Cost-effectiveness of tiotropium versus omalizumab for uncontrolled allergic asthma in US

We developed a probabilistic Markov model of asthma to project the costs and quality-adjusted life years (QALYs) of patient with uncontrolled allergic asthma under different treatments over 10 years. The setting for this evaluation is patients with severe allergic asthma in the US, and the study adopts a US societal perspective. As such, in the main analysis both direct and indirect medical costs were included. The choice of time horizon was in line with the similar cost-effectiveness studies in uncontrolled asthma [15]. Three interventions were modeled: continuation of standard therapy (high dose ICS + LABA), add-on therapy with tiotropium, and add-on therapy with omalizumab. The model consisted of seven health states including three asthma control states (controlled, partially controlled, uncontrolled), three exacerbation states (non-severe exacerbations, severe exacerbations without hospitalizations, and severe exacerbations requiring hospitalizations), and a death state. The cycle length of the study was 1 week. Previous studies that assessed the cost-effectiveness of omalizumab focused mainly on modelling the effect of treatment on transitioning between exacerbation free to exacerbation states [2, 3]. However, treatments can also result in improved symptom control. To enable a more comprehensive modeling of the treatment effects on both control and exacerbation states, we built our model structure in line with the previously published model of tiotropium, which modeled transition across both control levels and exacerbation states [10]. In particular, definition of heath states in our study were in line with the PrimoTinA-asthma trials [6] and the previously published cost-effectiveness study of tiotropium [10]. The control states represent the degree of severity of the daily symptoms of the disease. Non-severe exacerbations are defined as worsening of patient’s daily symptoms over and beyond the normal variations in the current level of symptom control. Severe exacerbations without hospitalizations represent a need for an oral corticosteroids use or doubling the doses of daily medications, and hospitalizations represent an episode of exacerbation that requires inpatient care [6]. Figure 1 represents the model structure. This probabilistic model enabled the projection of costs, QALYs, and total number of exacerbations for the three interventions under the study. All the future costs and QALYs were discounted at 3%.

Multiple outcome variables from published research were used to inform the model parameters, which are presented in Table 1. Model parameters were assigned probability distributions based on the degree of uncertainty reported by the original studies. The transition probabilities for moving between different health states for the standard therapy and add-on therapy with tiotropium were derived from a recent study by Willson et al. [18]. We could not find transition probabilities between the same health states for omalizumab in the published literature. Published research has mostly focused on the impact of omalizumab on the reduction of exacerbations as well as QALY gained [8, 19]. Thereby, using these data and the same methodology as in our earlier publications, we back-calculated the transition probabilities across different health states [4]. The details of this analysis are presented in the Additional file 1: Appendix 1.

Treatment costs were derived from previously published US-based studies [2, 3]. We derived the costs of standard therapy from a study by Campbell et al. [3]. These costs have been estimated from samples with a mix of asthma control levels. We estimated the costs per levels of control based on the frequency of asthma control in the study sample and the evidence on the ratio of costs across control states [20]. The details of these calculations are shown in the Additional file 1: Appendix 2. The additional costs of omalizumab and tiotropium, as well as exacerbations costs were derived from the published literature [2, 3, 10, 21, 22]. All costs were adjusted to 2013 US dollars, details of which can be found in Table 1.

The health state utility values (utilities) were derived from a recently published study by Willson et al. [10]. The EQ-5D scores for the control states were estimated from the PrimoTinA-asthma trials and were converted into the utility values using the UK EuroQol tariff [10, 23]. The utility values for exacerbations states were derived from two recent studies by Willson et al. and LIoyd et al. [10, 24]. The details of these values and their probabilistic distributions are shown in Table 1.

We performed a main (base case) analysis by running the model with the point estimates of each parameter. Cost-effectiveness was interpreted around the reference willingness-to-pay (WTP) value of $50,000/QALY. A Monte-Carlo simulation with 10,000 runs was used to quantify uncertainty around the base case results. In each Monte Carlo run, a random sample was drawn from each distribution. From the results of the probabilistic analysis, we calculated 95% credible intervals (CrI) around point estimates of model outputs. We also generated the cost-effectiveness plane and the cost-effectiveness acceptability curve. The former is a scatterplot representing the joint distribution of the difference in costs and QALYs between the add-on therapies. The latter represents the possibility of each of the three treatments being cost-effective at different WTP values. The model was implemented in statistical programming language R (version 3.2.2) [25].

To investigate the robustness of the outcomes with respect to changes in core model parameters, we performed a series of sensitivity analyses. The parameters evaluated were time horizon (changed to life time), baseline age (changed from 20 to 60 years), discount rate (0–5%), cost of omalizumab (changed by ± 25%), cost of tiotropium (changed by ± 25%), health state costs (changed by ± 25%), utilities (changed by ± 10%), and 30-day probability of death from severe asthma exacerbations (changed from 0.01 to 0.03).