Table 2.

Details of the Models and Model Inputs in the studies included in the Systematic Review

Authors (Year), Countries	Model type	Perspective	Time horizon	Cost included in the model	Source of cost and resource inputs	Currency, price year
Journal Articles
Batty AJ, et al. (2013), United Kingdom [16]	A Markov model with 13 health states, including death. The 12 states were split into two parallel stages: on treatment and off treatment. A 12-weeks cycle length was employed. The model also considered negative and positive stopping rule for the treatment.	National Health Service (NHS)	2 years	Cost of Botox; Consultant time to take participant history, tailor prophylactic and acute treatment; consultant time to administer the injections; cost of care including general practitioner (GP) visits, emergency department (ED) visits, hospitalisation and triptan costs.	Resource used was informed by International Burden of Migraine Study (IBMS), with unit costs taken from NHS reference cost, cost of triptans per attack was based on the weighted average costs in the UK in 2010.	UK £ 2010
Giannouchos TV, et al. (2019), Greece [17]	Decision tree model.	Payer and Societal perspective	1 year	Direct costs included the cost of the two drugs and administration, the use of acute drugs under usual care, and hospitalisation costs, physician, and ED visits. Indirect costs for the societal perspective analysis included wages lost on workdays.	Resource utilisation data were obtained from four previously published studies and the cost inputs were obtained from publicly available data for the Greek healthcare sector and on the governmental pricing system derived from a public Greek hospital.	Euro € 2019
Hansson-Hedblom A, et al. (2020), Norway and Sweden [18]	A Markov model with 13 health states, including death as mentioned in Batty AJ, et al.	Payer and Societal perspective	10 years	Direct cost included cost of Botox, Neurology consultant appointment, specialist nurse appointment, cost of care including GP visits, ED visits, hospitalisation and triptan costs. Indirect cost involved productivity cost.		SEK, 2018 for Sweden; NOK, 2018 for Norway
Hollier-Hann G, et al. (2020), United Kingdom [19]	CUA using Markov model with 13 health states, including death as mentioned in Batty AJ, et al.	NHS	2 years	Cost of Botox; Consultant time to take participant history, tailor prophylactic and acute treatment; consultant time to administer the injections; cost of care including GP visits, ED visits, hospitalisation and triptan costs.	Resource used was informed by IBMS, with unit costs taken from NHS reference cost, cost of triptans per attack was based on the weighted average costs in the UK in 2010.	UK £ 2010
Lipton RB, et al. (2018), United States of America [20]	A Markov model was implemented based on the clinical data from the Episodic Migraine (EM) and Chronic Migraine (CM) studies for the sub-groups of participants with prior treatment failures. The cycle length was 28 days.	US societal perspective	10 years	Direct medical costs included cost of medicine and its administration, GP visits, ED visits, hospitalisations, and specialist neurologist consultations based on published unit costs. Cost of medicines to treat acute attacks. Indirect costs included productivity cost associated with presentism and absenteeism.	Average annual medical resource use is taken from a published 2009 analysis of survey data from 7437 migraine participants in the US	USD $ 2017
Mahon R, et al. (2021), Sweden [21]	A hybrid decision-tree plus Markov model was developed	Swedish societal perspective	10 years	Direct cost included cost of medicine and its administration, ED visit, hospitalisation, GP visit, consultant visit, Nurse/physician visit, Triptan medication and other medications. Indirect cost related to absenteeism and presenteeism were included.	Resource utilisation and efficacy data were sourced from four trials (CM295, STRIVE, ARISE and LIBERTY). Study 178, which had an open-label phase of 256 weeks, was used to inform long-term assumptions regarding those who continued treatment. Resource usage cost were obtained from the price list of Southern Sweden. Productivity costs were included from the published literature.	SEK, 2018
Ruggeri et al. (2013), Italy [22]	A Markov model with 13 health states, including death as mentioned in Batty AJ, et al.	Italian National Health Service and a societal perspective.	2 years	Direct cost included cost of medicine and its administration, GP visit or outpatient cost, ED visit, hospitalisation and cost of Triptans. Indirect costs included productivity cost.	Resource utilisation data was derived from IBMS study. Costs were obtained from the local government data.	Euro € 2013
Sussman M, et al. (2018), United States of America [23]	A hybrid Monte Carlo participant simulation and Markov cohort model was constructed. Participants in both cohorts (EM and CM) must have failed at least one previous preventive therapy prior to model entry since Calcitonin gene-related peptide (CGRP) pathway antagonists are expected to be used as second-line therapies. Participants in the EM cohort must have had between 4 and 14 monthly migraine days (MMDs) and participants in the CM cohort must have had at least 15 MMDs at baseline.	US Societal and Payers perspective	2 years	Direct cost included - acute medication cost, physician visit, ED visit, adverse events, and hospitalisation cost. Indirect costs included productivity cost	Data inputs for the model were derived from the Erenumab pivotal and Open labelled extended trials, and Botox pivotal trial, published literature, and publicly available sources.	USD $ 2017
Vekov (2019), Bulgaria [24]	A hybrid model including a Monte Carlo simulation and a Markov cohort model. The input data to the model are the primary and secondary clinical endpoints in the randomized trials NCT02066415 and NCT02483585. They measure the change in the number of days with migraine per month at weeks 12 and 24, the number of days per month with symptomatic migraine therapy	Payers’ perspective	2 years	Only the cost of medicines was included; other healthcare costs were assumed to be equal for both therapies and hence excluded.	Resource utilisation (medicine usage) data was obtained from the randomised trial NCT02066415.	Bulgarian Rev. (BGN) 2019
Other reports
CADTH (Botox) (2019), Canada [25]	Hybrid model with decision tree for 12-week assessment period, classifying patients as responders and non-responders, and Markov model for post-assessment with 12-week cycle lengths.	Canadian public health care payer perspective	3 years	Direct costs included cost of medicine and its administration, GP visits or outpatient cost, ED visits, hospitalisation and cost of Triptans. Indirect costs included productivity cost.	Resource used was informed by IBMS, with unit costs taken from NHS reference costs, cost of triptans per attack was based on the weighted average costs in the UK in 2010.	CAD $ 2019
CADTH (Erenumab) (2019), Canada [26]	Hybrid model with decision tree for 12-week assessment period, classifying participants as responders and non-responders, and Markov model for post-assessment with 12-week cycle lengths.	Canadian public health care payer perspective	3 years	Direct costs included cost of medicine and its administration, GP visits or outpatient cost, ED visits, hospitalisation and cost of Triptans. Indirect costs included productivity cost.	Resource used informed by the trial and cost data was obtained from manufacturer and other local data resources	USD $ 2018
ICER (2018), United States of America [27]	Markov model comprising CGRP inhibitor versus no preventive treatment arms. The intervention arm of the model includes three health states: 1) CGRP inhibitor treatment, 2) no preventive treatment, and 3) death. The comparator arm includes two health states: 1) no preventive treatment and 2) death.	Health system payer perspective	2 years	Direct medical care cost including cost of medicine, GP visit, outpatient visit cost, ED visit and hospitalisation.	Resource used was informed by International Burden of Migraine Study (IBMS), with unit costs taken from the local data resources	CAD $ 2019
NICE: Erenumab (2019), United Kingdom [29]	A decision-tree plus Markov model included two health states - on treatment and discontinuation of treatment once patients were classified as responders or non-responders.	NHS perspective	Lifetime	Migraine specific cost related to hospitalisation and ED visits, health care professional visits and use of acute medication.	Resource used was informed by National Health and Wellness survey conducted in migraine population, with unit costs taken from the local data resources	UK £ 2018
NICE: Fremanezumab (2019), United Kingdom [28]	A decision-tree plus Markov model included two health states - on treatment and discontinuation of treatment once patients were classified as responders or non-responders.	NHS perspective	10 years	Migraine specific cost related to hospitalisation and ED visits, health care professional visits and use of acute medication.	Resource used was informed by National Health and Wellness survey conducted in migraine population, with unit costs taken from the local data resources	UK £ 2019
NICE: Galcenzenumab (2020), United Kingdom [30]	A decision-tree plus Markov model included two health states - on treatment and discontinuation of treatment once patients were classified as responders or non-responders.	NHS perspective	Lifetime	Migraine specific cost related to hospitalisation and ED visits, health care professional visits and use of acute medication.	Trial-specific (CONQUER) data and the resource utilisation data from Lipton et al. (2018) [20].	UK £ 2020
Warwick Evidence (2011), United Kingdom [31]	A Markov model with 13 health states, including death. The 12 states were split into two parallel stages: on treatment and off treatment. Each treatment state was sub-divided into categories based on the number of headache days per 28 days. Three health states for EM (0–3, 4–9, and 10–14 headache days per 28 days), and three health states for CM (15–19, 20–23, and 24+ headache days per 28 days). A 12-weeks cycle length was employed. The model also considered negative and positive stopping rule for the treatment.	NHS perspective	2 years	Migraine specific cost related to hospitalisation and ED visits, health care professional visit and use of acute medication.	Resource used was informed by IBMS, with unit costs taken from NHS reference cost, cost of triptans per attack was based on the weighted average costs in the UK in 2010.	UK £ 2011

CGRP Calcitonin gene-related peptide, CM Chronic migraine, CUA Cost utility analysis, ED Emergency department, EM Episodic migraine, GP General practitioner, IBMS International Burden of Migraine Study, MMD Monthly migraine days, NHS National Health Service