Stratified medicine in schizophrenia: how accurate would a test of drug response need to be to achieve cost-effective improvements in quality of life?

Objective Stratified medicine refers to the use of tests that predict treatment response to drive treatment decisions for individual patient. The pharmacoeconomic implications of this approach in schizophrenia are unknown. We aimed to assess the cost-effectiveness of a hypothetical stratified medicine algorithm (SMA) compared with treatment as usual (TAU), for patients with schizophrenia who failed a first-line antipsychotic. Methods A decision analytic model with embedded Markov process was constructed, which simulated the health and cost outcomes for patients followed SMA or TAU over a lifetime horizon, from healthcare and social care perspective. In the base-case analysis, sensitivity and specificity of the stratifier were both set as 60%. Extensive sensitivity analyses were conducted to test the impact of uncertainty around the value of important parameters. The primary outcome was the incremental cost per quality-adjusted life year (QALY) gained. Results When both sensitivity and specificity of the stratified test were set at 60%, SMA appeared to be the optimal strategy as it produces more QALYs and incurs lower costs than TAU. This is robust to all scenarios tested. At a willingness-to-pay threshold of £20,000 per QALY, the probability for SMA to be the optimal strategy is 82.4%. Conclusions Our results suggest that use of any stratifier with a sensitivity and specificity over 60% is very likely to be cost-effective comparing to TAU, for psychotic patients who failed a first-line antipsychotic. This finding, however, should be interpreted with caution due to lack of evidence for clozapine as a second-line antipsychotic. Electronic supplementary material The online version of this article (10.1007/s10198-019-01108-4) contains supplementary material, which is available to authorized users.

derived from the NICE guideline systematic review [2] and other published studies [16,17]. The probability of 1 developing neutropenia for patients on clozapine was obtained from Munro et al [18], which is a large-scale 2 cohort study following 12,760 clozapine-users for up to 7.6 years. The model takes the perspective of the NHS and Personal Social Services (PSS), as recommended by NICE [19].

6
The financial year is 2016. The total costs for the treatment strategies were estimated by multiplying the unit 7 costs with resource quantities. Unit costs were based on the NHS reference costs 2016- 17 [20], prescription cost 8 analysis (England) 2017 [21] or the Unit Costs of Health and Social Care 2017 [22]. The unit cost of the 9 stratified test is assumed to be £500 per patient (a conservative estimate of the cost of a magnetic resonance 10 spectroscopy scan) [23], with a range of £100-1000 tested in sensitivity analyses. Resources quantities were 11 informed by the NICE schizophrenia guideline 2014 [2] and clinician's estimates where these were unavailable.

14
HRQoL data was expressed as utilities from which quality adjusted life years (QALYs) were derived. Utility 15 weights are usually elicited on a 0-1 ratio scale of 0 (death) to 1 (perfect health). The average utilities, for 16 patients in different health states were taken from a UK study which reported separate utility data for stable or Clozapine (year 2 onwards) 0.0355 0-0.4000 As above As above Probability of relapse for patients discontinue antipsychotic treatment due to non-adherence Annual probability of relapse, first year following discontinuation of conventional antipsychotics 0.5650 0-0.8000 Beta distribution (SD assumed to be 50% of mean value) Mayoral-van et al [4] Annual probability of relapse, second year following discontinuation of conventional antipsychotics 0.2000 0-0.8000 As above As above Annual probability of relapse, year 3 onwards following discontinuation of conventional antipsychotics 0.0632 0-0.8000 As above As above  This appendix reports results of one-way and two-way sensitivity analysis in Section 2.1 and 2.2, respectively.

One-way sensitivity analysis results
The conclusion of the base case analysis (SMA being the most cost-effective strategy) was robust to all scenarios tested. The detailed results of one-way sensitivity analysis are reported in Table 3. 1. There are two scenarios under which SMA could be considered to be cost-effective: • Scenario 1. Compared with TAU, SMA is less costly and more clinically effective. In this case, SMA 'dominates' TAU and no further justification is necessary.
• Scenario 2. Compared with TAU, SMA is more effective, but also more expensive. In this case, the decision whether to implement the SMA would depend on how much the payer of healthcare (the NHS in the UK) is prepared to pay per additional unit of QALY. The incremental cost-effectiveness ratio (ICER) is the ratio of the difference in cost divided by the difference in QALYS and is expressed in UK pounds per additional QALY. In line with the NICE reference case, a willingness-to-pay (WTP) threshold of £20,000 per additional QALY was used. Thus, if the ICER of SMA versus TAU is less than £20,000 per QALY, then SMA is more cost-effective than TAU.

Two-way sensitivity analysis results
The results of two-way sensitivity analysis are presented in Figure 1 below. Figure 1A reports the combined effects of sensitivity and specificity of the stratifier on the results. It shows that: • If the sensitivity of the stratifier is 0%, as long as the specificity of the test is no less than 11.50%, SMA is more cost-effective than TAU • If the sensitivity of the stratifier is 50.00%, as long as the specificity of the test is no less than 6.00%, SMA is more cost-effective than TAU Figure 1B reports the combined effects of: proportion of clozapine responder in patients who failed a first-line antipsychotic, and proportion of AP2 responders who respond to clozapine. The results show that: • If 50% of the AP2 responders would respond to clozapine, as long as the proportion of clozapine responder is no less than 23.75%, SMA is more cost-effective than TAU • If none of the AP2 responders would respond to clozapine, as long as the proportion of clozapine responder is no less than 35.62%, SMA is more cost-effective than TAU