In previous sections, we described the significant burden of severe asthma and the efficacy of omalizumab in terms of significantly reduced exacerbation and emergency visit rates and significantly improved quality of life and day-to-day symptoms. The incremental therapeutic value of omalizumab, which is the sum of the effect on day-to-day asthma, clinically significant exacerbations, ER visits, risk of fatal asthma and resource utilization, has also been assessed against the incremental costs of therapy. The incremental cost per quality-adjusted life year (QALY) gained is a well accepted and widely used measure of the cost-effectiveness of therapies and can capture the overall treatment effect on the patient (65).
To date, the cost-effectiveness of omalizumab add-on therapy has been assessed in several analyses (58–60, 66, 67). Four analyses assessed cost-effectiveness in patients with inadequately controlled severe persistent asthma despite high-dose ICS plus a LABA (58–60). The analysis by Dewilde et al. (58) was based on the efficacy data from the INNOVATE study (48) (reference country Sweden). One analysis by Brown et al. (59) was based on the efficacy data from the ETOPA study (49, 62) (reference country Canada), and a second (containing two base-case analyses) (60) was based on data from INNOVATE and ETOPA (reference country the Netherlands). The analysis by Oba and Salzman (66) was based on data from two randomized, placebo-controlled trials in patients with inadequately controlled moderate-to-severe allergic asthma despite receiving ICS (51–54) (reference country: USA). The analysis by Wu et al. (67) was based on the medical care component of the US Consumer Price Index (68). The funding source of a cost-effectiveness analysis is also of interest when interpreting results. Indeed, the pharmaceutical industry must provide cost-effectiveness data to secure reimbursement from health authorities. In this review, three cost-effectiveness analyses were industry sponsored (58–60) and two were investigator sponsored (66, 67).
In the cost-effectiveness analyses of omalizumab in patients with inadequately controlled severe persistent allergic asthma despite high-dose ICS plus a LABA (58–60), the same Markov cohort model was used to estimate the cost-effectiveness of adding omalizumab to standard therapy. The Markov model has been described in detail by Dewilde et al. (58). Lifetime costs were determined by comparing lifelong standard therapy costs with 5 years of omalizumab add-on therapy followed by standard therapy. As only patients who are judged by physicians to have responded to omalizumab after 16 weeks of therapy should continue with treatment (EU label), nonresponders to omalizumab revert to standard therapy at 16 weeks. In the INNOVATE study, responders were defined as those patients with at least a marked improvement in asthma control (physician’s overall assessment). In the ETOPA study, responders were defined as those patients with ≥ 0.5-point improvement in Mini-AQLQ overall score. The benefits of treatment were expressed as QALYs gained and the cost-effectiveness of omalizumab was expressed as an incremental cost-effectiveness ratio (ICER). The ICER was calculated as the difference in total costs between the omalizumab and standard treatment arms per QALY (cost/QALY). Probabilistic analyses of cost-effectiveness over a range of willingness-to-pay values were also performed. The acceptance by the UK National Institute for Health and Clinical Excellence (NICE) of the submission from the manufacturers of omalizumab, which used this model and data from the INNOVATE primary-intention-to-treat population and a high-risk hospitalization subgroup, along with an additional analysis using ETOPA study data supports the validity of this model (69).
The analyses presented in the Oba and Salzman report (66) differed in several substantive ways from the research described above. The investigators constructed efficacy outcomes to assess cost-effectiveness, which was expressed as the daily cost to achieve an additional successfully controlled day or as the daily cost to achieve a >0.5-point improvement in AQLQ overall score. These outcomes, while important clinically, are of lower importance to resource decision makers who require estimates of the incremental cost/QALY. Importantly, the analyses did not account for whether or not a patient responded to omalizumab and used clinical data from a less severe patient population.
The most recent paper by Wu et al. (67) used data from the Asthma Policy Model (70) to estimate the 10-year cost-effectiveness of omalizumab added to ICS controller medication [with short-acting β2-agonist (SABAs) relievers when required] compared with ICS controller medication alone (with SABA relievers) in a severe adult asthma population. These results suggest that omalizumab, in combination with ICS, is not a cost-effective treatment strategy. Several concerns regarding the findings from this research should be noted. First, the analytic engine underlying the Wu et al. model relies heavily on the relationship between health care utilization and lung function parameters (FEV1). Omalizumab is known to exert its clinical benefits in asthma patients by reducing the risk of exacerbation and improving asthma-specific quality of life, not by improving lung function perse. Second, the analysis reports cost/QALY findings for an average cohort of severe patients. Omalizumab is recommended in the GINA guidelines as add-on treatment to combination therapy with ICS and LABA. A more relevant scenario for Wu et al. to have explored would have been to compute cost-effectiveness of omalizumab when used according to guidelines. That is, when omalizumab is added to a combination of ICS and LABA for patients who failed ICS and LABA combination. Finally, the clinical and economic value of omalizumab is seen most clearly in patients who respond to initial treatment. We describe this research in more detail below.
In the reference case scenario using INNOVATE data (Sweden), estimated total lifetime discounted costs were €52 702 for 11.6 QALYs for standard therapy (58). Add-on omalizumab therapy had an additional lifetime cost of €42 754 for an additional 0.762 QALYs (ICER of €56 091) (Table 2A). Using ETOPA data (Canada), total lifetime discounted costs and QALYs for patients on standard therapy were €27 403 and 6.49 (59). Add-on omalizumab therapy had an additional lifetime cost of €33 854 for 1.08 additional QALYs (ICER of €31 209) (Table 2A). INNOVATE data (the Netherlands) produced an estimated total lifetime discounted cost of €29 476 for 11.2 QALYs (60) for standard therapy and an additional lifetime cost of €44 641 for 1.0 QALY (ICER of €44 910) for add-on omalizumab (Table 2B). ETOPA data (the Netherlands) gave a total lifetime discounted cost of €30 144 for 9.7 QALYs (60) for standard therapy and an additional lifetime cost of €32 044 for 1.2 QALYs (ICER of €26 694) for add-on omalizumab (Table 2B).
Table 2A. One-way sensitivity analyses of omalizumab add-on vs standard therapy for severe persistent allergic asthma (58, 59)
|Analyses description||INNOVATE (reference country - Sweden)||ETOPA (reference country - Canada)|
|Incremental cost, €||Incremental QALY||ICER, €||Incremental cost, €||Incremental QALY||ICER, €|
|Base case||42 754||0.762||56 091||33 854||1.08||31 209|
|No discounting||48 340||1.078||44 858||40 096||1.69||23 762|
|Mortality = 0%||43 881||0.335||131 130||31 495||0.47||66 443|
|Mortality = upper range||41 634||0.900||46 268||33 546||1.00||33 578|
Table 2B. One-way sensitivity analyses of omalizumab add-on vs standard therapy for severe persistent allergic asthma (60)
|Analyses description||INNOVATE (reference country - the Netherlands)||ETOPA (reference country - the Netherlands)|
|Incremental cost, €||Incremental QALY||ICER, €||Incremental cost, €||Incremental QALY||ICER, €|
|Base case||44 641||1.0||44 910||32 044||1.2||26 694|
|No discounting||50 126||1.2||42 629||36 507||1.4||26 262|
|Mortality = 2.478%||43 715||0.3||127 286||30 691||0.5||60 650|
|Mortality = upper range||44 542||0.9||44 395||31 911||1.1||28 647|
The ICER range for one-way sensitivity analyses using INNOVATE data (Sweden) was €39 762 without discounting to €131 130 without inclusion of asthma-related mortality (Table 2A). The ICER range for one-way sensitivity analyses using ETOPA data (Canada) ranged from €23 762 without discounting to €66 443 without inclusion of asthma-related mortality (Table 2A). Using INNOVATE data (the Netherlands) the ICER range for one-way sensitivity analyses ranged from €42 629 without discounting to €127 286 without inclusion of asthma-related mortality (Table 2B). Using ETOPA data (the Netherlands), the ICER range for one-way sensitivity analyses ranged from €26 262 without discounting to €60 650 without inclusion of asthma-related mortality (Table 2B). The probabilistic cost-effectiveness planes for omalizumab add-on therapy vs standard therapy for INNOVATE (Sweden), ETOPA (Canada) and ETOPA (the Netherlands) data are shown in Fig. 5A–C, respectively. Each point is defined on the horizontal axis by the incremental difference in QALYs for omalizumab relative to standard therapy alone and the vertical axis depicts the incremental difference in cost. The horizontal dispersion is a reflection of the variation in QALYs.
Figure 5. Cost-effectiveness plane for omalizumab add-on therapy vs standard therapy; using (A) INNOVATE data (reference country - Sweden) (58), (B) ETOPA data (reference country - Canada) (59) and (C) ETOPA data (reference country - the Netherlands) (60).
Download figure to PowerPoint
The potential of input parameters to influence the results of the model should be considered in analyses of this kind. The analyses using the Markov model (58–60) relied on several assumptions. For example, it was assumed that future events (hospitalizations, exacerbations) are independent of previous events. Numerous published studies have shown that previous exacerbations or hospitalizations are positively correlated with future exacerbations or hospitalizations (17–23, 71). An analysis of INNOVATE data found that a history of clinically significant exacerbations in the year prior to the study was highly predictive of exacerbations during the study (72). Also, the INNOVATE trial had an 8-week run-in period during which asthma therapy was optimized (48). While the optimization on therapy in the INNOVATE trial may not always reflect the case in clinical practice, it is worth noting that optimization of therapy should be an ongoing treatment goal to maximize the benefits of standard therapy before more expensive options are considered. Both analyses assumed adherence to therapy although adherence in long-term illness is generally accepted to be poor. Dewilde et al. suggested that adherence to therapy was likely to be higher in omalizumab-treated patients given the increased physician contact for purposes of omalizumab administration (58). Brown et al. acknowledged that the assumption of adherence was optimistic, but considered it not to unduly favor one treatment group over the other (59). The analyses did not include any adverse event (AE) costs or utility decrements for either treatment group as AE rates were not statistically significantly different. AEs associated with long-term use of OCS in standard therapy were not considered, as no comparable long-term data for omalizumab exist.
Brown et al. (59) highlighted the extrapolation of 1-year trial data to lifetime and the lowering of the resulting ICER as a study limitation, but argued that this extrapolation is justified, given that severe persistent allergic asthma is a chronic condition. They also used non-trial data to estimate the risk of asthma-related mortality associated with clinically significant severe exacerbations, and assumed a mortality risk from severe exacerbations despite the fact that there were no deaths in the placebo arm of the 1-year ETOPA trial (there was one death in the Xolair-treated group, but post hoc analyses showed that it was unrelated to treatment). The authors considered the inclusion of the possibility of asthma-related death to be appropriate in a lifetime analysis of severe allergic asthma patients as asthma-related fatalities are known to occur, and the subgroup of patients included in this study was at a high-risk of asthma-related mortality based on prior medical history (14, 73). This inclusion of non-trial data may be seen as a potential study limitation, and the appropriateness of its source [a small observational study (74), which was not designed to examine the relationship between severe asthma exacerbations and death] has been questioned (75). However, the clinical trial environment is an exceptional one, with greater levels of patient care and medication compliance than is standard, and sudden deteriorations are more likely to be rapidly treated, thereby reducing the risk of death. Therefore estimates based on ‘real-world’ mortality data as reported by Lowhagen et al. (74) may be justified. Other published mortality data (76, 77) support the estimates used in the model. It should also be noted that the authors conservatively assumed that no risk of fatal events is associated with clinically significant exacerbations, and a range of estimated fatality rates were presented in the results for comparison. During the NICE assessment of the HTA submission from the manufacturers of omalizumab, the Committee was of the opinion that the mortality risk in the submission was overestimated. However they also considered that the high-risk hospitalization subgroup from the INNOVATE trial would plausibly have a higher risk of asthma mortality (78). Both the risk of mortality associated with severe asthma and the considerable uncertainty in estimating that risk are documented. The results of a confidential inquiry indicated that disease severity was usually a major factor in deaths attributed to asthma (25). The problems inherent in using currently available data to estimate asthma mortality are illustrated by a review of asthma-related deaths, which concluded that relying on underlying cause-of-death data may underestimate the actual number of deaths due to asthma (79). The review found that only 45% of over 135 000 investigator-identified asthma-related deaths had asthma recorded as the underlying cause of death. Another study into mortality in hospitalized patients in the US found the overall in-hospital mortality to be 0.5%, and this represented approximately one third of all asthma deaths reported in the US (80). Noting that research data on fatal asthma are limited, a study to determine the feasibility of creating a US national fatal asthma registry concluded that, as fairly consistent demographic and clinical data were available across states, the creation of a national fatal asthma registry could be feasible if legislative barriers could be overcome (81).
In Canada, the omalizumab label is not restricted to those with severe asthma (82). Therefore, the consistency of using Canada as the reference country in the analysis by Brown et al. (59) examining a severe population may be questioned. However, as a subgroup profile that mirrors the European omalizumab label population can be applied to any database regardless of country label specifics, this approach is acceptable. Dewilde et al. (58) also pointed out that data from a multi-country trial is adaptable to the Swedish database because of the wide acceptance of GINA treatment guidelines and the severe patient populations at GINA 2002 step 4 therapy are similar in their inability to achieve asthma control despite best standard care.
The definition of treatment response in Brown et al. (59) was based on improvements in the Mini-AQLQ, which was also used in the mapping algorithm to generate the utility values. Unsurprisingly, omalizumab responders by definition have markedly higher utility values than the standard therapy arm, and the question of whether results using the overall omalizumab group should be presented is raised. The authors argue that using the overall omalizumab group response after 16 weeks would not reflect the fact that only responders continue with therapy. The AQLQ, has been shown to correlate well with the physician’s 16-week evaluation of treatment (61) included in the model and was adequate to mimic responder identification in the ETOPA trial (data relating to the physician’s 16-week assessment were not collected). The model used the overall omalizumab group Mini-AQLQ scores and clinically significant exacerbation rates for the first 16 weeks. After this, when only responders would continue with treatment, responders had exacerbation rates and utility values based upon the responder Mini-AQLQ scores, while nonresponders (who stop omalizumab) were assumed to have those of the placebo group.
Utility values for day-to-day symptoms were obtained from AQLQ (Dewilde et al.) and Mini-AQLQ (Brown et al.) data from the clinical trials, but as there were insufficient observations of clinically significant and severe exacerbations in trials, utility values relating to these states were taken from a prospective four-centre study (74). The cost estimates included by Dewilde et al. in their model were yearly drug costs (standard therapy, based on patient usage in INNOVATE) and the additional weighted average cost of omalizumab therapy including administration costs, an assumed 3-monthly GP visit for asthma, and 16-week assessment. Exacerbation and severe exacerbation costs were based on INNOVATE data. Nonhealthcare related costs in added years of life were included as recommended by the Swedish Pharmaceutical Benefits Board. The effect of excluding this would have increased the calculated ICER. Brown et al. included drug costs and significant exacerbation costs based on ETOPA data and severe exacerbation costs based on INNOVATE. Administration costs were not included in the base case (as these are supported by Novartis in Canada), and only direct medical costs were included.
The analysis by Oba and Salzman reported a daily cost of $523 to achieve an additional symptom free day with omalizumab and a daily cost of $378 to achieve a ≥ 0.5-point improvement in AQLQ overall score in patients with moderate-to-severe allergic asthma (66). The authors concluded that omalizumab may be cost saving if given to nonsmoking patients who are hospitalized five or more times or 20 or more days or longer per year despite maximal therapy.
Wu et al. estimated a cost-effectiveness ratio of $821 000 per QALY gained for patients with acute event rates at baseline. A sensitivity analysis allowing the acute event rate to increase to five times that at baseline, they estimated a cost-effectiveness ratio of $491 000 per QALY gained. Other scenarios included in sensitivity analyses were efficacy of omalizumab (in terms of exacerbation rate, varied between 33% and 92%) which produced costs per QALY of $853 000 and $727 000 respectively; QoL improvement (health-related quality of life [HRQoL] improvement from 0% to 7.2%) which produced costs per QALY of $1.42 million and $207 000 respectively; and cost of omalizumab (monthly drug costs of $200) producing a cost per QALY of $100 000. The authors concluded that, at its current price, omalizumab is not cost-effective for patients with severe asthma in the US.
Given a willingness-to-pay of €60 000, the studies using country-specific data in the Markov model (58–60) support an indication that omalizumab is cost-effective in patients with severe persistent allergic asthma despite high-dose ICS plus LABA. Using ETOPA data applied to Canada gave an ICER of €31 209 for omalizumab (59), which was lower than that reported using data from the INNOVATE study applied to Sweden (€56 091) (58). Similarly, ETOPA data applied to The Netherlands gave in ICER of €26 694 which is lower than the €44 910 ICER from INNOVATE data applied to the same country (60). As the ETOPA study was a 1-year randomized open-label trial in a naturalistic setting and INNOVATE was a 28-week, randomized placebo-controlled trial it is likely that the ETOPA data may be more representative of what would be expected in clinical practice. Data reported by Oba and Salzman (66) are more difficult to interpret as these data were based on patients with less severe asthma, artificial endpoints were used to measure cost-effectiveness, and response to omalizumab was not evaluated. Although Wu et al. (67) based their estimate on a severe population, the Asthma Policy Model they used was developed with data from patients with mild-to-severe asthma and has not been validated in patients who are more severe. Improvement in HRQoL is one of the key drivers in their model and is based on a FEV1 prediction equation. However, as omalizumab has little impact on FEV1, and FEV1 does not relate to QALYs particularly well, in this context, the model does not fully reflect the important effects of Xolair. It is therefore not surprising that the authors found that omalizumab had a very low impact on HRQoL (0.9%) and that the estimated ICER increased accordingly. When HRQoL improvement is assumed to be 7.2% in the sensitivity analysis, the ICER is reduced significantly. Responders to omalizumab were not identified in this analysis which means that all patients receive omalizumab treatment for 10 years regardless of response status. While this is acceptable as the US label does not require the identification of responders, it should be noted that this has implications on the results, particularly where omalizumab is used according to the EU label. Some level of discontinuation based on nonresponse should have been assumed in the sensitivity analysis, based on the likelihood that clinicians would not let patients continue with treatment if there is no response. Additionally, in some cases, patients would not continue with treatment if they are not experiencing benefit. Drug costs for omalizumab are therefore overestimated, as in clinical practice nonresponders are likely to have treatment discontinued.