The current review is for satralizumab (Enspryng) for adolescents and adults with neuromyelitis optica spectrum disorder (NMOSD) who are anti–aquaporin 4 seropositive (AQP4+).
Economic Evaluation
Summary of Sponsor’s Economic Evaluation
Overview
The sponsor submitted 2 separate cost-utility analyses assessing satralizumab, with or without maintenance IST, for the treatment of adolescents and adults with NMOSD who are AQP4+. The 2 models reflected satralizumab as monotherapy and in combination with immunosuppressive therapy (IST).1 Satralizumab as monotherapy was compared with no treatment in one model while, in the other, satralizumab in combination with maintenance IST was compared with maintenance IST alone. Maintenance IST were aligned with the treatment administered to patients during Study 898 (i.e., azathioprine, mycophenolate mofetil, or corticosteroids). The modelled population was aligned with the AQP4+ subgroup populations from Study 898 and Study 900, which were aligned with the Health Canada–indicated population.
The recommended dose for satralizumab is 120 mg via subcutaneous injection at 0, 2, and 4 weeks followed by a maintenance dose of 120 mg every 4 weeks thereafter. At a submitted cost of $9,450 per pre-filled syringe, the annual cost of satralizumab in the first year is $132,300 and $113,400 in subsequent years.1 In the economic model for satralizumab as monotherapy, no drug costs were assumed for patients on no treatment while, for satralizumab in combination with IST, the total annual costs of maintenance IST (azathioprine, mycophenolate mofetil, oral corticosteroids) were $342.81, and these costs were assumed to be the same in both the satralizumab plus IST and IST alone arms. Patients were assumed to remain on maintenance IST over their lifetime.1
The economic analysis was conducted from the perspective of the public health care payer over a lifetime time horizon (defined as 50 years). Costs and clinical outcomes (i.e., QALYs) were discounted at a rate of 1.5% per annum.1
Model Structure
In both of the sponsor’s submitted models, a common Markov model structure was employed to capture the long-term costs and effects of a relapsing disease course that is associated with increasing disability, with each cycle reflecting 28 days. The disease course was modelled by health states defined by the Expanded Disability Status Scale (EDSS) to reflect the impact of relapses on disease severity and quality of life.1 In total, the model consisted of 10 EDSS health states (i.e., EDSS 0 to 9) along with an absorbing death health state. Patients entered the model across a distribution of baseline EDSS scores and could either remain in the same EDSS state (reflecting stable disease), experience a relapse (i.e., move to a higher EDSS state), or die after each cycle.1 Structurally, the model assumed that patients could accrue disability only by moving up to a higher EDSS score and could not improve by moving down to a lower EDSS score following recovery from a relapse.1 Patients were assumed to remain on satralizumab until discontinuation. Upon discontinuation, patients on satralizumab monotherapy were assumed to not receive any subsequent treatment, whereas patients on satralizumab plus IST were assumed to remain on maintenance IST. Treatment discontinuation was not modelled for the comparator arms (no treatment or maintenance IST).1
Model Inputs
The patient cohort was identical between the 2 analyses. It comprised NMOSD patients who were AQP4+ who were pooled from each of the AQP4+ subgroups within studies 898 and 900 (mean age = 42.70 years; mean weight = ▬ kg). Approximately 13% of the patient population included males.
In the economic model, the treatment efficacy of satralizumab was measured as time to first protocol-defined relapse. The comparative efficacy of satralizumab compared with no treatment and of satralizumab plus IST compared with IST alone was derived directly from studies 900 and 898, respectively. The sponsor assumed a constant relapse ratio between satralizumab and the comparator treatments until treatment discontinuation, whereupon these patients would then be modelled based on their expected natural history.1 To model the natural history of this condition (i.e., the comparator arms within the analyses), the probability of transitioning between EDSS states was informed by observational data of NMOSD patients obtained from a global disease-specific registry (NMOBase). EDSS transition matrices were derived for the comparator of each analysis (i.e., no treatment for satralizumab as monotherapy and IST for satralizumab as an add-on therapy) by identifying patient cohorts from NMOBase that matched the baseline characteristics of the comparator population from the pivotal clinical trials. For no treatment, the matching was performed against all patients in the NMOBase and was not limited to patients who were not previously exposed to relapse-prevention therapy. For maintenance IST, the matching was performed on only AQP4+ patients and was limited to patients with prior exposure to relapse-prevention therapy.1
Model-fit statistics were used to inform the best statistical distribution that fitted the Kaplan-Meier curves for time -to- off-treatment from studies 898 and 900 in order to estimate the proportion of patients who were alive and continued to receive satralizumab until discontinuation in each comparison.1 Treatment-related adverse events of grade 3 or higher and serious adverse events were considered in the model based on the incidence of adverse events in studies 898 and 900 and informed each comparison. Background mortality was modelled using all-cause age- and sex-specific mortality rates from the general population.1 Excess risk of mortality in patients with NMOSD compared with the general population was adjusted using standardized mortality ratios for patients with MS and an additional inflation factor to further account for the higher level of disability and death among NMOSD patients.1
Treatment-specific health state utility data for EDSS health states 1 to 6 were derived by pooling studies 898 and 900 to develop a linear mixed model for repeated measurements that predicted utility weight based on several covariates.1 As studies 898 and 900 did not have patients in the remaining EDSS health states (i.e., 0 and 7 to 9), published literature on the population of patients with MS was used to inform the utility weights.1 A short-term disutility was applied upon relapse to capture its impact on health-related quality of life. A disutility for visual impairment was applied to a proportion of patients with severe visual impairment in each EDSS state.2 A constant proportion of patients (22.2%) was assumed to be affected by severe visual impairment across the spectrum of relevant EDSS health states (i.e., 7 to 9).1 Disutilities from adverse events were excluded from the model, given that treatment-specific utilities were already incorporated in the model. Caregiver disutilities by EDSS state were included to reflect caregiver burden linked to the number of hours of care provided by a caregiver to a patient.
Several costs were considered, including costs of satralizumab and IST, drug administration costs, adverse event management costs, resource utilization costs for relapse management, and supportive care drug costs. Drug acquisition costs for satralizumab were submitted by the sponsor, while costs of maintenance IST and other medications (e.g., supportive care) were obtained from the Ontario Drug Benefit Formulary.3 Drug administration costs accounted for the administration cost per treatment cycle, pharmacy cost, diagnostic costs, and clinician consultation costs. Adverse events management costs were calculated based on the costs associated with the management of treatment-related adverse events (grade 3 severity or greater) obtained from the Ontario Schedule of Benefits: Physician Services and the Canadian Institute for Health Information Patient Cost Estimator,4,5 adjusted to reflect the same frequency of adverse events that occurred in studies 898 and 900. Resource utilization costs for relapse management by EDSS health state incorporated the costs of physician and health care professionals’ visits and the cost of diagnostic and laboratory tests. These were obtained from the Ontario Schedule of Benefits: Physician Services and the Canadian Institute for Health Information Patient Cost Estimator.4,5
Summary of Sponsor’s Economic Evaluation Results
The sponsor-submitted probabilistic analyses were based on 1,000 iterations. The sponsor’s deterministic results were consistent with its reported probabilistic findings. The probabilistic findings are presented subsequently.
Base-Case Results
In the sponsor’s base case, satralizumab monotherapy was associated with expected costs of $1,353,423 and 4.93 QALYs, and satralizumab plus IST was associated with expected costs of $1,602,989 and 4.76 QALYs, over a 50-year time horizon (see probabilistic results: and ). In both instances, satralizumab (monotherapy or in combination with IST) was more costly and produced more QALYs than the comparator treatments. The ICER for satralizumab monotherapy was $274,700 per QALY gained compared with no treatment, while the ICER for satralizumab plus IST was $337,109 per QALY gained compared with maintenance IST. In both economic models, satralizumab monotherapy or satralizumab plus IST had a 0% probability of being considered cost-effective at a WTP threshold of $50,000 per QALY ( and ).
Summary of the Sponsor’s Economic Evaluation Results for Satralizumab as Monotherapy Versus No Treatment.
Summary of the Sponsor’s Economic Evaluation Results for Satralizumab Plus IST Versus IST Alone.
From the sponsor’s submitted economic analysis, CADTH calculated that only 1% of the incremental clinical benefit (i.e., QALY gains) occurred during the trials’ observed period (i.e., at approximately 92 weeks), while the majority of the clinical benefit (approximately 99%) was observed beyond the trial period.
Sensitivity and Scenario Analysis Results
The sponsor conducted several sensitivity and scenario analyses for each of the study populations. These included applying alternate utility values based on an MS population, revising the natural history data based on a more restrictive cohort of patients (i.e., patients who are AQP4+ and naive to disease-modifying therapies), and allowing patients to improve in EDSS states following a relapse.
The sponsor also conducted a scenario analysis comparing satralizumab plus IST with eculizumab plus IST, specifically in adults with NMOSD who are AQP4+. This analysis was informed by a sponsor-commissioned indirect treatment comparison (ITC) that pooled the AQP4+ population from both studies to inform the satralizumab arm, while AQP4+ patients from the PREVENT trial informed the eculizumab arm. For this comparison, the clinical efficacy inputs from studies 898 and 900 were based on time to protocol-defined clinical relapse, and on-trial, investigator-determined relapse from the PREVENT trial. The comparative clinical efficacy for satralizumab versus eculizumab was based on a hazard ratio derived from the ITC. The incidence of adverse events for eculizumab was also based on published literature,6 while the incidence of adverse events for satralizumab was based on the AQP4+ population pooled from studies 898 and 900. Model inputs for treatment discontinuation and health state utility values for eculizumab were assumed to be the same as those utilized for satralizumab. The drug acquisition cost for eculizumab was derived from the Ontario Drug Benefit Formulary, while costs of IST were calculated using the proportion of patients who were on immunotherapies at baseline in the PREVENT trial for eculizumab and SAkura7 for satralizumab. This scenario analysis had the greatest impact, as the interpretation suggests satralizumab was less costly but also less effective. The ICER for eculizumab was $3,284,306 per QALY gained compared with satralizumab.
CADTH Appraisal of the Sponsor’s Economic Evaluation
CADTH identified several key limitations to the sponsor’s analysis that have notable implications on the economic analysis:
Cost-effectiveness of satralizumab versus IST is unknown. Satralizumab is indicated as monotherapy, or in combination with immunosuppressive therapy, for the treatment of NMOSD in adult and adolescent patients who are AQP4+.
8 To address its Health Canada indication, the sponsor submitted 2 cost-utility analyses to assess the cost-effectiveness of satralizumab as monotherapy and of satralizumab in combination with IST. The comparator in these analyses reflects the respective studies: placebo for Study 900 (versus satralizumab), and placebo plus IST in Study 898 (versus satralizumab plus IST). In both studies, placebo comprised commonly used pain medications or rescue therapy; however, in Study 900, patients did not receive the background relapse-prevention medication that would be reflective of the current standards of care in the Canadian context. Clinical trials that compare satralizumab with IST are lacking. The clinical expert consulted by CADTH noted that the submitted models do not fully reflect the anticipated place in therapy, as satralizumab monotherapy is most likely to be used in Canadian clinical practice as first-line therapy among patients who are newly diagnosed, or it could be used as second-line therapy in place of IST in patients experiencing incomplete disease treatment responses or adverse events due to IST. In these scenarios, a comparison of satralizumab versus IST monotherapy would be most appropriate, although no direct or indirect evidence exists comparing satralizumab with IST alone. Of note, according to the expert panel, IST are used off-label and, in Canadian clinical practice, the most common IST include azathioprine, mycophenolate mofetil, and rituximab.
CADTH was unable to address this limitation. In the absence of comparative clinical evidence for satralizumab versus IST, the cost-effectiveness of satralizumab versus IST remains unknown. The annual costs for individual IST were calculated by CADTH and can be found in
Appendix 1.
Relapse definition not representative of current clinical management. Progression in the model was based on relapse, with the sponsor selecting a relapse definition of “adjudicated protocol-defined relapse,” the trials’ primary outcome measure. As noted in CADTH’s Clinical Review, this approach for measuring a relapse is likely a more robust evaluation of a clinical end point because it reduces inter-site variability in assessments and over-reporting bias, thereby increasing the internal validity of this end point compared with relapses identified by the investigator or attending physicians (i.e., clinical relapse). However, this relapse definition does not reflect current clinical management practices. The definition of clinical relapse is preferred within the economic model because it is more representative of clinical practice, where relapses are determined by the physician. The clinical expert noted that any early signs and symptoms indicating a possible relapse event would require immediate action, as failing to treat an acute relapse could alter patient outcomes and may result in more severe outcomes. In practice, strict criteria for a change in EDSS or Functional System Scores would not be used to identify relapses. Rather, physicians must decide how to manage patients based on their clinical judgment in identifying relapses. Minor neurological changes attributed to a relapse would be treated early and aggressively to prevent disability accrual, with such clinical actions having resource implications. As noted within the CADTH Clinical Report, the sensitivity analyses conducted on different definitions of relapse included clinical relapse. Although the point estimates favoured satralizumab versus placebo, the confidence intervals included the null. However, uncertainty remains in interpreting these results because, although clinical relapse may more closely reflect clinical practice, the weekly monitoring of patients for relapses, and the option to move to the open-label satralizumab extension period for patients with a treated clinical relapse in Study 898, may have influenced the reporting of these events.
Conceptualization of model based on EDSS scores and uncertainty with the natural history data. The sponsor modelled the disease course for NMOSD by means of transitions between EDSS health states (scores 0 to 9) to reflect a patient’s change in disease severity following a relapse event. Higher EDSS scores (7 to 9) are indicative of a higher level of disability, while lower EDSS scores (0 to 6) are indicative of a lower level of disability. The clinical expert consulted by CADTH indicated that the conceptualization of the model structure may be limited by the use of EDSS scores as health states because the EDSS score may not fully capture the disability experienced by patients with NMOSD. In the context of a relapse event in NMOSD, a patient’s disability may involve different functional systems that may not be reflected in the EDSS score. For instance, the EDSS is less sensitive to the impacts of NMOSD on certain outcomes such as pain and disabilities in other functional systems. Additionally, the reliability and validity of the EDSS in NMOSD is unknown, as noted in the CADTH Clinical Review.
In addition, a structural assumption within the sponsor’s model was that patients could not improve EDSS scores following a relapse. This was implemented by setting any improvements in the EDSS stages observed in the natural history transition matrices to zero and reassigning the proportion of patients who had reported improvements to the previous EDSS state. CADTH’s clinical expert noted that it would be possible for a small proportion of patients to recover from relapse if immediate action with treatment was provided. By removing the improvement that was observed in the natural history data, the sponsor’s methodological approach resulted in a modelling of the disease that shows it to be more disabling than in reality. Although CADTH conducted a scenario analysis that permitted improvements in EDSS following recovery of a relapse, it was noted that the expected life-years predicted in this scenario analysis would have limited face validity, according to the clinical expert consulted (e.g., expected life-years, no treatment: 26.29 life-years; IST: 26.13 life-years) (
Table 19 and
Table 20). In fact, the clinical expert consulted by CADTH noted that, even in the sponsor’s base case, the life-years predicted are optimistic and unreasonable (e.g., expected life-years, no treatment: 21.81 life-years; IST: 22.93 life-years) (
Table 14 and
Table 15). Life expectancy is expected to be shorter, given that more relapses and certain types of relapse have a high mortality rate and would reduce life expectancy. This overestimation of the number of life-years predicted is likely due to the underlying natural history data used to inform the economic model. There were concerns with matching based on the NMOBase database. For instance, the no-treatment comparator was based on matching individuals in the NMOBase without limits to prior exposure to relapse-prevention therapy. It may have been more appropriate to limit the no-treatment comparator to individuals with no previous exposure to relapse therapy to more accurately reflect the first-line indication, as the disease course is expected to differ between treated and untreated patients. The sponsors justified this approach, as further exclusion to matching would have resulted in small patient numbers that would produce clinically implausible transitions. Given the rarity of the condition, it is unclear if better data would be available to inform the natural history of this condition.
CADTH was unable to address issues related to limits of the use of EDSS to conceptualize the model structure. Given the lack of natural history data, CADTH was further unable to address the concerns regarding the overestimation of expected life-years. It is difficult to identify or quantify the direction of bias introduced by these limitations. In a scenario analysis, CADTH explored the assumption that NMOSD patients could only accrue disability from relapse, and never improve.
Uncertainty associated with health state utility values. Health-related quality of life data were collected from NMOSD patients in studies 898 and 900 using the EuroQol 5-Dimensions 3-Levels (EQ-5D-3L) questionnaire and the UK value set. The sponsor derived health state utility values for a pooled AQP4+ population for EDSS health states 1 to 6 using a multivariate linear mixed model for repeated measurements. For the remaining EDSS health states (i.e., 0 and 7 to 9), utility values for these states were obtained from published literature, based on a relapsing-remitting MS (RRMS) patient population.
9 The clinical expert consulted by CADTH indicated that RRMS and NMOSD are not comparable diseases, as the quality of life of patients with NMOSD who are AQP4+ categorized in the same EDSS health state (or score) have a poorer quality of life than patients with RRMS categorized in the same health state.
Additionally, the sponsor incorporated caregiver disutilities to estimate caregiver impacts in the base case, which is not appropriate under a public payer perspective but may be suitable under a societal perspective. While a caregiver disutility was applied to all patients regardless of treatment, the inclusion of caregiver disutilities biased the results in favour of satralizumab, as caregiver disutilities were higher for patients in higher EDSS states. Patients on satralizumab were modelled to remain in lower EDSS states longer, with a lower risk of relapse and with a slower rate of disability accrual.
Differences in adverse event frequency between the satralizumab plus IST arm and IST alone arm. In the analysis for satralizumab as combination therapy with IST compared with IST alone, the frequency of adverse events was informed by Study 898 in which the unadjusted values suggested a higher frequency of adverse events in the IST group compared with the satralizumab plus IST group. This difference between groups resulted in substantial differences in the costs associated with adverse events. According to the CADTH Clinical Review, the rate of adverse events was comparable between patients who were assigned to IST and those who were assigned to satralizumab plus IST, after adjusting for follow-up time. Any numerical differences should be interpreted with caution, given the small sample size and differences in follow-up time between groups. The clinical expert consulted by CADTH noted that IST have a number of known toxicities, and the frequency of adverse events would not be expected to be lower for patients receiving satralizumab plus IST compared with IST alone. The expert noted that any observed differences within the clinical trial are likely due to chance and the small sample size of the AQP4+ subgroup
To address this limitation, CADTH assumed that the frequency of adverse events between groups in the combination-therapy analysis would be identical. Given that the trials are not powered to assess differences in safety, a scenario analysis was conducted by CADTH that used the unadjusted frequency rates of adverse events reported in Study 898.
Cost-effectiveness of satralizumab versus eculizumab is unknown. The sponsor submitted a scenario analysis to examine the cost-effectiveness of satralizumab compared with eculizumab using a sponsor-submitted ITC to inform comparative efficacy. Eculizumab received a CADTH Canadian Drug Expert Committee (CDEC) recommendation to be reimbursed with conditions and criteria.
10 Although this comparison may be relevant in some circumstances, given the broad Health Canada indication for satralizumab, the expert panel consulted by CADTH noted that eculizumab is most likely to be used as a last line of therapy in adult patients with NMOSD who are AQP4+ due to several factors (efficacy, safety, price, and mode of administration) presenting barriers to its use. However, as noted in the CADTH Clinical Review, the results of the sponsor’s ITC are highly uncertain and no conclusions could be drawn on the comparative efficacy and safety of satralizumab versus eculizumab in patients who are AQP4+ due to several methodological limitations, including the sparse network and the clinical heterogeneity between trials.
Additionally, the following key assumptions were made by the sponsor and have been appraised by CADTH ().
Key Assumptions of the Submitted Economic Evaluation.
CADTH Reanalyses of the Economic Evaluation
Base-Case Results
CADTH undertook the reanalyses outlined in to address, where possible, the limitations within the sponsor’s submitted economic model. CADTH was unable to address the limitations related to the comparative efficacy of satralizumab versus IST, the uncertainty associated with health state utility values, and the inherent limitations of the EDSS that was used to conceptualize the model structures in both indications. For the economic analysis where satralizumab plus IST was compared with IST alone, CADTH further addressed the limitation in adverse events frequency by assuming no between-group differences.
CADTH Revisions to the Submitted Economic Evaluation.
The results for the stepwise analyses can be found in . Results from the probabilistic analysis of the CADTH base case found that satralizumab was associated with an incremental benefit of 1.14 QALYs and incremental costs of $1,287,976 compared with no treatment. The ICER for satralizumab versus no treatment was $337,535 per QALY gained. Only 2% of the clinical benefit gains were estimated to occur during the trial’s observed period, while the majority of the clinical benefit gains (approximately 98%) were observed in the extrapolated period outside of the trial period for Study 900.
For satralizumab IST, the probabilistic results of the CADTH base case showed that satralizumab plus IST was associated with an incremental benefit of 0.76 QALYs and incremental costs of $1,769,537compared with IST alone. Similarly, the majority of the clinical benefit gains estimated (approximately 98%) occurred beyond the trial period of Study 898. The ICER for satralizumab plus IST compared with IST was $752,179 per QALY.
In both economic models, CADTH’s base-case reanalyses showed there was a 0% probability that satralizumab would be the most cost-effective strategy at a WTP threshold of $50,000 per QALY.
Summary of the Stepped Analysis of the CADTH Reanalysis Results for Satralizumab as Monotherapy Versus No Treatment.
Summary of the Stepped Analysis of the CADTH Reanalysis Results — Satralizumab Plus IST Versus IST Alone.
Scenario Analysis Results
CADTH undertook several scenario analyses on the CADTH base case to determine the impact of alternative assumptions on the cost-effectiveness of satralizumab versus no treatment, and satralizumab plus IST compared with IST alone. This included:
Adopting the societal perspective to incorporate the financial burden borne by patients (via which costs were included) and impact to caregivers.
Revising the structural assumption that relapse recovery for NMOSD patients following a relapse event is permissible.
Applying differences in unadjusted adverse event frequency, as per the sponsor’s base-case analysis (for satralizumab in combination with IST versus IST only).
Estimating the results over the entire population by weighting each analysis by its expected prevalence of use as monotherapy or combination therapy. According to the clinical expert consulted by CADTH, it was expected that the use of satralizumab would be 90% monotherapy and 10% combination therapy with IST.
The results of these analyses are presented in Table 15 to 20 (Appendix 4). The results were most sensitive to the alternate assumption that patients would not improve from relapse in the economic model where satralizumab was indicated as monotherapy. For the model that indicated satralizumab as combination therapy, results were most sensitive to the alternate assumption that patients would not improve from relapse.
CADTH undertook a series of price-reduction analyses on the price of satralizumab based on the sponsor’s submitted base case and CADTH’s base-case reanalyses. When offered as monotherapy or as a combination therapy, neither satralizumab compared with no treatment, nor satralizumab plus IST compared with IST plus no treatment, is cost-effective within the $50,000 per QALY threshold. Price reductions of 80% and 89% are scenarios in which satralizumab is cost-effective for each of the respective indications.
CADTH Price-Reduction Analyses — Satralizumab as Monotherapy.
CADTH Price-Reduction Analyses — Satralizumab in Combination With IST.
Issues for Consideration
Rituximab usage: The approach to treatment for NMOSD differs across Canada in the absence of formal treatment guidelines specifying which interventions should be used as first- or second-line therapies. The expert panel consulted by CADTH highlighted that rituximab, if available, is usually used as a second-line therapy for patients with NMOSD in Canada; otherwise, other off-label drugs (e.g., azathioprine and mycophenolate mofetil) may be used if rituximab is not available. Given the use of rituximab in NMOSD patients and satralizumab’s potential place in therapy, rituximab could be a relevant comparator even though it was not included in the sponsor’s economic model. The cost-effectiveness of satralizumab compared with rituximab is unknown in the absence of both direct treatment comparisons and ITCs. See
Appendix 1 for the costs associated with rituximab therapy for NMOSD.
Inebilizumab: The expert panel consulted by CADTH noted that inebilizumab, a CD19 monoclonal antibody (a B–cell specific antigen), has a mechanism of action similar to rituximab and has been approved in the US for adult patients with NMOSD who are anti-AQP4. It is unclear if and when this drug would be submitted to Health Canada for regulatory approval.
Drug administration differences: Satralizumab appears to be an attractive option because it fills a therapeutic gap for NMOSD patients who are AQP4+ and because it is a self-administered subcutaneous injection.
Off-label use: CADTH’s expert panel noted there is potential for satralizumab to be used off-label among patients who are AQP4 seronegative and patients under the age of 12 years, for whom existing off-label therapies are not well studied or have undesirable adverse side effects. The CADTH economic model is applicable only to the population indicated and it remains unclear how the cost-effectiveness results may change for a population of patients who are AQP4 seronegative or who are younger.
Overall Conclusions
In patients with AQP4+ NMOSD, fewer patients treated with satralizumab experienced an adjudicated relapse, relative to placebo, when it was administered as monotherapy or in combination with immunosuppressants. The between-group differences were considered clinically meaningful based on clinical expert input.
CADTH undertook reanalyses of the sponsor’s economic models for satralizumab administered as monotherapy and in combination with IST to address some of the identified limitations. In both models, CADTH’s base-case reanalysis included a relapse definition that is more reflective of clinical practice and removed caregiver impacts to align with a public payer perspective. In addition, in the economic model for satralizumab plus IST compared with IST alone, CADTH assumed no differences in the frequency of adverse events between groups. CADTH’s findings remained aligned with the sponsor, such that satralizumab is not cost-effective at a $50,000 per QALY WTP threshold for satralizumab as monotherapy and satralizumab as an add-on therapy. The ICER for satralizumab versus no treatment when administered as monotherapy was $337,535 per QALY gained, and the ICER for satralizumab plus IST versus IST alone was $752,179 per QALY gained. Price-reduction analyses suggest that for satralizumab to achieve an ICER below $50,000 per QALY gained, reductions in the price, by 80% when administered as monotherapy and by 89% when administered in combination with IST, would be required.
Relapse was incorporated into the model as the main treatment-effectiveness measure to define progression and movement within the sponsor’s economic model. As such, the model results were driven primarily by the definition of relapse and uncertainty remains as to the validity of both the sponsor’s and CADTH’s selected outcome for relapse; the sponsor’s definition within its analysis may be less generalizable, while the CADTH definition may have issues with internal validity, given protocol-driven monitoring. The incremental benefit of satralizumab in both economic models was minimal over the trial’s observed period, while the majority (approximately 98%) of the incremental benefits were obtained over the remainder of the extrapolated time horizon under the assumption of persistent treatment effects over time. CADTH was unable to address the inherent limitations of the EDSS scale that was used to conceptualize the economic model. Furthermore, limited face validity was noted with the modelled results that could not be adequately addressed, given the limitations with the natural history data that were incorporated into the economic model. Life-years were noted by the clinical expert to be overestimated in both the sponsor’s and CADTH’s reanalyses and the potential direction of bias with this limitation remains uncertain.
Importantly, both cost-effectiveness analyses do not address the most relevant comparison for satralizumab’s expected place in therapy due to the lack of comparative clinical evidence for satralizumab versus IST; therefore, the cost-effectiveness of satralizumab compared with IST (e.g., rituximab) is unknown. The cost-effectiveness of satralizumab compared with eculizumab also remains unknown due to several methodological limitations identified with the sponsor’s submitted ITC, including the sparse network and the clinical heterogeneity between trials.
Overall, the results of the economic analyses indicate that satralizumab is not cost-effective as a monotherapy nor in combination with IST as per the Health Canada indication, and the cost-effectiveness of satralizumab compared with other relevant comparators such as IST remains unknown at this time, given the lack of evidence on its comparative effectiveness.
