ISSN 0017-8748 doi: 10.1111/head.13199 Published by Wiley Periodicals, Inc. Headache C 2017 American Headache Society V Research Submissions AVP-825 (Sumatriptan Nasal Powder) Reduces Nausea Compared to Sumatriptan Tablets: Results of the COMPASS Randomized Clinical Trial Richard B. Lipton, MD; James S. McGinley, PhD; Kenneth J. Shulman, DO; Stephen D. Silberstein, MD; R. J. Wirth, PhD; Dawn C. Buse, PhD Background.—Migraine-related nausea is associated with significant disability, increased burden of disease, and personal distress. Nausea can lead to delays or avoidance of initiating oral migraine treatment, sometimes resulting in treatment failures and poor outcomes. Nausea is often a symptom of migraine, but nausea may also be a consequence of treatment (ie, treatment emergent nausea [TEN]). Relieving nausea and minimizing TEN are important goals in acute migraine therapy. Methods.—We analyzed data from the COMPASS study, a randomized, double blind, double-dummy, comparative efficacy study that contrasted two active treatments, AVP-825 (breath-powered intranasal delivery of powdered sumatriptan 22 mg) and oral sumatriptan tablets (100 mg). Three-level logistic multilevel models were used to examine longitudinal changes in nausea from three distinct perspectives across multiple attacks. Model 1 (Overall Nausea) examined longitudinal change in nausea from pre-dose through 120 minutes post-dose for the entire sample, independent of baseline nausea. Model 2 examined TEN from 10 minutes through 120 minutes post-dose in attacks free of nausea at baseline to investigate whether or not nausea developed following treatment. Model 3 examined Nausea Relief from 10 minutes through 120 minutes post-dose in eligible attacks with nausea at baseline to examine whether or not nausea was relieved over the first 2 hours post-dose. Models tested for differences in rate of change in nausea over time and odds of nausea at specific time-points. Results.—Longitudinal nausea trajectories differed for AVP-825 and oral sumatriptan in the Overall Nausea model (Model 1) and TEN model (Model 2), but were more comparable across treatments for the Nausea Relief (Model 3). More specifically, in the Overall Nausea model (Model 1), an individual treating an attack with AVP-825 had a significantly faster decrease in nausea through the first 60 minutes post-dose and reduced odds of nausea at each time-point from 30 minutes through 120 minutes post-dose compared to oral sumatriptan. In Model 2, an individual’s risk for TEN increased at a significantly faster rate through the first 45 minutes post-dose when treating an attack with oral sumatriptan, with significantly greater odds of experiencing TEN at 45, 60, and 90 minutes post-dose compared to AVP-825. The Nausea Relief model (Model 3) showed similar rates of change in nausea over time for the two treatments, but there was a constant difference in nausea level leading to reduced odds of nausea when treating with AVP-825 compared to oral sumatriptan. Conclusions.—All three longitudinal models showed that AVP-825 had more favorable nausea outcomes compared to oral sumatriptan. AVP-825 treatment led to more rapid early reductions in Overall Nausea rates during the first hour, From the Department of Neurology Albert Einstein College of Medicine and Montefiore Headache Center, Bronx, NY, USA (R.B. Lipton and D.C. Buse); Department of Epidemiology and Population Health, Albert Einstein College of Medicine, Bronx, NY, USA (R.B. Lipton); Montefiore Medical Center, Bronx, NY, USA (R.B. Lipton and D.C. Buse); Vector Psychometric Group, LLC, Chapel Hill, NC, USA (J.S. McGinley and R.J. Wirth); Avanir Pharmaceuticals, Inc., Aliso Viejo, CA, USA (K.J. Shulman); Thomas Jefferson University, Philadelphia, PA, USA (S.D. Silberstein). Address all correspondence to Richard Lipton, Albert Einstein College of Medicine, 1300 Morris Park Avenue, Van Etten 3C12, Bronx, NY, USA, email: email@example.com Accepted for publication August 22, 2017. 1 2 Month 2017 reduced odds of nausea from 30 minutes to 2 hours following treatment and reduced risk of TEN compared to oral sumatriptan. These results highlight the importance of separately assessing TEN and Nausea Relief in acute treatment trials of migraine. Key words: nausea, migraine, sumatriptan, triptan, disability (Headache 2017;00:00-00) INTRODUCTION Migraine is a neurobiological disorder characterized by recurrent episodes of headache and associated symptoms, including nausea, sensitivity to light, and to sound.1 Nausea is an important determinant of the migraine burden because it is common, disabling, and associated with poor outcomes for both individual attacks and for particular patients.2-4 Roughly half of people with migraine experience nausea with at least half of their attacks.2,5 Persons with migraine and nausea have substantially higher levels of attack-related disability than persons without nausea.2 A large population-based longitudinal study found that people with episodic migraine and frequent nausea progressed to chronic migraine at twice the rate of those with no or low frequency nausea.4 These results persisted after adjusting for headache-related disability, psychiatric comorbidities, and patterns of medication. Thus, nausea is important as a diagnostic feature of migraine, as a target of treatment, and as a factor which should be considered when choosing treatments for specific patients and attacks. The experience and timing of nausea in the migraine attack varies both within and among individuals. Nausea may develop before, after, or at the same time as headache.6,7 Potential mechanisms for migraine associated nausea include dysfunction of the central modulation of multisensory afferents and autonomic dysfunction, which may be associated with gastric paresis.6,8 Gastric paresis during migraine attacks is relevant to acute therapy because it may cause poor absorption of oral medications.9 Triptans are the most widely used acute prescription medications for the acute treatment of migraine and are the treatment of choice for people with significant migraine-related disability.10-12 Triptans are safe and effective; they relieve pain, Conflict of Interest: Dr. Richard B. Lipton is the Edwin S. Lowe Professor of Neurology at the Albert Einstein College of Medicine in New York. He receives research support from the National Institutes of Health (NIH): 2PO1 AG003949 (Program Director), 5U10 NS077308 (PI), 1RO1 AG042595 (Investigator), RO1 NS082432 (Investigator), K23 NS09610 (Mentor), K23AG049466 (Mentor). He also receives support from the Migraine Research Foundation and the National Headache Foundation. He serves on the editorial board of Neurology and as senior advisor to Headache. He has reviewed for the NIA and NINDS, holds stock options in eNeura Therapeutics; serves as consultant, advisory board member, or has received honoraria from: American Academy of Neurology, Alder, Allergan, American Headache Society, Amgen, Autonomic Technologies, Avanir, Biohaven, Biovision, Boston Scientific, Colucid, Dr. Reddy’s, Electrocore, Eli Lilly, eNeura Therapeutics, GlaxoSmithKline (formerly Novartis), Merck, Pernix, Pfizer, Supernus, Teva, Trigemina, Vector, Vedanta. He receives royalties from Wolff’s Headache, 8th Edition, Oxford Press University, 2009, Wiley and Informa. James S. McGinley, PhD, is an employee of Vector Psychometric Group, LLC (VPG). VPG received compensation for this work from Avanir Pharmaceuticals, Inc., Aliso Viejo, CA. Kenneth J. Shulman, MD, is an employee of Avanir Pharmaceuticals, Inc. Stephen D. Silberstein, MD, receives, or has received, honoraria from Alder Biopharmaceuticals. Allergan, Inc., Amgen. Avanir Pharmaceuticals, Inc., Curelator, Inc., Depomed. Dr. Reddy’s Laboratories. eNeura Inc., electroCore Medical, LLC. INSYS Therapeutics. Labrys Biologics. Lilly USA, LLC. Medscape, LLC. Medtronic, Inc., Neuralieve; NINDS. Pfizer, Inc., Supernus Pharmaceuticals, Inc., Teva Pharmaceuticals. Theranica. and Trigemina, Inc., as a consultant and/or advisory panel member. R.J. Wirth, PhD, is an employee of Vector Psychometric Group, LLC (VPG). VPG received compensation for this work from Avanir Pharmaceuticals, Inc., Aliso Viejo, CA. Dawn C. Buse, PhD, has received grant support and honoraria from Allergan, Amgen, Avanir, Biohaven, Eli Lilly and Promeius. She is an employee of Montefiore Medical Center, which has received research support from Allergan, Alder, Avanir, CoLucid, Dr. Reddy’s, Endo Pharmaceuticals, GlaxoSmithKline, Labrys, MAP Pharmaceuticals, Merck, NuPathe, Novartis, Ortho-McNeil, and Zogenix, via grants to the National Headache Foundation and/or Montefiore Medical Center. She is on the editorial board of Current Pain and Headache Reports, the Journal of Headache and Pain, Pain Medicine News, and Pain Pathways magazine. Funding: Manuscript Funding Provided by Avanir Pharmaceuticals, Inc., Aliso Viejo, CA. Headache 3 nausea, and other associated symptoms when given by oral and non-oral routes.9 People who experience nausea at migraine baseline are less likely to respond to oral triptans,13 and nausea associated with a migraine may lead individuals to delay or avoid oral treatment.9 In some cases, triptans may actually cause treatment emergent nausea (TEN).14,15 For these reasons, migraine treatment guidelines recommend that nonoral formulations should be considered for migraine attacks associated with nausea and/or vomiting.16 Most clinical trials assessing nausea compare the proportion of patients free of the symptom 2 hours post-treatment with an active drug or a comparator. Though this approach demonstrates the benefits of triptan therapy, it does not differentiate outcomes in attacks with nausea at baseline from those without nausea at baseline. In attacks with nausea at baseline, it is helpful to measure relief of nausea. Attacks free of nausea at baseline may remain nausea free or develop TEN. TEN may arise either as a side effect of medication or as a consequence of an evolving migraine attack. Investigators have attempted to disaggregate nausea relief in those who have it at baseline and in those free of it at baseline who experience the emergence of nausea following treatment. In a post hoc analysis of several clinical trials, Lipton et al examined the rates of TEN for a number of triptans and placebo.13 TEN occurred in about onefifth of patients treated with oral sumatriptan and in just one-tenth of placebo-treated patients.17 The hypothesis that this is a sumatriptan-related adverse effect which may arise from direct gastric effects of medication on the gut17 is consistent with the absence of TEN when sumatriptan is administered by non-oral routes, including subcutaneous injection and delivery via an iontophoretic patch.15,18 Under this hypothesis, we would predict that other nonoral routes of sumatriptan administration should also result in low rates of TEN. A non-oral delivery system for sumatriptan is provided by AVP-825 (OnzetraV XsailV ). This Breath PoweredV intranasal delivery system contains 22 mg of sumatriptan as a dry powder and uses the patient’s own breath to deliver medication beyond the nasal valve to the upper posterior nasal R R R cavity, a large area lined with richly vascular mucosa conducive to rapid drug absorption into the systemic circulation.19 Compared to sumatriptan 100 mg oral tablet and 20 mg nasal spray, AVP-825 has a shorter time to maximal plasma concentration (Tmax) and greater exposure in the first 15 minutes (AUC0-15 minutes) post-dose. This PK profile suggests that AVP-825 is a more efficient way of delivering sumatriptan, which may offer more rapid relief of migraine pain than oral sumatriptan, a conclusion directly supported by the COMPASS and TARGET trials.20-23 COMPASS was a randomized, double blind, double-dummy, comparative efficacy study that contrasted two active treatments, AVP-825 and oral sumatriptan tablets (100 mg). Participants treated up to five attacks with each treatment and assessed both pain and nausea at multiple early endpoints. We conducted a post hoc analysis of data from the COMPASS study to characterize the longitudinal trajectories of nausea from three unique perspectives, and to compare treatment differences between AVP-825 and oral sumatriptan using two comparative efficacy hypotheses: (1) AVP-825 will more rapidly reduce nausea compared to oral sumatriptan, and (2) Oral sumatriptan will produce higher rates of TEN than AVP-825. The COMPASS study is ideal to test these hypotheses for several reasons. First, AVP-825 is rapidly absorbed and so might be expected to rapidly relieve nausea. Second, AVP-825 is primarily absorbed across the mucosa of the posterior nasopharynx, avoiding TEN by circumventing the gut more than oral therapy. Third, the unique repeated attack, crossover study design instructed participants to treat up to five attacks with each active treatment. This enables us to obtain more stable estimates of treatment effects and to partition the total variation in nausea into components due to within person and between person differences. METHODS COMPASS Study Design.—Detailed methodology for the COMPASS study (NCT01667679, clinicaltrials.gov) has been previously published.20 Briefly, the COMPASS study was a randomized, 4 Month 2017 AVP-825 + Placebo Tablet AVP-825 + Placebo Tablet Fig. 1.—Study flowchart. double-blind, double-dummy, multiple attack, active-comparator crossover study conducted at 20 outpatient headache centers throughout the United States from August 2012 to March 2014. This study was approved by New England Independent Review Board (NEIRB.com, Newton, MA) as well as additional Institutional Review Boards at three institutions (Mayo Clinic, Cleveland Clinic, and Thomas Jefferson University). Written informed consent was obtained from each subject prior to protocol-related activities. Participants were randomized 1:1 to one of two blinded treatment sequences, each of which had two 12-week treatment periods. For treatment sequence 1, patients treated up to five attacks in the first treatment period with AVP-825 and a placebo tablet (Ranbaxy Pharmaceuticals Inc., Jacksonville, FL). After 3 months or five attacks treated (whichever was first), sequence 1 patients began the second treatment period and received an AVP-825 placebo (containing lactose powder) plus a 100 mg sumatriptan tablet. For participants in treatment sequence 2, treatment order was reversed as depicted in Figure 1. Qualifying migraine attacks had to meet ICHD-2 criteria for migraine24 and had to have pain of at least mild intensity. Treatment within 1 hour of attack onset was required, and subjects recorded outcomes for each attack in an electronic diary. A second dose of study drug was allowed after the 2-hour assessments were completed and up to 24 hours after the first dose if there was no relief or the headache worsened/recurred. Rescue medication was also permitted if there was no relief or headache worsened/recurred 2 hours or more after the second dose. Subjects were withdrawn from the study for failure to treat any migraine headaches over the 12 weeks of treatment period 1 or treatment period 2. Subjects.—Participants were recruited from outpatient headache centers in the US, 18-65 years old, met International Classification of Headache Disorders, 2nd Edition, 1st revision (ICHD-II) criteria for migraine with or without aura24 and had a migraine attack frequency from two to eight per month with <15 headache days per month for 1 year prior to screening. Participants were excluded if they were unable to distinguish migraine from other headache types, if they had 15 or more days per month, or if they had a history of nonresponse to an adequate dose and duration of treatment with sumatriptan or two other triptans. Persons whose Headache migraine preventive medications had been adjusted in the past 4 weeks were also excluded. Other medical and psychological conditions that may affect patient safety or study outcomes also led to exclusion. Variables.—The primary outcome for these analyses was a binary nausea measure (0 5 No nausea, 1 5 Nausea) reported at pre-dose and 10, 15, 30, 45, 60, 90, and 120 minutes post-dose for each attack. Covariates included in the statistical models were treatment, treatment period, treatment sequence, age, gender, race, and time. Analyses.—The current study utilized all available data to examine longitudinal change in nausea. All analyses were conducted using PROC GLIMMIX in SAS 9.4. Three-level logistic multilevel models, estimated with maximum likelihood, evaluated how nausea unfolds over time in individuals across multiple attacks. The three-level multilevel models explicitly accounted for the COMPASS study design, which had repeated measures (level 1) nested within each attack (level 2), and multiple attacks nested within each subject (level 3). Each of the fitted multilevel models included both random attack (level 2) and random subject effects (level 3). These random effects allowed nausea to vary both across attacks (within subjects) and between subjects. These models result in subjectspecific interpretations; see further details in the work by Fitzmaurice et al. and Hedeker and Gibbons.25,26 Three separate statistical models were used to examine unique aspects of nausea. Model 1 examined longitudinal change in Overall Nausea. This model tested how patients’ propensity for nausea changed over the first 2 hours (pre-dose through 2 hours post-dose) across treatments regardless of whether or not they reported baseline nausea. Model 2 examined TEN in eligible attacks free of nausea at baseline to explore whether or not nausea developed following treatment. This model tested treatment differences in TEN from 10 minutes through 2 hours post-dose, since eligible attacks were without baseline nausea. Model 3 examined Nausea Relief, testing treatment differences in nausea from 10 minutes through 2 hours 5 for attacks with baseline nausea. For all models, we examined several function forms for nausea trajectories over time including linear, polynomial (eg, quadratic, cubic), and spline (piecewise) trends. We ultimately used polynomial time trends for all three models: The Overall Nausea model (Model 1) used a quadratic time trend while the TEN (Model 2) and Nausea Relief models (Model 3) used a cubic functional form for time. The polynomial terms allowed for nonlinear changes in nausea over time. More specifically, the quadratic trend (squaredterm) permitted rate of change to differ over time (ie, acceleration/deceleration) and the cubic trend allowed for change in acceleration/deceleration (ie, in physics this is referred to as a “jerk”). Thus, these polynomial terms more accurately captured how nausea risk was changing throughout the course of attacks. A benefit of our modeling strategy was that the nonlinear longitudinal changes in nausea could be described using instantaneous rates of change (IROC), which represent how much nausea is changing for an individual at a given point in time. A superior treatment should show faster rates of decrease in nausea which is represented by a more negative IROC. Consistent with standard analyses of clinical trial data, we also test model-implied treatment differences of an individual’s odds of nausea at each fixed time-point (ie, level differences). Using multilevel modeling, the total variance in nausea could also be separated into that which is due to within attack (level 1), between attack (within subject) (level 2), and between subject (level 3) differences. For instance, the proportion of variance between attacks within subjects is ð2Þ ð3Þ ð2Þ p2 s00 = s00 1s00 1 3 and the proportion of variance 2 ð3Þ ð3Þ ð2Þ ð3Þ between subjects is s00 = s00 1s00 1 p3 , where s00 ð2Þ is the level 3 subject random effect variance, s00 is 2 the level 2 attack random effect variance, and p3 is the variance of the standard logistic distribution (ie, level 1 repeated measures); see the work by Hedeker and Gibbons26 or Raudenbush et al27 for more details on variance decomposition. In the current study, we computed these values based on the estimated variance components from the final conditional models. 6 Month 2017 Table 1.—Demographic Characteristics for Each Subsample Population N Model 2: TEN Model 3: Nausea Relief Individuals with no Individuals with nausea Model 1: Overall Nausea nausea at baseline at baseline for at least Total sample for at least one attack one attack N 5 259 N 5 232 N 5 167 Mean (SD) or % Mean (SD) or % Mean (SD) or % Sociodemographics Age Male White Historic symptoms prior to study Migraine attacks per month (past 12 mo.) Has monthly migraine moderate severity Historic pain intensity of treated migraine headaches None Mild Moderate Severe Migraine type (past 12 mo.) Aura only With aura Without aura Presence of (past 6 mo.) Vomiting Nausea None Mild Moderate Severe Photophobia None Mild Moderate Severe Phonophobia None Mild Moderate Severe Attacks treated attacks during the COMPASS study Number of attacks per patient Total attacks RESULTS Demographics.—Table 1 shows the demographic characteristics for the samples used for each of the three nausea models. The Overall Nausea model (Model 1) included 259 individuals with an average of 6.8 attacks each for a total of 1751 attacks. The TEN model (Model 2) included attacks free of nausea at baseline; there were 232 individuals with an average of 4.9 nausea free attacks each for a total of 1146 attacks. The Nausea Relief model (Model 40.0 (12.3) 15.4 78.4 40.2 (12.1) 15.9 78.4 40.5 (12.4) 12.0 79.0 4.9 (1.9) 100.0 4.9 (1.9) 100.0 4.9 (1.9) 100.0 5.0 43.0 40.7 11.2 5.2 42.0 41.6 11.3 5.4 39.8 42.8 12.1 1.2 32.8 77.2 0.8 33.6 75.9 1.2 33.5 77.8 35.9 33.6 43.7 15.8 32.4 37.8 13.9 17.7 34.5 35.3 12.5 4.8 29.3 45.5 20.4 1.5 12.7 42.1 43.6 1.7 14.2 40.1 44.0 1.2 12.0 38.3 48.5 5.0 17.8 43.6 33.6 5.2 17.2 44.8 32.8 5.4 16.2 41.9 36.5 6.8 (3.3) 1751 4.9 (3.0) 1146 3.5 (2.6) 579 3) included attacks with nausea at baseline; there were 167 individuals with an average of 3.5 attacks each for a total of 579 attacks. Across the three models, samples had similar distributions of demographic characteristics, where the mean age of participants was 40, 12%-16% were male, and 78%79% were white. All participants reported monthly migraines of at least moderate severity over the 12 months prior to entering the study, but only 52%55% of participants reported taking medication Headache 7 b: Treatment Emergent Nausea 1.5 1.5 0.5 0.5 Nausea Nausea a: Overall Nausea -0.5 -1.5 -0.5 -1.5 -2.5 -2.5 -3.5 -3.5 0 10 20 30 40 50 60 70 80 90 100 110 120 10 20 30 40 50 60 70 80 90 100 110 120 Time Time Oral Sumatriptan Oral Sumatriptan AVP-825 1.5 AVP-825 c: Nausea Relief Nausea 0.5 -0.5 -1.5 -2.5 -3.5 10 20 30 40 50 60 70 80 90 100 110 120 Time Oral Sumatriptan AVP-825 Fig. 2.—(a-c) Observed changes in log-odds of nausea by treatment pooling over subjects and attacks. when migraine pain intensity was moderate or severe. In the 6 months prior to entering the study, 48%-66% reported moderate or severe nausea, 84%-87% reported moderate or severe photophobia, and 77%-78% reported moderate or severe phonophobia. Model 1: Overall Nausea.—The observed logodds of pre-dose nausea were similar in the AVP825 and oral sumatriptan (see Fig. 2a). However, the change in nausea over time for attacks treated with AVP-825 appeared nonlinear, decreasing faster at earlier time points and then slowly decelerating as time approached 2 hours. For oral sumatriptan, nausea decreased in a linear fashion from pre-dose through 2 hours post-treatment. Consistent with the observed data shown in Figure 2a, change in nausea was modeled using a quadratic time trend. Table 2 shows the full set of results for the Overall Nausea model. There was significant between subject and between attack variability in nausea (P < .0001 for both random effect variances). Almost one-half of the total variability in nausea was between person, nearly one-third was between attacks (within person), and just under 20% was within attacks. Examining the fixed effects showed a significant treatment-by-time interaction, which indicated that an individual experienced a more rapid decline in their odds of nausea earlier during an attack when treating with AVP-825 compared to oral sumatriptan; v2(2) 5 34.67, P < .0001. This was demonstrated by significantly faster IROCs, where a subject’s nausea risk decreased at a greater rate for AVP-825 relative to oral sumatriptan at 10 minutes through 60 minutes (P <. 05 for all) (Fig. 3a and Table 3). It was not until 90 minutes that the two treatments had similar rates of change over time. By 2 hours, oral sumatriptan was decreasing at a significantly faster rate compared to AVP-825 (P < .05). Table 4 shows the model-implied treatment differences in nausea odds at each fixed time-point from pre-dose through 2 hours. Paralleling the findings regarding change in nausea over time, an individual’s odds of nausea during a typical attack were significantly reduced at all time-points from 30 minutes through 2 hours following treatment with AVP-825 vs oral sumatriptan (P < .05 for all 0.35 0.17 0.60 0.52 0.02 0.43 0.19 0.02 0.004 0.02 0.005 1.17 0.46 3.38 20.07 21.59 20.31 20.01 0.16 20.81 20.29 0.005 20.14 0.02 8.67 5.57 SE 8096.78 8122.78 8169.02 7.43 12.11 29.56 20.41 22.63 20.60 20.84 0.37 24.21 217.98 1.34 25.62 3.89 - Test stat. <.0001 <.0001 <.0001 .68 .009 .55 .40 .71 <.0001 <.0001 .18 <.0001 .0001 - P 2.85 6.10 24.58 20.37 20.88 20.39 20.01 0.64 21.07 0.09 20.08 0.006 20.33 0.05 0.0006 Est. .67 1.02 0.39 0.27 0.54 0.47 0.02 0.39 0.31 0.05 0.02 0.003 0.08 0.02 0.004 SE 2533.81 2563.81 2615.51 4.28 5.99 211.81 21.35 21.63 20.84 20.61 1.66 23.42 1.88 24.99 2.40 23.88 1.82 0.13 Test stat. <.0001 <.0001 <.0001 .18 .10 .40 .54 .10 .0006 .06 <.0001 .02 .0001 .07 .89 P Model 2: Treatment Emergent Nausea 9.39 4.91 2.79 0.36 21.13 20.87 20.02 0.67 20.57 21.58 0.15 20.006 - Est. 1.66 0.65 0.46 0.26 0.86 0.70 0.02 0.57 0.26 0.14 0.03 0.002 - SE 2985.84 3009.84 3047.26 5.66 7.55 6.03 1.39 21.31 21.24 20.93 1.16 22.18 211.65 4.90 22.97 - Test stat. Model 3: Nausea Relief <.0001 <.0001 <.0001 .16 .19 .22 .35 .24 .03 <.0001 <.0001 .003 - P Est. 5 Estimate; SE=Standard Error; 22LL 5 22 Log Likelihood; AIC 5 Akaike information criterion; BIC 5 Bayesian information criterion. Reference coding was used for categorical predictors. Age is centered at 40 years old (Age 5 Age-40) and time is centered at 45 minutes and divided by 10 ([Time – 45]/10). For the fixed effects (top portion of the table), the test statistics were t values and, for the random effects (bottom portion), the test statistics were z values. For Model 1, the intercept has df 5 254 and other effects had df 5 11,648. For Model 2, the intercept has df 5 227 and other effects had df 5 6584. For Model 3, the intercept has df 5 162 and other effects had df 5 3334. Fixed Effects Intercept Period 2 (ref 5 Period 1) Male (ref 5 Female) Minority (ref 5 White) Age Sequence 1 (ref 5 Seq. 2) AVP-825 (ref 5 Oral Suma) Time Time2 (squared) Time3 (cubed) AVP-825 3 Time AVP-825 3 Time2 (squared) AVP-825 3 Time3 (cubed) Random Effects Subject variance Attack variance Model Fit Statistics 22LL AIC BIC Est. Model 1: Overall Nausea Table 2.—Parameter Estimates, Standard Errors, Test Statistics, and P-Values for the Fitted Nausea Models 8 Month 2017 Headache 9 a: Overall Nausea 1 0.5 IROC 0.5 IROC b: Treatment Emergent Nausea 1 0 0 -0.5 -0.5 -1 -1 0 10 10 20 30 40 50 60 70 80 90 100 110 120 20 30 40 Oral Sumatriptan 50 60 70 80 90 100 110 120 Time Time Oral Sumatriptan AVP-825 AVP-825 Fig. 3.—(a,b) Treatment differences in IROCs. More negative IROC values on the y-axis indicate greater reductions (or slower increases) in nausea at that specific time-point; thus, lower values are better. For “Overall Nausea,” IROCs were significantly smaller at each time-point from 10 minutes through 60 minutes using AVP-825 compared to oral sumatriptan. For TEN, IROCs were significantly smaller at all time-points from 10 minutes through 45 minutes using AVP-825, and the treatment differences at 90 and 120 minutes were not statistically significant. Key: Smaller values are better. IROC 5 instantaneous rate of change. time-points). The odds of nausea were not significantly different from pre-dose through 15 minutes between treatments (P > .05). Thus, treating with AVP-825 produced faster IROCs earlier on, which led to significant treatment differences over time in odds of nausea at the fixed time-points compared to oral sumatriptan. Model 2: Treatment Emergent Nausea.—Figure 2b shows a plot of the observed TEN data over time stratified by treatment. TEN changed in a nonlinear fashion for both AVP-825 and oral sumatriptan, but the course of change differed. AVP-825 appeared to change in a cubic manner, with TEN increasing earlier (10-30 minutes) followed by reduction (30-60 minutes) and ultimately leveling out at a low level (60-120 minutes). However, the longitudinal change in TEN for attacks treated with oral sumatriptan had a more quadratic form, increasing through the first hour and decreasing through the second hour. Figure 2b shows that the odds of nausea in the first hour decreased for AVP-825 while oral sumatriptan increased. It was not until 2 hours post-dose that the treatments began to show more comparable levels of nausea. Results from the three-level TEN multilevel model (Table 2) showed significant between subject and between attack variability (P < .0001). About Table 3.—Model-Implied Treatment Differences in Instantaneous Rates of Change (IROC) in Nausea Model 1: Overall Nausea BL 10 minutes 15 minutes 30 minutes 45 minutes 60 minutes 90 minutes 120 minutes Model 2: Treatment Emergent Nausea Oral Suma IROC Est. AVP-825 IROC Est. t value P value Oral Suma IROC Est. AVP-825 IROC Est. t value P value 20.33 20.32 20.32 20.30 20.29 20.27 20.24 20.21 20.66 20.61 20.58 20.50 20.42 20.34 20.19 20.03 25.12 25.30 25.40 25.74 25.62 23.23 1.27 2.47 <.0001 <.0001 <.0001 <.0001 <.0001 .001 .20 .01 0.89 0.75 0.38 0.09 20.11 20.26 20.07 0.26 0.16 20.08 20.23 20.30 20.14 0.37 22.29 22.74 25.24 23.88 21.72 1.55 1.33 .02 .006 <.0001 .0001 .09 .12 .18 IROC 5 instantaneous rate of change. More negative IROC estimates implies that nausea is decreasing at faster rate (or increasing at a slower rate) at that point in time (eg, better outcome). For Model 1, the df for all statistical tests were 11,648. For Model 2, the df for all statistical tests were 6584. Model 3 is not listed because the treatment-by-time interaction was not statistically significant. 10 Month 2017 Table 4.—Model-Implied Treatment Differences in Odds of Nausea at Each Time-Point Model 1: Overall Nausea BL 10 minutes 15 minutes 30 minutes 45 minutes 60 minutes 90 minutes 120 minutes Model 2: Treatment Emergent Nausea OR 95% CI t value P value OR 95% CI t value P value 1.27 0.93 0.81 0.57 0.45 0.38 0.37 0.53 0.87-1.86 0.66-1.32 0.58-1.14 0.41-0.81 0.31-0.65 0.26-0.57 0.25-0.56 0.30-0.94 1.23 20.39 21.2 23.12 24.21 24.79 24.67 22.16 .22 .70 .23 .002 <.0001 <.0001 <.0001 .03 1.85 1.37 0.62 0.34 0.23 0.21 0.48 0.89-3.82 0.73-2.54 0.34-1.14 0.19-0.63 0.12-0.44 0.09-0.47 0.19-1.18 1.66 0.98 21.53 23.42 24.56 23.73 21.60 .10 .33 .13 .0006 <.0001 .0002 .11 OR 5 odds ratio. Odds ratios less than 1 indicate that AVP-825 was associated with reduced odds of being nauseous compared to oral sumatriptan (better outcome). For Model 1, the df for all statistical tests were 11,648. For Model 2, the df for all statistical tests were 6584. Model 3 is not listed because the treatment-by-time interaction was not statistically significant. The model-implied odds ratio was 0.57 in favor AVP-825 across all time-points. one-quarter of the total variance in TEN was between person, one-half was attributable to differences between attacks, and the remainder (25%) was due to within attack differences. The significant variance components demonstrated that TEN varied not only across individuals but also within individuals, across attacks. Consistent with the observed data plot in Figure 2b, there was also a significant treatment-by-time interaction, which indicated that change in nausea over time differed across treatments, v2(3) 5 35.64, P < .0001. Examination of IROC helps show the nature of treatment differences in longitudinal change in TEN (Fig. 3b and Table 3). Most notably, an individual’s propensity for TEN was increasing at a significantly faster rate with oral sumatriptan compared to AVP-825 at 10, 15, 30, and 45 minutes (P < .05 for all). Further, at 30 and 45 minutes the odds of experiencing nausea were still increasing for oral sumatriptan, but they were decreasing for AVP-825. From 60 minutes to 2 hours, TEN was changing in a comparable fashion across both treatment groups (P > .05). Table 4 shows that there were treatment differences in an individual’s overall odds of nausea at several time-points. Specifically, from 45 minutes through 90 minutes, an individual had significantly lower odds of TEN when treating an attack with AVP-825 compared to oral sumatriptan (P < .05 for all). There were no significant treatment differences in odds of TEN during the first 30 minutes and at the 2 hour assessment (P > .05 for all). Model 3: Nausea Relief.—The observed data suggested that AVP-825 had a constant level difference in nausea from 10 minutes through 2 hours for attacks with pre-dose nausea (Fig. 2c). However, there did not appear to be treatment differences in the rate of change in nausea over time (the lines run parallel). In Model 3, longitudinal change in Nausea Relief was modeled using a cubic time trend and demonstrated significant between subject and between attack (within subject) variability in nausea (P < .0001), with more than half of the total variability in nausea being attributable to between subject differences compared to just over onequarter being due to between attack (within subject) differences. As expected from the observed data, the treatment-by-time interaction was not statistically significant, v2(3) 5 2.40, P 5 .49. This suggested that the change in nausea over time was comparable for AVP-825 and oral sumatriptan. For parsimony, the treatment-by-time interaction terms were dropped for the final model displayed in Table 2. Although there were not significant Headache treatment differences in the rate of change in nausea over time, in individuals treating an attack with AVP-825, the odds of experiencing nausea were reduced by 43% from 10 minutes through 2 hours (OR 5 0.57, P < .05). These three statistical models demonstrated that an individual’s nausea risk differed by treatment. All three models showed that an individual’s odds of nausea (Overall, TEN, and Nausea Relief) were significantly lower at several time-points in the first 2 hours post-dose. Further, for Overall Nausea and TEN, rates of change in nausea over time differed significantly, with the more favorable longitudinal trends produced when attacks were treated with AVP-825 compared to oral sumatriptan. DISCUSSION We evaluated how nausea unfolds over time across multiple attacks treated with AVP-825 vs oral sumatriptan in a multi-attack, crossover study. We first examined the time course of nausea following treatment in the overall sample, independent of nausea status at baseline. Next, we examined TEN in attacks free of nausea at baseline. Finally, we examined relief of nausea in those attacks that had nausea at baseline. We will consider these results one at a time. In the overall sample, nausea declined more rapidly with AVP-825 than with oral sumatriptan. There was faster instantaneous rate of decline over the first hour post-treatment with AVP-825. When not stratifying by baseline nausea status, attacks treated with AVP-825 showed greater reductions in nausea than attacks treated with oral sumatriptan. Unlike most studies, the present report involved multiple attacks for each person with each treatment. Thus, we can partition the total nausea variance into between and within person components. We estimated that person to person variation accounted for about half of the total variation in nausea. Whereas between-attack variation within individuals accounted for about one-third of the total variation and just under 20% of the total variability in nausea was attributed to within attack differences. This means that the nausea trajectories were influenced by both within and between person 11 factors. While many prior reports show that triptans effectively relieve nausea, we are not aware of studies able to partition the total nausea variance in a multiple attack context.14 Next, we found that rates of TEN were higher with oral sumatriptan than with AVP-825. TEN trajectories for AVP-825 increased slightly through the first 15 minutes post-dose then decreased for the next hour (negative IROCs from 30 through 90 minutes). Sumatriptan tablets were associated with a more prolonged increase in odds of TEN, which lasted through the first 45 minutes and did not start to decline until 60 minutes following treatment. However, because there was no placebo treatment arm for comparison, it is not known whether this difference is due to the fact that AVP-825 treats the nausea associated with a developing migraine better than oral sumatriptan, or that oral sumatriptan increases TEN through a direct effect on the gut. Just under one-quarter of the total variation in TEN status was attributed to between person differences, about half was due to between attack differences, and the remainder (25%) was associated with within attack differences. Rates of TEN following oral sumatriptan 100 mg have rarely been studied. The rate of TEN was 20% for sumatriptan 100 mg vs 12% for placebo treated subjects in a published post hoc analysis.13 The authors suggested that oral sumatriptan may cause TEN through a direct effect on the gut. This effect is biologically plausible as sumatriptan is a sulfonamide, a medication class known to irritate the gut.17 Bigal et al examined the effect of a sumatriptan cutaneous patch and showed that it did not cause TEN.18 Similarly, there is no evidence that injectable sumatriptan causes TEN.15 Combining results from these studies, we suggest that oral sumatriptan tablets may cause TEN while sumatriptan administered through non-oral routes does not. Finally, the third model tested profiles of Nausea Relief only in persons who had it at baseline. Findings showed that nausea improved following either of the two treatments, but did not result in different rates of change over time (ie, similar nausea trajectories across treatments). However, there was a difference in overall level, suggesting that an 12 individual’s odds of experiencing nausea were consistently reduced through 2 hours with AVP-825 compared to oral sumatriptan. We have shown advantages of using multilevel modeling to estimate nausea trajectories. The models disaggregated nausea variability based on a three-level nesting structure that is reflective of the real world: repeated measures (level 1) nested within attacks (level 2) nested within persons (level 3). Further, these models revealed information that might not otherwise be apparent. Results provided insights on both the relative speed (eg, Is an individual’s nausea risk decreasing faster using Treatment A or B over time?) and level (eg, At 60 minutes, are an individual’s odds for experiencing nausea elevated when treating an attack with Treatment A compared to B?) differences across treatments. Taken together, these models are well suited to test clinical theory and produce insights that are directly applicable to practice. This study has several limitations. The COMPASS study did not include a placebo arm, so we do not have a contemporaneous control group with which to make contrasts. However, in a multiple attack, crossover study it might be difficult or even unethical to ask participants to treat up to five attacks with placebo. In addition, repeated placebo administration might attenuate participation even in the arms where active treatment is offered. When we partition attacks into those without nausea at baseline (model 2) and with nausea at baseline (model 3) the sample size in terms of persons and attacks are reduced with consequent reductions in statistical power. These reductions may impact the precision of our estimates, especially the random effects variance components. However, based on the model performance, the latter did not appear problematic. The results of this study combined with previous research may help clinicians optimize migraine treatment. Oral sumatriptan causes TEN while non-oral sumatriptan, whether administered by injection, cutaneous patch, or breath-powered intra-nasal delivery appears less likely to cause TEN. Results showed that nausea risk varied substantially across subjects and attacks. Future studies should investigate potential attack-level and subject-level covariates that may Month 2017 help explain this variation. Future research should also investigate how other migraine-related symptoms develop and resolve over time in conjunction with nausea. It would be of interest to evaluate the sequential unfolding of relief of pain, migrainerelated disability, and associated symptoms of migraine. That work may provide insight into the mechanisms through which triptans relieve the broader set of symptoms that define migraine. Acknowledgments: Authors would like to thank the COMPASS Study Site Primary Investigators: Roger Cady, MD, Andrew Cutler, MD, Marshall Freeman, MD, Gordon Gilson, MD, Jerome Goldstein, MD Rashmi Halker Singh, MD, Judith Kirstein, MD, David Kudrow, MD, Peter McAllister, MD, Laszlo Mechtler, MD, John Rubino, MD, Joel Saper, MD, Stephen Silberstein, MD, Timothy Smith, MD, Egilius Spierings, MD, PhD, Cynthia Strout, MD, Stewart Tepper, MD, Jonathan Wilson, MD, Alan Kravitz, MD, and Paul Winner, MD. The authors would like to thank Prescott Medical Communications Group, Chicago, IL, and Emmanuelle Hugentobler, MD, for editorial and administrative support. STATEMENT OF AUTHORSHIP Category 1 (a) Conception and Design Richard B. Lipton, James S. McGinley, neth J. Shulman, Stephen D. Silberstein, Wirth, Dawn C. Buse (b) Acquisition of Data Kenneth J. Shulman (c) Analysis and Interpretation of Data Richard B. Lipton, James S. McGinley, neth J. Shulman, Stephen D. Silberstein, Wirth, Dawn C. Buse KenR. J. KenR. J. Category 2 (a) Drafting the Manuscript Richard B. Lipton, James S. McGinley, Kenneth J. Shulman, R. J. Wirth, Dawn C. Buse (b) Revising It for Intellectual Content Richard B. Lipton, James S. McGinley, Kenneth J. Shulman, Stephen D. Silberstein, R. J. Wirth, Dawn C. Buse Headache 13 Category 3 (a) Final Approval of the Completed Manuscript Richard B. Lipton, James S. McGinley, Kenneth J. Shulman, Stephen D. Silberstein, R. J. Wirth, Dawn C. Buse 12. 13. REFERENCES 1. Headache Classification Committee of the International Headache Society (IHS). The International Classification of Headache Disorders, 3rd edition (beta version). Cephalalgia. 2013;33:629-808. 2. Lipton RB, Buse DC, Saiers J, Fanning KM, Serrano D, Reed ML. Frequency and burden of headache-related nausea: Results from the American Migraine Prevalence and Prevention (AMPP) study. Headache. 2013;53:93-103. 3. Lipton RB, Buse DC, Saiers J, Serrano D, Reed ML. Healthcare resource utilization and direct costs associated with frequent nausea in episodic migraine: Results from the American Migraine Prevalence and Prevention (AMPP) Study. J Med Econ. 2013;16:490-499. 4. Reed ML, Fanning KM, Serrano D, Buse DC, Lipton RB. Persistent frequent nausea is associated with progression to chronic migraine: AMPP Study results. Headache. 2015;55:76-87. 5. Bigal ME, Liberman JN, Lipton RB. Age-dependent prevalence and clinical features of migraine. Neurology. 2006;67:246-251. 6. Goadsby PJ, Charbit AR, Andreou AP, Akerman S, Holland PR. Neurobiology of migraine. Neuroscience. 2009;161:327-341. 7. Arruda MA, Guidetti V, Galli F, Albuquerque RC, Bigal ME. Childhood periodic syndromes: A population-based study. Pediatr Neurol. 2010;43: 420-424. 8. Aurora SK, Papapetropoulos S, Kori SH, Kedar A, Abell TL. Gastric stasis in migraineurs: Etiology, characteristics, and clinical and therapeutic implications. Cephalalgia. 2013;33:408-415. 9. Newman LC. Why triptan treatment can fail: Focus on gastrointestinal manifestations of migraine. Headache. 2013;53:11-16. 10. Silberstein SD. Clinical practice guidelines. Cephalalgia. 2005;25:765-766. 11. Lanteri-Minet M, Valade D, Geraud G, Lucas C, Donnet A. Revised French guidelines for the 14. 15. 16. 17. 18. 19. 20. 21. 22. diagnosis and management of migraine in adults and children. J Headache Pain. 2014;15:2. Antonaci F, Dumitrache C, De Cillis I, Allena M. A review of current European treatment guidelines for migraine. J Headache Pain. 2010;11:13-19. Lipton RB, Pascual J, Goadsby PJ, et al. Effect of rizatriptan and other triptans on the nausea symptom of migraine: A post hoc analysis. Headache. 2001;41:754-763. Ferrari MD, Goadsby PJ, Roon KI, Lipton RB. Triptans (serotonin, 5-HT1B/1D agonists) in migraine: Detailed results and methods of a metaanalysis of 53 trials. Cephalalgia. 2002;22:633-658. Pierce M. Oral triptans and nausea: Treatment considerations in migraine. Headache. 2013;53:1720. Silberstein SD. Practice parameter: Evidencebased guidelines for migraine headache (an evidence-based review): Report of the Quality Standards Subcommittee of the American Academy of Neurology. Neurology. 2000;55:754-762. Cipolla G, Sacco S, Crema F, Moro E, De Ponti F. Frigo G. Gastric motor effects of triptans: Open questions and future perspectives. Pharmacol Res. 2001;43:205-210. Bigal ME, Lipton RB, Newman LC, Pierce MW, Silberstein SD. Sumatriptan iontophoretic transdermal system reduces treatment-emergent nausea and is effective in patients with and without nausea at baseline – Results from a randomized controlled trial. Headache. 2015;55:1124-1132. Djupesland PG, Messina JC, Mahmoud RA. Breath powered nasal delivery: A new route to rapid headache relief. Headache. 2013;53:72-84. Tepper SJ, Cady RK, Silberstein S, et al. AVP-825 breath-powered intranasal delivery system containing 22 mg sumatriptan powder vs 100 mg oral sumatriptan in the acute treatment of migraines (the COMPASS study): A comparative randomized clinical trial across multiple attacks. Headache. 2015;55:621-635. Cady RK, McAllister PJ, Spierings EL, et al. A randomized, double-blind, placebo-controlled study of breath powered nasal delivery of sumatriptan powder (AVP-825) in the treatment of acute migraine (the TARGET study). Headache. 2015;55:88-100. Cady R. The pharmacokinetics and clinical efficacy of AVP-825: A potential advancement for 14 acute treatment of migraine. Expert Opin Pharmacother. 2015;16:2039-2051. 23. Cady R. A novel intranasal breath-powered delivery system for sumatriptan: A review of technology and clinical application of the investigational product AVP-825 in the treatment of migraine. Expert Opin Drug Deliv. 2015;12:1565-1577. 24. Headache Classification Subcommittee of the International Headache Society. The International Month 2017 Classification of Headache Disorders, 2nd ed. Cephalalgia. 2004;24:1-160. 25. Fitzmaurice GM, Laird NM, Ware JH. Applied Longitudinal Analysis. New York, NY: Wiley; 2012:998. 26. Hedeker D, Gibbons RD. Longitudinal Data Analysis. New York, NY: Wiley; 2006:451. 27. Raudenbush SW, Bryk AS. Hierarchical Linear Models: Applications and Data Analysis Methods. Thousand Oaks, CA: Sage; 2002:1.