ISSN 0017-8748 doi: 10.1111/head.12556 Published by Wiley Periodicals, Inc.
Headache © 2015 American Headache Society
Research Submission Adding Additional Acute Medications to a Triptan Regimen for Migraine and Observed Changes in Headache-Related Disability: Results From the American Migraine Prevalence and Prevention (AMPP) Study Dawn C. Buse, PhD; Daniel Serrano, PhD; Michael L. Reed, PhD; Shashi H. Kori, MD; Cedric M. Cunanan, MPH; Aubrey Manack Adams, PhD; Richard B. Lipton, MD
Background.—Though triptans are the most widely used acute treatments for migraine, response to treatment is sometimes suboptimal. Triptan therapy is often augmented by the addition of other acute treatments. The benefits of this practice have not been examined in large-scale, real-world observational studies. Objectives.—To assess changes in headache-related disability associated with adding additional acute treatments to a triptan regimen by category of added treatment including: a second triptan, nonsteroidal anti-inflammatory drugs (NSAID), opioids or barbiturates. Methods.—Subjects were participants in the American Migraine Prevalence and Prevention study, a longitudinal, US population-based study of individuals with “severe” headache. Respondents who met International Classification of Headache Disorders 3 beta criteria for migraine were on triptan therapy per respondent self-report, used the same triptan, and provided headache-related disability data for at least 2 consecutive years. Subjects were divided based on headache days per month into 3 groups: low-frequency episodic migraine (LFEM, 0-4), moderate-frequency episodic migraine (MFEM, 5-9), and high-frequency episodic migraine/chronic migraine (HFEM/CM, ≥ 10 headache days per month). HFEM and CM were combined into a single group for analyses because of sample size limitations. Patterns of acute treatment for migraine were monitored from one year to the next over the following couplets of years (2005-2006, 2006-2007, 2007-2008, and 2008-2009). The first eligible couplet was analyzed for each respondent. Medication regimens studied included: (1) maintaining current triptan use (consistent group); (2) adding a different triptan; (3) adding an NSAID; or (4) adding a combination analgesic containing opioids or barbiturates. We assessed change in Migraine Disability Assessment (MIDAS) score from the first to the second year of a couplet, contrasting scores of participants with consistent use with those who added an acute treatment to their triptan regimen.
From the Albert Einstein College of Medicine, Bronx, NY, USA (D.C. Buse, D. Serrano, and R.B. Lipton); Montefiore Headache Center, Bronx, NY, USA (D.C. Buse, R.B. Lipton); Endpoint Outcomes, Boston, MA, USA (D. Serrano); Vedanta Research, Chapel Hill, NC, USA (M.L. Reed); Formerly of Allergan Inc., Irvine, CA, USA (S.H. Kori, C.M. Cunanan); Autonomic Technologies Inc., Redwood City, CA, USA (S.H. Kori); Allergan Inc., Irvine, CA, USA (A.M. Adams). Sponsorship: The American Migraine Prevalence and Prevention Study was originally funded through a research grant to the National Headache Foundation from Ortho-McNeil Neurologics, Inc., Titusville, NJ. Additional analyses and manuscript preparation were supported through grants from MAP Pharmaceuticals, Mountain View, CA, and Allergan Inc., Irvine, CA, to the National Headache Foundation. Address all correspondence to D.C. Buse, Montefiore Headache Center, Hutch Metro 2 Ambulatory Care Center, 1250 Waters Place, 8th floor, Bronx, NY 10461, USA, email: [email protected] Accepted for publication February 20, 2015.
825
826
June 2015
Results.—The study sample (N = 2128) included 111 individuals who added another triptan, 118 who added an opioid or barbiturate, and 69 who added an NSAID, with referent groups of approximately 600 cases in each group who remained consistent. In general, MIDAS scores were higher among those who made changes from one year to the next compared with those who did not make changes in therapy. In fully adjusted models, adding triptans or NSAIDs was associated with increased disability for HFEM/CM cases at follow-up but decreased disability at follow-up for MFEM cases, resulting in significant interaction effects for both adding triptans and NSAIDs, respectively (15.88, 95% confidence interval [CI] 0.75, 31.01, 38.52, 95% CI 12.43, 64.61). Conclusions.—While the effects of adding vs staying consistent on the outcome of headache-related disability varied by medication type added and headache frequency strata, in general, these results suggest that for individuals with migraine, adding acute therapies to current triptan use is generally not associated with reductions in headache-related disability. The results were strongest among persons with HFEM and CM. These results identify important unmet medical needs in current migraine management, especially among patients with high-frequency migraine, and suggest that alternative treatment strategies are needed to improve patient outcomes. Key words: migraine, triptan, disability (Headache 2015;55:825-839)
INTRODUCTION There are a range of effective pharmacologic1 and non-pharmacologic therapies currently available for management of migraine.2 Acute treatments for migraine include migraine-specific medications (ie, triptans, ergotamine, and dihydroergotamine), dopamine blockers (metochlopramine, prochorperazine) as well as nonspecific agents (ie, simple and combination analgesics, nonsteroidal anti-inflammatories [NSAIDs], opioids, and barbiturates).1,3 When a patient does not respond to a particular medications regimen for headache, health-care professionals (HCPs) must make decisions about how to modify treatment. Options include changing the dose, timing, or route of administration of current therapy, switching medications or adding treatments to the currently prescribed medication. Many patients work their way through a series of treatments before finding a treatment regimen that is effective. Switching among triptans (ie, sumatriptan to eletriptan) or adding a second medication to the initial therapy (ie, adding an NSAID to a triptan) are common approaches in clinical practice.4 There is a wealth of data on the efficacy and clinical profiles of individual medications used for acute treatment derived from randomized, controlled clinical trials.4 However, there is a paucity of data on the functional and therapeutic consequences of combining treatments. As a consequence, treatment guidelines identify the acute treatment options with
proven efficacy but are relatively silent on what to do if an initial treatment does not deliver the desired results. In clinical practice, a high proportion of persons with migraine use more than one acute treatment. An analysis of 11,388 respondents from the American Migraine Prevalence and Prevention (AMPP) Study found that 91.5% of respondents with episodic migraine (EM) used acute medications for headache, and the majority of respondents (53.0%) reported using at least two classes of medication for acute headache management.5 Triptans were used by 18.3% of the sample, but only 21.7% of respondents endorsed using triptans used them as monotherapy, while 38.7% used one other class of medications and 39.6% used two or more additional classes of medication for headache.6 In addition, changes in medication regimes were common from one year to the next. Among AMPP Study respondents who used at least one triptan for acute treatment, 57.0% reported using that same medication one year later, 14.4% added additional medications, and 28.6% discontinued at least one type of medication from one year to the next, reflecting the ongoing practice of combining and changing acute treatments. Although polypharmacy or rational combination therapy is the rule in clinical practice, clinical trials usually examine medications in isolation. The few randomized trials conducted to support the development of combination treatments are a
Headache notable exception to this rule. These factorial studies are required by the Food and Drug Administration (FDA) for approval of combination products. For example, researchers conducted a factorial study that examined the separate and joint effects of acetaminophen, aspirin, and caffeine and showed that adding caffeine to acetaminophen and aspirin had better outcomes than using analgesics alone.7 The development program for a sumatriptan/naproxen combination product demonstrated that the combination of naproxen and sumatriptan together was more effective than either agent alone.8 To better understand patterns and consequences of acute treatment changes, we analyzed longitudinal data from the AMPP Study. In a previous manuscript, we examined the influence of switching between classes of acute treatments on change in headache-related disability using AMPP Study data.9 We found that in general, switching from a triptan to another class of medication or between triptans was not associated with improvements in headacherelated disability, and in some cases was associated with increased headache-related disability. We therefore decided to evaluate the alternative strategy of adding to current triptan therapy using similar methods. In the current manuscript, we identified individuals taking the same triptan for 2 consecutive years and examined the outcome of headache-related disability among those who added additional medications to established triptan therapy. Using individuals who did not change treatment from one year to the next as a reference group, we contrasted individuals who continued to take the same triptan and also added another triptan, NSAID, opioid, or barbiturate. Examining outcomes in the AMPP Study database provides a rich, longitudinal dataset, which allowed us to address questions of real-world combination acute treatment paradigms. We chose headache-related disability as the outcome variable because relieving headache-related disability is an important goal of treatment,3 and level of disability is associated with both direct and indirect costs.10,11 Therefore, reduction in headache-related disability and the restoration of function provides a useful external measure for evaluating outcomes related to therapies. We also
827 hypothesize that the effects associated with adding to triptan therapy may vary with headache frequency (headache days per month), therefore we stratified analyses based on headache frequency by dividing respondents into four groups.
METHODS Study Design.—The AMPP Study is a longitudinal, population-based study whose two phase methodology has been detailed elsewhere.12,13 In phase 1, a self-administered screening questionnaire was mailed to a stratified random sample of 120,000 households, selected to be representative of the US population. Households were drawn from a 600,000 household nationwide panel maintained by TNS Research (formerly known as National Family Opinion or NFO Inc.) a survey sampling firm. Of the 162,562 respondents to the screener, 30,721 reported at least one severe headache in the preceding year. A random sample of 24,000 adults (≥18 years of age) from among those reporting at least one severe headache in the preceding year were selected to participate in Phase 2: a longitudinal study with annual follow-up. Between the years 2005 and 2009, respondents were asked to complete a self-administered questionnaire assessing headache symptoms based on the International Classification of Headache Disorders (ICHD),14,15 socio-demographics, headache-related disability, productivity, medication use, neurological comorbidities, and psychiatric comorbidities among other data. Reasons for nonparticipation were not assessed during the screening phase or at annual follow-up. This study was approved by the Albert Einstein College of Medicine Institutional Review Board. Study Population.—To maximize sample sizes available for the definition of medication adding patterns, we employed an open or rolling-response design. Participants were invited each year but were not required to provide data in every year of the study (2005 to 2009). Respondents eligible for this analysis were required to provide data in 2005 and in any two adjacent years to qualify for inclusion. Pairs of adjacent years were defined as couplets and included 2005-2006, 2006-2007, 2007-2008, and 2008-2009.
828 Couplets were indexed by the first (baseline) year of the pairing. That is, the 2006 couplet characterized the association between adding treatment in 2006 (that couplet’s baseline year) and outcomes in 2007. Respondents were stratified into three groups based on headache frequency status in the couplet baseline year for each couplet contributed. Frequency strata were: low-frequency episodic migraine (LFEM, 0-4 headache days per month), moderate-frequency EM (MFEM, 5-9 headache days per month), or highfrequency EM or chronic migraine (CM) (HFEM/ CM, 10 or more headache days per month). The HFEM and CM groups were combined in order to have a large enough sample for analysis. AMPP Study Questionnaire and Variables.— The AMPP Study survey is a self-administered questionnaire comprising items assessing sociodemographics, headache characteristics, frequency, severity, other necessary information to assign a headache diagnosis, comorbidities, headache-related disability, medication use, health-related quality of life, and other information of interest. Headache classification was determined based on application of ICHD 2nd edition (ICHD-2) criteria using the American Migraine Study (AMS)/AMPP Study diagnostic module to respondent reported symptoms. The AMS/ AMPP Study diagnostic module has been demonstrated to have a sensitivity of 100% and specificity of 82% for the diagnosis of migraine,16 and sensitivity of 91% and specificity of 80% for the diagnosis of CM.17 This diagnostic module used ICHD-2 criteria, which are essentially unchanged for the diagnosis of migraine or episodic migraine. Chronic migraine is determined within the AMPP Study dataset using modified Silberstein–Lipton criteria, which basically assess criteria for migraine with an average of ≥15 days per month of headache.18,19 Due to the largescale and respondent self-report format of data collection, it is not possible to assess all ICHD-3beta (3b) criteria for CM such as response of an individual attack to treatment. The primary outcome of interest in this study was change in headache-related disability, which was assessed using the Migraine Disability Assessment Scale (MIDAS).20 The MIDAS is a five-item, selfadministered questionnaire that assesses days of
June 2015 missed activity or substantially reduced activity due to headache in the preceding 3 months in 3 domains: schoolwork/paid employment, household work or chores, and nonwork (family, social, and leisure) activities. The MIDAS has been used as an outcome measure in both observational studies and clinical trials. The MIDAS is sensitive to change over time.21-23 Controlled trials show that the MIDAS is sensitive to treatment effects in both episodic24,25 and chronic migraine.26 Responses to the five MIDAS items are summed to a total score where higher scores indicate greater headache-related disability. MIDAS change scores were calculated by subtracting raw MIDAS scores obtained in the couplet baseline year from that of the couplet follow-up year, for example, MIDAS 2006 minus MIDAS 2005 or MIDAS 2008 minus MIDAS 2007. Change scores were approximately normally distributed, with negative scores indicating decreased disability and positive values indicating increased disability. Change in MIDAS has been used to assess treatment quality in randomized trials of coordinated care.27,28 MIDAS scores roughly correspond (after accounting for double counting of days) to the number of days of disability per 3 months of disability. This finding is supported by Stewart et al, who demonstrated that the MIDAS disability day estimate (mean 14.53, median 9.0) was not significantly different from the diary-based disability day estimate (mean 13.5, median 8.4) in a daily diary study of n = 144 persons with migraine.29 Therefore, a MIDAS change score of +3 indicates 1 day per month of increased headache-related disability. A change score of −6 indicates 2 days per month decline in headache-related disability. Clinically meaningful change in MIDAS has not been explored. However, clinically relevant reductions of MIDAS may be inferred from prophylaxis and chronification prevention trials, wherein reductions of headache frequency and disability of 1-2 days have been associated with statistically significant and meaningful clinical change.30,31 Selecting and Assessing the Couplet of Interest.— The main exposure of interest in this study was the addition of other treatments to a triptan.
Headache Self-reported acute treatment regimen data were collected annually. Acute medications, either FDA approved for and/or commonly used for headache, were listed in surveys by both generic and brand names. Medications included in the questionnaires changed slightly across the years as the FDA approved new medications for acute migraine pharmacotherapy. Acute medications examined in these analyses included all triptans (naratriptan, almotriptan, frovatriptan, sumatriptan, rizatriptan, eletriptan, and zolmitriptan; in 2009, the sumatriptan naproxen sodium combination known as Treximet was added, but was not considered in this study as we restricted our attention to compounds assessed in every year), NSAIDs (naproxen, ibuprofen, ketaprofen, and ketorolac; in 2009, this list was expanded to include flurbiprofen, celecoxib, indomethacin, etodolac, meclofenamate, meloxicam, nabumetone, and diclofenac), and opioids/barbiturates (propoxyphene, meperidine, hydrocodone, oxycodone, butorphanol, and butalbital). Pairs of adjacent years (couplets) were monitored for additions to acute treatment regimens. Four medication addition types were considered: 1. Adding another triptan (eg, a respondent took rizatriptan in 2005 and rizatriptan plus sumatriptan in 2006). We do not have details on how participants were instructed to use triptans in combination but acknowledge that per label, triptans should not be combined within 24 hours of each other. 2. Adding an opioid or barbiturate (eg, a respondent took eletriptan in 2006 and took eletriptan plus a butalbital containing compound such as Fiorinal/ Fioricet in 2007). 3. Adding an NSAID (eg, a respondent took rizatriptan in 2008 and rizatriptan plus ibuprofen in 2009). 4. Consistent treatment control group (defined as the absence of any addition to triptan regimens from 1 year to the following year within a 2-survey couplet). Individuals who discontinued their current triptan and switched to other acute treatments
829 (rather than adding other agents) are considered in a separate manuscript.9 For each subject, we identified the first couplet in which any of the three additions was made and classified that addition into categories 1, 2, or 3 above. Once an addition was observed, subsequent couplets were not included in the analysis. Subjects were eligible for assignment to group 4 if they failed to meet the adding definitions (1-3) in all of their couplets. In that case, if there was more than one couplet with treatment consistency, we selected the first such eligible couplet. Because the adding could occur in any one of four couplets and the profile varied slightly across adding groups 1-3, the number of controls in the consistency group varied slightly from addition to addition resulting in slightly different sample sizes for consistent treatment control groups. Effectively, we created three largely overlapping consistent (ie, control) groups: Consistency group A: Did not add another triptan in any couplet and consistently used the same triptan across years in their first eligible couplet. Consistency group B: Did not add an opioid/ barbiturate in any couplet and consistently used the same triptan across years in their first eligible couplet. Consistency group C: Did not add an NSAID in any couplet and consistently used the same triptan across years in their first eligible couplet. In our primary models, we augmented the main effect of the adding vs consistency exposure with the inclusion of a headache frequency effect based on three headache frequency strata groups: LFEM, MFEM, and HFEM/CM. In addition to the main effect of headache frequency (days per month), the interaction between adding medications and headache frequency strata was also examined. Because of sample size limitations resulting from the rarity of adding any medications to triptan regimens (though adding to triptan regimens is more common than switching triptan regimens, which we explored in a previous paper9), we chose a parsimonious covariate set for adjusting model effects. Covariates used to adjust the exposure effect
830 included socio-demographic data such as age, sex, body mass index (BMI), and average annual household income were obtained via self-report. BMI was calculated from height and weight using the standard formula. Depression was also included as a covariate in order to adjust for the strong association between headache-related disability and depression. Depression was assessed with the Patient Health Questionnaire – depression module (PHQ-9),32 a validated measure of Major Depressive Disorder based on Diagnostic and Statistical Manual of Mental Disorders-4th Edition (DSM-IV) criteria.33 The PHQ-9 assesses symptoms of depression over the preceding 2 weeks based on items assessing anhedonia, sleep disturbance, appetite disturbance, thought disturbance, lethargy and mania, and suicidal ideation. Responses are scored as “0” (not at all), “1” (several days), “2” (more than half the days), and “3” (nearly every day). A diagnosis of major depression was assigned based on a sum score of ≥10.
ANALYSES Analyses were performed using SAS version 9.2 (Cary, NC, USA). Following convention, statistical significance was achieved when P values exceeded an alpha level of .05. Descriptive statistics were employed to understand adding definitions, MIDAS change scores, and baseline covariates. For descriptive analyses, categorical data were summarized using SAS’ FREQ procedure, while continuous variables were summarized using SAS’ MEANS procedure. To model change in MIDAS, a series of analysis of variance (ANOVA) and analysis of covariance (ANCOVA) models were fit using SAS’ general linear models (GLM) procedure. For each model, cell mean coding was employed. This permitted calculation of main effects as marginal mean differences, and adding by headache frequency interactions as differences in cell mean differences. For model results, cell means, marginal mean main effects, and interactions are given. Cell means are provided in order to allow interested readers to calculate interactions or components contributing to the calculations of interactions.
June 2015
RESULTS We first describe information about covariates and the headache-related disability (MIDAS) outcome measure for each adding and consistency definition followed by a series of unadjusted (ANOVA) and adjusted (ANCOVA) models of MIDAS change scores. Table 1 summarizes sociodemographic characteristics, headache features, and comorbidities among the adding and consistency groups. Pooling over time, there were 111 cases who added another triptan, with a referent of 622 who remained consistent, 118 cases who added an opioid or barbiturate, with a referent of 578 who remained consistent, and 69 cases who added an NSAID, with a referent of 630 cases who remained consistent. Reasons for variation in the size of the consistent treatment control groups are described in the statistical methods section. Baseline and follow-up MIDAS scores by treatment group are provided in Table 1. Headache frequency strata, though distributed similarly across consistency groups, varied across adding groups. Among those adding another triptan, n = 20 (18.0%) had LFEM, n = 16 (13.6%) of those adding an opioid or barbiturate had LFEM, and n = 3 (4.4%) of those adding an NSAID had LFEM. For HFEM/CM, the variation in headache frequency across adding definitions was comparable: n = 68 (61.0%), n = 77 (65.2%), and n = 53 (76.8%), for those adding another triptan, adding an opioid/barbiturate, and adding an NSAID, respectively. Corresponding numbers for MFEM are given in Table 1. Among those adding another triptan, opioid or barbiturate, or NSAID, n = 28 (25.2%), n = 38 (32.2%), and n = 16 (23.2%) met criteria for major depression based on the PHQ-9, respectively, whereas n = 110 (17.7%), n = 93 (16.1%), and n = 116 (18.4%) of those who were consistent met this criteria, respectively. Findings were similar across the different adding definitions for sex, age, income, and BMI. MODEL-BASED RESULTS In unadjusted and adjusted models, irrespective of adding pattern, adding never significantly differed from consistency with regard to change in headacherelated disability as measured by MIDAS. Similarly,
Headache
831
Table 1.—Socio-Demographic Characteristics, Headache Features, and Comorbidities Among Persons With Migraine by Treatment Status Group in the Second Year of the Couplet
Triptans
Variable
Malea Femalea Age BMI Income:
Headache © 2015 American Headache Society
Research Submission Adding Additional Acute Medications to a Triptan Regimen for Migraine and Observed Changes in Headache-Related Disability: Results From the American Migraine Prevalence and Prevention (AMPP) Study Dawn C. Buse, PhD; Daniel Serrano, PhD; Michael L. Reed, PhD; Shashi H. Kori, MD; Cedric M. Cunanan, MPH; Aubrey Manack Adams, PhD; Richard B. Lipton, MD
Background.—Though triptans are the most widely used acute treatments for migraine, response to treatment is sometimes suboptimal. Triptan therapy is often augmented by the addition of other acute treatments. The benefits of this practice have not been examined in large-scale, real-world observational studies. Objectives.—To assess changes in headache-related disability associated with adding additional acute treatments to a triptan regimen by category of added treatment including: a second triptan, nonsteroidal anti-inflammatory drugs (NSAID), opioids or barbiturates. Methods.—Subjects were participants in the American Migraine Prevalence and Prevention study, a longitudinal, US population-based study of individuals with “severe” headache. Respondents who met International Classification of Headache Disorders 3 beta criteria for migraine were on triptan therapy per respondent self-report, used the same triptan, and provided headache-related disability data for at least 2 consecutive years. Subjects were divided based on headache days per month into 3 groups: low-frequency episodic migraine (LFEM, 0-4), moderate-frequency episodic migraine (MFEM, 5-9), and high-frequency episodic migraine/chronic migraine (HFEM/CM, ≥ 10 headache days per month). HFEM and CM were combined into a single group for analyses because of sample size limitations. Patterns of acute treatment for migraine were monitored from one year to the next over the following couplets of years (2005-2006, 2006-2007, 2007-2008, and 2008-2009). The first eligible couplet was analyzed for each respondent. Medication regimens studied included: (1) maintaining current triptan use (consistent group); (2) adding a different triptan; (3) adding an NSAID; or (4) adding a combination analgesic containing opioids or barbiturates. We assessed change in Migraine Disability Assessment (MIDAS) score from the first to the second year of a couplet, contrasting scores of participants with consistent use with those who added an acute treatment to their triptan regimen.
From the Albert Einstein College of Medicine, Bronx, NY, USA (D.C. Buse, D. Serrano, and R.B. Lipton); Montefiore Headache Center, Bronx, NY, USA (D.C. Buse, R.B. Lipton); Endpoint Outcomes, Boston, MA, USA (D. Serrano); Vedanta Research, Chapel Hill, NC, USA (M.L. Reed); Formerly of Allergan Inc., Irvine, CA, USA (S.H. Kori, C.M. Cunanan); Autonomic Technologies Inc., Redwood City, CA, USA (S.H. Kori); Allergan Inc., Irvine, CA, USA (A.M. Adams). Sponsorship: The American Migraine Prevalence and Prevention Study was originally funded through a research grant to the National Headache Foundation from Ortho-McNeil Neurologics, Inc., Titusville, NJ. Additional analyses and manuscript preparation were supported through grants from MAP Pharmaceuticals, Mountain View, CA, and Allergan Inc., Irvine, CA, to the National Headache Foundation. Address all correspondence to D.C. Buse, Montefiore Headache Center, Hutch Metro 2 Ambulatory Care Center, 1250 Waters Place, 8th floor, Bronx, NY 10461, USA, email: [email protected] Accepted for publication February 20, 2015.
825
826
June 2015
Results.—The study sample (N = 2128) included 111 individuals who added another triptan, 118 who added an opioid or barbiturate, and 69 who added an NSAID, with referent groups of approximately 600 cases in each group who remained consistent. In general, MIDAS scores were higher among those who made changes from one year to the next compared with those who did not make changes in therapy. In fully adjusted models, adding triptans or NSAIDs was associated with increased disability for HFEM/CM cases at follow-up but decreased disability at follow-up for MFEM cases, resulting in significant interaction effects for both adding triptans and NSAIDs, respectively (15.88, 95% confidence interval [CI] 0.75, 31.01, 38.52, 95% CI 12.43, 64.61). Conclusions.—While the effects of adding vs staying consistent on the outcome of headache-related disability varied by medication type added and headache frequency strata, in general, these results suggest that for individuals with migraine, adding acute therapies to current triptan use is generally not associated with reductions in headache-related disability. The results were strongest among persons with HFEM and CM. These results identify important unmet medical needs in current migraine management, especially among patients with high-frequency migraine, and suggest that alternative treatment strategies are needed to improve patient outcomes. Key words: migraine, triptan, disability (Headache 2015;55:825-839)
INTRODUCTION There are a range of effective pharmacologic1 and non-pharmacologic therapies currently available for management of migraine.2 Acute treatments for migraine include migraine-specific medications (ie, triptans, ergotamine, and dihydroergotamine), dopamine blockers (metochlopramine, prochorperazine) as well as nonspecific agents (ie, simple and combination analgesics, nonsteroidal anti-inflammatories [NSAIDs], opioids, and barbiturates).1,3 When a patient does not respond to a particular medications regimen for headache, health-care professionals (HCPs) must make decisions about how to modify treatment. Options include changing the dose, timing, or route of administration of current therapy, switching medications or adding treatments to the currently prescribed medication. Many patients work their way through a series of treatments before finding a treatment regimen that is effective. Switching among triptans (ie, sumatriptan to eletriptan) or adding a second medication to the initial therapy (ie, adding an NSAID to a triptan) are common approaches in clinical practice.4 There is a wealth of data on the efficacy and clinical profiles of individual medications used for acute treatment derived from randomized, controlled clinical trials.4 However, there is a paucity of data on the functional and therapeutic consequences of combining treatments. As a consequence, treatment guidelines identify the acute treatment options with
proven efficacy but are relatively silent on what to do if an initial treatment does not deliver the desired results. In clinical practice, a high proportion of persons with migraine use more than one acute treatment. An analysis of 11,388 respondents from the American Migraine Prevalence and Prevention (AMPP) Study found that 91.5% of respondents with episodic migraine (EM) used acute medications for headache, and the majority of respondents (53.0%) reported using at least two classes of medication for acute headache management.5 Triptans were used by 18.3% of the sample, but only 21.7% of respondents endorsed using triptans used them as monotherapy, while 38.7% used one other class of medications and 39.6% used two or more additional classes of medication for headache.6 In addition, changes in medication regimes were common from one year to the next. Among AMPP Study respondents who used at least one triptan for acute treatment, 57.0% reported using that same medication one year later, 14.4% added additional medications, and 28.6% discontinued at least one type of medication from one year to the next, reflecting the ongoing practice of combining and changing acute treatments. Although polypharmacy or rational combination therapy is the rule in clinical practice, clinical trials usually examine medications in isolation. The few randomized trials conducted to support the development of combination treatments are a
Headache notable exception to this rule. These factorial studies are required by the Food and Drug Administration (FDA) for approval of combination products. For example, researchers conducted a factorial study that examined the separate and joint effects of acetaminophen, aspirin, and caffeine and showed that adding caffeine to acetaminophen and aspirin had better outcomes than using analgesics alone.7 The development program for a sumatriptan/naproxen combination product demonstrated that the combination of naproxen and sumatriptan together was more effective than either agent alone.8 To better understand patterns and consequences of acute treatment changes, we analyzed longitudinal data from the AMPP Study. In a previous manuscript, we examined the influence of switching between classes of acute treatments on change in headache-related disability using AMPP Study data.9 We found that in general, switching from a triptan to another class of medication or between triptans was not associated with improvements in headacherelated disability, and in some cases was associated with increased headache-related disability. We therefore decided to evaluate the alternative strategy of adding to current triptan therapy using similar methods. In the current manuscript, we identified individuals taking the same triptan for 2 consecutive years and examined the outcome of headache-related disability among those who added additional medications to established triptan therapy. Using individuals who did not change treatment from one year to the next as a reference group, we contrasted individuals who continued to take the same triptan and also added another triptan, NSAID, opioid, or barbiturate. Examining outcomes in the AMPP Study database provides a rich, longitudinal dataset, which allowed us to address questions of real-world combination acute treatment paradigms. We chose headache-related disability as the outcome variable because relieving headache-related disability is an important goal of treatment,3 and level of disability is associated with both direct and indirect costs.10,11 Therefore, reduction in headache-related disability and the restoration of function provides a useful external measure for evaluating outcomes related to therapies. We also
827 hypothesize that the effects associated with adding to triptan therapy may vary with headache frequency (headache days per month), therefore we stratified analyses based on headache frequency by dividing respondents into four groups.
METHODS Study Design.—The AMPP Study is a longitudinal, population-based study whose two phase methodology has been detailed elsewhere.12,13 In phase 1, a self-administered screening questionnaire was mailed to a stratified random sample of 120,000 households, selected to be representative of the US population. Households were drawn from a 600,000 household nationwide panel maintained by TNS Research (formerly known as National Family Opinion or NFO Inc.) a survey sampling firm. Of the 162,562 respondents to the screener, 30,721 reported at least one severe headache in the preceding year. A random sample of 24,000 adults (≥18 years of age) from among those reporting at least one severe headache in the preceding year were selected to participate in Phase 2: a longitudinal study with annual follow-up. Between the years 2005 and 2009, respondents were asked to complete a self-administered questionnaire assessing headache symptoms based on the International Classification of Headache Disorders (ICHD),14,15 socio-demographics, headache-related disability, productivity, medication use, neurological comorbidities, and psychiatric comorbidities among other data. Reasons for nonparticipation were not assessed during the screening phase or at annual follow-up. This study was approved by the Albert Einstein College of Medicine Institutional Review Board. Study Population.—To maximize sample sizes available for the definition of medication adding patterns, we employed an open or rolling-response design. Participants were invited each year but were not required to provide data in every year of the study (2005 to 2009). Respondents eligible for this analysis were required to provide data in 2005 and in any two adjacent years to qualify for inclusion. Pairs of adjacent years were defined as couplets and included 2005-2006, 2006-2007, 2007-2008, and 2008-2009.
828 Couplets were indexed by the first (baseline) year of the pairing. That is, the 2006 couplet characterized the association between adding treatment in 2006 (that couplet’s baseline year) and outcomes in 2007. Respondents were stratified into three groups based on headache frequency status in the couplet baseline year for each couplet contributed. Frequency strata were: low-frequency episodic migraine (LFEM, 0-4 headache days per month), moderate-frequency EM (MFEM, 5-9 headache days per month), or highfrequency EM or chronic migraine (CM) (HFEM/ CM, 10 or more headache days per month). The HFEM and CM groups were combined in order to have a large enough sample for analysis. AMPP Study Questionnaire and Variables.— The AMPP Study survey is a self-administered questionnaire comprising items assessing sociodemographics, headache characteristics, frequency, severity, other necessary information to assign a headache diagnosis, comorbidities, headache-related disability, medication use, health-related quality of life, and other information of interest. Headache classification was determined based on application of ICHD 2nd edition (ICHD-2) criteria using the American Migraine Study (AMS)/AMPP Study diagnostic module to respondent reported symptoms. The AMS/ AMPP Study diagnostic module has been demonstrated to have a sensitivity of 100% and specificity of 82% for the diagnosis of migraine,16 and sensitivity of 91% and specificity of 80% for the diagnosis of CM.17 This diagnostic module used ICHD-2 criteria, which are essentially unchanged for the diagnosis of migraine or episodic migraine. Chronic migraine is determined within the AMPP Study dataset using modified Silberstein–Lipton criteria, which basically assess criteria for migraine with an average of ≥15 days per month of headache.18,19 Due to the largescale and respondent self-report format of data collection, it is not possible to assess all ICHD-3beta (3b) criteria for CM such as response of an individual attack to treatment. The primary outcome of interest in this study was change in headache-related disability, which was assessed using the Migraine Disability Assessment Scale (MIDAS).20 The MIDAS is a five-item, selfadministered questionnaire that assesses days of
June 2015 missed activity or substantially reduced activity due to headache in the preceding 3 months in 3 domains: schoolwork/paid employment, household work or chores, and nonwork (family, social, and leisure) activities. The MIDAS has been used as an outcome measure in both observational studies and clinical trials. The MIDAS is sensitive to change over time.21-23 Controlled trials show that the MIDAS is sensitive to treatment effects in both episodic24,25 and chronic migraine.26 Responses to the five MIDAS items are summed to a total score where higher scores indicate greater headache-related disability. MIDAS change scores were calculated by subtracting raw MIDAS scores obtained in the couplet baseline year from that of the couplet follow-up year, for example, MIDAS 2006 minus MIDAS 2005 or MIDAS 2008 minus MIDAS 2007. Change scores were approximately normally distributed, with negative scores indicating decreased disability and positive values indicating increased disability. Change in MIDAS has been used to assess treatment quality in randomized trials of coordinated care.27,28 MIDAS scores roughly correspond (after accounting for double counting of days) to the number of days of disability per 3 months of disability. This finding is supported by Stewart et al, who demonstrated that the MIDAS disability day estimate (mean 14.53, median 9.0) was not significantly different from the diary-based disability day estimate (mean 13.5, median 8.4) in a daily diary study of n = 144 persons with migraine.29 Therefore, a MIDAS change score of +3 indicates 1 day per month of increased headache-related disability. A change score of −6 indicates 2 days per month decline in headache-related disability. Clinically meaningful change in MIDAS has not been explored. However, clinically relevant reductions of MIDAS may be inferred from prophylaxis and chronification prevention trials, wherein reductions of headache frequency and disability of 1-2 days have been associated with statistically significant and meaningful clinical change.30,31 Selecting and Assessing the Couplet of Interest.— The main exposure of interest in this study was the addition of other treatments to a triptan.
Headache Self-reported acute treatment regimen data were collected annually. Acute medications, either FDA approved for and/or commonly used for headache, were listed in surveys by both generic and brand names. Medications included in the questionnaires changed slightly across the years as the FDA approved new medications for acute migraine pharmacotherapy. Acute medications examined in these analyses included all triptans (naratriptan, almotriptan, frovatriptan, sumatriptan, rizatriptan, eletriptan, and zolmitriptan; in 2009, the sumatriptan naproxen sodium combination known as Treximet was added, but was not considered in this study as we restricted our attention to compounds assessed in every year), NSAIDs (naproxen, ibuprofen, ketaprofen, and ketorolac; in 2009, this list was expanded to include flurbiprofen, celecoxib, indomethacin, etodolac, meclofenamate, meloxicam, nabumetone, and diclofenac), and opioids/barbiturates (propoxyphene, meperidine, hydrocodone, oxycodone, butorphanol, and butalbital). Pairs of adjacent years (couplets) were monitored for additions to acute treatment regimens. Four medication addition types were considered: 1. Adding another triptan (eg, a respondent took rizatriptan in 2005 and rizatriptan plus sumatriptan in 2006). We do not have details on how participants were instructed to use triptans in combination but acknowledge that per label, triptans should not be combined within 24 hours of each other. 2. Adding an opioid or barbiturate (eg, a respondent took eletriptan in 2006 and took eletriptan plus a butalbital containing compound such as Fiorinal/ Fioricet in 2007). 3. Adding an NSAID (eg, a respondent took rizatriptan in 2008 and rizatriptan plus ibuprofen in 2009). 4. Consistent treatment control group (defined as the absence of any addition to triptan regimens from 1 year to the following year within a 2-survey couplet). Individuals who discontinued their current triptan and switched to other acute treatments
829 (rather than adding other agents) are considered in a separate manuscript.9 For each subject, we identified the first couplet in which any of the three additions was made and classified that addition into categories 1, 2, or 3 above. Once an addition was observed, subsequent couplets were not included in the analysis. Subjects were eligible for assignment to group 4 if they failed to meet the adding definitions (1-3) in all of their couplets. In that case, if there was more than one couplet with treatment consistency, we selected the first such eligible couplet. Because the adding could occur in any one of four couplets and the profile varied slightly across adding groups 1-3, the number of controls in the consistency group varied slightly from addition to addition resulting in slightly different sample sizes for consistent treatment control groups. Effectively, we created three largely overlapping consistent (ie, control) groups: Consistency group A: Did not add another triptan in any couplet and consistently used the same triptan across years in their first eligible couplet. Consistency group B: Did not add an opioid/ barbiturate in any couplet and consistently used the same triptan across years in their first eligible couplet. Consistency group C: Did not add an NSAID in any couplet and consistently used the same triptan across years in their first eligible couplet. In our primary models, we augmented the main effect of the adding vs consistency exposure with the inclusion of a headache frequency effect based on three headache frequency strata groups: LFEM, MFEM, and HFEM/CM. In addition to the main effect of headache frequency (days per month), the interaction between adding medications and headache frequency strata was also examined. Because of sample size limitations resulting from the rarity of adding any medications to triptan regimens (though adding to triptan regimens is more common than switching triptan regimens, which we explored in a previous paper9), we chose a parsimonious covariate set for adjusting model effects. Covariates used to adjust the exposure effect
830 included socio-demographic data such as age, sex, body mass index (BMI), and average annual household income were obtained via self-report. BMI was calculated from height and weight using the standard formula. Depression was also included as a covariate in order to adjust for the strong association between headache-related disability and depression. Depression was assessed with the Patient Health Questionnaire – depression module (PHQ-9),32 a validated measure of Major Depressive Disorder based on Diagnostic and Statistical Manual of Mental Disorders-4th Edition (DSM-IV) criteria.33 The PHQ-9 assesses symptoms of depression over the preceding 2 weeks based on items assessing anhedonia, sleep disturbance, appetite disturbance, thought disturbance, lethargy and mania, and suicidal ideation. Responses are scored as “0” (not at all), “1” (several days), “2” (more than half the days), and “3” (nearly every day). A diagnosis of major depression was assigned based on a sum score of ≥10.
ANALYSES Analyses were performed using SAS version 9.2 (Cary, NC, USA). Following convention, statistical significance was achieved when P values exceeded an alpha level of .05. Descriptive statistics were employed to understand adding definitions, MIDAS change scores, and baseline covariates. For descriptive analyses, categorical data were summarized using SAS’ FREQ procedure, while continuous variables were summarized using SAS’ MEANS procedure. To model change in MIDAS, a series of analysis of variance (ANOVA) and analysis of covariance (ANCOVA) models were fit using SAS’ general linear models (GLM) procedure. For each model, cell mean coding was employed. This permitted calculation of main effects as marginal mean differences, and adding by headache frequency interactions as differences in cell mean differences. For model results, cell means, marginal mean main effects, and interactions are given. Cell means are provided in order to allow interested readers to calculate interactions or components contributing to the calculations of interactions.
June 2015
RESULTS We first describe information about covariates and the headache-related disability (MIDAS) outcome measure for each adding and consistency definition followed by a series of unadjusted (ANOVA) and adjusted (ANCOVA) models of MIDAS change scores. Table 1 summarizes sociodemographic characteristics, headache features, and comorbidities among the adding and consistency groups. Pooling over time, there were 111 cases who added another triptan, with a referent of 622 who remained consistent, 118 cases who added an opioid or barbiturate, with a referent of 578 who remained consistent, and 69 cases who added an NSAID, with a referent of 630 cases who remained consistent. Reasons for variation in the size of the consistent treatment control groups are described in the statistical methods section. Baseline and follow-up MIDAS scores by treatment group are provided in Table 1. Headache frequency strata, though distributed similarly across consistency groups, varied across adding groups. Among those adding another triptan, n = 20 (18.0%) had LFEM, n = 16 (13.6%) of those adding an opioid or barbiturate had LFEM, and n = 3 (4.4%) of those adding an NSAID had LFEM. For HFEM/CM, the variation in headache frequency across adding definitions was comparable: n = 68 (61.0%), n = 77 (65.2%), and n = 53 (76.8%), for those adding another triptan, adding an opioid/barbiturate, and adding an NSAID, respectively. Corresponding numbers for MFEM are given in Table 1. Among those adding another triptan, opioid or barbiturate, or NSAID, n = 28 (25.2%), n = 38 (32.2%), and n = 16 (23.2%) met criteria for major depression based on the PHQ-9, respectively, whereas n = 110 (17.7%), n = 93 (16.1%), and n = 116 (18.4%) of those who were consistent met this criteria, respectively. Findings were similar across the different adding definitions for sex, age, income, and BMI. MODEL-BASED RESULTS In unadjusted and adjusted models, irrespective of adding pattern, adding never significantly differed from consistency with regard to change in headacherelated disability as measured by MIDAS. Similarly,
Headache
831
Table 1.—Socio-Demographic Characteristics, Headache Features, and Comorbidities Among Persons With Migraine by Treatment Status Group in the Second Year of the Couplet
Triptans
Variable
Malea Femalea Age BMI Income:
