Article Text

Cost-effectiveness of alternative strategies for interferon-γ release assays and tuberculin skin test in tuberculous uveitis
1. Marcus Ang1,2,3,
2. Hai V Nguyen4,
3. Sieh Yean Kiew1,
4. Shu Chen4,
5. Soon-Phaik Chee1,2,3,
6. Eric Finkelstein4
1. 1Singapore National Eye Centre, Singapore, Singapore
2. 2Singapore Eye Research Institute, Singapore, Singapore
3. 3Department of Ophthalmology, Yong Yoo Lin School of Medicine, National University of Singapore, Singapore, Singapore
4. 4Duke-NUS Graduate Medical School, Singapore, Singapore
1. Correspondence to Professor Soon-Phaik Chee, Singapore National Eye Centre, 11 Third Hospital Avenue, Singapore 168751, Singapore; chee.soon.phaik{at}snec.com.sg

## Abstract

Background Although tuberculous uveitis remains a major cause of ocular morbidity in the developing world, there is no consensus on which diagnostic test or testing strategy is the most cost effective. In this study we carried out a cost-effectiveness analysis to determine the most cost-effective diagnostic test strategy.

Methods In this prospective study, we recruited 102 patients from Singapore National Eye Centre with signs suggestive of tuberculous uveitis. Using prospective data from this cohort and from published meta-analyses, we modelled the incremental cost effectiveness of the following strategies: tuberculin skin test (TST) only; interferon-γ release assay (IGRA) only; IGRA following a positive TST result; and dual-test strategy, conducting TST and IGRA at presentation. Incremental cost-effectiveness ratios (ICERs) were calculated for each strategy and analysed using a willingness-to-pay threshold of $50 000 per quality-adjusted life year (QALY) gained. Results In our population, the least cost effective was the IGRA-only strategy. The dual-test strategy was the most cost effective, with an improvement of 0.017 QALY at an incremental cost of$190 relative to the TST-only strategy (ICER $11 500); while the TST-only strategy was more cost effective than the third strategy, using IGRA following a positive TST result (ICER$3610). This remained consistent while varying the costs of IGRA and TST, the incidence of tuberculosis and tuberculous uveitis, as well as the diagnostic accuracy of IGRA and TST found in previous studies in various populations.

Conclusions The dual-test strategy (performing TST and IGRA at presentation) was the most cost effective strategy for the diagnosis of tuberculous uveitis in our population.

• Diagnostic tests/Investigation
• Infection
• Inflammation

## Introduction

Tuberculous uveitis remains a significant cause of ocular morbidity in the developing world.1 However, making a confirmed diagnosis is difficult due to the low sensitivity of tests such as alcohol fast bacilli (AFB) smears and PCR analyses.2 ,3 Thus, tuberculous uveitis is often a clinical diagnosis—based on suggestive ocular features (such as broad-based posterior synechiae, serpiginious choroiditis, retinal vasculitis, granulomatous inflammation4 ,5) in patients with serological or radiographic evidence of latent or active Mycobacterium tuberculosis infection. However, the decision to treat patients with tuberculous uveitis using anti-tuberculosis (TB) therapy is controversial and difficult in view of the well known side effects, as well as the health economic implications of instituting this long-term therapy.6

Traditionally, this decision greatly depended on the ‘centuries-old’ tuberculin skin test (TST), a measure of the type IV hypersensitivity reaction to a crude mixture of mycobacterial antigens injected intradermally. This test is inexpensive and sensitive but non-specific, due to confounding by previous Bacille-Calmette-Guerin vaccination and by cross reactivity with environmental mycobacterial antigens.7 The advent of interferon-γ release assays (IGRAs) has provided new avenues for mycobacterial testing, offering a sensitive, specific, objective ex vivo assay for previous exposure to M tuberculosis.8 The US Communicable Diseases Center (CDC) guidelines recommend one-step IGRA-only testing for the diagnosis and treatment of M tuberculosis infection.9 However, the UK National Institute of Health and Care Excellence (NICE) guidelines recommend two-step testing with the TST first, followed by confirmatory IGRA testing in individuals with a positive TST.10

There is currently no clear consensus on which test to use and the most cost-effective testing algorithm, specifically for the diagnosis of tuberculous uveitis.11 Therefore, in this study we attempted to determine the most cost-effective testing strategy in the context of tuberculous uveitis for our population. We modelled the incremental cost effectiveness of the following four alternative strategies: TST only; IGRA only; IGRA following a positive TST result; and dual-test strategy of conducting the TST and IGRA simultaneously at presentation.

## Methods

In this prospective study we recruited patients who presented to the Singapore National Eye Centre with intraocular inflammation suggestive of tuberculous uveitis.5 Our study followed the principles of the Declaration of Helsinki, with ethical approval from the Singapore Health Services Institutional Review Board. All patients underwent a complete ocular examination and systemic review as previously described.12 In summary, investigations included the TST, QuantiFERON-Gold in-tube (Cellestis Inc, Australia), as well as full blood count, erythrocyte sedimentation rate, renal and liver panel. Patients with positive TST or IGRA were then referred to the infectious diseases physicians at Singapore General Hospital for review and consideration of anti-TB therapy. Investigations, diagnosis of tuberculous uveitis, follow-up and management are detailed in the online supplementary material due to the focus on cost-effectiveness modelling in this article; and also, have been previously described.6 ,12 ,13

### Model design

To estimate and analyse cost effectiveness of the four alternative testing strategies, we constructed a decision tree using TreeAge Pro 2013 software. The decision tree shown in figure 1 displays the four alternative screening strategies and subsequent treatment pathways. Each test yields either a positive or negative result, which can either be true or false. As described, all patients underwent a full ocular and systemic evaluation with investigators to diagnose the underlying cause of their uveitis and are comanaged with infectious disease physicians. A patient may not accept recommendations for anti-TB therapy, though all accepted ocular anti-inflammatory treatment. In this model, patients who refuse anti-TB therapy are assumed to not receive it in the future; and either therapy pathway may yield resolution or recurrence of disease. The model also allowed for the following adverse events: severe side effects such as hepatotoxicity from the anti-TB therapy requiring treatment, hospitalization, and in extreme cases, hepatic failure and death.14 Mild cases of hepatotoxicity or minor side effects such as rashes are assumed to stop after drug cessation. Patients who stop anti-TB therapy due to adverse effects continue on anti-inflammatory/corticosteroid treatment as required to control their ocular inflammation. We assumed adherence to anti-TB therapy if there are no severe adverse events. The model considers three possible outcomes: death from severe adverse events such as hepatotoxicity, patient becomes asymptomatic, or the patient has persistent or recurrent ocular inflammation.

Figure 1

Decision tree for each test strategy: tuberculin skin test (TST) only; interferon-γ release assay (IGRA) only; IGRA following a positive TST result (TST +ve IGRA); and dual-test strategy of conducting TST and IGRA simultaneously.

### Model inputs

We populated the decision tree using data from our prospective study cohort and from meta-analyses from published literature, all detailed in online supplementary tables S1–S3 (see online supplementary material). In summary, online supplementary table S1 describes the prevalence and test parameters used,15 obtained from studies with similar population dynamics to Singapore (ie, low to moderate TB burden, moderate to high levels of immigration).7 ,16–21 All costs are in 2010 Singapore dollars and based on standard charges from the local tertiary government hospital (Singapore General Hospital)—online supplementary table S2. The treatment costs for each patient depend on their subsequent treatment path. For example, costs ranged from $1050 (patients who refused or defaulted on anti-TB therapy) to$2070 (patients who completed the full course of anti-TB therapy but developed recurrences).

#### Utilities

For each clinical outcome, a utility (used as weights for calculating the number of quality-adjusted life years) was obtained from the literature—online supplementary table S3. Essentially, the utility of patients with uveitis at baseline was 0.80.22 The quality of life improved for all patients after treatment and those patients with tuberculous uveitis who were treated successfully enjoyed a higher quality of life than those who were not. Treated patients who had tuberculous uveitis with recurrent inflammation had a utility of 0.84.23 Meanwhile, the utility for treated patients without non-tuberculous uveitis but who had persistent inflammation was 0.86.23 ,24 Treated patients whose ocular inflammation resolved with no recurrence were assigned a utility of 0.89.24 Utilities were then modelled for a 30-year period (to reflect the average remaining lifespan of these patients with mean age of 48) and subject to a discount rate of 3%.

To determine whether a test strategy was cost effective, we first removed any dominated strategies, that is, strategies that have higher costs but are less effective compared with the next most costly alternative. Next, among the non-dominated strategies, we estimated the incremental cost-effectiveness ratio (ICERs) and compared them with the conventional willingness-to-pay (WTP) threshold of $50 000 per quality-adjusted life year (QALY). We then supplemented the baseline analysis with one-way and probabilistic sensitivity analyses. For the one-way sensitivity analysis, we allowed each parameter to change within an accepted reasonable range and for the probabilistic sensitivity analysis, we performed 10 000 iterations of a Monte Carlo simulation. We assumed β distributions for all probabilities and utilities whose values were bounded between 0 and 1; and a γ distribution for cost to capture its non-negative and skewed features. The moments of these distributions were based on the point estimates and on our choice of a relatively large SD (ie, 20% of the mean).25 We also tested the robustness of our findings by varying the diagnostic accuracy of each IGRA and TST test, which varied according to the prevalence of disease and populations. Finally, we contrasted our cost-effectiveness estimates with those in the literature that has thus far focused only on the three testing strategies: TST positive followed by IGRA testing (similar to UK NICE guidelines),10 IGRA only (similar to CDC guidelines),9 ,26 and the TST-only strategy. ## Results ### Summary of clinical outcome In this study, we recruited and followed up 102 patients, of which 52 tested positive with the TST and 50 were positive on QuantiFERON-Gold in-tube testing. Following consultation with the infectious diseases physicians, 24 patients agreed to anti-TB treatment, while understanding the controversies, risks and benefits of their choice in treating tuberculous uveitis.6 Of these, 22 completed 6–9 months of anti-TB treatment; 2 patients developed minor side effects of treatment (rash and mild elevation of liver enzymes, which resolved on cessation of treatment) and did not complete anti-TB treatment. There were no cases of severe hepatotoxicity or serious adverse effects that required hospitalisation, intensive care or surgical intervention in our cohort. The decision pathways and outcomes of these patients were included in the model (figure 1): among those who completed anti-TB treatment, 20 patients were asymptomatic at the end of the follow-up period, while 2 patients had recurrent/persistent inflammation requiring long-term treatment with systemic immunosuppression (9.1% recurrence rate). Of the two patients who began but did not complete anti-TB treatment, one was asymptomatic at the end of the follow-up period, while the other patient had persistent inflammation requiring long-term therapy. However, 78 patients received only ocular anti-inflammatory therapy, of which 65 were asymptomatic at the end of the follow-up period while 13 (16.7%) had recurrent or persistent inflammation despite treatment. ### Baseline analysis Table 1 displays the incremental cost-effectiveness results of the four testing strategies. In terms of actual costs, the ‘TST positive then IGRA test’ strategy is the least expensive, followed by the ‘TST only’ strategy and the ‘IGRA only’ strategy. The dual strategy of conducting TST and IGRA simultaneously is actually the most costly. Although the TST positive then IGRA strategy involves an additional test for some patients, it is less expensive than the TST alone because it results in fewer patients being put on anti-TB therapy. As for effectiveness, the dual-test strategy yielded the highest expected QALYs, followed by the TST positive then IGRA strategy and the TST-only strategy, with the IGRA-only strategy delivering the lowest QALYs. The IGRA-only strategy is thus the dominated strategy; that is, it costs more and is less effective than the TST-only strategy in our study. Among the three non-dominated strategies, the dual-test strategy delivers an incremental improvement of 0.017 QALY at an incremental cost of$190 relative to the TST, yielding an ICER of $11 500. The TST has an ICER of$3610 compared with its preceding more costly alternative, the TST-positive then IGRA strategy. Thus, the cost-effectiveness estimates suggest that the dual-test strategy (TST and IGRA simultaneously) is cost effective at the conventional WTP threshold of $50 000 per QALY gained. Table 1 Baseline incremental cost effectiveness results ### Uncertainty analysis Figure 2 depicts the cos-effectiveness acceptability curves derived from the probabilistic sensitivity analysis, with the WTP on the x-axis and the probability that each strategy is cost effective compared with all other strategies on the y-axis. When the WTP is less than$3000 (ie, the intersection between the TST positive then IGRA strategy and the TST alone), the TST positive then IGRA strategy is preferred, evidenced by its segment lying above other curves. If the WTP is between around $3000 and$11 500 per QALY (the intersection between the TST alone and the dual-test strategy), the TST-alone strategy is cost effective compared with all other strategies. When the WTP surpasses $11 500, the dual-test strategy becomes cost effective relative to the TST-alone strategy. In particular, when the WTP per QALY gained is$20 000 or more, the dual strategy is cost effective in 85% of iterations or more.

Figure 2

Sensitivity analysis analysing the cost effectiveness (CE) acceptability curve and willingness to pay. Comparing tuberculin skin test (TST) only; interferon-γ release assay (IGRA) following a positive TST result (TST +ve IGRA); and dual-test strategy of conducting TST and IGRA simultaneously.

### Sensitivity analysis

We explored the changes in ICER of the dual-test strategy relative to the TST-only strategy and relative to the TST positive then IGRA strategy due to changes in each input parameter. This showed that the ICERs are most sensitive to the prevalence of tuberculosis and tuberculous uveitis, the diagnostic test parameters, and the costs of performing each test (these are all reported in online supplementary figures S1 and S2). We also plotted the ICER as a function of changes in key input parameters. The dual-test strategy becomes less attractive (ie, higher ICER) when the cost of IGRA increases, when the prevalence of tuberculous uveitis decreases, when the sensitivity of IGRA decreases or when the sensitivity of the TST increases. But within the input ranges considered, the dual strategy remains cost effective relative to each of the two alternatives, with most ICERs varying between $7000 and$18 000 per QALY (see online supplementary figures S3–S6).

Notably, even when the prevalence rate of tuberculous uveitis is as low as 0.2%1 the dual-test strategy is still cost effective relative to the TST alone with an ICER of $37 400 per QALY. As an additional sensitivity check, we computed cost effectiveness using sensitivity and specificity parameter values from meta-analyses published. These cost-effectiveness results remain robust to this change and in particular, the dual-test strategy remains cost effective compared with the TST with an ICER of$13 000 per QALY gained (table 2). Finally, to improve the external validity of our study we took into consideration that IGRA in Singapore is seven times as expensive as the TST (ie, $156 vs$22), while in other countries the IGRA is normally threefold more expensive than the TST.15 ,27 However, even when we lowered the cost of the IGRA to a level comparable with that in the literature, the dual strategy remained cost effective (table 2).

Table 2

Sensitivity analysis for incremental cost effectiveness of strategies

## Discussion

Although there have been cost-effectiveness analyses comparing IGRA and TST in various testing strategies for latent TB,15 to our knowledge this is the first study that compares strategies specifically for tuberculous uveitis. Systematic reviews by Nienhaus et al28 and Oxlade et al29 generally recommend IGRA only or IGRA following TST positive test strategies for latent TB. However, these studies did not consider the dual-test strategy of performing TST and IGRA simultaneously. Our study suggests that the simultaneous testing of TST and IGRA for patients with suspected tuberculous uveitis would be the most cost-effectiveness strategy for decision makers willing to pay $11 500 or more per QALY gained (well below the accepted threshold of$50 000 per QALY commonly cited).30 However, if the WTP threshold is below $11 500, then the ‘TST only’ strategy would be recommended as the cheapest alternative for our population. In our study population, we observed that two main factors may have contributed to the cost effectiveness of the dual-test strategy. First, the dual-test strategy has higher predictive power, which helped to reduce unnecessary costly treatment from false positive results. Second, in patients with both tests positive and treated with anti-TB therapy, there was a decreased likelihood of recurrence of uveitis, and thus reduced ocular morbidity.6 However, patients with a single positive reading from either TST or IGRA alone, who chose not to receive anti-TB therapy, were more likely to have recurrent tuberculous uveitis. These two advantages led to larger benefits, which outweighed the increased costs, making the dual-test strategy incrementally cost effective in our population. An important note is that the results of our study should be taken in the context of our patient population, with factors such as prevalence of disease, costs of performing each test and the accuracy of each diagnostic test for the given population. For example, in our baseline analysis, the TST-only strategy dominated the IGRA-only strategy and is cost effective relative to the TST positive then IGRA strategy recommended by the NICE guidelines in the UK. However, this is due to the high cost of performing IGRA in Singapore compared with the UK. However, when the cost of IGRA was lowered (as previously published, table 2), the NICE guidelines which recommended the TST positive then IGRA strategy was then cost effective compared with a single test only strategy. Of note, the dual-test strategy was still the most cost effective, with an ICER of$8600 per QALY gained in our study population. Thus, we performed various sensitivity analyses to vary the prevalence of disease, diagnostic test parameters and costs to improve the external validity of this study and application to various other study populations, presented in detail in the online supplementary material. However, these results certainly require confirmation using similar modelled studies, in various populations.

Nonetheless, we recognise the limitations of our study, such as the lack of diagnostic confirmation with positive culture, smear or PCR results from any of the intraocular samples in our patients.3 First, due to the small numbers of patients who are culture positive or have evidence of AFB on smears or M tuberculosis DNA using PCR from ocular biopsies, it is difficult to evaluate the IGRA or TST as a diagnostic test using ‘classic’ hypothesis testing, which is a limitation of studies concerning tuberculous uveitis.13 Second, for the purposes of this cost-effectiveness study we assumed compliance to anti-TB therapy unless it was discontinued by the physician due to side effects. We found this to be reasonable in the context of our study population, in which patients are monitored closely for medication compliance. At the Singapore Tuberculosis Control Unit, directly observed therapy is implemented for all patients;31 while at tertiary hospitals, all visits are reported to the Ministry of Health for patients on anti-TB therapy, where the patient recall system is closely monitored.31 Third, the health states of patients with tuberculous uveitis are assumed to have remained relatively unchanged throughout the hypothetical period in our analyses; and we also assumed that patients who refused treatment did not receive anti-TB therapy during the study period—while in reality, some patients might change their mind and choose to undergo the treatment at a much later date after the study. However, the proportion of these patients is probably very small and thus unlikely to change the results significantly.

In the context of our study population, while recognising the difficulties of diagnosing tuberculous uveitis, our results suggest that the dual-test strategy of performing TST and IGRA simultaneously appears to be the most cost-effective strategy relative to the other strategies. Our sensitivity analyses suggest that this dual-test strategy is still the most cost effective while varying the incidence of TB and tuberculous uveitis, the diagnostic accuracy of the IGRA and TST tests, and also the costs of the IGRA and TST tests for comparison with other countries such as the UK. However, further studies within the target population are required to confirm these results.

• ## Supplementary Data

This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Files in this Data Supplement:

## Footnotes

• Contributors All authors contributed significantly to the conduct of the study, analysis and manuscript preparation.

• Competing interests None.

• Patient consent Obtained.

• Ethics approval Singapore Health Services Institutional Review Board.

• Provenance and peer review Not commissioned; externally peer reviewed.

• Data sharing statement All unpublished data may be obtained from the corresponding author upon request.

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.