Naturalistic assessment of patterns and predictors of acute headache medication use among women with comorbid migraine and overweight or obesity

Lay Summary

Obesity may contribute to more severe migraine symptoms and negatively impact migraine treatment outcomes. The present study aimed to understand patterns of acute medication use among 170 women with migraine and obesity who were seeking behavioral migraine treatment. Data were collected in participants’ natural environment using experience sampling methodology, during which participants reported daily migraine symptoms for 4 weeks. Approximately, 30% of attacks were not treated with any medications, and one in five attacks (i.e., 20%) was treated with migraine-specific medication. Participants were more likely to use medication during longer and more severe attacks that started earlier in the day. Participants were also more likely to use migraine-specific medication when attacks were precipitated by an aura and associated with work-related pain interference. Questionnaire-assessed factors were not related to medication use, although older age and higher educational attainment related to more frequent use. In general, these results also suggest that naturalistically assessed factors are more salient correlates of medication use compared to questionnaires. Additional investigation of barriers to medication use is needed among younger individuals and those of lower socioeconomic status.

Abstract

Given the potential for obesity to complicate migraine treatment outcomes, there is a need to understand patterns and correlates of acute medication use among individuals with this comorbidity. Experience sampling methodology (ESM) was used to characterize patterns of acute medication use among those with migraine and overweight/obesity and to examine individual and momentary factors related to medication use (both migraine-specific and nonspecific medications). Women with migraine and overweight/obesity (N = 170) seeking behavioral migraine treatment completed questionnaires followed by 28 days of daily ESM headache diaries. Participants used medications to treat 71.9% of attacks, 20% of which were treated with migraine-specific medications. Participants were more likely to use medication in the context of longer and more severe attacks that started earlier in the day. Presence of aura and greater work-related pain interference uniquely related to migraine-specific medication use. Questionnaire-assessed factors were not related to medication use, although older age and higher educational attainment related to more frequent use. A substantial proportion of attacks were left untreated, suggesting unmet treatment needs in this population. Results also suggest that ESM-assessed factors are more salient correlates of medication use compared to questionnaires. Additional investigation of barriers to medication use is needed.

Implications.

Practice: Mobile technology can be harnessed to monitor medication use patterns among women with comorbid migraine and obesity.

Policy: Policymakers who want to decrease the burden and cost associated with comorbid migraine and obesity should explore potential barriers to medication use in key demographic groups (e.g., younger age groups of lower socioeconomic status).

Research: Future research should be conducted to develop momentary interventions to enhance medication adherence in this population.

Introduction

Migraine and obesity are frequently comorbid, especially in reproductive-aged women. Obesity not only increases migraine complexity but also has the potential to undermine acute headache management [1–5]. For example, obesity is hypothesized to be an exacerbating factor for migraine, leading to more severe and disabling attacks through mechanisms such as heightened inflammation [2]. Given that early and appropriate use of nonspecific (e.g., nonsteroidal anti-inflammatory medications) and migraine-specific (e.g., selective serotonin receptor agonists, i.e., “triptans”) abortive therapies is key to effective management of acute migraine headache [6, 7], there is a need to understand patterns and correlates of abortive medication use in patients with migraine who have comorbid obesity. For these individuals who may already be at heightened risk for more severe attacks, delays in or failing to take medications could lead to even greater disability and suffering.

Utilizing naturalistic assessment methods, such as experience sampling methodology (ESM), offers unique advantages to understand medication use in daily life among people with migraine, both by reducing retrospective recall biases and allowing for the examination of trait-level and time-varying factors related to the occurrence and type of medication use. Doing so may ultimately serve to optimize intervention approaches by tailoring treatments to particular individuals as well as momentary circumstances that are linked to suboptimal medication use. Although prior research has applied ESM in the study of tension-type and migraine headaches [8, 9], only one ESM study has examined migraine medication use, which assessed treatment-seeking adults with migraine [10]. Results indicated that initially dosing with migraine-specific medication when pain was less severe was related to lower daily migraine disability and higher medication satisfaction [10].

Despite the importance of acute medication use in the management of migraine, adherence to medication guidelines tends to be inconsistent, which may be especially problematic among those with co-occurring obesity, who have a more complex clinical profile [11]. Both trait-level markers of migraine-related severity and burden (e.g., impact on usual daily activities, allodynia symptoms, and headache management self-efficacy) and momentary characteristics (e.g., pain intensity) may be relevant predictors of medication use, though no research has examined this. Therefore, the present study used ESM to: (a) characterize patterns of migraine medication use among women with comorbid migraine and overweight/obesity prior to initiation of treatment; (b) examine individual and momentary factors related to the occurrence and type of medication use; and (c) explore factors related to the use of migraine-specific and nonspecific medications.

Methods

Participants

Participants were 170 women with migraine and overweight/obesity recruited to participate in a randomized control trial that examined the efficacy of behavioral weight loss and migraine education on migraine headache frequency and severity (ClinicalTrials.gov identifier NCT01197196). Participants were eligible if they were female; 18–50 years old; experienced migraine as confirmed by a neurologist and in accordance with International Classification of Headache Disorders-3 (ICHD-3) criteria [12]; reported ≥3 migraine attacks and 4–20 headache days during each of the past 3 months; and had a body mass index (BMI) within the overweight to obese range (i.e., 25.0–49.9 kg/m²). Reporting of migraine and tension-type headaches were differentiated based on ICHD-3 criteria. Exclusion criteria were assessed prior to the ESM protocol and included medication overuse headache or headache disorder other than migraine or tension type (assessed during neurological screening); previous bariatric surgery; current participation in a weight loss program, use of prescription weight loss medication, or ≥5% weight loss within ≤6 months; pregnancy, breastfeeding, or plans to become pregnant during the trial; contraindication for weight loss or unsupervised exerciser; cancer diagnosis within 1 year; unable to read/comprehend study materials; any condition that in the opinion of the investigators would undermine adherence to the study protocol (e.g., terminal illness and relocation outside of the geographic region of the research center); and history of substance abuse, eating disorder diagnosis, or other severe psychiatric problem. Full study details have been previously described [13].

Procedure

Participants completed a baseline assessment that included informed consent, migraine diagnosis with a study neurologist, anthropometric measures, self-report questionnaires, and training on the ESM protocol. This was followed by a 28 day ESM protocol, which was administered by a web-based diary application designed by the researchers on a study-provided smartphone. During each day of ESM, participants completed an end-of-day recording (prior to bedtime) that included questions about the occurrence and characteristics of headaches and use of medications to treat headaches. After the ESM protocol, participants returned the smartphone and were randomized to the interventions. For the current study, only the baseline data were utilized. The protocol was approved by the Miriam Hospital Institutional Review Board.

Baseline measures

Anthropometric measurements

Height (centimeters) and weight (kilograms) were measured by using a wall-mounted Harpenden stadiometer (Holtain Ltd., Crymyh, UK) and a calibrated digital scale (Tanita BWB-800; Tanita Corporation of America, Inc., Arlington Heights, IL). BMI was calculated as weight in kilograms/height in meters squared.

Demographic characteristics

Age, race, ethnicity, and level of education were assessed using questionnaires. For the present study, education level was operationalized as a binary variable (i.e., completion/noncompletion of a college/university degree [or higher]).

Headache Impact Test 6 [14]

The Headache Impact Test 6 (HIT-6) is a six-item questionnaire that assesses the impact of headaches on usual daily activities. Responses are rated on a five-point Likert-type scale, with higher scores reflecting more severe headache disability. The HIT-6 has demonstrated good internal consistency and validity [14, 15].

Headache Management Self-Efficacy Scale [16]

The Headache Management Self-Efficacy Scale (HMSE) is a self-report measure of individuals’ confidence that they can take actions to prevent headaches or manage headache-related pain and disability. The HMSE includes 25 items that are rated on a seven-point Likert-type scale, with high scores indicating greater headache-specific self-efficacy. The HMSE has shown adequate reliability and validity [16].

Allodynia Symptom Checklist (ASC) [17]

The ASC is a 12-item questionnaire that assesses allodynia symptoms associated with headaches. Response categories for each symptom are scored as 0 (i.e., “never,” “rarely,” or “does not apply to me”), 1 (“less than half the time”), and 2 (“half the time or more”), with higher total scores indicating more severe allodynia. The ASC has demonstrated adequate validity [17, 18].

ESM measures

At each ESM recording, participants reported whether or not they had a migraine headache that day (Yes/No). If they indicated that a headache occurred, they reported the following characteristics: [1] when the headache began and ended; [2] if it was preceded by an aura (Yes/No); [3] their maximum headache pain intensity over the past 24 hr (on a scale from 0 “no pain” to 10 “pain as bad as you can imagine”); [4] the extent to which headache pain interfered with areas of functioning over the past 24 hr (i.e., general activity, mood, walking ability, normal work, relations with other people, and sleep, enjoyment of life), from 0 “does not interfere” to 10 “completely interferes; [5] the degree to which the headache pain changed with physical activity (1 “worse,” 2 “unchanged,” and 3 “better”); [6] the degree to which they experienced nausea (0 “no” to 3 “severe”); and [7] the degree to which they were bothered by light and sound (0 “no” to 3 “severe”) [19, 20].

If participants reported the occurrence of a headache, they were asked to indicate whether they took medications to treat the headache (Yes/No). If they responded “Yes,” they indicated what medications and how many doses they took. Medications were categorized as nonspecific (i.e., nonsteroidal anti-inflammatory drugs [NSAIDs], simple analgesics [alone or in combination, e.g., acetaminophen], opiates and butalbital-containing analgesics), or migraine-specific (i.e. triptans alone or in combination and other migraine-specific medication [e.g., sumatriptan/naproxen]).

Statistical analyses

Analyses were conducted using available data without imputation in SPSS version 26. Given that some migraine headaches may have persisted longer than 24 hr, multiple daily ESM recordings that corresponded to the same attack were identified based on onset time and then aggregated such that dimensional ratings (e.g., pain interference) were averaged, with the exception of the maximum pain intensity, for which the highest rating during the episode was extracted. The use of medication was based on any endorsement of use across ESM recordings for a given attack.

Aim 1

Descriptive statistics were used to identify the percentage of migraine attacks that were treated with medication; the number of participants who used each type of medication over the 28 days; the percentage of attacks that were treated with each type of medication; and the mean number of doses during attacks.

Aim 2

Generalized estimating equations (GEEs) were used to assess baseline and ESM variables as predictors of medication use when migraine attacks were reported. GEEs are an extension of the generalized linear model that apply a semi-parametric approach to longitudinal analysis. Each GEE employed an AR1 serial autocorrelation to account for the dependence within the nested data and a binary logistic function given the dichotomous nature of the dependent variable (occurrence/nonoccurrence of medication use). First, separate GEEs evaluated demographic characteristics (i.e., age, BMI, race, ethnicity, and education level) as predictors of the likelihood of medication use across the 28 days. Significant predictors were retained as covariates in subsequent analyses. Second, separate GEEs examined baseline variables (i.e., HIT-6, HMSE, and ASC scores) as predictors of medication use. Third, ESM-measured migraine characteristics were examined as predictors. Each ESM variable was assessed separately, with the exception of pain interference scores, which were entered as simultaneous predictors to identify unique contributions of each domain (i.e., general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment of life).

GEE models examining ESM-measured characteristics included within- and between-person effects of these variables. That is, within-person effects reflect the degree to which daily values of an independent variable differed from an individual’s average level of the variable (i.e., person-mean centered), whereas between-person effects reflect the degree to which an individual’s average level of an independent variable (across the ESM protocol) deviated from the sample mean (i.e., grand-mean centered). In addition, time trends were explored such that migraine onset time of day (i.e., coded continuously [0–24 hr] and categorically: morning [12 am–11:59 am], afternoon [12 pm–4:59 pm], evening [5 pm–11:59 pm]) and day of week (i.e., weekend vs. weekday) were assessed as predictors of medication use.

Aim 3

Analyses were analogous to the ESM analyses described in Aim 2, with the exception of the dependent variable. That is, among the migraine attacks that were treated with medication, these GEEs examined the likelihood of using a migraine-specific or nonspecific medication. Given that descriptive analyses indicated a small number of episodes were treated with both medication types (i.e., 7.2%), this category was not included as a separate outcome; therefore, the dependent variable was coded and interpreted such that 1 = use of migraine-specific medication (with or without nonspecific medication) and 0 = use of only nonmigraine-specific medication. Statistical significance was tested in all models as two-tailed p-values, with alpha set at .05.

Results

Table 1 displays descriptive statistics for study variables and medication use. Overall compliance to ESM monitoring was excellent (96% of daily recordings completed). The number of completed recordings was not significantly correlated with participants’ age (r = −.08, p = .285), BMI (r < −.01, p = .967), HIT-6 (r = −.02, p = .834), HMSE (r = .05, p = .522), or ASC scores (r = −.08, p = .333) and did not differ across race (t = 1.14, p = .258), ethnicity (t = 1.00, p = .318), or education level (t = .94, p = .347). Across the sample, migraine attacks were reported on 1,513 (32.1%) of the 4,711 days, with each participant reporting an average of 9.07 ± 4.93 migraine days out of the 28 days of ESM monitoring.

Table 1.

| Demographic and descriptive data

Baseline measures		M	SD	Minimum	Maximum
Age		38.70	8.13	21.00	51.00
BMI		35.36	6.51	24.46	51.85
HIT-6		65.29	4.54	51.00	78.00
HMSE		96.58	23.01	25.00	146.00
ASC		4.98	3.85	0.00	16.00
EMA measures		M	SD	Minimum	Maximum
Maximum pain intensity		5.85	2.34	0.00	10.00
Interference ratings
General		3.38	2.95	0.00	10.00
Mood		3.93	3.00	0.00	10.00
Walking Ability		1.56	2.54	0.00	10.00
Normal work		3.03	3.06	0.00	10.00
Relations with other people		2.87	2.96	0.00	10.00
Sleep		2.73	3.12	0.00	10.00
Enjoyment of life		3.60	3.16	0.00	10.00
Worsening with physical activitya		1.75	0.49	0.00	3.00
Nausea severity		0.63	0.83	0.00	3.00
Photophobia		1.20	0.90	0.00	3.00
Phonophobia		1.12	0.91	0.00	3.00
Duration of migraine attack (hours)		16.03	21.73	0.00	297.17
		Frequency	%
Preceded by an aura	Yes	274	18.1
	No	1236	81.9
Medications used during treated migraine attacks (total N = 1,076)
	Number of uses	Number of doses per use
	(% of treated attacks)b	M	SD	Minimum	Maximum
NSAID	420 (39.0%)	1.47	0.76	1.00	6.00
Analgesic	220 (20.4%)	1.39	0.77	1.00	6.00
Triptan	204 (19.0%)	1.42	0.78	1.00	6.00
Other migraine	16 (1.5%)	1.25	0.45	1.00	2.00
Opiates/butalbital	184 (17.1%)	1.51	0.80	1.00	6.00

analgesic analgesics alone or in combination; ASC Allodynia Symptom Checklist; BMI body mass index; EMA ecological momentary assessment; HIT-6 Headache Impact Test 6; HMSE Headache Management Self-Efficacy Scale; NSAID nonsteroidal anti-inflammatory drugs; opiates/butalbital opiates and butalbital-containing analgesics; SD standard deviation.

^aCoded such that 1 = worse; 2 = unchanged; 3 = better.

^bPercentages do not total 100 given that multiple medications were sometimes used to treat the same migraine attack.

Aim 1

Participants reported using medications to treat their attacks on 71.9% of days that an attack was reported. At the individual level, 60.0% of participants reported they used NSAIDs at least once over the course of the protocol; 37.6% reported using simple analgesics; 31.2% reported using triptans; 27.1% reported using opiates and butalbital-containing analgesics; and 2.4% reported using other migraine-specific medications. On average, each participant reported using simple analgesics on 3.50 ± 2.61 days (range: 1–11), NSAIDs on 4.17 ± 2.81 days (range: 1–15), triptans on 3.87 ± 2.06 days (range 1–8); opiates and butalbital-containing analgesics on 4.0 ± 4.42 days (range: 1–23), and other migraine-specific medications on 4.00 ± 2.58 days (range: 1–7). At the episode level, 64.1% of attacks were treated with only nonspecific medications; 12.8% were treated with only migraine-specific medications; 7.2% were treated with both migraine-specific and nonspecific medications; and the remaining 15.8% were treated with other types of medication or substances. As shown in Table 1, the most commonly used medications to acutely treat attacks were NSAIDs and simple analgesics alone or in combination, followed by triptans, opiates, and butalbital-containing analgesics, and other migraine-specific medications. On average, participants reported taking between one and two doses when they used each type of medication (see Table 1).

Aim 2

Baseline predictors

Table 2 displays results of GEEs examining baseline predictors of overall likelihood of medication use across the ESM protocol. Age (p < .001) and education level (p = .005) were the only demographic characteristics that were significantly associated with medication use such that older individuals and those who completed a college degree (or higher) were generally more likely to use medication to treat attacks compared to younger individuals and those without a college degree. These variables were included as covariates in all subsequent models. The relationship between participant age and education level was also explored to reduce potential collinearity of these covariates in subsequent models. Independent t-tests indicated age was not significantly different between those with and without college (or higher) degrees, t(168) = .50, p = .620. Therefore, each covariate was retained in subsequent models. Other baseline measures (i.e., ASC, HMSE, and HIT-6) were not significantly associated with likelihood of medication, ps = .164 to .507.

Table 2.

| Generalized estimating equations examining baseline predictors of medication use on days during which migraine attacks were reported (N = 1,513)

			95% confidence interval
	B	SE	Lower	Upper	Wald χ 2	p
Intercept	1.32	0.12	1.08	1.56	117.13	<.001
Age	0.06	0.01	0.04	0.09	23.51	<.001
Intercept	1.11	0.13	0.86	1.36	76.29	<.001
BMI	−0.01	0.02	−0.05	0.03	0.19	.661
Intercept	0.87	0.19	0.51	1.23	21.89	<.001
Education a	0.67	0.24	0.21	1.14	8.01	.005
Intercept	1.04	0.23	0.59	1.49	20.72	<.001
Raceb	0.20	0.27	−0.33	0.73	0.55	.459
Intercept	0.90	0.28	0.34	1.46	10.00	.002
Ethnicityc	0.37	0.32	−0.25	0.99	1.35	.245
Intercept	0.99	0.18	0.64	1.35	30.38	<.001
Education	0.65	0.24	0.18	1.12	7.48	.006
Age	0.06	0.01	0.04	0.09	24.44	<.001
ASC	0.05	0.03	−0.02	0.11	1.94	.164
Intercept	1.00	0.18	0.64	1.36	29.56	<.001
Education	0.65	0.25	0.17	1.13	6.97	.008
Age	0.06	0.01	0.04	0.09	24.14	<.001
HMSE	0.00	0.01	−0.01	0.01	0.62	.430
Intercept	1.01	0.18	0.65	1.37	30.75	<.001
Education	0.62	0.24	0.15	1.09	6.65	.010
Age	0.06	0.01	0.04	0.09	24.17	<.001
HIT-6	0.02	0.03	−0.03	0.07	0.44	.507

Bold values indicate p <.05.

ASC Allodynia Symptom Checklist; BMI body mass index; HIT-6 Headache Impact Test 6; HMSE Headache Management Self-Efficacy Scale; SE standard error.

^aCoded as college/university degree completion versus noncompletion (reference category).

^bCoded as White versus non-White (reference category).

^cCoded as non-Hispanic versus Hispanic (reference category).

Time trends

Both time of day and day of week were associated with medication use (Table 3). Attacks starting later in the day were less likely to be treated with medication (p = .013); specifically, those that started in the afternoon (p = .021) or evening (p = .034) were less likely to be treated with medication compared to attacks that began in the morning (i.e., before noon). Medication use was also more likely to occur on weekends compared to weekdays (p = .010).

Table 3.

| Generalized estimating equations examining EMA-measured migraine characteristics as predictors of medication use on days during which migraine attacks were reported (N = 1,513)

			95% confidence interval
Onset time of day (continuous)a	B	SE	Lower	Upper	Wald χ 2	p
Intercept	1.49	0.26	0.99	2.00	33.46	<.001
Education	0.62	0.24	0.15	1.09	6.69	.010
Age	0.06	0.01	0.04	0.09	22.91	<.001
Hours	−0.04	0.02	−0.07	−0.01	6.12	.013
Onset time of day (segmented)b
Intercept	1.22	0.20	0.83	1.61	37.35	<.001
Education	0.61	0.24	0.15	1.08	6.62	.010
Age	0.06	0.01	0.04	0.09	23.05	<.001
Evening	−0.47	0.22	−0.91	−0.04	4.52	.034
Afternoon	−0.39	0.17	−0.72	−0.06	5.36	.021
Onset day of weekc
Intercept	0.94	0.19	0.57	1.30	25.52	<.001
Education	0.61	0.24	0.14	1.08	6.48	.011
Age	0.06	0.01	0.04	0.09	25.36	<.001
Weekend	0.35	0.14	0.08	0.61	6.55	.010
Maximum pain intensity
Intercept	1.04	0.19	0.67	1.42	29.51	<.001
Education	0.85	0.27	0.32	1.38	9.72	.002
Age	0.07	0.01	0.04	0.10	24.00	<.001
Maximum pain (between)	0.22	0.08	0.06	0.38	7.53	.006
Maximum pain (within)	0.41	0.05	0.31	0.50	73.30	<.001
Pain interference
Intercept	1.13	0.19	0.76	1.50	35.71	<.001
Education	0.75	0.25	0.27	1.23	9.31	.002
Age	0.05	0.01	0.02	0.08	12.71	<.001
General (between)	0.23	0.19	−0.14	0.59	1.52	.218
Mood (between)	−0.48	0.17	−0.81	−0.14	7.91	.005
Walking (between)	−0.12	0.12	−0.35	0.11	1.02	.312
Work (between)	0.27	0.18	−0.07	0.62	2.38	.123
Relationship (between)	0.01	0.15	−0.27	0.30	0.01	.924
Sleep (between)	−0.04	0.09	−0.21	0.13	0.24	.627
Enjoyment (between)	0.26	0.16	−0.05	0.57	2.78	.095
General (within)	0.17	0.06	0.05	0.29	7.37	.007
Mood (within)	0.14	0.06	0.01	0.27	4.77	.029
Walking (within)	0.06	0.08	−0.09	0.21	0.66	.418
Work (within)	−0.02	0.06	−0.13	0.10	0.08	.771
Relationship (within)	−0.02	0.09	−0.19	0.15	0.05	.831
Sleep (within)	0.02	0.04	−0.05	0.10	0.32	.573
Enjoyment (within)	0.04	0.08	−0.12	0.20	0.24	.627
Worsening with activityd	B	SE	Lower	Upper	Wald χ 2	p
Intercept	1.03	0.19	0.66	1.40	29.68	<.001
Education	0.63	0.25	0.14	1.11	6.41	.011
Age	0.06	0.01	0.04	0.09	24.29	<.001
Activity (between)	−0.16	0.42	−0.99	0.67	0.14	.704
Activity (within)	−0.84	0.24	−1.32	−0.36	11.73	.001
Nausea severity
Intercept	1.03	0.19	0.65	1.40	29.15	<.001
Education	0.79	0.25	0.31	1.28	10.43	.001
Age	0.06	0.01	0.04	0.09	22.72	<.001
Nausea (between)	0.65	0.25	0.15	1.15	6.57	.010
Nausea (within)	0.87	0.18	0.52	1.21	24.22	<.001
Photophobia
Intercept	1.04	0.19	0.66	1.42	28.44	<.001
Education	0.72	0.26	0.22	1.22	7.95	.005
Age	0.07	0.01	0.04	0.09	23.34	<.001
Photophobia (between)	0.27	0.22	−0.17	0.71	1.44	.230
Photophobia (within)	0.86	0.13	0.61	1.12	44.80	<.001
Phonophobia
Intercept	1.03	0.19	0.66	1.40	29.44	<.001
Education	0.71	0.25	0.21	1.20	7.83	.005
Age	0.06	0.01	0.04	0.09	23.11	<.001
Phonophobia (between)	0.27	0.22	−0.17	0.70	1.45	.229
Phonophobia (within)	0.78	0.14	0.52	1.05	32.87	<.001
Preceded by an aura
Intercept	1.02	0.18	0.67	1.38	31.73	<.001
Education	0.62	0.24	0.15	1.09	6.77	.009
Age	0.06	0.01	0.04	0.09	24.90	<.001
Aura (between)	0.37	0.34	−0.29	1.03	1.23	.268
Aura (within)	0.56	0.30	−0.03	1.14	3.50	.061
Duration of migraine (hours)
Intercept	0.87	0.19	0.49	1.25	20.24	<.001
Education	0.66	0.25	0.17	1.14	7.08	.008
Age	0.06	0.01	0.03	0.09	20.24	<.001
Duration (between)	0.01	0.01	−0.01	0.03	1.09	.296
Duration (within)	0.02	0.01	0.01	0.04	11.41	.001

Bold values indicate p <.05.

“Between” reflects grand-mean-centered variables (i.e., individual differences); “within” reflects person-mean-centered variables (i.e., daily levels).

EMA ecological momentary assessment; SE standard error.

^aTime of day (continuous) was coded as 0–24 hr.

^bTime of day (segmented) was coded such that morning was the reference category. Follow-up analyses with alternate reference groups indicated that afternoon and evening onset did not differ significantly with respect to likelihood of medication use (p = .683).

^cDay of week was coded such that weekdays was the reference category.

^dLower values = worsening symptoms with activity.

Migraine characteristics

There were significant within-person (i.e., daily) effects with respect to maximum pain intensity (p < .001), general interference (p = .007), mood interference (p = .029), activity-related worsening of symptoms (p < .001), nausea severity (p < .001), photophobia (p < .001), phonophobia (p < .001), and duration (p = .001) as predictors of medication use. Participants were more likely to use medication on migraine days that were characterized by higher maximum pain intensity, higher general and mood-related interference, worsening symptoms with physical activity, more severe nausea, phonophobia, photophobia, and longer migraine duration (relative to their usual levels of these variables reported during attacks). The occurrence of an aura was not significantly related to medication use (p = .061).

Between-person effects emerged for pain intensity (p = .006), mood interference (p = .005), and nausea severity (p = .010) such that participants who reported higher pain intensity and more severe nausea symptoms yet lower interference with mood (relative to other participants in the sample) were more likely to use medication across the 28 days.

Aim 3

Time trends

Table 4 shows GEE analyses examining predictors of the use of migraine-specific (with or without nonspecific medication) compared to only nonspecific medications during treated attacks. Similar to Aim 2 findings, attacks starting later in the day were less likely to be treated with migraine-specific medication (p = .008). Specifically, those that started in the evening (p = .035) were less likely to be treated with migraine-specific medication compared to attacks that began in the morning (i.e., before noon). However, the use of migraine-specific versus only nonspecific medications did not differ between weekdays and weekends (p = .147).

Table 4.

| Generalized estimating equations examining EMA-measured migraine characteristics as predictors of migraine-specific medication use on days during which migraine attacks were treated with medication (N = 1,076)

			95% confidence interval
Onset time of day (continuous)a	B	SE	Lower	Upper	Wald χ 2	p
Intercept	−1.09	0.30	−1.68	−0.51	13.33	<.001
Education	0.23	0.33	−0.41	0.87	0.50	.481
Age	0.03	0.02	−0.01	0.07	1.88	.171
Hours	−0.04	0.02	−0.08	−0.01	6.99	.008
Onset time of day (segmented)b
Intercept	−1.38	0.26	−1.88	−0.88	29.21	<.001
Education	0.21	0.32	−0.42	0.85	0.43	.511
Age	0.03	0.02	−0.01	0.07	2.02	.155
Evening	−0.47	0.22	−0.91	−0.03	4.46	.035
Afternoon	−0.43	0.22	−0.85	0.00	3.86	.050
Onset day of weekc
Intercept	−1.58	0.26	−2.09	−1.08	37.60	<.001
Education	0.18	0.33	−0.46	0.82	0.31	.576
Age	0.03	0.02	−0.01	0.07	2.43	.119
Weekend	0.17	0.11	−0.06	0.39	2.11	.147
Maximum pain intensity
Intercept	−1.67	0.28	−2.22	−1.12	35.15	<.001
Education	0.21	0.36	−0.50	0.92	0.34	.560
Age	0.03	0.02	−0.01	0.08	2.42	.120
Maximum pain (between)	−0.08	0.12	−0.31	0.16	0.41	.523
Maximum pain (within)	0.29	0.07	0.16	0.42	18.58	<.001
Pain interference
Intercept	−1.70	0.29	−2.26	−1.13	34.99	<.001
Education	0.20	0.37	−0.54	0.93	0.28	.600
Age	0.03	0.02	−0.01	0.08	2.04	.154
General (between)	−0.08	0.28	−0.63	0.47	0.08	.774
Mood (between)	0.01	0.20	−0.38	0.40	0.00	.956
Walking (between)	−0.05	0.13	−0.30	0.20	0.15	.698
Work (between)	0.19	0.26	−0.32	0.70	0.54	.462
Relationship (between)	−0.24	0.21	−0.66	0.18	1.29	.256
Sleep (between)	−0.12	0.12	−0.35	0.12	0.97	.325
Enjoyment (between)	0.19	0.21	−0.21	0.60	0.88	.349
General (within)	0.04	0.05	−0.05	0.13	0.63	.427
Mood (within)	−0.03	0.05	−0.13	0.06	0.46	.496
Walking (within)	−0.01	0.05	−0.11	0.08	0.09	.770
Work (within)	0.11	0.04	0.02	0.20	6.31	.012
Relationship (within)	0.06	0.03	−0.01	0.12	2.77	.096
Sleep (within)	0.06	0.03	<−0.01	0.11	3.67	.055
Enjoyment (within)	0.05	0.06	−0.06	0.16	0.69	.407
Worsening with activityd	B	SE	Lower	Upper	Wald χ 2	p
Intercept	−1.58	0.26	−2.09	−1.06	36.28	<.001
Education	0.21	0.33	−0.44	0.85	0.39	.533
Age	0.03	0.02	−0.01	0.07	2.44	.118
Activity (between)	−0.34	0.52	−1.36	0.68	0.42	.516
Activity (within)	−0.44	0.20	−0.82	−0.06	5.06	.024
Nausea severity
Intercept	−1.71	0.28	−2.25	−1.17	38.69	<.001
Education	0.36	0.34	−0.31	1.02	1.11	.292
Age	0.03	0.02	−0.01	0.07	2.04	.153
Nausea (between)	0.44	0.27	−0.09	0.96	2.69	.101
Nausea (within)	0.55	0.11	0.33	0.77	23.39	<.001
Photophobia
Intercept	−1.73	0.29	−2.30	−1.16	35.27	<.001
Education	0.32	0.35	−0.37	1.01	0.82	.365
Age	0.03	0.02	−0.01	0.08	2.39	.122
Photophobia (between)	0.07	0.27	−0.46	0.60	0.07	.797
Photophobia (within)	0.77	0.13	0.52	1.02	35.70	<.001
Phonophobia
Intercept	−1.65	0.27	−2.18	−1.11	36.70	<.001
Education	0.17	0.35	−0.52	0.85	0.23	.631
Age	0.04	0.02	<−0.01	0.08	3.57	.059
Phonophobia (between)	−0.41	0.28	−0.96	0.14	2.10	.147
Phonophobia (within)	0.65	0.13	0.41	0.90	27.09	<.001
Preceded by an aura
Intercept	−1.56	0.25	−2.06	−1.06	37.83	<.001
Education	0.21	0.33	−0.43	0.84	0.40	.528
Age	0.03	0.02	−0.01	0.07	2.44	.118
Aura (between)	−0.10	0.52	−1.12	0.92	0.04	.849
Aura (within)	0.64	0.22	0.22	1.07	8.79	.003
Duration of migraine (hours)
Intercept	−1.44	0.26	−1.95	−0.92	29.85	<.001
Education	0.03	0.34	−0.63	0.69	0.01	.929
Age	0.03	0.02	−0.01	0.08	2.67	.102
Duration (between)	−0.01	0.01	−0.04	0.01	0.99	.321
Duration (within)	0.01	<0.01	<−0.01	0.01	2.75	.097

Bold values indicate p <.05.

“Between” reflects grand-mean-centered variables (i.e., individual differences); “within” reflects person-mean-centered variables (i.e., daily levels).

EMA ecological momentary assessment; SE standard error.

^aTime of day (continuous) was coded as 0–24 hr.

^bTime of day (segmented) was coded such that morning was the reference category. Follow-up analyses with alternate reference groups indicated that afternoon and evening onset did not differ significantly with respect to the likelihood of medication use (p = .835).

^cDay of week was coded such that weekdays was the reference category.

^dLower values = worsening symptoms with activity.

Migraine characteristics

With respect to within-person effects, maximum pain intensity (p < .001), work interference (p = .012), activity-related worsening of symptoms (p = .024), nausea severity (p < .001), photophobia (p < .001), phonophobia (p < .001), and precipitating aura (p = .003) were predictors of migraine-specific medication use. Participants were more likely to use migraine-specific medication during attacks that were characterized by higher maximum pain intensity, work-related pain interference, worsening symptoms with physical activity, more severe nausea, phonophobia, photophobia, and when there was a preceding aura (relative to their usual levels of these variables). The within-person effect of attack duration was not significant (p = .097); there were no significant between-person effects.

Discussion

The present study characterized patterns of medication use among treatment-seeking women with migraine and co-occurring overweight/obesity using near real-time data collection. Approximately 30% of attacks were not treated with any medications, which may reflect unmet treatment needs or barriers to medication use in this population. Of the treated attacks, the majority (i.e., 64%) were treated only with nonspecific medications, particularly NSAIDs and simple analgesics alone or in combination. One in five attacks (i.e., 20%) was treated with migraine-specific medication (alone or in combination with nonspecific medication). Taken together, such data are in line with findings of prior studies that have reported that a substantial proportion of individuals with migraine remain undertreated and do not adhere to recommended guidelines regarding acute migraine-specific medications, which include taking medication-specific medication early during the episode when the pain is less severe [21, 22]. Participants who were older and had higher educational attainment were also generally more likely to use medication, though baseline measures of headache impact, disability, and self-efficacy were not related to medication use.

In contrast, ESM data revealed both individual-level and momentary predictors of medication use and type, which further supports the clinical utility of this methodology to understand medication adherence in this population. In particular, ESM results revealed that participants were more likely to use medications (either migraine-specific or nonspecific) in the context of longer and more severe attacks, particularly when they involved higher than usual pain intensity, general and mood-related interference, worsening symptoms with physical activity, nausea, phonophobia, and photophobia. Similarly, participants were more likely to take migraine-specific medication (relative to use only nonspecific medication) during episodes characterized by higher than usual pain intensity, worsening symptoms with physical activity, nausea, phonophobia, and photophobia. Experiencing a precipitating aura and greater than usual work-related interference emerged as unique predictors of migraine-specific medication use. At the individual level, overall medication use was greater among those who had higher pain intensity and more severe nausea symptoms across the ESM monitoring period. Taken together, these results are consistent with prior cross-sectional research demonstrating that more frequent migraine medication use is associated with a more severe clinical profile [22, 23]. Importantly, our findings add to this literature by showing how daily fluctuations (i.e., within-person effects) in symptoms can predict when medication use will occur. Furthermore, daily symptom fluctuations were generally more predictive of medication use compared to individual difference factors, as reflected by limited between-person ESM effects and no effects of baseline questionnaires.

Interestingly, overall medication use was greater among individuals with lower mood-related interference, which contrasts with the episode-level effect. Individuals who generally have less mood-related interference may be more likely to exhibit proactive coping and self-efficacy to manage migraine attacks with medication when they occur. While this speculative interpretation is supported by previous research on medication adherence in other conditions, confirmation is warranted among individuals with migraine [24, 25].

The use of ESM also allowed for nuanced understanding of time trends related to medication use. It is notable that individuals were more likely to treat attacks with medication, and particularly with migraine-specific medication, when attacks began earlier in the day. This is encouraging given that this pattern of use is consistent with existing recommendations for acute migraine treatment—that is, taking migraine-specific medications first and treating attacks when pain is less severe [6, 26]. However, attacks that began on weekends versus weekdays were more likely to be treated with medication (either migraine specific or nonspecific), which could reflect a barrier related to anticipated side effects of some migraine-specific medications interfering with the workday [27, 28].

The finding that age and education level (a recommended index of socioeconomic status [29]); were related to medication use is consistent with an earlier study that found the majority of individuals who never used acute prescription medication were younger adults (i.e., mid-30s and younger) and that higher household income was related to more frequent medication use [22]. Furthermore, lower socioeconomic status and younger age has been associated with premature dropout from headache treatment [30]. These factors may reflect decreased access to and utilization of health care and/or lower health literacy among younger adults and those of lower socioeconomic status. Thus, potential barriers to medication use will be important to explore and address among these individuals.

Despite strengths of the current study (e.g., use of real-time, naturalistic data collection within a large treatment-seeking sample), there are limitations to note. Data were based on daily recalls and thus may be subject to recall biases (e.g., effort after meaning), particularly with respect to pain intensity and interference. To this end, future research using ecological momentary assessment (EMA) would be helpful to capture pain levels at the actual time of migraines and repeatedly throughout the day. The sample was limited to women with overweight or obesity. Although this represents a strength of the current study given that this population may be at heightened risk for more severe symptoms, additional research that directly compares medication use patterns and correlates among individuals with migraine who do and do not have obesity is needed to evaluate independent contributions of weight status. Thus, results may not generalize across the weight spectrum or to men. Furthermore, it should be noted that the present study excluded individuals with medication overuse headache (i.e., headache occurring on 15 or more days/month and developing as a consequence of regular overuse of acute or symptomatic medication for more than 3 months), which affects 1%–2% of the general population and 30%–50% of patients who present to specialty headache clinics [31]. It was also not possible to determine whether any participants developed medication overuse headache given that our measurement period was only 28 days, and diagnosis of medication overuse headache requires overuse for a period of 3 months or more. As such, while the results of the present study support early use of medications, future work should examine patterns of medication use in this population, as well as factors that may predispose individuals with migraines to develop this condition over time. Although variance was decomposed into between- and within-person effects, the study was cross-sectional in design and cannot speak to causal relationships. While the use of ESM enhances ecological validity, all daily reports are nevertheless subject to retrospective recall biases. Time of dosing or pain intensity at the time of medication use was not assessed, which precludes nuanced understanding of the extent to which participants were able to identify attacks and initiate treatment early during the episode. In addition, the present study did not examine the extent to which patterns of medication use or adherence were predictive of future symptom change, which would be important questions for future research. Finally, the dose of each medication or formulation type was not assessed. Given that medication doses vary in duration (e.g., some are dosed once daily while others require more frequent dosing), the number of doses reported should be interpreted with caution.

In conclusion, this study provides the first evidence to our knowledge to characterize patterns of migraine-specific and nonspecific medication use in daily life among women comorbid migraine and overweight or obesity. Potential barriers to medication use will be important to explore, particularly among younger populations and those of lower socioeconomic status. Importantly, results also suggest that naturalistically assessed factors are more salient predictors of medication use compared to traditional questionnaires. Ultimately, such data may be harnessed to inform the development of tailored momentary interventions (e.g., just-in-time adaptive interventions) to enhance medication adherence in this populations [32]. Doing so may facilitate adherence to medication recommendations at the precise moments when support is needed and among individuals most at risk for migraine-related burden.

Funding

This work was supported by the National Institutes of Health (R01 NS077925; Primary Investigator: D.S.B.).

Compliance with Ethical Standards

Conflicts of Interest: None.

Ethical Approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The protocol was approved by the Miriam Hospital Institutional Review Board.

Informed Consent: Informed consent was obtained from all individual participants included in the study.

Transparency Statement:

This study was not formally registered. The analysis plan was not formally preregistered. The primary study was registered at ClinicalTrials.gov (NCT01197196). Deidentified data from this study are not available in a public archive. Deidentified data from this study will be made available (as allowable according to institutional review board standards) by emailing the corresponding author. Analytic codes used to conduct the analyses presented in this study are not available in a public archive. They may be available by emailing the corresponding author. Materials used to conduct the study are not publicly available.