G indicates gamma-aminobutyric acid analogs; SC, sodium channel blockers; SV2, synaptic vesicle protein 2A binding; and M, multiple mechanisms.
Margolis JM, Chu B, Wang ZJ, Copher R, Cavazos JE. Effectiveness of Antiepileptic Drug Combination Therapy for Partial-Onset Seizures Based on Mechanisms of Action. JAMA Neurol. 2014;71(8):985-993. doi:10.1001/jamaneurol.2014.808
To our knowledge, the current study is the first to describe antiepileptic drug (AED) combination therapy patterns according to their mechanism of action (MOA) in a real-world setting and to evaluate the differences in outcomes comparing different-MOA combination therapy with same-MOA combination therapy for patients with partial-onset seizure.
To compare treatment persistence and health care use with AED combinations categorized by MOA in patients with partial-onset seizures.
Design, Setting, and Participants
Using the Truven Health MarketScan Commercial Claims Database containing 96 million covered lives from July 1, 2004, through March 31, 2011, adults with concomitant use of 2 different AEDs and a recent partial-onset seizure diagnosis were selected. Antiepileptic drugs were categorized by MOA: sodium channel blockers (SC), gamma-aminobutyric acid analogs (G), synaptic vesicle protein 2A binding (SV2), and multiple mechanisms (M). Patients were assigned a combination category based on their concomitant AED use.
Main Outcomes and Measures
Treatment persistence was measured from the start of AED combination therapy until the end of the combination. Health care resource use was measured during the combination treatment duration. Multivariate analyses evaluated AED discontinuation risk and health care use according to MOA combinations.
Distribution of 8615 selected patients by combination was 3.3% for G+G, 7.5% for G+SV2, 8.6% for G+M, 13.9% for SC+SC, 19.0% for G+SC, 21.5% for SC+M, and 26.3% for SC+SV2. The same-MOA (G+G and SC+SC) combinations had the shortest persistence (mean [SD], 344  days and 513  days, respectively) and greater hazard of discontinuation compared with different-MOA combinations. Patients with different-MOA G combinations had a significantly lower risk for inpatient admission (odds ratio, 0.716; 95% CI, 0.539-0.952; P = .02) compared with G+G combinations. Patients with different-MOA SC combinations had significantly lower risks for emergency department visits (odds ratio, 0.853; 95% CI, 0.742-0.980; P = .03) compared with SC+SC combinations.
Conclusions and Relevance
The findings suggest that AED combinations with different MOAs have greater effectiveness as measured by treatment persistence and lower risks for hospitalization and emergency department visits. Further research is needed to more fully understand the role of the MOA in achieving optimal outcomes.
Quiz Ref IDEpilepsy affects approximately 3 million people in the United States, resulting in substantial health care economic burden (an estimated $17.6 billion in direct and indirect costs in the United States in 2010).1 Seizures are typically classified as generalized, with synchronized cortical manifestation in both hemispheres, or partial, with focal cortical onset.2 Partial-onset seizures (POS) occur in about two-thirds of patients with epilepsy.3
Quiz Ref IDAlthough antiepileptic drug (AED) therapy is the mainstay of treatment, seizure-free outcomes are attained in only 60% to 70% of patients with initial monotherapy.2,4 Treatment options for the 30% to 40% of refractory patients include other AED monotherapies or combination therapy using an additional AED as adjunctive therapy. Data from AED combination therapy trials in patients with POS have shown elimination of seizures in 15% to 35% of patients and reduction of seizure frequency by more than half in 12% to 29% of patients. In addition to the clinical benefits from adequate seizure control, there is significant cost savings potential with effective treatment, with as much as an 8-fold higher cost of illness in patients with refractory epilepsy.5,6
More than 20 AEDs are now available in the Unites States, creating considerable challenges for physicians in choosing among drugs that synergistically control seizures while minimizing adverse effects. The hypothesis of rational polytherapy is that combining drugs with different mechanisms of action (MOAs) can be more effective because they act on multiple drug targets rather than combining drugs with the same MOA that also risk having additive adverse-effect profiles.4,7,8 Although there is theoretical appeal to the use of different-MOA vs same-MOA combinations in AED therapy, to our knowledge, there are no randomized clinical studies and little population-level or real-world evidence for the comparative effectiveness of this approach.
The objectives of the current study were to test whether MOA-based AED combination therapies for treatment of POS are more effective than non-MOA–based therapies, specifically whether different-MOA–based combinations are more effective than same-MOA–based combinations. Using real-world data from a large population sample, combination effectiveness was assessed by persistence on therapy as a measure of tolerability and seizure control and by comparing health care use between same-MOA and different-MOA combinations.
Study patients were selected from the Truven Health MarketScan Commercial Claims and Encounters (Commercial) Database using data from July 1, 2004, through March 31, 2011. This database contains the health care experience of more than 96 million privately insured individuals and their dependents covered under a variety of health plans. Detailed enrollment, use, cost, and outcomes data are available for services performed in inpatient and outpatient settings including medical and pharmacy claims. All database records are de-identified and fully compliant with US patient confidentiality requirements; therefore, in accordance with the Office for Human Research Protections of the US Department of Health and Human Services these data are exempt from institutional review board evaluation.
Patients aged 18 years or older with concomitant use of 2 oral AEDs initiated between January 1, 2005, and March 31, 2010 (enrollment period), were selected. Patients had been receiving AED monotherapy prior to the initiation of a concomitant second AED. Concomitance was identified by overlap of the days of supply of the AEDs, with continuous therapy of the 2 drugs for a minimum of 90 days following initiation of the second AED. Drug therapy was considered continuous based on receipt of a subsequent prescription claim for the same medication prior to the end of the supply of the current prescription plus 30 days or half the days’ supply of the current prescription, whichever was greater. The index date was the date of the first prescription claim for the second AED. Continuous medical and prescription coverage was required for a minimum of 6 months prior to the index date (pre-index period), during which no AED combination therapy was found, and during follow-up (postindex period) for a minimum of 12 months after the index date, until either the end of enrollment or end of available data (March 31, 2011).
Epilepsy and a POS diagnosis were determined by screening all medical claims during both pre-index and postindex periods for at least 1 inpatient claim or 2 outpatient claims on different service dates for a diagnosis of epilepsy or epileptic syndromes with partial seizures (International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes 345.4x, 345.5x, 345.9x, and 780.39). Codes 345.9x (epilepsy, unspecified) and 780.39 (convulsions, not elsewhere classified) were included to account for physicians not using clinically specific codes. Patients were also required to have at least 1 claim with a diagnosis code of 345.4x or 345.5x, indicating POS. Patients meeting these diagnosis criteria may have had other epilepsy diagnoses during the study and were not excluded if other epilepsy codes were found. The resulting sample size remained robust (10 667 patients) and enabled focusing on a more homogeneously POS-diagnosed cohort.
Quiz Ref IDAntiepileptic drugs were assigned to 4 MOA categories based on their presumed primary MOA: sodium channel blocker (SC), gamma-aminobutyric acid analog (G), synaptic vesicle protein 2A binding (SV2), and multiple mechanisms (M).4Table 1 shows AEDs assigned to these categories.
Patients were assigned to an AED combination category based on their index combination. Same-MOA combination categories included SC+SC and G+G. Different-MOA combinations included SC+SV2, SC+M, G+SC, G+SV2, G+M, and M+SV2. Analysis was limited to each category’s top 10 specific AED combinations. The top 10 specific drug combinations for SC+SC included 99% of patients in this MOA combination category; for SC+M, 90% of patients; for G+SC, 77% of patients; for G+M, 85% of patients; and for G+G, 97% of patients. The categories of SC+SV2 and G+SV2 had fewer than 10 specific AED combinations so all were included. The categories of M+M and M+SV2 were not included in the analysis because the heterogeneity of mechanisms for the combinations made implications of their outcomes difficult to interpret. Furthermore, with only 1 drug in the SV2 class (levetiracetam), the M+SV2 combination had no same-MOA comparators.
Treatment duration with the index AED combination was measured as the number of days from the index date to the end of the index combination, the end of enrollment, or the end of available data (March 31, 2011), whichever occurred first. The index AED combination ended when there were no further claims for 1 of the AEDs prior to its last day of supply, allowing for a gap of 30 days or half the days’ supply from the current prescription claim.
All-cause health care use was measured during the time of persistence with the index AED combination. For inpatient admissions, emergency department visits, and physician office visits, the proportion of patients using each resource and the mean number of services per patient were reported along with lengths of stay for inpatient admissions.
Demographic variables measured as of the index date included age, sex, geographic region, urban/rural residence, insurance plan type, and index year.
Clinical variables were measured for the 6 months prior to the index date. Comorbidity burden was evaluated using 3 indices: the Deyo-Charlson Comorbidity Index, the mean number of unique 3-digit International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes, and the mean number of unique outpatient medications.9,10 Specific comorbid conditions and concomitant medications (other than AEDs) were also reported.
Descriptive statistics for demographic and clinical variables were reported by AED combination category. Appropriate statistical tests of significance for differences between treatment groups included χ2 tests for categorical variables and analysis of variance for quantitative variables. A critical value of 0.05 was specified a priori as the threshold for statistical significance for all tests.
Multivariate logistic models were used to examine the relationship between the various health care use outcomes with AED combination categories. Both the odds ratios (ORs) and their 95% CIs are presented along with tests of significance of the coefficients or log-odds. Cox proportional hazards models were applied to assess the effects of independent predictors on persistence with AED combination therapy, providing hazard ratios (HRs) with their 95% CIs. The proportionality assumption was tested for each predictor and no violation was detected. Analyses were conducted using Stata statistical software version 11.2 (StataCorp).
The final study cohort consisted of 8615 patients who met all study criteria (Figure 1). An SC combination was received by 81% of patients and 38% received a G combination. The most common combination was SC+SV2 (26%) and the least common was G+G (3.3%) (Table 2).
Mean (SD) age ranged from 41.7 (12.7) years (SC+M) to 48.4 (11.2) years (G+G) (Table 2). Cohorts receiving G combinations were generally older (mean [SD] range, 45.0 [12.2] years to 48.4 [11.2] years) than SC combination cohorts (mean [SD] range, 41.7 [12.7] years to 43.7 [12.3] years). Combinations of G had higher percentages of female patients, with women comprising 71.6% of G+G patients vs 51.2% of SC+SV2 patients.
The relative comorbidity burden, measured with the pre-index Deyo-Charlson Comorbidity Index, appeared highest for G+SV2 and G+G patients (Table 2). The G+SV2 cohort had the highest mean (SD) number of unique diagnoses during the pre-index period (13.0 [10.3] 3-digit International Classification of Diseases, Ninth Revision, Clinical Modification codes), followed by the G+G cohort, with a mean (SD) of 11.9 (9.6) diagnoses. The latter cohort also had the highest mean (SD) number of unique outpatient medications pre-index (15.9 [10.2]). The SC+SC cohort had the lowest comorbidity burden, with a mean (SD) Deyo-Charlson Comorbidity Index score of 0.3 (0.9), pre-index diagnoses of 6.0 (5.5), and unique pre-index medications of 6.7 (5.1). Comorbid conditions occurring in more than 10% of at least 1 cohort included depression, anxiety/panic disorders, hyperlipidemia, and fibromyalgia. The G+G cohort had the highest percentages of patients with these conditions and with prescriptions for concomitant medications (Table 2).
Cohorts receiving different-MOA combinations had longer mean and median persistence compared with cohorts receiving same-MOA combinations (Table 3). Patients receiving an AED combination of G+G remained taking therapy for the shortest period of time (mean [SD], 344  days; median, 218 days) compared with other G combinations. The SC+SC patients similarly had the shortest persistence (mean [SD], 513  days; median, 313 days) compared with other SC combinations (Figure 2).
The Cox proportional hazards models examining time to discontinuation showed different-MOA combinations with a lower hazard of discontinuation compared with same-MOA combinations (Table 4). Using SC+SC as the reference group, SC+SV2 patients were significantly less likely to discontinue therapy (HR, 0.810; 95% CI, 0.749-0.877; P < .001). Patients receiving G+SV2 or G+SC were significantly less likely to discontinue their combination compared with G+G patients (HR, 0.780; 95% CI, 0.670-0.907 and HR, 0.777; 95% CI, 0.678-0.890, respectively; both P ≤ .001). Although findings with other SC or G combinations were not significant, the direction of these findings was consistent with lower hazards associated with different-MOA combinations.
Patients taking G+G combinations had significantly greater health care resource use (unadjusted) compared with those taking other G combinations for inpatient admissions, admissions per patient, emergency department visits, emergency department visits per patient, and physician office visits per patient (Table 5). The SC+SC cohort showed significantly more admissions and emergency department visits per patient than SC+SV2. The SC+SC group also had longer lengths of stay and more physician visits per patient than the SC+SV2 and SC+M cohorts.
Logistic regression models examining the risk for inpatient admissions for G combinations found patients receiving different-MOA G combinations had a significantly lower risk for inpatient admission (Table 6, model 1; OR, 0.716; 95% CI, 0.539-0.952; P = .02). Model 2 found significantly lower risks with the different-MOA combinations of G+M and G+SC (OR, 0.724; 95% CI, 0.529-0.993; P = .05 and OR, 0.683, 95% CI, 0.509-0.916; P = .01, respectively), and while findings for G+SV2 were not significant, the point estimate was consistent with lower risk (OR, 0.802; 95% CI, 0.578-1.112; P = .19). Logistic regression models for SC combinations did not find statistically significant differences between SC combinations (Table 6).
Logistic regression models examining the risk for an emergency department visit found no significant differences between G combination cohorts; however, point estimates were suggestive of lower risk for emergency department visits with combinations based on different MOAs (Table 7). Logistic regression models of emergency department visits for patients with SC combinations found significantly lower risks for emergency department visits for different-MOA combinations (Table 7, model 1; OR, 0.853; 95% CI, 0.742-0.980; P = .03) and with the specific combination of SC+SV2 (Table 7, model 2; OR, 0.767; 95% CI, 0.657-0.896; P = .001) compared with SC+SC.
To our knowledge, based on our research and a review of the current literature, the current study is the first to describe AED combination therapy patterns by MOA in patients with POS in a real-world setting and to evaluate differences in outcomes comparing different-MOA combination therapy to same-MOA combination therapy. Many of the previous studies describing AED use were performed prior to the availability of the newer AEDs, did not reflect use in typical practice, and did not provide detail on specific AED combinations and outcomes resulting from their use.11,12 Although these previous studies certainly added to what was known about AED use and combination therapy, our current study provides greater detail of outcomes needed to inform drug-choice decisions.
Combinations containing an SC were prescribed for 80.7% of patients, with only 17.2% of these patients receiving an SC+SC combination vs an SC combination with a different MOA. Of the patients receiving a combination containing G, only 8.5% of the combinations were a G+G same-MOA combination. The findings suggest that different-MOA–based combination AED therapies have been the most common in clinical practice whether or not by conscious awareness of the MOAs by the prescriber.
Patients receiving G combinations had higher comorbidity burdens, as evidenced by the differences in comorbidity indices (Table 2) and by the pre-index prevalence of diagnoses for depression, hyperlipidemia, anxiety/panic disorders, and fibromyalgia. These differences in chronic illnesses were particularly evident in the percentages of patients treated concomitantly with antidepressants, anxiolytics, antihypertensives, and statins. All differences between SC and G combination cohorts for these comorbidities and concomitant medications were statistically significant. Prescribers of AEDs, or of any medication class, use comorbidity as one of the important factors in drug choice and therefore our results may also reflect that selection bias.
Quiz Ref IDIn this study, persistence with therapy was used as a proxy for effectiveness, having been an established proxy in both randomized clinical trials and prior epilepsy studies, assuming patients continued therapy owing to adequate seizure control and the absence of intolerable adverse events.13- 15 Patients treated with SC+SV2 had the longest treatment duration, with 19% lower discontinuation risk. Combinations of G+G resulted in the shortest treatment duration and overall higher risk for discontinuation. One possible explanation is the potential for noncompliance because of additive toxicity if the drugs in the combination exhibited similar adverse-effect profiles.16 Conversely, SC+SC combinations had the third longest treatment duration, suggesting that there are other determinants of persistence.
Logistic regression of emergency department visits found significantly lower risks for emergency department visits for SC combinations of different MOAs, while G combinations showed no differences in the odds of an emergency department visit. With lamotrigine accounting for 76% of SC+SC combinations, it is possible that the 6- to 8-week titration period for this drug may be influencing this finding despite the 90-day minimum for coadministration of the AED combination, although other factors, such as drug interactions among enzyme-inducing SC drugs, may also be involved.
Logistic regression of inpatient admissions suggested a lower risk with different-MOA G combinations, while differences in hospitalization risk for SC combinations were not significant. Different-MOA G combinations showed a 28% lower hospitalization risk vs G+G overall and significantly lower risk for G+SC and G+M combinations specifically. In observing that clonazepam or phenobarbital combinations accounted for 84% of G+G cases, differences found for G+G may be a result of selection bias owing to these patients being more difficult to treat and therefore resulting in increased hospitalization risk. As with treatment duration, other factors related to drug choices may impact outcomes. Additional research is needed to better understand the potential for the characteristics of a specific drug in a given MOA class to impact outcomes.
Differences in health care resource use between SC and G combination cohorts were associated with their relative comorbidity burden and may be indicative of differential practice patterns for selecting AED combinations based on patients’ comorbidities and clinical characteristics, possibly contributing to differences in outcomes across MOA combination categories. Because this study could not establish causation, further research may more closely examine associations of comorbidities with AED selection and resulting outcomes.
Several study limitations should be noted. Drugs categorized in a particular MOA class may have different clinical, pharmacokinetic, and safety profiles that influence their use and so may limit the clinical implications of our findings. In view of other published works describing MOA-based combination therapy, the MOA categories selected for this study may not have been sufficiently detailed to detect differences between AED combinations.4 This study used administrative claims data, for which the completeness and accuracy of medical coding is subject to data-coding restrictions, diagnosis misclassifications, and data entry error. The observed relationships could reflect the effects of unmeasured variables, such as severity of illnesses and seizure frequency, or incomplete capture of variables that determine the optimal AED combination for a given patient with POS. It was assumed that prescriptions were taken as directed, with accurate days’ supply noted on the claim. Lastly, this study was limited to working-aged individuals (or their dependents) enrolled in commercial health care plans, so results may not be generalizable to other age groups or health care coverages.
Quiz Ref IDThese findings suggest that AED combinations with different MOAs have greater effectiveness, as measured by treatment persistence, and lower risks for hospitalization and emergency department visits, suggesting a strategy for achieving optimal AED combination therapy outcomes that is based on different MOAs rather than the same MOAs. Further research is needed to more fully understand the role of the MOA in achieving optimal outcomes in AED combination therapy and to identify additional factors that may influence outcomes when implementing an MOA-based combination therapy approach to POS management.
Corresponding Author: Jay M. Margolis, PharmD, Truven Health Analytics, 332 Bryn Mawr Ave, Bala Cynwyd, PA 19004 (firstname.lastname@example.org).
Accepted for Publication: March 20, 2014.
Published Online: June 9, 2014. doi:10.1001/jamaneurol.2014.808.
Author Contributions: Dr Margolis had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Margolis, Wang, Copher, Cavazos.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: All authors.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Margolis, Chu, Copher, Cavazos.
Obtained funding: Margolis, Wang.
Administrative, technical, or material support: Margolis, Wang.
Study supervision: Margolis, Wang.
Conflict of Interest Disclosures: Drs Margolis and Chu are employees of Truven Health Analytics, which was paid by Eisai Inc in connection with the research described in this article. Drs Wang and Copher are employees of Eisai Inc. No other disclosures were reported.
Funding/Support: This research was sponsored by Eisai Inc, Woodcliff Lake, New Jersey.
Role of the Sponsor: Eisai Inc had a role in the design and conduct of the study; analysis and interpretation of the data; preparation, review, and approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: Editorial assistance was provided by Michele Shaw, PharmD, who was funded by Truven Health Analytics. We acknowledge the key contributions of Larry Radbill, MA (Truven Health Analytics), for his tireless work in defining, extracting, assembling, and analyzing the data. He did not receive funding from the study sponsor for his contributions.