An assessment of nucleic acid amplification testing for active mycobacterial infection

Time horizon
20 months
Outcomes	QALYs
Costs	Australian dollars, 2014 prices
Methods used to generate results	Decision tree analysis
Discount rate	5% costs and outcomes accrued beyond 1 year
Software packages used	Microsoft Excel

QALY = quality-adjusted life-year

The structure of the decision tree is presented in Figure (AFB model arm) and Figure (AFB plus NAAT model arm).

Currently, patients with true TB are likely to be mixed across the populations that have a high or low clinical suspicion of TB. The prevalence of TB in each of these patient groups is likely to vary, as it would be expected that those with a high clinical suspicion of TB would have a higher prevalence than those with a low clinical suspicion of TB. These assumptions are used to inform the base-case scenario (‘TB mixed scenario’). However, given the influence of clinical judgement on the treatment management pathways, and the uncertainties associated with estimating the relative mix of patients across these groups (see ‘Prevalence of TB’), the influence of clinical judgment on the cost-effectiveness of NAAT will be explored through the addition of the following scenarios:

• 
  TB low suspicion scenario: all patients (including all with true TB) are treated as though they have a low clinical suspicion of TB (i.e. clinical judgment is not used as a basis to initiate treatment)
  
• 
  Perfect clinical judgment scenario: all patients with true TB are treated as though they have a high clinical suspicion of TB (i.e. clinical judgement is used as a basis to initiate treatment, and it is assumed that this has 100% sensitivity and specificity in identifying TB), and all patients without TB are treated as though they have a low clinical suspicion of TB (i.e. treatment initiation decisions are based on results of AFB ± NAAT)
  
• 
  TB high suspicion scenario: all patients are treated as though they have a high clinical suspicion of TB (i.e. treatment is initiated in all patients on the basis of clinical judgment).
  

These additional scenarios are considered to be extreme cases. NAAT is expected to be most cost-effective in the TB low scenario, as it is associated with more benefits in those considered to have a low clinical suspicion of TB. In contrast, NAAT is also expected to be least cost-effective in the scenarios in which all patients with TB, and with or without TB, respectively, are managed as though they have a high clinical suspicion of TB. In these scenarios treatment initiation decisions are based on clinical judgement, with the benefit of NAAT restricted to identifying drug resistance to initiate an appropriate treatment earlier. In the perfect clinical judgement scenario all true TB-negative patients are treated as though they have a low clinical suspicion of TB, and so treatment decisions are based on the results of AFB ± NAAT and only false-positive patients will receive treatment (determined by the specificity of testing). The relative cost-effectiveness of NAAT between these extreme high and low scenarios is likely to be determined by the relative specificity of NAAT compared with AFB.

The cost-effectiveness of the TB mixed scenario, which is thought best to reflect current practice, is likely to lie between the extreme additional scenarios.

Model assumptions

• 
  When AFB and NAAT are discordant, the treatment decision is based on NAAT (consistent with PASC protocol)
  
• 
  C&S testing (the reference standard) is assumed to be 100% sensitive and specific, as all patients have C&S testing and at the end of 2 months all will have correct diagnosis (i.e. MDR-TB, TB or no TB)
  
• 
  To simplify the model structure, rifampicin resistance is used as a surrogate marker of MDR-TB (Lumb 2000), as the majority (37/40) of Australian bacteriologically confirmed cases in 2010 with rifampicin resistance were also MDR (Lumb et al. 2013)
  
• 
  Once the decision to initiate or delay treatment has been made, the model assumes there will be no change in treatment until the results of C&S are available; this assumption may favour NAAT, as the earlier initiation of resistant drugs in the comparator arm would reduce the benefit of introducing NAAT
  
• 
  Cost and utility penalties associated with the secondary transmission of TB are applied for each index case in the model, but the consequences (cost or health outcome) of further ongoing transmissions (e.g. tertiary transmissions and beyond) are not included in the base-case.
  

Figure Decision analytic structure of the economic evaluation, comparator (AFB) model arm

AFB = acid-fast bacilli test; C = culture; High_pretest = proportion of patients considered to have high clinical suspicion of TB; MDR = multidrug-resistant; RIF res = rifampicin resistant; RIF suscept = rifampicin susceptible; Rif_res = prevalence of rifampicin resistance in TB; sensAFB = sensitivity of AFB for TB; specAFB = specificity of AFB for TB; S = susceptibility; TB = tuberculosis; TB_high = prevalence of TB in high clinical suspicion population; TB_low = prevalence of TB in low clinical suspicion population

Figure Decision analytic structure of the economic evaluation, intervention (AFB plus NAAT) model arm

AFB = acid-fast bacilli test; C = culture; MDR = multidrug-resistant; NAAT = nucleic acid amplification test; R = resistance; RIF res = rifampicin resistant; RIF suscept = rifampicin susceptible; Rif_res = prevalence of rifampicin resistance in TB; sensAFB = sensitivity of AFB for TB; sensNAAT_AFBn = sensitivity of NAAT for TB in AFB-negative; sensNAAT_AFBp = sensitivity of NAAT for TB in AFB-positive; sensNAAT_rif_res = sensitivity of NAAT for rifampicin resistance; specAFB = specificity of AFB for TB; specNAAT_AFBn = specificity of NAAT for TB in AFB-negative; specNAAT_AFBp = specificity of NAAT for TB in AFB-positive; specNAAT_rif_res = specificity of NAAT for rifampicin resistance; S = susceptibility; TB = tuberculosis; TB_high = prevalence of TB in high clinical suspicion population; TB_low = prevalence of TB in low clinical suspicion population

Implications for false-positive and false-negative results

The decision trees presented in Figure and Figure culminate in nine different categories according to whether a true or false result is initially concluded (referred to as ‘outcome states’). These are summarised in Table .

Table Summary of decision tree outcome states in the economic evaluation

True status	Treated status	Implication
No TB	Untreated (TBTN)	Correct no treatment
No TB	Standard treatment (TBFP, TRN)	Standard treatment initiated, stop treatment on C&S results
No TB	MDR treatment (TBFP, FRP)	MDR treatment initiated, stop treatment on C&S results
TB	Untreated (TBFN)	No treatment initiated, begin standard treatment on C&S results
TB	Standard treatment (TBTP, TRN)	Correct standard treatment
TB	MDR treatment (TBTP, FRP)	MDR treatment initiated, switch to standard treatment on C&S results
MDR-TB	Untreated (TBFN)	No initial treatment initiated, begin MDR treatment on C&S results
MDR-TB	Standard treatment (TBTP, FRN)	Standard treatment initiated, switch to MDR treatment on C&S results
MDR-TB	MDR treatment (TBTP, TRP)	Correct MDR treatment

C&S = culture and sensitivity; FRN = false resistance negative; FRP = false resistance positive; MDR = multidrug-resistant; TB = tuberculosis; TBFN = tuberculosis false negative; TBFP = tuberculosis false positive; TBTN = tuberculosis true negative; TBTP = tuberculosis true positive; TRN = true resistance negative; TRP = true resistance positive

False-negative results (i.e. initially untreated TB (± MDR) or initial standard treatment in MDR-TB)

As there was no indication from the clinical assessment that a treatment delay of up to 2 months leads to an increase in disease severity (van der Oest, Kelly & Hood 2004), the economic modelling will assume treatment duration and QoL (from the time of correct diagnosis) as for those correctly treated. However, there is some indication that a delay in treatment leads to an increased risk of TB transmission (Ponticiello et al. 2001). A cost and utility penalty are applied to account for the treatment costs and utility decrement associated with secondary infections; see ‘TB transmissions’ and Utility penalty for active TB transmissions’ for further details.

Treatment outcomes in MDR-TB patients treated initially with the standard regimen are assumed to be poorer than for those initially untreated, as treatment is ineffective and associated with AEs (i.e. outcomes equal to those untreated who then have a disutility associated with treatment applied).

False-positive results (i.e. initial TB (± MDR) treatment in true-negative patients or MDR-TB treatment in susceptible TB)

Patients that are truly negative for TB who undergo initial TB (± MDR) treatment are assumed to have the cost and disutility of 2 months of the applicable treatment applied. As these patients may have a range of alternative diagnoses that present with similar symptoms (associated with differing costs and outcomes), the delay to treatment for the alternative diagnosis is not considered in the assessment.

Patients with true susceptible TB that are treated initially with MDR regimen are assumed to be effectively treated but have poorer overall health outcomes than those treated with the standard regimen because of the increased AEs associated with MDR treatment.