Opioid Use among People with Migraine: Results of the OVERCOME (US) Study
https://ror.org/03vek6s52grid.38142.3c000000041936754XDepartment of Neurology and Department of Anesthesia, Critical Care, and Pain Medicine, Harvard Medical School, Beth Israel Deaconess Medical Center, Boston, MA USA
https://ror.org/01qat3289grid.417540.30000 0000 2220 2544Eli Lilly and Company, Indianapolis, IN USA
https://ror.org/05cf8a891grid.251993.50000 0001 2179 1997Department of Neurology, Albert Einstein College of Medicine, Bronx, NY USA
Vedanta Research, Chapel Hill, NC USA
https://ror.org/0155zta11grid.59062.380000 0004 1936 7689Department of Neurological Sciences, Larner College of Medicine, University of Vermont, Burlington, VT USA
Haven Headache and Migraine Center, San Francisco, CA USA
https://ror.org/044ntvm43grid.240283.f0000 0001 2152 0791Montefiore Headache Center, Bronx, NY USA
https://ror.org/01qat3289grid.417540.30000 0000 2220 2544Medical Affairs, Neuroscience Business Unit, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN 46225 USA
Abstract
Introduction
Despite expert recommendations against using opioids for migraine treatment, their use remains common in the USA. We aimed to evaluate the use of opioids among people with active migraine using data from the Observational Survey of the Epidemiology, Treatment, and Care of Migraine (OVERCOME) (US) study.
Methods
This observational, longitudinal, web-based survey study included a demographically representative sample of adults with migraine in the USA (2018–2020). Participants with migraine (International Classification of Headache Disorders, third edition [ICHD-3]) and ≥ 1 headache in the previous 12 months were identified via a questionnaire and/or self-reported diagnosis. Information on opioid use for acute migraine treatment was collected. Demographics, clinical, and migraine-related characteristics among those with current opioid use and those with non-use were evaluated in the cross-sectional analysis using standardized mean difference (SMD). Multivariable analysis was conducted using machine learning (least absolute shrinkage and selection operator regression, random forest) and logistic regression models to assess factors associated with current opioid use.
Results
Of 61,932 respondents with active migraine, 13,331 (21.5%) reported currently using opioids to treat migraine. Among those using opioids, 68.0% were female, 64.3% identified as White, and 13.7% identified as Hispanic. Those currently using opioids differed from those not using opioids in various characteristics, including higher tobacco/marijuana use, more comorbidities, higher migraine-related disability, and higher interictal burden (all SMD > 0.2). The factors most associated with current opioid use were “currently taking recommended acute medications for migraine” (odds ratio [OR], 10.1; confidence interval [CI], 9.47, 10.78), “currently taking barbiturates for migraine” (OR, 2.2; CI, 2.03, 2.34), and “sought care at an Emergency Department/Urgent Care for migraine in the previous 12 months” (OR, 1.7; CI, 1.67, 1.85).
Conclusions
This study shows that opioid use for migraine is associated with using recommended acute medications, barbiturates, and emergency department care for migraine. Understanding how to limit these factors is key to developing interventions to reduce opioid use in migraine.
Supplementary Information
The online version contains supplementary material available at 10.1007/s40122-025-00774-y.
Key Summary Points
| Why carry out this study? |
| Opioid abuse is a significant public health concern, and despite expert recommendations against their use for treating migraine, opioids remain commonly prescribed and/or used in the USA. |
| This indicates a potential gap in effective migraine management and preventive options, and the current study aimed to identify the factors associated with opioid use for migraine among participants of the Observational Survey of the Epidemiology, Treatment, and Care of Migraine (OVERCOME) (US) study. |
| What was learned from the study? |
| Among participants with active migraine from OVERCOME (US), an observational web-based survey among a demographically representative sample of more than 60,000 adults in the USA from 2018 to 2020, 21.5% reported currently using opioids to treat migraine. |
| The factors most associated with using opioids for migraine were taking recommended acute migraine treatment, currently taking medications containing barbiturates, and having sought care at an emergency department for migraine in the previous 12 months. |
| There is a need for more effective and informed migraine prevention strategies to reduce reliance on acute medication. |
Introduction
Migraine, a neurological disease, affects as many as 16% of the US adult population [1] and is a leading cause of disability [2, 3]. Pharmacologic management of migraine includes acute medications to abort headache attacks, as well as preventive medications aimed at reducing the frequency of headache attacks [4–7]. Current evidence-based guidelines and recommendations from scientific societies regarding the acute treatment of migraine include simple analgesics, caffeinated analgesic combinations, and migraine-specific medications including triptans, ditans, and gepants [5, 8]. Opioids are not recommended treatments for migraine but are often prescribed and/or used in the USA [9–16]. Historically, opioids were often used in the emergency department setting or as a last resort when acute medications had failed [17]. However, previous studies have shown that opioid use can be a risk factor for chronification of migraine [18]. Opioid abuse, dependence, and misuse are leading causes of disability-adjusted life years across pain conditions [19, 20] and are major public health issues [21–24]. Opioid use is associated with risk of addiction [25] and, in patients with migraine, increases the risk of experiencing more headache days per month (MHDs) [9, 13, 18, 26], more severe headache-related disability [9], greater healthcare resource utilization for headaches [9], risk of medication overuse, and medication overuse headache [18, 27]. There is a need to understand the factors that continue to drive opioid use in people with migraine. Prior studies have evaluated the use of opioids among those with migraine and characterized those using opioids in more detail [6, 10, 22, 28]. However, some of these studies were unable to ascertain whether opioids were being used specifically for migraine or a comorbid condition (e.g., chronic back pain). While these studies utilize a candidate approach for examining patient characteristics among those using opioids, their sample size did not allow for the opportunity to utilize more advanced modeling (e.g., machine learning) to consider a more expansive set of factors that may be associated with opioid use. More advanced modeling further permits consideration of the complex, nonlinear relationships and interactions of factors that can reasonably be accomplished with a standard regression-based approach. Moreover, given the historical emphasis on reducing unnecessary opioid use in clinical practice [9, 13, 26, 29–31], assessing and understanding current opioid use for migraine may provide further opportunities for improving migraine treatment. Through increasing clinicians’ understanding of the risks of opioid use in the treatment of migraine, the updated analysis may provide insight that can help clinicians better evaluate their patients’ needs and aid in the treatment decision-making. The Observational Survey of the Epidemiology, Treatment, and Care of Migraine (OVERCOME) (US) study characterized medication use in people with migraine. The current analysis aimed to evaluate the current use of opioids in the treatment of migraine and utilize machine learning methods to determine factors associated with opioid use, with the overall aim to better understand risk factors for opioid use and migraine care needs in the USA. We hypothesized that those using opioids exhibit greater medical needs compared with those who do not.
Methods
Study Design
The OVERCOME (US) study, a longitudinal, web-based survey that annually (2018–2020) recruited a demographically representative adult sample in the USA. The current analysis of the OVERCOME (US) study was a cross-sectional evaluation of pooled baseline cohorts, which were recruited using identical methods in 2018, 2019, and 2020. Full details of the methods used to recruit, identify, ensure demographic representativeness, and administer the survey have been previously published [32]. Briefly, it assessed acute and preventive medication use for migraine, including opioid use. Representativeness of the US population was achieved via quota sampling to reflect the marginal distribution of geography, age, race, and sex. Study inclusion criteria were that the respondent (1) was 18 years or older, (2) was US resident, (3) was online survey panel member, (4) had internet access, and (5) had the ability to read/write English. Among the demographically representative sample, participants were identified as having active migraine if they reported having had ≥ 1 headache in the previous 12 months and met the criteria for migraine (assessed via a validated migraine diagnostic questionnaire [33, 34] based on criteria described in Part 1 of the International Classification of Headache Disorders, third edition [35]) and/or self-reported having received a medical diagnosis of migraine from a healthcare professional. Those with active migraine then completed the full migraine survey. Figure 1 provides a diagram of the respondent flow. The sample size requirements were based on the precision of the percentage of individuals with a certain attribute (e.g., percent eligible for preventive migraine treatment); the inclusion of 40,000 people with migraine provided a precision of at least 0.75%. The study was approved by the Sterling Institutional Review Board (IRB; ID #6425-001), and the participants provided informed consent prior to the start of the study. The study was performed in accordance with the Helsinki Declaration of 1964 and its later amendments.
Measures
Current Opioid Use
Participants reported whether they used specific oral or nasal spray opioid prescription medication(s) by selecting “all that apply” from a list of acute medications that they were “currently using (or typically keep on hand) to treat or relieve migraine or severe headaches” when they had an attack or felt the onset of a headache. Opioids included in this list were butorphanol nasal spray, codeine combinations (including acetaminophen plus codeine and butalbital plus codeine), hydrocodone combinations, meperidine, oxycodone with or without acetaminophen, and tramadol with or without acetaminophen. Study participants were characterized as having current use or non-current opioid use (non-use).
Demographic and Lifestyle Characteristics and Comorbidities
Demographic characteristics included age, sex assigned at birth, geographic region, ethnicity, race, education, household income, employment, and health insurance status. Lifestyle characteristics were collected via self-report and included tobacco use, marijuana use, presence of cardiovascular comorbidities (including prediabetes, diabetes, high cholesterol level, and hypertension), joint or pain comorbidities (including chronic back pain, fibromyalgia, osteoarthritis, and rheumatoid arthritis), psychiatric comorbidities (anxiety, depression, panic disorder), digestive comorbidities (including acid reflux/ gastroesophageal reflux disease, constipation, and inflammatory bowel disease/Crohn’s disease/ulcerative colitis), and allergies, asthma, or sleep apnea. The frequency of anxious and depressive symptoms was measured using the four-item Patient Health Questionnaire (PHQ-4) [36].
Healthcare Setting
Participants reported where they had sought care for migraine or severe headaches in the previous 12 months; response options were grouped into “Emergency Department/Urgent Care,” “Primary Care,” “Specialist” (general neurologist or headache specialist), or “Other” (other specialist, including obstetrician and gynecologist [OB/GYN]). Consulting with a specialist was considered the most specialized level of care, followed by primary care and emergency department/urgent care. Respondents were also asked at which of these care settings they had received their most recent opioid prescription.
Statistical Analysis
Sociodemographic, clinical, and migraine-related factors were summarized with means (± standard deviation [SD]) for continuous variables and with proportions for dichotomous or ordinal categorical variables. Sociodemographic, clinical, and migraine-related characteristics were compared between the opioid use and non-use groups using standardized mean differences (SMDs) to provide a sample size index invariant to the size of the difference. SMD values < 0.2 were considered not different and 0.2–0.49 indicated a small difference, 0.5–0.79 a moderate difference, and values > 0.8 a large difference [45]. Multivariable analysis was conducted to assess the predictors of current opioid use. Variables were identified using two machine learning methods. First, a grouped least absolute shrinkage and selection operator (LASSO) regression with operational definitions for sociodemographic, clinical, and migraine-related characteristic variables selected variables based on Akaike information criterion [46]. Second, a random forest (RF) consisting of 1000 individual trees with bootstrap aggregation (bagging) was applied to provide variable importance measures (using the RF algorithm in the Statistical Analysis Software [SAS]) that rank variables in order of predictive importance. The out-of-bag (OOB) Gini measure of variable importance was used, and variables with an OOB Gini Index < 0.004 were not selected for further evaluation. Finally, a multivariable logistic regression with the final variables identified by LASSO and RF was run, modeling current opioid use yes versus no. Odds ratios (ORs) are presented with 95% confidence intervals (CIs) and were considered statistically significant at the 5% level of significance if the 95% CI for the OR did not include 1 using a two-tailed hypothesis test. Analyses were completed in SAS (Version 9.4; Cary, NC), using PROC HPGENSELECT for LASSO regression and the PROC HPFOREST for RF.
Results
Study Population
Among a total of 61,932 respondents with migraine in OVERCOME (US), 13,331 (21.5%) participants reported currently using opioids for migraine, and 48,601 (78.5%) did not (Table 1). Several sociodemographic characteristics were different between opioid use and non-use groups (Table 1). In particular, tobacco (43.1% versus 30.6% [SMD 0.26]) and marijuana use (36.9% versus 23.7% [SMD 0.29]) were higher among those currently using opioids. The mean PHQ-4 sum score was also higher among those currently using opioids (5.1 versus 4.2 [SMD 0.26]), as was the presence of self-reported cardiovascular comorbidities (63.6% versus 47.6% [SMD 0.33]), joint and pain comorbidities (40.8% versus 28.9% [SMD 0.25]), psychiatric comorbidities (63.8% versus 53.9% [SMD 0.20]), and presence of comorbidity that contraindicates triptan use (29.2% versus 14.0% [SMD 0.38]).Characteristic Current opioid use for migrainea Standardized mean differenceb Yes
(n = 13,331)No
(n = 48,601)Age, mean (SD) 41.3 (14.0) 41.8 (14.9) 0.04 Sex at birth, n(%) Female 9065 (68.0) 37,057 (76.2) 0.18 Male 4266 (32.0) 11,544 (23.0) 0.18 Race, n (%) White 8568 (64.3) 34,996 (72.0) 0.17 Black 1394 (10.5) 3413 (7.0) 0.12 Others 3369 (25.3) 10,192 (21.0) 0.10 Hispanic, n (%) Yes 1824 (13.7) 4824 (9.9) 0.12 No 11,060 (83.0) 42,067 (86.6) 0.10 Prefer not to answer 1824 (13.7) 4824 (9.9) 0.01 Health insurance, n (%) Yes 11,788 (88.4) 41,098 (84.6) 0.11 No 1543 (11.6) 7503 (15.4) 0.11 Level of education, n(%) College graduate or above 4679 (35.1) 17,123 (35.2) 0.00 Some college 5403 (40.5) 19,690 (40.5) 0.00 High school or less 3159 (23.7) 11,530 (23.7) 0.00 Prefer not to answer 90 (0.7) 258 (0.5) 0.02 Household income, n (%) < $50,000 6748 (50.6) 24,008 (49.4) 0.02 $50,000–$99,999 4092 (30.7) 15,294 (31.5) 0.02 $100,000 or over 2264 (17.0) 7722 (15.9) 0.03 Prefer not to answer 227 (1.7) 1577 (3.2) 0.10 BMI group, n (%) Underweight/normal 4620 (34.7) 16,895 (34.8) 0.00 Overweight 3667 (27.5) 12,842 (26.4) 0.02 Obese 4484 (33.6) 16,343 (33.6) 0.00 Not reported 560 (4.2) 2521 (5.2) 0.05 Tobacco use (yes), n (%) 5750 (43.1) 14,883 (30.6) 0.26 Marijuana use (yes), n (%) 4921 (36.9) 11,506 (23.7) 0.29 PHQ-4, mean (SD) 5.1 (3.4) 4.2 (3.5) 0.26 Presence and number of self-reported cardiovascular comorbidity/eventsc No 4855 (36.4) 25,489 (52.4) 0.33 Yes 8476 (63.6) 23,112 (47.6) 0.33 1 3188 (23.9) 10,677 (22.0) 0.05 2+ 4540 (34.1) 10,590 (21.8) 0.28 Presence and number of self-reported joint or pain comorbiditiesd No 7888 (59.2) 34,565 (71.1) 0.25 Yes 5443 (40.8) 14,036 (28.9) 0.25 1 3235 (24.3) 9367 (19.3) 0.12 2+ 2208 (16.6) 4669 (9.6) 0.21 Presence and number of self-reported psychiatric comorbiditiese No 4821 (36.2) 22,395 (46.1) 0.20 Yes 8510 (63.8) 26,206 (53.9) 0.20 1 2748 (20.6) 8786 (18.1) 0.06 2 3028 (22.7) 9766 (20.1) 0.06 3 2734 (20.5) 7654 (15.7) 0.12 Presence of comorbidity that is a contraindication for triptans, n (%)f 3899 (29.2) 6807 (14.0) 0.38
Migraine-related characteristics also differed between these groups (Table 2). Those using opioids were more likely to have a higher migraine-related disability (severe MIDAS: 38.5% versus 22.4% [SMD 0.35]), higher interictal burden (severe MIBS-4: 59.9% versus 32.4% [SMD 0.57]), and lower quality of life (mean MSQ-RFR score: 46.2 versus 57.0 [SMD 0.46]). Those using opioids were also less likely to report low headache frequency (0–3 MHDs: 51.7% versus 61.6% [SMD 0.20]).Characteristic Current opioid usea Standardized mean differenceb Yes
(n = 13,331)No
(n = 48,601)MHD, n (%) 0–3 6892 (51.7) 29,927 (61.6) 0.20 4–7 2856 (21.4) 9120 (18.8) 0.07 8–14 1701 (12.8) 4795 (9.9) 0.09 15+ 1882 (14.1) 4759 (9.8) 0.13 Years with migraine, mean (SD) 17.8 (14.4) 19.2 (14.9) 0.09 Pain severity, mean (SD) 6.8 (1.9) 7.5 (1.8) 0.0 MIDAS, n (%) I. Little or no disability: 0–5 3102 (23.3) 21,718 (44.7) 0.46 II. Mild disability: 6–10 2156 (16.2) 8116 (16.7) 0.01 III. Moderate disability: 11–20 2945 (22.1) 7866 (16.2) 0.15 IV. Severe disability: 21+ 5128 (38.5) 10,901 (22.4) 0.35 MIBS-4, n (%) Score = 0: no interictal burden 1852 (13.9) 17,052 (35.1) 0.51 Score = 1–2: mild interictal burden 1639 (12.3) 8700 (17.9) 0.16 Score = 3–4: moderate interictal burden 1856 (13.9) 7108 (14.6) 0.02 Score = 5+: severe interictal burden 7984 (59.9) 15,741 (32.4) 0.57 MSQ-RFR, mean (SD) 46.2 (22.6) 57.0 (24.0) 0.46 PHQ-4, mean (SD) 4.2 (3.5) 5.1 (3.4) 0.26 Most specialized level of care, n(%) None 2622 (19.7) 27,781 (57.2) 0.84 Emergency department/urgent care 627 (4.7) 2321 (4.8) 0.00 Primary care 4179 (31.3) 10,990 (22.6) 0.20 Specialist 5760 (43.2) 6760 (13.9) 0.69 Other 143 (1.1) 749 (1.5) 0.04 Provider of opioid prescription, n(%) Emergency department/urgent care 3857 (28.9) N/A N/A Primary care 6936 (52.2) N/A N/A Specialist 3023 (22.7) N/A N/A Other 355 (2.7) N/A N/A Treatment related, n(%) Currently taking recommended acute medication for migrainec 10,696 (80.2) 11,023 (22.7) 1.41 Currently taking medications that contain barbiturates for migrained 2927 (22.0) 1997 (4.1) 0.96 Currently taking triptanse 7633 (57.3) 7616 (15.7) 0.55 Currently taking recommended preventive medication for migrainef 3857 (28.9) 4903 (10.1) 0.49
Healthcare Setting and Opioid Prescriber
When asked whether and where respondents had sought care, 80.3% among those currently using opioids versus 42.8% of those not currently using opioids indicated that they had sought medical care for migraine in the previous 12 months (SMD 0.84) (Table 2). Our results further showed that those currently using opioid were more likely to seek care at primary care (31.3% versus 22.6% [SMD 0.20]) and through a specialist (43.2% versus 13.9% [SMD 0.69]) compared with those not using opioids. The rate of those who visited an Emergency department/urgent care clinic for migraine did not differ between those currently using opioids and those not currently using opioids ([SMD 0.00]). This is particularly interesting, as 28.9% of those currently using opioids stated that they had received a prescription for an opioid-containing medication to treat their migraine from emergency department/urgent care settings, compared with 52.2% from a primary care provider, and 22.7% from a specialist.
Predictors of Opioid Use
Machine learning models (LASSO, RF) were used to identify the most relevant factors among an a priori identified list of 51 sociodemographic, clinical, and migraine-related characteristics (Supplemental Table 1). Among the total set of variables, LASSO identified six as associated with current opioid use (Table 3A), and RF identified five (Table 3B). Two of the variables, “currently taking recommended acute treatment for migraine” and “sought care at the Emergency Department/Urgent Care in the previous 12 months” were identified by both models.(A) LASSO Variables associated with opioid use among participants with migraine PHQ-4a Years with migraine Average headache pain intensity scoreb MSQ-RFR scorec Sought migraine care at an emergency department/urgent care in the previous 12 months*d Currently taking recommended acute treatment for migraine*e
A total of nine unique variables identified by LASSO and/or RF were then entered into a multivariate logistic regression model to understand the directionality of the association between these variables and the outcome of current opioid use for migraine. This model showed that the most important factor was whether respondents were currently taking a recommended acute treatment for migraine, with those taking such treatment having much higher odds of currently also taking opioids (OR, 10.1; 95% CI, 9.47, 10.78) (Fig. 2). Those currently taking medications containing barbiturates were also at higher odds of current opioid use (OR, 2.18; 95% CI, 2.03, 2.34). The third most significant factor showed that those who had sought care at an emergency department/urgent care in the previous 12 months were also more likely to be those currently using opioids (OR, 1.75; 95% CI, 1.67,1.85). Also, at higher odds of opioid use were those with higher MIBS-4 total score (OR, 1.25; 95% CI, 1.22, 1.27), higher PHQ-4 score (OR, 1.01; 95% CI, 1.00, 1.02), and higher average pain intensity (OR, 1.06; 95% CI, 1.05, 1.08). Higher MSQ-RFR (OR, 1.00; 95% CI, 1.00, 1.00) and currently using triptans (OR, 0.92; 95% CI, 0.86, 0.97) were associated with lower odds of current opioid use (Fig. 2).
Discussion
The current analysis among participants with migraine showed that, despite recommendations against opioid use for the treatment of migraine, a large proportion still use opioids to manage their migraine. We found that more than one in five (21.5%) participants with migraine in OVERCOME (US) (2018–2020) reported using opioids to treat their migraine, up from 16.0% reported in the American Migraine Prevalence and Prevention (AMPP) study (2005–2009) [9]. Of note, the AMPP study defined those currently using opioids as those who reported opioid use for 3 months and only asked about what medications were used to treat their most severe type of headache, whereas OVERCOME (US) defined those with current use as those who reported currently using opioid medication for acute migraine treatment for all headaches [9]. These findings are concerning given that the use of opioids can potentially lead to increased tolerance, physical dependence, and possible opioid use disorder [47–49].
In this study, the single factor that was most strongly associated with current opioid use for migraine was “currently taking recommended acute treatments for migraine” (OR, 10.10). Other associated factors included currently taking barbiturate-containing medications that are not recommended for migraine (OR, 2.18) and having sought care at the emergency department/urgent care in the previous 12 months (OR, 1.75). The association of current opioid use with taking recommended acute treatment might be the result of not experiencing adequate relief from other acute medications, and thus, either opioid medications were added to other acute medications or vice versa. In fact, 80.2% reported currently using recommended acute treatments for migraine. Moreover, current opioid use was associated both with currently taking triptans (SMD 0.55), as well as comorbidities that contraindicate prescribing them (SMD 0.38), which might limit recommended prescribing options. Patients may be utilizing opioids as rescue therapy, as indicated by the Chronic Migraine Epidemiology and Outcomes (CaMEO) study, which showed that more than one third of respondents kept opioids “on hand” [13]. Finally, it is striking that 22.7% of those using opioids were prescribed opioids for migraine by specialists who should be most knowledgeable that they are not recommended for migraine. This suggests that more than one in five patients with active migraine may have attacks that are either unresponsive to or have contraindications/intolerance to recommended medications or that some specialists continue to prescribe opioids that were initiated by a prior prescriber. While health insurance access was similar among those using and not using opioids, patients with attacks that are unresponsive to generic triptans might still be prescribed opioids if step care insurance barriers limited access to other recommended medications (i.e., gepants, ditan, and dihydroergotamine).
Previous studies have shown that using opioids for the acute treatment of migraine is associated with a high risk of medication overuse and medication overuse headaches [27, 50–52]. While many people keep opioid medications in their toolbox of acute treatment options, for many patients, overusing opioids can lead to migraine progression, transformation from episodic to chronic migraine, or an increase in monthly headache day frequency [53, 54]. This risk may be limited if effective and recommended acute and preventive treatments for migraine are used instead of opioids.
The utilization of barbiturates and opioids concurrently is especially concerning, as neither is a preferred treatment for migraine and both have been associated with adverse effects and acute medication overuse [51, 55, 56]. Our results indicate that 22% of the those using opioids were also taking barbiturate-containing medications compared with 4.1% of those not using opioids. The concurrent utilization may reflect ineffective acute migraine treatment as previously reported [15]. While our data do not indicate the timeline of prescribing, the results suggest that individuals with migraine may benefit from being prescribed and taking effective preventive medication rather than merely adding additional acute medications. In this study, 28.9% of those using opioids and 10.1% of those not using opioids are currently taking recommended preventive treatments. This presents an unmet need, as many of these individuals could be considered for preventive treatment, per the AHS consensus statement [8], given their level of disability (60.6% moderate to severe MIDAS) and number of monthly headache days (48.3% with 4 or more MHDs).
The association between opioid use and utilization of emergency care is consistent with previous findings that opioids are prescribed for migraine in the emergency department setting [15, 57]. Overall, 28.9% of those currently using opioids reported receiving their most recent opioid prescription in an emergency department/urgent care clinic setting. This may be because many emergency departments might lack effective options, such as triptans, and until recently, most emergency department clinicians and staff had limited training on effective migraine management [58, 59]. Once medications are prescribed, emergency departments may not have staff to assist with authorization and/or options to monitor adherence and address other concerns [60, 61]. Additionally, there are concerns associated with prescribing triptans to patients with a history of cardiovascular conditions [62]. Our data also show a high rate of opioid prescriptions from primary care providers (52.2%), which is likely due to a higher number of primary care visits overall. Previous data from OVERCOME evaluating the likelihood of receiving an opioid relative to a triptan in various care settings showed the opioid/triptan ratio to be 0.8 in primary care and 2.4 in emergency departments [57]. These findings highlight an opportunity to optimize migraine management in the emergency department/urgent care settings through increased awareness of evidence-based guidelines for acute treatment of migraine and initiation of preventive treatment whenever appropriate [63].
Previous studies that evaluated factors associated with opioid use for migraine had identified sociodemographic (male sex), clinical (body mass index, anxiety, depression, and cardiovascular comorbidity), and migraine-related characteristics (allodynia and increased MHDs) that the current study did not highlight [9, 13]. This is likely a reflection of how the factors were identified. Traditionally, these studies utilized factors selected by clinician-scientists on the basis of their understanding of what factors were theoretically relevant. By using machine learning, the current study was able to ascertain the importance across a larger set of factors. In addition to the associations discussed above, we found a higher likelihood of current opioid use for migraine among those who reported current tobacco or marijuana use, had a higher PHQ-4 score, and reported the presence of various comorbidities, including cardiovascular, joint, and/or pain, psychiatric, and comorbidities that contraindicate triptan use. It is possible that some people used opioids originally prescribed for other conditions to treat migraine; this may explain why those using opioids in this study more frequently reported joint and/or pain comorbidities [55]. Furthermore, our results demonstrated that opioid use is greater among those with higher migraine-related disability, higher interictal burden, and lower migraine-specific quality of life, reflecting similar findings in the AMPP study.
Overall, our data are consistent with previous population-based data, demonstrating that using opioids to treat migraine is associated with increased levels of disability, decreased quality of life, and higher rates of psychiatric symptoms in people with migraine [9]. These outcomes may be mitigated by utilizing more effective non-opioid acute treatments for migraine and/or optimizing preventive treatment as per recommendations and guidelines [64, 65]. These state that opioid treatment should be reserved for people with contraindications for first-line migraine treatments, and the focus should be on preventive treatment strategies for eligible candidates [66]. Notably, having a cardiovascular comorbidity or a contraindication for triptan therapy was associated with opioid use in our study, which may have influenced the prescribing of opioids. Additionally, having a psychiatric comorbidity was also associated with opioid use, which may be due to concerns of serotonin syndrome with triptans and many psychiatric medications. With the availability of newer acute therapies since this study was conducted, individuals with contraindications for triptans have more non-opioid options. It is important for clinicians to adhere to these recommendations and guidelines when discussing treatment options with their patients, especially those who may be seeing patients with migraine when they first seek care for migraine. Additionally, since many patients continue to use opioids despite their limited effectiveness and risks, public awareness campaigns discouraging opioid use for migraine and promoting evidence-based alternatives may be useful.
Strengths and Limitations
This study has many strengths. Firstly, OVERCOME (US) surveyed the largest population of adults with migraine to date and collected data as novel therapeutics entered the market. Secondly, this study only included respondents who reported taking opioids specifically for migraine. Finally, by using supervised machine learning approaches, this study allowed for robust evaluation of > 50 important factors potentially associated with opioid use, more than in previously conducted studies.
This study has several limitations. The sample matched the US census through web-based consumer panels, not through random sampling, potentially under-representing specific subpopulations, such as those not fluent in English, those without reliable or consistent internet access, or those who elect not to participate in web panels. Variations in how variables were coded in this large dataset may have led to inconsistencies in the data analyzed. The self-reported survey data were not validated by medical professionals, claims, or electronic health records, making them susceptible to recall and selection bias. Observational studies often lack control over confounding variables, potentially introducing bias. Additionally, the potential bias of confounding by indication must be considered when interpreting results. Causality cannot be determined from a cross-sectional study, leaving unresolved whether frequent migraines lead to opioid use or vice versa, and whether opioid use drives extracephalic pain or vice versa. Adverse events of medications and timing and manner of opioid use were not collected in the survey. For example, a patient may “keep them on hand” to use as a rescue medication or backup if they run out of other prescribed acute treatments, a nuance this study did not capture. Finally, some opioid prescription medications (e.g., morphine and fentanyl) were excluded in this study; however, their infrequent prescription for migraine likely minimized any significant impact on findings. Future studies should assess the frequency and reasons for opioid use and consider other influences on opioid use, including migraine symptoms and preventive medication use.
Conclusions
The current analysis conducted in more than 60,000 adults with migraine added to the evidence that, despite recommendations against opioid use for the treatment of migraine, many still use them to manage their migraine. This highlights the need for improvements in migraine care to ensure patients are receiving appropriate and recommended treatments for their migraine. This study further highlighted key factors associated with opioid use, including current use of recommended acute treatments or barbiturate-containing medications for migraine, as well as seeking care for migraine in an emergency setting. Understanding the factors associated with opioid use may help clinicians improve interventions to reduce opioid use for migraine in the population and may provide important insights for advocacy groups trying to address opioid use more broadly.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
We thank the participants of the study.
Medical Writing, Editorial, and Other Assistance
We thank Shonda Foster, an employee of Eli Lilly and Company, and Keerthana Muthiah and Maanasa Surampally, employees of Eli Lilly Services India Pvt. Ltd., for writing and/or editorial support, and Armen Zakharyan from TechData Service Company for statistical support.
Funding
Sponsorship for this study was funded by Eli Lilly and Company, Indianapolis, IN, USA. The sponsor also agrees to find the publication charges associated with this manuscript.
Data Availability
Lilly will provide access to anonymized individual participant data collected during the study. The data will be available to request on vivli.org after the study team has completed analyses and publications. Access will be provided after a proposal has been approved by an independent review committee identified for this purpose and after receipt of a signed data-sharing agreement. After a proposal is approved, data and documents, including the study protocol, will need to be provided in a secure data-sharing environment. For details on submitting a request, see the instructions provided at www.vivli.org.
Declarations
Conflict of Interest
Sait Ashina, MD, has served as a consultant and/or advisory board member or has received honoraria from Allergan/AbbVie, Amgen, Biohaven Pharmaceuticals, Eli Lilly and Company, Impel NeuroPharma, Novartis, Satsuma, Supernus, Percept, Pfizer, Teva, and Theranica. Elizabeth Johnston, PharmD, is an employee and minor stockholder of Eli Lilly and Company. E. Jolanda Muenzel, MD, Ph.D., is an employee and minor stockholder of Eli Lilly and Company. Gilwan Kim, PharmD, is an employee and minor stockholder of Eli Lilly and Company. Dawn C. Buse, Ph.D., has received research support from the Food and Drug Administration (FDA) and the National Headache Foundation; she serves as consultant, advisory board member or has received honoraria or research support from AbbVie/Allergan, Amgen, Biohaven, Eli Lilly and Company, Lundbeck, Novartis, and Teva and Theranica. Michael L. Reed, Ph.D., has received research support from the National Headache Foundation; he serves as a consultant and/or advisory board member or has received honoraria or research support from AbbVie/Allergan, Amgen, Dr. Reddy’s Laboratories (Promius), and Eli Lilly and Company. Robert E. Shapiro, MD, Ph.D., serves as consultant and/or advisory board member or has received honoraria or research support from Eli Lilly and Company, AbbVie, Theranica, and Lundbeck. Susan Hutchinson, MD, has consulted for and/or spoken at or received honoraria from Alder/Lundbeck, AbbVie/Allergan, Amgen, Biohaven, Currax, electroCore, Eli Lilly and Company, Impel, Novartis, Teva, Theranica, and Upsher-Smith. Anthony Zagar, MS, is an employee and minor stockholder of Eli Lilly and Company. Robert A. Nicholson, Ph.D., is an employee and minor stockholder of Eli Lilly and Company. Richard B. Lipton, MD, has received research support from the National Institutes of Health, the FDA, and the National Headache Foundation. He serves as a consultant or advisory board member or has received honoraria or research support from AbbVie/Allergan, Amgen, Biohaven, Dr. Reddy’s Laboratories (Promius), electroCore, Eli Lilly and Company, GlaxoSmithKline, Lundbeck, Merck, Novartis, Pfizer, Teva, Vector, and Vedanta Research. He receives royalties from Wolff’s Headache, eighth edition (Oxford University Press, 2009) and Informa. He holds stock/options in Axon, Biohaven, CoolTech, and Manistee Health.
Ethical Approval
The study was approved by Sterling Institutional Review Board (IRB ID #6425–001) and the participants provided informed consent prior to the start of the study. The study was performed in accordance with the Helsinki Declaration of 1964 and its later amendments.