Economic evaluations in migraine: systematic literature review and a novel approach

Abstract

Background: For novel migraine therapies, economic evaluations will be required to understand the trade-offs between additional health benefit and additional cost. The purpose of this study was to conduct a systematic literature review (SLR) to identify previous economic evaluations in migraine from the United Kingdom or Irish perspective to critically appraise these evaluations and to propose, if necessary, a novel modelling approach that can be used for future economic evaluations of migraine therapies.

Methods: An SLR was conducted to identify previous economic evaluations of preventive migraine treatments. Key opinion leaders were consulted to determine the criteria for a robust migraine economic evaluation. Economic evaluations identified in the SLR were appraised against these criteria, and a novel cost-effectiveness model structure was then proposed.

Results: Eight records reporting on published economic evaluations were identified and critically appraised for general quality. Expert consultation provided 6 recommendations on the ideal model structure for migraine that is both clinically and economically meaningful. A decision-tree plus Markov structure was then developed as a cost-effectiveness model for migraine therapies where each health state is associated with a patient distribution across monthly migraine day (MMD) frequencies.

Conclusions: Future migraine economic evaluations should allow for assessments across the full spectrum of migraine, a response-based stopping rule, and the estimation of benefits and resource costs based on MMD frequency. The approach proposed in this paper captures all of the desired elements for an economic evaluation of migraine therapy and is suitable to assess new migraine therapies.

Keywords:

• Model
• economic evaluation
• systematic literature review
• episodic migraine
• chronic migraine
• headache
• CGRP

JEL classification codes:

• I11
• I15
• C52

Introduction

Migraine is a distinct, common neurological disease characterized by recurrent, often unilateral, throbbing head pain of moderate to severe intensity accompanied by symptoms including nausea, vomiting, photophobia, or phonophobia.¹ Migraine is a source of significant burden, as it is estimated to affect >10% of the adult population across the world.²

Migraine has a variable burden driven by attack frequency and is sometimes broadly classified as episodic migraine (EM) or chronic migraine (CM) based on the number of migraine days and headache days per month.² Episodic migraine is defined as fewer than 15 headache days per month, with or without aura, and accounts for more than 90% of persons with migraine.³ Chronic migraine is defined as at least 15 headache days per month (of which ≥8 are migraine days with or without aura), and it affects approximately 5% to 8% of persons with migraine.⁴

To understand better the value of treatments for migraine, health technology assessment (HTA) agencies and other resource allocation decision makers may utilize economic evaluations. It is important to consider the relevance of existing economic evaluations and modelling approaches to ensure that the costs and benefits are appropriately captured for novel migraine therapies such as those designed to block the trigeminovascular calcitonin gene-related peptide (CGRP) pathway. The purpose of this study was to conduct a systematic literature review (SLR) to identify previous economic evaluations in migraine from the United Kingdom (UK) or Ireland perspective, to critically appraise these evaluations, and finally, to propose a novel modelling approach that can be used for future economic evaluations of new migraine therapies, including recently approved CGRP antagonists.

Methods

SLR overview

An SLR was conducted to identify economic evidence to support the development of a cost-effectiveness model for the prevention of chronic or episodic migraine. The SLR was conducted in July 2017, and subsequently updated in January 2018 and September 2018, to identify all literature published since database inception on any of the following topics:

• Economic evaluations of pharmacological interventions for the treatment of chronic or episodic migraine.

• Health state utility values for chronic or episodic migraine patients.

• Cost and resource use data for chronic or episodic migraine patients.

The SLR and updates were performed in accordance with the methodological principles of conduct for systematic reviews as detailed in the University of York Centre for Reviews and Dissemination’s “Guidance for Undertaking Reviews in Health Care”.⁵ No ethics review board was required for tertiary research. The focus of this manuscript is the economic evaluations of pharmacologic interventions for the treatment of chronic or episodic migraine that were identified in the SLR.

SLR search strategy

MEDLINE, Embase, The Cochrane Library, and EconLit electronic databases were searched for economic evaluations in migraine over 3 phases during July 2017, with subsequent updates through September 2018 (Appendix Table 1). A list of search terms used in MEDLINE, MEDLINE Daily, MEDLINE In-Process, and Epub Ahead of Print electronic databases for both the original review and the updates is provided in Appendix Table 2. Search terms used in the Embase database, Cochrane Library of databases, and EconLit search for both the original review and the updates are presented in Appendix Table 3, Appendix Table 4, and Appendix Table 5, respectively. Results from the database searches were downloaded into an Endnote database and de-duplicated before being transferred into a bespoke Microsoft Excel-based platform designed to enable record screening. In addition to the electronic database searches, the conference proceedings of the major migraine and neurological congresses held over the prior 3 years (2015–2018) were manually searched (Appendix Table 6).

The National Institute for Health and Care Excellence (NICE), Scottish Medicines Consortium (SMC), All Wales Medicines Strategy Group (AWMSG), and National Centre for Pharmacoeconomics (NCPE) websites were manually searched for previous, relevant HTA submissions. In addition, The Cost-Effectiveness Analysis (CEA) Registry, managed by Tufts Medical Center, and EconPapers at Research Papers in Economics (RePEc) websites were also manually searched to ensure that no relevant publications were missed.

Finally, the bibliographies of all relevant SLRs, meta-analyses, HTA submissions, and economic evaluations identified through the electronic database, conference, and HTA agency website searches were also manually searched to identify any additional studies of relevance.

SLR study selection

To be included in the review, economic evaluation articles had to meet predefined eligibility criteria that are detailed using the PICOS method (Table 1).

Table 1. Inclusion Criteria for the Economic Evaluations SLR and Updates.

Population	Adult patients with chronic or episodic migraine
Interventions	Select preventive pharmacological interventions including angiotensin converting-enzyme inhibitors, angiotensin receptor blockers, anticonvulsants, beta-blockers, Botox, calcium channel blockers, CGRP inhibitors and receptor inhibitors, leukotriene receptor antagonists, selective serotonin reuptake inhibitors, serotonin norepinephrine reuptake inhibitors, serotonin agonists, tricyclic antidepressants, and vitamins, minerals, and co-enzymes
Comparators	Any comparator
Outcomes	Outcomes of relevant study designs, including: Costs Life-years gained QALYs Incremental costs and QALYs Incremental cost-effectiveness ratios
Study design	Original economic evaluations considering both the costs and benefits of alternative interventions. Specifically, the following types of analysis: Cost-effectiveness Cost-utility Cost-benefit Cost-minimisation Cost-consequence
Other	Economic evaluations from United Kingdom or Irish perspectives English language Human subjects only In addition, SLRs of economic evaluations were included at the abstract  review stage, then excluded following hand-searching of their reference  lists at the full-text review stage

Abbreviations. CGRP, calcitonin gene-related peptide; QALY, quality-adjusted life-year; SLR, systematic literature review.

The citations found through the searches were first assessed against the eligibility criteria by 2 independent reviewers based on abstract and title. Where the applicability of the inclusion criteria was unclear, the article was included at this stage in order to ensure that all potentially relevant studies were captured. Full-text copies of publications potentially meeting the eligibility criteria were then obtained and reviewed against the same eligibility criteria by 2 independent reviewers. In cases where the article did not give enough information to be sure it met the inclusion criteria at the full-text screening stage, the article was excluded to ensure that only relevant articles were ultimately included in the review. At both the title/abstract and full-text review stages, any disagreements between the reviewers were resolved by discussion until a consensus was met, with a third independent reviewer making the final decision if necessary. For studies meeting the eligibility criteria after the second (full-text) screening stage, data were extracted by a single reviewer into a prespecified data extraction grid and verified by a second individual.

Quality appraisal of economic evaluations

Critical appraisals of each published economic evaluation included in the SLR were conducted using the checklist adapted from Drummond et al.⁶. The checklist for quality appraisal of economic evaluations includes 25 items in the categories of study design, data collection, and analysis and interpretation of results.

Expert opinion consultation of Migraine-Specific economic evaluation approaches

Three key opinion leaders, within both the medical and health economics and outcomes research (HEOR) fields, were consulted in order to provide the criteria for a suitable migraine economic evaluation. Key opinion leaders were also consulted for recommendations and insights on the ideal model structure for migraine that is both clinically and economically meaningful. Recommendations for a model structure were further informed by relevant clinical guidelines (e.g. British Association of the Study of Headache⁷ and NICE⁸) and previous economic evaluations in order to understand clinical practice and previous modelling approaches in migraine prevention.

Results

SLR findings

A total of 3,914 unique articles were identified (3,410 in the original SLR, 187 more in the January 2018 update, and 317 more in the September 2018 update) from the electronic database searches and reviewed at the title/abstract review stage. After title/abstract review in the second update, 238 articles were reviewed for the full-text stage and 37 articles ultimately met the inclusion criteria. An additional 6 articles to those captured through the database searches were identified through congress, website, and hand searching of bibliographies in the original review. No extra articles were identified through hand searches in the first update, but 4 were identified in the second update. The flow of studies through the systematic review process is presented in Appendix Figure 1.

Identified economic evaluations

In total, 8 records reporting on published economic evaluations were identified in the original SLR (see Table 2), and no further economic evaluations were identified during the update. Importantly, several of the identified publications were published on a similar core model.^8–12 All of the publications were cost-utility analyses, and the majority utilized a state-transition (Markov) model structure. Models commonly included health states stratified by the number of migraine or headache days and allowed for treatment discontinuation or treatment on/off periods. The economic evaluations most commonly utilized a 2-year time horizon, and length varied from 1 year to 3 years.

Table 2. Details of the Relevant Economic Evaluations Identified in the Systematic Review.

Reference	Country and Perspective	Summary of Model Structure	Patient Population	QALYs (Interventions and Comparators)	Costs and Outcomes	ICER (Cost per QALY Gained)
NICE [8] Royle et al. [9]	England and Wales; NHS perspective	Model type: Cost-utility analysis using a state-transition (Markov) model to estimate the cost-effectiveness of onabotulinumtoxinA vs placebo (triptans plus pain relief only) in patients with chronic migraine Health states: 6 health states according to number of headache days/month: 0–3; 4–9; 10–14; 15–19; 20–23; ≥24. Patients could enter the model in any of the 3 chronic migraine states Inputs: Clinical data were incorporated from the pooled phase 3 PREEMPT clinical trials programme Health state utility values were collected in PREEMPT using the MSQ and mapped to the EQ-5D Costs included medicine acquisition and administration costs; cost of care consisting of GP visits, ED visits, hospitalization costs, and triptan costs The model used 12-week cycles with a time horizon of 2 years Deterministic and probabilistic sensitivity analyses were conducted The cost year was 2009/2010 Costs and QALYs were discounted at 3.5% per annum	Patients with chronic migraine, defined as ≥15 headache days/month for ≥3 months, of which at least 8 days are with migraine	Discounted totals ≥1 Prior oral prophylactic population: OnabotulinumtoxinA: 1.31 Placebo: 1.22 Incremental QALYs: 0.09 ≥3 prior oral prophylactic population: OnabotulinumtoxinA: 1.29 Placebo: 1.20 Incremental QALYs: 0.09 Totals ≥1 prior oral prophylactic population: OnabotulinumtoxinA: 1.34 Placebo: 1.24 ≥3 prior oral prophylactic population: OnabotulinumtoxinA: 1.32 Placebo: 1.23	Discounted totals ≥1 Prior oral prophylactic population: OnabotulinumtoxinA total costs: £2,388 Placebo total costs: £1,839 Incremental total costs: £549 ≥3 prior oral prophylactic population: OnabotulinumtoxinA total costs: £2,438 Placebo total costs: £1,895 Incremental total costs: £543 Totals ≥1 prior oral prophylactic population: OnabotulinumtoxinA total costs: £2,419 Placebo total costs: £1,879 ≥3 prior oral prophylactic population: OnabotulinumtoxinA total costs: £2,471 Placebo total costs: £1,936	≥1 prior oral prophylactic population: £5,828/QALY gained ≥3 prior oral prophylactic population: £6,083/QALY gained
Batty et al. [10]	England and Wales; NHS perspective	Model type: Cost-utility analysis using a state-transition (Markov) model to estimate the cost-effectiveness of onabotulinumtoxinA vs placebo in patients with chronic migraine Health states: 13 health states: 6 on treatment, 6 off treatment, and death. The on- and off- treatment states were divided by the number of headache days per 28 days Inputs: Clinical data were incorporated from the pooled phase 3 PREEMPT clinical trials program Health state utility values were collected in PREEMPT using the MSQ and mapped to the EQ-5D Costs included cost of care, consisting of GP visits, ED visits, hospitalization costs, and triptan costs The model used 12-week cycles with a time horizon of 2 years Scenario and probabilistic sensitivity analyses were conducted (5,000 simulations were performed) The cost year was 2010 Costs and QALYs were discounted at 3.5% per annum	Patients with chronic migraine, defined as ≥15 headache days/month	Discounted totals OnabotulinumtoxinA: 1.30 Placebo: 1.20 Incremental QALYs: 0.09 Totals OnabotulinumtoxinA: 1.34 Placebo: 1.24	Discounted totals Costs: OnabotulinumtoxinA total costs: £2,997 Placebo total costs: £1,630 Incremental total costs: £1,367 Totals Costs: OnabotulinumtoxinA total costs: £3,077 Placebo total costs: £1,680 Incremental total costs: £1,367 Life-years (undiscounted): OnabotulinumtoxinA: 2.06 Placebo: 2.06 Incremental LYs: 0.00	£15,028/QALY gained (discounted)
SMC [11]	Scotland; NHS perspective	Model type: Cost-utility analysis using a state-transition (Markov) model to estimate the cost-effectiveness of onabotulinumtoxinA injections given every 12 weeks vs BSC in patients with chronic migraine who had previously failed on oral prophylactic therapy due to side effects or lack of efficacy and were in the care of a headache specialist in a secondary care centre Health states: Clinical trial data were used to categorize patients into various health states in the model using mean headache days over a 28-day period. There were 7 health states in total Inputs: Clinical data were incorporated from patient-level data from the post-hoc pooled analysis of key clinical trials BSC was assumed to encompass use of off-label treatments such as gabapentin, venlafaxine, and greater occipital nerve blocks or, alternatively, patients may not have been given prophylaxis but were managed by acute treatments only Health state utility values were derived using EQ-5D data from the IBMS Costs included medicine acquisition and administration costs. Other resource use was estimated from the IBMS The model used a time horizon of 2 years Sensitivity analyses were conducted The model cycle length was not reported No discounting was reported	Patients with chronic migraine, defined as headaches on at least 15 days per month, of which at least 8 days are with migraine, who have failed oral prophylactic therapy due to side effects or lack of efficacy and were in the care of a headache specialist at a secondary care centre	QALY gain: 0.08 (onabotulinumtoxinA vs BSC)	Incremental total costs: £1,394	£17,436/QALY gained
SMC [12]	Scotland; NHS perspective	Model type: Cost-utility analysis using a state-transition (Markov) model to estimate the cost-effectiveness of onabotulinumtoxinA injections given every 12 weeks vs BSC in patients with chronic migraine who had failed ≥3 oral prophylactic therapies, where medication overuse has been appropriately managed, and who are in the care of a headache specialist in a secondary care centre Health states: Clinical trial data were used to categorize patients into on-treatment and off-treatment health states within the model using mean headache days over a 28-day period Inputs: Clinical data were incorporated from the post-hoc pooled analysis of the phase 3 PREEMPT clinical trials program Health state utility values were derived from estimating EQ-5D values by transforming MSQ scores collected from the clinical studies Costs included medicine acquisition and administration costs and cost of care, consisting of GP visits, ED visits, and hospitalization costs The model used a time horizon of 2 years Scenario analyses were conducted The model cycle length was not reported No discounting was reported	Patients with chronic migraine, defined as headaches on at least 15 days per month, of which at least 8 days are with migraine, who have failed ≥3 oral prophylactic therapies, where medication overuse has been appropriately managed and who are in the care of a headache specialist in a secondary care centre	QALY gain (onabotulinumtoxinA vs BSC): 0.08	Incremental total costs (onabotulinumtoxinA vs BSC): £1,012	£12,176/QALY gained
SMC [13]	Scotland; NHS perspective	Model type: Cost-utility analysis using a state-transition (Markov) model to estimate the cost-effectiveness of onabotulinumtoxinA vs BSC in patients with chronic migraine who had failed ≥3 oral prophylactic therapies and for whom medication overuse is appropriately managed Health states: Clinical trial data were used to categorize patients into on-treatment and off-treatment health states in the model using mean headache days over a 28-day period Inputs: Clinical data were incorporated from the post-hoc pooled analysis of the phase 3 PREEMPT clinical trials program Health state utility values were based on EQ-5D values from a European observational study of onabotulinumtoxinA in patients with chronic migraine Costs included medicine acquisition and administration costs, cost of care, consisting of GP visits, ED visits, hospitalization costs, and triptan costs The model used 12-week cycles with a time horizon of 3 years Scenario analyses were conducted No discounting was reported	Patients with chronic migraine, defined as headaches on at least 15 days per month of which at least 8 days are with migraine, who have failed ≥3 oral prophylactic therapies and for whom medication overuse is appropriately managed	QALY gain (onabotulinumtoxinA vs BSC): 0.12	Incremental total costs (onabotulinumtoxinA vs BSC): £1,301	£10,816/QALY gained
SMC [14], Brown et al. [15]	UK; UK NHS and societal perspectives	Model type: Cost-utility analysis using a decision-tree model to estimate the cost-effectiveness of topiramate vs no preventive treatment (acute treatment only) in patients with migraine Health states: On entering the model, patients could either receive topiramate or no treatment. Those receiving topiramate treatment could have one of the following clinical responses: ≥75%, 75%–50%, or <50% for mean reduction in monthly migraine frequency Inputs: Clinical data were incorporated from topiramate clinical trials The model assumed 40% of topiramate patients drop out in the first month of treatment and receive no clinical benefit for the 1-year period Health state utility values were derived from transformations of SF-36 data collected as part of topiramate clinical trials Costs included cost of care, consisting of GP visits, ED visits, hospitalization costs, and triptan costs The time horizon was 1 year Sensitivity analyses were conducted The cost year was 2005 No discounting was reported	Patients with migraine (base-case frequency is 6/month)	QALY gain (topiramate vs no preventive treatment): 0.0384 No treatment: NR Incremental QALYs: NR	Costs/month: Topiramate total costs: £37.13 No treatment total costs: £18.80 Incremental total costs per migraine averted: £10.13	£5,728/QALY gained

Abbreviations. BSC, best supportive care; ED, emergency department; EQ-5D, EuroQol 5-dimensions; GP, general practitioner; IBMS, International Burden of Migraine Study; ICER, incremental cost-effectiveness ratio; MSQ, Migraine Specific Quality of Life Questionnaire; NHS, National Health Service; NR, not reported; QALY, quality-adjusted life-year; UK, United Kingdom.

Quality appraisal of economic evaluations

Critical appraisals of each published economic evaluation included in the SLR were conducted using the checklist adapted from,⁶ as recommended by NICE. Economic evaluations were appraised on 35 items across the categories of study design, data collection, and analysis and interpretation of results. The results of these critical appraisals are presented in Table 3. All of the economic evaluations fulfilled the items for proper disclosure of the study design. However, several items in the data collection category were missing including the details of the subjects from whom valuations were obtained. Most of the items within the analysis and interpretation of results section were included, but information on the use of a discount rate and rationale and details were often not published.

Table 3. General Quality Appraisal of Economic Evaluations.

Reference	NICE^8 Royle et al.^9	SMC^11	SMC^12	SMC^13	Batty et al.^10	SMC^14, Brown et al.^15
Study design
Was the research question stated?	Y	Y	Y	Y	Y	Y
Was the economic importance of the research question stated?	Y	Y	Y	Y	Y	Y
Was/were the viewpoint(s) of the analysis clearly stated and justified?	Y	Y	Y	Y	Y	Y
Was a rationale reported for the choice of the alternative programs or interventions compared?	Y	Y	Y	Y	Y	Y
Were the alternatives being compared clearly described?	Y	Y	Y	Y	Y	Y
Was the form of economic evaluation stated?	Y	Y	Y	Y	Y	Y
Was the choice of form of economic evaluation justified in relation to the questions addressed?	Y	Y	Y	Y	Y	Y
Data collection
Was/were the source(s) of effectiveness estimates used stated?	Y	Y	Y	Y	Y	Y
Were details of the design and results of the effectiveness study given (if based on a single study)?	N/A	N/A	N/A	N/A	N/A	N/A
Were details of the methods of synthesis or meta-analysis of estimates given (if based on an overview of a number of effectiveness studies)?	Y	N	N	N	Y	N
Were the primary outcome measure(s) for the economic evaluation clearly stated?	Y	Y	N	N	Y	Y
Were the methods used to value health states and other benefits stated?	Y	Y	Y	Y	Y	Y
Were the details of the subjects from whom valuations were obtained given?	N	N	N	N	N	N
Were productivity changes (if included) reported separately?	N	N/A	N/A	N/A	Y	N
Was the relevance of productivity changes to the study question discussed?	Y	N	N	N	Y	Y
Were quantities of resources reported separately from their unit cost?	N	N/A	N	N	Y	Y
Were the methods for the estimation of quantities and unit costs described?	Y	N/A	Y	Y	Y	Y
Were currency and price data recorded?	Y	Y	Y	Y	Y	Y
Were details of price adjustments for inflation or currency conversion given?	N/A	N/A	N/A	N/A	N/A	Y
Were details of any model used given?	Y	Y	Y	Y	Y	Y
Was there a justification for the choice of model used and the key parameters on which it was based?	Y	Y	Y	Y	Y	Y
Analysis and interpretation of results
Was the time horizon of cost and benefits stated?	Y	Y	Y	Y	Y	Y
Was the discount rate stated?	Y	N	N	N	Y	N
Was the choice of rate justified?	N	N/A	N/A	N/A	Y	N/A
Was an explanation given if cost or benefits were not discounted?	N/A	N	N	N	N/A	N
Were the details of statistical test(s) and confidence intervals given for stochastic data?	N/A	N/A	N/A	N/A	N/A	N/A
Was the approach to sensitivity analysis described?	Y	Y	Y	Y	Y	Y
Was the choice of variables for sensitivity analysis justified?	Y	Y	Y	Y	Y	Y
Were the ranges over which the parameters were varied stated?	Y	N	Y	N	Y	Y
Were relevant alternatives compared in the incremental analysis?	Y	Y	Y	Y	Y	Y
Was an incremental analysis reported?	Y	Y	Y	Y	Y	Y
Were major outcomes presented in a disaggregated and aggregated form?	Y	Y	Y	Y	Y	Y
Was the answer to the study question given?	Y	Y	Y	Y	Y	Y
Did conclusions follow from the data reported?	Y	Y	Y	Y	Y	Y
Were conclusions accompanied by the appropriate caveats?	Y	Y	Y	Y	Y	Y

Abbreviations. N, no; N/A, not applicable; Y, yes.

Quality appraisal of economic evaluations with respect to evaluating migraine treatments specifically

The economic evaluations identified through the SLR were carefully assessed and provided valuable insights to better understand alternative ways of economically evaluating migraine interventions and to determine if other approaches were needed.

Consultation with medical and HEOR experts identified clinical meaningfulness and scientific robustness as the 2 key areas for the design of economic evaluations. For a migraine model to be clinically meaningful, it should have face validity and ensure each model component reflects current clinical practice and understanding. For migraine interventions, these include:

• Capture migraine as a spectrum disease (MMDs between 0 and 28; ie, chronic and episodic together).

• Include a discontinuation rule based on treatment response (treatment response being defined by % reduction in MMDs).

• Include forms of negative discontinuation (non-responders, adverse event [AE]-related discontinuation, and long-term general discontinuation) and positive discontinuation (patients with well-managed migraine no longer requiring treatment).

HEOR and medical experts also concluded that scientific robustness is essential to have strong internal, external, cross, and predictive validity. The key criteria items for a developing a scientifically robust economic evaluation for migraine treatments are:

• The evaluation should closely reflect trial results as a form of validation.

• The model should simultaneously track both MMD frequency (in order to estimate quality-adjusted life-years [QALYs] gained and resource utilization incurred) and response status (to facilitate a response-based discontinuation rule).

• The model should reflect how patients are distributed across MMD frequencies in order to capture the non-linear impact of MMD frequency change on both QALYs and resource utilization.

Modelling MMD frequency as a continuous outcome, rather than as a series of categories, has several advantages over the use of arbitrary health states which have been previously reported.^16,17 The use of continuous MMD frequency distribution retains all of the observed information to accurately estimate the difference in outcomes between groups. This approach also more readily allows for indirect comparisons based on study primary endpoints (e.g. probability of 50% response, or change in MD frequency per 28 days). In addition, the number of MMDs and associated events (e.g. emergency department visits) can be directly quantified.

In consideration of clinical guidelines (International Headache Society [IHS] and NICE), input from the KOLs’ previous economic evaluations were reviewed against the criteria outlined above (Table 4). Several models^8–13 included disease attack frequency via headache days per month but not MMDs. Similarly, these models included a discontinuation rule based on reduction in headache days or rates derived from clinical trials, but not explicitly defined as percentage reduction in MMDs. Previous economic evaluations were parameterized with clinical trial data, but reviewers identified areas of concern. In the 2012 appraisal of botulinum toxin, reviewers noted that the model and trial datasets for headache days per month at week 24 corresponded poorly and that this overestimated the net impact of botulinum toxin type A by 53% and there were no clinical data to support the use of the 2 stopping rules. A modelling approach in which patients are distributed across MMD frequencies may more closely reflect migraine trial results as a form of validation. None of the identified evaluations had a structure that allowed accurate and simultaneous tracking of both MMD frequency and response status.

Table 4. Migraine-Specific Quality Appraisal of Economic Evaluations.

Reference	NICE^8 Royle et al.^9	SMC^11	SMC^12	SMC^13	Batty et al.^10	SMC^14, Brown et al.^15
Clinically meaningful criteria
The model structure captures MMD frequencies between 0 and 28, reflecting the nature of migraine as a spectrum disease	N	N	N	N	N	N
Include a discontinuation rule based on treatment response (treatment response being defined by % reduction in MMDs)	N	N	N	Y	N	N
Include other forms of discontinuation (AE-related, positive)	Y	Y	Y	Y	Y	Y
Scientifically robust criteria
The evaluation should closely reflect migraine trial results as a form of validation	N	N	N	N	N	N
The structure allows accurate and simultaneous tracking of both MMD frequency for QALY estimation and response status for response-based discontinuation rule	N	N	N	N	N	N
The model reflects how patients are distributed across MMD frequencies in order to capture the non-linear impact of migraine day loss/gain	Y	Y	Y	Y	Y	N

Key, N, no; Y, yes.

A novel Cost-Effectiveness model structure

In light of the critical appraisal described above, a novel model structure was developed with the aim of meeting all criteria simultaneously. A decision-tree plus Markov structure was developed as a cost-effectiveness model for migraine therapies. Reflecting clinical practice, the decision-tree component represents an assessment period, allowing for treatment discontinuation based on safety and clinically relevant response criteria. The Markov component represents a post-assessment period, where treatment responders and non-responders follow distinct treatment pathways (Figure 1)¹⁸. Responders to treatment without safety or tolerability issues continue on that treatment over the post-assessment with an optional re-evaluation period, which may lead to positive discontinuation, while non-responders discontinue and do not reinitiate treatment. The underlying assumption of the model is that both costs and QALYs can be estimated based on the MMD frequency being experienced by patients. Thus, each health state in the model is associated with a patient distribution across MMD frequencies, which contrasts with defined ranges of migraine day or headache day frequency used for health states used in previous decision tree or Markov model approaches.

Figure 1. Illustration of CE Model Structure.

A key component of the model is the expression of treatment efficacy over the assessment period. The assessment period in the model simply replicates, as faithfully as possible, the MMD distributions, per treatment, as observed in the available clinical trial data. Given the selected population of interest, the model estimates (for each treatment):

1. The MMD distribution for all patients at baseline (Week 0).

2. The proportion of responders at Week 12 (based on given response threshold).

3. The MMD distributions for responders and non-responders at Week 12.

The overall MMD distribution (for each treatment) at Week 12 is implied by the combination of the responder and non-responder MMD distributions (weighted by the estimated proportions of responders and non-responders). The response threshold is defined as a specified percentage reduction in MMDs. This process is illustrated in Figure 2. Note that the MMD distributions displayed here are not based on real data and are for illustrative purposes only.

Figure 2. Detail of Assessment Period.

Discussion

The burden patients experience each day of migraine is meaningful. In a large retrospective, cross-sectional study, an increase of 1 headache-free day (HFD) was associated with an average decrease in absenteeism of 3.9% and presenteeism of 2.1%¹⁹. Resource utilization was also reduced, with an increase of 1 HFD being associated with expected decreases in healthcare provider visits and neurologist visits of 1.0% and 4.7%, respectively. The benefits of each additional HFD corresponded to average increases of 0.171, 0.306, 0.003, and 0.008 points for the mental composite summary (MCS) score, physical component summary (PCS) score, SF-6D utility score, and EQ-5D index score, respectively (p < 0.001 for all). The burden of migraine is worse for people who need to change preventive therapy. For these patients, 87% reported that migraine had a negative impact on professional, private, or social domains of life²⁰. New therapies for migraines, particularly those for treatment-refractory patients, have the opportunity to bring meaningful improvement for patients, and economic evaluations can be useful tools for assessing their value.

A review of previous modelling approaches of economic evaluations, clinical guidelines, and expert opinion has identified key aspects that should be considered in the development of future economic evaluations for migraine therapies for the UK and Ireland.

To avoid the use of arbitrary established MMD cut-offs, the use of a full MMD distribution is likely to better quantify the impact of associated health benefits and costs. This approach also better reflects the fact that migraine is a spectrum disease with both chronic and episodic characteristics.

A discontinuation rule is required based on treatment response where treatment response is defined as a percent reduction in MMDs. Other forms of discontinuation to cater to tolerability issues (AE-related, positive) should be included to provide flexibility in terms of discontinuation rules, reasons for discontinuation, and consequences for discontinuation. It is particularly important that tolerability be included in any economic evaluation for appropriate risk-benefit assessment where tolerability is the limiting factor for sustained efficacy with prior standard of care in a chronic disorder such as migraine.

Lastly, if comparative data are available, the economic evaluation should include the ability to compare the treatment of interest with other preventive treatments.

Compared to other disease areas, the number of economic evaluations in migraine was limited. Many of these evaluations utilized a similar structure, but each approach has unique strengths and limitations. Importantly, some evaluations with perspectives outside of the UK or Ireland were not in the scope of our research and not included. A review of these evaluations may provide additional insights for developing a novel economic evaluation of migraine therapies. In the time since the SLR was completed, we identified an update to the cost-effectiveness model of onabotulinumtoxinA for the prevention of headache in adults with chronic migraine²¹. However, this update was to incorporate new EQ-5D utility estimates, and there were no novel changes to the model structure.

The proposed modelling approach that estimates both MMD frequencies and response to treatment is reflective of clinical practice in Europe and is appropriate in assessing the cost-effectiveness of preventive treatments of migraine. In this study, the proposed approach is reflective of migraine as a spectrum disorder, and it is likely to appropriately capture the costs and health benefits of migraine therapies vs appropriate comparators.

Similar to previous economic evaluations, there are some limitations inherent with the proposed model. Limitations of this modelling approach include the assumption that treatment response can be based on MMD frequency alone and the assumption that both health-related quality of life and resource utilization can be estimated based on MMD frequency alone. While it is envisioned that response to therapy in the initial version of this model will be based on reduction in MMD frequency alone, response in later versions could be based on a combination of multiple clinically relevant endpoints and efficacy parameters such as frequency, severity, enhanced effect of acute migraine medication, and benefits in the pro-dromal and post-dromal periods.

A reliable cost-effectiveness assessment utilizing the proposed approach will require careful consideration of inputs and definitions to parameterize the model. This includes establishing responder and non-responder definitions, defining the duration of treatment benefit for responders, determining the frequency of patients re-evaluated, and the most appropriate time horizon. Extensive sensitivity analyses will be important for addressing these issues, but the proposed modelling approach provides a basis for transparency regarding these assumptions.

Conclusion

Based on the findings of the SLR, a limited number of economic evaluations for migraine therapies have been published. The design of any economic evaluation should be informed by clinical expert opinion and clinical guidelines. As new therapies are being developed and approved for use in patients with high unmet needs, such as those patients with migraine experiencing at least 4 MMDs, new economic models developed in this disease area should allow for the assessments of cost-effectiveness across the full spectrum of migraine. The approach proposed in this paper captures all of the desired elements to an economic evaluation of migraine therapy and is suitable to assess new migraine therapies such as CGRP antagonists. Future research is needed to better characterize the data to adequately structure and parameterize an economic model to support decision making for migraine therapies.

Transparency

Declaration of funding

This study was sponsored by Novartis Pharma AG.

Declaration of financial/other interests

RM, JH, VH, AD, VU, SV, PV are employees of Novartis and RM, VH and AD owns shares in the company. FM reports personal fees from Novartis, Teva and Allergan outside the submitted work. SP reports personal fees from Amgen, outside the submitted work. PG reports personal fees from Alder Biopharmaceuticals, personal fees from Allergan, grants and personal fees from Amgen, personal fees from Autonomic Technologies Inc., personal fees from Biohaven Pharmaceuticals Inc., grants and personal fees from Eli Lilly and Company, personal fees from Electrocore LLC, personal fees from eNeura Inc, personal fees from Impel Neuropharma, personal fees from MundiPharma, personal fees from Novartis, personal fees from Teva Pharmaceuticals, other from Trigemina, personal fees from WL Gore, outside the submitted work. In addition, PG has a patent Magnetic stimulation for headache licensed to eNeura without fee and fees for publishing from Oxford University Press, Massachusetts Medical Society, Wolters Kluwer, and for medico-legal work.

JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgements

The authors gratefully acknowledge the contributions of Rose Wickstead, Eleanor Ferns, and Amy Buchanan-Hughes, who conducted the systematic review and performed the initial data extractions for the referenced economic models. The authors acknowledge David Campbell, Clémence Arvin-Berod, and Marie-Josée Martel of Xcenda for their editorial contributions to this work.