Abstract
Background
Rimegepant improves both headache pain and physical function in clinical studies for the acute treatment of migraine. It is not clear whether rimegepant has direct effects on physical function or if improvements in function are an indirect treatment effect mediated through reductions in headache pain. The objective of this study was to estimate the extent to which rimegepant-mediated reductions in headache pain intensity contribute to improvements in physical functioning during a migraine attack.
Methods
Data were pooled from four placebo-controlled trials of rimegepant 75 mg for acute treatment of migraine. Participants rated current headache pain and functional disability on four-point scales prior to dosing and from 15 min to 48 h post-dose. A cross-sectional mediation model including variables for pain, function, and treatment (rimegepant or placebo) was run separately at each timepoint. Based on results of cross-sectional modeling, a longitudinal pseudo-steady-state mediation model was constructed to estimate relationships between variables using all available data during the period when all processes in the model were considered to have achieved and maintained equilibrium. The pseudo-steady-state model assumed relationships between variables were the same at all timepoints.
Results
Across all studies and treatment groups, participants (N = 3975) were mostly female (75.2–89.2%) and mean age was 40.0 to 41.9 years. Cross-sectional models (N = 3441–3834 across timepoints) indicated that the period represented by measurements at 3, 4, 6, and 8 h can be considered a pseudo-steady-state period for a mediation model representing inter-relationships among headache pain, function, and treatment. The pseudo-steady-state model (N = 3281) incorporated all available data from 3, 4, 6, and 8 h post-dose and estimated that 76.1% of the effect of treatment on function is mediated indirectly via improvements in headache pain, with 23.9% representing a direct effect (representing all other effects/mediators not included in the model).
Conclusion
The effect of rimegepant on functioning was largely mediated through reductions in headache pain intensity during the migraine attack. Approximately one-quarter of the improvement in functioning was mediated through factors other than pain intensity.
This is a visual representation of the abstract.
Background
Migraine affects over 1 billion individuals worldwide, with an estimated 1-year period prevalence of 14% to 15%.1,2 The hallmark characteristic of a migraine attack is typically the headache phase, which is usually unilateral, pulsating, and moderate-to-severe in intensity. 3 Other common symptoms include nausea, photophobia, and phonophobia (which may be present in the headache phase as well as other phases of migraine). 3 Migraine headache pain is also typically aggravated by physical activity, leading patients to avoid such activity. 3 A migraine attack can also have pre-headache and post-headache phases that are also associated with symptoms causing negative health impact and disability. 4 The impact of migraine on daily functioning, work productivity, and overall quality of life is substantial, with migraine representing the leading cause of years lived with disability among women aged 15 to 49 years and the second leading cause across all ages and genders.5–7
Headache pain is a key concern among patients with migraine,8–10 and reduction of this pain is a primary endpoint in clinical trials. 11 As a result, recommended acute treatments for migraine include non-specific analgesics (e.g. acetaminophen and non-steroidal anti-inflammatory drugs (NSAIDs)) for attacks of mild pain severity, and migraine-specific agents (e.g. triptans, gepants, ditans) for moderate-to-severe pain. 12 Opioids are also commonly used, although their use is explicitly recommended against by organizations such as the American Headache Society. 13
Rimegepant is an oral small-molecule inhibitor of the calcitonin gene-related peptide (CGRP) receptor and the only agent currently approved for both acute treatment of migraine (with or without aura) and preventive treatment of episodic migraine (migraine with headache on ≤14 days per month) in adults. 14 CGRP plays a key role in the pathogenesis of migraine attacks and headache pain.15,16 Rimegepant has been shown to improve both pain and physical function compared with placebo in phase 3 clinical studies for the acute treatment of migraine.17–21 However, it is not clear whether rimegepant has direct effects on physical function or if improvements in function are an indirect treatment effect mediated through improvements in headache pain. Examining the relationships between treatment, pain, and function is key to understanding the mechanisms through which rimegepant leads to improvement in patient outcomes.
Mediation modeling can be used to explore relationships between different variables, including relationships between treatment and different patient outcomes in clinical trials.22–26 The objective of the current study, therefore, was to apply mediation modeling to estimate the extent to which rimegepant-mediated reductions in headache pain intensity contribute to improvements in physical functioning during a migraine attack.
Methods
Data sources
This was a pooled analysis of data from four previously published, randomized, placebo-controlled trials of rimegepant for the acute treatment of migraine (NCT03235479, NCT03237845, NCT03461757, NCT05399459).17–19,21 Studies NCT03235479, NCT03237845, and NCT03461757 were conducted in the United States, and study NCT05399459 was conducted in Japan. All studies were conducted in compliance with the Declaration of Helsinki and all International Council on Harmonisation Good Clinical Practice Guidelines. Study protocols were reviewed and approved by institutional review boards and/or independent ethics committees. All participants provided written informed consent. This article was written in accordance with STROBE reporting guidelines.
Briefly, each trial included a screening period of 3 to 28 days, an acute treatment period of up to 45 days, and an end-of-treatment visit. In each study, participants were aged ≥18 years and had a ≥1-year history of migraine (with or without aura), two to eight attacks of moderate or severe headache pain intensity per month in the past 3 months, and <15 headache (migraine or non-migraine) days per month during the 3 months before screening. In studies NCT03235479, NCT03237845, and NCT03461757, participants were randomized in a double-blind 1:1 manner to rimegepant 75 mg or matching placebo. In study NCT05399459, participants were randomized in a double-blind 1:1:1 manner to rimegepant 25 mg, rimegepant 75 mg, or matching placebo; only participants from the rimegepant 75 mg and placebo arms from study NCT05399459 were included in the mediation analyses. Participants were provided a single dose of rimegepant 75 mg (conventional oral tablet in two studies, orally disintegrating tablet (ODT) in two studies) or placebo to treat a single migraine attack of moderate or severe headache pain intensity within the next 45 days. The ODT (marketed rimegepant formulation) and conventional tablet formulations are bioequivalent. 27
Upon experiencing a migraine attack with moderate or severe headache pain, and prior to dosing of study medication, participants recorded pain intensity (0 = none, 1 = mild, 2 = moderate, 3 = severe) and level of functional disability (0 = normal, 1 = mildly impaired, 2 = severely impaired, 3 = requires bedrest) in an electronic diary (eDiary). Following administration of study medication, participants used the eDiary to rate pain intensity and functional disability at 15, 30, 45, 60, and 90 min and at 2, 3, 4, 6, 8, 24, and 48 h post-dose. Preventive migraine treatments were permitted during each study if dosing was stable for at least 3 months prior to screening. Other headache treatments were prohibited for 2 h after dosing with study medication. Acute rescue medication (aspirin, ibuprofen, acetaminophen up to 1000 mg/day (2000 mg/day in study NCT05399459), NSAIDs, anti-emetics, or baclofen) was allowed if pain was not relieved at 2 h post-dose or if pain that was relieved at 2 h returned to moderate or severe intensity between 2 and 48 h post-dose. These were the only medications allowed for acute rescue treatment between 2 and 48 h post-dose.
Mediation modeling
Cross-sectional and longitudinal pseudo-steady-state mediation models were used to examine the relationships among treatment, headache pain (simply referred to as pain hereafter), and physical function. The models included observed variables of pain, function, and treatment. Treatment was represented by a binary variable, which was rimegepant (value of 1) versus placebo (value of 0). Pain was evaluated as the mediator of the effect of treatment on function, meaning that treatment leads to improvement in pain and, in turn, having less pain leads to improvement in function. Pain intensity and level of functional disability were measured using the eDiary as described in the previous section. No imputation of missing data was performed. Error terms were included in our mediation models to represent and account for unexplained variance in the dependent variable (function in our models, with error represented by e_func) and the mediator (pain in our models, with error represented by e_pain). Data analysis was performed using SAS v9.4 (SAS Institute, Cary, NC, USA).
First, a cross-sectional mediation model (Figure 1) was run separately at each data collection timepoint from 15 min post-dose to 48 h post-dose. Each cross-sectional model included only participants with data available for all variables at the timepoint being assessed (i.e. participants needed to have treatment, pain, and function data at 15 min post-dose to be included in the analysis at 15 min). The cross-sectional model was used to identify a timeframe when the system was in a pseudo-steady-state, meaning a period when all processes in the model were considered to have achieved and maintained equilibrium. Estimated standardized path loadings were plotted at each timepoint for the path from pain to function and for the path from treatment to pain.
Cross-sectional mediation models. Treatment is a binary variable: rimegepant (value of 1) or placebo (value of 0). Function represents functional disability. Pain represents headache pain severity. e_pain and e_func represent error terms. Curved two-headed arrows pointing to the same variable represent variance.
A longitudinal pseudo-steady-state mediation model (Figure 2) was then constructed to estimate relationships between variables using data from every timepoint when all processes in the cross-sectional model were considered to have achieved and maintained equilibrium. The longitudinal pseudo-steady-state model only included participants with data available for all variables at all timepoints during the pseudo-steady-state period (i.e. if cross-sectional modeling identified the period from 1 to 3 h post-dose as the pseudo-steady-state period, then participants needed to have treatment, pain, and function data available at all timepoints from 1 to 3 h to be included in the longitudinal simulations). As a result, because of (a) missing data and (b) natural attrition of participants in the studies, a smaller number of observations were available in the longitudinal pseudo-steady-state model than the cross-sectional model.
Longitudinal pseudo-steady-state mediation model. Treatment is a binary variable: rimegepant (value of 1) or placebo (value of 0). Function3, 4, 6, and 8 represent functional disability at 3, 4, 6, and 8 h post-dose, respectively. Pain3, 4, 6, and 8 represent headache pain severity at 3, 4, 6, and 8 h post-dose, respectively. e_pain3, 4, 6, 8 and e_func3, 4, 6, 8 represent error terms at the specified times post-dose. Curved two-headed arrows pointing to the same variable represent variance. Curved two-headed arrows pointing to different variable represent covariance.
Additionally, unlike the cross-sectional model, the longitudinal pseudo-steady-state model assumed that relationships between variables were the same at all timepoints. This was done by constraining the corresponding path coefficients to be the same at all timepoints. Further, the longitudinal pseudo-steady-state model does not assume independence between the measurements of pain and function at every timepoint (as cross-sectional models do) by allowing to covary the related error terms (i.e. covariances between error terms associated with pain measurements and covariances between error terms associated with function measurements).
Direct and indirect effects of treatment on function were calculated and were presented as a percentage of the overall effect of treatment on function. The indirect effect is defined as the proportion of the total effect the independent variable (“treatment” in our models) has on the dependent variable (“function” in our models) that can be explained by the mediator variable (“pain”). More simply, it represents how much of the total effect of treatment on function is due to improvements in pain. The indirect effect is represented by dashed lines (treatment → pain → function) in our model schematics. The direct effect can, in principle, be from the treatment itself but, in practice, it typically represents all other effects/mediators not included in the model and is represented as solid lines (treatment → function) in our model schematics.
Results
Participants
Participant (N = 3975) demographics and baseline clinical characteristics were similar in the rimegepant and placebo groups in each of the four studies included in the analysis (Table 1). Across all groups, 75.2% to 89.2% of participants were female, mean age was 40.0 to 41.9 years, 65.0% to 87.8% of participants had primarily migraine without aura, and the mean number of attacks of moderate or severe pain intensity per month in the 3 months prior to screening was 4.3 to 4.8.
Demographics and clinical characteristics of participants in each study.
Rim: rimegepant.
Not collected in study NCT05399459; all participants were from Japan/living in Japan.
Includes Asian, American Indian or Alaska Native, Native Hawaiian or other Pacific Islander, or multiple responses.
Participants, n = 540.
Cross-sectional model
The number of participants included in the cross-sectional model ranged from 3441 to 3834 across timepoints. Figure 3 shows standardized path loadings derived from cross-sectional modeling at every timepoint from 15 min to 48 h post-dose. The path from pain to function (blue line) can be considered stable for the period starting at 3 h post-dose to the end of the observation period (48 h post-dose). This path can be considered to be representative of the general relationship between pain and function and, as such, it is not that dependent on treatment effect. Paths from treatment to pain show a different pattern. Initially (15–30 min post-dose), paths from treatment to pain are very close to zero. Path loadings then began to increase (in absolute value; negative value here shows that treatment reduces pain) and were considered stable from 3 to 8 h post-dose. Standardized path loadings then declined and were again close to zero at 24 and 48 h post-dose.
Estimated standardized path loading (cross-sectional models). The upper blue curve represents the pain-to-function standardized path loading. The lower red curve represents the treatment-to-pain standardized path loading. The time periods where each curve was considered to be stable is represented by a straight red (upper curve) or blue (lower curve) line. NS: not significant
Longitudinal pseudo-steady-state model
The longitudinal pseudo-steady-state model incorporated data only from 3, 4, 6, and 8 h post-dose (the period when cross-sectional modeling was considered to have achieved and maintained equilibrium). The number of participants included in the longitudinal pseudo-steady-state model was 3281. The longitudinal pseudo-steady-state model estimated that 76.1% (P < 0.0001) of the effect of treatment on function is mediated indirectly via improvements in pain and the remaining 23.9% (P < 0.0001) is a direct effect (representing all other effects/mediators not included in the model; Figure 4).
Estimated direct and indirect effects (longitudinal pseudo-steady-state model). Function represents functional disability. Pain represents headache pain severity.
Discussion
Previous studies of rimegepant have shown that acute treatment improves both headache pain and function in patients with migraine.17–21 In the current pooled analysis of data from four trials of rimegepant, mediation modeling analyses in participants with migraine indicated that a large proportion (76%) of the acute treatment effects of rimegepant on physical function can be explained by improvements in pain. A smaller proportion (24%) of the overall effect on function was a direct effect (representing all other effects/mediators not included in the model).
Given the assumed causal relationship between pain and function, the mechanism of action of rimegepant (CGRP receptor antagonist), 14 and the documented role of CGRP signaling in migraine pain pathogenesis, 15 it is not surprising that such a large proportion of rimegepant's effect on function is being mediated through improvements in pain. Any mediators responsible for the direct effect of rimegepant on function observed in this study, accounting for 24% of the total effect, are unknown. As such, discussion regarding any potential mediators is speculative. Nausea, phonophobia, and photophobia are common symptoms of migraine that can occur during and outside of the headache phase. 3 It is possible, therefore, that the direct effects of rimegepant on function observed in this study are due to improvements in one or more of these symptoms. For example, CGRP plays an important role in gastrointestinal function 28 and pre-clinical findings have demonstrated that, in contrast to triptans, use of small molecule CGRP receptor antagonists attenuates motion-induced nausea. 29 Nausea occurs as a pre-monitory symptom (independent of pain and trigeminal activation) in some patients and specific brain regions (including the periaqueductal gray within the brain stem nuclei) are activated during the pre-monitory phase in such individuals. 30 It is possible that CGRP action at these sites underlies migraine-associated nausea. Pre-clinical evidence has also shown that CGRP causes photophobia and phonophobia through both peripheral and central neuronal mechanisms.31,32 CGRP may also play a role in mood, with pre-clinical evidence suggesting that CGRP causes anxiety by depleting dopamine in the dorsal hippocampus via the heterochromatin protein 1 gamma—Krüppel-like factor 11—monoamine oxidase B pathway. 33 Additionally, anti-CGRP antibodies have been shown to reduce symptoms of depression in patients with migraine, independent of migraine reduction. 34 Our study, however, examined the acute use of a single dose of rimegepant and it is unclear whether any effects of CGRP-inhibition on nausea, photophobia, phonophobia, or mood would contribute to improvement in function during the pseudo-steady state period (3–8 h post-dose) identified by our models.
Previous studies have explored the association between pharmacologic treatment, pain, and function in individuals with migraine. For example, patients with migraine who achieved pain freedom at 2 h following acute treatment with lasmiditan were more likely to have other positive outcomes, including freedom from disability, than patients with mild pain at 2 h post-dose. 35 Our study attempted to quantify how much of the observed effect of rimegepant on physical function is indirectly due to improvements in pain. Other studies have attempted to do this in other chronic pain conditions, with similar findings. In participants with osteoarthritis, for example, mediation modeling estimated that at least 75% of the treatment effect of the nerve growth factor inhibitor tanezumab on physical functioning could be attributed to improvements in pain. 26
Mediation models, like all causal models, assume that all processes with the system being modeled achieve and maintain equilibrium (i.e. in a pseudo-steady-state). Our approach initially used cross-sectional modeling to establish the time period that can be considered as the pseudo-steady-state. Cross-sectional models, however, can be biased if processes are happening over time. 36 To address this bias, we also conducted an additional longitudinal model using data from timepoints where the cross-sectional models indicated that the system achieved and appeared to maintain an equilibrium (3, 4, 6, and 8 h post-dose). This longitudinal pseudo-steady-state model incorporates all available data across each of these timepoints and was the model used to estimate direct and indirect effects of treatment on function.
Headache pain is generally considered the hallmark symptom of a migraine attack and a key concern among patients.3,8,9 In our study, pain was the only mediator analyzed to understand the relationship between rimegepant treatment and improvements in function. This was based on the importance of pain in the migraine experience and an assumed causal relationship between improvements in pain and improvements in function. Our findings were consistent with, and support, this assumed causal relationship, as it was estimated that 76% of the overall improvements in function were an indirect result of improvements in pain. Thus, while the results of the models aligned with such a relationship, they do not inherently or necessarily prove that the causal relationship exists (i.e. that rimegepant-mediated improvements in pain cause improvements in function). For example, it is possible that other mediators not evaluated in this study (e.g. nausea, photo- and phonophobia, allodynia, mood, anxiety, stress, sleep) could covary with migraine pain intensity and contribute to improvements in function. Given the large proportion of the treatment effect on function being mediated by pain (76%), it is likely that at least some of the explained variance is due to relationships between pain and other related mediators. It is also likely that adding additional mediators to the model could increase the proportion of explained variance in the treatment effect on functioning. Future research examining relationships among treatment, pain, and function could consider other mediators (in both parallel and serial mediation with pain) to better understand their independent and shared contribution to the overall effect of treatment on function.
Our models used observed data without imputation for missing values. As such, results are conditional on the participants who had pain and function evaluations in a given model. However, the analyses are strengthened by the large sample sizes (N = 3441 to 3834 across timepoints in cross-sectional modeling and N = 3281 in the longitudinal model). Another strength is that all studies included in the model shared similar designs, inclusion and exclusion criteria, and outcome measures. A notable exception is that study NCT05399459 was conducted in Japan and, therefore, all participants were of Asian race; the other three studies were conducted in the United States and had few Asian participants. Study NCT05399459 also included a rimegepant 25 mg treatment arm (not included in the three other studies) that was excluded from the mediation analyses.
Conclusions
Cross-sectional and longitudinal pseudo-steady-state mediation modeling using data pooled from four randomized, placebo-controlled trials in participants with migraine indicated that the effect of rimegepant on functioning was largely mediated through reductions in pain intensity during the acute migraine attack phase. However, approximately one-quarter of the improvement in functioning was mediated through factors other than pain intensity. These findings highlight the key role pain reduction plays in overall migraine treatment response. Other mediators (e.g. nausea, photophobia, phonophobia, cognition) may also play a role in overall response, and their relative contribution could be the focus of future research.
Clinical implications
Mediation modeling indicated that approximately three-quarters of the effect of rimegepant on functioning are mediated through reductions in pain intensity during the acute migraine attack phase.
Approximately one-quarter of the improvement in functioning was mediated through factors other than pain intensity (e.g. nausea, photophobia, phonophobia, cognition) and the relative contribution of these factors could be the focus of future research.
Footnotes
Abbreviations
Acknowledgments
Medical writing support was provided by Matt Soulsby, PhD, CMPP, of Engage Scientific Solutions and was funded by Pfizer.
Ethical considerations
Study protocols were approved by an institutional review board or independent ethics committee for each participating investigational center. Full details can be found in the original study publications.
Consent to participate
All participants provided written informed consent.
Consent for publication
The authors agree to publish with Ceph Reports.
Author contributions
LA: data interpretation, writing and revising manuscript, and approval of manuscript. AGB: study design, data analysis, data interpretation, writing and revising manuscript, and approval of manuscript. DCB: data interpretation, writing and revising manuscript, and approval of manuscript. JCC: study design, data analysis, data interpretation, writing and revising manuscript, and approval of manuscript. DCS: data interpretation, writing and revising manuscript, and approval of manuscript. EKS: data interpretation, writing and revising manuscript, and approval of manuscript.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by Pfizer.
Declaration of conflicting interests
The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: LA, JCC, and DCS are employees of, and hold stock/options in, Pfizer. AGB is a former employee of, and holds stock in, Pfizer. DCB has been a consultant for AbbVie/Allergan, Amgen/Novartis, Eli Lilly, Lundbeck, Pfizer, Theranica, and Teva; serves on the editorial board of Current Pain and Headache Reports; and has received grant support from the FDA and the National Headache Foundation. EKS has consulted or served on advisory boards for Abbvie, Pfizer, and Click Therapeutics.
Data availability statement
Trial registration
ClinicalTrials.gov NCT03235479, NCT03237845, NCT03461757, and NCT05399459.