Preventive drug treatments for adults with chronic migraine: a systematic review with economic modelling

Search strategy

The search strategy for the clinical effectiveness review (see Chapter 2) and the AEs review (see Chapter 3) was constructed by an information specialist (AB), in consultation with the project team. The search strategy was initially constructed in MEDLINE, using both free text keywords and thesaurus (MeSH) terms for migraine/headache and the prophylactic drug interventions of interest, with the addition of a search filter for RCTs. No date or language limits were applied. The MEDLINE strategy was checked by another information specialist (not involved in the project) for any omissions or errors in spelling, search syntax, structure and use of MeSH, before being translated for the other bibliographic databases. Full search strategies can be found in Appendix 1, Table 22.

The following databases and clinical trials registers were searched between 8 and 15 September 2021:

• MEDLINE All, 1946 to 7 September 2021 (via Ovid);

• EMBASE Classic + EMBASE, 1947 to 7 September 2021 (via Ovid);

• Cochrane Central Register of Controlled Trials (CENTRAL), Issue 9 of 12, September 2021 (via Cochrane Library);

• Science Citation Index Expanded, 1970 to present (via Web of Science);

• Global Index Medicus (all regional indexes, via WHO website);

• ClinicalTrials.gov;

• International Clinical Trials Registry Platform (ICTRP) (via WHO).

Records retrieved by the database and the trials register searches were exported into EndNote X9, to enable identification and systematic removal of duplicates. 54

An additional pragmatic search in MEDLINE, EMBASE and the Cochrane Database of Systematic Reviews was performed to identify recent systematic reviews of prophylactic migraine treatments. The reference lists of the outputs of this search, and those also of the NICE, SIGN and American Headache Society guidelines, were checked for relevant literature. Authors of key studies were contacted and asked for details of any articles (e.g. reports, papers published or unpublished) that may not have been captured in our search. We performed forward and backward citation tracking from all included papers using Web of Science Core Collection (and Google Scholar where papers were not available in Web of Science).

Following consultation with the study team clinical experts, we conducted a further round of searches for three medicines not previously included in the search strategies: riboflavin, magnesium and CoQ-10. Our clinical experts suggested that these medicines are currently used within the UK. This supplemental search followed the same process as initial searches and was conducted in February 2022. Searches for all the prophylactic drug interventions of interest were updated in November 2022 to identify any additional publications that had become available. Searches to check for any retractions, errata or similar, relating to included studies, were also undertaken at this time. Full details of all searches are provided in Appendix 1.

Assessing relevance and inclusion of studies

The first round of screening was based on title and abstract and was conducted by two reviewers (AB, SN). The second phase of screening was performed according to PICO criteria (see Box 1 for inclusion criteria and Box 2 for exclusion criteria). At this stage, the abstracts of the retrieved studies were reviewed independently by two out of four reviewers (MU, SN, AA, ND). The full texts of the remaining studies were retrieved, and the same combination of the reviewers conducted an additional round of full-text screening according to the pre-specified inclusion/exclusion criteria. For both the clinical effectiveness review (see Chapter 2) and AEs review (see Chapter 3), the screening process was the same.

We excluded studies with fewer than 100 participants per arm to ensure that we included better-quality studies and to avoid loss of precision on our NMA by including heterogenous studies. 55,56 Studies with fewer than 200 participants will not have been adequately powered to show a standardised mean difference of less than 0.5. Smaller studies are also typically older and do not use an adequate definition of chronic migraine and are of poor quality.

Data extraction for systematic review and network meta-analysis

Data for included studies were extracted by one reviewer (SN) and 20% were randomly checked for accuracy by another reviewer (SK). Data extraction forms were developed in Microsoft Excel (Microsoft Corporation, Redmond, WA, USA) to capture the following information: ClinicalTrials.gov identifier (National Clinical Trial number), study name, study characteristics, patient demographics including baseline characteristics, intervention and comparator details, outcome(s) of interest with relevant data and measure of variability, time point of outcome measurements, duration of treatments, AEs and serious adverse events (SAEs).

Means and standard deviations (SDs) for continuous outcomes were extracted. If SDs were not provided, we calculated them from standard errors, CIs or other measures. 57 We also contacted authors by e-mail to ask for the original data in the event of any missing data.

Assessment of risk of bias for included trials

The Cochrane risk-of-bias (RoB 2) tool for randomised trials58 was applied for assessing the risk of bias of all trials independently by two members (SN, SK). The tool was used to determine whether there was high, some, or low risk of bias in the following domains: (1) arising from the randomisation process, (2) due to deviations from the intended interventions (effect of assignment to intervention), (3) missing outcome data, (4) measurement of the outcome and (5) selection of the reported result. In this approach, the rating low risk of bias ‘is judged to be at “low risk of bias” for all domains’, and the trial ‘is judged to raise “some concerns” in at least one domain for this result, but not to be at “high risk of bias” for other domains, whereas the trial ‘is judged to be at “high risk of bias” in at least one domain’ or the trial ‘is judged to have “some concerns” for multiple domains in a way that substantially lowers the confidence in the result’. 58

Assessment of certainty in evidence for included trials

We assessed the degree of certainty of evidence, all comparisons for each outcome, by using the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) framework independently by two members (SN, SK). 59 Any discrepancies in any of the screening steps were referred to the third reviewer (MU). In GRADE, RCTs were considered as a high quality of evidence (where authors have a lot of confidence that the true effect is similar to the estimated effect) to very low-quality evidence (where authors believe that the true effect is probably markedly different from the estimated effect). There are five domains for rating down GRADE including: study limitations (risk of bias), imprecision, inconsistency, indirectness and publication bias. There are three domains for rating up GRADE including: large magnitude of effect, dose–response gradient, and all plausible confounding.

Outcomes of interest

1. Monthly headache days: As reported in the original papers.

2. Monthly migraine days: As reported in the original papers.

3. Migraine-specific quality of life (MSQ): The MSQ version 2.1 is a 14-item questionnaire that measures a patient’s quality of life over the last 4 weeks across three domains: migraine-specific quality of life-restrictive role function (MSQ-RR), seven items that assess the functional impact of migraine through limitations on a patient’s daily work and social activities; migraine-specific quality of life-preventive role function (MSQ-PR), four items that measure the impact of migraine through prevention of daily work and social activities; and migraine-specific quality of life-emotional function (MSQ-EF), three items that evaluate the emotional impact on migraine. The score ranges from 0 to 100, with a higher score indicating better quality of life. 60

4. The headache impact test-6 (HIT-6): The HIT-6 consists of six items: pain, social functioning, role functioning, vitality, cognitive functioning and psychological distress. There are five responses to each of the six items: ‘never’, ‘rarely’, ‘sometimes’, ‘very often’ or ‘always’. These responses are summed together to produce a total score for the HIT-6. A lower score (49 or less) is categorised as having little or no impact and a higher score (60–78) is categorised as having a severe impact. 61

5. EuroQol-5 Dimensions, five-level version (EQ-5D-5L): The EQ-5D-5L descriptive system is a preference-based health-related quality of life (HRQoL) measure with five dimensions that include mobility, self-care, usual activities, pain/discomfort and anxiety/depression, with each dimension assessed at five levels: from no to extreme problems. 62

6. Migraine Disability Assessment (MIDAS): The instrument was developed to assess headache-related disability for migraine patients over a 3-month recall period. 63 The questionnaire contains five questions regarding the number of days of missed work/school, reduced productivity at work/school, missed household work, reduced productivity in household work, and missed family and/or social activities. The MIDAS score is calculated by summing the five items. Higher scores depict increased disability due to headache. The total MIDAS score can be categorised according to disability: 0–5, minimal or infrequent disability; 6–10, mild or infrequent disability; 11–20, moderate disability; and 21+, severe disability. 63

7. Work Productivity and Activity Impairment – Specific Health Problem (WPAI:SHP). The WPAI:SHP questionnaire measures the effect of different health conditions on work productivity, generating scores for absenteeism, presenteeism, absenteeism plus presenteeism, and activity impairment outside work. 64 WPAI:SHP is a version of Work Productivity and Activity Impairment that can be modified for use with a specific disease, such as migraine. 64

8. Patient Health Questionnaire 9-item (PHQ-9): The PHQ-9 is a self-administered instrument for screening, diagnosing, monitoring and measuring the severity of depression over the last 2 weeks. Each question is scored from ‘0’ (not at all) to ‘3’ (nearly every day). The total score for PHQ-9 is obtained by summing the score for each question. 65

9. Functional Impact of Migraine Questionnaire (FIMQ): The FIMQ is a 20-item questionnaire that measures patient-relevant impacts of migraine in the past 7 days across three domains: activity impairment (14 items), emotional functioning (3 items) and cognitive functioning (3 items). Individual items are then transformed to a 0–100 scale, with higher scores indicating a greater level of migraine impact. 66

Selection of data and synthesising data for the network meta-analysis

Network meta-analysis is a methodology that simultaneously combines three or more interventions in a single analysis using data from several studies. Results for each pairwise comparison combine both the direct evidence in primary studies and the indirect evidence, which has not been made directly within the studies to form a network. 67 In other words, a NMA provides an estimation of treatment effect size for each pair of interventions, regardless of whether they have been compared directly in a RCT or not. 67 In addition to facilitating comparisons between interventions, NMA provides a ranking of the interventions based on their effectiveness. This can help decision-makers choose the most suitable and effective treatments. 68

The stepwise feasibility framework was used to ensure that the underlying assumptions are systematically explored, and also to ensure that pooling and comparing the treatment effects for a particular research question are transparent. 69 It comprises of four steps to illustrate the heterogeneity and differences within or between direct treatment comparisons in terms of treatment and outcome characteristics, the study and patient characteristics, baseline risk and observed treatment effects. 69

We conducted a NMA for those outcomes which were provided data for more than three interventions to obtain the treatment effect size for the clinical effectiveness review.

Two different NMA models were conducted:

• Fixed-effects model – this model assumes that all studies included in a NMA are estimating a single true underlying effect. However, if there is significant variation in the effect sizes, known as statistical heterogeneity, then the fixed-effects model may not be appropriate.

• Random-effects model – this model assumes that the estimated treatment effects observed across studies can differ due to both actual differences in the treatment effect in each study, as well as differences in sampling. 70

We chose between the different NMA models by using the posterior mean deviance as an indicator of model fit and the deviance information criterion (DIC). DIC is a metric used to assess the goodness of fit of a statistical model while also taking into account its complexity. It penalises models for their complexity and therefore favours simpler models over more complex ones. 71 DIC differences of three or more are considered meaningful between models. 72 When both model results were similar, we chose the results from the most parsimonious model.

Network plots were created for each analysed outcome. The node sizes of the network plots are proportional to the number of participants randomised to each of the interventions, whereas the thickness of the edges (lines) is proportional to the number of participants contributing to that comparison. 73 Stata SE17 was used to generate the forest plots for each intervention’s comparison with placebo as the reference treatment. 74 The comparisons of all interventions were interpreted using leagues tables showing all pairwise comparisons with associated 95% credible intervals (CrIs).

In our review, all outcomes are presented in continuous format (change from baseline). The calculated point estimates were mean differences (MDs) with their associated 95% CrIs. We considered follow-up periods of 12 and 16 weeks as a measurement time point for all outcomes, because most of the interventions in the included trials had reported outcomes at week 12 or 16. The only exception to this was the data for BTA as most of their outcomes were reported at week 24. Hence, data for BTA are evaluated in a longer time frame of 6 months. Where outcome data were presented for multiple time points, we took the data closest to 3 months follow-up as the main time point.

We excluded studies that had insufficient information about the mean change from baseline (SD) for each outcome per arm [e.g. in the absence of the mean change from baseline, we calculated it by subtracting the post-treatment value from the baseline. However, we were not able to produce the related SDs for those mean change because calculating the SDs requires some more data (e.g. 95% CI, or at least p-value according to the Cochrane Library guidance)]. We used a fixed-effects approach to the meta-analyses. 75 Statistical heterogeneity was quantified using the between-study SD and Tau2 or I²-statistic. The between-study SD gives a direct measure of variance in the treatment effect across studies,76,77 while the I²-statistic is used to quantify the percentage of variation in effect estimates across studies that is due to heterogeneity rather than chance. In other words, it measures the proportion of variance across studies that can be attributed to differences in population characteristics. 78,79

The statistical analyses were conducted within a Bayesian framework using multinma package80 in R software version 4.1.3. 81 We estimated the posterior densities using Markov Chain Monte Carlo (MCMC) simulations; there were four Markov chains with 4000 iterations for each chain. All baseline and intervention effect parameters were given flat (uninformative) normal (0, 1000) priors and the between-study SD flat uniform distributions with an appropriately large range given the scale of measurement. The generalised linear model settings for continuous was a normal link. 71 We assessed the convergence of the Markov chains by using the potential scale reduction factor and examining the history and autocorrelation plots for each estimated parameter. 82

Sensitivity analysis

Based on discussions with our clinical experts, we conducted sensitivity analyses for the mean change in MHDs, and the mean change in MMDs, by excluding the lower doses of eptinezumab (10 and 30 mg), since these doses are currently not available in the UK.

Included studies

Study selection

The PRISMA flow diagram in Figure 1 summarises the results of our searches for the clinical effectiveness review. The electronic searches yielded 18,528 records after the removal of duplicates. Of these, 18,184 citations were excluded at the title and abstract sifting phase. Three hundred and forty-four records were obtained for screening. We found that many articles provided a poor definition of migraine in the abstract. Of these, 293 studies were excluded based on full-text screening. A list of excluded papers and their reasons for excluding them are presented in Report Supplementary Material 1. Seven full-text articles were not available to be sifted, despite an extensive search by the University of Warwick Library Document Supply service. Thus, these seven papers were excluded. We identified 51 articles which described data from 11 trials for the clinical effectiveness review and/or NMA. Although these linked articles were cited, we used the main trial paper for the main citation, as the other linked papers only reported some subgroup analyses, were either repetitive or combined the data.

FIGURE 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram summarising the flow of studies for the clinical effectiveness review.

Narrative synthesis of results by primary outcome(s) of interest

We present the summary of evidence from 11 included RCTs for each outcome of interest narratively.

1. Monthly headache days: Eight trials reported the change in MHDs from baseline. 37,88,89,92–95,110 Two double-blind RCTs evaluating BTA versus placebo in 1384 chronic migraine participants for 24 weeks, followed by a 32-week open label phase in the USA. 92,93 Reduction in headache days from baseline (mean change; 95% CI) in BTA groups in both trials were [9 (−9.69 to −8.31) and 7.8 (−8.5 to −7.1)] while for placebo groups were [6.7 (−7.39 to −6) and 6.4 (−7.11 to −5.69)]. 92,93 The efficacy and safety of BTA 155U every 12 weeks for 3 cycles was assessed in comparison with topiramate ‘immediate release’ 50–100 mg/day in 282 chronic migraine participants for 36 weeks in the open label trial. 88 After week 12, participants initially randomised to topiramate could cross over to BTA group. BTA was significantly superior to topiramate in reduction of headache days at week 32 [8.3 (−9.77 to −6.83) and 2.1 (−3.02 to −1.18), respectively]. 88

   A double-blind trial evaluated the efficacy and safety of topiramate 100 mg (twice daily) with 306 chronic migraine participants in 46 clinics in the USA for 16 weeks. 28,110 Topiramate produced a statistically significant reduction in headache days compared with placebo treatment [least square mean change from baseline (95% CI); 5.8 (−6.69 to −4.91) and 4.7 (−5.59 to −3.81), respectively]. 28,110

   Two double-blind trials comparing the efficacy and safety of different doses of eptinezumab against placebo in the chronic migraine population. 89,94,108 One of the trials was conducted in 128 sites across the USA and Europe with 1072 participants and outcomes were measured at weeks 12 and 24. The reduction in headache days (mean change from baseline and 95% CI) for eptinezumab 100 and 300 mg were 8.2 (−8.8 to −7.6) and 8.8 (−9.44 to −8.16), respectively versus placebo 6.4 (−7.01 to −5.79) at week 12. The reduction in headache days for 100 and 300 mg of eptinezumab at week 24 was 9.6 (−10.27 to −8.91) and 10.6 (−11.3 to −9.88), respectively compared with placebo 8.1 (−8.07 to −7.4). 94,108 Another trial was performed at 92 sites across the USA, Australia, New Zealand and the Republic of Georgia with 558 participants. Treatment duration was measured at 12 weeks. 89 The results for reduction in headache days [mean change from baseline (95% CI)] for 100, 300, 30 and 10 mg of eptinezumab were 8.9 (−10.12 to −7.67), 9.6 (−10.87 to −8.33), 9.2 (−10.35 to −8.05) and 7.5 (−8.72 to −6.28), respectively in comparison with placebo 6.9 (−8.06 to −5.74). 89

   The efficacy and safety of fremanezumab was assessed in 1121 chronic migraine subjects for 12 weeks in 132 sites in 9 countries across the USA and Europe. 37 In this double-blind RCT, participants received fremanezumab quarterly (a single dose of 675 mg at baseline and placebo at weeks 4 and 8), fremanezumab monthly (675 mg at baseline and 225 mg at weeks 4 and 8) or matching placebo. 37 Headache days’ reduction from baseline was measured at 4 weeks after first dose and at 12 weeks. Fremanezumab resulted in a lower frequency of headaches than placebo in this trial. Fremanezumab quarterly reduced mean headache days per month by 4.3 (95% CI −4.89 to −3.71) and fremanezumab monthly decreased mean headache days per month by 4.6 (95% CI −5.18 to −4.01), while the reduction in MHDs for the placebo group was 2.5 (95% CI −3.09 to −1.91). 37

   A double-blind RCT compared the efficacy and safety of two doses of galcanezumab in a sample of 1085 chronic migraine for a period of 12 weeks (3 injections) in 116 headache and clinical research centres across 13 countries. 95 Both doses of galcanezumab were superior to placebo in reducing the number of MHDs. The mean change in headache days from baseline for 120 and 240 mg of Galcanezumab were −4.8 (95% CI −5.58 to −4.01) and −4.6 (95% CI −5.38 to −3.8) compared with placebo −3 (95% CI −4.1 to −1.9). 95

   In summary, eight trials showed that all included medications – the CGRP MAbs (fremanezumab, eptinezumab and galcanezumab), BTA and topiramate were superior to reduction in headache days in comparison with placebo. The headache days’ reduction ranged from 2.5 days for placebo37 to 10.6 days for eptinezumab 300 mg. 108

2. Monthly migraine days: Eleven studies from ten trials investigated MMDs. 28,37,45,89–95,108 Two double-blind RCTs evaluating BTA versus placebo in 1384 chronic migraine participants for 24 weeks followed by a 32-week open label phase in the USA. 92,93 Reduction in migraine days from baseline mean change (95% CI) in BTA groups in both trials were 8.7 (−9.4 to −8) and 7.6 (−8.29 to −6.91), while for placebo groups were 6.3 (−7 to −5.6) and 6.1 (−6.82 to −5.38). 92,93

   • Three trials evaluated the efficacy of fremanezumab. One of them was performed in 1121 chronic migraine subjects for 12 weeks in 132 sites in 9 countries across the USA and Europe. 37 In this double-blind RCT, participants received fremanezumab quarterly (a single dose of 675 mg at baseline and placebo at weeks 4 and 8), fremanezumab monthly (675 mg at baseline and 225 mg at weeks 4 and 8) or matching placebo. 37 MMDs reduction from baseline was measured at 12 weeks. Fremanezumab resulted in a lower frequency of migraine days than placebo in this 12-week trial. Fremanezumab quarterly reduced mean migraine days per month by 4.9 (95% CI −5.68 to −4.12) and fremanezumab monthly decreased mean migraine days per month by 5 (95% CI −5.78 to −4.22), while the reduction in MMDs for the placebo group was 3.2 (95% CI −3.98 to −2.4). 37 The other double-blind RCT which compared the efficacy of fremanezumab was conducted in 104 sites (including hospitals, medical centres, research institutes and group practice clinics) across European countries and the USA. 90 The trial population included both episodic and chronic migraine patients who had documented failure to 2 to 4 classes of migraine preventive medications in the past 10 years, although the results for reduction in MMDs was provided separately for the 837 chronic migraine participants. Fremanezumab quarterly (month 1: 675 mg; months 2 and 3: placebo), fremanezumab monthly (month 1: 675 mg; months 2 and 3: 225 mg) and matched monthly placebo for 12 weeks were administered. 90 Reductions from baseline in mean MMDs over 12 weeks were greater versus placebo; 3.9 (95% CI −4.56 to −3.23) for quarterly, −4.5 (95% CI −5.16 to −3.83) for monthly and 0.7 (−1.35 to −0.04) for placebo. 90 The double-blind trial for evaluating the efficacy and safety of fremanezumab monthly (675 mg at baseline and 225 mg at weeks 4 and 8), fremanezumab quarterly (675 mg at baseline and placebo at weeks 4 and 8) or matching placebo in Japan and Korea was conducted in 571 chronic migraine participants. 91 The change in migraine days from baseline (95% CI) for monthly and quarterly administration were −4.9 (−5.56 to −4.23), −4.1 (−5.07 to −3.12) respectively compared with placebo −2.8 (−3.78 to −1.82). 91

   • A double-blind RCT comparing the efficacy and safety of two doses (120 and 240 mg) of galcanezumab in a sample of 1085 chronic migraine participants for a period of 12 weeks (3 injections) in 116 headache and clinical research centres across 13 countries. 95 The mean change in migraine days from baseline for 120 and 240 mg of galcanezumab were superior [−4.8 (95% CI −5.58 to −4.01) and −4.6 (95% CI −5.38 to −3.8)] compared with placebo [−2.7 (95% CI −3.48 to −1.91)]. 95

   • A double-blind trial compared different doses of erenumab efficacy and safety in 69 headache and clinical research centres in North America and Europe. 667 chronic migraine patients were randomly assigned to be administered monthly 70 mg, 140 mg of erenumab or matched placebo for 12 weeks. 45 The results demonstrated that erenumab 70 and 140 mg reduced the number of MMDs compared with placebo: the mean change from baseline (95% CI) −6.64 (−7.05 to −6.23), −6.63 (−7.04 to −6.22) and −4.18 (−4.51 to −3.85), respectively. 45

   • A double-blind trial evaluated the efficacy and safety of topiramate 100 mg (twice daily) for 306 chronic migraine participants in 46 clinics in the USA for 16 weeks. 28 Topiramate treatment resulted in a mean (95% CI) reduction from baseline of 5.6 (−6.56 to −4.63) migraine days per month compared with 4.1 (−5.07 to −3.13) for the placebo group. 28

   • Two double-blind trials comparing the efficacy and safety of different doses of eptinezumab against placebo in the chronic migraine population. 89,94 One trial was conducted in 128 sites across the USA and Europe with 1072 participants and outcomes were measured at week 12. Treatment with eptinezumab 100 mg [7.7, 95% CI (−8.41 to –6.99)] and 300 mg [8.2, 95% CI (−9.13 to −7.26)] was associated with significant reductions in MMDs across weeks 1 to 12 compared with placebo [5.6, 95% CI (−6.42 to −4.78)]. The MD (95% CI) from placebo for 100 and 300 mg during 24 weeks were −1.98 (−2.94 to −1.01) and −2.65 (−3.62 to −1.68), respectively. 94 The other trial was performed across 92 sites in the USA, Australia, New Zealand and the Republic of Georgia with 558 participants. Treatment duration and time point measurement was 12 weeks. 89 Participants were assigned in eptinezumab 100, 300, 30, 10 mg or placebo, administered as a single IV infusion. The results for reduction in migraine days [mean change from baseline (95% CI)] for 100, 300, 30 and 10 mg of eptinezumab were 7.7 (−8.94 to −6.46), 8.2 (−9.48 to −6.91), 7.9 (−9.06 to −6.74) and 6.7 (−7.9 to −5.5) respectively in comparison with placebo 5.6 (−6.78 to −4.41). 89

   • In summary, 10 trials investigated different doses of CGRP MAbs drugs (including fremanezumab, erenumab, eptinezumab and galcanezumab), BTA and topiramate. These trials illustrated data from different time points in comparison with placebo. The results demonstrated superiority of pharmacological medications versus placebo in the reduction of migraine days from baseline. Migraine days were reduced ranging from 0.7 days for placebo90 to 8.7 days for BTA. 93

Narrative synthesis of results by secondary outcomes of interest

1. Migraine-specific quality of life: Ten studies from five trials used the MSQ questionnaire at multiple time points. 45,92,93,95,97,102,110,111,113,119 A double-blind trial compared the efficacy and safety of two doses of galcanezumab in a sample of 1085 chronic migraine patients for a period of 12 weeks (3 injections) in 116 headache and clinical research centres across 13 countries. 102 At week 12, the least-squares mean change (95% CI) in total MSQ for galcanezumab-treated patients were 20.51 (20.33 to 20.69) (120 mg) and 20.49 (20.31 to 20.67) (240 mg), both statistically significantly greater than the placebo-treated patients 14.55 (14.44 to 14.66). 102 Improvement in all domains of MSQ for both doses were significantly greater than placebo; restrictive role function [120 mg: 21.8 (19.48 to 24.12), 240 mg: 23.1 (20.62 to 25.58) than placebo 16.8 (14.65 to 18.95)], preventative role function [120 mg: 18 (15.69 to 20.32), 240 mg: 16.1 (13.77 to 18.43) than placebo 11 (8.56 to 13.14)], and emotional function [120 mg: 21 (18.3 to 23.7), 240 mg: 20.7 (17.99 to 23.41) than placebo 14.1 (11.62 to 16.58)]. 102

2. A double-blind trial evaluated efficacy and safety of topiramate 100 mg (twice daily) with 306 chronic migraine participants in 46 clinics in the USA for 16 weeks. 111 The MSQ analysis demonstrated significant improvements at week 4 in all three domains, and at weeks 8 and 16 in both restrictive role function and emotional function domains. The preventative role function closely approached, but did not reach statistical significance at week 8. 111 The mean improvement from baseline (95% CI) for topiramate-treated subjects was 23.7 (20.04 to 27.36), 16.1 (12.69 to 19.51) and 26.3 (21.9 to 30.71) for MSQ-RR, MSQ-PR and MSQ-EF, respectively. The mean improvement from baseline (95% CI) for placebo-treated subjects was 18.8 (15.22 to 22.38), 12.6 (9.27 to 15.93) and 21.0 (16.22 to 25.78) for MSQ-RR, MSQ-PR and MSQ-EF, respectively. The differences between treatment groups were statistically significant for MSQ-RR and MSQ-EF but were not statistically significant for MSQ-PR at week 16. 110

   Fremanezumab efficacy and safety was assessed in 1121 chronic migraine subjects for 12 weeks in 132 sites in 9 countries across the USA and Europe. In this double-blind trial, participants received fremanezumab quarterly (a single dose of 675 mg at baseline and placebo at weeks 4 and 8), fremanezumab monthly (675 mg at baseline and 225 mg at weeks 4 and 8) or matching placebo. 113 Fremanezumab quarterly and monthly was associated with significant improvements over placebo in mean change from baseline in MSQ in all 3 domains to week 12. 113 Least square mean change in MSQ-RR (95% CI) from baseline for quarterly and monthly group was 20.3 (18.33 to 22.27) and 21 (18.77 to 23.23), respectively versus 14.7 (12.55 to 16.85) for placebo. Improvement in MSQ-PR for quarterly, monthly and placebo group was 15.9 (14.16 to 17.64), 15.5 (13.79 to 17.21) and 11.6 (9.86 to 13.34), respectively; and improvement in MSQ-EF was 20.9 (18.75 to 23.05), 20.3 (18.53 to 22.07) and 17 (15.08 to 18.92) for quarterly and monthly administration of fremanezumab and placebo. 113

   A double-blind trial evaluating BTA versus placebo in 1384 chronic migraine participants for 24 weeks followed by a 32-week open label phase in the USA92,93,97 found that the improvement in MSQ-RR [mean change from baseline (95% CI)] at week 12 for BTA was 16.2 (13.55 to 18.85) against placebo 9.9 (7.26 to 12.54). For MSQ-PR, the mean change from baseline (95% CI) favoured BTA [13 (10.89 to 15.11)] rather than placebo [13 (12.41 to 13.59)]. MSQ-EF improvement was superior in the BTA group, 18.3 (15.23 to 21.37) rather than placebo 11 (7.95 to 14.05). 97

   A double-blind trial conducted in 69 headache and clinical research centres in North America and Europe randomly assigned 677 chronic migraine patients to be administered monthly 70 or 140 mg of erenumab or matched placebo for 12 weeks. 45,119 Participants in the lower dose (70 mg) of erenumab experienced less improvement in MSQ-RR function [mean change from baseline (95% CI)] than the higher dose (140 mg) participants, 17.7 (14.77 to 20.63) versus 19.1 (16.15 to 22.53), while the mean change from baseline for the placebo group was 11.8 (9.25 to 14.35). The results showed participants in the 70 mg, 140 mg and placebo group had improvement in MSQ-PR function, 13 (10.51 to 15.49), 13.8 (11.31 to 16.29) and 8.9 (6.87 to 10.93), respectively. Improvement in the MSQ-EF for the 70 mg, 140 mg and placebo group was 18.2 (13.15 to 23.24), 18.8 (14.73 to 22.87) and 9.9 (5.98 to 13.82), respectively. 119

   In summary, five trials reported MSQ data for three dimensions separately, including MSQ-RR, MSQ-PR and MSQ-EF. Galcanezumab, erenumab, fremanezumab, topiramate and BTA were investigated in this diverse time window in the included trials. All these drugs were associated with a better improvement in quality of life compared with placebo.

3. The HIT-6: Eleven studies from six trials evaluated headache disability through HIT-6. 37,88,89,91–94,97,100,109,119 Two trials were associated with efficacy and safety of fremanezumab. The first trial which was double blind was conducted in 1121 chronic migraine subjects for 12 weeks in 132 sites in 9 countries across the USA and Europe. The participants received fremanezumab quarterly (a single dose of 675 mg at baseline and placebo at weeks 4 and 8), fremanezumab monthly (675 mg at baseline and 225 mg at weeks 4 and 8) or matching placebo. 37 The degree of headache-related disability decreased between baseline and the 4-week period after the last dose, with significantly greater reductions [mean change from baseline (95% CI)] in HIT-6 scores with quarterly [6.4 (−7.38 to −5.42)] and monthly [6.8 (−7.58 to −6.02)] rather than with placebo [4.5 (−5.48 to −3.52)]. 37 The second trial found a greater reduction with quarterly or monthly administration of Fremanezumab compared with placebo at 4 weeks after the final (third) trial medication administration [4.1 (−4.89 to −3.31), 4.1 (−4.90 to −3.3) and 2.4 (−3.21 to −1.59), respectively]. 91 This double-blind trial assessed 571 participants with chronic migraine who received subcutaneous fremanezumab monthly (675 mg at baseline and 225 mg at weeks 4 and 8), fremanezumab quarterly (675 mg at baseline and placebo at weeks 4 and 8) or matching placebo in Japan and Korea. 91

   • Two double-blind trials evaluated BTA versus placebo in 1384 chronic migraine participants for 24 weeks followed by a 32-week open label phase in the USA. 92,93,97 The pooled results showed a statistically significant and clinically meaningful difference for BTA versus placebo at all time points starting at the first post-treatment study visit (week 4) and including week 24 for the mean change from baseline in total HIT-6 score. 97 Mean change from baseline (95% CI) at week 12 for BTA was −4.7 (−5.58 to −3.82) compared with placebo −2.6 (−3.48 to −1.72). 97 Efficacy and safety of BTA 155U every 12 weeks for 3 cycles was assessed in comparison with topiramate ‘immediate release’ 50–100 mg/day in 282 chronic migraine participants for 36 weeks in the open label trial. 88,109 After week 12, participants initially randomised to topiramate could cross over to BTA group. At week 30, BTA resulted in a mean (95% CI) reduction in HIT-6 scores from baseline of 5.6 (−6.95 to −4.25) compared with 1.3 (−2.01 to −0.59) for topiramate, with a significant between-group difference favouring BTA. 88,109

   • Two double-blind trials compared the efficacy and safety of different doses of eptinezumab against placebo in the chronic migraine population. 89,94 The first trial was conducted in 128 sites across the USA and Europe with 1072 participants, and outcomes were measured at weeks 12 and 24. 94 Patients in the eptinezumab 300 mg group demonstrated a statistically significant improvement on the HIT-6 at week 12, with an estimated MD from placebo. 94 Reduction [mean change from baseline (95% CI)] in HIT-6 for 100 and 300 mg was 6.2 (−6.92 to −5.48) and 7.3 (−8.34 to −6.26) versus placebo 4.5 (−5.27 to −3.73). 94 The second trial was performed at 92 sites across the USA, Australia, New Zealand and the Republic of Georgia with 558 participants. Participants were assigned to eptinezumab 100, 300, 30, 10 mg or placebo, administered as a single IV infusion. 89 The greatest effect of eptinezumab, as measured by the HIT-6, was observed at week 12, with changes in baseline scores of −10.0 (−11.54 to −8.46), −6.9 (−8.24 to −5.56), −6.5 (−7.83 to −5.16) and −6.5 (−7.91 to −5.09) for the 300, 100, 30 and 10 mg groups, respectively, compared with −5.8 for the placebo group. 89

   • A double-blind trial comparing different doses of erenumab in 69 headache and clinical research centres in North America and Europe had 677 chronic migraine patients who were randomly assigned monthly 70 or 140 mg of erenumab or matched placebo for 12 weeks. 45,119 The change from baseline (95% CI) in HIT-6 score was greater in the erenumab groups than in placebo as early as month 1 and this improvement was sustained throughout the trial [70 and 140 mg 5.6 (−6.80 to −4.40) and placebo 3.1 (−4.04 to −2.17)]. 119

   • In brief, six trials aimed to explore the change of disability measured by HIT-6. All pharmacological medications (BTA, fremanezumab, erenumab and eptinezumab) were more effective in the reducing the disabilities score compared with placebo. Reduction in HIT-6 score ranged from 1.3 for placebo88 to 17.4 for BTA. 109

4. EuroQol-5 Dimensions, five-level version (EQ-5D-5L): A double-blind, placebo RCT assessed the effect of treatment with fremanezumab on HRQoL in 1130 participants with chronic migraine. 113 Fremanezumab quarterly (675 mg at baseline, placebo at weeks 4 and 8) or monthly (225 mg at baseline, weeks 4 and 8) led to statistically significant improvements in the EQ-5D-5L visual analogue scale score, compared with placebo. Differences were reported as least-mean squares changes which were 4.6 and 4.8 for fremanezumab quarterly and monthly respectively, compared with 2.2 for placebo.

5. Migraine disability assessment: Three trials reported the MIDAS at different time points. 102,110,119 The first trial reported the MIDAS total score in a study which aimed to assess topiramate for 306 participants. 110 The MIDAS score [mean (95% CI)] decreased from baseline, indicating that improvement was greater in the topiramate group [31.4 (22.87 to 39.92)] compared with the placebo group [21.0 (12.73 to 29.27)]. The second trial evaluated the effect of erenumab in 667 participants with chronic migraine. 119 Reductions from baseline to month 3 in MIDAS total score was greater in the erenumab group compared to the placebo group, indicating better improvement. Respective differences from baseline [least-squares mean (CI)] were −11.9 (−19.3 to −4.4) and −12.2 (−19.7 to −4.8) for erenumab 70 and 140 mg. The final trial assessed galcanezumab in 1117 chronic migraine participants. 102 At week 12, the difference in the least-squares mean (CI) from baseline in the MIDAS total score for galcanezumab indicated a decrease in disability that was significantly greater for the 120 mg dose only [8.74 (−16.4 to −1.1)] and similar for the 240 mg dose [5.49 (−13.1 to 2.1)] compared with placebo.

   In summary, the three trials found that there was improvement in MIDAS score for erenumab, galcanezumab and topiramate in comparison with placebo.

6. Work Productivity and Activity Impairment Questionnaire – Specific Health Problem (WPAI:SHP): An open label trial compared BTA with topiramate 100 mg for ≤ 36 weeks in people with chronic migraine. 109 Overall, 85.7% participants in the BTA group completed the study, while only 19.7% of the participants randomised to topiramate completed their initial treatment. 56.3% of those participants who discontinued topiramate, from week 12 switched to BTA. Work productivity assessed by the WPAI:SHP scores reported at week 36 revealed significant improvements with BTA versus topiramate in work productivity loss [MD: 0.67 (–1.25 to –0.09)] and activity impairment [MD: 1.53 (–2.07 to –1.0)] domains. In summary, this trial found that there was an improvement in work productivity measured by WPAI:SHP which favoured BTA compared to topiramate.

7. Patient Health Questionnaire 9-item (PHQ-9): The same trial that used the WPAI:SHP questionnaire109 also compared BTA with topiramate and reported outcomes for depression at week 36. Improvements in depression were observed via larger changes in PHQ-9 scores with BTA than topiramate [MD: 1.86 (–2.63 to –1.10)]. In summary, BTA led to a better reduction on depression in comparison with topiramate.

8. Functional Impact of Migraine Questionnaire (FIMQ): The open-label trial comparing BTA with topiramate reported FIMQ at week 30. 109 The FIMQ total score showed a greater reduction from baseline with BTA versus topiramate [MD: 11.38 (−16.01 to −6.75)] and also a greater reduction in the following domains: activity impairment [MD: 0.75 (–15.38 to –6.13)]; emotional functioning [MD: 10.81 (–15.76 to –5.86)]; and cognitive functioning [MD: 14.49 (–19.90 to –9.07)]. In brief, BTA had a favourable profile in reduction of activity and functional impairment.

Feasibility of a network meta-analysis

From the eight studies which reported MHDs, seven trials were eligible for inclusion in the NMA. Following guidance from our clinical experts, they recommended that 12 weeks can be used as the measurement time point for the NMA. They also agreed that the 16 weeks measurement time point for topiramate was comparable and can be pooled with the 12 weeks time point. The project team also decided to pool the BTA data which was measured at the 24 weeks time point. However, we excluded the open label trial evaluating BTA efficacy and safety versus topiramate88 for the NMA as the data were reported at 32 weeks. We planned to perform a sensitivity analysis to reflect the effect of the study design (open-label vs. double-blind), but it was not possible because there was insufficient information for MHDs at week 12. The other studies included in the NMA were comparable in terms of participants characteristics, treatment dosing and schedules, baseline risk and observed treatment effects.

For MMDs, 10 studies were eligible for the NMA. Five studies evaluated the change in MSQ score from baseline and were eligible in another separate NMA. Only the 12 weeks time point was included for this NMA and any other time points were excluded. From the seven trials which reported HIT-6 score, six studies were eligible to be included in NMA. We used the same reasoning for excluding the open label trial as we did for MHDs. In summary, we conducted an NMA for those outcomes which were reported in at least three trials.

Network meta-analysis results

We performed a NMA on two primary outcomes: mean change in MHD from baseline, and the mean change in MMDs from baseline.

We also performed NMA on two QoL outcomes: the mean change in MSQ score from baseline for three dimensions – (1) MSQ-RR function; (2) MSQ-PR function; and (3) MSQ-EF and the mean change in HIT-6 score from baseline.

We fitted both fixed and random-effects NMA models based on the model fit indices; we selected the fixed-effects NMA model for all outcomes (see Appendix 3, Tables 24, 28, 32, 36, 40 and 44). We found no indirect evidence in the results, as all trials included in the analysis were placebo-controlled, where no two active treatments were directly compared (Figures 2–19); thus the direct evidence and NMA estimates are the same for each outcome.

FIGURE 2.

Summary of the change in MHD from baseline – network plot. Nodes and edges are weighted according to the number of participants and studies evaluating each treatment and direct comparison, respectively. EPT10 mg, eptinezumab 10 mg IV infusion; EPT30 mg, eptinezumab 30 mg IV infusion; EPT100 mg, eptinezumab 100 mg IV infusion; EPT300 mg, eptinezumab 300 mg IV infusion; FRE-Quarterly, fremanezumab 675 mg (quarterly) SC single dose; FRE-Monthly, fremanezumab 675 + 225 + 225 mg SC; GAL120 mg, galcanezumab 120 mg SC; GAL240 mg, galcanezumab 240 mg SC; TOP100 mg, topiramate 100 mg oral; BTA, onabotulinumtoxinA 155 + 40U SC; IV, intravenous; PBO, placebo; SC, subcutaneous.

FIGURE 3.

Summary of the change in MHD from baseline – forest plot. The forest plot shows the MDs and 95% CrI compared with placebo as reference treatment. MDs lower than 0 indicate favoured results for the intervention.

FIGURE 4.

Summary of the change in MHD from baseline – rankings and SUCRA. Ranking probabilities graph (hatching blue bars) of each treatment and red bars are the SUCRA values for each treatment. The probabilities ranking ranged from 0 to 1. Due to rounding in R software these do not equate to one. SUCRA, surface under the cumulative ranking curve.

FIGURE 5.

Summary of the change in MMD from baseline – network plot. Nodes and edges are weighted according to the number of participants and studies evaluating each treatment and direct comparison, respectively. EPT10 mg, eptinezumab 10 mg IV infusion; EPT30 mg, eptinezumab 30 mg IV infusion; EPT100 mg, eptinezumab 100 mg IV infusion; EPT300 mg, eptinezumab 300 mg IV infusion; ERE70 mg, erenumab 70 mg SC; ERE140 mg, erenumab 140 mg SC; FRE-Quarterly, fremanezumab 675 mg (quarterly) SC single dose; FRE-Monthly, fremanezumab 675 + 225 + 225 mg SC; GAL120 mg, galcanezumab 120 mg SC; GAL240 mg, galcanezumab 240 mg SC; TOP100 mg, topiramate 100 mg oral; BTA, onabotulinumtoxinA 155 + 40U SC; IV, intravenous; PBO, placebo; SC, subcutaneous.

FIGURE 6.

Summary of the change in MMD from baseline – forest plot. The forest plot shows the MDs and 95% CrI compared with placebo as reference treatment. MDs lower than 0 indicate favoured results for the intervention.

FIGURE 7.

Summary of the change in MMD from baseline – rankings and SUCRA. Ranking probabilities graph (hatching blue bars) of each treatment and red bars are the SUCRA values for each treatment. The probabilities ranking ranged from 0 to 1. Due to rounding in R software these do not equate to one. SUCRA, surface under the cumulative ranking curve.

FIGURE 8.

Summary of the change in MSQ-RR from baseline – network plot. Nodes and edges are weighted according to the number of participants and studies evaluating each treatment and direct comparison, respectively.

FIGURE 9.

Summary of the change in MSQ-RR from baseline – forest plot. The forest plot shows the MDs and 95% CrI compared with placebo as reference treatment. MDs lower than 0 indicate favoured results for the intervention.

FIGURE 10.

Summary of the change in MSQ-RR from baseline – rankings and SUCRA. Ranking probabilities graph (hatching blue bars) of each treatment and red bars are the SUCRA values for each treatment. The probabilities ranking ranged from 0 to 1. Due to rounding in R software these do not equate to one. MSQ-RR, migraine-specific quality of life – restrictive role; SUCRA, surface under the cumulative ranking curve.

FIGURE 11.

Summary of the change in MSQ-PR from baseline – network plot. Network plot shows that the nodes and edges are weighted according to the number of participants and studies evaluating each treatment and direct comparison, respectively.

FIGURE 12.

Summary of the change in MSQ-PR from baseline – forest plot. The forest plot shows the MDs and 95% CrI compared with placebo as reference treatment. MDs lower than 0 indicate favoured results for the intervention.

FIGURE 13.

Summary of the change in MSQ-PR from baseline – rankings and SUCRA. Ranking probabilities graph (hatching blue bars) of each treatment and red bars are the SUCRA values for each treatment. The probabilities ranking ranged from 0 to 1. Due to rounding in R software these do not equate to one. MSQ-PR, migraine-specific quality of life – preventative role; SUCRA, surface under the cumulative ranking curve.

FIGURE 14.

Summary of the change in MSQ-EF from baseline – network plot. Nodes and edges are weighted according to the number of participants and studies evaluating each treatment and direct comparison, respectively.

FIGURE 15.

Summary of the change in MSQ-EF from baseline – forest plot. The forest plot shows the MDs and 95% CrI compared with placebo as reference treatment. MDs lower than 0 indicate favoured results for the intervention.

FIGURE 16.

Summary of the change in MSQ-EF from baseline – rankings and SUCRA. Ranking probabilities graph (hatching blue bars) of each treatment and red bars are the SUCRA values for each treatment. The probabilities ranking ranged from 0 to 1. Due to rounding in R software these do not equate to one. MSQ-EF, migraine-specific quality of life – emotional function; SUCRA, surface under the cumulative ranking curve.

FIGURE 17.

Summary of the change in HIT-6 from baseline – network plot. Nodes and edges are weighted according to the number of participants and studies evaluating each treatment and direct comparison, respectively.

FIGURE 18.

Summary of the change in HIT-6 from baseline – forest plot. The forest plot shows the MDs and 95% CrI compared with placebo as reference treatment. MDs lower than 0 indicate favoured results for the intervention.

FIGURE 19.

Summary of the change in HIT-6 from baseline – rankings and SUCRA. Ranking probabilities graph (hatching blue bars) of each treatment and red bars are the SUCRA values for each treatment. The probabilities ranking ranged from 0 to 1. Due to rounding in R software these do not equate to one. HIT-6, Headache Impact Test; SUCRA, surface under the cumulative ranking curve.

Sensitivity analysis results

The results for two sensitivity analyses for the mean change in MHDs and MMDs from baseline, excluding eptinezumab 10 and 30 mg as they are not used in routine practice, are presented below. Further results are presented at Appendix 3, Tables 48 and 49 and Figures 46–49.

Mean change in monthly headache days

The results of the MHD sensitivity analysis suggest that excluding eptinezumab 10 and 30 mg from the base-case analysis changed the observed effect size from −0.02 to +0.02. This has meant some of the ranking of treatments in the SUCRA have changed. However, the top two treatments (eptinezumab 300 mg and fremanezumab monthly) have remained the same. The reordering of the new treatments is provided in Figure 20.

FIGURE 20.

Illustrative sensitivity analysis results for mean change in MHDs from baseline.

Mean change in monthly migraine days

The results of the MMD sensitivity analysis suggest that the observed effect after excluding eptinezumab 10 and 30 mg changed from −0.02 to +0.02. This has meant some of the ranking of treatments in the SUCRA have changed, including for the top two treatments: the SUCRA for eptinezumab 300 mg and fremanezumab monthly switched from 0.77 to 0.73 and 0.70 to 0.73, respectively. The reordering of the new treatments is provided in Figure 21.

FIGURE 21.

Illustrative sensitivity analysis results for mean change in MMDs from baseline.

Risk of bias in included studies

RoB assessments were undertaken using the Cochrane RoB 2 tool for randomised trials. The results of the RoB ratings by trial are summarised across the studies below and are presented in Figure 22 by RoB category. Overall, there were no major concerns that the studies were not applicable to the research question for this assessment.

FIGURE 22.

Risk of bias assessment result.

Certainty of evidence assessment by GRADE approach

Using the GRADE approach, we found that the certainty of evidence for each estimate was judged to be low to high (see Report Supplementary Material 2). Some estimate points were rated down by one level when the studies with high risk of bias in at least one domain contributed to the comparisons. Imprecision of some results were downgraded when the null value (zero for continuous outcomes) lies within the 95% CrI. The summary of GRADE results for outcomes of interest separately is presented in Table 8. In brief, the GRADE assessments indicated a relative certainty of evidence. The robust certainty degree was particularly highlighted for those estimations when drugs were compared with placebo.

Comparison to existing literature

To the best of our knowledge, this study is the first comprehensive NMA for pharmaceutical treatments currently available in the UK for adults with chronic migraine. Our findings for MHDs and MMDs are largely in line with a previous NMA. 72 The authors in this previous review aimed to investigate the effects of CGRP MAbs on 5164 chronic migraineurs in seven randomised trials. 72 Their focus was solely on CGRP MAbs drugs, whereas we took into account all pharmacological medications available in the UK. Our eligibility criteria allowed for the inclusion of not only CGRP MAbs, but also other drugs, such as BTA and topiramate.

In another paper, erenumab was more effective than BTA in the reduction of MMDs,120 which is in line with our results. The effectiveness of different CGRP MAbs for 3052 adult migraine patients with prior treatment failure was investigated in another review. 121 Galcanezumab 240 mg was ranked first in reducing MMDs, followed by fremanezumab monthly and then eptinezumab 300 mg. However, these findings were not in line with ours and it seems that the population with the previous treatment failures may have resulted in this discrepancy. Moreover, erenumab in our findings was ranked as the second best treatment (jointly with fremanezumab monthly) in reduction of MMDs, while in their analyses erenumab was ranked as the least effective treatment for those participants with previous treatment failures. 121 Another review and NMA aimed to assess the effect of CGRP MAbs on disability related to migraine in 7095 adult patients from nine randomised trials. 122 Fremanezumab depicted slightly better improvement in disability compared with other CGRP MAbs at 12 weeks. 122 However, our finding in improvement of MSQ score in all dimensions had the same result. Although we also compared other medications including different doses of fremanezumab, eptinezumab and BTA in our analyses, this provides a comprehensive picture of the effectiveness of different classes of drugs on participants’ QoL. Based on our results, erenumab 140 mg was the most effective treatment in the improvement of MSQ-RR and MSQ-EF but was ranked in fourth place in the effectiveness of the MSQ-PR. Our results illustrate that there are no significant differences between the two doses of erenumab in decreasing disability scores (measured by HIT-6); however, they were the second most effective treatment for MMDs.

Strengths and limitations

The main strength of this analysis is the range of migraine treatment classes including the latest treatments, such as CGRP MAbs, namely fremanezumab, eptinezumab, galcanezumab and erenumab, which are commonly used after other concurrent preventive treatments, such as BTA and topiramate have failed in the UK. This diversity can provide a comprehensive picture of the effectiveness profile of medications for decision-makers to compare migraine treatment alternatives. Therefore, this may better reflect current clinical practice. Another strength of this review is the comprehensiveness of the search strategy which was used. The search was run and updated across a broad range of electronic databases to ensure all relevant trials were included. Furthermore, we did not allow for any date or language restrictions.

However, the results of this analysis should be interpreted with caution due to its limitations. All trials included in the NMA were placebo-controlled; thus, we were not able to estimate any indirect comparisons and, hence, assess the local inconsistency. This means that there were no direct drug-to-drug comparisons in our included trials. We also included a trial which included participants who failed up to four migraine preventive drug classes90 which might result in bias in our results. Finally, we excluded studies with fewer than 100 participants per arm to include better-quality studies and to avoid loss of precision on our NMA by including heterogenous studies;55,56 hence, this excluded all other trials on oral migraine preventatives and restricted the analysis to topiramate, BTA and CGRP MAbs. This may have limited the NMA to more recently investigated treatments where the trial methodology is more precise, and which were undertaken after chronic migraine was introduced as a classification in 2007. Older trials did not separate out chronic migraine from episodic migraine or even define a difference – and including them would have resulted in a large degree of heterogeneity (e.g. between participant baseline characteristics) and results would have been at a high risk of bias and, thus, a NMA would not have been possible. Nevertheless, we might have also missed some important data by excluding these smaller trials. The quality of these older trials may be limited by a variety of factors, such as inadequate sample size, inadequate control groups and outdated methodologies. Conducting newer trials with adequate power, larger sample size and more rigorous designs can help improve our understanding of a treatment’s effectiveness and can help address these limitations of the older trials and provide more reliable and accurate results. Due to the above-mentioned restrictions to the older trials or even newer trials with no efforts to distinguish between migraine subtypes, we believe our results may have less heterogeneity and, subsequently, more precise results.

After completion of the study, we reviewed papers that we had excluded on the size criterion and we identified just one that might have been included. This study randomised 72 people to BTA or amitriptyline. It did not report on MMD, MHD or headache-related QoL. No between-group difference was found in the measures reported. 123 Another trial, that randomised 191 participants from an original target of 250, tested the addition of propranolol to topiramate in people after failure of topiramate monotherapy. This trial was stopped early on the advice of the data safety and monitoring board for futility and provides conclusive evidence that it is not worth using propranolol in this situation. 124 In our protocol design we did not consider the inclusion of trials where additional drugs were added and by default we might have included this, if the trial had reached its recruitment target. Nevertheless, it would not have fitted in our NMA, and its effect estimate does not tell us what the effect of propranolol might be when used as monotherapy. This post hoc review of excluded studies does not indicate that any relevant data have been excluded from our NMA by setting a size criterion for inclusion. The total number of trials for which we eventually extracted data was substantially fewer than we anticipated at the scoping stage because of reporting different aspects of the trials across multiple, sometimes overlapping papers, with 51 individual papers reporting just 11 trials.

In summary, the NMA findings from the included 11 RCTs indicated that pharmacological treatments are more effective than placebo in managing chronic migraine across all outcomes of interest. This review provides supportive evidence for using prophylactic medications to improve both effectiveness and QoL in chronic migraine management. According to our results, some (but not all) MAbs are better than BTA and the remainder roughly equivalent to BTA. Topiramate is worst overall. However, it is important to consider some limitations in the analyses that may affect the certainty of the results, including the lower quality of some of the included trials and the focus on larger-scale trials.

Search strategy

The search strategies for the AEs review were conducted jointly with those for the clinical effectiveness review and have already been reported in the previous chapter (see Chapter 2).

Assessing the relevance and inclusion of studies

Title and abstract screening was conducted by two reviewers (AB, SN). Screening was performed according to PICO criteria (see Box 3 for inclusion criteria and Box 4 for exclusion criteria). At this stage, the abstracts of the retrieved studies were reviewed independently by two out of four reviewers (MU, SN, AA, ND). The full texts of the remaining studies were retrieved, and the same combination of the reviewers conducted an additional round of full-text screening according to the pre-specified inclusion/exclusion criteria.

Data extraction

Data for included studies were extracted by one reviewer (SN) and 20% randomly checked for accuracy by another reviewer (SK). Data extraction forms were developed in Microsoft Excel to capture the following information: ClinicalTrials.gov identifier (NCT number), study name, study characteristics including first author, year, purpose, design, date, setting and country, treatments details, participant demographics, key inclusion and exclusion criteria, AEs and SAEs definition, and information on AEs, TAEs, SAEs and TSAEs.

Assessment of risk of bias for included trials

The Cochrane RoB 2 tool for RCTs58 was applied for assessing the risk of bias of all trials independently by two members (SN, SK). The details of the tool in classifying the risk of bias in the various domains have been provided in Chapter 2.

Data synthesis

Information extracted from the included studies was summarised and tabulated. We applied the CTCAEs v5.0125 to classify the events. In addition, AEs and SAEs were pooled and the proportion of AEs and SAEs for each system organ class (SOC) for each drug was calculated where the original paper used the standard definition for AEs and SAEs.

We reported the adverse and serious adverse events from the rest of the studies (AEs from 11 studies and SAEs from 6 trials) separately, as these studies did not report events according to standard definitions for AEs and SAEs.

Included studies

Study selection

The PRISMA flow diagram in Figure 23 summarises the results of our searches for the AE review. Of the 344 records which were assessed for eligibility, 277 records were excluded at full text. We identified 67 articles which described data from 40 trials22,28,35–37,45,88–95,97,107,108,117,118,126–152 for the AEs review. Although these linked articles were cited, we used the main trial paper for the main citation, as the other linked papers only reported some subgroup analyses, were either repetitive or combined the data.

FIGURE 23.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram summarising the flow of studies for the AEs review.

Risk of bias in included studies

Risk-of-bias ratings by trial are summarised across the studies below and are presented in Figure 24. For this purpose, the Cochrane RoB 2 tool for RCTs58 was applied.

FIGURE 24.

Risk of bias assessment result.

Adverse events results

Of the 40 included trials, 29 trials with 20,694 participants applied a standard definition for AEs which we use in our review. We used SOC to classify and illustrate the proportion of attributed AEs to each drug. A list of classified AEs are presented in Appendix 5, Table 51. Appendix 5, Table 68 shows the percentage of total incidence of any AEs reported for 20 different dosing regimens of 9 drugs. Among them, the most reported AEs belonged to amitriptyline 25–100 mg and galcanezumab 150 mg with 89%133,135 and 72.0%,133 respectively. The lowest number of any AEs is for erenumab 140 mg (33%). 36,45,142–144 Arm level AEs incidence are presented in Appendix 5, Tables 52–67.

Table 10 provides a detailed summary of classified AEs reported in 29 trials. The table illustrates the percentage of attributed AEs of each SOC.

• Investigations: amitriptyline 25–100 mg (14%), atogepant 60 mg (4%), atogepant 30 mg and galcanezumab 240 mg (2%), fremanezumab quarterly and galcanezumab 120 mg (1%).

• Skin and subcutaneous: galcanezumab 150 mg (5%), galcanezumab 240 mg (2%), galcanezumab 120 mg and Fremanezumab monthly (1%).

• Gastrointestinal disorders: amitriptyline 25–100 mg (59%), topiramate 100 mg (27%), galcanezumab 150 mg (14%), atogepant 60 and 10 mg (13%), erenumab 140 mg (12%), atogepant 30 mg (11%), eptinezumab 300, 30 and 10 mg (5%), erenumab 70 mg, galcanezumab 120 and 240 mg and fremanezumab quarterly (4%), eptinezumab 100 mg, erenumab 7 mg, rimegepant 75 mg and placebo (3%), erenumab 21 mg and fremanezumab monthly (2%).

• Ear and labyrinth disorders: topiramate 100 mg (3%), erenumab 140 mg, galcanezumab 120 and 240 mg (1%).

• Eye disorders: onabotulinumtoxinA 150U and galcanezumab 150 mg and Topiramate 100 mg (3%).

• Psychiatric disorders: topiramate 100 mg (13%), erenumab 140 mg (3%), atogepant 60 mg, fremanezumab monthly and quarterly, atogepant 10 mg and OnabotulinumtoxinA 150U (1%).

• Metabolism and nutrition disorders: topiramate 100 mg (9%), amitriptyline 25–100 mg (5%), erenumab 140 mg (1%).

• Vascular disorders: galcanezumab 150 mg (5%).

• Renal and urinary disorders: galcanezumab 120 mg (1%).

• Musculoskeletal and connective tissue disorders: galcanezumab 150 mg (17%), onabotulinumtoxinA 150U (16%), erenumab 7 mg (4%), erenumab 140 mg, galcanezumab 120 and 240 mg, eptinezumab 100 mg and placebo (2%), eptinezumab 300 and 30 mg, erenumab 70 mg, fremanezumab monthly and quarterly (1%).

• Nervous system disorders: topiramate 100 mg (60%), amitriptyline 25–100 mg (44%), eptinezumab 10 mg (10%), erenumab 140 mg (7%), onabotulinumtoxinA 150U, galcanezumab 150 mg and erenumab 7 mg (5%), eptinezumab 30 mg and erenumab 21 mg (4%), atogepant 10 mg, galcanezumab 120 mg and eptinezumab 100 mg (3%), atogepant 30 and 60 mg, eptinezumab 300 mg, fremanezumab monthly, galcanezumab 240 mg, erenumab 70 mg and placebo (2%), fremanezumab quarterly (1%).

• Infection and infestation: galcanezumab 150 mg (26%), amitriptyline 25–100 mg (24%), eptinezumab 300 and 30 mg (19%), atogepant 30 mg and eptinezumab 10 mg (18%), atogepant 60 mg (17%), galcanezumab 120 mg (15%), fremanezumab quarterly (14%), erenumab 70 mg (13%), eptinezumab 100 mg, fremanezumab monthly, galcanezumab 240 mg and placebo (12%), atogepant 10 mg, erenumab 7 and 21 mg (11%), erenumab 140 mg (9%), rimegepant 75 mg (8%), topiramate 100 mg (7%), onabotulinumtoxina 150U (2%).

• General disorders and administration site conditions: fremanezumab monthly (63%), fremanezumab quarterly (61%), galcanezumab 240 mg (32%), galcanezumab 150 mg (26%), amitriptyline 25–100 mg (24%), galcanezumab 120 mg (22%), topiramate 100 mg (16%), placebo (13%), erenumab 140 mg (6%), erenumab 7 and 70 mg (5%), atogepant 60 mg (4%), atogepant 30 mg and onabotulinumtoxina 150U (3%), eptinezumab 30, 100 and 300 mg and erenumab 21 mg (2%), atogepant 10 mg (1%).

• Respiratory, thoracic and mediastinal disorders: topiramate 100 mg (9%), fremanezumab monthly (5%), amitriptyline 25–100 mg (4%), eptinezumab 10 and 30 mg (3%), atogepant 60 mg, eptinezumab 100 and 300 mg, erenumab 7 mg and galcanezumab 240 mg (2%), atogepant 10 and 30 mg, erenumab 21 mg, galcanezumab 120 mg, fremanezumab quarterly and placebo (1%).

Eleven of the forty included trials did not mention what criteria they considered as AEs and just reported the incidence of AEs accrued. Appendix 5, Tables 69 and 75 present the AEs reported in these trials as per publication.

Serious adverse events results

From the 40 RCTs, 30 trials reporting data from 21,529 participants have applied a standard definition for SAEs. These trials evaluated 20 different dosing regimens of 9 drugs. Among them, one study (a series of three sequential, randomised, controlled studies of repeated treatments with BTA for migraine prophylaxis)145 did not explicitly report the number of people with SAEs, however, the results showed that there were no treatment-related SAEs. Thus, SAEs from 29 trials with 20,557 participants were combined. Appendix 6, Table 98 shows the percentage of any SAEs reported for each dosing regimen of these drugs. Table 11 provides more details of classified SAEs and illustrates the percentage of attributed SAEs of each SOC.

From the 40 trials included to assess the safety data, 4 trials have not reported any SAEs data. These four trials evaluated efficacy and safety of topiramate for the treatment of chronic migraine,28 flunarizine versus propranolol in the prophylaxis of migraine,128 amitriptyline versus divalproate in migraine,137 and a comparative study of topiramate versus sodium valproate in the prevention of migraine headaches. 147

Six trials did not provide any definitions used to identify SAEs. We have reported them separately from those trials with a standard definition:

1. A trial evaluated BTA (105–260U) prophylactic treatment of episodic migraine for 369 participants. 132 Only four participants (three in the BTA group and one in the placebo group) experienced four SAEs, of which none were reported by the investigator to be related to study medication.

2. A study of multiple treatments of BTA (75, 150 and 225U) for the prophylaxis of episodic migraine headaches in 495 participants138 reported that seven participants experienced SAEs. No further details were provided.

3. Amitriptyline 25–100 mg in the prophylactic treatment of migraine in 393 participants found that no serious events occurred. 22

4. Efficacy and tolerability in migraine prophylaxis of flunarizine (5 and 10 mg) in comparison with propranolol 160 mg daily were evaluated in 808 participants with episodic migraine. 136 The results depicted one participant in the flunarizine 5 mg group experienced malaise and vertigo. In the flunarizine 10 mg group, five participants reported a SAE: urinary incontinence (n = 1), injury (n = 1), cholelithiasis (n = 1), breast neoplasm (n = 1) and depression (n = 1). In the propranolol group, two participants reported a SAE: one injury and one menstrual disorder.

5. A trial evaluated whether topiramate (100 mg) would prevent the transformation of episodic migraine to chronic daily headache (CDH) in 361 participants with a high-frequency episodic migraine (HFEM). 139 Eight participants (three in the topiramate group and five in the placebo group) reported a total of nine SAEs including spontaneous abortion (x2), bradycardia, bipolar disorder, suicidal thoughts, neuropathy, fractured pelvis secondary to a motor vehicle accident, chest pain and worsening of migraine.

6. A trial assessed the effects of discontinuation of topiramate (200 mg) after a treatment period of 6 months in 512 participants with episodic migraine. 152 Six of the 25 reported SAEs were judged by investigators to be possibly (urinary calculus, dyspnoea, pyrexia and urticaria), probably (depressed mood), or very likely to be (nephrolithiasis) related to the use of topiramate.

List of classified SAEs are presented in Appendix 6, Table 76. Arm level SAE incidence of the 36 trials can be found in Appendix 6, Tables 77–97.

Comparison with previous literature

When comparing with other studies, we have found some review studies that support our findings and a few reviews which may not be aligned with the conclusions we have reached about the AEs and SAEs in this review. Comparisons by each of the drugs are:

• Topiramate 100 mg: One open-label trial was assessed as having a high risk of bias,88 and the rest were considered as having some concerns. 28,135,139,142,147,152 Overall, topiramate was reported as well tolerated with the most common AEs related to the nervous system and gastrointestinal disorders, although erenumab demonstrated a favourable tolerability profile compared to topiramate in a trial. 142 In another crossover trial, 51% of patients discontinued topiramate due to AEs and swapped to the BTA group. 88 Although these trials have been conducted with relatively similar inclusion and exclusion criteria and design, AEs incidences reported for topiramate 100 mg in a trial are meaningfully lower than others. 147 This gap might be justified by the short treatment duration, 4 weeks compared with at least 12 weeks. In another meta-analysis, the safety profile was in favour of the CGRP MAbs, with a higher likelihood to help than to harm compared with topiramate. 153

• BTA at different ranges of doses: The safety profile of BTA at doses between 7.5 and 260 mg was investigated in five RCTs with 2237 subjects with chronic or episodic migraine. Two studies were rated as having a high88 or a low risk of bias,97 and three were rated as having some concerns. 132,138,145 The results showed that lower doses of BTA have a safer AEs profile than higher doses. It is worth mentioning that the lower doses are only prescribed for episodic migraine. Musculoskeletal and connective tissue disorders were the most common AEs for BTA. A pairwise meta-analysis shows that total AEs for BTA was higher than placebo with a relative risk ratio of 1.22 (95% CI 1.07 to 1.14). 154 This is in line with our results which show that BTA has a higher rate of AEs compared with placebo.

• Eptinezumab at different ranges of doses: The safety profile of eptinezumab at doses 10, 30, 100 and 300 mg was investigated in five RCTs with 2696 subjects with chronic or episodic migraine. Two studies were rated as being at low risk of bias,89,94 and three were rated as having some concerns. 131,149,151 However, all doses of eptinezumab were generally reported to be tolerable and acceptable. Eptinezumab 100 mg showed a more desirable AE profile (a smaller proportion of AEs), which may be due the short treatment duration (4 weeks). 149 Hou et al. synthesised results from five trials and found that total AEs in migraine patients with CGRP MAbs therapy were not significantly different from those observed in placebo groups (OR 1.17, 95% CI 0.91 to 1.51). 155 The most common AEs for all doses were depicted for infection and infestation in the SOC. These results align with the other review results, which presented upper respiratory tract infection and urinary tract infection as the frequent AEs. 155 For SAEs, it appears from the data reviewed that a lower dose has a more favourable profile rather than higher doses.

• Erenumab at different ranges of doses: The results of two meta-analyses, one by Lattanzi et al. and another by Zhu et al. , align with our review and concluded that there were no differences in the occurrence of AEs and SAEs between the erenumab and placebo groups. 156,157 In our results for erenumab 21 mg, no SAEs were reported. 130 Our results showed that the least AEs incidence belonged to erenumab 140 mg by SOC, although AEs for gastrointestinal disorders were high. While participants who underwent erenumab 70 mg reported a higher number of infections and infestations, the results were in line with another review. 155 Two erenumab RCTs had some concerns regarding the risk of bias;142,143 the rest were rated as having a low risk of bias. 36,45,130,134,144

• Fremanezumab monthly and quarterly: The safety profile of fremanezumab was investigated in five RCTs with 2514 subjects with chronic or episodic migraine. Four studies were assessed as being at low risk of bias,35,90,91,126 and one had some concerns. 37 There were differences in terms of participants who were included in the trials, that is, subjects with medication overuse, history of failed treatment, and those using preventive migraine medications. AEs incidence in monthly groups was reported to be lower than in quarterly groups.

• Galcanezumab at three doses: The results for evaluating the safety of galcanezumab 120, 150 and 240 mg were reported in seven trials with 2264 participants with chronic or episodic migraine. Of these studies, one trial was rated as being low risk of bias,127 and six trials as having some concerns. 95,133,140,141,146,150 There were differences in eligibility criteria, for example, including participants with a history of documented treatment failure of two to four migraine preventive medications,146 while another trial excluded participants having failed treatment with three or more migraine prevention medications. 141 Overall, galcanezumab for all doses was tolerable and accepted, although it appears from the data reviewed that the AEs incidence in the studies with 12 weeks of treatment occurred in a lower proportion than at 24 weeks. General disorders and administration site conditions, followed by infection and infestations, were the most frequent AEs for all doses. However, for Hou et al. they presented upper respiratory infections and viral infections (infections and infestations) as their most common AEs. 155 This discrepancy may be because they only included one trial.

• Rimegepant 75 mg: The results for rimegepant 75 mg were reported in one trial with 375 participants with episodic migraine which had some concerns in terms of risk of bias, although it showed similar tolerability to placebo, and there were no unexpected or serious safety issue noted. 148,158 Gao et al. included four RCTs (3827 subjects) and their results showed that rimegepant 75 mg had good safety for episodic migraine. Similar to our finding, there was no statistically significant increase in AEs compared with the placebo. 159

• Atogepant at different ranges of doses: The safety profile of atogepant at doses 10, 30 and 60 mg was investigated in a low risk of bias trial with 680 episodic migraine subjects. 129 The AEs for all doses were approximately the same and well tolerated, although atogepant 30 mg had fewer treatment-related AEs incidences. No SAEs were reported for atogepant 30 and 60 mg. A systematic review found that atogepant was well tolerated and had a low frequency of AEs. 160

• Amitriptyline at two doses: The results for evaluating the safety of amitriptyline 50 and 100 mg were reported in three trials with 504 subjects with episodic migraine. An open label trial was assessed as being at high risk of bias,137 and the other two trials as having some concerns. 22,135 AEs experienced the most by the participants were for gastrointestinal disorders followed by nervous system disorders. The results showed that a lower dose of amitriptyline had more AEs. We could not find any evidence for the safety profile of amitriptyline that had been synthesised through systematic review or meta-analysis.

• Divalproate extended release (250–1000 mg) and sodium valproate (200 mg): Two RCTs with 428 episodic migraine subjects. One trial was rated as having a high risk of bias,137 and the other trial was rated as having some concerns. 147 Both were found to be tolerable, which is supported by two reviews. 161,162

• Flunarizine at two doses: Two RCTs with 698 episodic migraine subjects investigated the safety profile of flunarizine at two doses (5 and 10 mg). One trial was rated as having a high risk of bias,128 and the other was rated as having some concerns. 136 Both doses were well tolerated and acceptable. There were no considerable safety differences between the doses. Anker et al. ’s systematic review on flunarizine efficacy and safety for episodic migraine supports our findings. 163

• Propranolol at two doses (40 and 160 mg): The results from two trials with 440 subjects suffering from episodic migraine showed that the lower dose had a more desirable safety profile than the higher dose. One of these trials was rated as having a high risk of bias,128 and the other was rated as having some concerns. 136 Gastrointestinal disorders, general disorders and administration site conditions were reported as the most frequent AEs.

Wang et al. 144 compared the different MAbs against CGRP or its receptor for adult patients with migraine; however, this NMA was limited to direct comparisons between placebo and eptinezumab, erenumab, fremanezumab and galcanezumab, for both AEs and SAEs. 164 The results of this NMA showed that galcanezumab ranked the highest for causing at least one SAE, followed by eptinezumab, erenumab and fremanezumab. However, this result is constrained by massive variations in reported SAEs among the included RCTs. Also, there was no available indirect comparisons between the trials. 164

Strengths and limitations

Almost all the available evidence from the systematic reviews focused on one drug or fewer drugs than what we have considered in this review. Therefore, one of the key strengths of this analysis is the inclusion of a range of migraine treatments, including the latest therapies, such as CGRPs, specifically fremanezumab, eptinezumab, galcanezumab and erenumab, which are commonly used after other concurrent preventive treatments, such as BTA and topiramate have failed. As the inclusion criteria for this chapter comprised both episodic and/or chronic migraine, we have also included some oral medications which were not included in the clinical effectiveness chapter. This diversity provides a comprehensive overview of the medication effectiveness profile, allowing decision-makers to compare alternative treatments and obtain a better reflection of clinical practice. Another main strength of this review is the comprehensiveness of the search strategy employed. The search was conducted and updated across a wide range of electronic databases, without any restrictions on dates or language, to ensure that all relevant trials were included in the analysis.

It is important to note that all the systematic reviews that we have compared to our review have mentioned that there are shortcomings in their included RCTs and that further head-to-head RCTs are required for more robust results for AEs. We recommend further head-to-head RCTs to assess the safety profile of oral medications in the chronic migraine population because our review only found data evaluating the AEs incidence in chronic migraine participants which is limited to newer CGRP treatments, BTA and topiramate. In addition, for some of the considered drugs in our review including amitriptyline, divalproate, sodium valproate and Propranolol, our searches couldn’t identify any relevant systematic reviews to enable us to compare our results. Hence, further studies are needed to have a clear understanding of the safety profile of these drugs.

We should also note that there are further limitations of some of the trials included in this review. For example, the trials which included the drugs atogepant and rimegepant – even though they have product licences, they have not yet been approved by NICE or SIGN; the trial for BTA for episodic migraine patients used non-standard doses, whereas the current standard dose for chronic migraine patients is 155U; and galcanezumab 150 mg dose is not used in standard practice and had a significantly higher AEs profile.

Finally, the results of this analysis should be interpreted with caution due to its limitations. We used CTCAE Version 5.0 for classifying the AEs and SAEs. However, there are some reported AEs and SAEs in the included studies which are not in the CTCAE, thus as a solution for this issue our clinical experts discussed what would be the best respective category for those events. For instance, panic attack was categorised as a psychiatric disorder. The included studies were not consistent in terms of the reporting of AE and SAE definitions, and due to this limitation, our clinical advisers reached a consensus on pooling the results for those studies with the same definitions and reporting the results for the other studies by each study narratively.

In summary, all medications were tolerable, but they had different side effects. Rimegepant had a favourable AEs profile. Amitriptyline and topiramate were associated with a higher occurrence of nervous system disorders and gastrointestinal disorders. Topiramate also was linked to a higher prevalence of psychiatric disorders. All medications had infections and infestations as a side effect. However, the medications did not follow a similar incidence pattern; BTA had the least infection rate, while it had a higher incidence of musculoskeletal and connective tissue disorders than other medications. Participants taking fremanezumab and galcanezumab experienced more general disorders and administration site conditions, while erenumab and eptinezumab had a higher rate of infection and infestation, similar to atogepant.

Search strategy

The search strategy was constructed by an information specialist (AB), in consultation with the project team. MEDLINE and EMBASE searches were based on those used in the searches for Chapters 2 and 3, with the addition of filters for economic and costs studies (instead of a RCTs filter) and search terms for prophylactic drug treatments of migraine in general (as well as specific named interventions). Search strategies in economics/Health Technology Assessment (HTA) specific sources included only terms for migraine/headache, with the addition of general terms for drug treatment or prevention in some cases. No language or date limits were applied. Full search strategies including bibliographic database names and dates searched can be found in Appendix 7, Table 99, along with the targeted internet searches using Google and Google Scholar. Furthermore, Appendix 7 also contains information on the websites of the government agencies which were also searched for publications relating to migraine or headache.

Records retrieved by the database searches were exported into EndNote X9, to enable systematic removal of duplicates. 54 In addition, we did forward and backward citation tracking from included journal articles, using Web of Science Core Collection (and the Citation Finder tool or Google Scholar where articles were not available in Web of Science). Searches were re-run in November 2022 to identify any new studies or publications since the original searches, and we also ran searches to check for any retractions, errata or similar relating to included journal articles. Additional searches for utility data to inform the economic model were also undertaken at this time.

Assessment of eligibility

The citations including title and abstracts were first assessed against the eligibility criteria by two reviewers (HM, SK). Full-text articles meeting the eligibility criteria were then obtained and reviewed. Any disagreements between the reviewers were resolved by discussion or by a third reviewer if necessary (MU). No language restrictions were applied.

Inclusion criteria

Only studies meeting the following inclusion criteria were included:

• Study type: Full economic evaluations in which both the costs and the outcomes of interventions and comparators are examined, including both trial-based and model-based evaluations.

• Population: Adults with chronic migraine, where the headache occurred for 15 or more days/month for more than 3 months.

• Interventions: Prophylactic drugs to treat chronic migraine as listed in Box 1 (see Chapter 2).

• Comparators: Placebo, usual care, or other prophylactic drugs as in Box 1 (see Chapter 2).

• Outcomes: Measures included headache/migraine days, headache-related QoL, MSQ and incremental cost-effectiveness ratios (ICERs) amongst others.

Exclusion criteria

Studies meeting the following exclusion criteria were excluded:

• Partial economic evaluations

• Systematic reviews and/or meta-analyses

• Qualitative studies

• Study protocols

• Conference abstracts

• Editorials and short commentaries

• Articles comparing pharmacotherapy with non-pharmacological interventions.

Data extraction

Using a pre-specified data extraction form, data extraction for full-text studies was carried out by one reviewer (SK) and then checked by a second reviewer (SN). Data extracted included:

• Study context – authors, publication year, country, setting, study population, intervention and comparators.

• Economic evaluation methods – economic evaluation type, model type, study perspective, time horizon, currency and price year, discount rate, resource use/costs, outcome measures and analytical methods.

• Economic evaluation results – study results, sensitivity analyses, generalisability and conclusion.

• Other – funding sources and conflicts of interest.

Data synthesis

Information extracted from the included studies was summarised and tabulated. Findings from individual studies were compared narratively. We summarised the published journal articles separately to the reports, as the latter will not have had a formal peer-reviewed process.

Quality appraisal of economic evaluations

The quality of both the trial-based and model-based economic evaluations was assessed using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) 2022 checklist. 165 The Philips checklist was also used to assess the quality of each model based economic evaluation. 166 Quality assessment was undertaken by one reviewer (SK) and was checked for completeness and accuracy by a second reviewer (SN).

Search results

A total of 7309 citations were retrieved from the database searches. After deduplication and title and abstract review, 88 articles were reviewed at the full-text stage. Nine articles met the inclusion criteria for published peer-reviewed journal articles. 167–175 The excluded studies with their reasons for exclusions are provided in Report Supplementary Material 3. Two articles were translated into English language. 173,175 The targeted internet and website searches identified an additional 72 reports, and after the screening we included 7 of these reports. 176–182 The PRISMA diagram is shown in Figure 25. We have narratively synthesised the reporting of the nine published journal articles separately from the seven reports.

FIGURE 25.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram for cost-effectiveness studies. CADTH, Canadian Agency for Drugs andTechnology in Health.

Journal articles (n = 9)

All journal based studies were model based cost-utility analyses and were from high-income countries including two from the UK. 167,170 Four studies167,169,170,173 evaluated BTA and the remaining five studies evaluated erenumab. 168,171,172,174,175 All studies evaluating BTA compared the treatment with placebo or best supportive care; and studies evaluating erenumab compared their outcomes with placebo or BTA (see Appendix 8, Table 100).

Five studies167,169–171,173 utilised a Markov state-transition model structure (see Appendix 8, Table 101). The Markov models included health states which were stratified by the number of migraine or headache days. Of the other four studies, one used a decision tree,168 and the other three developed hybrid models. 172,174,175 The majority of studies used a 2-year time horizon in the economic models. An NHS perspective was adopted for the UK studies, the European studies were based on the societal perspective and the American studies were based on societal and payers’ perspectives.

The majority of studies included the direct costs of the prophylactic and comparator drugs plus administration costs, and any associated costs for general practitioner (GP) visits, consultant, neurology or specialist nurse appointments and/or accident and emergency department visits (see Appendix 8, Table 101). The study by Vekov and Izmaylov175 only included the costs of the drugs as they assumed all other healthcare costs for both arms were equal. For studies adopting a societal perspective, indirect costs were reported, such as wages lost as a result of not working. Some studies explicitly presented the productivity costs associated with presenteeism and absenteeism. 171,172 Resource use data for three studies167,170,173 were obtained from the International Burden of Migraine Studies (IBMS), and for other studies, resource use was obtained from published studies including trial data or other local databases. For the UK studies,167,170 the NHS reference costs were the main source of cost inputs, and for the non-UK studies the main source of cost inputs were previously published studies, publicly available local databases or published price lists. All nine studies reported the currency and price year in which the unit costs were calculated and reported (see Appendix 8, Table 101).

All studies longer than the 1-year time frame used discounting. The two UK studies in line with the NICE guidelines183 used a 3.5% discount rate for both costs and outcomes. 167,170 All other studies discounted costs and outcomes at a rate of 3%, except one study which used a 5% discount rate. 175 The two UK studies167,170 used a £20,000–30,000 per quality-adjusted life- year (QALY) as the willingness-to-pay (WTP) threshold, whereas the other studies had different WTP thresholds (see Appendix 8, Table 102 for more information).

All studies used the EQ-5D as the main outcome measure to estimate QALYs (see Appendix 8, Table 102). The EQ-5D scores were mapped from the MSQ to produce utility values in six studies. 167–169,171–173 Hollier-Hann et al. 170 administered the EQ-5D-5L as part of the REPOSE trial, and utility values were estimated using a UK tariff. Sussman et al. 174 used EQ-5D-5L scores from the erenumab and BTA trials to estimate utility values; however, they did not state whose values or what tariff was used to estimate them. 174 Other outcomes measures included the number of headache days or migraine days avoided.

All studies167,169,170,173 which used BTA as the treatment found BTA to be more cost-effective than placebo, with ICERs ranging from £15,028 (€17,720) to £16,598 (€19,572). Erenumab was found to be cost-effective in two studies in participants for whom previous preventive treatments had not worked,168,171 and erenumab dominated placebo in two studies. 172,174 When erenumab was compared to BTA from a societal perspective, the ICER was above the most common WTP thresholds £182,128 (€218,870). 168 One study found erenumab not cost-effective compared to placebo. 175 All of the studies conducted deterministic and/or probabilistic sensitivity analysis (PSA). In most cases, the results were sensitive to changes in MMDs, health utilities and treatment costs, but were cost-effective overall – see Appendix 8, Table 101 for more information.

Other reports (n = 7)

Four of the seven reports were from the UK,179–182 two from Canada176,177 and one from the USA178 (see Appendix 8, Table 100). All reports were cost-utility model based analyses. The key interventions assessed for cost-effectiveness were: BTA,176,182 erenumab,177,178,180 fremanezumab178,179 and galcanezumab. 181 All reports compared the intervention to placebo (best supportive care) and four reports also compared the main intervention to BTA. 177,179–181

Two main models were employed (see Appendix 8, Table 101): a Markov model with health states stratified by number of headache days per 28 days or hybrid model with decision tree and Markov model using 12-week cycle lengths. The time horizon ranged from 2 years to a lifetime. The UK-based studies adopted an NHS perspective, and a healthcare payer perspective was adopted for the three North American studies.

All studies reported similar resource usage data. All studies reported the currency and price year for unit costs and also the discount rate (see Appendix 8, Table 101). All the studies used the MSQ scores which were mapped on to the EQ-5D to estimate utility values (see Appendix 8, Table 102). 176–182

All CGRP inhibitors like erenumab, galcanezumab and fremanezumab were found to be cost-effective in the chronic migraine population for whom the previous treatments did not work under the widely accepted WTP thresholds177,179–181 and have been recommended for such group of participants (see Appendix 8, Table 102 for more information). The sensitivity analyses reported in each of the reports was more comprehensive than what was reported in the journal articles.

Generalisability

To assess the level of generalisability, all studies were classified as: (1) generalisable; (2) transferable; and (3) context-specific. Three journal articles167,170,173 were transferable, and the remaining six studies were considered to be context-specific. Two journal articles did not report the source of funding. 168,175 All of the reports were considered to be context-specific and none of them declared any conflicts of interest, although they were all funded by the pharmaceutical industries except one report182 (see Appendix 8, Table 103).

Quality appraisal of economic evaluations

None of the included studies fulfilled all of the quality criteria for the CHEERS 2022 checklist165 (see Appendix 8, Table 104); however, the majority of studies fulfilled a large number of quality criteria. The criteria that were the least well addressed were the items on heterogeneity and generalisability. Most of the studies fulfilled a large number of the quality criteria according to the Phillips checklist. 166 The criteria that were least well addressed were whether the data has been assessed appropriately, the principles of uncertainty, heterogeneity, and assumption about the continuity of treatment and its effect, including sensitivity analysis around the assumption of different alternatives of treatment effect.

Transition probabilities

For the base-case analysis, the transition probabilities were calculated in the following way: for the placebo group, the transition probabilities were calculated by digitising the transition probability image (see Figure 2 in the Batty et al. paper) which showed a visual representation of transition probabilities. 167 These transition probabilities were based on the PREEMPT trial. 92,93

For the different prophylactic medications, transition probabilities were estimated using treatment effect estimates from the NMA (see Figure 2). Since treatment effects were parameterised as MD in MHDs, we derived transition probabilities from these as follows: we assumed that the number of headache days was uniformly distributed across the range covered by each state, and that the mean and variance of the treatment effect was independent of baseline disease severity. This allowed us to derive the post-treatment distribution of MHDs for each state, and therefore the proportion arriving in each state post-treatment. Based on advice from clinical members of the team and following NICE guidance, we also applied a discontinuation rate for each medication. For those who were on BTA we applied a 10% discontinuation rate and for all other medications we applied a 20% discontinuation rate. 39–41

These probability matrices for each prophylactic medication were then multiplied with the placebo’s transition probability matrix mentioned earlier (a two-step transition) to obtain transition probabilities for each individual prophylactic drug. The transition probabilities used for each drug are shown in Appendix 9, Table 105.

Utilities

The utility values for each of the health states were based on the EQ-5D-5L data from a RCT for educational and supportive self-management intervention for people with chronic headaches (CHESS). 12 We categorised participants in the CHESS trial regardless of their treatment group, based on the ‘number of days over the last 4 weeks that they had a migraine/headache’ to obtain MHD health states. The EQ-5D-5L data from the CHESS trial were obtained from the participants at 4 time points (baseline, 4, 8 and 12 months). These EQ-5D-5L responses were converted into health state utilities based on values mapped on to the EQ-5D-3L descriptive system185 using the Hernandez-Alava crosswalk algorithm186 for the baseline time point. In consultation with our clinical colleagues, we assumed that the utility is the same for all prophylactic drugs, but they differ by the MHD health states that the patient is in (Table 12).

Resource use and costs

The costs of the drugs for a 3-monthly cycle were obtained from the BNF. 51 For each of the drugs (except for BTA and eptinezumab as these are administered in hospitals/clinics and topiramate as this is an oral drug) we assumed that the first injection/infusion would be administered by a nurse (30 minutes) and in this appointment they would show the patient how to administer the drug by themself. In the NICE appraisals for these drugs, the manufacturers for these prophylactic drugs assumed that patients would then be able to self-administer these drugs; however, based on the information and guidance provided by NICE we have assumed that 10% of patients would not be able to self-administer and we have included a cost for this in each subsequent cycle. 179,181 For the drugs which are administered in a hospital/clinic setting we have assumed that this would be a 15-minute appointment with a nurse. The hourly cost of the nurse’s time was obtained from the Unit Costs of Health and Social Care 2021. 187 These unit costs are shown in Table 13 and further information is provided in Appendix 9, Table 106. The price year for costs is 2021/22 and any costs not in this financial year were bought in line using the NHS cost inflation index. 187

For each 12-week cycle (regardless of the prophylactic medication), we assumed that there was a cost of care associated for each health state. This included GP visits, accident and emergency (A&E) visits, hospital admissions and triptan use. The frequency of usage for these resource items was obtained from the IBMS for UK patients. 190 In addition, we checked the NICE guidance for the various prophylactic medications and we also included any additional neurology consultant and nursing visits (see Table 13 and Appendix 9, Table 107).

All-cause mortality

Age-specific mortality rates used in the model were based on the UK general population lifetime tables from the Office for National Statistics (ONS). 191 Using the ONS data, the average probability of death for males and females were combined. As the cohort ages, mortality rates generally increase throughout the time horizon in the model.

Base-case analysis: cost-effectiveness results

In the base-case analysis we compared the cost-effectiveness of the different medications for chronic migraine using data from the NMA based on MHDs. The time horizon was 2 years, with a starting age of 30 years for the patient cohort. Costs are in 2021–2 prices and utility was estimated using the Hernandez-Alava crosswalk algorithm. Table 14 shows the deterministic (undiscounted and discounted) and probabilistic (discounted) results. The results are presented in terms of increasing costs.

When comparing each of the medications separately against placebo, the deterministic discounted results showed that topiramate was cheaper (£104 less expensive) and more effective (0.0464 more QALYs) than placebo, therefore topiramate dominated placebo. Each of the other medications, when compared separately, were more expensive than placebo, however, they generated more QALY gains than placebo. In terms of the cost per QALY gained, BTA was more cost-effective than placebo with an ICER of £25,238 per QALY gained. The other five medications (fremanzumab monthly, fremanzumab quarterly, eptinezumab 100 mg, eptinezumab 300 mg and galcanuzmab) when compared with placebo had ICERs which would not be considered cost-effective if using a £20,000–30,000 per QALY threshold. Probabilistic results were in line with deterministic results (see Table 14).

The cost-effectiveness planes for each of the medications versus placebo are shown in Appendix 10, Figures 50–56. For topiramate versus placebo, the ICER points are scattered across the four quadrants, with the majority of points in the bottom two quadrants (indicating that toprimate is cheaper but the effectiveness is varied in terms of topiramate either being less or more effective than placebo). For the other medications versus placebo, the ICER points were in the top two quadrants, indicating that each medication was more expensive than placebo and the effectiveness also varied (being less or more effective).

Figures 57–63 show the CEACs for each medication against placebo. For any amount (up to £50,000 maximum as shown in the graph) that a decision-maker is willing to pay for an additional QALY, topiramate was always 60% more cost-effective than placebo. When comparing BTA with placebo, if a decision-maker is willing to pay anything above £23,700 per QALY, BTA was the more cost-effective option. For the other medications, when comparing with placebo, placebo always remained the most cost-effective option (up to £50,000 maximum a decision-maker is willing to pay as shown in the graph).

Table 15 shows the base-case results when comparing all medications for the 2-year time horizon, ranked by the least costly option. For the discounted deterministic results, topiramate was the least costly option and had slightly more QALY gains than placebo, whereas eptinezumab 300 mg was the more costly option but generated the most QALY gains. Options placebo (dominated by topiramate), fremanezumab quarterly, eptinezumab 100 mg and galcanezumab (all dominated by fremanezumab monthly) were all eliminated as they were dominated by other medications. Then we compared topiramate, BTA, fremanezumab monthly and eptinezumab 300 mg. Fremanezumab monthly was extendedly dominated by a linear combination of BTA and eptinezumab 300 mg and therefore was eliminated. The ICER between BTA and topiramate and the ICER between eptinezumab 300 mg and BTA were not within plausible cost-effectiveness thresholds.

For the discounted probabilistic results, again topiramate was the least costly option and had slightly more QALY gains than placebo, whereas eptinezumab 300 mg was the more costly option but generated the most QALY gains, in line with the deterministic results. Options placebo (dominated by topiramate), fremanezumab quarterly, eptinezumab 100 mg and galcanezumab (all dominated by fremanezumab monthly) were all eliminated as they were dominated by other medications. Then we compared topiramate, BTA, fremanezumab monthly and eptinezumab 300 mg. Fremanezumab monthly was extendedly dominated by a linear combination of BTA and eptinezumab 300 mg and therefore was eliminated. The ICER between BTA and topiramate and the ICER between eptinezumab 300 mg and BTA were not within plausible cost-effectiveness thresholds. This is represented graphically using a CEAF, where topiramate is the most cost-effective medication if the decision-maker is willing to pay up to £50,000 per QALY (Figure 27).

FIGURE 27.

Base-case CEAF.

Ethical approval

We obtained ethical approval for the consensus workshop from the University of Warwick’s Biomedical and Scientific Research Ethics Committee (BSREC 49/22-23).

Design

Nominal Group Technique (NGT) is a method used in health research to enable a large diverse group to generate ideas and make decisions quickly. 196–199 The method facilitates everyone to participate and contribute to the decision-making. The workshop was designed together with our patient and public involvement (PPI) representative who suggested that we add a breakaway session, wherein people with migraine and headache experts meet separately, to share thoughts and reflect on any challenges in the mixed groups. This was an additional way of ensuring that all voices were heard at the workshop. Small group work facilitated discussion between patients and healthcare professionals, moderated by experienced facilitators. Facilitators were not members of the study team, minimising the possibility of influencing the conversation in a particular direction. At the end of the workshop, participants voted anonymously using an online polling software (Vevox). 200

Sample and recruitment

We aimed to recruit around 15 people with chronic migraine and 10 healthcare professionals. People with migraine were approached through the National Migraine Centre (NMC) mailing list. An invitation was sent out by administrators of the NMC containing information about the workshop and a link to an online expression of interest (EoI) form and contact details for the study team. Healthcare professionals were approached directly by the study team using personal networks. We aimed for a mix of specialties and backgrounds, such as neurologists, GPs with a special interest and headache nurses. Clinicians were invited to express an interest by contacting the study team.

Participants

We received 147 EoIs in response to the invitation shared by NMC. Nineteen people were sampled for maximum variation in terms of age, years living with chronic migraine and ethnicity, and were invited to the workshop. Eight people with chronic migraine attended on the day. Fourteen clinicians expressed an interest in response to information circulated by the team to their networks, and all were invited to the workshop. Eleven attended on the day; all of the eight neurologists worked in a secondary care setting and four of these worked solely as headache specialists (tertiary referral headache neurologists). Although we invited more people with migraine than health professionals, on the day the balance of attendees was tipped in favour of health professionals. Demographics of our sample are shown in Table 16.

The workshop

Prior to the workshop, we sent out a summary of the study findings (see Report Supplementary Material 4). The workshop took place online using Microsoft Teams. The workshop began with a presentation to summarise the research and the findings of work packages 1–4. We explained the aim of the workshop, and the scope of research recommendations. Our PPI representative spoke briefly about the importance of equal voice within the small groups. Following this, we split into three breakout groups with around seven participants (approximately four healthcare professionals and three people with migraine).

In the small groups, participants were asked to agree on their top five drug-placebo comparisons, and their top five drug-drug comparisons. They were asked to consider:

• how much evidence we have on the drug

• safety (side effects)

• efficacy (how effective the drugs were found to be in this study)

• feasibility (cost, availability, ease of administration).

We provided a crib sheet reporting the study findings to support the discussion and decision-making (see Report Supplementary Material 4). Participants could suggest comparisons of drugs not included in this study, and they were reminded that they held valid knowledge and perspectives to bring to the discussion.

We then split into two groups: one group with people with migraine and the other group with healthcare professionals, to reflect on the success of/any issues with equal voice in the small group sessions. During a break, the scribes sent their notes from the breakout group sessions to the team. We then held a plenary session to discuss the outcomes of the group discussions, took a break and prepared for the voting (to take place using Vevox, an anonymous polling website). 200 Voting took place in the last part of the workshop, followed by a brief discussion of the results and explanation of the next steps for the team. All attendees were provided with a certificate of attendance (health professionals) or thank you letter and payment (patients) by e-mail after the meeting.

Results

Each group provided ten top comparisons [five drug vs. placebo (Table 17), and five drug vs. drug (Table 18)]. We removed duplicate questions to create the following two lists of top comparisons:

Participants then anonymously voted for their top five choices of the drug versus placebo (Table 19) and drug versus drug (Table 20) comparisons. The results were:

We then combined these 10 and asked participants to rank them in order of priority (Table 21). The results were:

Discussion

The results indicate that comparisons of CGRP MAbs and BTA were a top priority for our group. They also raised the question of whether there might be additive effects of combining these medications, which was not something we anticipated. The effect sizes, in terms of MHDs/MMDs days for each of the drugs we included in our reviews, are at best modest; the largest being 2.76 days for fremanezumab monthly dose. As these drugs work through different pathways, it might be that more substantial effects are possible. Adding together the effects of BTA and a CGRP MAb, assuming no negative interaction, might have a mean effect size of 4–5 days that would be transformative for many people with chronic migraine. Candesartan and flunarizine were the top drugs the group wanted compared against placebo. There was no evidence for these drugs in our clinical and cost-effectiveness study, and the group felt strongly that these were important drugs to study.