Sage Journals: Discover world-class research

Abstract

Introduction

This study quantified risks of cardiovascular, cerebrovascular, and mortality events among patients with migraine receiving prophylaxis.

Methods

Patients with migraine aged 18–65 years were identified from 2010 through 2015 within a United States administrative claims database. Topiramate initiators during follow-up were propensity score-matched separately to anticonvulsant, cardiovascular treatment, antidepressant, and other prophylactic treatment initiators. Incident outcomes were identified, and hazard ratios were calculated comparing outcome occurrence among topiramate initiators relative to each comparator. A case-control analysis was nested within the full migraine cohort, and odds ratios quantified the association between outcomes and use or non-use of individual prophylactic treatments (anticonvulsants, serotonin norepinephrine reuptake inhibitors, beta blockers, antihypertensives, tricyclic antidepressants, and other prophylactic treatments).

Results

The cohort included 119,243 patients with migraine. The matched topiramate initiators had a lower mortality risk versus antidepressant (hazard ratio: 0.44, 95% CI: 0.24, 0.83) and anticonvulsant initiators (hazard ratio: 0.45, 95% CI: 0.25, 0.84). In the case-control analysis, increased risks of several outcomes were observed with all prophylactic treatments relative to non-use of that treatment (odds ratios range from 1.54 to 7.90, and 95% CIs exclude 1.0) except for topiramate and calcium channel blockers.

Conclusions

Although increased risks for several outcomes were observed with certain prophylactic treatments, the treatments other than topiramate likely represent markers for outcome risk factors that developed or progressed after cohort entry, rather than being a direct effect of the treatments. Factors including migraine severity, frequency, and other treatment indications should be considered in future migraine prophylactic treatment safety assessments.

Keywords

Migraine prophylactic treatment cardiovascular events cerebrovascular events mortality

Introduction

Migraine is associated with the development of cardiovascular and cerebrovascular diseases (1 –6), particularly among young women with migraine with aura (7). More frequent migraines may be associated with increased risk of ischemic stroke in women (8 –10). Associations between migraine and mortality due to cardiovascular and cerebrovascular or other causes have also been observed (11,12). The risk of these outcomes among patients with migraines who use prophylactic treatments to reduce migraine frequency and severity is unclear (4). Topiramate, valproate, valproic acid, beta blockers, calcium channel blockers, tricyclic antidepressants, and botulinum toxins have been prescribed for this purpose, but also have primary indications for other conditions such as epilepsy, obesity, depression, and hypertension. Monoclonal antibodies against calcitonin gene-related peptide (CGRP) or CGRP receptor have demonstrated positive efficacy and safety profiles in clinical trials for migraine prevention (13). The recent approval of the first inhibitor of the CGRP pathway in the US for migraine prevention (14) presents a need for expanded evidence relating to prophylactic treatments and their potential effects on the association between migraines and cardiovascular and cerebrovascular disease and mortality. This observational study quantifies the risks of cardiovascular and cerebrovascular events and mortality among patients with migraine receiving currently-available prophylactic treatments.

Methods

Study design

This observational study included cohort and nested case-control designs. A cohort of patients with migraine was identified from January 2010 through September 2015 using data from a US health care insurer. Incident cardiovascular events (myocardial infarction, myocardial ischemia, unstable angina/coronary revascularization), cerebrovascular events (ischemic stroke, hemorrhagic stroke, transient ischemic attack), and mortality were identified within the cohort. In the cohort analysis, outcome risk was determined among the subset of cohort members who initiated prophylactic treatment during follow-up. In the nested case-control analysis, outcome risk was determined among the full cohort who received prophylactic treatments in the 12 months prior to the outcome or control date.

Data sources

The study was sourced from the Optum Research Database, which includes demographics and pharmacy, medical, and facility claims that provide dates of services, procedures, and their accompanying diagnoses, for members with commercial health plan or Medicare Advantage insurance from United HealthCare. The underlying insured population from which the data are drawn is geographically diverse across the US and comprises approximately 3–4% of the US population. Information is available dating back to 1993 for commercial health plan members and 2006 for Medicare Advantage members. For a subset of insured members with administrative approvals from their health plan, patient identifying information may be accessed to enable medical record requests or linkage to external data sources. Personal identifying information may only be accessed in limited instances where applicable law allows the use of patient-identifiable data, and when appropriate approvals are obtained.

This study was supplemented with data from the National Death Index (NDI) for mortality outcome ascertainment. The NDI, which is maintained by the National Center for Health Statistics (15), is a computerized index of mortality that contains date and cause of death reported from state vital statistics offices. Primary and underlying causes of death are recorded using ICD-10 diagnosis codes. Use of the NDI allowed for a more complete ascertainment of fact and cause of mortality than use of the claims data alone, including deaths that do not generate a corresponding insurance claim (e.g. deaths occurring outside a hospital).

Study population

Cohort eligibility criteria are summarized in Table 1. The cohort entry date was the earliest date on which a member met all eligibility criteria.

Table 1.

Eligibility criteria for migraine cohort entry.

Presence of at least one of the following five criteria:
Criterion 1	At least one inpatient medical claim with a migraine diagnosis (ICD-9-CM diagnosis code 346.xx) AND one of the following, 7–180 days apart: • At least one outpatient or emergency department medical claim with a migraine diagnosis • At least one claim for a dispensing/administration of a triptan or ergotamine
Criterion 2	At least two outpatient or emergency department medical claims with a migraine diagnosis, 7–180 days apart
Criterion 3	At least two claims for a dispensing/administration of a triptan or ergotamine, 7–180 days apart
Criterion 4	At least one outpatient or emergency department medical claim with a migraine diagnosis and one claim for a dispensing/administration of a triptan or ergotamine, 7–180 days apart
Criterion 5	At least one medical claim with a migraine diagnosis associated with a visit to a neurologist
Additional eligibility criteria:
• 18–65 years of age at the time of meeting one of the above criteria
• At least 12 months of continuous enrollment in a commercial health plan prior to meeting one of the above criteria (baseline period starting in 2009 for patients entering the cohort in 2010)
• Complete medical and pharmacy benefits
• Accessible patient identifying information (required for the linkage to the NDI database).
Exclusion criterion:
Medical claim for epilepsy (ICD-9-CM diagnosis code 345.xx) during the baseline period, in order to not include patients who might be using topiramate and other anticonvulsant medications for epilepsy rather than migraine prophylaxis.

ICD-9: International Classification of Diseases, 9^th Revision, Clinical Modification; NDI: National Death Index.

Outcome identification and follow-up

Myocardial infarction, myocardial ischemia, unstable angina, ischemic stroke, hemorrhagic stroke, and transient ischemic attack cases were identified based on ICD-9-CM diagnosis codes in the first (principal) diagnosis position on an inpatient medical claim (Supplemental Table 1). Coronary revascularization procedure codes were used in addition to ICD-9-CM diagnosis codes for identification of unstable angina cases. As risk factors for recurrent events are expected to differ from risk factors for initial events, only initial events were evaluated. Cohort follow-up was censored upon the occurrence of an outcome for subsequent occurrences of the same outcome. The potential for transient ischemic attack to be a precursor to ischemic stroke, and for unstable angina and myocardial ischemia to be precursors to myocardial infarction was taken into consideration. The occurrence of ischemic stroke censored follow-up for subsequent occurrences of transient ischemic attack, and the occurrence of myocardial infarction censored follow-up for subsequent occurrences of myocardial ischemia or unstable angina. Myocardial ischemia and unstable angina outcomes occurring during follow-up were not counted in the analysis if a baseline myocardial infarction claim was present, and transient ischemic attack outcomes occurring during follow-up were not counted if a baseline ischemic stroke claim was present.

Mortality (overall, from cardiovascular or cerebrovascular causes, from trauma, ascribed to drug overdose), was ascertained from the NDI database. The NDI linkage is made on the basis of a probabilistic score, which is the sum of the weights assigned to each of the identifying data elements used in the NDI record match. The weights reflect the degree of agreement between the information on the record submitted by the researchers and the NDI record (16). Linked deaths were included in the study analysis if they occurred within 60 days of disenrollment from the health insurer. ICD-10-coded primary and underlying causes of death were used for cause-specific mortality ascertainment.

Outcome follow-up was censored upon health plan disenrollment ( + 60 days for mortality), death, or the end of the study period (30 September 2015).

Migraine prophylactic treatment exposure

Cohort members were characterized based on exposure to the following migraine prophylactic treatment groups during follow-up: Topiramate, other anticonvulsants (valproate, valproic acid, divalproex, pregabalin, gabapentin, levetiracetam, zonisamide, and carbamazepine), beta blockers, tricyclic antidepressants, serotonin norepinephrine reuptake inhibitors, calcium channel blockers, antihypertensives (lisinopril and candesartan), and other prophylactic treatments (botulinum toxins, cyproheptadine, methysergide, memantine, carisoprodol, clonidine, and guanfacine).

Cohort members included both newly and previously diagnosed migraine patients who may start prophylactic treatments at different stages during follow-up. The use of prophylactic treatments was assessed in the following time frames after cohort entry to allow for flexibility in assessing outcome risk associated with time-varying exposures: a) based on the first prophylactic treatment received during the course of follow-up, provided there was no use of that drug group in the preceding 12 months (prophylactic treatment initiators), and b) in the 12 months (and 3 months) prior to the occurrence of an outcome during the follow-up period (nested case-control study).

Propensity score modeling and matching process for prophylactic treatment initiators

Among the subset of prophylactic treatment initiators, four separate logistic regression models were used to estimate propensity scores predicting initiation of topiramate compared to initiation of a) cardiovascular treatments (consisting of beta blockers, calcium channel blockers, and antihypertensives), b) other anticonvulsants, c) antidepressants (consisting of tricyclic antidepressants and serotonin norepinephrine reuptake inhibitors), and d) other prophylactic treatments. The primary (topiramate) and comparator treatment groups were selected based on preliminary numbers of cohort members using these treatments, with topiramate being the more frequently used, and clinical reasonableness (prophylactic treatments used for similar types of migraines/patients). Additionally, it was possible to restrict the analysis of topiramate users to those without a diagnosis code for epilepsy (the other indication for topiramate treatment) without a substantial reduction in sample size, which created a group of topiramate users who would be most representative of patients with migraine with respect to comorbidities and other risk factors. For each comparison, cohort members who initiated both topiramate and the comparator drug on the same day were excluded from the propensity score process. Cohort members with use of prophylactic treatments other than the initiating drug in the prior 12 months were included in the propensity score process, and the other treatments were accounted for in the propensity score model.

Propensity scores were estimated based on a wide range of covariates ascertained in the 12 months prior to prophylactic treatment initiation. Pre-specified (demographics, comorbidities, measures of healthcare utilization, and treatment history) and empirically-identified (based on the 100 most frequently occurring diagnoses, procedures, and medications dispensed) covariates were considered for inclusion in the logistic regression models as independent variables, and exposure status (topiramate or comparator) as the outcome. Propensity scores were estimated using a two-step approach, where first the pre-specified covariates most predictive of topiramate initiation based on their univariate c-statistic (age, hypertension and hypercholesterolemia) and interaction terms of those variables with calendar year were forced into the regression model, along with sex, geographic region, calendar month, and calendar year. In the second step, other variables, including previous prophylactic treatment use, were chosen from the remaining pre-specified and empirically-identified covariates using stepwise selection with a p-value cut-off of 0.1 for model entry and 0.3 for retention into the model. A list of variables included in the propensity score models is included in Supplemental Table 2.

Topiramate initiators were subsequently matched based on propensity score to initiators in each comparator cohort in a 1:1 ratio, without replacement, using a greedy matching algorithm with a variable caliper (17,18).

Case and control selection for nested case-control study

A case-control study was nested within the migraine cohort to examine prophylactic treatments and risk factors at or near the time of an outcome event. Cases were the incident outcomes arising within the cohort. Controls were randomly selected from the cohort in a way that aimed to sample the person-time giving rise to the cases. A cohort member and a person-day from the full range of cohort follow-up (1 January 2010 through 30 September 2015) were each randomly selected. A person-day was retained as a control if the patient was enrolled in the health plan on the day selected and had not previously developed the outcome of interest (on or prior to the randomly selected day). Controls were chosen in a 100:1 ratio to cases. This ratio was selected to increase the likelihood that even analyses stratified by comorbidities or other treatments would have adequate statistical power.

Analysis

Cohort analysis

Migraine cohort characteristics, including demographics, treatment history, measures of healthcare utilization, and co-morbid conditions recorded in the 12 months leading up to cohort entry are described.

The subcohorts of propensity score-matched topiramate and comparator prophylactic treatment initiators were characterized in the same way as the migraine cohort but anchored to prophylactic treatment initiation. The balance of the distribution of each covariate between the matched topiramate and comparator drug initiators was assessed with absolute standardized differences (difference between the two mean values divided by the standard deviation). Covariates with an absolute standardized difference of ≤0.10 were considered to be balanced and not to represent plausible confounders in the outcome analysis (19,20). Cox proportional hazards regression models were used to estimate the hazard ratio of each outcome among the matched initiators of topiramate versus each of the four comparator drug group initiators. The outcome analysis was restricted to prophylactic treatment initiators without epilepsy or cardiac arrhythmia in the 12 months prior to initiation. Cohort member characteristics are described for the prophylactic treatment initiators who were unable to be matched on propensity score.

Nested case-control analysis

The nested case-control analysis compared cases to controls with regard to comorbidities, treatment history, and each of the migraine prophylactic treatment exposures occurring in the 12 months prior to the case or control date. The analysis was restricted to cases and controls without a claim for epilepsy or cardiac arrhythmia in the 12 months prior to the case or control date to reduce the potential for confounding that might have arisen since cohort entry if patients had received study treatments for these indications rather than migraine prophylaxis. Logistic regression was used to estimate crude and adjusted odds ratios for each outcome according to each prophylactic treatment. Separate multivariate logistic regression models were built to assess the risk of each outcome associated with exposure to each prophylactic treatment relative to no use of that treatment in the 12 months prior to the case or control date. The variables included in the multivariable regression models had strong associations with case/control status based on univariate odds ratios and also represented plausible confounding variables based on a priori knowledge. All of the regression models were adjusted for age group (18–24, 25–34, 35–44, 45–54, 55–65, > 65 years) and sex. Age groups were used in the models in order to account for the nonlinear relationship between the outcomes of interest and age. Regression models comparing topiramate exposure to non-exposure were adjusted for age group, sex, hypertension, depression, and use of the remaining prophylactic treatments occurring within the 12-month period prior to the case or control date, but on or prior to the topiramate dispensing date. The regression models comparing the remaining treatments to non-use of those treatments were adjusted for age group and sex due to the small number of cases within treatment strata.

Results

We identified 119,243 members who met the cohort eligibility criteria (Figure 1) and contributed 212,180 person-years of follow-up time (an average of 1.78 years per person). The cohort members were predominantly female (82.2%) with an average age of 42 years at cohort entry (Table 2). The distributions of the characteristics of the propensity score-matched prophylactic treatment initiators were well balanced among the topiramate group relative to each comparator group (Table 3), with absolute standardized differences for all measured characteristics of less than or equal to 0.10. Among initiators unable to be matched on propensity score, topiramate initiators were generally younger and had fewer treatments with other medications, healthcare utilization encounters, and comorbidities than the initiators of the comparator prophylactic treatment groups (Supplemental Table 3).

Figure 1.

Formation of the migraine study cohort.

Table 2.

Select characteristics of the study cohort with migraine.

	Migraine study cohort (n = 119,243)
Age in years* (mean, standard deviation)	42.4 (11.8)
Age group in years* (n, %)
18–24	9927 (8.3)
25–34	22,155 (18.6)
35–44	32,517 (27.3)
45–54	33,972 (28.5)
55–65	20,672 (17.3)
Female* (n, %)	97,976 (82.2)
Healthcare utilization† (n, %)
Number of physician visits
0–2	13,934 (11.7)
3–6	39,598 (33.2)
≥ 7	65,711 (55.1)
Number of emergency department visits
0	70,865 (59.4)
1	20,516 (17.2)
≥ 2	27,862 (23.4)
Number of hospitalizations
0	108,515 (91.0)
≥1	10,728 (9.0)
Treatment history† (n, %)
Antidepressants	14,905 (12.5)
Opioids	53,834 (45.1)
Prescription NSAIDs	35,786 (30.0)
Antihyperlipidemics	13,484 (11.3)
Hormone replacement therapy	10,967 (9.2)
Co-morbidities† (n, %)
Depression	23,059 (19.3)
Hypertension	25,679 (21.5)
Hypercholesterolemia	30,196 (25.3)
Cardiac arrhythmia	5192 (4.4)
Obesity	10,453 (8.8)
Asthma	11,005 (9.2)
Pregnancy^‡
Delivery indicators	2829 (2.9)
Non-delivery indicators (e.g. ultrasound, obstetrical care)	4399 (4.5)

Assessed on the day of study cohort entry.

†

Assessed within the 12 months prior to and including the day of study cohort entry.

‡

Percentages were calculated using the number of female migraine patients as the denominator.

Table 3.

Retrospective cohort analysis: Select characteristics of the subset of the migraine study population initiating prophylactic treatment occurring in the 12 months prior to treatment initiation, after propensity score matching.

	Comparison 1		Comparison 2		Comparison 3		Comparison 4
	Topiramate (n = 8054)	Cardiovascular (n = 8054)	Topiramate (n = 9337)	Antidepressants (n = 9337)	Topiramate (n = 6940)	Anticonvulsants (n = 6940)	Topiramate (n = 4487)	Other (n = 4487)
Age in years* (mean, standard deviation)	41.7 (11.6)	41.7 (11.7)	41.2 (11.7)	41.2 (11.7)	43.4 (11.2)	43.2 (11.4)	44.5 (11.0)	44.1 (11.3)
Female* (n, %)	1295 (16.1)	1342 (16.7)	7975 (85.4)	7977 (85.4)	5777 (83.2)	5761 (83.0)	3802 (84.7)	3795 (84.6)
Prophylactic treatment history (n, %)
Naïve new user	4119 (51.1)	4090 (50.8)	4859 (52.0)	4828 (51.7)	2773 (40.0)	2920 (42.1)	1403 (31.3)	1457 (32.5)
Switch new user	3935 (48.9)	3964 (49.2)	4478 (48.0)	4509 (48.3)	4167 (60.0)	4020 (57.9)	3084 (68.7)	3030 (67.5)
Healthcare utilization (mean, standard deviation)
Number of physician visits	11.5 (9.9)	11.5 (9.6)	11.6 (9.9)	12.0 (10.7)	12.8 (11.1)	13.3 (11.0)	13.6 (11.6)	14.7 (13.3)
Number of emergency department visits	2.1 (5.0)	2.1 (5.0)	2.1 (5.1)	2.2 (5.3)	2.4 (5.4)	2.5 (5.9)	2.4 (5.5)	2.8 (6.7)
Number of hospitalizations	0.2 (0.5)	0.2 (0.6)	0.2 (0.6)	0.2 (0.5)	0.2 (0.7)	0.2 (0.6)	0.2 (0.7)	0.2 (0.6)
Other medications (n, %)
Opioids	4376 (54.3)	4362 (54.2)	5166 (55.3)	5189 (55.6)	4208 (60.6)	4508 (65.0)	2883 (64.3)	3079 (68.6)
Prescription NSAIDs	2997 (37.2)	3003 (37.3)	3504 (37.5)	3569 (38.2)	2866 (41.3)	2982 (43.0)	1937 (43.2)	2035 (45.4)
Antihyperlipidemics	1021 (12.7)	950 (11.8)	1084 (11.6)	1046 (11.2)	1011 (14.6)	993 (14.3)	722 (16.1)	607 (13.5)
Hormone replacement therapy	761 (9.4)	776 (9.6)	891 (9.5)	916 (9.8)	794 (11.4)	743 (10.7)	606 (13.5)	593 (13.2)
Triptan use†
Eletriptan	1041 (12.9)	1058 (13.1)	NR	NR	NR	NR	NR	NR
Zolmitriptan	287 (3.6)	297 (3.7)	NR	NR	NR	NR	NR	NR
Sumatriptan	NR	NR	3523 (37.7)	3496 (37.4)	2419 (34.9)	2418 (34.8)	1588 (35.4)	1562 (34.8)
Rizatriptan	NR	NR	1156 (12.4)	1185 (12.7)	841 (12.1)	832 (12.0)	585 (13.0)	569 (12.7)
Co-morbidities (n, %)
Depression	1966 (24.4)	1919 (23.8)	2407 (25.8)	2502 (26.8)	1959 (28.2)	1895 (27.3)	1347 (30.0)	1256 (28.0)
Hypertension	2219 (27.6)	2184 (27.1)	2102 (22.5)	2144 (23.0)	1923 (27.7)	1871 (27.0)	1281 (28.5)	1275 (28.4)
Hypercholesterolemia	2136 (26.5)	2105 (26.1)	2323 (24.9)	2324 (24.9)	1985 (28.6)	1997 (28.8)	1347 (30.0)	1328 (29.6)
Cardiac arrhythmia	455 (5.6)	442 (5.5)	485 (5.2)	498 (5.3)	440 (6.3)	448 (6.5)	267 (6.0)	300 (6.7)
Obesity	991 (12.3)	942 (11.7)	1028 (11.0)	1012 (10.8)	835 (12.0)	839 (12.1)	471 (10.5)	486 (10.8)
Asthma	838 (10.4)	801 (9.9)	1004 (10.8)	1012 (10.8)	825 (11.9)	830 (12.0)	549 (12.2)	511 (11.4)
Pregnancy
Delivery indicators	196 (2.4)	181 (2.2)	229 (2.5)	214 (2.3)	166 (2.4)	140 (2.0)	112 (2.5)	69 (1.5)
Non-delivery indicators‡	277 (3.4)	275 (3.4)	330 (3.5)	313 (3.4)	32 (0.5)	29 (0.4)	160 (3.6)	121 (2.7)

NSAIDs: nonsteroidal anti-inflammatory drugs; NR: not reported.

Age and sex were assessed on the day of prophylactic treatment initiation.

†

The triptan medications reported were identified empirically from the most frequently occurring medications dispensed and were selected for inclusion in the propensity score models via stepwise selection.

‡

Non-delivery indicators include indicators of pre-natal care (e.g. ultrasound and other obstetrical care).

Propensity score-matched topiramate initiators had a lower risk of overall mortality versus the matched antidepressant initiators (hazard ratio: 0.44, 95% confidence interval (CI): 0.24, 0.83) and matched anticonvulsant initiators (hazard ratio: 0.45, 95% CI: 0.25, 0.84) (Table 4). No cases of cardiovascular or cerebrovascular mortality were observed among the matched prophylactic treatment initiators.

Table 4.

Retrospective cohort analysis: Outcome incidence rates and relative hazard ratios among the propensity score-matched prophylactic treatment initiators.

Outcomes by prophylactic treatment comparisons	Number of events	Person-years	IR* (95% CI)	Relative HR (95% CI)
Myocardial infarction
Topiramate	12	10,506	1.14 (0.62, 1.93)	0.63 (0.31, 1.30)
Cardiovascular treatments	19	10,512	1.81 (1.12, 2.76)	Reference
Topiramate	11	12,222	0.90 (0.48, 1.56)	0.63 (0.29, 1.34)
Antidepressants	17	12,010	1.42 (0.86, 2.22)	Reference
Topiramate	7	8610	0.81 (0.36, 1.60)	0.58 (0.23, 1.48)
Anticonvulsants	12	8575	1.40 (0.76, 2.37)	Reference
Topiramate	5	5646	0.89 (0.34, 1.94)	0.44 (0.15, 1.27)
Other prophylactic treatments	11	5632	1.95 (1.04, 3.38)	Reference
Myocardial ischemia
Topiramate	3	10,506	0.29 (0.08, 0.76)	0.74 (0.17, 3.33)
Cardiovascular treatments	4	10,510	0.38 (0.13, 0.90)	Reference
Topiramate	4	12,222	0.33 (0.11, 0.78)	3.92 (0.44, 35.06)
Antidepressants	1	12,010	0.08 (0.01, 0.39)	Reference
Topiramate	2	8610	0.23 (0.05, 0.74)	1.99 (0.18, 21.92)
Anticonvulsants	1	8575	0.12 (0.01, 0.54)	Reference
Topiramate	2	5646	0.35 (0.07, 1.14)	1.99 (0.18, 21.90)
Other prophylactic treatments	1	5630	0.18 (0.02, 0.83)	Reference
Unstable angina
Topiramate	8	10,506	0.76 (0.36, 1.44)	0.99 (0.37, 2.65)
Cardiovascular treatments	8	10,511	0.76 (0.36, 1.44)	Reference
Topiramate	7	12,222	0.57 (0.26, 1.12)	1.71 (0.50, 5.84)
Antidepressants	4	12,010	0.33 (0.11, 0.79)	Reference
Topiramate	5	8610	0.58 (0.22, 1.27)	1.67 (0.40, 7.00)
Anticonvulsants	3	8575	0.35 (0.10, 0.93)	Reference
Topiramate	4	5646	0.71 (0.24, 1.68)	0.96 (0.24, 3.85)
Other prophylactic treatments	4	5630	0.71 (0.24, 1.69)	Reference
Ischemic stroke
Topiramate	33	10,478	3.15 (2.21, 4.37)	1.14 (0.69, 1.88)
Cardiovascular treatments	29	10,494	2.76 (1.89, 3.91)	Reference
Topiramate	23	12,208	1.88 (1.23, 2.78)	0.87 (0.50, 1.53)
Antidepressants	26	12,001	2.17 (1.45, 3.13)	Reference
Topiramate	26	8591	3.03 (2.02, 4.37)	0.97 (0.57, 1.66)
Anticonvulsants	27	8561	3.15 (2.13, 4.52)	Reference
Topiramate	17	5636	3.02 (1.82, 4.72)	1.00 (0.51, 1.95)
Other prophylactic treatments	17	5624	3.02 (1.82, 4.73)	Reference
Hemorrhagic stroke
Topiramate	7	10,512	0.67 (0.30, 1.31)	1.16 (0.39, 3.46)
Cardiovascular treatments	6	10,523	0.57 (0.24, 1.18)	Reference
Topiramate	5	12,226	0.41 (0.16, 0.90)	0.41 (0.14, 1.16)
Antidepressants	12	12,019	1.00 (0.55, 1.69)	Reference
Topiramate	6	8614	0.70 (0.29, 1.44)	0.74 (0.26, 2.15)
Anticonvulsants	8	8581	0.93 (0.44, 1.76)	Reference
Topiramate	5	5644	0.89 (0.34, 1.94)	0.72 (0.23, 2.26)
Other prophylactic treatments	7	5635	1.24 (0.55, 2.44)	Reference
Transient ischemic attack
Topiramate	11	10,477	1.05 (0.56, 1.82)	2.76 (0.88, 8.67)
Cardiovascular treatments	4	10,493	0.38 (0.13, 0.91)	Reference
Topiramate	8	12,207	0.66 (0.31, 1.24)	2.62 (0.70, 9.89)
Antidepressants	3	12,000	0.25 (0.07, 0.67)	Reference
Topiramate	9	8589	1.05 (0.52, 1.91)	2.31 (0.71, 7.52)
Anticonvulsants	4	8561	0.47 (0.16, 1.11)	Reference
Topiramate	3	5636	0.53 (0.15, 1.42)	1.47 (0.25, 8.79)
Other prophylactic treatments	2	5624	0.36 (0.07, 1.14)	Reference
Overall mortality
Topiramate	12	10,522	1.14 (0.62, 1.93)	0.52 (0.26, 1.05)
Cardiovascular treatments	23	10,534	0.50 (0.17, 1.47)	Reference
Topiramate	14	12,236	1.14 (0.66, 1.87)	0.44 (0.24, 0.83)
Antidepressants	31	12,032	2.58 (1.78, 3.61)	Reference
Topiramate	15	8618	1.74 (1.02, 2.80)	0.45 (0.25, 0.84)
Anticonvulsants	33	8592	3.84 (2.69, 5.33)	Reference
Topiramate	11	5651	1.95 (1.03, 3.37)	0.46 (0.22, 0.94)
Other prophylactic treatments	24	5646	4.25 (2.79, 6.22)	Reference
Mortality from trauma
Topiramate	2	10,521	0.19 (0.04, 0.61)	0.25 (0.05, 1.17)
Cardiovascular treatments	8	10,534	0.76 (0.36, 1.43)	Reference
Topiramate	3	12,236	0.25 (0.07, 0.65)	0.30 (0.08, 1.08)
Antidepressants	10	12,032	0.83 (0.43, 1.47)	Reference
Topiramate	5	8618	0.58 (0.22, 1.27)	0.50 (0.17, 1.46)
Anticonvulsants	10	8592	1.16 (0.60, 2.06)	Reference
Topiramate	2	5651	0.35 (0.07, 1.13)	0.25 (0.05, 1.19)
Other prophylactic treatments	8	5646	1.42 (0.67, 2.67)	Reference
Mortality ascribed to drug overdose
Topiramate	2	10,521	0.19 (0.04, 0.61)	0.66 (0.11, 3.97)
Cardiovascular treatments	3	10,534	0.28 (0.08, 0.76)	Reference
Topiramate	2	12,236	0.16 (0.03, 0.52)	0.39 (0.08, 2.03)
Antidepressants	5	12,032	0.42 (0.16, 0.91)	Reference
Topiramate	2	8618	0.23 (0.05, 0.74)	0.50 (0.09, 2.71)
Anticonvulsants	4	8592	0.47 (0.16, 1.11)	Reference
Topiramate	1	5651	0.18 (0.02, 0.83)	0.17 (0.02, 1.40)
Other prophylactic treatments	6	5646	1.06 (0.44, 2.19)	Reference

IR: incidence rate; CI: confidence interval; HR: hazard ratio.

Incidence rates are reported per 1000 person-years.

Table 5 presents the odds ratios and corresponding 95% CIs for each outcome according to migraine treatments received in the 12 months prior to the case or control date. Following multivariable adjustment of the odds ratios, increased risks of cardiovascular events, cerebrovascular events, and mortality were observed with exposure to anticonvulsants, serotonin norepinephrine reuptake inhibitors, beta blockers, antihypertensives, tricyclic antidepressants, and other prophylactic treatments, as indicated by the adjusted odds ratios in Table 5 where the corresponding 95% CIs excluded 1. Depression, hypertension, hypercholesterolemia, obesity, and opioid use were also independent risk factors for several outcomes (Supplemental Table 4).

Table 5.

Nested case-control analysis: Unadjusted and adjusted outcome odds ratios by migraine prophylactic treatment use in the 12 months prior to the case or control-defined event date.

Outcomes by prophylactic treatment*	Cases	Controls	Unadjusted OR (95% CI)	Adjusted OR (95% CI)^†,‡
Myocardial infarction	182	21,682
Topiramate	29 (15.9%)	3605 (16.6%)	0.95 (0.64, 1.42)	0.97 (0.60, 1.58)
Anticonvulsants	37 (20.3%)	2458 (11.3%)	2.00 (1.39, 2.87)	1.78 (1.23, 2.56)
Beta blockers	39 (21.4%)	2914 (13.4%)	1.76 (1.23, 2.51)	1.43 (1.00, 2.05)
Tricyclic antidepressants	20 (11.0%)	2373 (10.9%)	1.00 (0.63, 1.60)	1.05 (0.66, 1.67)
SNRIs	30 (16.5%)	2065 (9.5%)	1.88 (1.26, 2.78)	1.85 (1.24, 2.76)
Calcium channel blockers	10 (5.5%)	700 (3.2%)	1.74 (0.92, 3.31)	1.41 (0.74, 2.69)
Antihypertensives	22 (12.1%)	917 (4.2%)	3.11 (1.98, 4.89)	2.09 (1.32, 3.30)
Other prophylactic treatments	21 (11.5%)	1258 (5.8%)	2.12 (1.34, 3.35)	2.05 (1.29, 3.25)
Myocardial ischemia	31	4012
Topiramate^§	7 (22.6%)	679 (16.9%)	1.43 (0.61, 3.34)	1.37 (0.46, 4.07)
Anticonvulsants	5 (16.1%)	447 (11.1%)	1.53 (0.59, 4.01)	1.32 (0.50, 3.50)
Beta blockers	6 (19.4%)	528 (13.2%)	1.58 (0.65, 3.88)	1.21 (0.49, 2.99)
Tricyclic antidepressants	3 (9.7%)	435 (10.8%)	0.88 (0.27, 2.91)	0.94 (0.28, 3.13)
SNRIs	6 (19.4%)	371 (9.2%)	2.36 (0.96, 5.78)	2.42 (0.97, 6.02)
Calcium channel blockers	3 (9.7%)	132 (3.3%)	3.15 (0.95, 10.49)	2.43 (0.72, 8.19)
Antihypertensives	5 (16.1%)	168 (4.2%)	4.40 (1.67, 11.60)	2.90 (1.08, 7.79)
Other prophylactic treatments	4 (12.9%)	195 (4.9%)	2.90 (1.00, 8.37)	2.79 (0.96, 8.14)
Unstable angina	84	8485
Topiramate^¶	14 (16.7%)	1428 (16.8%)	0.99 (0.56, 1.76)	1.38 (0.77, 2.49)
Anticonvulsants	16 (19.0%)	924 (10.9%)	1.93 (1.11, 3.33)	1.77 (1.02, 3.09)
Beta blockers	16 (19.0%)	1162 (13.7%)	1.48 (0.86, 2.57)	1.15 (0.66, 2.00)
Tricyclic antidepressants	7 (8.3%)	908 (10.7%)	0.76 (0.35, 1.65)	0.83 (0.38, 1.82)
SNRIs	11 (13.1%)	802 (9.5%)	1.44 (0.76, 2.73)	1.51 (0.79, 2.88)
Calcium channel blockers	5 (6.0%)	238 (2.8%)	2.19 (0.88, 5.46)	1.50 (0.59, 3.80)
Antihypertensives	11 (13.1%)	344 (4.1%)	3.57 (1.88, 6.78)	2.10 (1.09, 4.07)
Other prophylactic treatments	10 (11.9%)	475 (5.6%)	2.28 (1.17, 4.44)	2.20 (1.12, 4.33)
Ischemic stroke	328	39,318
Topiramate	76 (23.2%)	6603 (16.8%)	1.49 (1.15, 1.93)	1.26 (0.91, 1.73)
Anticonvulsants	80 (24.4%)	4445 (11.3%)	2.53 (1.96, 3.26)	2.38 (1.85, 3.07)
Beta blockers	70 (21.3%)	5310 (13.5%)	1.74 (1.33, 2.27)	1.54 (1.18, 2.01)
Tricyclic antidepressants	45 (13.7%)	4259 (10.8%)	1.31 (0.95, 1.80)	1.35 (0.98, 1.85)
SNRIs	54 (16.5%)	3953 (10.1%)	1.76 (1.31, 2.37)	1.69 (1.26, 2.28)
Calcium channel blockers	14 (4.3%)	1265 (3.2%)	1.34 (0.78, 2.30)	1.19 (0.70, 2.05)
Antihypertensives	35 (10.7%)	1576 (4.0%)	2.86 (2.01, 4.08)	2.22 (1.55, 3.18)
Other prophylactic treatments	42 (12.8%)	2302 (5.9%)	2.36 (1.70, 3.27)	2.25 (1.62, 3.12)
Hemorrhagic stroke	106	12,904
Topiramate	15 (14.2%)	2127 (16.5%)	0.84 (0.48, 1.44)	0.54 (0.26, 1.12)
Anticonvulsants	20 (18.9%)	1489 (11.5%)	1.78 (1.09, 2.91)	1.68 (1.03, 2.75)
Beta blockers	16 (15.1%)	1727 (13.4%)	1.15 (0.67, 1.96)	1.04 (0.61, 1.79)
Tricyclic antidepressants	20 (18.9%)	1362 (10.6%)	1.97 (1.21, 3.22)	2.01 (1.23, 3.28)
SNRIs	15 (14.2%)	1312 (10.2%)	1.46 (0.84, 2.52)	1.39 (0.80, 2.42)
Calcium channel blockers	6 (5.7%)	407 (3.2%)	1.84 (0.80, 4.22)	1.68 (0.73, 3.87)
Antihypertensives	7 (6.6%)	516 (4.0%)	1.70 (0.78, 3.67)	1.39 (0.64, 3.03)
Other prophylactic treatments	13 (12.3%)	773 (6.0%)	2.19 (1.22, 3.94)	2.08 (1.16, 3.75)
Transient ischemic attack	71	8562
Topiramate	20 (28.2%)	1441 (16.8%)	1.94 (1.15, 3.26)	1.46 (0.75, 2.81)
Anticonvulsants	13 (18.3%)	957 (11.2%)	1.78 (0.97, 3.26)	1.60 (0.87, 2.93)
Beta blockers	11 (15.5%)	1144 (13.4%)	1.19 (0.69, 2.27)	1.03 (0.54, 1.98)
Tricyclic antidepressants	6 (8.5%)	994 (11.6%)	0.70 (0.30, 1.63)	0.69 (0.30, 1.61)
SNRIs	11 (15.5%)	807 (9.4%)	1.76 (0.92, 3.36)	1.56 (0.81, 3.00)
Calcium channel blockers	3 (4.2%)	271 (3.2%)	1.35 (0.42, 4.32)	1.23 (0.38, 3.95)
Antihypertensives	8 (11.3%)	358 (4.2%)	2.91 (1.38, 6.12)	2.25 (1.06, 4.79)
Other prophylactic treatments	7 (9.9%)	487 (5.7%)	1.81 (0.83, 3.98)	1.72 (0.78, 3.78)
Overall mortality	234	36,974
Topiramate	35 (15.0%)	6278 (17.0%)	0.86 (0.60, 1.23)	0.65 (0.40, 1.04)
Anticonvulsants	103 (44.0%)	4158 (11.2%)	6.21 (4.78, 8.05)	5.83 (4.49, 7.58)
Beta blockers	62 (26.5%)	5030 (13.6%)	2.29 (1.71, 3.07)	2.00 (1.49, 2.68)
Tricyclic antidepressants	44 (18.8%)	4110 (11.1%)	1.85 (1.33, 2.57)	1.90 (1.37, 2.65)
SNRIs	61 (26.1%)	3735 (10.1%)	3.14 (2.34, 4.21)	3.05 (2.27, 4.11)
Calcium channel blockers	6 (2.6%)	1245 (3.4%)	0.76 (0.34, 1.70)	0.64 (0.28, 1.44)
Antihypertensives	11 (4.7%)	1535 (4.2%)	1.14 (0.62, 2.09)	0.82 (0.45, 1.51)
Other prophylactic treatments	40 (17.1%)	2190 (5.9%)	3.28 (2.32, 4.62)	3.11 (2.20, 4.39)
Cardiovascular and cerebrovascular mortality	27	4551
Topiramate	4 (14.8%)	757 (16.6%)	0.87 (0.30, 2.53)	0.14 (0.01, 1.40)
Anticonvulsants	12 (44.4%)	541 (11.9%)	5.93 (2.76, 12.74)	5.47 (2.53, 11.83)
Beta blockers	7 (25.9%)	625 (13.7%)	2.20 (0.93, 5.22)	1.78 (0.74, 4.30)
Tricyclic antidepressants	6 (22.2%)	487 (10.7%)	2.38 (0.96, 5.94)	2.39 (0.95, 5.98)
SNRIs	7 (25.9%)	467 (10.3%)	3.06 (1.29, 7.28)	2.96 (1.23, 7.10)
Calcium channel blockers	0 (0.0%)	153 (3.4%)	–	–
Antihypertensives	1 (3.7%)	196 (4.3%)	0.85 (0.12, 6.33)	0.68 (0.09, 5.12)
Other prophylactic treatments	3 (11.1%)	33 (0.7%)	2.04 (0.61, 6.81)	1.91 (0.57, 6.42)
Mortality from trauma	77	9,090
Topiramate	16 (20.8%)	1540 (16.9%)	1.29 (0.74, 2.24)	1.13 (0.58, 2.19)
Anticonvulsants	35 (45.5%)	1065 (11.7%)	6.28 (3.99, 9.88)	6.40 (4.06, 10.11)
Beta blockers	21 (27.3%)	1193 (13.1%)	2.48 (1.50, 4.11)	2.47 (1.49, 4.12)
Tricyclic antidepressants	18 (23.4%)	915 (10.1%)	2.73 (1.60, 4.64)	2.88 (1.68, 4.91)
SNRIs	24 (31.2%)	913 (10.0%)	4.06 (2.49, 6.60)	4.50 (2.74, 7.38)
Calcium channel blockers	3 (3.9%)	286 (3.1%)	1.25 (0.39, 3.98)	1.10 (0.34, 3.54)
Antihypertensives	2 (2.6%)	348 (3.8%)	0.67 (0.16, 2.74)	0.58 (0.14, 2.40)
Other prophylactic treatments	15 (19.5%)	511 (5.6%)	4.06 (2.29, 7.19)	4.22 (2.38, 7.51)
Mortality ascribed to drug overdose	45	5506
Topiramate	11 (24.4%)	928 (16.9%)	1.60 (0.81, 3.16)	0.71 (0.25, 2.00)
Anticonvulsants	22 (48.9%)	598 (10.9%)	7.85 (4.35, 14.18)	7.90 (4.37, 14.29)
Beta blockers	11 (24.4%)	752 (13.7%)	2.05 (1.03, 4.05)	2.03 (1.02, 4.05)
Tricyclic antidepressants	13 (28.9%)	641 (11.6%)	3.08 (1.61, 5.91)	3.11 (1.62, 5.96)
SNRIs	20 (44.4%)	551 (10.0%)	7.20 (3.97, 13.04)	7.95 (4.33, 14.60)
Calcium channel blockers	2 (4.4%)	177 (3.2%)	1.40 (0.34, 5.83)	1.35 (0.32, 5.63)
Antihypertensives	1 (2.2%)	215 (3.9%)	0.56 (0.08, 4.08)	0.49 (0.07, 3.60)
Other prophylactic treatments	13 (28.9%)	349 (6.3%)	6.00 (3.12, 11.54)	5.99 (3.10, 11.59)

OR: odds ratio; CI: confidence interval; SNRIs: serotonin norepinephrine reuptake inhibitors.

Patients may have been exposed to multiple prophylactic medications prior to the case or control-defined event date. The odds ratios compared any use of the specified prophylactic medication to no use of the medication in the 12 months prior to the case or control-defined event date.

†

Except where noted, regression models comparing topiramate exposure to non-exposure were adjusted for age group, sex, and presence of hypertension, depression, other anticonvulsants, beta blockers, tricyclic antidepressants, serotonin norepinephrine reuptake inhibitors, calcium channel blockers, antihypertensives, and other prophylactic treatments prior to the topiramate dispensing. The regression models comparing the remaining treatments to non-use of those treatments were adjusted for age group and sex.

‡

Adjustment for age group in the regression models was done using the following categories (in years): 18–24, 25–34, 35–44, 45–54, 55–65, > 65. Due to the small number of cases within certain age strata, age groups were collapsed for select outcome regression models as follows: Myocardial infarction, unstable angina, transient ischemic attack: 18–34, 35–44, 45–54, 55–65, > 65; myocardial ischemia: 18–44, 45–54, 55–65, > 65.

Due to sample size constraints, the regression model was adjusted for age group, sex, and presence of hypertension, depression, other anticonvulsants, cardiovascular treatments (consisting of beta blockers, calcium channel blockers, antihypertensives), antidepressants (consisting of tricyclic antidepressants, serotonin norepinephrine reuptake inhibitors) and other prophylactic treatments prior to the topiramate dispensing.

The regression model was adjusted for age group and sex due to sample size constraints.

In a secondary analysis, topiramate use in the 3 months prior to the case or control date was associated with an increased risk of transient ischemic attack (adjusted odds ratios: 2.08, 95% CI: 1.11, 3.91) (Supplemental Table 5).

Discussion

This observational study evaluated the risk of cardiovascular, cerebrovascular and mortality events in a cohort of patients with migraine receiving prophylactic treatment. The cohort demographics are consistent with the known epidemiology of migraine (21). Among the prophylactic treatment initiators, the observed differences in the risk of non-fatal cardiovascular and cerebrovascular events across the propensity score-matched cohorts were consistent with chance. Topiramate initiators had a significantly lower risk of overall mortality compared to matched antidepressant initiators and matched anticonvulsant initiators. In the nested case-control analysis, no increased risk of any of the outcomes was observed with topiramate or calcium channel blocker use compared to no use of that specific drug in the previous 12 months. An increased risk of transient ischemic attack was associated with topiramate use in the previous 3 months (compared to no use of topiramate in the previous 3 months), suggesting that topiramate use might be associated with risk factors for transient ischemic attack. Use of other anticonvulsants, serotonin norepinephrine reuptake inhibitors, beta blockers, antihypertensives, tricyclic antidepressants, and other prophylactic treatments in the previous 12 months was associated with an increased risk of several outcomes compared to no use of that specific drug.

This study has a number of strengths. Although the prophylactic treatments assessed are approved for indications other than migraine prevention, we assessed the treatments in a cohort formed on the basis of migraine diagnoses and treatments indicated for acute migraine. Members with a prior diagnosis for epilepsy (indicated for topiramate) were excluded from the cohort. In order to avoid a substantial reduction in cohort size, patients with other outcome risk factors, such as cardiac arrhythmia, hypertension, depression, or obesity, were not excluded from cohort entry. We accounted for some of these risk factors, which may have developed or worsened during the course of follow-up (epilepsy and cardiac arrhythmia), by restricting the outcome analyses to patients without these conditions in the 12 months prior to prophylactic treatment initiation or in the 12 months prior to the case or control date. Recall bias, a frequently-cited limitation of case-control studies, is excluded in this study with exposure ascertained from administrative pharmacy records. Additionally, as outcome sample size allowed, the regression models in the nested case-control analysis with topiramate as the primary exposure were adjusted for several potential confounding factors in addition to age group and sex. The propensity score-matched topiramate initiators were comparable to the matched comparator drug initiators with respect to demographics, comorbidities, treatments, and healthcare utilization variables. Mortality outcomes were ascertained from the NDI database, which is considered the gold standard for death identification in medical and health research (22). A prior study evaluating the accuracy of NDI-coded causes of death showed a 93% exact match after confirmation with death certificates from state vital statistics offices (23). Even among the mismatches, there was still agreement on the major organ systems involved, indicating that the NDI-coded causes of death are comparable to other cause of death identification procedures (23).

Several factors should be considered when interpreting the study results. The linkage to the NDI for ascertainment of mortality required that the study population members come from health plans that allow access to patient identifiers (particularly name and social security number). Although the study excluded patients from health plans that do not allow this access, this exclusion is for administrative reasons rather than clinical ones, so would not be expected to introduce bias. The small numbers of certain outcomes limit the interpretation of the outcome effect measures. Due to sample size constraints in the nested case-control analysis, the regression models for the prophylactic treatments other than topiramate only included age group and sex. Although increased risks for several of the outcomes were observed with many of the prophylactic treatments, it is likely that the prophylactic treatments other than topiramate represent markers for outcome risk factors that developed or progressed after cohort entry, rather than being a direct effect of the treatments. Indeed, several co-morbidities assessed in the nested case-control analysis were risk factors for most of the outcomes independent of prophylactic treatment. Topiramate has demonstrated efficacy as a weight loss medication (24) and is available as a combination drug with phentermine for the management of obesity. The use of topiramate in our study population for obesity management (separate from or in addition to migraine prophylaxis) cannot be ruled out. Prophylactic treatments may be prescribed to patients with more severe or frequent migraine attacks. As severity and frequency of migraine attacks may influence the risk of the outcomes assessed, the prophylactic treatments may act as a marker of increased migraine severity and/or frequency. Use of triptans and ergotamines was part of the definition for migraine cohort entry. The vasoconstrictive properties of these medications may increase the risk of ischemic events (25). In the nested case-control analysis, the odds ratios for myocardial ischemia, ischemic stroke, transient ischemic attack, all-cause mortality, and cardiovascular and cerebrovascular mortality with prior triptan use ranged from 0.41 to 0.69, with upper bounds of the 95% CIs less than one, and ergotamine use was rare. This suggests that prescribing decisions for these medications are influenced by existing risk factors for these outcomes.

Although an increased risk of transient ischemic attack associated with recent topiramate use was observed, a recent worsening of migraine attacks, particularly with the presence of migraine aura, may be misdiagnosed as a transient ischemic attack due to the similarity of the clinical features of the two conditions, resulting in an overestimation of transient ischemic attack risk (26). Topiramate is also associated with adverse events that may mimic the clinical features of a transient ischemic attack, such as numbness or tingling of extremities, dizziness, blurred vision, and speech problems. The ability to distinguish certain features of migraines (e.g. presence of aura, chronic versus episodic migraines) on the basis of ICD-9-CM diagnosis codes within claims data was limited. The intent-to-treat approach used for incidence rate estimation among the prophylactic treatment initiators has the advantage of preserving the randomization-like features of the propensity score matching but risks misclassification of exposure. Such misclassification might provide biased effect estimates as exposure status changes throughout the course of follow-up. The nested case-control analysis estimated outcome risks for a given prophylactic treatment compared to no use of that specific treatment in the previous 12 months but did not take into account use of multiple prophylactic treatments.

There are also limitations that are inherent to administrative claims data sources. While claims data are valuable for the examination of health care outcomes and treatment patterns, claims are collected for the purpose of reimbursement. Presence of a claim for a filled prescription does not indicate that the medication was consumed or taken as prescribed. Medications obtained over the counter or provided as samples by the physician will not be observed in the claims. As such, the assessment of NSAID use in this study only includes those prescribed by a healthcare provider. Presence of a diagnosis code on a medical claim may be incorrectly coded or included as rule-out criteria rather than indicate the presence of true disease. To mitigate this potential source of misclassification, the study cohort was identified based on various medical care patterns corresponding to a migraine diagnosis, requiring two indicators 7–180 days apart. Additionally, VH and JS examined the cohort entry definition by reviewing de-identified patient-level chronological listings of all diagnosis, service, procedure, and pharmacy claims surrounding the migraine diagnosis date for a sample of cohort members meeting each of the five criteria to ensure the pattern of medical care was consistent with a migraine diagnosis.

Although the cardiovascular and cerebrovascular outcomes in this study were not validated through medical record review, our case ascertainment method was based on published algorithms for cardiovascular and cerebrovascular events that demonstrated high validity (27 –29). Follow-up duration can be limited due to individuals changing health insurance plans. The average length of follow-up in this study (1.8 years) is consistent with the average length of enrollment for all patients in the health insurance plan (2 years). The length of follow-up allowed for the evaluation of the short-term effects of prophylactic treatments on the risk of cardiovascular and cerebrovascular disease and mortality, but additional studies with longer duration of follow-up would be warranted for the evaluation of long-term effects. The required length of continuous enrollment did not capture comorbidities and past diagnoses that occurred more than 12 months prior to cohort entry. As such, prevalent outcomes may have been misclassified in the analysis as incident outcomes. Additionally, the prophylactic treatment initiators identified in this study likely represent a mix of first-time users of that specific treatment and those who re-initiated that specific treatment after an extended period of non-use. However, given the average length of patient enrollment in the health plan, extension of the continuous enrollment period beyond 12 months would have led to a substantial reduction in the size of the cohort. As cohort members had insurance coverage through a US commercial health plan through their employer or as a dependent, the findings of this study may not be generalizable to patients with non-commercial or no health insurance coverage and who therefore have a different pattern of medical care.

This study provides an initial assessment of the influence of prophylactic treatments on the estimated effect between migraine and cardiovascular and cerebrovascular disease and mortality. The cohort and nested case-control designs focused on different aspects of the migraine cohort follow-up, including patient characteristics preceding an outcome event and preceding prophylactic treatment initiation. Additional risk factors, including migraine type, severity, frequency, and other treatment indications should be considered in future migraine prophylactic treatment safety assessments.

Supplemental Material

Supplemental material for Risk of cardiovascular and cerebrovascular events and mortality in patients with migraine receiving prophylactic treatments: An observational cohort study

Supplemental Material for Risk of cardiovascular and cerebrovascular events and mortality in patients with migraine receiving prophylactic treatments: An observational cohort study by Veena Hoffman, Fei Xue, Stephen M Ezzy, Akeem Yusuf, Edward Green, Osa Eisele, Tobias Kurth and John D Seeger in Cephalalgia

Footnotes

Article highlights

This study provides an initial assessment of the influence of prophylactic treatments on the association between migraine and cardiovascular disease, cerebrovascular disease, and mortality.

Additional risk factors, including migraine severity, frequency, and other treatment indications should be considered in future studies evaluating the safety of prophylactic treatments in patients with migraine.

Acknowledgements

The authors wish to acknowledge the contributions of the following colleagues from Optum Epidemiology: Michael Doherty, Laura Yochum, Yuewen Sara Gao, and Ling Li for their assistance with the study analysis, and Nicole Brooks, Julia Cocozza, and Laura Karslake for their assistance with study operations.

Institutional Review Board approval

This study used demographic and claims data of insured members in the Optum Research Database, in accordance with applicable laws and regulations. Limited personal identifying information was used to link cohort member data to the NDI database. All study analyses were conducted using de‐identified data. We obtained approval of the study protocol and a waiver of patient authorization from the New England Institutional Review Board (NEIRB 120160605) and affiliated Privacy Board (PB 16‐027). No patients were directly contacted as part of the study.

Author contributions

VH, FX, OE, TK, and JS designed the study. VH and JS directed the analyses, which were conducted by SE and EG. All authors contributed to the interpretation of the results. VH drafted the manuscript, and all authors critically revised the manuscript for intellectual content and approved the final version.

Declaration of conflicting interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: All authors have completed the ICMJE uniform disclosure form at (available on request from the corresponding author), and declare the following: VH, SE, and JS are employed by Optum, and VH and JS hold stock/stock options in the parent company of Optum (United HealthGroup, Inc.). EG was employed by Optum at the time of study conduct. FX, AY, and OE are employees and stockholders of Amgen, Inc. TK provided methodological expertise to Amgen and CoLucid, for which the Charité – Universitätsmedizin Berlin has received financial compensation. TK further received honoraria from Novartis, Daiichi Sankyo, and Total for a scientific presentation and from Lilly and Newsenselab for methodological advice.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded with a research contract between Optum and Amgen, Inc.

References

Bigal

Kurth

Santanello

, et al. Migraine and cardiovascular disease: A population-based study. Neurology 2010; 74: 628–635.

Guidetti

Rota

Morelli

, et al. Migraine and stroke: “Vascular” comorbidity. Front Neurol 2014; 5: 193.

Kurth

Winter

Eliassen

, et al. Migraine and risk of cardiovascular disease in women: Prospective cohort study. BMJ 2016; 353: i2610.

Sacco

Kurth

. Migraine and the risk for stroke and cardiovascular disease. Curr Cardiol Rep 2014; 16: 524. DOI: 10.1007/s11886-014-0524-1 .

Adelborg

Szepligeti

Holland-Bill

, et al. Migraine and risk of cardiovascular diseases: Danish population based matched cohort study. BMJ 2018; 360: k96.

Mahmoud

Mentias

Elgendy

, et al. Migraine and the risk of cardiovascular and cerebrovascular events: A meta-analysis of 16 cohort studies including 1 152 407 subjects. BMJ Open 2018; 8: e020498.

Linstra

Ibrahimi

Terwindt

, et al. Migraine and cardiovascular disease in women. Maturitas 2017; 97: 28–31.

Donaghy

Chang

Poulter

. Duration, frequency, recency, and type of migraine and the risk of ischaemic stroke in women of childbearing age. J Neurol Neurosurg Psychiatry 2002; 73: 747–750.

Kurth

Schurks

Logroscino

, et al. Migraine frequency and risk of cardiovascular disease in women. Neurology 2009; 73: 581–588.

10.

MacClellan

Giles

Cole

, et al. Probable migraine with visual aura and risk of ischemic stroke: The stroke prevention in young women study. Stroke 2007; 38: 2438–2445.

11.

Schurks

Rist

Shapiro

, et al. Migraine and mortality: A systematic review and meta-analysis. Cephalalgia 2011; 31: 1301–1314.

12.

Gudmundsson

Scher

Aspelund

, et al. Migraine with aura and risk of cardiovascular and all cause mortality in men and women: Prospective cohort study. BMJ 2010; 341: c3966.

13.

Tso

Goadsby

. Anti-CGRP monoclonal Antibodies: The next era of migraine prevention?. Curr Treat Options Neurol 2017; 19: 27. DOI: 10.1007/s11940-017-0463-4 .

14.

US Food and Drug Administration. FDA approves novel preventive treatment for migraine (news release), https://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm608120.htm?utm_campaign=05172018_PR_FDA%20approves%20migraine%20prevention%20treatment&utm_medium=email&utm_source=Eloqua (2018, accessed 3 August 2018).

15.

National Centers for Health Statistics. National Death Index, http:\\www.cdc.gov/nchs/ndi.htm (7 March 2017, accessed 23 May 2018).

16.

National Centers for Health Statistics. National Death Index user's guide, https://www.cdc.gov/nchs/data/ndi/ndi_users_guide.pdf (2013, accessed 23 May 2018).

17.

Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques. In: Proceedings of the Twenty-Sixth Annual SAS Users Group International Conference, Long Beach, California, USA, 22–25 April 2001, paper 214-26, Cary, NC: SAS Institute, Inc., http://www2.sas.com/proceedings/sugi26/p214-26.pdf (accessed 23 May 2018).

18.

Parsons LS. Performing a 1:N case-control match on propensity score. In: Proceedings of the Twenty-Ninth Annual SAS Users Group International Conference, Montreal, Canada, 9–12 May 2004, paper 165-29, Cary, NC: SAS Institute, Inc., http://www2.sas.com/proceedings/sugi29/165-29.pdf (accessed 23 May 2018).

19.

Austin

Grootendorst

Anderson

. A comparison of the ability of different propensity score models to balance measured variables between treated and untreated subjects: A Monte Carlo study. Stat Med 2007; 26: 734–753.

20.

Austin

. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med 2009; 28: 3083–3107.

21.

Lipton

Bigal

. The epidemiology of migraine. Am J Med 2005; 118: S3–S10.

22.

Fillenbaum

Burchett

Blazer

. Identifying a national death index match. Am J Epidemiol 2009; 170: 515–518.

23.

Doody

Hayes

Bilgrad

. Comparability of national death index plus and standard procedures for determining causes of death in epidemiologic studies. Ann Epidemiol 2001; 11: 46–50.

24.

Cosentino

Conrad

Uwaifo

. Phentermine and topiramate for the management of obesity: A review. Drug Des Devel Ther 2013; 7: 267–278.

25.

Roberto

Raschi

Piccinni

, et al. Adverse cardiovascular events associated with triptans and ergotamines for treatment of migraine: Systematic review of observational studies. Cephalalgia 2015; 35: 118–131.

26.

Amort

Fluri

Schafer

, et al. Transient ischemic attack versus transient ischemic attack mimics: Frequency, clinical characteristics and outcome. Cerebrovasc Dis 2011; 32: 57–64.

27.

Andrade

Harrold

Tjia

, et al. A systematic review of validated methods for identifying cerebrovascular accident or transient ischemic attack using administrative data. Pharmacoepidemiol Drug Saf 2012; 21: S100–S128.

28.

Roumie

Mitchel

Gideon

, et al. Validation of ICD-9 codes with a high positive predictive value for incident strokes resulting in hospitalization using Medicaid health data. Pharmacoepidemiol Drug Saf 2008; 17: 20–26.

29.

Varas-Lorenzo

Castellsague

Stang

, et al. Positive predictive value of ICD-9 codes 410 and 411 in the identification of cases of acute coronary syndromes in the Saskatchewan Hospital automated database. Pharmacoepidemiol Drug Saf 2008; 17: 842–852.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.67 MB