An assessment of nucleic acid amplification testing for active mycobacterial infection

Structure and rationale of the economic evaluation

Economic literature review

A literature search was conducted to identify published economic evaluations of NAAT for active TB infections (in those who can have an AFB) and to inform the structure of and inputs to the economic model (see Appendix ).

Five studies were identified that investigated the cost-effectiveness of NAAT in low-prevalence populations, as these are the most relevant to the Australian population (Table ) (Choi et al. 2013; Dowdy et al. 2003; Hughes et al. 2012; Millman et al. 2013; Rajalahti et al. 2004).

Table Economic evaluations identified that investigate NAAT for active TB in low-prevalence countries

Study	Setting	Results
Millman et al. (2013)	Decision tree analysis of adult inpatients in US hospital setting who have presumed TB and are in isolation until results of diagnostic tests (AFB compared with NAAT) become available. Differences in health outcomes were not anticipated, and so net costs were determined, which considered savings associated with the reduction in unnecessary hospitalisations and isolations. The cost implications of FPs and FNs were additionally considered as a cost penalty.	NAAT was associated with cost savings due to reduced hospital isolation and reduced overall length of stay.
Choi et al. (2013)	Decision tree cost–utility analysis of individuals with suspected pulmonary TB in the USA. A single-year time horizon was used for mapping the decision analytic, after which extrapolation extended the time horizon to the life expectancy of the patients. Models are for HIV-negative and HIV-positive patients (considering different epidemiological and accuracy estimates, but same utility weights), and include outcomes of resistance mutation testing. Costs included lab testing, hospitalisation, isolation and treatment. Implications for FPs were considered. Multiple testing algorithms were modelled. Algorithm 1 (no molecular testing) is relevant to comparator, with algorithms 3 and 5 relevant to proposed NAAT in Australia. Treatment may be initiated in AFB-negative, NAAT-negative if clinical suspicion is high (i.e. clinical diagnosis) (any algorithm)	Testing without NAAT was dominated by the strategies that included NAAT.
Hughes et al. (2012)	Decision tree cost–utility analysis of NAAT for people with a clinical suspicion of TB in the UK setting. Time horizon chosen of 1 year. Model incorporates resistance testing and outcomes in FP and FNs. Costs include testing, treatment and follow-up outpatient consultations; isolation costs were not considered. The model did identify the number of people infected by unidentified TB, but attributed neither their costs nor outcomes into the results. Strategies relevant to this model include #3: AFB and culture every time (for the comparator); and #11: AFB, NAAT and culture every time (for NAAT).	Strategy #11 was unlikely to be cost-effective compared with #3 (ICER £64,723). If secondary infections were incorporated fully into the model, the authors conclude that it could be conceivable that #11 would be optimal, as it was associated with the fewest secondary infections.
Rajalahti et al. (2004)	Decision tree cost-effectiveness analysis of AFB and culture ± NAAT in patients with a clinical suspicion of TB in the Finnish setting. Effectiveness was measured in terms of correct treatment and isolation decisions. Costs included isolation, treatment, lab tests and inpatient/outpatient visits. Decision tree parameters populated based on observed data.	NAAT was associated with additional costs when applied in all patients, but cost savings were only in AFB-positive patients.
Dowdy et al. (2003)	Decision tree cost-effectiveness analysis of NAAT in AFB-positive patients in the US setting. Effectiveness was measured in terms of ‘early exclusion of TB’. Costs included testing, isolation and treatment. Unclear if all patients were subject to C&S testing.	NAAT was not considered cost-effective, as costs did not offset those of isolation and treatment averted.

AFB = acid-fast bacilli; C&S = culture and sensitivity; FN = false negative; FP = false positive; HIV = human immunodeficiency virus; NAAT = nucleic acid amplification testing; TB = tuberculosis

All 5 studies were generally consistent in structure (decision tree) and time horizon (up to 1 year). Three of the studies considered the implications of false-positive and false-negative results, and most considered the cost of hospital isolation. However, the outcomes of the models varied; 2 studies (Choi et al. 2013; Hughes et al. 2012) measured outcomes in terms of cost per QALY, whereas the other 3 investigated cost per correct treatment or isolation decision, or just costs as no change in outcomes were anticipated. The studies additionally varied in their results, with NAAT considered cost-effective in 3 studies and not cost-effective in 2.

None of the identified studies were conducted in the Australian setting. In Australia treatment initiation decisions take the clinical suspicion of TB into consideration. However, clinical suspicion was not considered in any of the studies identified, and therefore the applicability of the identified economic evaluations to the Australian context is uncertain. A modelled economic evaluation will be presented to determine the cost-effectiveness of NAAT (as an add-on test) in the population who can currently have an AFB.