Estimating the clinical effectiveness and value-based price range of erenumab for the prevention of migraine in patients with prior treatment failures: a US societal perspective

Abstract

Background: Frequent migraine with four or more headache days per month is a common, disabling neurovascular disease. From a US societal perspective, this analysis models the clinical efficacy and estimates the value-based price (VBP) for erenumab, a fully human monoclonal antibody that inhibits the calcitonin gene-related peptide receptor.

Methods: A Markov health state transition model was developed to estimate the incremental costs, quality-adjusted life-years (QALYs), and value-based price range for erenumab in migraine prevention. The model comprises “on preventive treatment”, “off preventive treatment”, and “death” health states across a 10-year time horizon. The evaluation compared erenumab to no preventive treatment in episodic and chronic migraine patients that have failed at least one preventive therapy. Therapeutic benefits are based on estimated changes in monthly migraine days (MMD) from erenumab pivotal clinical trials and a network meta-analysis of migraine studies. Utilities were estimated using previously published mapping algorithms. A VBP analysis was performed to identify maximum erenumab annual prices at willingness-to-pay (WTP) thresholds of $100,000–$200,000 per QALY. Estimates of VBP under different scenarios such as choice of different comparators, assumptions around inclusion of placebo effect, and exclusion of work productivity losses were also generated.

Results: Erenumab resulted in incremental QALYs of 0.185 vs supportive care (SC) and estimated cost offsets due to reduced MMD of $8,482 over 10 years, with an average duration of treatment of 2.01 years. The estimated VBP at WTP thresholds of $100,000–$200,000 for erenumab compared to SC ranged from $14,238–$23,998. VBP estimates including the placebo effect and excluding work productivity ranged from $7,445–$13,809; increasing to $12,151–$18,589 with onabotulinumtoxinA as a comparator in chronic migraine.

Conclusion: Erenumab is predicted to reduce migraine-related direct and indirect costs, and increase QALYs compared to SC.

Keywords:

• Value based-price
• episodic migraine
• chronic migraine
• economic evaluation
• productivity
• indirect costs
• CGRP
• erenumab
• cost-effectiveness analysis

JEL classification codes:

• I10, I19

Introduction

Frequent migraine is a highly disabling neurovascular disease characterized by severe, typically unilateral headache, commonly accompanied by nausea, photophobia, phonophobia, and aura¹. Migraine prevalence is 3-times higher in women than in men^2–7 and is most common between the prime productive working ages of 18 and 59, with the peak prevalence of migraines occurring at ∼40 years of age^8–10.

Migraine can be broadly classified as episodic (EM) or chronic migraine (CM) based on the number of migraine days and headache days per 28 days (defined as monthly migraine days (MMD); monthly headache days (MHD)). EM is characterized by <15 MHD and accounts for more than 90% of migraine in the US population. In contrast, CM is defined by ≥15 MHD, including at least 8 days with migraine, and accounts for ∼5–8% of migraine¹¹. Previous studies have indicated that ∼90% of migraine patients are functionally impaired during an attack, 53% are severely impaired and require bedrest, and subjects have reported being only about half as productive while working with migraine^9,12.

Preventive therapies are recommended by US guidelines for people who experience four or more MMD who are overusing acute medication, or who have headache-related disability¹³. The mainstay of migraine prevention has been re-purposed anti-epileptic drugs (topiramate and divalproex), antidepressants (amitriptyline), and beta-blockers (propranolol), but only 13% of eligible patients reported current use of preventive therapy in published survey data⁸. In addition to not being specifically designed to alter the underlying physiology of migraine, existing treatments are associated with significant side-effects, and it is estimated that more than 80% of treated patients discontinue their preventive medication within 12 months of initiation¹⁴. OnabotulinumtoxinA was approved by the US Food and Drug Administration (FDA) in 2010 for preventative use, but is restricted to use in CM patients only. There is no recommended standard of care or published data in patients who try current prevention and fail, either because of tolerability, lack of effectiveness, or both. There is, therefore, an unmet need in these migraine patients. This analysis deploys a US societal perspective, since migraine is atypical in that indirect costs (absenteeism/disability) and presenteeism (being less productive while at work) account for up to ∼70% of total costs¹⁵. Each employee with frequent migraine costs employers thousands of dollars every year, with estimates between $2,400 and $7,000 for women and $4,000 and $13,000 for men^16,17. Developing novel treatments for migraine prevention with better efficacy or tolerability profiles is a priority for improving migraine outcomes. One promising approach targets the calcitonin gene-related peptide (CGRP, a sensory neuropeptide implicated in migraine pathogenesis) pathway. Erenumab is the only fully human monoclonal antibody in development targeting the CGRP pathway, and the only fully human monoclonal antibody in development that targets the CGRP receptor. Pivotal studies in EM and CM have been completed, and the data package is under review by regulatory agencies at the time of this writing (February 2018). The efficacy of erenumab 140 mg was demonstrated vs placebo in pivotal studies in EM and CM^18,19. The primary efficacy end-point in both pivotal studies was the change from baseline to the end of the double-blind treatment period in the mean number of MMD. In the EM study, the mean number of MMD at baseline was 8.3 in the overall population; by months 4–6, the number of days was reduced by 3.7 in the 140 mg erenumab group, as compared with 1.8 days in the placebo group (p < .001). Linear mixed-effects regression models predicted a least squares mean change from baseline vs placebo for the erenumab 140 mg group of –1.85 MMD (95% CI = –2.33, –1.37; p < .001) over the final 12 weeks of the double-blind period¹⁸. In the CM study, erenumab 140 mg reduced MMD vs placebo (–6.6 days vs placebo –4.2 days). Least squares mean change from baseline for erenumab 140 mg vs placebo at week 12 was –2.45 MMD (95% CI = –3.51, –1.38; p < .001)¹⁹. In addition, the proportion of patients in the erenumab 140 mg group with at least a 50% reduction in MMD from baseline to the end of double-blind period ranged from 50% in EM (26% for placebo; odds ratio =2.8 (2.0–3.9)) to 41% in CM (23% for placebo; odds ratio =2.3 (1.6–3.5)). In pre-specified sub-group analysis in the clinical studies, erenumab demonstrated a numerically greater reduction in MMD compared to placebo in patients who had previously failed ≥1 prior preventive treatment, than was observed in the overall trial populations. Erenumab has, therefore, demonstrated efficacy in patients who have tried and failed preventive therapies, a population of patients with greater unmet medical need²⁰.

The value of novel health technologies is typically assessed via cost-effectiveness modeling, comparing the ratio of incremental health outcomes to incremental costs, known as the incremental cost-effectiveness ratio (ICER). Erenumab is not approved for use, and pricing is not known at the time of this writing (February 2018), so a direct analysis of its cost-effectiveness is not possible. However, it is useful to consider what level of price is justifiable given the additional benefits of erenumab over current options and the potential to displace sub-optimal therapies. To do this, one can estimate the value based-price (VBP) based on incremental costs and quality adjusted life years (QALY)²¹. The VBP is the maximum price at which the drug would still be considered cost-effective vs a comparator, when using a defined willingness-to-pay (WTP) threshold for additional benefits. In the US, WTP thresholds per incremental QALY that have been commonly used to assess the cost-effectiveness of novel medical interventions are $100,000–$200,000.

The objective of this study is to estimate VBP ranges for erenumab 140 mg, administered subcutaneously every 4 weeks, in migraine patients who have failed at least one prior preventive treatment, compared to supportive care (SC), by evaluating the incremental costs and QALYs within a cost-effectiveness modeling framework.

Methods

We built a Markov model, implemented in Microsoft Excel, based on the clinical data from the EM and CM pivotal studies for the sub-groups of patients with prior treatment failures. The model comprises health states accounting for patients who are “on preventive treatment”, “off preventive treatment”, and “dead” (accruing no costs or health outcomes). In addition to the primary clinical outcome of MMD frequency, the model predicts the costs and health-related quality-of-life outcomes associated with erenumab as preventive treatment of migraine in patients with ≥1 prior failed treatment, compared to SC. EM and CM cohorts are modeled independently based on the clinical trial data, but outcomes are combined based on a split of the overall treated migraine population between EM and CM, based on available literature²². A comparison of erenumab to onabotulinumtoxinA in exclusively CM patients is presented as scenario analysis. Based on this output, ranges of the VBP of erenumab are estimated based on commonly used WTP thresholds.

The cycle length of the model is 28 days, consistent with the primary efficacy outcome (MMD) and the frequency of administration of erenumab. Cost and QALY outcomes are discounted at an annual rate of 3%, in line with published US recommendations²³. Clinical outcomes (number of migraine days, life years) are not discounted. The analysis is performed from a US societal perspective, including the direct medical costs of treating migraine and the indirect costs of missed work days and lost workplace productivity. This reflects the working age of the migraine population^18,19. The model evaluates cost outcomes in 2017 US dollars.

Time horizon

The time horizon in these analyses spans 10 years. The erenumab studies were reflective of the migraine prevalent population, with mean age at baseline for the pivotal studies ranging from 40–43 years. The prevalence of migraine after age 60 falls to ∼5%, and is less than <1% in CM²⁴. Published guidance on the design of economic evaluations also state that the time horizon of analyses should be long enough to capture all relevant differences between treatment strategies compared²³. The model assumes that the clinical and economic outcomes of erenumab patients are equal to those in the SC arm after they have discontinued treatment. This means that there are no further differences between arms once all patients have discontinued, so incremental outcomes are limited to the duration of erenumab treatment. Based on the disease epidemiology and the erenumab time on treatment predicted by the model (full details provided in Supplementary Material, section A), a 10-year time horizon is sufficiently long to capture the lifetime impact of the decision problem. As over 99% of patients discontinue erenumab by the end of the simulation, further extrapolation of the clinical trial data is not required.

Patient population

Erenumab studies enrolled subjects that were either naïve to preventive treatment or previously treated with preventive medication, but failed due to lack of efficacy or intolerability. However, it is anticipated that erenumab and other CGRP and monoclonal antibodies will be restricted for use to patients who have failed prior preventive therapies. Therefore, the migraine populations considered in the model are the sub-groups of patients who have previously failed ≥1 prior preventive therapy. In the clinical studies, a patient was considered to have failed a preventive therapy if they were recorded to have discontinued due to lack of efficacy or intolerability, at any time. In addition, chronic patients are more likely to seek treatment and, therefore, in the base case analysis, the migraine population is modeled as 33% EM and 67% CM²². A scenario analysis is presented in which the migraine types are evenly split (50% EM, 50% CM).

Intervention and comparators

The intervention evaluated in the model is erenumab 140 mg, self-administered every 28 days by subcutaneous injection.

In patients for whom currently available preventive treatments can be efficacious and tolerable, use of these treatments represents maximum value to the patient and the healthcare system. However, there is currently no defined standard of care for patients with four or more MMD who have tried and failed either topiramate or propranolol, due to the lack of published evidence from clinical trials or observational studies. Sequencing these treatments with either one or other generics is also not supported by evidence-based guidelines. Clinicians resort to sequencing simply due to the lack of other pharmacologic options. Therefore, neither topiramate nor propranolol are appropriate comparators in patients with four or more MMD who have failed prior preventive treatment. This gap in the data may be addressed by erenumab. Multiple clinical and insurer sources suggest that, in clinical practice, erenumab will be used after failure of topiramate or propranolol or a similar beta blocker or anti-hypertensive, addressing the high unmet need of migraine patients who have experienced a lack of efficacy or tolerability from prior preventives.

Although these previously failed patients are likely to have failed multiple preventives, the clinical trial sub-groups of patients who had failed at least one prior preventive were used as a proxy in this analysis. This assumption retains the sample size available in these sub-groups, but is also supported by published analyses which have shown that the number of prior failed therapies does not substantially affect the absolute MMD reductions of erenumab^20,25.

In clinical practice, most of these patients are typically managed with acute treatments only. As such, the comparator against which erenumab is assessed in patients who have previously received preventive therapy is SC, in which patients receive only acute treatment for migraine. OnabotulinumtoxinA is the only migraine preventive exclusively indicated for CM patients, and is commonly used after the failure of prior preventive treatments. To reflect this, a scenario analysis is presented in which erenumab is compared to onabotulinumtoxinA in an entirely CM population¹³.

Clinical trials in migraine prevention have typically observed strong placebo effects²⁶, but the administration of placebos, such as sham injections, does not represent a plausible treatment option in clinical practice. Therefore, we do not consider placebo a relevant comparator in the model. There is an absence of reliable real-world data on the natural history of migraine. In our modeling, we examine two scenarios. In the base case, the placebo effect attributable to enrollment into the clinical studies and the administration of sham injections are excluded. It is assumed that patients in the SC cohort of the model remain at the MMD observed during the 4-week pre-randomization period in the clinical studies, prior to the start of the double-blind phase. This assumption is tested in a scenario where the placebo effect is included.

Model structure

The model is comprised of two primary health states: “on preventive therapy” and “off preventive therapy” (Figure 1). Patients are at risk of death in each cycle, based on US general population mortality rates²⁷. The risks of death are assumed to be unaffected by MMD or treatment, and life expectancy is identical in both arms of the model.

Figure 1. Model schematic. MD, migraine days. Patients can transition to an absorbing death state due to all-cause mortality at any point. (A) Time- and treatment-dependent discontinuation rates determine time on preventive therapy, during which patients experience the MMD reduction attributed to treatment. (B) The cohort of patients achieves the reduction in mean MMD from baseline, based on clinical trial end-points. (C) Parametric distributions represent the variation of patients around the mean MMD, and allow outcomes linked to the number of migraine days to be estimated. Hypothetical time points (1) and (2) indicate how the distribution of patients is estimated, based on the mean MMD of the cohort at different time points.

In each cycle, patients on treatment are at risk of discontinuation (Figure 1(A)), after which they withdraw from treatment and lose the associated treatment effect. In the absence of real world discontinuation data for erenumab, baseline persistence rates were taken from US claims data, using onabotulinumtoxinA as the closest analog to a novel preventive. An exponential function was fitted to the proportion of patients remaining on onabotulinumtoxinA treatment over a follow-up period of 52 weeks²⁸. A discontinuation rate ratio of erenumab compared to onabotulinumtoxinA was derived from a network meta-analysis (NMA) of all-cause discontinuation data reported in nine clinical studies of preventives in CM (Supplementary Material, section A). The predicted time on treatment curve for erenumab was used to drive transitions between the “on preventive treatment” and “off preventive treatment” health states in each cycle. The approach is described in greater detail in the Supplementary Material. Discontinued patients are assumed to remain untreated for the remainder of the simulation. Transitions between all three model health states were half-cycle corrected.

In each 28-day cycle, the mean MMD is modeled for patients in the living health states (only “on treatment” shown in Figure 1(B)). Patients are distributed based on the mean MMD, across the range of possible MMD counts (between 0–28 migraine days in each cycle), using previously validated parametric models (Figure 1(C))^29,30. As shown in hypothetical time points (1) and (2), the shape of the distribution of individual patients by MMD changes to account for both the mean MMD and the asymmetric spread of individual patients.

The parametric models used in the calculation steps in Figure 1(B,C) is described in greater detail in the Supplementary Material.

Costs

Drug and administration costs

Preventive therapy and acute migraine medication costs are accounted for in the model (Table 1). Erenumab is currently undergoing regulatory review by the FDA and, as such, is not yet available for purchase. In the absence of a list price, VBP ranges are evaluated based on the model. For the scenario analysis, onabotulinumtoxinA is estimated to cost $5,035 in drug acquisition costs, and $649 in administration costs per year (CMS Physician Fee Schedule CPT 99212).

Table 1. Preventive therapy costs, migraine resource use costs, and acute medication costs.

Medical resource	Unit cost (2017 USD)	Average use per year*	Use per migraine day†
Physician visits (CPT99212)	$44.14^32	0.720	0.0379
Emergency room visits	$939.59^28	0.167	0.0088
Hospitalization (DRG 102 and 103)	$4,298.35^33	0.075	0.0039
Specialist consultations (CPT 99215)	$146.43^32	0.221	0.0116
Acute medication‡		Cost per day of use - EM (2017 USD)	Cost per day of use - CM (2017 USD)
Non-migraine-specific		0.99	1.76
Migraine-specific		4.94	3.99
Preventive therapy	Cost per year (2017 USD)	Frequency of occurrence or dosing	Annual cost
OnabotulinumtoxinA	$1,158.00^34	12-weekly	$5,035

* Annual use reported in Munakata et al.³¹, migraine patient cohort.

† Patients reported an average of 19 headache days over the previous 12 months.

‡ Estimation of costs per day of use based on published breakdown of medication types by frequency of use and 2017 unit costs. Migraine-specific medication comprised of triptans and ergot derivatives. Non-migraine-specific medication comprised of acetaminophen, non-steroidal anti-inflammatory drugs [NSAIDs], barbiturates, opioids, isometheptene compounds, and other over-the-counter medication.

Medical resource use costs

Medical resource use in the model consists of physician office visits (primary care doctor), emergency room visits, hospitalizations, and specialist neurologist consultations based on published unit costs (Table 1). Average annual medical resource use is taken from a published 2009 analysis of survey data from 7,437 migraine patients in the US³¹. The mean patient-reported medical resource use over 12 months was divided by the reported annual number of headache days to estimate the medical resource cost per migraine day in the model³¹. The resource use per migraine day and the unit costs are combined in the model to estimate the weighted average costs of medical resource use for each cohort of patients.

Acute migraine day medication costs

The distribution of the drug classes by usage and the dosages used to treat acute migraine were obtained from three studies in the literature^35–37. Using acute medication use data collected in the erenumab clinical studies, the model differentiates between migraine-specific acute medication (comprised of triptans and ergot derivatives), and non-migraine-specific acute medication (comprised of acetaminophen, non-steroidal anti-inflammatory drugs [NSAIDs], barbiturates, opioids, isometheptene compounds, and other over-the-counter medications)³⁵. Weighted average costs per day of use are shown in Table 1, and the numbers of days of acute medication use by migraine day frequency are presented in the Supplementary Material.

Indirect costs of lost work productivity

The substantial impact on a patient’s ability to function and associated lost productivity accounts for the greatest proportion of total costs attributed to migraine^31,38. The productivity cost of migraine is split into two types. Absenteeism days are days on which patients are unable to attend work or school due to their migraine. Presenteeism days are days on which patient productivity at work or school is reduced by at least 50% (but less than 100%). The number of days of productivity losses in the model are based on erenumab clinical trial data, and reflect the sex, age, and employment status of the clinical trial populations. The average costs of absenteeism and presenteeism days are calculated assuming the median hourly gross wage obtained from the US Bureau of Labor Statistics³⁹, assuming an 8-h working day. As the degree of productivity loss on each presenteeism day (i.e. days where productivity is reduced by at least 50%) is not known⁴⁰, the model assumes lost productivity of 50%. The costs per absenteeism and presenteeism day used in the model are presented in Table 2, and a scenario excluding productivity costs is presented in Supplementary Materials.

Table 2. Estimated indirect costs per absenteeism and presenteeism day.

Parameter	Value	Source
Median hourly wage	$26.00	Bureau of Labor Statistics, Private sector December 2016
Number of working hours per day	8	Assumption
Proportion of productivity loss on presenteeism days	50%	Assumption
Estimated cost per absenteeism day	$208.00	Calculated
Estimated cost per presenteeism day	$104.00	Calculated

The number of absenteeism and presenteeism days are estimated based on patient responses to the Migraine Disability Assessment questionnaire collected in the erenumab EM and CM pivotal studies^19,41. Question 1 of the Migraine Disability Assessment questionnaire refers to absenteeism, and question 2 refers to presenteeism⁴⁰. Patient responses from both the EM and CM studies were combined to generate one complete migraine data-set, in which the relationship between MMD and productivity was analyzed. Zero-inflated Poisson regression models were fitted and used to predict the average number of absenteeism and presenteeism days for each possible migraine day frequency (0–28 MMD). As an example, a person experiencing 15 migraine days in a 28-day period is estimated to have 3.94 presenteeism days and 1.40 days absence, at a total lost productivity cost of $702. The predicted values by migraine day frequency used to estimate absenteeism and presenteeism costs in the model are presented in the Supplementary Material, section B.

Health-related quality-of-life

Utility values in the model were estimated as a function of MMD. Patient responses to the Migraine Specific Questionnaire version 2.1, collected in the pivotal EM and CM clinical studies, were mapped to the UK tariff set of the EuroQoL 5-dimension 3-level instrument (EQ-5D-3L) using previously published algorithms for EM and CM⁴². Gillard et al.⁴² report algorithms for mapping between the Migraine Specific Questionnaire and EQ-5D-3L generated based on datasets of 5,770 and 338 participants from 10 countries in the International Burden of Migraine Study survey in EM and CM, respectively. Migraine Specific Questionnaire responses from the erenumab EM and CM pivotal studies were mapped to the EQ-5D-3L using the respective algorithm, then pooled to generate one complete migraine dataset. A longitudinal beta regression model was fitted, with mapped EQ-5D-3L as the response variable, controlling for MMD and key patient characteristics. The regressions were used to generate predicted EQ-5D-3L values for each frequency of MMD, which are used in the model to estimate the mean utility of the patient cohort, weighted by the distribution of patients by migraine day frequency in each cycle. As treatment status (erenumab 140 mg compared to placebo) was significantly predictive of utility, with higher utility values predicted for erenumab, the predicted values applied in the model are separated for actively-treated (erenumab, onabotulinumtoxinA) and untreated patients (SC, post-discontinuation). This approach is consistent to the assumptions made in the previous economic model for onabotulinum-toxinA³⁸, which also assumed an additional treatment effect on utility of active treatment compared to SC. As an example, a person with 15 migraine days in a 28-day period would have an estimated utility value of 0.589 on erenumab 140 mg and 0.571 whilst untreated. The values applied in the model are reported in the outcomes table presented in the Supplementary Material, section B.

Results

In the base case analysis, patients receiving SC were estimated to experience an average of 1,949 migraine days over 10 years (Table 3). By comparison, erenumab-treated patients were estimated to experience 1,805 migraine days, meaning a reduction of 144 migraine days. Because of discontinuation, this reduction is based on a mean duration of erenumab treatment of ∼2 years. As a result of the migraine day frequency reductions, erenumab was associated with increased total discounted QALYs per person of 0.1849 over the 10-year horizon.

Table 3. Base case model results per person by comparison and treatment arm, over 10 years.*

Comparison	Erenumab	SC	Incremental
Mean duration of treatment (years)	2.01	N/A	N/A
Mean migraine days	1,805	1,949	–144
Mean discounted QALYs	5.1437	4.9588	0.1849
Mean discounted migraine day-related costs**	$121,407	$129,889	–$8,482
Societal value based price***	$14,238–$23,998	–	–

* Migraine population in the base case model is made up of 33% EM and 67% CM patients²³.

** Cost estimates do not include the costs of providing preventive medication, as a price for erenumab is not available.

*** Maximum acceptable price at a willingness-to-pay threshold of $100,000–$200,000 per QALY.

The discounted cost associated with the burden of migraine in patients on SC was estimated to be $129,889 over 10 years. By reducing the number of migraine days, erenumab was expected to reduce the total migraine day-related cost by $8,482. This does not include the incremental acquisition costs of erenumab. Disaggregated incremental migraine day-related costs, showing the contribution of the different cost types, are presented in Table 4.

Table 4. Disaggregated incremental costs by comparison and treatment arm, over 10 years.

Cost category	Erenumab	SC	Incremental
Physician visits	$2,443	$2,631	–$188
Emergency room visits	$12,061	$12,988	–$927
Hospitalizations	$24,779	$26,684	–$1,904
Specialist consultations	$2,487	$2,679	–$191
Migraine-specific acute medication	$2,599	$2,820	–$221
Non-migraine-specific acute medication	$673	$708	–$36
Absenteeism	$31,339	$32,997	–$1,658
Presenteeism	$45,025	$48,382	–$3,357
Total	$121,407	$129,889	–$8,482

Based on the clinical effectiveness of erenumab predicted by the model, VBP ranges were estimated. These prices represent the maximum annual treatment costs at which erenumab would be considered cost-effective at WTP thresholds ranging from $100,000–$200,000 per incremental QALY. Calculation of the VBP incorporates both the cost reduction and the QALY gain associated with erenumab in the quantification of the potential monetary value of erenumab treatment. The estimated VBP of erenumab ranged from $14,238–$23,998 per year.

Deterministic sensitivity analysis

To explore the sensitivity of VBP estimates to key input parameter values, deterministic sensitivity analysis (DSA) was performed, in which upper and lower bounds of individual model parameters were tested to identify model drivers in each of the comparisons assessed. The results of this analysis were quantified as the percentage deviation from the base case VBP estimate, calculated based on a WTP threshold of $150,000 per incremental QALY. The estimate of the VBP was driven mostly by the relative reduction in migraine days of erenumab, reflecting uncertainty in the NMA outcomes parameterizing this. There was a smaller influence of migraine day-related outcomes, primarily utility estimates, productivity costs, and hospitalization frequency. The maximum variation in the VBP was within ±50% of the base case estimate (Figure 2).

Figure 2. DSA results. *Relative MMD reduction for erenumab based on NMA end-points, combined uncertainty for EM and CM data. **Utility and reference change in MMD are vectors of parameters based on regression models.

Scenario analyses

In addition to the base case results, four scenarios are presented to test major model assumptions.

The first includes the reduction from baseline in MMD in the placebo cohorts of the clinical studies. Patients in the SC arm are assumed to achieve this reduction, and patients who discontinue erenumab are assumed to retain the proportion of the reduction observed in the placebo groups. In this scenario, the VBP ranged from $8,886–$15,250.

The second scenario also includes the placebo reduction, but also excludes the indirect costs of lost productivity, considering only costs that would be incurred by a healthcare payer. By combining the exclusion of these costs with the placebo reduction, this is expected to be the most conservative scenario with respect to the cost-effectiveness of erenumab. In this scenario, the VBP estimates ranged from $7,445–$13,809.

The third scenario assumes that the migraine population is split evenly between EM and CM, assuming 50% EM and 50% CM. Under this assumption, the VBP estimates ranged from $13,331–$22,553.

The final scenario considers only CM patients, and compares erenumab to onabotulinumtoxinA in previously treated CM patients. Compared to onabotulinumtoxinA in exclusively CM patients, the VBP estimates ranged from $12,151–$18,589.

The ranges of VBP estimated in the base case and scenarios are presented graphically in Figure 3, along with the assumptions defining each scenario. Full results for each scenario are presented in the Supplementary Material, section C.

Figure 3. Summary of VBP estimates, assuming a 33% EM, 67% CM split. Abbreviation. WTP, willingness-to-pay.

Discussion

To achieve efficient allocation of healthcare resources under budget constraints, cost-effectiveness analysis is increasingly used by healthcare decision-makers to prioritize societal preferences for changes in health status across competing healthcare interventions²³. The MMD reductions and QALY improvements with erenumab presented here estimate the value of this novel migraine therapy compared to current practices in migraine patients who have failed prior preventive therapy. In people with frequent migraine, there are no published data supporting preventive treatment for patients that have failed at least one prior preventive therapy; therefore, this represents an important QALY gain of ∼0.184.

At the time of launching a new therapy, there is a necessity to satisfy not only safety and efficacy requirements, but increasingly the need to highlight economic value in relation to costs to satisfy paying organizations. Accomplishing this is challenging, considering the full economic value of a new intervention cannot be fully established before launch, due to the absence of real-word data. Attempts to estimate economic value of new interventions using only the regulatory data package (i.e. FDA filing) is limited by this data availability. The analysis described here highlights the challenges of demonstrating economic value for a new product when no price has been established and real-world evidence is not available. To circumvent the challenges of conducting an economic value demonstration on a pre-launch preventive migraine therapy, we have conducted an analysis which seeks to evaluate the annual cost of treatment that reflects the estimated clinical and economic value of erenumab, using acceptable value standards (i.e. WTP thresholds). From a US societal perspective, these are the maximum estimated “prices” below which erenumab would be cost-effective at a WTP of $100,000–$200,000 given the framework of a cost-effective analyses for patients who have failed at least one prior treatment and against appropriate comparators.

The modeling approach applied in this study is different to that used in previous economic evaluations in migraine prevention^38,43,44, which have adopted decision tree approaches or Markov models based on health states based on defined ranges of migraine day or headache day frequency. Modeling MMD as a continuous outcome better captures the outcomes of patients, by accounting for variability in migraine day frequency without relying on compartmentalizing patients based on response status or arbitrary categories of MMD, which have been shown to introduce bias into migraine day estimates⁴⁵. The approach allows cost and quality-of-life outcomes to be linked to individual migraine frequency, rather than average outcomes for compartmentalized health states. In this way, the model, therefore, spans the range of migraine frequency, across EM and CM, and is consistent with patient presentation in clinical practice. This also permits the same model structure to accommodate combined assessments of EM and CM and for estimating the impact of each individual migraine day event.

Scenarios presented in this paper excluding indirect costs, such as those associated with absenteeism and presenteeism, lower the VBP range compared to the base case analyses. Consistent with US guidelines on economic evaluation²³, the analysis here includes missed work days and lost productivity. In migraine, these costs represent a significant proportion of the economic burden of migraine, and are often paid by employers due to reduced productivity of people with migraines. We recognize that healthcare payers may not always consider these costs in assessing the value of novel preventives, despite their importance to patients and employers and, hence, VBP were also generated based on this scenario. Even when the monetary value of QALY gains are ignored, migraine day related costs off-sets with erenumab (ignoring erenumab drug costs) are still ∼ $8,500 over the mean treatment duration of 2.01 years. These VBP estimates represent one of several factors considered in pricing decisions, and other factors, such as affordability, also impact the final price. Cost-effectiveness models by definition do not factor in affordability and, typically, do not address other considerations important to payers, such as the size of the treated patient population and unmet need.

The results presented here should be interpreted within the context of the study limitations. This analysis is based on erenumab treatment practices defined by treatment protocols used in the pivotal randomized controlled trials in the pre-launch phase of drug development. However, in clinical practice, physicians and patients may adjust treatment practices to optimize outcomes, and, in some cases, introduce strategies for when to discontinue therapy. It is likely that when erenumab enters treatment practice, and prior to the establishment of clinical guidelines, clinicians will adjust erenumab use to meet patient treatment goals. This may include treatment discontinuation in cases of non- or partial-clinical response. The discontinuation of patients experiencing smaller reductions in MMD will likely improve estimates of the clinical effectiveness and VBP ranges presented here. In a cohort of treated subjects, as non-responders or low-responders discontinue, the average MMD reduction of the patients remaining on treatment will increase, the total number of erenumab-treated patients will reduce, and, thus, cost-effectiveness will be more favorable.

The model is also limited by the consideration of MMD as the only metric of disease status, and other dimensions of migraine, such as duration and severity, are not explicitly considered beyond their contribution to the definition of a migraine day. Any residual impact during non-migraine days such as interictal burden, prodromal symptoms, anxiety, and depression is not captured in our analysis, and should be assessed in the future⁴⁶. Improvement in the other dimensions may be indirectly captured by the application of utility values stratified by treatment (i.e. separate values for patient on erenumab/onabotulinumtoxinA vs SC), but these are not isolated as separate treatment effects. The model is also subject to limitations in available data. In particular, there is no evidence of time to discontinuation for patients treated with erenumab in clinical practice, and the comparative discontinuation rates applied in the model are derived from available clinical trial data. Furthermore, the use of cost data from Munakata et al.³¹ is likely to result in an under-estimation of medical resource use costs. First, the source data reported resource use across the US migraine population, and the resource use among patients who have failed a previous preventive therapy is likely to be greater. Second, the study reported only headache days, only a proportion of which will be migraine days, so the resource use per migraine day will also be an under-estimation.

The model is also limited by several simplifying assumptions, most notably the assumption that patients remain untreated after discontinuation. Whilst this may not be reflective of clinical practice, the lack of long-term, sequential treatment data prevents other scenarios from being explored. Finally, it is not certain that the MMD of patients treated only with acute medication would be constant over time. Whilst the inclusion of the placebo reduction is essential in assessing the treatment effect of erenumab in a clinical trial context, its relevance to economic evaluation as a potential comparator is limited. It is also possible that patients whose migraines are not controlled with preventive therapy, and instead rely only on acute medication, may experience increased MMD over time, due to pain medication over-use¹⁴.

Conclusion

The VBP ranges presented in this study represent the value of erenumab, as assessed within the scenarios described under a cost-effectiveness framework. However, cost-effectiveness is just one criterion against which value can be assessed, and affordability and other factors also impact final price. In this study, erenumab showed consistent and meaningful improvements in migraine day frequency and QALY compared to SC for patients who have failed at least one prior generic preventive therapy. The results presented provide the range of prices at which erenumab would be considered a valuable addition as migraine prevention in people with migraine, based on established WTP thresholds in the US. The value demonstration framework based on willingness to pay for health gains offers a meaningful approach to understand product value in relation to potential prices. Our analysis also highlights potential cost savings that can be achieved for people with migraine attributed to acute migraine day treatment costs, physician costs, and improved productivity output, suggesting benefits for both health services and broader societal impact. In the post-launch period, the economic results described here can be enriched to more accurately define clinical and economic value.

Transparency

Declaration of funding

The study was funded by Amgen Inc.

Declaration of financial/other relationships

RBL has received research grants from NIH, the Migraine Research Foundation, and the National Headache foundation, and has received consulting fees from the American Academy of Neurology, Alder, Allergan, American Headache Society, Amgen, Avanir, Biohaven, Biovision, Boston Scientific, Dr. Reddy’s, Electrocore, Eli Lilly, eNeura Therapeutics, GlaxoSmithKline, Merck, Pernix, Pfizer, Supernus, Teva, Trigemina, Vector, and Vedanta. RBL serves on the editorial board of Neurology, as an Associate Editor of Cephalalgia, and as a senior advisor to Headache. RBL receives royalties from Wolff’s Headache and Other Head Pain, 8th Edition (Oxford Press University, 2009), Wiley, and Informa, and holds stock in eNeura and Biohaven. DD has received consulting fees from Allergan, Amgen, Alder, Pfizer, Colucid, eNeura, Eli Lilly, Autonomic Technologies, Supernus, Tonix, Novartis, Teva, Zosano, Insys, GBS/Nocira, Acorda, Biohaven, Nocira, Ipsen, and Promius. DD receives royalties from Oxford University Press and Cambridge University Press, and holds stock in Nocira, Epien, King-Devick, Migraine Buddy, and Mobile Health (Second Opinion). ST has received research grants (no personal compensation) from Alder, Allergan, Amgen, ATI, Dr. Reddy’s, ElectroCore, eNeura, Scion Neurostim, Teva, and Zosano. ST serves as a consultant or on Advisory Boards, Scientific Advisory Boards, or Trial Steering Committees for Acorda, Alder, Allergan, Amgen, ATI, Cefaly, Charleston Laboratories, DeepBench, Dr. Reddy’s, ElectroCore, Eli Lilly, eNeura, GLG, Guidepoint Global, Impax, Neurolief, Pfizer, Scion Neurostim, Slingshot Insights, Supernus, Teva, and Zosano. ST sits on the board of the American Headache Society, receives royalties from Springer, and holds stock in ATI. ST is an employee of Dartmouth Hitchcock Medical Center and receives a salary for editorship from the American Headache Society. AB has received research grants from the National Institute of Health Research, Public Health England, the National Institute of Health (US), and the Department of Health (UK), and has received consulting fees from Amgen, GlaxoSmithKline, RTI, and TeamDRG. SP has received consulting fees from Amgen. AJH was an employee of BresMed Health Solutions, who have received consulting fees from Amgen, when the study was conducted. JKP is an employee of Amgen. SS, GV, and NS are employees of Amgen and hold stock. Peer reviewers on this manuscript have received an honorarium from JME for their review work, but have no other relevant financial relationships to disclose.

Acknowledgments

We are grateful to Gertruud Haitsma, Saswat Panda, and Dr. Mark Connolly for their technical and editorial contributions to this work.