. Berlin, Germany:.
 
Salpeter, EE, Salpeter, SR Mathematical model for the epidemiology of tuberculosis, with estimates of the reproductive number and infection-delay function.Am J Epidemiol1998;142,398-406
 
Behr, MA, Warren, SA, Salamon, H, et al Transmission ofMycobacterium tuberculosisfrom patients smear-negative for acid-fast bacilli.Lancet1999;353,444-449. [PubMed]
 
Nolan, CM, Goldberg, SV, Buskin, SE Hepatotoxicity associated with isoniazid preventive therapy: a 7-year survey from a public health tuberculosis clinic.JAMA1999;281,1014-1018. [PubMed]
 
Weinstein, M, Siegel, J, Gold, M, et al Recommendations of the Panel on Cost-effectiveness in Health and Medicine.JAMA1996;276,1253-1258. [PubMed]
 
Owens, DK Interpretation of cost-effectiveness analyses.J Gen Intern Med1998;13,716-717. [PubMed]
 
Snyder, DC, Paz, EA, Mohle-Boetani, JC, et al Tuberculosis prevention in methadone maintenance clinics: effectiveness and cost-effectiveness.Am J Respir Crit Care Med1999;160,178-186. [PubMed]
 
Dasgupta, K, Schwartzman, K, Marchand, R, et al Comparison of cost-effectiveness of tuberculosis screening of close contacts and foreign-born populations.Am J Respir Crit Care Med2000;162,2079-2086. [PubMed]
 
Porco TC, Lewis B, Marseille E, et al. Cost-effectiveness of tuberculosis evaluation and treatment of newly-arrived immigrants. BMC Public Health [serial online] 2006; 6:157.
 
Dasgupta, K, Menzies, D Cost-effectiveness of tuberculosis control strategies among immigrants and refugees.Eur Respir J2005;25,1107-1116. [PubMed]
 
NOTE:
Citing articles are presented as examples only. In non-demo SCM6 implementation, integration with CrossRef’s "Cited By" API will populate this tab (http://www.crossref.org/citedby.html).

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging & repositioning the boxes below.

CHEST Journal Articles
PubMed Articles
  • CHEST Journal
    Print ISSN: 0012-3692
    Online ISSN: 1931-3543